# THE ROLE OF COMPLEMENT IN HEALTH AND DISEASE

EDITED BY : Maciej Cedzyński, Nicole M. Thielens, Tom Eirik Mollnes and Thomas Vorup-Jensen PUBLISHED IN : Frontiers in Immunology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers.

> The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence. ISSN 1664-8714 ISBN 978-2-88963-159-9 DOI 10.3389/978-2-88963-159-9

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# THE ROLE OF COMPLEMENT IN HEALTH AND DISEASE

Topic Editors:

Maciej Cedzyński, Institute of Medical Biology, Polish Academy of Sciences, Poland

Nicole M. Thielens, University Grenoble Alpes, CEA, CNRS, Institut de Biologie Structurale (IBS), France

Tom Eirik Mollnes, Oslo University Hospital and University of Oslo, Norway; Bodø Hospital and University of Tromsø, Norway; Norwegian University of Science and Technology, Norway

Thomas Vorup-Jensen, Department of Biomedicine, Aarhus University, Demnark

Citation: Cedzyński, M., Thielens, N. M., Mollnes, T. E., Vorup-Jensen, T., eds. (2019). The Role of Complement in Health and Disease. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-159-9

# Table of Contents

#### I. EDITORIAL

*09 Editorial: The Role of Complement in Health and Disease* Maciej Cedzyński, Nicole M. Thielens, Tom Eirik Mollnes and Thomas Vorup-Jensen

#### II. BASIC MECHANISMS OF COMPLEMENT PHYSIOLOGICAL AND PATHOLOGICAL FUNCTIONS


Eija Nissilä, Pipsa Hakala, Katarzyna Leskinen, Angela Roig, Shahan Syed, Kok P. M. Van Kessel, Jari Metso, Carla J. C. De Haas, Päivi Saavalainen, Seppo Meri, Angeliki Chroni, Jos A. G. Van Strijp, Katariina Öörni, Matti Jauhiainen, T. Sakari Jokiranta and Karita Haapasalo


Mariusz Z. Ratajczak, Mateusz Adamiak, Magda Kucia, William Tse, Janina Ratajczak and Wieslaw Wiktor-Jedrzejczak

*78 Asparaginyl Endopeptidase (Legumain) Supports Human Th1 Induction via Cathepsin L-Mediated Intracellular C3 Activation*

Simon Freeley, John Cardone, Sira C. Günther, Erin E. West, Thomas Reinheckel, Colin Watts, Claudia Kemper and Martin V. Kolev

# III. COMPLEMENT ASSAYS


Michele Mutti, Katharina Ramoni, Gábor Nagy, Eszter Nagy and Valéria Szijártó

*123 Development of a Quantitative Assay for the Characterization of Human Collectin-11 (CL-11, CL-K1)*

Rafael Bayarri-Olmos, Nikolaj Kirketerp-Moller, Laura Pérez-Alós, Karsten Skjodt, Mikkel-Ole Skjoedt and Peter Garred

#### *137 CL-L1 and CL-K1 Exhibit Widespread Tissue Distribution With High and Co-Localized Expression in Secretory Epithelia and Mucosa*

Soren W. K. Hansen, Josephine B. Aagaard, Karen B. Bjerrum, Eva K. Hejbøl, Ole Nielsen, Henrik D. Schrøder, Karsten Skjoedt, Anna L. Sørensen, Jonas H. Graversen and Maiken L. Henriksen

*149 Reference Intervals of Factor H and Factor H-Related Proteins in Healthy Children*

Anna E. van Beek, Angela Kamp, Simone Kruithof, Ed J. Nieuwenhuys, Diana Wouters, Ilse Jongerius, Theo Rispens, Taco W. Kuijpers and Kyra A. Gelderman

*155 Complement Factor H-Related Protein 4A is the Dominant Circulating Splice Variant of* CFHR4

Richard B. Pouw, Mieke C. Brouwer, Anna E. van Beek, Mihály Józsi, Diana Wouters and Taco W. Kuijpers

#### IV. THERAPEUTIC TOOLS: COMPLEMENT SUBSTITUTION AND INHIBITION

*164 Substitution of Mannan-Binding Lectin (MBL)-Deficient Serum With Recombinant MBL Results in the Formation of new MBL/MBL-Associated Serine Protease Complexes*

Mischa P. Keizer, Angela Kamp, Gerard van Mierlo, Taco W. Kuijpers and Diana Wouters


# V. COMPLEMENT AND INFECTION

*220 Epithelial C5aR1 Signaling Enhances Uropathogenic* Escherichia coli *Adhesion to Human Renal Tubular Epithelial Cells*

Yun Song, Kun-Yi Wu, Weiju Wu, Zhao-Yang Duan, Ya-Feng Gao, Liang-Dong Zhang, Tie Chong, Malgorzata A. Garstka, Wuding Zhou and Ke Li

*234 Decreased Ficolin-3-mediated Complement Lectin Pathway Activation and Alternative Pathway Amplification During Bacterial Infections in Patients With Type 2 Diabetes Mellitus*

László József Barkai, Emese Sipter, Dorottya Csuka, Zoltán Prohászka, Katrine Pilely, Peter Garred and Nóra Hosszúfalusi

*244 The Lectin Complement Pathway is Involved in Protection Against Enteroaggregative* Escherichia coli *Infection*

Camilla Adler Sørensen, Anne Rosbjerg, Betina Hebbelstrup Jensen, Karen Angeliki Krogfelt and Peter Garred

*255 Interaction of Mannose-Binding Lectin With Lipopolysaccharide Outer Core Region and its Biological Consequences*

Aleksandra Man-Kupisinska, Anna S. Swierzko, Anna Maciejewska, Monika Hoc, Antoni Rozalski, Malgorzata Siwinska, Czeslaw Lugowski, Maciej Cedzynski and Jolanta Lukasiewicz

*269 Virulence Associated Gene 8 of* Bordetella pertussis *Enhances Contact System Activity by Inhibiting the Regulatory Function of Complement Regulator C1 Inhibitor*

Elise S. Hovingh, Steven de Maat, Alexandra P. M. Cloherty, Steven Johnson, Elena Pinelli, Coen Maas and Ilse Jongerius

*280 Complement Component C1q as Serum Biomarker to Detect Active Tuberculosis*

Rosalie Lubbers, Jayne S. Sutherland, Delia Goletti, Roelof A. de Paus, Coline H. M. van Moorsel, Marcel Veltkamp, Stefan M. T. Vestjens, Willem J. W. Bos, Linda Petrone, Franca Del Nonno, Ingeborg M. Bajema, Karin Dijkman, Frank A. W. Verreck, Gerhard Walzl, Kyra A. Gelderman, Geert H. Groeneveld, Annemieke Geluk, Tom H. M. Ottenhoff, Simone A. Joosten and Leendert A. Trouw

*292 Active Human Complement Reduces the Zika Virus Load via Formation of the Membrane-Attack Complex*

Britta Schiela, Sarah Bernklau, Zahra Malekshahi, Daniela Deutschmann, Iris Koske, Zoltan Banki, Nicole M. Thielens, Reinhard Würzner, Cornelia Speth, Guenter Weiss, Karin Stiasny, Eike Steinmann and Heribert Stoiber

*301 Ficolin-1 and Ficolin-3 Plasma Levels are Altered in HIV and HIV/HCV Coinfected Patients From Southern Brazil*

Maria Regina Tizzot, Kárita Cláudia Freitas Lidani, Fabiana Antunes Andrade, Hellen Weinschutz Mendes, Marcia Holsbach Beltrame, Edna Reiche, Steffen Thiel, Jens C. Jensenius and Iara J. de Messias-Reason


Marcin Okrój and Jan Potempa

#### VI. COMPLEMENT AND CANCER

*329 Complementing Cancer Metastasis*

Dawn M. Kochanek, Shanawaz M. Ghouse, Magdalena M. Karbowniczek and Maciej M. Markiewski

*340 MASP-1 and MASP-2 Serum Levels are Associated With Worse Prognostic in Cervical Cancer Progression*

Carlos Afonso Maestri, Renato Nisihara, Hellen Weinschutz Mendes, Jens Jensenius, Stephen Thiel, Iara Messias-Reason and Newton Sérgio de Carvalho

#### *345 The Role of Complement Activating Collectins and Associated Serine Proteases in Patients With Hematological Malignancies, Receiving High-Dose Chemotherapy, and Autologous Hematopoietic Stem Cell Transplantations (Auto-HSCT)*

Anna S. Świerzko, Mateusz Michalski, Anna Sokołowska, Mateusz Nowicki, Łukasz Eppa, Agnieszka Szala-Poździej, Iwona Mitrus, Anna Szmigielska-Kapłon, Małgorzata Sobczyk-Kruszelnicka, Katarzyna Michalak, Aleksandra Gołos, Agnieszka Wierzbowska, Sebastian Giebel, Krzysztof Jamroziak, Marek L. Kowalski, Olga Brzezińska, Steffen Thiel, Jens C. Jensenius, Katarzyna Kasperkiewicz and Maciej Cedzyński

*360 Spontaneous Remission in Paroxysmal Nocturnal Hemoglobinuria—Return to Health or Transition Into Malignancy?* Eva-Stina Korkama, Anna-Elina Armstrong, Hanna Jarva and Seppo Meri

#### VII. COMPLEMENT AND RENAL DISEASES

*365 Overactivity of Alternative Pathway Convertases in Patients With Complement-Mediated Renal Diseases*

Marloes A. H. M. Michels, Nicole C. A. J. van de Kar, Marcin Okrój, Anna M. Blom, Sanne A. W. van Kraaij, Elena B. Volokhina and Lambertus P. W. J. van den Heuvel On Behalf of the COMBAT Consortium

*378 Alternative Pathway is Essential for Glomerular Complement Activation and Proteinuria in a Mouse Model of Membranous Nephropathy* Wentian Luo, Florina Olaru, Jeffrey H. Miner, Laurence H. Beck Jr,

Johan van der Vlag, Joshua M. Thurman and Dorin-Bogdan Borza


Sophie Chauvet, Lubka T. Roumenina, Pierre Aucouturier, Maria-Chiara Marinozzi, Marie-Agnès Dragon-Durey, Alexandre Karras, Yahsou Delmas, Moglie Le Quintrec, Dominique Guerrot, Noémie Jourde-Chiche, David Ribes, Pierre Ronco, Frank Bridoux and Véronique Fremeaux-Bacchi

*413 Unraveling the Molecular Mechanisms Underlying Complement Dysregulation by Nephritic Factors in C3G and IC-MPGN*

Roberta Donadelli, Patrizia Pulieri, Rossella Piras, Paraskevas Iatropoulos, Elisabetta Valoti, Ariela Benigni, Giuseppe Remuzzi and Marina Noris

*438 Complement Activation During Ischemia/Reperfusion Injury Induces Pericyte-to-Myofibroblast Transdifferentiation Regulating Peritubular Capillary Lumen Reduction Through pERK Signaling*

Giuseppe Castellano, Rossana Franzin, Alessandra Stasi, Chiara Divella, Fabio Sallustio, Paola Pontrelli, Giuseppe Lucarelli, Michele Battaglia, Francesco Staffieri, Antonio Crovace, Giovanni Stallone, Marc Seelen, Mohamed R. Daha, Giuseppe Grandaliano and Loreto Gesualdo

#### *455 Complement C3 Produced by Macrophages Promotes Renal Fibrosis via IL-17A Secretion*

Yanyan Liu, Kun Wang, Xinjun Liang, Yueqiang Li, Ying Zhang, Chunxiu Zhang, Haotian Wei, Ran Luo, Shuwang Ge and Gang Xu

#### *472 The Complement System in Dialysis: A Forgotten Story?* Felix Poppelaars, Bernardo Faria, Mariana Gaya da Costa, Casper F. M. Franssen, Willem J. van Son, Stefan P. Berger, Mohamed R. Daha and Marc A. Seelen

*484 Intradialytic Complement Activation Precedes the Development of Cardiovascular Events in Hemodialysis Patients*

Felix Poppelaars, Mariana Gaya da Costa, Bernardo Faria, Stefan P. Berger, Solmaz Assa, Mohamed R. Daha, José Osmar Medina Pestana, Willem J. van Son, Casper F. M. Franssen and Marc A. Seelen

#### VIII. COMPLEMENT AND OTHER DISEASES


Olivia May, Nicolas S. Merle, Anne Grunenwald, Viviane Gnemmi, Juliette Leon, Cloé Payet, Tania Robe-Rybkine, Romain Paule, Florian Delguste, Simon C. Satchell, Peter W. Mathieson, Marc Hazzan, Eric Boulanger, Jordan D. Dimitrov, Veronique Fremeaux-Bacchi, Marie Frimat and Lubka T. Roumenina

*568 The Lectin Pathway of Complement in Myocardial Ischemia/Reperfusion Injury—Review of its Significance and the Potential Impact of Therapeutic Interference by C1 Esterase Inhibitor*

Anneza Panagiotou, Marten Trendelenburg and Michael Osthoff


Sandra Jeremias Catarino, Fabiana Antunes Andrade, Angelica Beate Winter Boldt, Luiza Guilherme and Iara Jose Messias-Reason

#### *606 IL-6 Receptor Inhibition by Tocilizumab Attenuated Expression of C5a Receptor 1 and 2 in Non-ST-Elevation Myocardial Infarction*

Hilde L. Orrem, Per H. Nilsson, Søren E. Pischke, Ola Kleveland, Arne Yndestad, Karin Ekholt, Jan K. Damås, Terje Espevik, Bjørn Bendz, Bente Halvorsen, Ida Gregersen, Rune Wiseth, Geir Ø. Andersen, Thor Ueland, Lars Gullestad, Pål Aukrust, Andreas Barratt-Due and Tom E. Mollnes

#### *617 Analysis of C3 Gene Variants in Patients With Idiopathic Recurrent Spontaneous Pregnancy Loss*

Frida C. Mohlin, Piet Gros, Eric Mercier, Jean-Christophe Raymond Gris and Anna M. Blom

# Editorial: The Role of Complement in Health and Disease

Maciej Cedzynski ´ 1 \*, Nicole M. Thielens <sup>2</sup> \*, Tom Eirik Mollnes 3,4,5 \* and Thomas Vorup-Jensen6,7 \*

<sup>1</sup> Laboratory of Immunobiology of Infections, Institute of Medical Biology, Polish Academy of Sciences, Łódz, Poland, <sup>2</sup> Univ. Grenoble Alpes, CNRS, CEA, IBS, Grenoble, France, <sup>3</sup> Department of Immunology, Oslo University Hospital, University of Oslo, Oslo, Norway, <sup>4</sup> Research Laboratory, Bodø Hospital, K.G. Jebsen TREC, University of Tromsø, Tromsø, Norway, <sup>5</sup> Centre of Molecular Inflammation Research, Norwegian University of Science and Technology, Trondheim, Norway, <sup>6</sup> Biophysical Immunology Laboratory, Department of Biomedicine, Aarhus University, Aarhus, Denmark, <sup>7</sup> Interdisciplinary Nanoscience Center, Aarhus University, Aarhus, Denmark

Keywords: complement, classical pathway, alternative pathway, lectin pathway, membrane-attack complex, infection, cancer, renal disease

#### **Editorial on the Research Topic**

#### **The Role of Complement in Health and Disease**

#### Edited and reviewed by:

Francesca Granucci, University of Milano-Bicocca, Italy

#### \*Correspondence:

Maciej Cedzynski ´ mcedzynski@cbm.pan.pl Nicole M. Thielens nicole.thielens@ibs.fr Tom Eirik Mollnes t.e.mollnes@medisin.uio.no Thomas Vorup-Jensen vorup-jensen@biomed.au.dk

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 16 July 2019 Accepted: 24 July 2019 Published: 07 August 2019

#### Citation:

Cedzynski M, Thielens NM, ´ Mollnes TE and Vorup-Jensen T (2019) Editorial: The Role of Complement in Health and Disease. Front. Immunol. 10:1869. doi: 10.3389/fimmu.2019.01869 The complement system is a crucial mediator of the innate immune response, contributing to cell homeostasis, tissue development, and repair, reproduction, and cross-talk with other endogenous systems. Each of the three major complement activation pathways (classical, CP; alternative, AP; and lectin, LP) employs specific recognition molecules and initiating serine proteases. All three pathways converge into a common sequence. This leads to the formation of the biologically highly active anaphylatoxin, C5a, and the C5b-9 membrane attack complex (MAC). The latter forms transmembrane channels which either induces "sub-lytic" activation of the cell, or results in target cell lysis.

Primarily, as an important branch of first-line defense, complement protects the host from invading pathogens and abnormal self-derived components. This can be achieved through several mechanisms: (i) opsonization by activation products, (ii) attracting immune cells through chemotaxis, and (iii) direct lysis after incorporation of the MAC into the cell envelope of the invading pathogen. Therefore, complement plays key roles in (i) preventing the spread of infection to other cells and tissues, (ii) participating in the clearance of damaged cells and tissues, and (iii) preventing the development of chronic inflammation and/or cancer. On the other hand, uncontrolled complement activation may lead to life-threatening effects such as systemic inflammation and shock, dysregulation of coagulation/fibrinolysis, and auto-aggression. Furthermore, under certain conditions, deregulated complement activation may also promote tumorigenesis.

The 49 articles of this issue summarize recent achievements and provide timely reviews in the field of complement research.

#### BASIC MECHANISMS OF COMPLEMENT PHYSIOLOGICAL AND PATHOLOGICAL FUNCTIONS

The complement cascade needs to be tightly regulated to avoid over activation and inflammatory pathologies. The knowledge on complement regulators has greatly evolved recently, and Sánchez-Corral et al. provided an update on the family of factor H related proteins (FHRs). Their review includes quantitation of circulating FHR proteins, differential

**9**

interaction of FH and FHR proteins with various ligands, and association of genetic variants and abnormal protein rearrangements with diseases involving renal or ocular pathologies. FHRs are emerging as FH antagonists able to enhance complement activation by outcompeting FH for ligand binding. FH interaction with apolipoprotein E (apoE) has been revisited by Nissilä et al. to decipher its possible role in the antiatherogenic effects of apoE. FH bound to monocyte cell surface complement receptor (CR3) in the absence of deposited C3b and increased ApoE binding to monocytes and THP-1 cells, possibly through simultaneous interaction with sialic acids. Incubation of FH with THP-1 macrophages and cholesterol-labeled cells enhanced cholesterol efflux and modulated the transcription of inflammatory genes. In addition, binding of FH to THP-1 macrophages decreased complement activation, suggesting an overall contribution of FH apoE interaction in reducing inflammation in atherosclerotic lesions.

Complement receptors provide a connection between soluble complement components and immune cells functions. The review by Vorup-Jensen and Jensen highlights similarities and differences in the structure and function of CR3 and CR4, two members of the family of β2 integrins with similar ectodomain organization. Available X-ray crystal structures and comparison of the wide variety of identified ligands suggest that CR3 and CR4 selectively recognize polycationic and polyanionic species, respectively, with complementary functional implications. The potential role of CR3 and CR4 in human therapy is raised, since existing drugs may target their contribution to both innate and adaptive immunity.

The review by Ratajczak et al. focuses on an emerging role of mannose-binding lectin- (MBL)-dependent complement activation in the release of effector cells from bone marrow (BM) in response to tissue injury, pathogens, and pharmacological inducers of hematopoietic stem/progenitor cells (HSPCs). This process is triggered by MBL binding to ATP, an alarmin released from activated BM cells and contributes to sterile inflammation in the BM microenvironment. ATP is an important signaling molecule, providing an emergent link between purinergic signaling and complement, which participates in homeostasis, mobilization and aging of HSPCs, and possibly in pathologies including myelodysplasia and graft-vs.-host disease in transplanted recipients.

Intracellular C3 activation mediated by cathepsin L (CTSL) plays a central role in the regulation of human T cell responses through signals delivered by C3a and C3b binding to complement receptors C3aR and CD36, respectively. Freeley et al. showed that asparaginyl endopeptidase (AEP), also called legumain, is required for normal Th1 induction and IFN-γ production in both human and mouse CD4<sup>+</sup> T cells. They also identified AEP as responsible for CTSL processing to an active form, required for CTSL-mediated C3 activation and autocrine CD46 and C3aR activation in human CD4<sup>+</sup> T cells, but not in mice cells.

This section highlights that complement basic knowledge is continuously emerging with the description of novel functions relying on recently identified ligands of complement components or on cross-talk with different signaling pathways.

# COMPLEMENT ASSAYS

The review by Ekdahl et al. summarizes the current indications for complement diagnostics and the specific methods used to determine individual complement status, and provides clues to interpret serological complement biomarkers in pathologic conditions. The described assays measure the levels of circulating complement components and activation products, and the function of the different activation pathways, allowing to distinguish between lack of a single component and functional deficiency. The main indications for determination of patient's complement status are deficiencies (inherited and acquired) and pathologies with abnormal complement activation. Additionally serum biomarkers should be useful for the follow-up of patients with complement therapeutics.

Complement biomarkers are especially useful to follow the activity of diseases if they are compared with baseline values under physiological conditions. Gaya da Costa et al. have investigated the serum concentrations of 19 complement components and the activity of the three complement pathways in a healthy Caucasian population of 120 individuals (50% women/men, 20–69 years old). They observed that sex and age significantly impact the levels and functionality of complement, suggesting that these two population parameters should be considered in complement-related pathologies and -targeted therapies.

Mutti et al. addressed the problem of low reproducibility of assays employing normal serum. With the help of affinity chromatography, antibodies may be removed from the serum; however, this procedure depletes serum from complement components associated with CP, LP, and AP. The reconstitution of complement activity (at least its classical pathway) is possible by supplementation with corresponding factors at proper concentrations and in strictly defined order. The proposed method makes possible better standarization which may be helpful in studies involving CP activity.

Clinical tools are still rare for analysis of collectin-11 (CL-11, alias CL-K1), the most recently discovered member of the complement defense collagens family. This protein has been found in the circulation as heteromeric complexes with collectin-10 (CL-10, alias CL-L1) that are able to efficiently activate the lectin complement pathway. Bayarri-Olmos et al. developed a quantitative sandwich ELISA for measuring circulating CL-11, based on two newly generated mAbs. Mean plasma concentrations from 126 Danish blood donors (289.4 ng/ml) were in agreement with the literature and highly correlated to the levels of CL-10/11 complexes. In addition, zymosan was identified as a novel CL-11 ligand able to trigger activation of CL-11/MASP-2 complexes. With a view to understand the mechanism of CL-10/11 complex formation and cellular localization, Hansen et al. studied the tissue distribution of CL-10 and CL-11. Tissues from endocrine organs/glands including liver, kidney, adrenals, and pancreas, as well as exocrine tissues of the digestive system and sex-specific organs showed immunoreactivity for both collectins. The mRNA levels showed accordance between the synthesis site and protein localization. In addition, MASP-3 expression appeared to overlap greatly with the localization of CL-10 and CL-11, which might have functional implications.

The discrimination between normal and abnormal circulating levels of serum components is based on availability of reference intervals (RI) from healthy donors. In the case of FH and FHR-1 to−5, variations in serum levels in adults are associated with multiple diseases. As complement-mediated diseases can have early onset, van Beek et al. assessed the circulating levels of FH and FHR in a population of 110 healthy children (7 months−20.9 years old) and found differences in some, but not all of the RIs of these proteins by comparison with adults. In the case of FHR4, the level of FHR-4A was tested using an assay setup by Pouw et al. in healthy donor serum samples using novel specific mAbs. These authors showed indeed that FHR-4A is the dominant circulating variant with a level in the range of those reported for all other FHRs whereas FHR-4B could not be detected in serum. In addition, FHR-4A was found to be considerably variable among healthy individuals.

This section illustrates the wide variety of tools available to measure the levels of individual complement components and quantify complement function. The need for standardized tools is highlighted, given the growing number of pathologies involving complement and the development of complementbased therapeutics.

# THERAPEUTIC TOOLS: COMPLEMENT SUBSTITUTION AND INHIBITION

MBL deficiency is a common condition associated with several pathologies, including increased susceptibility to infections in childhood. Substitution therapies with plasma-derived MBL (pdMBL) proved to be safe with restoration of MBL levels but not of a functional LP due to inactivation of preactivated MBL-MASPs complexes present in pdMBL. Keizer et al. demonstrated that recombinant MBL (rMBL) can associate with non-complexed circulating forms of proenzyme MASPs and restore LP activity of MBL-deficient serum, providing a rationale for new clinical rMBL substitution studies.

The review by Dobó et al. presents a timely overview of complement therapeutics preventing or inhibiting unwanted/excessive activation of LP and AP. The potential drug targets described encompass LP pattern recognition proteins, serine proteases (MASPs, factors B and D), the central component C3, and complement regulators (factors H and I, properdin). Emphasis is put on recently described cross-talks between LP and AP involving active MASP-3 as the primary physiological activator of pro-FD to FD, and MASP-1 required for AP initiation on LPS containing surfaces. Drug candidates include antibodies, peptides, and selective protease inhibitors for use in the growing number of complement-mediated pathologies.

In a search for new complement inhibitors acting at different levels of the complement cascade, Hertz et al. developed chimeric proteins consisting in MAP-1, a MASP competitor for binding to LP recognition proteins, fused N- or C-terminally to CCP1- 5 of C4b-binding protein, a regulator of both CP and LP. Both chimers formed dimers able to associate with MBL or CL-11 while retaining the cofactor activity of C4BP. They could inhibit both CP and LP, exhibiting unique complementregulatory and anti-inflammatory properties and thus contribute to the emerging field of complement therapeutics.

With a view to develop specific inhibitors of AP and CP C3/C5 convertases, Zwarthoff et al. used C3b-and DNP-coated beads models with purified AP and CP components, respectively, and C3aR and C5aR cell reporter systems to functionally characterize surface-bound reconstituted AP and CP C3/C5 convertases. Screening of therapeutic complement inhibitors and known C3bbinding proteins from human, bacteria or ticks revealed that all molecules tested inhibited C5 conversion in both AP and CP, but only some of them inhibited the C3 convertase. These models open the way to the identification and development of specific convertase inhibitors for treatment of complementmediated diseases.

#### COMPLEMENT IN INFECTION

Urinary tract infections belong to the most common infectious diseases worldwide. Uropathogenic Escherichia coli are aetiological agents of ∼80% of cases. Song et al. demonstrated that stimulation of C5aR1-expressing renal tubular epithelial cells (RTEC) with C5a results in elevated expression of mannose residues, depending on the ERK1/2/NF-κB signaling pathway and upregulation of TNF-α production. That facilitates UPEC adhesion and invasion of RTEC, mediated by E. coli type 1 fimbriae. Furthermore, higher urine C5a concentrations were found in UTI patients compared with controls. Therefore, it was concluded that C5a/C5aR1 interaction may contribute to the pathogenesis of UTI. Infections, including UTI, are particularly common and often severe in patients suffering from diabetes mellitus. Barkai et al. studied type 2 diabetes mellitus (T2DM) patients with acute bacterial infections, largely UTI caused by E. coli. They found higher functional activity of ficolin-3-dependent lectin and alternative pathways, accompanied by lower concentrations of C4d and sC5b-9, in plasma samples, compared with those of non-diabetic patients. The authors suggested generally lower consumption of ficolin-3 dependent LP/AP in T2DM patients resulting in impaired complement-mediated protection from infections. Indeed, lack of ficolin-3-dependent activation and AP amplification were found associated with higher 3-months mortality in diabetic patients.

Another E. coli pathotype, EAEC (enteroaggregative) is a worldwide-spread agent of diarrhea. Adler Sørensen et al. reported that ficolin-2 recognized more or less efficiently three of four prototypic EAEC strains but only 5/56 of clinical isolates. The recognition was followed by complement activation, partially inhibitable when ficolin-2 or factor D inhibitors were used and completely blocked when used together. Inhibition of complement activation protected bacteria from direct killing. Therefore, again, although LP activation accompanied by AP amplification is an effective mechanism of elimination of pathogens, many EAEC strains have evolved to evade that. Prototypic EAEC were not recognized by ficolin-1 or ficolin-3 while MBL binding was noted for two strains.

Man-Kupisinska et al. found recognition of enterobacterial lipopolysaccharides by MBL to be common. For most strains, the lectin target was LPS core oligosaccharide. N-acetyl-Dglucosamine and L-glycero-D-mannoheptose were identified as the ligands. They are accessible to MBL in vivo when constituting the terminal outer core sugars and the O-polysaccharide is relatively short or when LPS is released due to bacterial cell damage. Using Hafnia alvei as a model, it was demonstrated that MBL-LPS interaction not only initiates the complement cascade but also induces early-phase (lipid A-independent) endotoxic shock in mice. Thus, MBL may be involved in life-threatening events in the course of Gram-negative sepsis.

Pertussis (whooping cough), caused by Bordetella pertussis, is still one of the most deadly childhood diseases. Hovingh et al. focused on the interaction of Vag8, one of its virulence factors, with C1-inhibitor (C1-INH). That interaction results not only in rescuing bacteria from complement-dependent elimination (as suggested previously) but also in contact system activation. Such an effect was observed when recombinant Vag8 or wild-type B. pertussis strain were used but not for the Vag8-knockout bacteria. Sequestration of C1-INH by Vag8 is supposed to hamper its inhibitory activity against βFXIIa and plasma kallikrein. Contact system activation may be additional to the complement evasion mechanism of B. pertussis pathogenicity/virulence.

Another major public health problem is tuberculosis. It is estimated that a quarter of the global population is infected with Mycobacterium tuberculosis. The lack of sufficiently specific and sensitive biomarkers makes diagnosis difficult as well as differentiation between active and latent infection. Lubbers et al. suggested C1q to be a good candidate TB marker basing on data from several geographically distinct cohorts. The C1q serum concentrations were significantly higher in TB patients, compared not only with healthy controls but also with individuals with latent M. tuberculosis infection, leprosy, pneumonia, or sarcoidosis. Moreover, the expression of several complement-associated genes (essentially C1qB and SERPING1) was upregulated in active TB. Importantly, after half-yearly treatment, C1q levels in the TB group dropped to within the control range. Furthermore, the potential usefulness of that molecule as an active disease marker was supported by data from the non-human primate, Macaca mulatta.

Zika virus infections have spread rapidly in recent years. Schiela et al. studied the interaction of Zika virus with complement. The natural (specific against insect components but not against virus) IgM-dependent classical pathway was found to be the major player in complement activation, although direct binding of C1q to NS1 (regulator of viral RNA transcription) and ZIKV envelope were influential as well. Twice diluted serum efficiently reduced number of virions (2 log) while lower (10%) serum concentration had no effect. Based on data from C9 depleted/reconstituted serum, MAC formation was proven to be responsible for viral neutralization.

HIV/AIDS affects nearly 40 million people globally. Coinfections with other pathogens, including HBV and HCV are clinically important, aggravating factors, associated with increased mortality. Tizzot et al. demonstrated that plasma ficolin-1 concentrations were lower in HIV/HCV co-infected patients compared with those infected with HIV alone or controls (the median for controls was in turn lower than for HIV patients). In contrast, average serum ficolin-3 was found to be highest in HIV group and differed significantly from those noted for HIV/HCV patients and healthy subjects. However, ficolin-3 levels within the HIV group correlated with CD4<sup>+</sup> T cell counts and lower concentrations were found associated with AIDS. Such findings suggest that ficolin-1 and ficolin-3 play different roles in the host response to HIV and/or HCV infections.

Plasmodium falciparum, the major agent of another lifethreatening disease, malaria (>400 thousand deaths yearly), has developed a variety of pathways to evade host immune responses. Larsen et al. elucidated one of them, associated with P. falciparum erythrocyte membrane protein 1 (PfEMP1), expressed on the surface of infected red blood cells. Although recombinant PfEMP1, opsonised by IgG was demonstrated to activate complement via the classical pathway efficiently, no activation was observed when native protein was exposed on the surface of erythrocytes. As ability to initiate complement activation depends on antigen orientation, the authors suggested that exposure of PfEMP1 in electron-dense protrusions (knobs) on the cell surface prevents on-target IgG hexamerization and thus consequent C1q binding. Therefore, although antibody response against PfEMP1 is dominated by IgG1 and IgG3 subclasses, it does not lead to complement-dependent lysis of infected erythrocytes.

Other mechanisms of microbial complement evasion were reviewed by Okrój and Potempa. A special attention was paid to Porphyromonas gingivalis, the keystone pathogen of periodontitis. P. gingivalis expresses outer membrane-anchored proteases, gingipains, able to cleave C5, C3, and C4 components. Moreover, those molecules contribute to the increase of C1 deposition on the cell surface. Interaction of gingipains with C5 results in C5a anaphylatoxin production, generally associated with antimicrobial events. However, C5a suppresses intracellular killing of P. gingivalis by macrophages and corrupts the C5aR-TLR-2 cross-talk leading to release of proinflammatory cytokines, contributing thus to bone resorption. Furthermore, gingipains (at higher concentrations) degrade C5b (as well as C3 and C4) which prevents complement-dependent cell lysis. Finally, bacterial peptidyl arginine deiminase (PPAD) is able to citrullinate C5a Cterminal Arg residue leading to loss of its chemotactic activity. Therefore, it is supposed that P. gingivalis strategy consists in active controlling (promoting or inhibiting) complement activation at various stages and this "inflammophilic" pathogen utilizes that system to disturb host's homeostasis.

This section includes papers clarifying a variety of microbialcomplement interactions, both beneficial and detrimental for the host, and discussion of application of new disease biomarkers and treatment strategies. Last but not least it offers an overview of contemporary methodology for complement research.

#### COMPLEMENT IN CANCER

Multiway involvement of the complement system in cancer is widely discussed in the literature. Carcinogenesis is generally associated with changes in the expression of surface antigens, therefore cancer cells may be recognized by the immune system as "abnormal self." However, numerous reports revealed variety of mechanisms of escape from complement-dependent elimination, utilizing complement system to establish favorable conditions for tumor survival or to metastasis. Such strategies were depicted in two review papers included in this issue. The first one, cited above (by Okrój and Potempa), specifies excessive expression of membrane-bound/production of soluble complement inhibitors, hijacking the latter from plasma by tumor cells and their ability to remove MAC from the surface. Some cancer cells take advantage of local inflammation mediated by anaphylatoxins which may be intensified by endogenous C5 production and C5a generation with their own serine proteases. The detrimental effects of anaphylatoxins, contributing to the proliferation and invasiveness of cancer cells expressing C3a and C5a receptors, was underlined as well in another review, by Kochanek et al. This paper is focused on complement-mediated promotion of cancer metastasis. The invasion-metastasis cascade, leading to spreading of tumor cells through blood or lymph is a complex process involving events in primary tumor, circulation and target sites. It is associated with majority (∼90%) of fatal outcomes of disease. The authors delineate contribution of C3a and C5a/C5aR1 to epithelial-tomesenchymal transition (EMT). Further, the formation of the premetastatic niche (including vascular alterations, activation of resident cells, remodeling of extracellular matrix, and recruitment of immunosuppressive cells) as well as proven experimentally or suspected involvement of complement system in that process are discussed in details. Therefore, targeting some complement factors seem to be promising for development of new life-saving therapeutic strategies.

Two other, experimental papers, concern associations of complement lectin pathway factors with malignancy. Cervical cancer is one of the most common cancers affecting women, evolving from persistent infection with oncogenic types of human papilloma virus (HPV). Maestri et al. found significantly higher serum concentrations of MASP-1, MASP-2, and MAp19 in patients with invasive cervical cancer compared with women diagnosed with low or moderate grade cervical intraepithelial neoplasia (CIN-1, CIN-2, CIN-3). Higher MASP-2 and MAp19 levels were moreover associated with CIN/cancer relapse and deaths. The ROC analysis revealed good sensitivity and specificity of high MASP-2 concentration for worse prognosis of cervical lesions. These data suggest an involvement of aforementioned serine proteases and related regulatory protein in the pathogenesis of cervical cancer and/or increase of their concentrations in response to carcinogenesis.

Hematologic malignancies derive from various cells of the immune system. Patients are severely immunocompromised due to disease and therapy what is a reason for high susceptibility to infections and mortality. Swierzko et al. found MBL deficiency to be associated with multiple myeloma while "gt1" genotype (based on analysis MBL2 gene 3'-UTR) seemed to be protective from lymphomas. Furthermore, serum concentrations of CL-LK (a complex of collectin-10 and collectin-11) were higher in multiple myeloma patients before chemotherapy than in healthy controls. High MBL levels were unexpectedly a risk factor for hospital infections. In contrast, MBL deficiency had no impact, at least during cytopenia. On the other hand, data from follow-up (although from limited number of patients) suggested its association with the most severe infections. Therefore, it was supposed that MBL is fully effective when acts together with phagocytes recovered thanks to autologous stem cell transplantation.

Paroxymal nocturnal hemoglobinuria is caused by a mutation of the PIG-A gene, resulting in absence of regulatory CD55 and CD59 on affected cells. It is associated with intravascular hemolysis, thrombosis and bone marrow failure. Previously, it was supposed that the spontaneous remission is relatively common. However, Korkama et al., who extensively investigated Finish cohort of patients, found that disappearance of PNH clone may be related not only to patient's recovery (which occurred to be less common as believed earlier) but also to development of leukemia. Therefore, patients have to be followed-up carefully for a potential of emergence of cancer.

This section demonstrates complex interplay between complement and cancer, from carcinogenesis through metastasis, infectious complications, impact of complement on effectiveness of therapy (and vice versa—effect of therapy on complement activity), to positive or fatal outcome.

# COMPLEMENT IN RENAL DISEASES

The role of complement in the pathophysiology of renal diseases, and particular the potential for treating several of these diseases, have been one of the main focuses of the complement literature the last years. Furthermore, the different triggering events and subsequent pathophysiological development of the renal diseases varies substantially. It is therefore critically important to understand the mechanisms behind the different renal diseases in order to develop optimal treatment regimens for each of them. Some of the diseases definitely are strongly complement driven, whereas other are less dependent on complement, though still complement activation may occur as a secondary process worsening the condition.

In the present issue, nine papers deals with renal diseases including dialysis. The atypical hemolytic uremic syndrome and the C3 glomerulopathies (C3GN), the latter with its two main subtypes, the dense deposits disease (DDD) and the C3 glomerulonephritis (C3GN), are caused by primary dysregulation of the alterative pathway due to mutations in the components. Secondary, destabilization of the alternative convertase may be due to autoantibodies (nephritic factors). In all cases, there are a stabilization or dysregulation of the alternative pathway convertase, leading to enhanced activation of C3 with, in most cases, subsequent C5 activation and destruction of the glomeruli mainly by terminal pathway activation with release of C5a and C5b-9 release. Michels et al. described a novel hemolytic assay to study the alternative convertase in whole serum, implying that both mutations, here exemplified by factor B, and autoantibodies to the convertase can be studied simultaneously. They documented different effects on the convertase by the different conditions and followed the convertase over time in the patients. Thus, they provided a

new tool to study the alternative pathway activity in a broad range of renal diseases. The authors postulated targets in the alternative pathway for therapeutic inhibition. The important role of the alternative pathway in renal disease is followed up by Luo et al. who studied the role of the alternative pathway in a mice model of membranous nephropathy, a disease which has been shown to be mechanistically driven by the C5b-9 destruction of the podocytes with subsequent proteinuria. The disease is caused by autoantibodies, mainly of the IgG4 subclass, which does not activate the classical pathway. The role of the lectin and alternative pathway in triggering complement activation is unclear. Interestingly, the authors fond the same amount of subendothelial IgG in wild type as in factor B deficient mice, whereas the latter was protected from proteinuria. Proteinuria was markedly reduced in C5 deficient mice, consistent with C5b-9 being the mediator of the proteinuria. Notably, the authors document that the alternative pathway is crucial for the membrane damage and proteinuria in this model. Future studies will reveal whether this is a direct alternative activation or is due to a strong amplification by a weak triggering signal from another initial pathway.

Different renal diseases with complement-dependent pathophysiology clearly differ with respect to the initial triggering event and the alternative pathway is not necessarily primary, as mentioned above, although it contributes with the amplification mechanism. Limited data are available on the role of the lectin pathway in lupus glomerulonephritis. In a study by Machida et al., it was documented that the lectin pathway molecules MASP-1/3, using knock-out mice for this gene, were essential for development of glomerulonephritis secondary to the lupus-like mouse MRL/lpr model. They concluded that the activation was either by the lectin and/or the alternative pathway. To further illustrate the complexity of initial activation of complement, Chauvet et al. showed that not only the monoclonal antibodies contributes to C3G (MIg-C3), but a number of patients also presented with polyclonal antibodies. This might be of importance for the strategy of treatment of these patients and the authors suggest this as an example of how to design personalized therapy in the future.

Nephritic factors, as mentioned above for the alternative pathway, has for a long time been associated with renal disease. Recent research has revealed several different forms of these autoantibodies, not only intervening with the alternative pathway C3 convertase, but C4 nephritic factors, intervening with the classical pathway, and C5 nephritic factors, which are C3 nephritic factors also stabilizing the C5 convertase, and limited to the C3 convertase, have been described. The mechanism by which the nephritic factors acts has not been fully elucidated. In the study by Donadelli et al., a significant contribution to unraveling the different molecular mechanism of which C3 nephritic factors acts is described.

Another general pathophysiological condition, the ischemiareperfusion injury, is of importance in kidney damage, e.g., trauma and shock, and not least during kidney transplantation. Castellano et al. showed a mechanism by which ischemiareperfusion induces trans-differentiation from pericytes to myofibroblasts by regulating the peritubular capillary lumen through pERK signaling. The data from a pig model suggest that C1-inhibitor might be a future treatment option for renal ischemia-reperfusion injury. Notably, C1-INH is not a specific complement inhibitor, but acts potently both in the kallikreinkinin systems and in hemostasis in general. Since there is a considerable cross-talk between the cascade systems, named thromboinflammation, this is a very interesting approach for the future of attenuating the renal ischemia-reperfusion injury.

For any renal disease, it should be remembered that C3 is locally produced and contributes to the complement activity and pathogenesis in the tissue. Kidney macrophages serve as an important producer of C3, as an additional and important source to the liver derived plasma C3. Here, Liu et al. showed that this local C3 production might lead to the serious condition renal fibrosis via IL-17A secretion, using inhibitors of C3 and C3aR.

Finally, two papers by Poppelaars et al. and Poppelaars et al. focus on the consequences of hemodialysis on complement activation and subsequent undesired effects, not only on the already damaged kidneys, but also on the systemic endothelial and vessel homeostasis. In the first one, the authors timely and importantly reminds us in a review that we should always be aware of the consequences of hemodialysis. The biocompatibility of the devices is of outmost importance. Complement has been used as valuable biomarker for the compatibility, as well as it has been documented that devices that activate complement lead to a series of secondary adverse effects. In the second paper, the authors follow up this reminder with the important notice that intradialytic complement activation precedes the development of cardiovascular events in hemodialysis patients. This is a clinical study supplied with experimental data supporting their conclusions. Both papers highlight the importance of improving biocompatibility of dialysis devices, reducing complement activation, and not least increase the number of donors with reduced damage of their kidneys before, during and after the transplantation. This would shorten the time for patients suffering during dialysis waiting for a kidney graft and a substantially improved quality of life.

# COMPLEMENT IN OTHER DISEASES

Autoinflammation is the activation of the innate immune system to produce a condition where the response damages host tissue by a release of cytokines and other overactivation of immune effector mechanisms. As such, there is a similarity to the autoimmunity of the adaptive immune system except that no major antigenic stimulants seems to be involved. Rather, more elusive initiators are at play, for instance involving underlying mutational defects in signaling pathways, deficiency in certain critical homeostatic mechanisms or, at least for the acute responses, violent inflammatory responses triggered by traumas or other marked insults to the body. Unsurprisingly, the complement system has long been implicated in these events. In the past, however, the contributions of the complement system was mostly described at observational level, while the present volume includes several papers that outlines more clearly mechanistic associations.

Systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) are usually considered autoimmune diseases. There is a strong association with autoreactive antibodies, which may serve as biomarkers for the disease. Nevertheless, the innate immune system and its autoinflammatory responses are also an important part of the diseases. As reviewed by Holers and Banda, complement activation in RA may happen through the classical pathway and antibodies to collagen, but other means of activations also occur. Importantly, an animal model of RA deficient in C3 does not develop disease and several other modifications of complement activations delay onset of disease. Taken together, there can be little doubt that complement is an important part of RA. In SLE, complement is also activated although the precise mechanisms remain unclear. In consequence, split products of C3 are generated. Troldborg et al. showed that this enables a detection of C3dg as a diagnostic marker for SLE. Apparently, this methodology is superior to measuring C3. As in the case of RA, it implicates complement activation in the disease. These perspectives on SLE, a disease with marked skin manifestations, are further generalized by Giang et al. in a review on complement and skin disease. Indeed, a new wave of attention to complement has occurred after the development pharmaceutically useful complement inhibitors. This could likely alter the treatment of psoriasis and several other skin diseases with the known influences of complement. In psoriasis vulgaris, the most common form of psoriasis, several complement split products can be found in psoriatic lesions and in the skin scales. These inflammatory foci generate enough complement activation products to cause a rise in systemic levels through a "spill-over." It remains unknown, however, what exact mechanisms are responsible for complement activation in psoriasis.

Ischemic injuries are another setting where the autoinflammatory response acts to enhance morbidity. Nauser et al. explained the role of a relatively recently discovered member of the complement-activating molecules, namely the collectin-11, in renal transplant injury. Here, ischemia upregulates a glycan structure, which becomes a damage-associated molecular pattern, recognized by collectin-11. Yet, our possibilities of reducing such a response still need the elucidation of the precise glycan structure recognized. The endothelium of kidneys may also be target of overactivation of complement during atypical hemolytic uremic syndrome. As shown by May et al., the release of heme in consequence of the hemolysis is correlated to complement-deposition on the microvascular endothelial cells while the macrovascular endothelium is protected by upregulation of heme- and complement-degrading molecules. A similar topic of tissue damage was pursued by Panagiotou et al. with a focus on ischemia in the myocardial injury. The lectin pathway is a contributor to such injury during reperfusion although other means of complement activation also may play a role. Perhaps for this reason, the multifaceted inhibition of complement by C1-inhibitor, which embodies both the MASP as well as C1 proteases, offers protection in myocardial ischemia/reperfusion. However, the authors conclude that there is a lack of high-quality clinical studies of complement inhibition in human acute myocardial infarction. In some ways, the dysregulated activation of complement induced by trauma resemble the ischemic/reperfusion responses with a significant formation of complement split products. However, the review by Chakraborty et al. describes how the inflammatory response in trauma also involves central parts of the adaptive immune system, from antigen presentation to regulatory signaling in both B and T cells. These findings should encourage similar considerations in other setting with dysregulated complement activation.

Organ damage through dysregulated complement activation may also have an origin in infectious diseases. Catarino et al. present evidence of tissue-damaging complement activation in rheumatic fever after infection with Streptococcus pyogenes. In this way, ficolins play a double role by both protecting against the infection as well as raising the tissue damage by complement activation also reminiscent of the report on collectin-11.

In asking the question how to handle complement-mediated inflammation in a clinical setting, the study by Orrem et al. provided interesting insight on the relationship between IL-6 inhibition and the expression of the highly inflammatory C5a and C3a receptors. Tocilizumab, a function-blocking antibody to IL-6 receptor, also down-regulates C5a receptor expression in a context of coronary artery disease.

Complement also serves function in physiology outside the usual venues of immunology. Mohlin et al. identified a set of C3 gene variation, which they showed in subsequent expression studies were associated with low cellular secretion or functional defects. Interestingly, these defects may affect functions of this molecule in early phase of pregnancy and in the development of the placenta.

#### AUTHOR CONTRIBUTIONS

MC wrote introduction, Complement in Infection, and Complement in Cancer sections. NT wrote Basic Mechanisms of Complement Physiological and Pathological Functions and Complement Assays sections. TM wrote Complement in Renal Diseases section. TV-J wrote Complement in Other Diseases section.

# ACKNOWLEDGMENTS

We would like to thank all authors for their contributions to this Research Topic. We are also grateful to all reviewers for their evaluation of manuscripts submitted.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Cedzynski, Thielens, Mollnes and Vorup-Jensen. This is an open- ´ access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Self-Damage Caused by Dysregulation of the Complement Alternative Pathway: Relevance of the Factor H Protein Family

*Pilar Sánchez-Corral1†, Richard B. Pouw2†, Margarita López-Trascasa1,3† and Mihály Józsi 4,5\*†*

*1Complement Research Group, Hospital La Paz Institute for Health Research (IdiPAZ), La Paz University Hospital, Center for Biomedical Network Research on Rare Diseases (CIBERER), Madrid, Spain, 2Department of Pharmaceutical Sciences, University of Basel, Basel, Switzerland, 3Department of Medicine, Universidad Autónoma de Madrid, Madrid, Spain, 4Complement Research Group, Department of Immunology, ELTE Eötvös Loránd University, Budapest, Hungary, 5MTA-SE Research Group of Immunology and Hematology, Hungarian Academy of Sciences and Semmelweis University, Budapest, Hungary*

#### *Edited by:*

*Maciej Cedzynski, Institute for Medical Biology (PAN), Poland*

#### *Reviewed by:*

*Cees Van Kooten, Leiden University, Netherlands Lubka T. Roumenina, INSERM UMRS 1138, France*

> *\*Correspondence: Mihály Józsi mihaly.jozsi@ttk.elte.hu*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 31 May 2018 Accepted: 28 June 2018 Published: 12 July 2018*

#### *Citation:*

*Sánchez-Corral P, Pouw RB, López-Trascasa M and Józsi M (2018) Self-Damage Caused by Dysregulation of the Complement Alternative Pathway: Relevance of the Factor H Protein Family. Front. Immunol. 9:1607. doi: 10.3389/fimmu.2018.01607*

The alternative pathway is a continuously active surveillance arm of the complement system, and it can also enhance complement activation initiated by the classical and the lectin pathways. Various membrane-bound and plasma regulatory proteins control the activation of the potentially deleterious complement system. Among the regulators, the plasma glycoprotein factor H (FH) is the main inhibitor of the alternative pathway and its powerful amplification loop. FH belongs to a protein family that also includes FH-like protein 1 and five factor H-related (FHR-1 to FHR-5) proteins. Genetic variants and abnormal rearrangements involving the FH protein family have been linked to numerous systemic and organ-specific diseases, including age-related macular degeneration, and the renal pathologies atypical hemolytic uremic syndrome, C3 glomerulopathies, and IgA nephropathy. This review covers the known and recently emerged ligands and interactions of the human FH family proteins associated with disease and discuss the very recent experimental data that suggest FH-antagonistic and complement-activating functions for the FHR proteins.

Keywords: age-related macular degeneration, atypical hemolytic uremic syndrome, C3 glomerulopathy, complement activation, complement de-regulation, factor H, factor H-related protein, opsonization

#### INTRODUCTION

While initially only regarded as a supporting factor for the effectivity of immunoglobulins, the complement system is nowadays widely recognized as a crucial part of the innate immune system involved in many different processes (1). In addition to acting as a first line of defense by directly targeting and killing invading pathogens, with or without the help of immunoglobulins, its role in inflammation, immune cell recruitment, and clearance of immune complexes, apoptotic cells, and necrotic cells places complement at the center of the human immune system. The relevant role of complement is corroborated by the variety of pathological situations associated with complement deficiency or dysfunction.

Three complement activation pathways have been defined, each comprised of various proteins forming an intricate cascade of activation events (**Figure 1**). Both the classical and the lectin pathways are initiated when pattern recognition molecules (PRMs) that are complexed with zymogens of serine proteases, bind to their ligand. The classical pathway is activated by the binding of the C1 complex to immunoglobulins and pentraxins, while the lectin pathway uses various PRMs, including mannose-binding lectin and ficolins, which bind to specific carbohydrate moieties. These ligands are normally not present on healthy human cells. In contrast, the alternative activation pathway is initiated through the constitutive low rate hydrolysis of the internal thioester bond of C3, allowing binding of various activating complement proteins. All three pathways lead to the cleavage of C3 into C3a and C3b. C3b contains a highly reactive thioester group that is exposed upon C3 cleavage, resulting in the deposition of C3b onto virtually any molecule or cell surface in close proximity. When left unchecked, C3b on its own will again initiate the alternative pathway. As both the classical and the lectin pathway will also activate the alternative pathway once C3b is formed, thus enhancing complement activation, the alternative pathway has a pivotal role as an amplification loop within the complement system. Up to 80% of total complement activation has been ascribed to this amplification loop (2). Due to the spontaneous nature of the alternative pathway, it must be tightly controlled to prevent unwarranted and dangerous complement activation.

Complement regulation takes places both on the human cell surface and in the fluid phase. Several regulators, like most complement components, are found in the circulation. In addition, human cells express a wide array of membrane-bound complement regulators that control the system at various steps. Especially due to the activating proteins of the alternative pathway, regulation in the fluid phase is crucial, as unchecked, spontaneous C3 activation would lead to complete consumption of C3 and loss of complement activity. The 155-kDa glycoprotein complement factor H (FH) is the major regulator of the alternative pathway, inhibiting C3 activation both in the fluid phase as well as on human cell surfaces (**Figure 1**) (3). Similar to other complement regulators encoded in the regulators of complement activation (RCA) gene cluster, FH is composed of complement control protein (CCP) domains, often also referred to as short consensus repeat domains. FH is composed of 20 CCPs (**Figure 2**) (4). The first four N-terminal domains contain the complement inhibiting activity, such as decay accelerating activity and co-factor activity (5). The two most C-terminal CCP domains (19 and 20), together with a region located in CCPs 6–8, are crucial for binding of FH to surfaces, such as human cell membranes, as well as for mediating binding to several host and non-host ligands (discussed below) (6, 7). FH is highly abundant in plasma, with circulating levels of 233–400 µg/mL on average, although it has to be noted that some of the assays used might detect other FH family members as well (8–11).

Factor H-like protein 1 (FHL-1) is a splice variant derived from the *CFH* gene. Serum levels of FHL-1 are estimated to be 10–50 µg/mL (12, 13). FHL-1 is identical to the first seven CCP domains of FH, with an unique, four amino acid long C-terminus (14, 15). Thus, FHL-1 shares the C3b binding and regulatory domains CCPs 1–4 with FH and, like FH, it has complement inhibiting activity (16). Likewise, due to the shared CCPs 6–7 domains, FHL-1 and FH bind some common ligands, such as heparin, the pentraxins C-reactive protein (CRP) and pentraxin 3 (PTX3), and malondialdehyde epitopes (**Figure 2**). However, there are also differences in ligand interactions between FHL-1 and FH, not only because of the extra domains in FH but also due to the difference in their conformation and the unique SFTL tail at the C-terminus of FHL-1. For example, it was recently reported that the SFTL tail increases the interaction of FHL-1 with CRP and PTX3 (17).

Figure 1 | Overview of the role of factor H (FH) within the complement system. (I) The complement system is activated *via* binding of C1q (classical pathway), or mannan binding lectin/ficolins (MBL/FCN) (lectin pathway) in complex with serine proteases to specific molecules, or through the spontaneous activation of C3 into C3(H2O) (alternative pathway). Upon activation, the three pathways form C3 convertases (C4b2a or C3(H2O)Bb) resulting in the generation and deposition of C3b on the activating surface. (II) C3b forms new C3 convertase molecules (C3bBb) that enhance C3b deposition and amplify complement activation. (III) C3b can also bind to C3 convertases to generate C5 convertases (C4b2a3b or C3bBb3b); this process initiates the terminal pathway of complement activation, and the formation of the lytic C5b–C9 complex. FH keeps the spontaneous activation of C3 under control, and it also inhibits the complement system at both the activation and amplification stages. FH binds to deposited C3b and C3bBb complexes on human cell surfaces and inhibits further activation by three mechanisms: it competes with factor B (FB) for C3b binding and C3bBb generation; it increases the decay of C3bBb complexes, and it acts as a cofactor for factor I (FI), which in turn cleaves C3b into inactive C3b (iC3b).

is shown, with CCPs aligned vertically to the homologous domains in FH. The N-terminal CCPs 1–4 of FH and FHL-1 mediate the complement regulatory functions of these proteins (shown in yellow). CCPs 7 and 19–20 (shown in blue) harbor the main ligand- and host surface-recognition sites; selected ligand binding sites are indicated by horizontal lines. The CCPs 1–2 of factor H related protein 1 (FHR)-1, FHR-2, and FHR-5 are closely related to each other and mediate dimerization of these FHRs. The CCPs in FHRs with high sequence identity to the homologous FH domains are indicated by identical/similar colors.

Next to FH and FHL-1, humans (and several non-human species; not discussed here) possess FH-related (FHR) proteins, homologous to FH. They are encoded separately, with their genes (*CFHR1* to *CFHR5*) lying in tandem next to *CFH* at 1q31.3. The *CFHR* genes originate from *CFH* through gene duplication events (18). The *CFH–CFHRs* loci contain several segmental duplications, making them prone to genetic structural rearrangements due to nonallelic homologous recombination (NAHR) events. This has led to copy number polymorphisms (CNPs), with the very common 86.3-kb deletion (CNP147) that results in loss of *CFHR3–CFHR1* (Δ*CFHR3–CFHR1*), and the very rare 122-kb deletion (CNP148) resulting in loss of *CFHR1–CFHR4* (Δ*CFHR1–CFHR4*) (19). Like FH, the FHRs are entirely composed of CCP domains (**Figure 2**), which display high sequence similarity with CCP domains of FH known to be involved in ligand and surface binding. Remarkably, none of the human FHR proteins possess CCP domains homologous to FH CCPs 1–4. Thus, based on their primary structure, FHR proteins are not expected to have any direct complement inhibiting activity similar to FH. Nonetheless, several reports have observed direct complement inhibitory activity for some of the FHRs, albeit often weak compared to FH (20–24). However, other studies have not found such activity for FHRs, questioning whether this is truly the physiological role of the FHR proteins (25–30). Instead, the FHRs are currently hypothesized to have an antagonistic function over FH, competing with binding to FH ligands and cell surfaces. By lacking direct complement inhibiting activity, binding of FHRs instead of FH would allow complement activation to proceed (31). This process has also been termed complement de-regulation. Indeed, binding to various (FH) ligands has been reported for all FHRs, which will be discussed below. In addition, some FHRs were reported to promote alternative pathway activation by binding C3b and serving as a platform for the assembly of the C3 convertase (27, 32, 33). Recent characterization of some of the mouse FHRs supports a role of these proteins as positive regulators in the modulation of complement activation (34, 35).

In this review, we outline and provide an update on the recent developments regarding the FH protein family. New insights regarding circulating levels of FHRs, ligand binding, and disease associations allow re-assessing the role of FHRs in the complement system. Together, these results shed light on the balance of the FH–FHRs axis, and the role of FHRs in non-pathological and pathological conditions.

#### QUANTITATION OF FHR PROTEINS

Factor H, FHL-1, and the FHR proteins are mainly synthesized by hepatocytes, but synthesis by other cells and tissues has also been reported, particularly for FH and FHL-1 (36–38). FH production has been detected in endothelial cells, platelets, mesangial cells, keratinocytes, fibroblasts, retinal pigment epithelial cells, monocytes, and dendritic cells, among others (39–46). On the other hand, little information on the extrahepatic expression of the FHR proteins is available. Both *CFHR3* mRNA and FHR-3 protein have been identified in retinal macrophages, while no FHR-3 expression was found in other retinal cell types (47). Extrahepatic synthesis of FH/FHRs most likely contributes to an efficient control of complement activation locally, but a relevant contribution to the plasma levels of these proteins is unlikely, considering the relative low expression compared to the hepatic source.

Accurate quantification of the FHR proteins has been a great challenge since their discovery. Due to the high sequence similarity among FH and FHR proteins, it has proven to be very difficult to obtain specific reagents for each of the FHR proteins. For some time, only concentration estimates were available for most of the FHR proteins (21, 48). However, with recently renewed and successful efforts in generating highly specific antibodies, specific immunoassays for each of the FHR proteins are now becoming available, although some discrepancy about their actual physiological levels still remains (**Table 1**).


*a FHL-1 levels were determined indirectly, by subtracting the values of FH measurements from those of FH* + *FHL-1 measurements. N: number of samples; SD: standard deviation; IQR: interquartile range.*

#### Factor H-Related Protein 1 (FHR-1)

Factor H-related protein 1 is composed of five CCP domains, and circulates in two forms (37 and 42 kDa), with either one or two *N*-linked carbohydrate moieties (30, 49, 50). Two genetic variants of FHR-1 have been described, FHR-1\*A and FHR-1\*B, the difference being three amino acids in CCP3 (51). FHR-1\*B CCP3 is identical to FH CCP18, whereas FHR-1\*A CCP3 shares 95% sequence identity with FH CCP18. FHR-1 CCPs 4 and 5 share high sequence identity (100 and 97%) with FH CCPs 19 and 20, respectively. FHR-1 has a dimerization motif located in CCPs 1–2 that are highly similar (>85% sequence identity) to CCPs 1–2 of FHR-2 and FHR-5, and allow the formation of FHR-1 homodimers and heterodimers with FHR-2 (26, 52, 53). While identified *in vitro*, the existence of FHR-1/FHR-5 heterodimers *in vivo* is still controversial (26, 52, 53). Similarly, FHR-1 quantification also remains controversial. In 2017, several groups determined FHR-1 levels. Tortajada et al. reported an average of 122 µg/mL in 44 healthy controls with two copies of *CFHR1*, and an overall average of 90.4 µg/mL in 76 controls (including eight homozygous Δ*CFHR3*–*CFHR1* carriers and 24 heterozygous Δ*CFHR3*–*CFHR1* carriers) (54). Using the same immunoassay, Medjeral-Thomas et al. reported 94.4 µg/mL FHR-1 in 158 controls (of whom 133 were genotyped: 3 Δ*CFHR3*–*CFHR1* homozygous, 45 Δ*CFHR3*–*CFHR1* heterozygous, and 85 without Δ*CFHR3*–*CFHR1*) (55). Of note, the immunoassay described by Tortajada et al. does not distinguish between FHR-1 homodimers or heterodimers. In contrast, using immunoassays specific for FHR-1 homodimers and FHR-1/-2 heterodimers, van Beek et al. reported ~10-fold lower levels (averages of 11.33 and 5.48 µg/mL, respectively), in 115 healthy donors (2 homozygous Δ*CFHR3*–*CFHR1,* 36 heterozygous Δ*CFHR3*–*CFHR1* carriers, and 77 without Δ*CFHR3*–*CFHR1*) (53).

#### Factor H-Related Protein 2

Factor H-related protein 2 is the smallest FHR protein, composed of four CCP domains (56). FHR-2 circulates either non-glycosylated (24 kDa) or with one *N*-linked carbohydrate moiety in CCP2 (29 kDa). FHR-2 CCP1 and CCP2 are nearly identical to FHR-1 CCP1 and CCP2 (100 and 98%), respectively, including all residues comprising the dimerization motif (26). Similar to the proposed FHR-1/FHR-5 dimers, FHR-2/FHR-5 dimers remain to be identified *in vivo*, while FHR-2 homodimers and FHR-1/ FHR-2 heterodimers have been confirmed (52, 53). FHR-2 homodimer levels have been shown to be around 3 µg/mL; with these relatively low levels, FHR-2 seems to be the limiting factor in the formation of FHR-1/FHR-2 heterodimers and, indeed, most FHR-2 is found dimerized with FHR-1 (53).

#### Factor H-Related Protein 3

Factor H-related protein 3 is composed of five CCP domains, of which CCP1 and CCP2 have high sequence similarity with FH CCP6 and CCP7 (94 and 86%), respectively (57). The C-terminal CCPs 3–5 are virtually identical to the C-terminal domains of FHR-4A and FHR-4B (93–100%). FHR-3 contains four *N*-linked glycosylation sites, and it circulates in plasma as multiple glycosylation variants ranging from 37 to 50 kDa. A quantitative FHR-3-specific immunoassay was first described by Pouw et al., reporting levels of 0.38 and 0.83 µg/mL for healthy individuals carrying either one or two *CFHR3* copies, respectively (58). These results were later confirmed in a similar assay, reporting mean levels of 1.06 µg/mL (47). Two major genetic variants of *CFHR3* (*CFHR3\*A* and *CFHR3\*B*) have been described (59); interestingly, these are quantitative variants, with *CFHR3\*B* determining higher FHR-3 levels than *CFHR3\*A* (60). The FHR-3\*A and FHR-3\*B allotypes differ at aminoacid 241 in CCP3 (Pro/Ser), but its functional relevance has not been determined.

#### Factor H-Related Protein 4

*CFHR4* is the only known *CFHR* gene that expresses two splice variants, FHR-4A and FHR-4B (61, 62). FHR-4A is composed of nine CCP domains (86 kDa), while FHR-4B has five CCP domains (43 kDa). All FHR-4B domains are also present in FHR-4A, with FHR-4B CCP1 being identical to FHR-4A CCP1, and FHR-4B CCPs 2–5 being identical to FHR-4A CCPs 6–9. FHR-4A CCPs 2–4 seems to have arisen from internal gene duplication, and have high sequence similarity (85–93% amino acid identity) with the other CCPs in FHR-4A/B (61). Thus, obtaining specific reagents to distinguish FHR-4A from FHR-4B is challenging on first sight. Quantification by using an immunoassay that in principle measures both FHR-4A and FHR-4B resulted in average levels of 25.4 µg/mL (27). However, FHR-4A-specific antibodies have been described recently and used in an FHR-4A-specific ELISA which shows 10-fold lower levels for FHR-4A (2.55 ± 1.46 µg/mL) (63). In line with the complete sequence identity of FHR-4B with several FHR-4A domains, no specific antibodies for FHR-4B could be obtained. Strikingly, FHR-4B was not detected in plasma using various antibodies that did react with recombinant FHR-4B (63). This indicates that free FHR-4B must be in an extremely low concentration or even absent from plasma.

#### Factor H-Related Protein 5

Factor H-related protein 5 is composed of nine CCPs and is the only FHR with domains (CCPs 3–7) homologous to FH CCPs 10–14 (64). FHR-5 CCPs 1–2 are highly similar (85–93% amino acid identity) to CCPs 1–2 of FHR-1 and FHR-2, although not all residues identified in the FHR-1/2 dimerization motif are present in FHR-5 (26). This could explain why the presence of FHR-5 heterodimers *in vivo* is still controversial (26, 52, 53). FHR-5 seems to circulate predominantly as homodimer *in vivo* (53), making quantification a bit more straightforward. FHR-5 serum levels were reported to be 3–6 µg/mL (24), which was later confirmed in 13 healthy individuals, with median levels of 5.5 µg/mL (65). Similar FHR-5 levels (median 2.46 µg/mL) were found in a larger group of 158 healthy controls using the same immunoassay (55). More recently, an average concentration of 1.66 µg/mL was shown in 115 controls by using a newly developed FHR-5 ELISA (53).

#### Other Quantifications

In addition to the specific immunoassays described above, mass spectrometry has also been used to quantify the FHR proteins (66). While this approach allows specific measurement of FHRs based on unique peptide sequences, quantification of FHR dimers is not possible. Results similar to the immunoassays were obtained for FHR-2 (3.64 ± 1.2 µg/mL), FHR-4A (2.42 ± 0.18 µg/mL), and FHR-5 (5.49 ± 1.55 µg/mL). However, much lower levels were found for FHR-1 (1.63 ± 0.04 µg/mL) and FHR-3 (0.020 ± 0.001 µg/mL). It is unclear why such lower concentrations were found for FHR-1 and FHR-3, although the frequency of Δ*CFHR3–CFHR1* in the studied population (*n*= 344, Icelandic origin) was not determined. Of note, the peptide used for FHR-4 quantification is only present in FHR-4A, thus providing no extra information whether FHR-4B exists *in vivo*.

Kopczynska et al. measured FHR-1, FHR-2, and FHR-5 altogether in one immunoassay, finding a total FHR-1/2/5 concentration of 10.67 µg/mL (±5.42) in 42 healthy individuals (67). This result is in great contrast to previously reported levels of approximately 100 µg/mL for FHR-1 (54, 55), but is comparable to a combined mean FHR-1/2/5 concentration of 19.27 (53) and 10.76 µg/mL (66).

The reasons for the huge differences in FHR levels outlined above are unclear. Moreover, the existence of homo- and heterodimers, and the fact that the frequency of the Δ*CFHR3*–*CFHR1* polymorphism is highly population-dependent (19, 68, 69), further complicate the accuracy and assessment of measurements. To exclude any possible cross-reactivity that interferes with FH or FHR quantifications, it is crucial to extensively characterize antibodies generated against FH or any of the FHRs. FH immunoassays should ideally use at least one antibody targeting an epitope located in domains absent from the FHRs, such as CCPs 15–17. Furthermore, when quantifying FHR proteins, it is highly recommended to stratify protein levels based on *CFHR* CNPs, as well as distinguishing between heteroand homodimers. This would aid in comparison of control and patient groups, as CNP frequencies and dimer formation might be altered in patients. CNPs should be determined at the genetic level, as stratification based only on protein levels seems not to be possible due to the wide range in protein concentration within each CNP group (53, 58). CNPs are most commonly determined using multiplex ligation-dependent probe amplification (MLPA), although there is currently no commercial kit available that also covers *CFHR4*. In addition, while normal levels of FHR proteins are now being reported, further data are necessary to reach consensus on their actual concentrations in circulation.

#### LIGANDS OF FH AND THE FHR PROTEINS AND THEIR RELEVANCE

As outlined above, FH is a major inhibitor of the alternative pathway in plasma and when bound to cells and surfaces like the glomerular basement membrane. This complement regulatory activity is due to the interaction of FH with C3b (70). In addition, FH binds to several other ligands (**Figure 2**) and, when ligand-bound, in many cases maintains its complement inhibitory activity. These FH interactions ensure proper regulation of complement activation, as well as the resulting opsonization and inflammation.

Complement activation can be initiated on modified, dangerous self surfaces, which are recognized by PRMs within (C1q, ficolins, MBL, and properdin) and outside the complement system (e.g., pentraxins). FH along with other regulators may ensure targeted but restricted complement activation and an optimal degree of opsonization, while preventing overt inflammation and damage resulting from cascade over-activation (71, 72). The FHR proteins appear to counter-balance this activity of FH and enhance complement activation by binding to the same or similar ligands and outcompeting FH (**Figure 3**), and in some instances also by interacting with C3b and other ligands independent of FH (31). This section briefly summarizes the main ligand interactions of FH and the FHR proteins (**Figure 2**), and indicates their relevance in the regulation and modulation of complement activation.

We would like to briefly note that tumor cells and microbes can bind FH in an attempt to avoid their destruction by host complement. In addition, the main microbial ligand binding sites of FH are in CCPs 6–7 and 19–20, and homologous domains are conserved in the FHRs, thus these proteins may modulate opsonization/killing of microbes. These aspects have been reviewed in detail (6, 31, 73–75).

#### C3b

The main ligand of FH is the active C3 fragment C3b, which can be generated by fluid phase and surface-bound C3 convertases. Since C3b is the central component that promotes complement amplification *via* the alternative pathway, and is also required for the assembly of C5 convertases and the initiation of the terminal pathway, its regulation is key to maintain the proper balance of complement activation and inhibition. FH interacts with C3b at two main sites, harbored by CCPs 1–4 and 19–20 (76). The N-terminal C3b binding site is active when FH is in the fluid phase (e.g., in blood plasma) and also when FH is bound to cells or other surfaces [*via* glycosaminoglycans (GAGs), sialic acid, or a specific receptor—see below] (**Figure 3A**). FH may also bind C3b by CCPs 1–4 when already bound to other ligands, such as pentraxins, because these interactions typically involve CCPs 6–7 and 19–20 (74, 77–79). Thus, FH maintains its complement regulatory activity when bound to cells or other ligands.

Structural studies revealed that FH engages surface-deposited C3b in the context of host GAGs/sialic acid, i.e., CCPs 19–20 bind to these ligands at the same time, which allows avid interaction of FH with a host surface under complement attack. The FH C-terminal site also binds C3d, the final C3b degradation product that remains covalently attached to the surface (80, 81).

The FHR proteins also bind to C3b, but the nature of these interactions is inherently different from that of FH because FHRs lack domains homologous to FH CCPs 1–4. Thus, FHRs lack FH-like cofactor activity and decay accelerating activity, although some residual activity may be present due to the interaction of the C-terminal domains of these proteins with C3b. This should be investigated in detail in the future to clarify the currently contradicting reports in this regard (20, 23, 24, 26, 27, 32, 33).

In contrast to possible inhibitory activities, FHR-1, FHR-4, and FHR-5 were reported to activate the alternative pathway, by

relative FH surplus, FH can potently regulate complement activation on the surface. (F) Increased FHR levels and/or (G) ligand densities, and (H) formation of higher order oligomers (e.g., due to duplicated dimerization domains) can cause enhanced competition with FH and tip the balance to increased complement activation.

binding C3b through their C-terminal domains and forming a platform for the assembly of an active C3bBb convertase (27, 32, 33). This activity could take place on surfaces where these FHRs are bound directly, or *via* another ligand, such as pentraxins (33). FHR-1 and FHR-5 were shown to enhance complement activation on the extracellular matrix (ECM) and on the surface of apoptotic or necrotic cells (32, 33, 82).

Additionally, FHRs may compete with FH for binding to C3b deposited on surfaces, a process termed complement deregulation, because FHRs can enhance complement activation by inhibiting FH binding (**Figure 3D**). This activity of the FHRs may only be significant—considering their relative serum concentrations and avidity for C3b—if increased amounts of FHRs or altered FHR forms (such as higher order oligomers) are present (**Figures 3E–H**) (25, 26, 52, 83, 84). For FHR-2, it was described that, despite binding to C3b, it cannot effectively compete with FH for binding to surface-bound C3b (20).

Altogether, based on these data the FHRs can be regarded as positive complement regulators.

# Other C3 Fragments

While interacting sites for other C3 fragments were described, current evidence strongly supports the physiologically relevant binding of FH to C3b *via* CCPs 1–4 and 19–20, as well as to C3d *via* CCPs 19–20 (76). Interaction of FHR-1 and FHR-2 with C3d was also shown, but without functional analyses (20, 84). Binding of FHR-3, but not of FHR-1, to C3d was shown to prevent the binding of C3d to its receptor on B cells, thus modulating B cell activation (85). FHR-5 was reported to bind to iC3b and C3d with affinities similar to C3b; in contrast, FH bound very weakly to iC3b and C3d compared with FHR-5, indicating that despite its lower serum concentration FHR-5 can be an efficient competitor of FH for binding to deposited C3 fragments (26).

# Glycosaminoglycans (GAGs), Sialic Acid, and Heparin

Distinction between self, non-self, or altered self surfaces relies in part on the recognition of host-specific GAGs and sialic acid by FH (and FHL-1). This allows complement activation to proceed unhindered on microbial ("activator") surfaces, but prevents activation on host ("non-activator") surfaces (86). This has been a subject of intensive research, often using heparin as a model for polyanionic molecules. The main heparin-binding sites were identified in FH (and FHL-1) CCP7 and FH CCP20 (87, 88). This allows recognition of, and attachment to, host glycomatrix and cells, such as platelets and endothelial cells (89, 90). Recent studies revealed some functional differences indicating that while some GAGs are recognized by FH and FHL-1 *via* CCP7, the sialic acid binding site is in CCP20 (91), also targeting these host regulators to different surfaces and explaining the different consequences of mutations affecting these domains (89, 92, 93).

Factor H-related protein 1 can also bind to host surfaces *via* its FH-homologous C-terminus (22, 29), and FHR-3 binds heparin through CCP2, which is homologous to CCP7 of FH (23, 87). In addition, FHR-5 has a heparin-binding site in CCPs 5–7 (24, 94). The functional relevance of these interactions needs to be investigated further, but they could anchor these proteins on certain cells and surfaces.

#### ECM as a Non-Cellular Surface

Extracellular matrices occur in many tissues and can have different functions, the most important ones being the physical support of cells and acting as barriers and filters. The composition of ECMs differs at distinct anatomic sites and is dynamically regulated. Under certain conditions, e.g., endothelial cell activation or injury, ECMs can be exposed to body fluids and plasma proteins; in addition, the Bruch's membrane in the eye and the kidney glomerular basement membrane are also exposed because the lining cell layer is fenestrated. To prevent overt complement activation, such ECMs rely largely on soluble complement regulators, such as FH and FHL-1, which can bind *via* their GAG binding sites and locally regulate complement (95). As noted above, differences in ECMs and in domain composition of the FH family proteins may target FH and FHRs toward distinct sites, such as FH to the glomerular basement membrane (*via* CCPs 19–20) and FHL-1 to the Bruch's membrane (*via* CCP7) (95). FHR-5 was shown to bind to MaxGel, an ECM extract, and de-regulate complement on this surface (32); a recent study identified laminin as an ECM ligand of FHR-5 (94).

#### Ligands on Dead Cells

Complement is largely involved in the immunologically safe and silent disposal of apoptotic and necrotic cells *via* opsonophagocytosis (96). The soluble regulators FH and C4b-binding protein bind to dead cells and prevent excessive complement activation and potential deleterious effects when membrane-anchored regulators are down-regulated on the cells (97). FH can bind to Annexin-II, DNA, and histones (98), as well as malondialdehyde epitopes on apoptotic cells (94, 99). In addition, the pentraxins CRP and pentraxin 3 (PTX3) also bind to dead cells and recruit FH (77, 100). For FHR-1 and FHR-5, binding to necrotic cells and enhancement of complement activation have been shown (33, 82), suggesting that these FHRs modulate opsonization of dead cells.

#### Pentraxins

The pentraxins are soluble PRMs of the innate immune system and, based on their structure, categorized as short and long pentraxins. Pentraxins have numerous ligands and functions, reviewed in detail elsewhere (101); of note, they participate in the opsonization of microbes and dead cells, and they also bind to components of the ECM. For the prototypic short pentraxin CRP and the long pentraxin PTX3, interactions with both complement activators (C1q, MBL) and inhibitors (FH, C4b-binding protein) were described (74, 77, 79, 101–107).

C-reactive protein circulates in its native, pentameric form (pCRP) in body fluids, but it can adopt an altered conformation exposing neoepitopes upon pH change or binding to membranes, and it can even decay to its monomeric form (mCRP) *in vitro* by chelation of the Ca2<sup>+</sup> ions or adsorption on plastic. FH was described to bind primarily to mCRP *via* CCPs 7, 8–11, and 19–20 (79, 108, 109), but interaction with pCRP *via* CCPs 7 and 19–20 at acute phase concentrations was also reported (110). The binding to mCRP allows targeting of the complement inhibitor FH to certain surfaces, including apoptotic cells (71, 100, 109). Among the FHRs, FHR-1 binds to mCRP *via* CCPs 4–5 (33) and FHR-5 *via* CCPs 5–7 (24, 32). The FHR-1/ mCRP interaction enhanced classical and alternative pathway activation, and FHR-5 efficiently competed with FH for mCRP binding, resulting in enhanced complement activation on mCRP (32, 33). In contrast, FHR-4 binds to pCRP *via* CCP1, and this interaction results in enhanced classical pathway activation (111, 112).

PTX3 forms a complex, octameric structure stabilized in part by covalent bonds (113). PTX3 binds to FH *via* CCPs 7 and 19–20, and recruits it to apoptotic cells to downregulate complement activation (77). PTX3 also binds to FHR-1 (weaker than FH) and FHR-5 (stronger than FH); FHR-5 competes with FH and enhance complement activation on PTX3 (32, 33, 74).

#### Malondialdehyde Epitopes

Malondialdehyde (MDA) and malondialdehyde-acetaldehyde (MAA) adducts of proteins and lipids may be generated upon oxidation as oxidation-related neoepitopes, and induce inflammatory responses. FH was shown to bind to MDA/ MAA epitopes and inhibit complement activation and the proinflammatory effects of such MDA/MAA epitopes (99). Two binding sites, within CCP7 and CCPs 19–20 of FH, were identified to bind to MDA/MAA epitopes (99, 114). Recently, FHR-5 was also shown to bind to MAA epitopes (MAA-BSA) *via* CCPs 5–7 and to compete with FH for MAA-BSA binding, thus increasing complement activation. In addition, binding of FHR-5 to necrotic cells was mediated by the same domains, possibly in part *via* the MDA/MAA epitopes that appear on dead cells (94).

# Other, Less Characterized Ligand Interactions of FH

Factor H binds to other ligands that are implicated in certain diseases, particularly in the thrombotic microangiopathy atypical hemolytic uremic syndrome (aHUS). One of these ligands is thrombomodulin, a transmembrane glycoprotein present in endothelial cells, which is involved in the regulation of coagulation and inflammation; thrombomodulin soluble fragments can also be released upon endothelial cell activation or injury. Thrombomodulin was shown to bind to FH and the FH–C3b complex with nanomolar affinity and to enhance FH cofactor activity, which would be reduced in the case of thrombomodulin mutations in aHUS (115–117). These data suggest a role for thrombomodulin in inhibiting alternative pathway activation locally *via* its interaction with FH, but thrombomodulin was also found to inhibit complement hemolytic activity in a FH-independent mechanism (116). An additional, complement-activating function of thrombomodulin by enhancing C3 cleavage into C3b has also been described (117).

Similarly, binding of von Willebrand factor (vWF) to FH enhances FH cofactor activity and also modulates the vWF prothrombotic status (118–120). FH was found co-localized with vWF in the Weibel–Palade bodies in human umbilical vein endothelial cells, and the complex was also detected in human plasma. Purified FH and vWF were shown to interact with nanomolar affinity, and to influence their respective functions; vWF enhanced the cofactor activity of FH, whereas FH inhibited ADAMTS13-mediated cleavage of vWF and facilitated platelet aggregation (120). However, another investigation found that FH binds *via* its C-terminus to the vWF A2 domain, and enhances its cleavage by ADAMTS13 (118). FH was also reported to reduce large soluble vWF multimers (119). Thus, further studies are needed to clarify the functional relevance of the complex interaction between FH and vWF, and its potential role in disease.

Recently, hemolysis-derived heme was shown to activate the alternative pathway in serum and on endothelial cells, and to bind both C3 and FH. Heme-exposed C3 and endothelial cells displayed increased FH binding, and FH was shown to be a major serum factor that regulates C3 deposition on heme-treated endothelial cells (121).

Factor H was also reported to bind to apolipoprotein E *via* domains CCPs 5–7, and to regulate alternative pathway activation on high density lipoprotein particles (122). Complement regulation by FH on such lipoprotein particles could be potentially impaired in diseases characterized by immune deposits containing also apolipoprotein E, such as age-related macular degeneration (AMD) and dense deposit disease (DDD) (122).

In addition, FH binds to myeloperoxidase (MPO) released from activated neutrophil granulocytes, and FH and MPO colocalize in neutrophil extracellular traps. Interestingly, the binding site for MPO in FH was determined to be CCPs 1–4 and, thus, MPO inhibited FH binding to C3b, as well as FH decay accelerating activity and cofactor activity (123).

# Binding to Cellular Receptors—Non-Canonical Roles of the FH Family Proteins

Factor H and some of the FHRs can also bind to cells *via* specific receptors, and may modulate the cell activation and response, as well as inflammatory processes. These aspects are reviewed in detail elsewhere (124); here, we summarize only some major FH/FHR-receptor interactions and their role, particularly those described very recently.

Complement receptor type 3 (CR3; CD11b/CD18; or integrin αMβ2) was identified as a main FH receptor on neutrophils and macrophages (125, 126). FH maintains its cofactor activity when receptor bound, but it also directly affects cellular functions, such as adhesion, cell spreading, migration, and cytokine production (125–127). Interestingly, FH was able to inhibit the release of extracellular traps by human neutrophils activated with immobilized fibronectin plus fungal β-glucan, or with phorbol 12-myristate 13-acetate (127). FH can also enhance the interaction of certain pathogens with human macrophages and neutrophils, and modulate the response of the phagocytes (128, 129). This was also shown for FHR-1 which, by binding to CR3, could enhance neutrophil responses to *Candida albicans* (129). In addition, FHL-1 was shown to mediate cell adhesion and spreading (129, 130).

Described functional effects of FH on monocytes include enhancement of IL-1β secretion, respiratory burst, and chemotactic effect (131–134). FH was shown to induce an antiinflammatory and tolerogenic phenotype in monocyte-derived dendritic cells *in vitro* (135). Very recently, in the context of inflammation in AMD, FH, and its two variants Y402 and H402 were investigated in a mouse model. FH was shown to inhibit the resolution of inflammation by binding to CR3 and thus blocking thrombospondin-1–CD47 signaling that would normally promote the elimination of macrophages. The AMD-associated H402 FH variant displayed a stronger inhibitory effect compared to FH Y402, causing increased accumulation of macrophages in the inflamed tissue (136).

Factor H was shown to bind to B cells and may modulate some B cell functions, such as proliferation and immunoglobulin secretion, but no specific receptor has been identified to date (137–140). A recent report described an indirect modulation of B cell activation by FHR-3, which was shown to bind to C3d and inhibit its binding to complement receptor type 2, a coreceptor of the B cell receptor complex; FH and FHR-1 had no such effect (85).

These non-canonical functions of the FH family proteins deserve further investigation, because they may play roles in inflammation and anti-microbial defense that are currently underappreciated. Clarification of their cell-mediated effects may provide additional insights into disease mechanisms.

#### DISEASE ASSOCIATIONS

Studies in patients and controls have shown a variety of common *CFH/CFHRs* genetic variants that predispose to autologous damage, which is predominantly organ-specific. Prevalent kidney damage occurs in the rare diseases aHUS and C3 glomerulopathies (C3G), and in the more frequent IgA nephropathy (IgAN), while destruction of the retinal pigment epithelium by autologous complement contributes to AMD. A defective regulation of complement activation on the renal microvasculature endothelium occurs in aHUS, while in C3G uncontrolled complement activation in plasma gives rise to massive deposition of C3b breakdown products (iC3b, C3dg, and C3c) in the glomeruli (141–143). IgAN is characterized by mesangial cell proliferation and hypoglycosylated IgA1 deposits in the glomeruli, and it is likely that complement defects contribute, at least in part, to its clinical heterogeneity (144, 145). A defective control of complement activation in the retina is most relevant in AMD pathogenesis, and enhances the inflammatory response (146).

Extremely rare and pathogenic *CFH/CFHRs* variants have been mainly found in aHUS and C3G patients. Some of these variants result from gene conversion events between *CFH* and *CFHR1*, and they give rise to mutated FH or FHR-1. Other variants are intragenic duplications or hybrid genes resulting from gene rearrangements, and generate abnormal proteins; some of these proteins have distinct molecular weights and can be detected by Western blot analysis. It is interesting that abnormal rearrangements involving FH/FHRs associate with aHUS, while in C3G patients only FHR proteins are affected.

#### *CFH* Variants Associated With Renal or Ocular Damage

Common SNPs in *CFH* give rise to different haplotypes that can be disease neutral, predisposing, or protecting. Thus, haplotype *CFH(H1)* predisposes to membranoproliferative glomerulonephritis (MPGN) and AMD, haplotype *CFH(H3)* predisposes to aHUS, and haplotype *CFH(H2)* is protective against these three diseases (147–149). Haplotype *CFH(H2)* generates the FH62Ile variant, which shows increased binding to C3b and cofactor activity in the fluid phase and on cellular surfaces (150), thus favoring protection against autologous complement damage.

The common variant FH402His, which is present in FH and its shorter isoform FHL-1, is a major predisposing factor in AMD (151). The functional relevance of FH402His in C3b, CRP, or heparin binding has been analyzed in several studies. Reduced binding of FH402His to polyanionic surfaces has been found (152), but the pathogenic mechanism may also depend on FHL-1, which can regulate complement activation similarly to FH. It has been shown that FHL-1, but not FH, is present in the retinal Bruch's membrane, a major target in AMD pathogenesis, and that binding of the AMD-FHL-1402His variant was lower than binding of the FHL-1402Tyr variant (93). Nonetheless, *CFH* intronic variants show stronger association with AMD than FH402His (153). In an analysis of seven common *CFH* haplotypes, haplotypes H1, H6, and H7 were found to confer increased risk to AMD; these haplotypes share a 32-kb region downstream of rs1061170 (FH Tyr402His) that must be critical for AMD development (19), and that includes a 12-kb block 89% similar to a noncoding region in CNP148 (see below).

Other disease-predisposing FH variants are very rare. One of the most relevant is Arg1210Cys (FH1210C), which was initially identified in aHUS patients (154), and shown to be covalently bound to albumin in plasma (155); the presence of albumin most likely prevents the interaction of FH1210C with its physiological ligands, generating a partial, pathogenic FH deficiency. FH1210C has been also associated with C3G (156), and it highly increases AMD-risk and predisposes to early disease onset (157, 158). It has been suggested that in individuals with the FH1210C variant it is the concurrence of other genetic predisposing factors what ultimately determines the clinical phenotype (159).

#### *CFHR1* and *CFHR3* Variants Associated With Renal or Ocular Damage

As happens with the common *CFH* haplotypes, the two main *CFHR1* alleles show differential disease associations. *CFHR1\*B*, displaying increased similarity with *CFH*, increases aHUS risk (51), and *CFHR1\*A* predisposes to AMD (160). The molecular bases for these associations have not been determined, but they will most likely depend on subtle functional differences among the FHR-1\*A and FHR-1\*B allotypes. *CFHR1\*A* is in strong linkage disequilibrium with the AMD-risk *CFH402His* allele, and *CFHR1* genotyping has similar predictive value of developing AMD as *CFH402His;*Δ*CFHR3–CFHR1* genotyping (160); these findings are suggestive of a direct role of FHR-1 in AMD pathogenesis, most likely by interfering with the interaction of FH with specific ligands and promoting complement activation (33).

The *CFHR3* gene also has two major variants, *CFHR3\*A*, more frequent in healthy controls, and *CFHR3\*B*, which predisposes to aHUS but not to C3G (59). Because the aHUS risk *CFHR3\*B* allele generates higher FHR-3 levels than the non-risk *CFHR3\*A* allele (60), it seems that increased competition of FHR-3 and FH for certain ligands could favor aHUS development. FHR-3 is also produced in the retina, and its contribution to retinal degeneration by inhibiting FH binding to C3b and modified surfaces has been suggested (47); nonetheless, the relevance of the *CFHR3\*A* and *CFHR3\*B* variants in AMD has not been addressed.

The two CNPs in the *CFHR* genes have been shown to be disease-relevant (19). The common variant Δ*CFHR3–CFHR1* is protective against AMD (161), and IgAN (162), but it predisposes to aHUS (163) and to systemic lupus erythematosus (SLE) (69) because it is associated with generation of anti-FH autoantibodies (discussed on page 13). The rare variant Δ*CFHR1–CFHR4* was initially identified in a few aHUS patients with anti-FH autoantibodies (51), and is present in 1.4% of aHUS patients and 0.9% of controls (164).

The protective effect of the Δ*CFHR3*–*CFHR1* haplotype against AMD was first described in 2006 (161), and it is the more common copy number variation in the *CFH/CFHRs* region (165). Δ*CFHR3*–*CFHR1* is tagged by *CFH* rs6677604A with 99% accuracy (166), and strongly correlates with the 86.4-kb deletion CNP147 and high FH levels (8). Because protection conferred by Δ*CFHR3*–*CFHR1* was independent of the FH Tyr402His polymorphism, a direct effect of FHR-1 and FHR-3 in AMD pathogenesis was suggested (21). Nonetheless, the strong association of Δ*CFHR3*–*CFHR1* with high FH levels, together with the finding that FHR-1 levels were lower in AMD patients than in control individuals, suggests that Δ*CFHR3*–*CFHR1* is actually tagging an allele expressing high FH levels, but it is not causal in protection against AMD (8). The much rarer Δ*CFHR1*–*CFHR4* deletion (also referred to as CNP148) also confers protection against AMD independent of SNPs in *CFH* (19); because Δ*CFHR1*–*CFHR4* also removes non-coding flanking regions, its protective effect against AMD could either be due to the reduction of FHR-1 and/or FHR-4A levels, or to the absence of regulatory regions relevant for disease pathogenesis.

The first evidence for a direct complement role in IgAN pathogenesis was the finding that the common variant Δ*CFHR3*–*CFHR1* protects against IgAN when it is in homozygosis (162), pointing out to a possible role of FHR-3 and/or FHR-1 levels in the pathogenic mechanism. However, because the Δ*CFHR3*–*CFHR1* allele generates high FH levels which associate with lower mesangial C3 deposition, the actual contribution of FHR-1 and/or FHR-3 levels to IgAN is unclear (167). Two studies in different IgAN cohorts have recently shown that FHR-1 levels and FHR-1/FH ratios are increased in patients with disease progression, thus providing evidence for a direct role of FHR-1 in the disease mechanism. One of these studies reported that high FHR-5 levels were also slightly elevated in the IgAN patients, but without any correlation with progressive disease (55). The other study also reported low FH levels associated with *CFH* or *CFI* mutations in a few IgAN patients (54).

### FH::FHR-1, FHR-1::FH, and FH::FHR-3 Hybrid Proteins Associate With aHUS

*CFH* exons 18–20 and *CFHR1* exons 4–6 have a high degree of sequence similarity, that result in only five amino acid difference between CCPs 18–20 of FH (Y1040-V1042-Q1058-S1191-V1197) and CCPs 3–5 of FHR-1 (H157-L159-E175-L290-A296). Studies in aHUS patients have revealed that these differences determine higher binding of FH than FHR-1 to cell surfaces. Amino acids S1191 and V1197 in FH seem to be particularly important, and single mutations involving these amino acids (FHS1191L and FHV1197A) have been found in a number of aHUS patients from different geographical origins. The double mutant (FHS1191L-V1197A) was observed in two unrelated aHUS patients with early disease onset, showed a defective capacity to control complement activation on cellular surfaces, and had been generated by gene conversion (168).

FHS1191L-V1197A can also be generated by NAHR events that give rise to *CFH::CFHR1* hybrid genes. A *CFH(Ex1*–*21)::CFHR1(Ex5*–*6)* hybrid gene was first described in a family with many cases of aHUS along several generations, and a clinical history of disease recurrence in affected individuals (169), demonstrating that FHS1191L-V1197A is highly pathogenic. This hybrid gene has also been found in other non-related aHUS patients. A slightly different *CFH(Ex1*–*22):: CFHR1(Ex6)* gene which also generates FHS1191L-V1197A has been found in another patient with a prompt aHUS onset (170). The reverse situation (i.e., the existence of *CFHR1::CFH* hybrid genes) has also been reported. A *CFHR1(Ex1*–*3)::CFH(Ex19*–*20)* hybrid gene generated by "*de novo*" NAHR was identified in one sporadic case of aHUS (171), and a *CFHR1(Ex1*–*4)::CFH(Ex20)* hybrid gene was found in a family with two members affected with aHUS (172). These two *CFHR1::CFH* hybrid genes generated a double-mutated FHR-1 protein that carries the homologous amino acids in FH CCP20 domain (FHR-1L290S-A296V); these amino acids most likely confer the mutated FHR-1 increased competition with FH for endothelial cell binding, and result in reduced protection against complement damage (173). Screening of *CFH::CFHR1* and *CFHR1::CFH* hybrid genes is normally done by MLPA analysis of copy number variations. The *CFH::CFHR1* alleles lack a normal copy of *CFHR3* and *CFHR1*, while the *CFHR1::CFH* allele contains two copies of *CFHR3*; it cannot be ruled out that these additional factors also contribute to the pathogenic mechanism.

A FH::FHR-3 hybrid protein containing CCPs 1–19 of FH and the five CCPs of FHR-3 was identified in a large family with aHUS (174). This protein resulted from an abnormal rearrangement that deleted the last exon of *CFH*, which was then fused to the adjacent *CFHR3* gene by the genetic mechanism microhomology mediated end joining (MMEJ). The absence of FH CCP20 domain in the hybrid protein and/or the presence of the FHR-3 CCPs does not affect complement regulation in the fluid phase, but cellular surface regulation seemed to be highly reduced. Estimation of aHUS penetrance in carriers of the hybrid gene is 33%. Another FH::FHR-3 hybrid protein containing CCPs 1–17 of FH and the five CCPs of FHR-3 was found in an aHUS patient with a very early disease onset (175). The hybrid protein resulted from a *"de novo"* 6.3 kb-deletion of exons 21–23 of the *CFH* gene through a MMEJ mechanism, and it showed impaired cell surface complement regulation.

### Abnormal FHR Proteins in C3G

The abnormal rearrangements that predispose to C3G thus far described involve exclusively the *CFHR* genes, but not the *CFH* gene. This is a distinctive feature from aHUS that suggests a more important contribution of FHRs in the protection of the glomerular basement membrane and mesangium than in protection of endothelial cells. Abnormal rearrangements include intragenic duplications in *CFHR1* or *CFHR5*, and *CFHR2::CFHR5* and *CFHR3::CFHR1* hybrid genes.

#### FHR-5 and FHR-1 Proteins With Additional Dimerization Domains

Larger forms of FHR-1 and FHR-5 with duplicated dimerization domains have been observed in a few C3G patients. These proteins circulate in plasma together with the normal FHR-1 and FHR-5 proteins, but disease penetrance in mutation carriers is very high, strongly suggesting a dominant negative effect of the larger, abnormal protein. This is particularly evident for a partially duplicated FHR-5 protein initially observed in two families of Cypriot ancestry in which renal disease was consistent with autosomal dominant transmission (176). All affected individuals were heterozygous for a *CFHR5* gene in which exon 2 (coding for CCP1) and exon 3 (coding for CCP2) were duplicated, giving rise to an abnormal FHR-5 protein containing two extra dimerization domains (FHR-512123-9). *In vitro* studies with patient's sera showed reduced binding of the FHR-512123-9 to the cell surface, and increased FI cofactor activity, but the relevance of these findings for the pathogenic mechanism is unknown. Patients carrying FHR-512123-9 had a high risk of progressive renal disease, particularly males. This renal phenotype, which histologically corresponds to a C3 glomerulonephritis, is clinically characterized by continuous macroscopic hematuria, and was denominated as "CFHR5 nephropathy." These seminal observations were further extended to 16 pedigrees of Cypriot origin in a study that also provided a thorough description of histological, molecular, and clinical findings (177). Recurrence of CFHR5 nephropathy in a kidney allograft has been reported in one patient, although it did not occur in two other cases (178). The same duplicated FHR-5 protein observed in patients of Cypriot ancestry was found in a familial case of C3 glomerulonephritis with a different ethnic origin (179). Of note, this protein was generated from a different genomic rearrangement, reinforcing the relevance of the duplicated FHR-5 protein for the pathogenic mechanism, and the authors proposed that all patients with clinical suspicion of CFHR5 nephropathy should be screened for the abnormal protein by Western blot.

Another FHR-5 protein with two additional dimerization domains was found in a familial case of C3G with DDD (C3G-DDD) (83). In this family, a genomic 24.8 kb-deletion from intron III of the *CFHR2* gene to *CFHR5* gives rise to a hybrid *CFHR2::CFHR5* gene which generates a so-called FHR-21,2-FHR-5-hybrid protein very similar to the FHR-512123-9 protein previously described. This hybrid protein shows increased binding to C3b and stabilization of the AP C3 convertase, which would explain the low C3 and increased Ba levels detected in the patients' sera; in addition, reduced regulation of the AP C3 convertase by FH will result in increased generation of iC3b molecules which will deposit on the glomerular basement membrane and favor the pathogenic mechanism.

To understand how the presence of two extra dimerization domains in FHR-512123-9 has pathogenic consequences it is necessary to recapitulate that the dimerization domains in FHR-1, FHR-2, and FHR-5 confer these proteins the ability to generate homo- and hetero-dimers physiologically (26). Although a recent study has not found evidence of the presence of FHR-5 heterodimers with FHR-1 or FHR-2 (53), the additional dimerization domains in FHR-512123-9 and FHR-21,2-FHR-5 will most likely give rise to higher order oligomeric forms with increased avidity for surface-bound C3b, and these multimeric proteins will compete more efficiently with FH and favor autologous tissue damage, as illustrated in **Figure 3**. In this context, it is intriguing that two very similar FHR-5 proteins result in different clinical entities (CFHR5 nephropathy or DDD). FHR-512123-9 and FHR-21,2-FHR-5 contain the nine CCPs of FHR-5 preceded by the two dimerization domains of FHR-5 or FHR-2, respectively, that present 85% aminoacid identity. Functional studies with the recombinant forms of these two proteins (referred to as FHR-5Dup and FHR-2-FHR-5Hyb) revealed that they exacerbate local complement activation by recruiting the complementactivating protein properdin, and that properdin binding is mediated by the FHR-5 dimerization domains, and not by the FHR-2 dimerization domains (82). Therefore, local complement activation would be higher in patients with FHR-5Dup than in patients with FHR-2-FHR-5Hyb, and this could explain the different clinical phenotype. Another, non-exclusive, explanation is that the pathogenic mechanism is much dependent on the plasma levels of these FHR-5 proteins. In line with this hypothesis, Western blot analyses of patient serum samples showed that the FHR-5 band has similar intensity to a normal serum, while the intensity of the FHR-512123-9 (32) or the FHR-21,2-FHR-5 band is much higher, suggesting highly increased levels (82). The latter study also showed that FHR-5 binds to necrotic human endothelial cells, but not to normal endothelial cells, strongly suggesting a role for FHR-5 in complement-mediated elimination of damaged cells.

An abnormal, large FHR-1 protein was identified in a Spanish family with C3G by Western blot analysis (52). This protein was generated by an internal duplication of the *CFHR1* gene, and contains two copies of domains CCPs 1–4. Purification of the normal and the duplicated FHR-1 proteins allowed biochemical, functional, and structural studies that illustrated that normal FHR-1 circulates in plasma as homo- and hetero-oligomers (with FHR-2 and FHR-5), and that the duplicated FHR-1 (containing nine CCP domains) organized into much larger oligomers with increased binding to C3b, iC3b, and C3dg. These findings provided the first evidence for the existence of oligomeric forms of FHR-1, FHR-2, and FHR-5 in normal plasma, and confirmed that duplication of their homologous CCPs 1–2 is pathogenic and associates with C3G. The authors proposed that multimerization of FHR-1 strongly inhibits FH binding to certain cell surfaces, but not to endothelial cells, the target surface in aHUS. A different FHR-1 protein containing two copies of domains CCPs 1–2 has been described in another Spanish patient with a C3G clinical phenotype, but further characterization of this duplicated FHR-1 (containing seven domains) has not been provided (180).

#### FHR-3::FHR-1 Hybrid Protein

A hybrid *CFHR3::CFHR1* gene associated with C3G-MPGN III has been described in an Irish family (181). This hybrid gene contains exons 1–3 of *CFHR3* and exons 2–6 of *CFHR1*, and generates a protein containing CCPs 1–2 of FHR-3 followed by the five CCPs of FHR-1. The protein was detected in the patients' plasma by Western blot, and it was apparently at a much lower concentration than normal FHR-1. Because patients with the *CFHR3::CFHR1* gene also has two copies of *CFHR3* and *CFHR1*, the authors propose a dominant effect of the hybrid FHR-3::FHR-1 protein in the pathogenic mechanism. It is of note that plasma C3 levels in all affected individuals were normal, as opposed to the reduced levels observed in the C3G-DDD patient with FHR-21,2-FHR-5-hybrid protein (83); this fact suggests that the potential pathogenic effect of FHR-3::FHR-1 on complement activation or regulation is surface-restricted. The clinical data and outcome of the five patients from this family who received renal transplantation has been reported (182); disease recurrence in the kidney allograft was high, but the overall graft survival was good.

#### Anti-FH Autoantibodies Predispose to Renal Diseases

Disorders related with these autoantibodies are mainly present in aHUS and C3 glomerulonephritis patients and secondary in other autoimmune diseases. The anti-FH autoantibodies cause a functional FH defect, resulting in impaired complement regulation by FH (**Figure 3C**).

The existence of anti-factor H autoantibodies in aHUS and the resulting functional deficiency of FH were first described in 2005 (183). The frequency of the anti-FH antibodies associated with aHUS is approximately 10% of the pediatric patients in the European series and occasionally in patients with adult onset (184). These autoantibodies form complexes with FH and induce a functional FH deficiency. Characterization of these autoantibodies showed that they recognized the C-terminal region of FH, involved in the binding to cell surfaces (185, 186). Moreover, it has been shown that, especially in the acute phase, these antibodies are also capable of blocking the activity of FH as cofactor of FI and the acceleration of the dissociation of the convertases of the alternative pathway (187).

The presence of anti-FH autoantibodies is associated with homozygous Δ*CFHR3–CFHR1* in several aHUS cohorts (51, 164, 188, 189). The Δ*CFHR1–CFHR4* has also been found in a few patients (51, 164), suggesting a relevant role for the absence of FHR-1 in autoantibody generation. In this context, it has been found that most anti-FH autoantibodies also bind to FHR-1, which presents high similarity with FH CCPs 19–20 (29, 164).

The anti-FH autoantibodies in aHUS patients are able of forming immune complexes that can be detected in serum. The amount of these complexes correlates better with the clinical evolution than the total autoantibody titer (187), because FH bound to the complexes cannot regulate the AP on cell surfaces. The use of two monoclonal antibodies binding to different parts of FH allowed the quantitation of total and free FH, which depends on the concentration of circulating anti-FH immune complexes (29, 186, 190). In some cases, the concentration of total FH was within the normal range, but the amount of free FH was practically undetectable, indicating that the anti-FH autoantibodies almost completely blocked the ability of FH to protect cell surfaces from complement activation, although its regulatory activity in the fluid phase was conserved (190).

The epitope recognized by anti-FH autoantibodies has been defined more precisely using recombinant fragments of CCPs 19–20 containing point mutations (191). In this work, it was found that in patients with FHR-1 deficiency, anti-FH antibodies recognize a region that acquires a different conformation in FH and FHR-1 after binding to certain ligands, including various bacterial proteins. This suggests a model in which the absence of FHR-1 plays a role in the loss of tolerance to FH and in the generation of anti-FH autoantibodies, thus explaining the frequent association between the presence of anti-FH antibodies and homozygous Δ*CFHR3–CFHR1* in aHUS. By using the same mutated FH recombinant fragments in our series of patients with anti-FH autoantibodies, we have obtained concordant results, at least in the patients with FHR-1 deficiency, which supports the proposed model for the generation of autoantibodies in these patients (192). However, the mechanism of anti-FH autoantibody generation in aHUS patients without FHR-1 deficiency remains to be determined.

Anti-FH autoantibodies have also been described in patients with C3G (193–197). This association is much less frequent than in the case of aHUS despite having been described for the first time (196). In cases in which the effect of these anti-FH autoantibodies has been studied, it has been shown that they inhibit the regulatory activity of FH by recognizing and blocking its N-terminal region (193, 194, 197), which is a difference with the anti-FH autoantibodies from aHUS patients.

In patients with SLE and other autoimmune diseases, a greater frequency of anti-FH autoantibodies has been described with respect to healthy controls (198). Unlike the anti-FH autoantibodies present in aHUS, the epitopes that are recognized by the autoantibodies seem to be distributed throughout the entire protein, and they are not associated with FHR-1 deficiency.

#### CONCLUSION

The FH protein family remains an intriguing group of proteins. FH is well-known for its protecting role against self-damage from complement, and the FHRs are emerging as FH antagonists that act as an additional regulatory mechanism to control where and when FH protects human cells and/or surfaces. With the recent development of FHR-specific assays, quantification of the whole protein family has now become possible. This has elucidated the intricate balance between FH and the FHR proteins, showing that overall the balance is in favor of FH. However, this balance can shift on altered self, and also genetic variations have a major impact on FH and FHRs. This includes decreased FH function due to mutations, altered expression levels, as well as hybrid FH::FHR and FHR::FHR proteins and unusual FHR multimers with abnormal function that disturb complement regulation. Associations of increased FHR levels, as a result of genetic variations, with diseases like aHUS and IgAN are highly suggestive of a pathological role for the FHRs. It remains to be seen whether the FHRs are indeed causative in these diseases, but it is likely that they at least contribute to altered complement regulation on host surfaces.

# AUTHOR CONTRIBUTIONS

PS-C, RBP, ML-T, and MJ prepared the text and the figures. All authors have revised and approved the manuscript.

# FUNDING

PS-C and ML-T are funded by grants PI16/00723 and PI15/00255 (Spanish Ministerio de Economía y Competitividad/ISCIII, and European Program FEDER) and B2017/BMD3673 (Complement II-CM network from the Comunidad de Madrid). MJ is supported by the National Research, Development and Innovation Fund of Hungary (NKFIA grants K 109055 and K 125219), the Kidneeds Foundation, Iowa, US, and by the Institutional Excellence Program of the Ministry of Human Capacities of Hungary.

### REFERENCES


heterodimers, while FHR-5 circulates only as homodimer in human plasma. *Front Immunol* (2017) 8:1328. doi:10.3389/fimmu.2017.01328


differentiation of immune responses. *Nat Struct Mol Biol* (2017) 24(8):643–51. doi:10.1038/nsmb.3427


carboxy-terminal glycosaminoglycan binding site. *Am J Pathol* (2005) 167(4): 1173–81. doi:10.1016/S0002-9440(10)61205-9


with atypical hemolytic uremic syndrome. *Blood* (2010) 115(2):379–87. doi:10.1182/blood-2009-05-221549


of autoimmunity in this disease. *J Biol Chem* (2015) 290(15):9500–10. doi:10.1074/jbc.M114.630871


**Conflict of Interest Statement:** RBP is co-inventor of a patent describing potentiating anti-FH antibodies and uses thereof. The other authors declare no conflict of interest.

*Copyright © 2018 Sánchez-Corral, Pouw, López-Trascasa and Józsi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

,

# Complement Factor H and Apolipoprotein E Participate in Regulation of Inflammation in THP-1 Macrophages

#### Edited by:

Eija Nissilä<sup>1</sup>

Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France

#### Reviewed by:

Lubka T. Roumenina, INSERM U1138 Centre de Recherche des Cordeliers, France Harald F. Langer, Universität Tübingen, Germany

#### \*Correspondence:

Karita Haapasalo karita.haapasalo@helsinki.fi

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 21 August 2018 Accepted: 01 November 2018 Published: 21 November 2018

#### Citation:

Nissilä E, Hakala P, Leskinen K, Roig A, Syed S, Van Kessel KPM, Metso J, De Haas CJC, Saavalainen P, Meri S, Chroni A, Van Strijp JAG, Öörni K, Jauhiainen M, Jokiranta TS and Haapasalo K (2018) Complement Factor H and Apolipoprotein E Participate in Regulation of Inflammation in THP-1 Macrophages. Front. Immunol. 9:2701. doi: 10.3389/fimmu.2018.02701 Kok P. M. Van Kessel <sup>2</sup> , Jari Metso<sup>3</sup> , Carla J. C. De Haas <sup>2</sup> , Päivi Saavalainen<sup>1</sup> , Seppo Meri <sup>1</sup> , Angeliki Chroni <sup>4</sup> , Jos A. G. Van Strijp<sup>2</sup> , Katariina Öörni <sup>5</sup> , Matti Jauhiainen<sup>3</sup> , T. Sakari Jokiranta<sup>1</sup> and Karita Haapasalo<sup>1</sup> \*

, Pipsa Hakala1†, Katarzyna Leskinen1†, Angela Roig1†, Shahan Syed1†

<sup>1</sup> Department of Bacteriology and Immunology, and Research Programs Unit, Immunobiology, University of Helsinki, Helsinki, Finland, <sup>2</sup> Medical Microbiology, University Medical Center Utrecht, Utrecht, Netherlands, <sup>3</sup> Minerva Foundation Institute for Medical Research, Helsinki, Finland, <sup>4</sup> Institute of Biosciences and Applications, National Center for Scientific Research "Demokritos," Athens, Greece, <sup>5</sup> Wihuri Research Institute, Helsinki, Finland

The alternative pathway (AP) of complement is constantly active in plasma and can easily be activated on self surfaces and trigger local inflammation. Host cells are protected from AP attack by Factor H (FH), the main AP regulator in plasma. Although complement is known to play a role in atherosclerosis, the mechanisms of its contribution are not fully understood. Since FH via its domains 5–7 binds apoliporotein E (apoE) and macrophages produce apoE we examined how FH could be involved in the antiatherogenic effects of apoE. We used blood peripheral monocytes and THP-1 monocyte/macrophage cells which were also loaded with acetylated low-density lipoprotein (LDL) to form foam cells. Binding of FH and apoE on these cells was analyzed by flow cytometry. High-density lipoprotein (HDL)-mediated cholesterol efflux of activated THP-1 cells was measured and transcriptomes of THP-1 cells using mRNA sequencing were determined. We found that binding of FH to human blood monocytes and cholesterol-loaded THP-1 macrophages increased apoE binding to these cells. Preincubation of fluorescent cholesterol labeled THP-1 macrophages in the presence of FH increased cholesterol efflux and cholesterol-loaded macrophages displayed reduced transcription of proinflammatory/proatherogenic factors and increased transcription of anti-inflammatory/anti-atherogenic factors. Further incubation of THP-1 cells with serum reduced C3b/iC3b deposition. Overall, our data indicate that apoE and FH interact with monocytic cells in a concerted action and this interaction reduces complement activation and inflammation in the atherosclerotic lesions. By this way FH may participate in mediating the beneficial effects of apoE in suppressing atherosclerotic lesion progression.

Keywords: complement, complement system, Factor H, apolipoprotein E, atherosclerosis, inflammation

# INTRODUCTION

Complement (C) system is part of the humoral innate immune response. It quickly attacks microbes and foreign particles invading the human body. Activation of complement through any of the three pathways, the classical, alternative (AP), and lectin pathways, leads to cleavage of C3 and covalent surface deposition of the C3b fragment that is capable of forming an enzyme with factor B. Surface deposited C3b and its fragments, inactive C3b (iC3b) and C3dg, are important opsonins that can be recognized by complement receptors expressed by phagocytes. Further activation of the C cascade leads to formation of membrane attack complexes and release of proinflammatory chemotactic and anaphylatoxic protein fragments C3a and C5a that mediate their effects by binding to receptors on phagocytes.

AP is constantly active in plasma leading to low grade challenge to all plasma-exposed particles and surfaces by spontaneous hydrolysis or enzymatic activation of C3. Activation proceeds rapidly through amplification on surfaces that are missing efficient regulatory mechanisms. Factor H (FH) is the main AP regulator as it keeps this spontaneous activation in control (1). This is obvious since depletion of plasma/serum from FH or blockage of FH by autoantibodies leads to activation of the AP leading to overconsumption and loss of active complement within <30 min (2, 3). FH is an elongated molecule composed of 20 domains. The N-terminal domains 1–4 are responsible for FH regulatory activity while domains 19–20 on the C-terminal end are responsible for the self surface recognition (4). Domain FH19 binds to surface-deposited C3b, while FH20 binds sialic acids and glycosaminoglycans (GAGs) present on self surfaces (5). In this way, FH discriminates host self surfaces from non-self ones. Also domains 6–7 of FH are important for self surface recognition. These mediate interaction with sulphated GAGs, heparin, and C-reactive protein (6, 7). When bound to C3b, the cofactor activity of FH helps in inactivation of C3b to iC3b by factor I and simultaneous inhibition of factor B binding to C3b (8, 9). Moreover, FH prevents further AP amplification by accelerating the decay of formed AP convertases (10). AP activation does not need a trigger as it is based on continuous low-grade activity. However, imbalance between activation and regulation e.g., when numerous C3b deposits are formed on protein complexes can lead to enhanced AP activation in serum/plasma (11). Recently, it has become clear that AP dysregulation is a central event in development of several complement-related diseases to which mutations or polymorphisms in domains FH5-7 and FH19-20 predispose. While mutations in FH19- 20 cause atypical hemolytic uremic syndrome (aHUS), the Y402H polymorphism in domain 7 is associated with age-related macular degeneration (AMD) (12, 13) and dense deposit disease (DDD) (14).

Atherosclerosis is a chronic multifactorial inflammatory disease caused by the subendothelial accumulation of lipids, immune cells and fibrous elements in arteries leading to thickening and hardening of the arterial wall. Low-grade inflammation is a key mediator of the disease. Both adaptive and innate immune responses are crucial for initiation and progression of atherosclerosis, and these mechanisms have been exploited to develop new diagnostic biomarkers and therapies for patients recently (15). The hallmark of early atherosclerotic lesion is the formation of fatty streaks composed of cholesterolladen macrophages, which are formed when low density lipoproteins (LDL) are modified by oxidation or proteolytic modification and accumulated in the subendothelium of arteries leading to monocyte recruitment and differentiation into macrophages (16). The balance between proinflammatory M1 type macrophages and anti-inflammatory M2 type macrophages plays a crucial role in the pathogenesis of atherosclerotic plaques.

The antiatherosclerotic activity of apoE is based on its ability to regulate lipoprotein metabolism and to promote cholesterol efflux from cells (17, 18). Endogenous production of apoE by macrophages in blood vessel walls has been shown to be critical in the prevention and healing of atherosclerotic plaques. Importantly, apoE modulates macrophage polarization into the anti-inflammatory M2 phenotype (17) and promotes reverse cholesterol transport from peripheral cells to high density lipoprotein (HDL) for further transportation of cholesterol to the liver for excretion (19).

Genetic variations in the APOE gene coding for apolipoprotein E constitute important risk factors both for AMD and atherosclerosis. Interestingly, similar underlining mechanisms including disturbances in lipid metabolism, oxidative stress and the inflammatory process are closely associated in the pathogenesis of both diseases. It has also been shown that in human eyes with AMD, FH co-localizes with and binds to oxidized lipids in drusen, fatty deposits under the retina. It seems that the common FH variant 402Y has a higher affinity for oxidized lipids than the risk allele 402H suggesting a stronger FH-mediated complement inhibition of the effects of oxidized lipids on macrophages (20).

We have shown before that FH binds both lipid-free and high density lipoprotein (HDL) associated apoE via domains 5–7 and thereby regulates AP activation in plasma (21). The present study was set up to investigate whether FH and apoE interaction could play a role in the induction and progression of atherosclerosis by macrophages. We show here that FH increases apoE binding to monocytes and THP-1 macrophages possibly via simultaneous interaction between cell surface sialic acids and apoE and thereby regulates local complement activation. Moreover, FH interaction with THP-1 macrophages and cholesterol-labeled cells increases macrophage-mediated cholesterol efflux and modulates the expression of inflammatory genes suggesting a yet unexplored anti-inflammatory mechanism for FH.

#### MATERIALS AND METHODS

#### Proteins

Cloning and Pichia pastoris expression of the recombinant fragments FH5-7, FH19-20, and FH1-4 has been described earlier (22, 23). If necessary, fragments were further purified by passing through a HiLoad 16/60 Superdex 200 prep-grade gel filtration column (GE healthcare) in phosphate buffered saline (NaCl 300mM, KCl 5.4 mM, Na2HPO4 20 mM, KH2PO4 3.6 mM, pH 7.4), and concentrated using heparin affinity chromatography. Labeling of proteins was performed using Nhydroxysuccinimide-reactive Red dye (NT647, catalog no. L001) following the manufacturer's instructions (NanoTemper).

#### Expression of apoE Proteins

The preparation of vectors, expression of recombinant apoE2, apoE3, and apoE4 in the Escherichia coli BL21-Gold (DE3) bacterial system following induction by IPTG, and purification by immobilized metal affinity chromatography has been described elsewhere (24–26).

#### Isolation of HDL and LDL and Acetylation of Human LDL

LDL (d = 1.019–1.050 g/mL) and HDL (d = 1.063–1.210 g/mL) were isolated from plasma of healthy volunteers obtained from the Finnish Red Cross Blood Service by sequential ultracentrifugation using KBr for density adjustment (27). LDL was acetylated by repeated additions of acetic anhydride (28). Briefly, LDL (10 mg as LDL protein) in 1.5 ml LDL buffer (150 mM NaCl, 1 mM EDTA, pH 7.4) was mixed 1:1 (vol/vol) with saturated sodium acetate and stirred in ice-water bath for 10 min. Next 30 µl acetic anhydride was added four times with 10 min stirring intervals. After the fourth addition of acetic anhydride, the incubation was continued for 60 min with continuous stirring. Finally, the mixture was dialysed for 24 h at 4 ◦C against LDL buffer.

# Isolation of Peripheral Blood Cells, Cell Cloning, and Culturing

For peripheral blood cell isolation blood was drawn to tubes containing hirudin (Roche Diagnostics, Mannheim, Germany) from healthy human volunteers after informed written and signed consent (Ethical Committee decision 406/13/03/00/2015, Hospital district of Helsinki and Uusimaa). The blood samples were diluted 1:1 (v/v) with PBS and centrifuged through a gradient (Histopaque <sup>R</sup> 1.119 and 1.077; Sigma-Aldrich) at 320 x g for 20 min at 22◦C. The PBMC layer was collected, washed once with RPMI 1640 (Gibco <sup>R</sup> ) containing 0.05% (w/v) HSA (RPMI-HSA) and diluted with RPMI-HSA. U937 human monocytic cells were obtained from the ATCC (American Type Culture Collection), cultured in RPMI medium supplemented with penicillin/streptomycin and 10% (v/v) FCS. CR3 was stably expressed in U937 cells using a lentiviral expression system as described (29). We cloned the CD11b (NM\_001145808.1) and CD18 (NM\_000211.4) cDNA in the dual promoter lentiviral vectors, RP137 (BIC-PGK-Zeo) and RP-139 (BIC-PGK-Puro), respectively. These vectors were constructed by replacing the Zeo-T2a-mAmetrine cassette from the BIC-PGK-Zeo-T2a-mAmetrine (RP172) vector (30) with either a Zeocin or Puromycin resistance gene. First CD11b was stably expressed in U937 cells, subsequently these cells were used for stable expression of CD18. THP-1 monocytes were transformed to macrophages by incubating the cells for 48–72 h in the presence of 100 nM phorbol 12-myristate 13-acetate (PMA) and for generating cholesterol loaded cells mimicking foam cells the macrophages were further incubated in the presence of 100µg/ml (as LDL protein) acetylated LDL for 24–48 h in serum free media.

# Detection of CR3 and Sialic Acid Expression on Cells and Binding of FH on Cell Surface Sialic Acids

The expression of CR3 was analyzed by incubating the cells in 50 µl RPMI-HSA at 5 × 10<sup>6</sup> cells/ml concentration for 45 min with FITC-conjugated anti-CD11b (sigma). The cells were washed once with RPMI-HSA and expression of CR3 was detected by flow cytometry. To detect the specificity of sialic acid expression on different cells and FH binding to cell surface sialic acids the cells were preincubated with 100U α2-3,6,8 Neuraminidase (New England biolabs) or PBS for 30 min at 37◦C. Next, 45 nM NT647 labeled Maackia Amurensis Lectin II (MAL II<sup>∗</sup> ) or 200 nM NT647-FH was added to the wells and the plate was further incubated for 30 min at 37◦C. The cells were washed once with 200 µl of icecold RPMI-HSA, centrifuged at 300 × g for 10 min, fixed with 100 µl of 1% (v/v) paraformaldehyde (Thermo) RPMI-HSA. Next, 2,000–10,000 cells were run by BD LSR Fortessa flow cytometer (Lazer 640 nm, filter 670/30) and analyzed using FlowJo V10 software, where the gating of the cells was performed by using forward scatter (FSC) and side scatter (SSC) to find viable, single cell events. Mean fluorescence intensities were calculated for the gated cells.

# ApoE, FH5-7 and Factor H Binding to Cells

Binding of FH or FH5-7 to peripheral blood monocytic cells, U937 and THP-1 cells was studied by incubating 2 × 10<sup>5</sup> cells with 200 nM of NT647-FH or 1,3µM of NT647 FH5-7 for 45 min at 4◦C in round bottom 96-well plates. For inhibition assays, cells were incubated for 30 min at 4◦C with 1.5µM of apoE, FH1-4 or 9 and 3µM of FH5-7 prior adding the labeled protein. The cells were washed once with 200 µl of icecold RPMI-HSA, centrifuged at 300 × g for 10 min and fixed with 100 µl of 1% (v/v) paraformaldehyde RPMI-HSA. To study the effect of apoE-FH interaction and their binding to monocytic cells a dilution series of FH (Complement technologies), FH19-20 or apoE was incubated in dark for 5 min at 37◦C with a constant 200 nM concentration of NT647-labeled apoE or NT647-labeled FH. Then 40 µl of hirudin blood isolated monocytes were added in each well and incubated in dark for 25 min at 37◦C. The incubation was stopped by adding 300 µl of ice-cold PBS to the tubes and the cells were run by flow cytometry and analyzed using the gating strategy described earlier.

#### FH5-7 Binding to a Panel of Leukocyte Receptors

Monocytes isolated from peripheral blood (PBMCs) cells (5 × 10<sup>6</sup> cells/ml) in RPMI-HSA 0.05% (w/v) were incubated in the presence of 14µg/ml of FH5-7, or PBS for 15 min on ice followed by incubation with a panel of receptor specific FITC-, PE- or APC-labeled antibodies using a 96-well U-plate (greiner, Bio one) for 45 min at 4◦C. The cells were washed with 200 µl of RPMI/HSA 0.05% (w/v) and fixed with 150 µl of 1% (v/v) paraformaldehyde in RPMI-HSA 0.05% for 30 min at 4◦C before counting the fluorescence of 10,000 cells using on FACS flow cytometer (Lazers 405, 488, and 640 nm; filters 450/50, 525/50, and 670/30) and analyzed using the gating strategy described earlier. The inhibition value of FH5-7 on receptor specific antibody binding was calculated by dividing the mean fluorescence of FH5-7 treated cells by PBS treated cells.

### C3b/iC3b Deposition on FH Incubated THP-1 Cells

Untreated THP-1 monocytes, PMA activated THP-1 macrophages, and PMA activated THP-1 macrophages loaded with cholesterol by using acLDL (100/ml µg of LDL protein) were incubated in the presence or absence of FH (320 nM) for 24 h in 24-well tissue culture plates at 37◦C in 5% CO<sup>2</sup> in serum free THP-1 medium using 4 x 10<sup>5</sup> cells/well/200 µl. A 10 µl sample from the supernatant was taken at 0, 5 and 24 h timepoints after the media was centrifuged for 10 min at 300 x g to remove any cellular debris. These culture media aliquots were stored at −20 ◦C before analysis for the apoE ELISA method. After 24 h the cells were detached using 200 µl Cellstripper (Corning) for 45 min at 37◦C, 5% CO2, harvested by centrifugation as before and diluted with RPMI-HSA. Cells from this assay were used to analyze NT647-FH binding (described earlier), mRNA expression, apoE secretion (described later), ABCA1 protein expression and complement activation assays. To detect ABCA1 protein expression the cells were incubated with 0.3 µg/4 x 10<sup>5</sup> cells/well/150 µl mouse anti-ABCA1 antibody (abcam) washed once by centrifugation and incubated with 1:200 dilution of Alexa Fluor 488 conjugated goat antimouse IgG (Invitrogen). To compare serum complement activity in the presence or absence of preincubated FH 5 x 10<sup>4</sup> of THP-1 cells were incubated for 15 min in a 50 µl volume of 20% (v/v) serum. Cells were washed once by centrifugation and incubated with 2 µl of FITC conjugated anti-C3b (Cederlane) for 45 min at 4 ◦C. The cells were washed, fixed, run by flow cytometry (Lazer 488, filter 525/50) and analyzed as described earlier.

# Effect of FH on apoE Secretion and Binding to THP-1 Cells

The cells detached from the tissue culture 24-well plates were incubated in the presence of 2 µl of rabbit anti-human apoE (600µg/ml, non affinity purified) in a 50 µl volume of RPMI-HSA 0.05% (w/v) for 45 min at 4◦C. The cells were washed once by centrifugation and incubated in the presence of 1:100 diluted 488 goat anti-rabbit antibody (Life technologies) for 45 min at 4 ◦C. After incubation the cells were washed, fixed and analyzed by flow cytometry as described earlier (Lazer 488, filter 525/50). Secretion of apoE by THP-1 cells was analyzed from 8 µl of culture media collected at different time points using apoE ELISA protocol as previously described (31).

#### Cholesterol Efflux Assay

The cholesterol efflux assay was performed according to manufacturer's instructions (Abcam) using THP-1 cells activated for 24 h with PMA. The cells were labeled overnight with fluorescently-labeled cholesterol in the presence or absence of 650 nM FH. After overnight labeling the PMA activated and cholesterol labeled THP-1 macrophages were washed gently with 200 µL of RPMI media and incubated then in the presence of 50 µg of HDL (as HDL protein) as cholesterol acceptor. After 5 h incubation the media and supernatant of lysed cells were measured separately for fluorescence (Ex/Em = 482/515 nm). The ratio between fluorescence intensity of media and fluorescence intensity of cell lysate plus media were calculated as percentage of cholesterol efflux.

# RNA-Sequencing

RNA sequencing method was designed based on the Drop-seq protocol described earlier (32). Briefly, the cells were mixed with lysis buffer (0.3% (v/v) triton, 20 mM DTT, 2 mM dNTPs) in wells of U-bottomed 96-well plate. Magnetic Dynabeads (M-270 Streptavidin, Thermo Fisher Scientific) coated with Indexing Oligonucleotides (Integrated DNA Technologies, **Table 1**) were added to each well. After 5 min of incubation at ambient temperature the magnetic beads were separated from the supernatant and washed twice with 6X SSC buffer. Subsequently, the beads were combined with RT mix, containing 1 x Maxima RT buffer, 1 mM dNTPs, 10 U/µl Maxima H- RTase (all ThermoFisher Scientific), 1 U/µl RNase inhibitor (Lucigen), and 2.5µM Template Switch Oligo (Integrated DNA Technologies). Samples were incubated in a T100 thermal cycler (BioRad) for 30 min at 22◦C and 90 min at 52◦C. The beads were washed twice with 6X SSC buffer and once in PCR-grade water. The constructed cDNA was amplified by PCR in a volume of 15 µl using 5 µl of RT mix as template, 1x HiFi HotStart Readymix (Kapa Biosystems) and 0.8µM SMART PCR primer. The samples were thermocycled in a T100 thermocycler (BioRad) as follows: 95◦C 3 min; subsequently four cycles of: 98◦C for 20 s, 65◦C for 45 s, 72◦C for 3 min; following 13 cycles of: 98◦C for 20 s, 67◦C for 20 s, 72◦C for 3 min; and with the final extension step of 5 min at 72◦C. The PCR products were pooled together and purified with 0.6X Agencourt AMPure XP Beads (Beckman Coulter) according to the manufacturer's instructions. They were eluted in 10 <sup>µ</sup>l of molecular grade water. The 3′ end cDNA fragments for sequencing were prepared using the Nextera XT (Illumina) tagmentation reaction with 600 pg of the PCR product serving as an input. The reaction was performed according to manufacturer's instruction, with the exception of the P5 SMART primer that was used instead of S5xx Nextera primer. Subsequently, the samples were PCR amplified as follows: 95◦C for 30 s; 11 cycles of 95◦C for 10 s, 55◦C for 30 s, 72◦C for 30 s; with the final extension step of 5 min at 72◦C. Samples were purified twice using 0.6X and 1.0X Agencourt AMPure Beads (Beckman Coulter) and eluted in 10 µl of molecular grade water. The concentration of the libraries was measured using a Qubit 2 fluorometer (Invitrogen) and the Qubit DNA HS Assay Kit (ThermoFisher Scientific). The quality of the sequencing libraries was assessed using the LabChip GXII Touch HT electropheresis system (PerkinElmer), with the DNA High Sensitivity Assay (PerkinElmer) and the DNA 5K/RNA/Charge Variant Assay LabChip (PerkinElmer). Samples were stored at −20◦C. The libraries were sequenced on Illumina NextSeq500, with a custom primer (**Table 1**) producing read 1 of 20 bp and read 2 (paired

#### TABLE 1 | Primers used in this study.


\* = phosphorothioate bond added

end) of 55 bp (32). Sequencing was performed at the Functional Genomics Unit of the University of Helsinki, Finland.

#### Read Alignment and Generation of Digital Expression Data

Raw sequence data was filtered to remove reads shorter than 20 bp. Subsequently, the original pipeline suggested in Macosko et al. (32) for processing of drop-seq data was used. Briefly, reads were additionally filtered to remove polyA tails of length 6 or greater, then aligned to the human (GRCh38) genome using STAR aligner (33) with default settings. Uniquely mapped reads were grouped according to the 1–12 barcode, and gene transcripts were counted by their Unique Molecular Identifiers (UMIs) to reduce the bias emerging from the PCR amplification. Digital expression matrices (DGE) reported the number of transcripts per gene in a given sample (according to the distinct UMI sequences counted).

#### Statistical Methods

Statistical analyses between multiple samples were performed using one-way ANOVA supplemented with non-parametric Tamhane's multiple-comparison test. Statistical differences between two independent samples were calculated using non-parametric Mann-Whitney U-test (SPSS for Windows, Analytical Software).

#### RESULTS

#### Factor H Binds to Monocytic Cell CR3 via CD11b

FH interacts with human cell surfaces mainly via C3b and cell surface glycosaminoglycans, like heparan sulfate, and sialic acids. Previous studies have shown that especially on endothelial cells and platelets the presence of sialic acids is crucial for efficient FH-mediated complement regulation (34). In addition to surface sialic acids, FH has been shown to interact with CR3 (CD11b/CD18) on human neutrophils in the absence of C3b deposits (35). We studied binding of FH to different ligands on monocytes and monocytic cell lines to find out how these interactions could be altered by apoE that is known to be secreted by macrophages and to interact with FH domains 5–7 (21, 36). By using NT647-labeled FH we could detect clear binding of FH to monocytes even in the absence of C3b, while binding of FH to lymphocytes was closer to the background values (**Figure 1A**). To find out the receptor that interacts with the apoE binding domains FH5-7 (location of the apoE binding domains is shown in **Figure 1B**) on monocytes we performed a screening assay where the cells were preincubated with the recombinant FH5-7 fragment prior to adding the anti-receptor antibody. Incubation of the cells with FH5-7 resulted in a clear reduction in anti-CD11b binding that is the alpha-chain of the integrin-type CR3 receptor heterodimer (CD11b/18; **Figure 1C**). In addition, similar inhibition was also detected with antibodies against CD35 and CD89. Binding of FH5-7 on CR3 was further analyzed using CR3 overexpressing U937 monocytic cells (**Figure 1D**). Binding of NT647-labeled FH5-7 on these cells was reduced in the presence of increasing concentrations of non-labeled FH5-7 (3 and 9µM) and on empty U937 cells that do not express CR3.

# Binding of apoE to Monocytes Is Increased by Factor H

To study further the interaction between the apoE binding ligand FH and CR3 on human monocytes, we next performed an assay to see whether apoE could alter FH-CR3 interaction on U937 monocytic cells overexpressing the receptor. We found that on CR3 overexpressing U937 cells binding of NT647-FH was clearly inhibited by apoE2 to the level of binding of FH

the mean fluorescence of FH5-7 treated cells by PBS treated cells (FH5-7 vs. control). The CD11b part of CR3 dimer is marked. The dashed line shows the level of anti-CD11b binding in the presence of FH5-7 (< 0.8) compared to binding of the antibody without FH5-7 (1.0). (D) CR3 expressing and empty U937 mononuclear cells were incubated with NT647 labeled FH5-7 in the presence or absence of unlabeled FH5-7 (x-axis). Statistical significance between multiple samples was calculated using one-way ANOVA supplemented with non-parametric Tamhane's post-hoc multiple comparison test. Error bars indicate SD. \* = p < 0.05.

to cells devoid of CR3. A slight inhibition was also detected by apoE3 and apoE4 but the levels of inhibition were not statistically significant (**Figure 2A**). As the apoE2 isotype showed significant inhibition we used the apoE2 isoform in the further assays. Surprisingly, when the same inhibition assay was performed using peripheral blood monocytes and lymphocytes we did not detect any inhibition of NT647-FH binding in the presence of increasing concentrations of unlabeled apoE (**Figure 2B**). On the contrary, a clear increase in binding of NT647-apoE was detected when the cells were incubated with increasing concentrations of unlabeled FH, while FH19-20 did not have the same effect. This indicates that FH increases rather than attenuates apoE binding to monocytes. To further study the cell surface receptors leading to increased binding of apoE to these cells, and because both apoE and FH are known to interact with cell surface heparan sulfate, we incubated monocytes with NT647-apoE and FH and increasing concentrations of antiheparin antibody. Here, binding of NT647-apoE in the presence of anti-heparin antibody was reduced to the level of NT647 apoE only (**Figure 2C**) suggesting that increased binding of apoE to these cells is dependent on FH and cell surface heparan sulfate.

Because we could detect inhibition of FH binding by apoE only to U937 cells overexpressing CR3 but not to monocytes

increasing concentrations of anti-heparin antibody. Anti-TREM-2 antibody was used as a negative control. Control (Neg. Cntrl. NT647-apoE) shows binding of apoE on the cells without FH incubation and Control (Pos. Cntrl. NT647-apoE+FH) binding of apoE in the presence of FH and apoE only (n = 3). Statistical significance was calculated using one-way ANOVA supplemented with non-parametric Tamhane's post-hoc multiple comparison test. Error bars indicate SD. Percentages of mean fluorescence intensities are shown as relative to the maximum intensity in each individual experiment. \*= p < 0.05.

we hypothesized that on monocytes, where CR3 expression is lower, FH could simultaneously bind to apoE and cell surface sialic acids. This is because FH sialic acid binding domains are located within domains FH19-20 and not on apoE interacting domains FH5-7. This could also explain why apoE binding was increased by FH to monocytes. As expected, the human peripheral blood monocytes showed high sialic acid expression and low CR3 expression while the U937 cells showed very low expression of sialic acids and very high CR3 expression (**Figures 3A–C**). The high expression levels of CR3 and low levels of sialic acids on U937 cells may explain why inhibition of FH binding to CR3 in the presence of apoE could only be detected on these cells. To study the effect of sialic acids on binding of FH different cell types were preincubated with neuraminidase that removes cell surface sialic acids. All tested cells bound FH at different levels, but only monocytes showed significant reduction in FH binding after neuraminidase treatment. No decrease in FH binding was detected on other tested cell types with lower sialic acid expression (**Figures 3D,E**). These data suggest that FH interacts with several different ligands on cells but binding of FH to sialic acids via domains FH19-20 enables simultaneous binding of apoE via domains FH5-7 and cell surface heparan sulfate.

#### Binding of FH to THP-1 Cells Reduces Complement Activation

Since we found that FH interacts directly with human peripheral blood cells and because activation of complement is known to play a role in the induction and progression of atherosclerosis (37) we next studied whether FH binding to macrophages and cholesterol-loaded macrophages could have effect on local complement activation. To study this, we used THP-1 monocytes stimulated with PMA and acetyl LDL as model cells of early stage atherosclerosis. To avoid measuring FH binding to damaged cells the viability was determined using Trypan blue staining. These cell populations had cell viability of approximately 70%. When THP-1 monocytes, THP-1 macrophages and acLDL (i.e., cholesterol) loaded THP-1 macrophages were studied for NT647-FH binding in the absence of serum, a significant increase in FH binding was detected on activated cholesterol loaded THP-1 macrophages compared to THP-1 monocytes (**Figure 4A**). When the cells were incubated with NT647-FH in the presence of 20% serum a similar trend in FH binding could be detected indicating that the increase in FH binding due to THP-1 activation is independent from surface C3b deposition. Both THP-1 macrophages and cholesterol loaded macrophages showed a significant increase in FH binding compared to THP-1 monocytes (**Figures 4A,B**). To study whether increased FH binding to these cells has functional significance in reducing local complement activation, the serum incubated cells were also analyzed for C3b deposition. Surprisingly, preincubation of THP-1 cells in the presence of FH showed a clear reduction in cell surface C3b deposition only in the case of THP-1 macrophages (**Figure 4C**). Incubation of THP-1 monocytes or cholesterol-loaded THP-1 cells in 20% serum did not lead to an increase in cell surface C3b deposition. Therefore, the presence of additional FH had an effect on the cell targeted complement activity only on THP-1 macrophages.

multiple samples was calculated using one-way ANOVA supplemented with Tamhane's post-hoc multiple comparison test. Statistical significance between two

#### FH Increases apoE Binding to THP-1 Cells and Macrophage-Mediated Cholesterol Efflux

samples was calculated using Mann-Whiney U-test. Error bars indicate SD. \* = p < 0.05.

We used anti-apoE antibody to detect surface bound apoE on THP-1 cells incubated with or without FH under cell culture conditions. Similarly to the human peripheral monocytes apoE binding was detected to activated THP-1 cells from which cholesterol loaded THP-1 macrophages demonstrated significant increase in apoE binding in the presence of FH, while binding of apoE on THP-1 monocytes was low (**Figure 5A**). This correlated well with the previous results, where cholesterol loading of the cells resulted in most significant FH binding. This indicates that FH binding to THP-1 cell surfaces increases apoE binding as well. Because endogenous production of apoE by macrophages in blood vessel wall has been suggested to promote healing of atherosclerotic plaques and efficient transport of cholesterol out from the cell (17), we analyzed the amount of secreted apoE in the culture media. After a 24 h incubation apoE secretion increased significantly from activated THP-1 cells among which cholesterol-loaded THP-1 cells showed highest apoE concentrations in the culture media (**Figure 5B**). These results also correlated well with the transcriptome data obtained from the apoE mRNA sequencing performed from the isolated cells after 24 h incubation (**Figure 5C**). No difference between cells incubated either in the presence or absence of FH could be observed in apoE expression. ApoE is a good cholesterol acceptor among many other apolipoproteins having the potential to bind phospholipids ensuring cholesterol removal (38). Therefore, we next studied whether elevated apoE binding via FH to THP-1 cells could affect cholesterol efflux. Here, a significant increase in cholesterol efflux could be detected from

each individual experiment.

cholesterol-labeled THP-1 macrophages that were incubated in the presence of FH when compared to cells in the absence of FH (**Figure 5D**).

in transcription of complement receptor C5aR2 by FH in macrophages was detected although this was not statistically significant (p = 0.053).

# FH Increases Transcription of Anti-inflammatory Genes in Macrophages

While cholesterol loaded THP-1 macrophages showed highest apoE secretion FH did not have any effect on this. We, however, hypothesized that FH could display anti-inflammatory effects on macrophages as it interacts with both macrophages and cholesterol loaded cells, inhibits complement activity on macrophages, increases cholesterol efflux from cholesterol labeled THP-1 macrophages and increases apoE interaction with these cells as well. After 24 h incubation the THP-1 cell mRNAs were isolated and subjected to sequencing-based transcriptome analysis covering over 27,000 RNA sequences of different genes. Transcriptome analysis resulted in a selection of transcripts that are associated with inflammation, atherosclerosis and the complement system [**Table 2**, (18, 39– 46)]. When these transcriptomes were compared between the FH and PBS incubated cells some clear effects/differences could be seen between THP-1 macrophages and cholesterol loaded THP-1 macrophages. As expected and based on the above FH and apoE binding assays THP-1 monocytes remained unresponsive to FH treatment. FH treatment resulted in increase of ABCA1 transcription in cholesterol loaded macrophages (foam cells) and in a significant increase in transcription of a regulator of ABCA1 expression, PPARα (40). Similarly, FH increased ABCA1 expression as well and correlated with the significantly increased transcription levels of the protein (**Figure 5E**). In THP-1 macrophages the expression of proinflammatory factors CX3CR1, CCL5, and SAAL1 was significantly decreased by FH. Moreover, a reduction

# DISCUSSION

In the present study we showed that complement FH increases apoE binding to macrophages and leads to an increased cholesterol efflux and reduced inflammation. The FH/apoE interaction is hypothesized to limit the progression of atherosclerosis as complement regulation by FH is critical in the prevention self-cell damage and exacerbated inflammation.

We previously found that FH interacts with apoE and reduces complement activation in plasma HDL particles (21). The current study shows that this interaction apparently limits inflammation in the atherosclerotic lesions by affecting macrophage activation and cholesterol efflux. FH was found to interact with cell surface sialic acids and CR3 on monocytic cells in the absence of surface deposited C3b. In the presence of apoE the interaction between FH and CR3 was inhibited, while apoE had no effect on cells abundantly expressing sialic acids. In addition to CR3, our screening assay suggested that domains 5–7 of FH could also interact with CD35 (C3b/C4b receptor, CR1) and CD89. From these, CR3 and CR1 are receptors directly involved in Cmediated clearance and suppression of inflammation. We have previously shown that FH blocks binding of CR1 to C3b (47) as both molecules compete for the same binding site on C3b. Like FH, CR1 is also a C regulator and therefore should be separately analyzed for its role in C regulation during the atherosclerotic lesion development. CR1 is mostly present on cell surfaces but may also occur in soluble form. Its main function is the

supernatants to THP-1 monocytes, THP-1 macrophages and cholesterol-loaded THP-1 cells detected by anti-apoE antibody. Cells were incubated with and without FH for 24 h. Next, the cells were washed and detached from the tissue culture plates. Presence of apoE on cell surfaces were detected using anti-human apoE and Alexa 488 labeled goat anti-rabbit antibody in flow cytometry. ApoE binding is shown as relative to the maximum intensity in each individual experiment (n = 3). (B) Secretion of apoE to cell culture media by THP-1 monocytes, THP-1 macrophages and cholesterol-loaded THP-1 cells detected by ELISA (n = 4). (C) Number of apoE mRNA transcripts analyzed by sequencing the cell isolated mRNA (n = 4) (D) Cholesterol efflux from non-loaded THP-1 macrophages labeled with fluorescent cholesterol in the presence or absence of FH (n = 3). Cholesterol efflux in the presence of equal molarity of FH19-20 is shown. The positive (Pos. cntrl. from Abcam) and non HDL treated controls (Neg. cntrl.) were included in the assay. (E) Protein expression and transcription levels of ABCA1 in THP-1 monocytes, THP-1 macrophages and cholesterol-loaded THP-1 cells detected by anti-ABCA1 antibody and mRNA sequencing. Levels are calculated as relative to the protein expression and transcription in THP-1 macrophages (n = 4). Statistical significance calculated using Mann-Whiney U-test or one-way ANOVA supplemented with Tamhane's multiple comparison test. Error bars indicate SD.

transport of immune complexes and other unwanted materials for clearance by macrophages in the spleen and liver.

Factor H binding to cell surface C3b increases in the presence of self cell glycosaminoglycans and sialic acids. The binding of FH to CR3 has still been regarded as controversial, probably because of the possibility of C3b contamination in the sample. The novelty we show here in the regard of FH and CD11b interaction is that in the absence of C3b FH binding to CR3 expressing cells can be inhibited by apoE2 indicating a common binding site between apoE and CR3 on FH domains 5–7. Inhibition of FH binding to these cells by apoE3 and apoE4 was not that obvious indicating that the single amino acid differences (Cys and Arg in positions 112 and 158) between the apoE isotypes could alter binding to FH. Analogous differences between these isoforms have been observed earlier due to their structural variation. For instance, apoE3 and apoE4 protein isoforms bind well to LDL-receptors, whereas apoE2 displays defective binding (48). CR3 is known to interact with several ligands. A major ligand that promotes phagocytosis is iC3b on opsonised microbes and other particles. According to the mRNA expression data obtained from FH stimulated cells it is unlikely that FH could trigger CR3-mediated signaling, although a weak but not significant increase in CLEC7A (dectin-1) expression in cholesterol-loaded THP-1 macrophages could be detected in the presence of FH (**Table 2**). This C-type lectin is upregulated in anti-inflammatory M2 macrophages (49). However, earlier studies have shown that binding of FH on CD11b suppresses acute subretinal inflammation in mice indicating that FH-CR3 interaction could reduce inflammation in humans as well (50).

In contrast to the lacking inhibitory effect of apoE on FHmonocyte interaction, the presence of FH increased apoE binding to these cells significantly. This observation could be due to the

#### TABLE 2 | Changes in the transcriptome in response to FH in human THP-1 cells.


\*Panel of genes known to play a role in inflammation or atherosclerosis with the description of their function in inflammation. The numbers show the amount of mRNA transcripts (=mRNA expression level). \*\*Up- and downregulation marked as + and – and filled with dark or light gray colors, respectively. Statistical significance (p-values < 0.05 marked in bold numbers) calculated between cells incubated with or without FH using Student's t-test. Gray numbers = no statistical significance between FH and PBS treated cells.

high abundance of sialic acids and low expression levels of CR3 on these cells. Sialic acids favor FH binding to the cell surface via domains 19–20 enabling simultaneous interaction with apoE via domains 5–7. The effect of sialic acids on FH binding was demonstrated by neuraminidase treatment of different cell types, where only monocytes showed a clear reduction in FH binding at different cell surface sialic acid densities. Moreover, binding of FH on THP-1 macrophages clearly reduced C3b deposition on these cells suggesting that FH binding to these cells prior to C3b deposition has an effect on local complement activation. As demonstrated by our mRNA expression data, FH expression on all THP-1 cell types was almost or completely absent, while apoE expression was clearly increased both by THP-1 macrophages and cholesterol-loaded cells.

We observed that FH did not induce secretion or expression of apoE by the studied cells but slightly reduced apoE concentrations in the culture media. However, FH significantly increased binding of apoE especially to cholesterol-loaded THP-1 cells that also showed the highest FH binding. According to trypan blue staining the increase in FH binding was not due to cell damage as the cell viabilities were the same in the absence and presence of FH. It is possible that this kind of rebinding or apoE capturing on the macrophage surfaces could promote apoE recycling that has been suggested to reduce intracellular

cholesterol accumulation and thereby prevent formation of foam cells. Previous studies have shown that apoE can be spared from degradation in lysosomes and recycled to the cell surface in order to maximize cholesterol removal from the cell (41). The intracellular ATP binding cassette transporter ABCA1 is an important regulator of cellular cholesterol homeostasis. It is involved in apoE secretion and in lipidating apoE-containing particles secreted by macrophages. ABCA1 is suggested to promote apoE recycling as well. Our results on the effect of FH on cholesterol efflux suggest that the increased binding of apoE caused by FH could promote apoE recycling and thereby maximize cholesterol efflux from macrophages. Importantly, the mRNA expression data showed increase in ABCA1 expression on FH-treated cholesterol-loaded THP-1 macrophages and a significant increase in the expression of PPAR-α transcription factor that is known to induce ABCA1 expression. This clearly suggests an anti-atherogenic function for FH-apoE interaction (42).

We found that treatment of THP-1 cells for 24 h clearly altered transcription of several genes associated with inflammation, atherosclerosis, and the complement system (**Table 2**). These changes occurred mainly within PMA-differentiated THP-1 macrophages and cholesterol loaded THP-1 macrophages while THP-1 monocytes as well as non-loaded THP-1 macrophages were less responsive to the treatment. These data did not suggest any clear M1/M2 polarization but general trend was that FH reduced significantly transcription of proinflammatory genes in macrophages (such as CXCR3, CCL5, SAAL1) but increased transcription of antiatherogenic genes that are involved in enhanced lipid metabolism (such as ABCA1, PPAR-α) mainly in cholesterol loaded cells. FH did not only result in increase in transcription of ABCA1 in cholesterol loaded macrophages but also in ABCA1 protein expression and PPAR-α transcription. This indicates that FH may affect in macrophage cholesterol efflux via this transcription factor known to activate ABCA1-mediated cholesterol efflux in human macrophages (42). Therefore, the difference between the transcription levels cannot be explained by the instability of mRNA transcripts that does not always correlate with the corresponding levels of protein expression (51).

Certain mutations and polymorphisms of FH are associated with AP dysregulation mediated diseases such as AMD, DDD, and aHUS. AMD and DDD are characterized by formation of deposits in the eye (drusen) or kidney (glomerular basement membrane) that are rich in complement activation products and, most importantly, also contain high amounts of apoE (52, 53). Therefore, in addition to atherosclerosis the pathogenesis of diseases such as AMD and DDD could be related to the FH and apoE-macrophage interactions. While FH interaction with apoE launches a concerted action leading to several antiinflammatory responses on macrophages (**Figure 6**) genetic or acquired disturbances in this homeostatic mechanism could promote the progression of atherosclerotic and other analogous lesions.

# AUTHOR CONTRIBUTIONS

EN helped in data interpretation and manuscript evaluation, wrote the paper, and performed analysis. PH, KL, AR, SS, and JM performed analysis. JvS, KVK, CDH, and PS contributed data or analysis tools. KÖ, MJ, AC, SM, and SJ helped in data interpretation, helped to evaluate and edit the manuscript, contributed data or analysis tools. KH supervised development of work, helped in data interpretation and manuscript evaluation, designed the analysis, wrote the paper, and performed analysis.

#### FUNDING


#### REFERENCES


Foundation (to MJ and JM) and Sigrid Juselius Foundation.

# ACKNOWLEDGMENTS

We thank Bachelor student Mikael Gromyko for excellent technical assistance.


E3 ligase involved in ER-associated protein degradation. Nat Commun. (2014) 5:3832 doi: 10.1038/ncomms4832


ATP and activation of P2Y11 receptor. PLoS ONE (2013) 8:e59778. doi: 10.1371/journal.pone.0059778


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Nissilä, Hakala, Leskinen, Roig, Syed, Van Kessel, Metso, De Haas, Saavalainen, Meri, Chroni, Van Strijp, Öörni, Jauhiainen, Jokiranta and Haapasalo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Structural Immunology of Complement Receptors 3 and 4

Thomas Vorup-Jensen1,2 \* and Rasmus Kjeldsen Jensen<sup>3</sup>

*<sup>1</sup> Biophysical Immunology Laboratory, Department of Biomedicine, Aarhus University, Aarhus, Denmark, <sup>2</sup> Interdisciplinary Nanoscience Center, Aarhus University, Aarhus, Denmark, <sup>3</sup> Department of Molecular Biology and Genetics—Structural Biology, Aarhus University, Aarhus, Denmark*

Complement receptors (CR) 3 and 4 belong to the family of beta-2 (CD18) integrins. CR3 and CR4 are often co-expressed in the myeloid subsets of leukocytes, but they are also found in NK cells and activated T and B lymphocytes. The heterodimeric ectodomain undergoes considerable conformational change in order to switch the receptor from a structurally bent, ligand-binding in-active state into an extended, ligand-binding active state. CR3 binds the C3d fragment of C3 in a way permitting CR2 also to bind concomitantly. This enables a hand-over of complement-opsonized antigens from the cell surface of CR3-expressing macrophages to the CR2-expressing B lymphocytes, in consequence acting as an antigen presentation mechanism. As a more enigmatic part of their functions, both CR3 and CR4 bind several structurally unrelated proteins, engineered peptides, and glycosaminoglycans. No consensus motif in the proteinaceous ligands has been established. Yet, the experimental evidence clearly suggest that the ligands are primarily, if not entirely, recognized by a single site within the receptors, namely the metal-ion dependent adhesion site (MIDAS). Comparison of some recent identified ligands points to CR3 as inclined to bind positively charged species, while CR4, by contrast, binds strongly negative-charged species, in both cases with the critical involvement of deprotonated, acidic groups as ligands for the Mg2<sup>+</sup> ion in the MIDAS. These properties place CR3 and CR4 firmly within the realm of modern molecular medicine in several ways. The expression of CR3 and CR4 in NK cells was recently demonstrated to enable complement-dependent cell cytotoxicity toward antibody-coated cancer cells as part of biological therapy, constituting a significant part of the efficacy of such treatment. With the flexible principles of ligand recognition, it is also possible to propose a response of CR3 and CR4 to existing medicines thereby opening a possibility of drug repurposing to influence the function of these receptors. Here, from advances in the structural and cellular immunology of CR3 and CR4, we review insights on their biochemistry and functions in the immune system.

Keywords: innate immunity, complement, complement receptors, integrins, cell adhesion, von willebrand facor A (VWA) domain, divalent metal ions, drug repurposing

#### Edited by:

*Robert Braidwood Sim, University of Oxford, United Kingdom*

#### Reviewed by:

*Annette Karen Shrive, Keele University, United Kingdom Jean van den Elsen, University of Bath, United Kingdom*

> \*Correspondence: *Thomas Vorup-Jensen vorup-jensen@biomed.au.dk*

#### Specialty section:

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

Received: *07 July 2018* Accepted: *05 November 2018* Published: *26 November 2018*

#### Citation:

*Vorup-Jensen T and Jensen RK (2018) Structural Immunology of Complement Receptors 3 and 4. Front. Immunol. 9:2716. doi: 10.3389/fimmu.2018.02716*

# INTRODUCTION

Complement receptors (CRs) make an important link between cellular functions, notably—but not exclusively—between functions of the leukocytes and soluble complement components (C), factors (F), and several related proteins. They also create a strong connection between those parts of the immune system often termed the innate immune system and other parts forming the adaptive immune system. Among these receptors, especially CR3 has been the subject of several studies at essentially all levels of modern biology including its biochemistry to in vivo analysis by use of transgenic mice (1). Nevertheless, in spite of more than 40 years of research, the full versatility of CR3 seems not to have been captured as yet, not to mention the structurally similar CR4, which is even less understood.

The present review focuses on highlighting both a few past and some more recent insights on the structural biology and functions of CR3 and CR4. The focus is on extracellular biology of these receptors, comparing their ligand recognition and how to put their structural biology into a context of immunology. The equally important, and quickly developing, topic, of intracellular signaling by CR3 and CR4 is only briefly touched upon. The Reader is referred to other authoritative reviews for a more comprehensive elucidation of this topic (2–4).

It is not a new idea to review the literature on CR3 and CR4 together (5–7). The present paper aims to make a critical contribution by addressing the question why we have come to think of these receptors as particularly similar. To this end, the present section includes a brief historical summary on the discovery of CR3 and CR4, followed by a broader introduction to their family of β<sup>2</sup> (CD18) integrins. Section the structure, conformational regulation, and ligand recognition by CR3, and CR4 addresses the conundrum of CR3 and CR4 ligand recognition in the context of advances in the structural biology of these receptors. In Section Therapeutic interventions targeting CR3 and CR4, an obvious, yet in the literature surprisingly absent, theme is brought up, namely what role CR3 and CR4 play in human medical therapy. The current situation is paradoxical as no medicines in use are directed to these receptors, but several pharmacological agents may nevertheless target CR3 and CR4 functions, at least as conjectured from primarily biochemical and cellular investigations. Finally, Section Conclusion: CR3 and CR4, significant contributors to both innate and adaptive immunity concludes by looking ahead to the next important steps in the investigations of CR3 and CR4.

#### Functions of CR3 and CR4 and the Family of CD18 Integrins

CR3, at the time named Mac-1, was discovered by Springer et al. (8). They immunized rats with a human leukocyte cell membrane extract and thereby produced a monoclonal antibody (Ab), the M1/70, which was the first to react with a "discrete molecule specific to phagocytes" (8). The activity toward phagocytes prompted the question of the M1/70 impact on complementopsonized phagocytosis. Indeed, M1/70 blocked the interaction of neutrophils with iC3b (9), an activity assigned before as constituted by CR3 but with no molecule "in hand" (10).

The discovery of CR4 was more convoluted. Originally characterized as a part (p150,95) of the product in pull-down experiments with Ab to CD18, little information was obtained on its function initially (11). By the use of affinity matrices coupled with iC3b, it was possible to pull down the p150,95 antigen (12, 13). The similarity in terms of ligand specificity with CR3 was striking (13), also including the inability of CR4 to react with C3d, an observation, which has received further support recently (14). CR4 has been useful as a widely employed marker of murine dendritic cells (with the nomenclature CD11c/CD18) following the observation that this molecule is the most abundant in the cell membrane of these cells (15). A remarkable property of both CR3 and CR4 is the intracellular location of receptors stored in neutrophil granula (11). Upon activation of neutrophils, for instance using the bacterial product N-formylmethionineleucyl-phenylalanine, CR3 is particularly mobilized from these storages to the cell membrane functionally enabling these cells to respond to iC3b deposited on targets (16–18). This provides an ∼17-fold upregulation of expression in the membrane through a mechanism which has no immediate transcriptional component. CR3 provides an example sometimes overlooked in the age of transcriptomics, that not all protein expression is regulated by mRNA synthesis and decay, at least in the cell membrane.

From a protein phylogenetic standpoint, CR3, and CR4 belong to the β<sup>2</sup> integrin family of adhesion molecules (19, 20). The family contains four members, namely integrins αLβ<sup>2</sup> (lymphocyte function-associated antigen [LFA]-1 or CD11a/CD18), αMβ<sup>2</sup> (Mac-1, CR3, or CD11b/CD18), αXβ<sup>2</sup> (CR4, p150,95 or CD11c/CD18), and αDβ<sup>2</sup> (CD11d/CD18). CD18 integrin expression is restricted to leukocytes and one or more types are found on nearly all leukocytes (19). An interesting exception is human and murine mast cells, which as part of the maturation process lose CD18 expression (21, 22). The functional consequences of this situation remain unknown.

LFA-1 is expressed in both lymphocytes and myeloid cells, while CR3 and CR4 is strongly expressed in macrophages and non-classical monocytes, which are usually considered a precursor cell of the tissue-embedded macrophages. Neutrophil granulocytes are also prominently expressing CR3 and CR4 (23). The expression in lymphocytes is more varying and probably dependent on activation. Natural killer (NK) cells express high levels of LFA-1, CR3 and CR4 (19). A less strong expression of CR3 and CR4 is also found in some T and B cells as reported in a few studies (24–26). The integrin αDβ<sup>2</sup> is not particularly well studied in any respect, but the expression and functions seems to share some properties with CR3 and CR4 (27, 28).

CD18 integrins serve important roles in leukocyte cell contacts. They are involved in the extravasation of leukocytes through the endothelium to zones of inflammation, the contact between lymphocytes and antigen presenting cells, and phagocytosis of complement opsonized targets.

LFA-1 is the key molecule in regulating the contacts of leukocytes with intercellular adhesion molecule (ICAM)-1 expressed on activated endothelium (29–31). CR3 may also interact with ICAM-1(32) and LFA-1, CR3, and CR4 have been reported to bind ICAM-4 (33, 34). However, it remains unclear whether these interactions are auxiliary to LFA-1-mediated adhesion or serve purposes that are more specialized. The LFA-1/ICAM-1 interaction also serves the important task of the formation of the immunological synapse (19, 35), crucial the contact between antigen presenting cells (APC) and T lymphocytes. On the surface of the APC, ICAM-1 molecules will form bonds to LFA-1 on the T lymphocyte, essentially in an outer circle surrounding the T cell receptor (TCR)-major histocompability complex (MHC) molecule. This organization easily follows from the curvature of the cells and the longer stretch of the LFA-1/ICAM-1 compared to the TCR-MHC complex (36). Interestingly, it is not known if CR3 and CR4 forms similar structures as part of their contact with target surfaces. If their ligands on such a surface is complement, the deposition could present less order in the spatial organization of CR3 and CR4 ligands than what is seen for the classic immunological synapse. On the other hand, results from nanomicrobiology have pointed to a high level of surface structure of the microbial cell wall. For instance, the peptidoglycan cell wall of Staphylococcus aureus was shown to be built in concentric circles (37). Likewise, the fungal human pathogen Aspergillus fumingatus also present woven textile-like surface pattern (38). The patterns probably affects the binding of certain polyvalent immune effector molecules such as IgM and mannan-binding lectin (MBL) (39, 40). However, it seems plausible that deposition of molecules such as the complement component C3-fragment iC3b could be guided by the surface structure. With the concentric ridges on the surface of S. aureus, it is not inconceivable that this would impose ring-like organization of CR3 or CR4 in the leukocyte cell membrane upon contact with the complement-opsonized bacterial surface. Investigations on these questions are lacking.

As mentioned above, NK cells are carrying high levels of CR3 and CR4. A few studies in the past documented their role in complement-dependent cell cytotoxicity (CDCC), but the role of this NK cell effector mechanism was unclear, at least compared to the better understood antibody-dependent cell cytotoxicity (ADCC). A recent study has now shown that CDCC may account for as much as 50% of NK cell cytotoxicity to anti-CD20 (rituximab)-covered B cell targets (41). Of course, this finding opens up for a better understanding of complement in antibodybased biological therapy and certainly highlights the role of CR3 and CR4 in this process. Likewise, it is also possible that the contribution of complement to certain pathologies can be now be thought of as involving NK cell cytotoxicity, including diseases with autoreactive antibodies. Especially in the latter case, the means of actually targeting the function of CR3 and CR4 appears equally important as discussed further in Section Therapeutic interventions targeting CR3 and CR4.

#### Soluble CD18 Complexes

An increasing number of reports have now identified shed ectodomains of CD18 integrins in the blood of humans and in mice. Mechanistically the shedding probably involves matrix metalloproteinases (MMPs), notably MMP-9 (42), although experiments in mice suggest a more complex situation, probably with several sources of pericellular proteolysis involved (43).

In humans, an initial study identified soluble (s) CD18, mainly in the form of sLFA-1, in fluid from induced blisters with large influx of neutrophil granulocytes. On these cells, a stub remained of the CD11a chain, which apparently was more degraded than the CD18 chain (44). Gjelstrup et al. published the analysis of three groups of arthritis patients, namely, rheumatoid, spondylo, and osteoarthritis. The sCD18 in synovial fluid from the inflamed rheumatoid and spondyloarthrtitis patients was clearly higher than the plasma concentration (44, 45). Interestingly, the plasma sCD18 concentration has turned out often to be lower in chronically inflamed patients (45, 46). This probably connects to the observation that the sCD18 species are ligand binding active to a level where they may compete with cellular adhesion as shown in several experiments with ICAM-1 as a ligand for sLFA-1 (45, 46). The most abundant type of sCD18 species in humans seem to contain the CD11a chain (45), meaning that ICAM-1 is likely the major ligand for sCD18 (47). One important observation made, so far only by Gjelstrup et al. is the oligomeric state of the sCD18 species (45). It is likely that the oligomerization enables a strong, polyvalent interaction with ligand-coated surfaces such as the tested surfaces with ICAM-1 (44–46) and iC3b (48). It was observed that recombinant sCR3 fragments oligomerize (49), but the relationship between these oligomers (45) with the oligomer forms found in plasma remains unclear. Furthermore, the structure of sCD18 oligomers, is not well-understood, not even at a level of understanding the stoichiometry of alpha and beta chains. Direct detection of sCR3 and sCR4 forms in human plasma was attempted by Gjelstrup et al. (45). Probably as the first, this study reported on barely detectable amounts of sCR3 in human plasma, later supported by reports by others demonstrating the shedding of CR3 (50). Recently, it was possible to demonstrate the antagonistic influence of full plasma on cell adhesion to iC3b, i.e., a CR3 and CR4 ligand, with reduction of the antagonism when sCD18 species were depleted (48). CR3 appears to bind the zymogen form of MMP-9 and also associates with the active enzyme in the cell membrane (51), a finding which undoubtedly has implications for receptor shedding. In murine serum, sCR3 is considerable easier to detect (52). Other studies demonstrated that shedding of CR3 is critical for the efflux of macrophages in an experimental murine model of peritonitis and presented vidence that sCR3 may act as soluble antagonist to CD18 integrin binding to ICAM-1, fibrin, and collagen (43). Attempts to measure sCR4 in human plasma failed and no reports on such species are apparently available (45). Both in this case as well as concerning the issues in making strong detection of sCR3 in human plasma, factors such as proteolytic degradation or affinity of the tested antibody recognition could explain the lack of signal.

# THE STRUCTURE, CONFORMATIONAL REGULATION, AND LIGAND RECOGNITION BY CR3 AND CR4

Both CR3 and CR4 have been helpful molecules in understanding the structural biology of integrins. The atomic-resolution structure of the CR3 ligand binding domain and the CR4 ectodomain explained critical aspects of integrin ligand binding activity and the large conformational changes enabling ligand binding. With the recent structure of a complex between the CR3 ligand binding domain and C3d, new light has been shed on how this receptor binds what is likely its most prominent ligand. In direct structural comparison between the ligand binding domains of CR3 and CR4, it is also evident why CR4 is not able to bind C3d similar to CR3, this way distinguishing the binding of C3 fragments by CR3 and CR4.

#### Structure of CR3 and CR4 Ectodomains

As members of the CD18 integrins, CR3 and CR4 form a heterodimeric complex containing one CD18 beta chain (β2) and either of the alpha chains α<sup>M</sup> or αX, respectively. The β<sup>2</sup> chain is a moderately glycosylated molecule with a M<sup>r</sup> of 95,000. The alpha chains vary between Mrs of 150–170,000, with the α<sup>X</sup> being notably less glycosylated than the other chains (19). It is not known if and how the reduced glycosylation of CR4 affects its function, and the topic is not pursued further here. The CD18 integrins also contain multiple metal ion binding sites, some with significant implications for the function of the integrins (53, 54).

The structural organization of the CD18 integrins follows a widely conserved domain organization, also found in other integrins (**Figure 1**). The CD18 integrins belongs to the class of inserted (I) domain-carrying receptors. As suggested by the name, the I domain is inserted between blade W2 and W3 of the seven-bladed beta-propeller domain (58). It belongs to the family of von VWA domains, taking the Rossmann fold (59). In the CD18 integrins, the domain contains seven amphipathic helices surrounding a hydrophobic β-sheet core. This domain is found in GTPases as well as in several other molecules with adhesive functions, and notably some parasite proteins found in Plasmodium falciparum and Toxoplasma gondii are considered for use as vaccine antigens (60, 61).

The I domain is the best and most widely characterized part of the CD18 integrin structure. In those integrins carrying an I domain, it is the major ligand binding site. Isolated domains from α<sup>L</sup> (62, 63), α<sup>M</sup> (64–67), and α<sup>X</sup> (68) have all been characterized at high resolution with X-ray crystallography (XRC). So far, a highresolution solution structure has only been obtained for the α<sup>L</sup> I domain (αLI) was by nuclear magnetic resonance spectroscopy (69, 70). Earlier, NMR was also used to confirm the folded nature of the αMI (71). Several of the key structural findings came from analysis of the αMI. The first structure, referred to as the "open" conformation identified the metal-ion dependent adhesion sites (MIDAS), which chelates a Mg2<sup>+</sup> ion in the primary coordination sphere through the hydroxyl groups of the residues Ser142, Ser144, and Thr209 (**Figure 2**). Two water molecules and Glu314 from a neighboring αMI completed the Mg2<sup>+</sup> coordination sphere (65). Another structure of the αMI, referred to as the "closed" conformation, showed a more compact packing of the C-terminal α7 helix and a primary coordination sphere consisting of Ser142, Ser144, Asp242, and three water molecules (64) (**Figure 2**). Evidently, this suggests a mechanism for regulation of ligand binding, where the side chains in the coordination sphere of the "open" conformation enables the chelation of external anionic ligands, e.g., a glutamate side chain carboxylate. Experimental evidence from both computational stabilization of the domain in the open conformation (72), a

structure-guided mutation in a hydrophobic pocket in the wildtype domain keeping the C-terminal alpha helix in position (67), as well as from engineered disulphide bridges locking the domain in the open conformation confirmed that the open-conformation αMI had a several fold higher affinity for ligand than the closed conformation (73). It is noteworthy that the hydrophobicpocket mutation also stabilized the αXI in the ligand-binding conformation (68), while the αLI requires a different set of mutations to be stabilized in this conformation (74).

with permission from Elsevier.

Evidence for ligand contacts outside the I domain in CD18 integrins is limited to studies on CR3 and iC3b binding. Mutations in the alpha chain beta-propeller domain showed a reduced binding to this ligand (75). Deletion of the αMI from the alpha chain produces a construct that more moderately supports iC3b binding, which, however, was not completely ablated by deletion of the I domain, again supporting a ligand binding site outside the I domain (76). The involvement of the CR3 alpha chain beta-propeller domain in binding iC3b recently received further support from analysis by electron microscopy (EM) (49). A more complex aspect is the apparent ability of CR3 to interact with certain carbohydrate chains, notably β-glucan (77, 78). A well-defined binding site for this interaction has not been characterized, even though some evidence from a functionblocking antibody suggest a location in the membrane proximal part of the α<sup>M</sup> chain (77). The interaction seems to be able to prime certain anti-cancer responses (77), but the mechanistic

part remains uncertain, including the possible involvement of a lectin co-receptor in complex with CR3.

With regard to the function of the CD18 integrin ectodomain outside the I and beta propeller domains, a wide range of experimental work on several types of integrin receptors has now shown how conformational changes are transmitted through alpha and beta chains (79). Briefly, in their resting state, integrins are kept in a bent conformation with the head piece in close proximity to the transmembrane part of the alpha and beta chain, and close to the cell membrane. Upon activation, there is a Swissblade like opening of the receptor to take a more elongated state (**Figure 1**). Studies on the integrin αLβ<sup>2</sup> identified the β<sup>2</sup> chain I-like domain, with structure highly resembling the I domain, as critical in forming a contact to the C-terminal helix of the αLI, thereby exerting a pull sufficient to open the conformation of the I domain (80). A similar mechanism would be expected for both CR3 and CR4. The critical interplay between the alpha and beta chains in transmitting the conformational signal to regulate ligand binding was demonstrated earlier by studies on CR4.

CR4 is probably the most difficult CD18 integrin to activate. As one part of the challenge to enable ligand binding by CR4, it should be noted that CD18 integrins in most expression system require co-expression of both the alpha and beta chain to be presented on the cell surface or secreted in a well-folded state. The reasons for this requirement are not clear, although it may be speculated that the chains exert a mutual, and critical, chaperone-like activity, which ensures that only correctly paired heterodimers reaches the compartments for CD18 integrin function. In principle, mutations in the human chains could enable activation, but prior to the detailed structural information now available, such a strategy would face the dual problem of making constructs that enabled ligand binding and maintaining sufficient integrity to permit heterodimer formation. Bilsland et al. (81) tested the elegant hypothesis that co-expression of the human α<sup>X</sup> chain with the chicken β<sup>2</sup> chain would produce an expressible construct with sufficient alterations in the pairing between the two chains to enable ligand binding activation. In effect, since this construct bound iC3b, while a construct with the native human chain did not, is clear evidence that the alphabeta chain pairing is important in regulating the activity of the CD18 integrins. This is also of direct consequence to the studies on the ligand binding sCD18 species, discussed in Section Soluble CD18 complexes Evans et al. (44) noted that, in the case of sCD11a/CD18, portions of the alpha chain was probably degraded, but ICAM-1 binding activity was retained. With the insight from CR4 on how contacts between the alpha and beta chains restrain activation (81), it seems likely that proteolytic removal of some alpha (or beta chain) domains would unleash the ligand binding activity of the soluble ectodomains.

CR4 was the first ectodomain of an I-domain carrying integrin to be studied with XRC by Xie et al. (55). In addition to adding further insight to the nature of the conformational lability of the ecto domain, it was clearly demonstrated that the αXI is loosely attached to the remainder of the ectodomain body through long loop regions. Xie et al. explained this finding as logically offering some structural freedom in the ability to form contacts with ligands. Indeed, at least on speculative grounds, one would think that such freedom would be usable to solicit further interactions with the beta-propeller domain as experimentally found for the CR3:iC3b interaction. Curiously, however, at least in the case of CR4, there seems not to be such interactions (49). Another possibility for the need of I domain flexibility, if not often addressed in CD18 integrin ligand binding studies, concerns the involvement of the divalent metal ion of the MIDAS in the contact. In the case of Ctype lectins, which binds carbohydrate hydroxyl groups through a chelated Ca2<sup>+</sup> ion, in many respects chemically similar to Mg2+, an important paper showed by NMR that the permitted stereochemistry of this interaction constrains the position of the carbohydrate (82). In I domains, the stereochemistry of ligands in the primary coordination sphere of the Mg2<sup>+</sup> is likely to restrict the movements to pivoting around the chelated anion, similar to what have more recently been observed for the αMI:simvastatin complex (see Section Mechanistic basis for αM and αX I domain recognition of structurally diverse ligands). Accordingly, when structures of the ligand or ligand mimetics are compared, the fixed stereochemistry of Mg2<sup>+</sup> coordination sphere is striking (**Figures 3A–D**). When the CR4 ectodomain is compared with the ectodomain of LFA-1, it seems that the loop regions connecting either of the alpha chain I domains are longer in CR4 (**Figures 4A,B**). With the MIDAS requirements for a certain orientation of the Mg2<sup>+</sup> coordination sphere ligands, the flexible attachment of the αXI probably serve to enable successful chelation of acidic groups in ligands even with considerable

variation in the structural environment of these groups. The study by Sen & Springer (85) concluded that, at least in the case of LFA-1 and CR4, the I domain flexibility is only structurally limited by the contact with the headpiece platform and that both integrins probably permit large movements of their I domains. However, the LFA-1 carries four sites for attachment of large, Nlinked glycosylations in the vicinity of the I domain, while the CR4 has none such. These differences in features could explain how CR4 may bind multiple ligands, even with multiple sites within the same molecule (49, 86–88), while LFA-1 is far more restricted in its interactions.

#### Structural Insights on CR3 and CR4 Ligand Binding

Considering the long list of ligands for CR3 and CR4, the number of structural studies on ligand interactions is disappointingly limited. There is no doubt that one fascinating part of CR3 and CR4 biology is how they accommodate binding to such a large inventory of chemically highly diverse ligands. Fortunately, recent progress in structural studies on especially CR3 offers valuable data.

As mentioned above, indirect evidence of the MIDAS function was produced from XRC on the αMI. On one hand, these structures indicated an open conformation, which enables the contact with a glutamate side chain from a neighboring domain in the crystal lattice, while the closed conformation would not support such an interaction. On the other hand, in this homotypic interaction, the glutamate was the only contact between the domains, which seemed to exclude this interaction as reflecting a proper protein-protein interaction (65), usually requiring larger surface areas to form stable contacts. This was later found for the αLI in complex with ICAM-1 producing a buried surface area of 1,250 Å<sup>2</sup> (63). Even so, the homotypic interaction is a quite persistent property of αMI and CR3. Recent EM studies clearly show that the homotypic interactions also can be found with the CR3 headpiece, in this case forming an abundance of dimers (49). Due to limitations in the structural resolution, this interaction is not clarified at the atomic level.

Two interesting reports detailed the inhibitory potential of the antibody mAb 107, to the αMI. Surprisingly, in the authors' terms, the antibody acts as a ligand mimetic (84, 89). mAb 107 stabilized the αMI in the closed conformation, even when using αMI constructs mutated to favor the open conformation (**Figures 3E,F**). This stabilization occurred with a Ca2+, rather than Mg2+, in the MIDAS. Further separating this structure from others complexed integrin I domain structures was the finding of bidentate involvement of aspartate side chains as part of the Ca2<sup>+</sup> coordination sphere. This work highlights the surprisingly multifaceted nature of the MIDAS in regulating CR3 ligand binding, especially because the structure could have natural, but so far elusive, ligand correlates. From earlier metal ion affinity measurements directly on the αMI, it is clear that, in the isolated domain, Mg2<sup>+</sup> is strongly favored over Ca2+, although none of the affinities would permit the MIDAS to be saturated with metal ions at the physiological concentrations (66, 90). This was also found for the αLI, where hypo or hyper saturation with Mg2<sup>+</sup> compared to physiological levels, strongly changed the interaction with ICAM-1 (91). Taken together, this work suggests that the CR3 MIDAS metal ion binding is part of both the conformational dynamics and potentially contributing some regulation of the ligand binding.

The first structure revealing details of CR3 complement binding was made by Bajic et al. (14), who characterized the complex between αMI and C3d (**Figures 3A**, **5A,B**). C3d essentially constitutes the minimal binding site for the domain. The complex interface was formed by an aspartate side chain chelating the MIDAS, occupied by a Ni2<sup>+</sup> ion available in the mother liquid generating the crystal. As judged from ligand binding measurements by surface plasmon resonance (SPR), this binding site is hidden in the C3b structure, but exposed in iC3b. This is fully consistent with necessity of FI cleavage of C3b to produce the CR3 ligand iC3b (9).With the CR3 binding site located in C3d, the considerable conformational change induced in C3b's conversion into iC3b involving a partial detachment of C3d now offer a structural rationale for the classic characterization of the CR3 recognition of C3 fragments (14, 93). The affinity (KD) for C3d is in the sub micromolar-range on a par with the αLI:ICAM-1 complex, further corroborated by the size of the interface area at 491 Å<sup>2</sup> (14), also close to the value for αLI:ICAM-1 and α2I in complex with synthetic collagen-like peptide [Ac-(GPO)2GFOGER(GPO)3-NH2], both at 609 Å<sup>2</sup> (63, 94).

A quite important finding by Bajic et al. (14) was the possibility of CR2 and CR3 to bind the same C3d molecule. CR2 is mainly expressed in B lymphocytes, but is also found in follicular dendritic cells, at least in mice (1). This opens for a quite interesting handling of C3d-opsonized antigens in the lymph nodes. Here, of course, several subsets of CR3-expressing leukocytes reside, including the subcapsular sinus macrophages. As indicated by the name, these cells are in contact with the draining lymph and bordering the leukocyte-dense area of the lymph node, which enables the delivery of antigens to especially B cells. The ability of B of cells to bind the CR3-presented complement-opsonized antigen through CR2, essentially a "hand-over" of antigen (**Figure 6**), readily extend an important aspect of how the complement system is a part in the formation of antigen stimulation of B cells, and hence antibody formation. With the involvement of CR3 on the cell surface of macrophages, the process becomes essentially an "antigen presentation" to B cells (99). The molecular structures involved are, of course, different from the way antigens are usually presented to T lymphocytes through MHC molecules. On the other hand, it was previously thought that B cell antigen recognition involved mainly events on the B cell surface alone, with complement adding to support the binding through co-binding to CR2 while the B cell receptor engaged an epitope in the opsonized antigen (100). Not excluding the likelihood of these events as well, the CR2:C3d:CR3 complex enables the presentation of antigen in a close contact between the antigen-presenting cell and the lymphocyte. It seems that the large dimensions of particularly CR2 are such that even a quaternary complex with the B cell receptor may be permitted through the C3d-opsonized antigen. It is a classic demonstration that the essentially two-dimensional confinement of juxtaposing receptors in the cell membranes of T cells and APCs greatly enhances the resulting affinity of the receptors for each other (101), compared with a situation where the affinity was measured in (free) solution (102), in effect a three-dimensional compartment. Considering that both the CR2 and CR3 bind C3d with affinities in the micromolar and submicromolar range respectively (103), the principle of 2D affinity is undoubtedly significant in producing results in the cellular context.

FIGURE 5 | Contacts between the αMI and C3d. (A,B) The structure of αMI in complex with C3d is shown as determined by X-ray crystallography [4M76; (14)]. The αMI domain is indicated in turquoise and the C3d fragment indicated in purple. Residues involved in the interaction are represented as sticks, and the polar interactions are indicated by dotted lines. (C) The structure of αMI with residues involved in the C3d or iC3b interaction shown as spheres. Residues implicated from mutagenesis studies in αMI and binding to iC3b are indicated in green (92), residues implicated by XRC (14) are indicated in red, and residues implicated both methods are indicated in blue.

# How Are CR3 and CR4 Capable of Binding Multiple Ligands?

Many reports, with a wide distribution in both time and methodologies, have now shown that CR3 and CR4 bind a vast inventory of ligands (40). Quite a few of these ligands are natural occurring substances, including many proteins, nucleic acids and negatively charged glycosaminoglycans (GAG). In addition, multiple engineered molecules are also on the list, including several peptides and small molecules.

The question on how CR3 and CR4 accommodate such binding has fundamental roots in our understanding on the mechanistic workings of the immune system in at least two ways. First, the concept of immune recognition of the body's foes require a level of specificity in the recognition to avoid undue inflammation in non-infected or otherwise normal tissue. Although the many homeostatic roles of the immune system is now well-established, and hence the need for receptors which can interact with several "self " or altered "self molecules," receptors on leukocytes should logically be restricted in their ligand binding to avoid autoinflammatory responses. The relevance of such restriction was recently emphasized by the contribution of CR3 to pathological inflammation (104). Nevertheless, at the outset, CR3 and CR4 seems to challenge this concept. Second, it was one of the great scientific accomplishments of twentieth century to explain how "an apparently infinite range of antibodycombining specificity associated with what appeared to be a nearly homogeneous group of proteins" (105) leading to the discovery of the complex somatic genetic rearrangements and mutations encoding these molecules. This is, however, not an option to rationalize how the CR3 and CR4 I domains manage their binding of many ligands, since the I domains are subject to neither genetic nor post-translational modifications. In effect, there is an unresolved matter concerning a type of proteinprotein interaction permitting binding of a broad range of ligands.

Below, these questions are further addressed with support from the past two decades of research on the CR3 and CR4 ligand binding.

#### The Need for Multi-Ligand Receptors in the Immune System

In understanding the functions of CR3 and CR4, it is probably fair to state that there has been a tendency toward placing CR3 and CR4 in the biological context of their ligands one-by-one. Ligands of the coagulation cascade provide an example. Both CR3 (106, 107) and CR4 (108) bind fibrinogen, and CR3 interacts well with its coagulated form, fibrin. Characteristically, there is large number of reports detailing this interaction, identifying the responsible residues in both ligands and receptors (109). In mice, a binding site for CR3 identified in fibrinogen is necessary for the role of this molecule in limiting staphylococcal infections in vivo (110). Both CR3 and CR4 were reported to also bind heparin (111, 112), which act to limit coagulation. Add to this list kininogen and plasminogen as CR3 ligands (40), and the receptor would very reasonably seem a part of the wider functions of the coagulation system.

The role in the complement system follows a similar path. There can be little doubt that the C3d fragment is one of the strongest ligands for CR3, hence the crystallization of this ligand—receptor complex (14). The binding site covers a relatively small interface area, which, as mentioned above, is not unusual among integrins and certainly still on par with many other interactions considered specific (113). From investigations of this ligand, CR3 appears a complement receptor in its own right and in ways unrelated to its function in binding coagulation factors.

With these examples in mind, and with a list of many other, less characterized ligands (40), our efforts to understand the multiple ligand interactions by CR3 and CR4 face a significant conundrum from a point of view of structural biology. CR3, and maybe CR4, are clearly able to form classic receptor-ligand interactions, which involves a number of critical side chains in both the receptor and ligand. Nonetheless, among the multitude of ligands reported for each receptor, there is no evidence of any particular shared structural element, at least to a level typical for integrins. Paraphrasing the famous lock-and-key analogy by Emil Fischer (1852–1919) originally addressing the specificity of enzymes but used in many other context of protein interactions since including immunology (114), CR3 and CR4 appear to be "locks" with very definite and distinct structural characteristics, but nevertheless permitting the fit of almost any "key," in spite of these keys not sharing any obvious similarities themselves.

To understand the immunological relevance of receptors with such properties, it is worthwhile mentioning that one group of receptors seems to share characteristics with CR3 and CR4 with regard to ligand binding, namely the so-called scavenger receptors. Indeed, CR3 has for several years been on the list of scavenger receptors (115). Scavenger receptors, such as CD36, enable cellular removal of decayed macromolecules in extracellular space (116). This decay can be mediated by sources such as oxidation of low density lipoproteins. Both CD36 and the receptor for advanced glycation end products (RAGE) bind many different biomacromolecular coining the designation of multiligand receptors (117). Adding CR3 and CR4 to this group is easily justified, especially as evidence suggest CR3 and RAGE to act in consort with regard to cellular signaling in leukocytes of the innate immune system. A recent paper identified CR3 as reacting with proteins modified by oxidations products of polyunsaturated fatty acids (118). Such modifications as well as several other processes, including proteolysis, impacts protein structure, sometime causing denaturation. This has a special interest in the case of CR3 and CR4, which bind denatured protein well (76, 88, 119). The concept of CR3 and CR4 being scavenger receptors is quite attractive and avoids any too tight association with distinct physiologic processes from simple binding of the associated proteins. The special role of complement, at least in the case of CR3, also fits this proposal well. Complement deposition on apoptotic cells, immunoaggregates and many other plasma-exposed molecular species is a known and important mechanism of cellular clearance (120). Failure of such clearance, for instance through complement component deficiency or defects in CR3, are associated with autoimmune responses such as systemic lupus erythematosus (120, 121). An increasing literature now shows that CR3 outside-in signaling, i.e., the cellular signaling following ligation, serves to down-regulate inflammation by several leukocyte subsets (23). For a receptor on leukocytes involved in clearance of decayed or "altered self " molecular species, both the broad ability to react with many ligands as well as anti-inflammatory regulation are prerequisites for successful—and harmless—completion of this process.

With the many shared ligands, including denatured proteins, it would be simple to claim that CR4 also serve as a scavenger receptor like CR3. There is, however, no evidence that ligand binding of CR4 is anti-inflammatory. Unlike CR3, CR4 is capable of binding highly proteolyzed fibrinogen, increasing the adhesion by neutrophil granulocytes (88). This result was obtained with the proteases plasmin and subtilisin, which mainly share the ability to profoundly degrade many protein substrates. This capability of CR4 was suggested to enable a "danger signal" from proteolytically damaged tissues, for instance as inflicted by certain microbial infections. In such a scenario, there is a coupling between the use, or perhaps more precisely overuse, of a scavenger receptor function and the triggering of a proinflammatory response. Other evidence seems to suggest that proteolysis of other ligands may convert these into better ligands for CR4. The role of proteolysis in converting non-ligands into ligands is reminiscent of both the complement and coagulation systems, although these cases usually are being understood as far more regulated. Again, as in the case of CR3 as a scavenger receptor, recognition of, on one hand, highly proteolyzed species and, on another, species probably of multiple origins, would seem to involve a principle quite different from a more standard binding interface in protein complexes.

#### Mechanistic Basis for α<sup>M</sup> and α<sup>X</sup> I Domain Recognition of Structurally Diverse Ligands

Surprisingly little effort has been spend on explaining how the CR3 and CR4 recognize structurally diverse ligands, at least compared to the number of reports simply focusing on identifying one or another ligand. As mentioned above, their list of ligands spans not only proteins but also other classes of biomacromolecules, including nucleic acid, GAGs, and lipopolysaccharide (LPS).

One model, which here will be referred to as the "mosaic model" by Ustinov and Plow, embodies the claim from recombinant engineering that the same loop structures on the MIDAS face of the αMI apparently are used in recognizing many ligands (122, 123). The logical strength of this model is its classic approach to what is required for formation of a proteinligand interaction site by clearly providing for a sufficiently large surface area to produce a reasonably strong interaction. As the model predates the αMI:C3d structure (**Figures 5A,B**), the data involved were based on mutagenesis in the MIDAS face of the αMI domain. Many, if not all, of the supporting data were generated by mutating selected αMI residues into their equivalents in αLI, which cannot bind iC3b (122). The lack of binding introduced by these mutations in vicinity of the MIDAS was interpreted as direct engagement of the affected residues in ligand contacts. However, from the mutational investigations on the interaction with iC3b, only one residue was identified, which was also corroborated by the structure of the αMI:C3d complex (**Figure 5C**). This residue was furthermore only involved in a backbone interaction with C3d. As judged from the recent studies by EM, it is unlikely that the αMI forms contacts with iC3b outside the C3d fragment (49). The mutational approach probably failed to distinguish direct contacts from indirect loss-of-function through structural alterations of the αMI. This prompts a concern over the experimental evidence for the "mosaic model."

Another model, here named the "anion chelation model," also makes a starting point with the αLI, which is different from both the α<sup>M</sup> and α<sup>X</sup> I domains, in so far as the αLI has been reported to only bind the structurally highly conserved ICAMs (40). Any model explaining why the α<sup>M</sup> and α<sup>X</sup> I domains bind many ligands should, in consequence, also embody the αLI in explaining why this, otherwise highly similar domain, will not. A central inspiration is here the above mentioned αMI structure with a crystal lattice contact producing a glutamate side chain coordinating the MIDAS of an open-conformation I domain (64) (**Figure 2B**). Nothing similar was reported for the αLI in spite of several available crystal structures (62, 63). In a simple inhibition experiment using surface plasmon resonance, Vorup-Jensen et al. showed that free glutamate acts as an antagonist of fibrinogen binding by αMI and αXI (88). Calculations on the solution affinity of these domains for free glutamate came to a K<sup>D</sup> of ∼2 × 10−<sup>4</sup> M. Similar experiments with the αLI and ICAM-1, estimated the affinity of the αLI for free glutamate to be a 100-fold lower with a K<sup>D</sup> of ∼2.5 × 10−<sup>3</sup> M. Similar findings for the αXI could be made with compounds such as acetate. This identifies anionic compounds, most likely in the form of carboxylates, being unusually strong ligands for the αMI and αXI, but not the αLI. At the same time, anionic moieties are present in most, if not all, of the reported ligands for αMI and αXI, pointing to a shared, if yet minimal, structural motif in the ligands of αMI and αXI.

At this point, support for the anionic chelation model can be found from several sources of investigations.

First, not only will proteins and other macromolecules often carry anionic moieties; these will also be present in different sites within the same molecule, giving rise to multiple binding sites. This was easy to demonstrate with the highly quantitative SPR experiments, where such multiple binding have been directly observable in several experiments from calculations on the moles of immobilized ligand and moles of bound analyte, i.e., αMI or αXI (86–88, 124). As discussed elsewhere (40, 87), this can furthermore be made as a model-free calculation, which avoids the usual reservations regarding extrapolated values. Any influence from immobilization of the ligand on the stoichiometry by destroying binding sites hardly applies in this context, where the apparent stoichiometry exceeds 1:1. In the case of CR4, the concept of multiple binding sites within a single molecule was recently confirmed by EM, which showed a class of interaction with two recombinant CR4 ectodomains bound to the same iC3b (49).

A second aspect also derives from the multiplicity of binding sites within a single protein species. While the binding sites share the carboxylated side chains as the central part, nearby side chains or other structural features may still affect binding of the I domain. In effect, this means that the binding kinetics to the sites may differ, producing a heterogeneous interaction between the αMI and αXI and their ligands. This phenomenon is clearly observable in SPR or similar experiments, where the sensorgrams reflect a composite of binding reactions, unlikely to be accounted for by single exponentials as it would be expected for simple 1:1 reactions. A robust solution to analyzing such experimental data has been provided by Schuck et al. with an algorithm enabling determination of the minimal ensemble of 1:1 reactions required to explain the experimental data set (125, 126). This algorithm has now been applied to analysis of multiple ligands (14, 83, 86–88, 127, 128). Typically, the experimental design used the ligand coupled onto surfaces with the I domains applied in the flow stream. However, in a recent experiment studying the αMI binding to the antimicrobial peptide LL-37, it was possible to show that the ensembles determined in the reverse orientation with immobilized I domain were equivalent to those with I domain in the flow stream (128). An example of the interaction between αMI and iC3b and C3d are provided in **Figures 7A–D**. From the analysis, it is clear that the compactly folded C3d provides an almost homogenous interaction (**Figure 7D**). Although C3d presents more than one carboxylate on its surface, the compact folding would limit the access to the relevant side chains. As expected, this interaction is discernible in the ensemble of interactions characterizing also iC3b (**Figure 7B**). However, the much larger iC3b molecules, with several regions less compactly folded than the C3d part (93), provides additional types of interactions, in particular some with a K<sup>D</sup> at 10−<sup>4</sup> M, i.e., close to the K<sup>D</sup> for the interaction with free glutamate (**Figure 7B**). Similar findings have been made for fibrinogen (88) and the intrinsically unstructured myelin basic protein (MBP) and the likewise unstructured antimicrobial peptide LL-37 (127, 128).

A third consequence of the anion chelation model suggest that at least carboxylates would be ligands for the αMI and αXI irrespective of their "mounting." It was already shown that acetate and propionate inhibited CR4 ligand binding as efficiently as glutamate (88). This was further confirmed by structural studies over the interaction between the cholesterollowering drug simvastatin and the αMI (83). Interestingly, while the binding between the simvastatin carboxylate (**Figure 8A**) and MIDAS Mg2<sup>+</sup> is relatively stable and almost fixed in geometry as discussed above (Section Structure of CR3 and CR4 ectodomains and **Figures 3C,D**), both the molecular dynamics calculations and the lack of resolution of the simvastatin decalin ring in XRC point to rotation of other parts of this ligand when it is chelated to the MIDAS (83) (**Figure 8B**). Further calculations showed that this rotation acts to solicit interaction with side chains in vicinity of the MIDAS. The possibility of soliciting interactions in the MIDAS area through such rotation may contribute necessary binding energy, in particular for small molecules or those ligands not forming a large number of interactions.

Fourth and finally, an important question pertain to if interactions with KDs at ∼10−<sup>4</sup> M play a role in cellular adhesion. Other single-amino acid interactions, such as between plasmin and lysine, take values a 100-fold lower, in the µM range. However, here it is necessary again (see also Section How are CR3 and CR4 capable of binding multiple ligands?) to consider the special situation governing membrane-bound receptors as discussed by Vorup-Jensen (40). A well-established case is the interaction between CD2 and LFA-3. Measured in free solution, similar to the experiments with αMI and αXI, CD2 and LFA-3 bind each other with a K<sup>D</sup> at ∼1.5 × 10−<sup>5</sup> M (129). When the receptors are confined in the membrane, however, the principle of 2D affinity applies (101). Although the 2D affinity constant (2D KD) with units in molecules·µm−<sup>2</sup> is difficult to compare with the solution-based KD, a point in the studies is, that the weak interaction as measured in solution translated into ∼90% binding saturation when the CD2-expressing T lymphocytes adhered to the LFA-3 expressing surfaces (101). By analogy, the apparently weak interaction between αMI and αXI and some of their ligands, as recorded in solution by SPR, is probably strong enough to support meaningful molecular interaction between surface-confined receptors and ligands. It also needs to be taken into account that the interactions from clustering of the receptors in the membrane (130) gain a polyvalent structure, which may further strengthen cellular adhesion even with weak, monovalent interaction as the basis (40).

Concerning the physiological significance of the CR3 and CR4 binding of carboxylates, a somewhat overlooked aspect also involves the availability of free glutamate for αMI and αXI binding. In plasma, the free glutamate concentration is ∼100µm, or 50% of the K<sup>D</sup> of αMI and αXI for this compound (131). From first principles in chemistry, one would expect a 33% saturation of open-conformation I domains. Even more compelling for at least an occasional role of free

glutamate in binding these receptors, in experimental models of staphylococcal brain infections, it was shown that glutamate released from damaged neurons increases the cerebrospinal fluid concentration to 500µm, corresponding to 70% saturation of the I domains of CR3 and CR4 expressed especially on microglial cells (131). How and when free glutamate affect CR3 and CR4 remains unexplored.

#### Ligand Binding Selectivity of CR3 and CR4

From the discussion above, it would be tempting to conclude that the presence of carboxylates is the single most important property characterizing the ligands of α<sup>M</sup> and α<sup>X</sup> I domains, and hence CR3 and CR4. Indeed, a standard negative control on integrin involvement in any binding involves testing the binding in a buffer containing EDTA, removing the MIDAS Mg2<sup>+</sup> ion. However, with the crystallization of the αXI, enabling a direct structural comparison with the αMI, it became clear that their surfaces in vicinity of the MIDAS are quite different with regard to their presentation of hydrophobic elements and electrostatic charge (**Figure 9**). Notably, the αXI present a ridge of positively charged residues, which is not found in the αMI (68). This may well explain some of the ligand binding differences now reported between these domains.

Vorup-Jensen et al. found that the affinity of αXI for polyglutamate was higher than for free glutamate (88). Although chemically very different, heparin sulfate also binds the αXI strongly (88, 112). Heparin is a random co-polymer of repeating disaccharide residue of D-glucosamine and uronic acids with varying levels of sulfation, making heparin amongst the most negatively charged compound in the human body. It is unclear if sulfate groups may act as ligands for the MIDAS. However, in experiments with purified heparin fragments, there was a clear correlation between the level of sulfation of heparin fragment and their affinity for the αXI. By contrast, the αMI showed a relatively poor affinity for these species (112). Another highly negatively charged molecule is osteopontin (OPN) (135). The negative charge is contributed both by aspartate and glutamate residues as well as multiple phosphorylations. Surprisingly, the phosphorylation seems to play no role in the interaction with the αXI (86, 136). This leaves the high density of negatively charged side chains as the primary source of polyanionicity. In addition to the work with isolated αXI, the preference for such ligands was also demonstrated with the intact CR4 in cell membranes (86).

In striking contrast, polyanionic molecules do not bind well αMI or the intact CR3. Rather, evidence suggests a much better interaction with cationic species. For instance, MBP, thought to constitute an important autoantigen in MS, binds CR3 (127). Due to its role in forming contact with negatively charged phospholipid membranes, it carries a high positive charge with a resulting pI of 10. The antimicrobial peptide LL-37, a highly positively charged proteolytic split product from human cathelicidin, is also a ligand for CR3 (128, 137, 138) with an affinity comparable to C3d (128). These reports are further supported by a recent analysis suggesting a degenerate protein motif of positively charged residues as binding αMI (137).

The difference in ligand binding selectivity seems to add a complementary aspect to CR3's and CR4's function. Both polycationic and polyanionic species are fairly abundant species in the body. Cationic molecules often seem to share a significant role in interacting with cell membranes, which could enable functions of a receptor clearing such species in situations where membrane damage is involved. As one example, we have already suggested that CR3 serves a role in interacting with damaged oligodendrocyte membranes in a manner, which could, however, be exacerbated as part of MS pathology (127). A receptor such as CR4 recognizing polyanioinc species may serve quite different functions. Many microbial organisms carry a high negative surface charge, contributed by cell surface constituents such as peptidoglycan and LPS. As well-known from development of nanoparticles, negative charge add to the colloidal stability (139). In this way, the negative charge on microbial organisms become a pattern in the sense of Janeway's concept of innate immunity, namely a chemical trait that the microorganism cannot survive well without and may serve for recognition by the immune system by germ-line encoded receptors (140). The duality between scavenger and immune receptors noted by Gordon (115) also applies here. Many plasma proteins, including fibrinogen, carries a net negative charge. Making these negative charges more accessible through damage to the protein structure, prompts CR4 recognition. Such damage is a consequence of both normal physiological processes such as coagulation as well as excessive proteolysis induced by microbial organisms, most notably bacteria procuring amino acids from the environment. In this way, a receptor with the ligand binding preferences of CR4 will act as both a scavenger to collect damaged proteins as well as potentially alerting the immune system to microbial threats (88).

### THERAPEUTIC INTERVENTIONS TARGETING CR3 AND CR4

Currently, several immunomodulatory therapies aim to manipulate the function of receptors in the cell surface of leukocytes. They usually induce immunosuppression to reduce symptoms in inflammatory disease or, more recently, supporting immune activation, which enable the elimination of malignancies. A large number of both in vitro and in vivo experiments suggest that therapeutic targeting of both CR3 and CR4 could potentially produce effects such as lowering of autoimmune inflammation (141) or enhance the effects of anti-cancer vaccination (142). Even so, apart from a clinical trial aiming to improve the outcome in stroke by blocking the function of CR3 (see below), to our knowledge no other attempts were made to target either CR3 or CR4. It is beyond the scope of the present review to discuss the pharmacological and clinical challenges in doing this. Below, a perhaps more surprising point is made, namely that both CR3 and CR4-binding molecules are almost routinely used in current medical treatments. This is, of course, a consequence of the broad range of ligands bound by these receptors discussed in Section The structure, conformational regulation and ligand recognition by CR3 and CR4. In an era where drug repurposing is increasingly seen as a convenient way to improve therapy without the costs of full-scale clinical trials, the short list of drugs made below is meant as a thought-provoking tool box on how to hit some of the arguably most versatile receptors in the immune system.

# CR3 as Target in Clinical Trials and Target for Multifunctional Drugs

With only one, failed, clinical trial attempting to block the function of CR3, our possibilities of knowing the impact of such therapy is very limited. Below, two other examples of clinically used formulation that may hit CR3 functions, namely glatiramer acetate (GA; CopaxoneTM ) and simvastatin is brought up. Detailed insight on their CR3-directed functions in vivo is not available, but in vitro experiments may still provide functional evidence on their known anti-inflammatory properties involving CR3.

#### Neutrophil Inhibitory Factor in Amelioration of Stroke

It is increasingly evident that inflammatory responses play a significant role in adding to the morbidity of stroke. The role of ischemic reperfusion injuries in stroke has long been clear, while the molecular details of the complement system in aggravating such diseases is rather recent (143). Particular diseases of the central nervous system have benefitted from the use of magnetic resonance imaging (MRI) scanners. Recently, a study using ultra-small super paramagnetic iron oxide (USPIO) particles demonstrated that macrophages, or macrophage-like cells, in the stroke lesion are activated, and additional literature point to these CR3-positive cells as being aggravators of the disease (144, 145). Indeed, early studies in a rat model showed that administration of antibodies against CR3 lessened symptoms of experimentally induced strokes (146) and CR3-deficient mice are less susceptible to such injury (147). In humans, clinical trials were made with the compound UK-279,276, a recombinant analog of the hookworm protein neutrophil inhibitory factor (NIF). NIF is a relatively specific inhibitor of CR3 (148) and bind the αMI (149). In this way, the trials with UK-279,276 became the first, and to our knowledge the only, study to aim for direct inhibition of CR3. Although interactions with CR3 apparently was discernible in both preclinical models and in humans (150), the trials supported no evidence, unfortunately, of any benefit in stroke therapy (151). A significant reason was liver clearance and the formation of inhibitory antibodies to this non-human protein (150, 151). However, the study suggests that targeting of CR3 is well-tolerated, pointing to other pharmaceutical agents as a way forward.

# Glatiramer Acetate as an Antagonist of CR3 Function

GA is an effective drug in treatment of relapsing-remitting MS (152). It was among the earliest such treatments, in many ways paving the way for later anti-inflammatory therapies used for this disease. The drug itself is among the most complex formulations on the market. The active ingredient, glatiramer, is a mixture of random copolymers made from bulk synthesis by polymerization of the acetic anhydrides of glutamic acid, lysine, alanine, and tyrosine. After polymerization, chromatography is used to provide a heterogeneous mixture of copolymers with a narrow distribution in M<sup>r</sup> around 8,000, or 50–60 residues. This leads to the astonishing observation that the formulation, in principle, may contain any of 10<sup>30</sup> different co-polymers, while the pre-filled syringes with 20 mg only delivers 10<sup>17</sup> such copolymers (153). Although the theoretical number of co-polymers in the clinical formulations may be curbed by complex aspects of the polymerization process (154), there is little doubt that the patients rarely, if ever, receives the same co-polymer twice. The ratios of amino acid anhydrides were mixed to mimic the properties of MBP, one of the used autoantigens in experimental autoimmune encephalitis (EAE), and maybe an autoantigen in human MS as well (153). In effect, this means that GA also carries an excess positive charge from the high abundance of lysine residues. GA is capable of inhibiting EAE and reduces the frequency of attacks in relapsing-remitting MS with ∼30%. The pharmacological mode of action (PMA) remains enigmatic. Strong support for the copolymers acting as activators of a polyclonal Th2 type response has been provided (155). T cell proliferation may proceed even in the absence of professional antigen-presenting cells, suggesting that other, extracellular loading of MHC II molecules is a possibility. Stapulionis et al. considered the potential role of CR3 in contributing to the PMA of GA (127). Both cellular adhesion to MBP and iC3b were inhibited by the addition of GA to the medium in a concentration of 3µg/ml. This nicely matches the resulting concentration from a distribution of the applied clinical dosage of 20 mg in 6 liters of plasma. In agreement with difference in ligand binding selectivity discussed above, CR4 was not capable of binding GA. Experiments with the αMI in SPR with the methodologies mentioned above showed a K<sup>D</sup> of ∼10−<sup>4</sup> M. Finally, circular dichroism spectroscopy suggested a significant portion of unfolded polypeptide sequence in GA. Taken together, the mode of CR3 interaction with these co-polymers is probably very similar to the vast range of peptides reported to bind this receptor, with the binding supported by the unfolded character of at least some of the material. CR3 was already shown in animal models to be a factor in development of EAE (141). More recent evidence suggest that onset of the PMA in MS is fast, within hours of the first injection (156). This points away from the adaptive immune response as responsible for all of the effects in MS. However, GA is a complex drug, as demonstrated by the recent observations that it may directly kill T lymphocytes in process similar to LL-37, which is likewise known to possess immunomodulatory properties (157). The cytotoxicity toward prokaryotes only expands the possible therapeutic influences in MS (158). However, with the significant role of CR3-expressing macrophages and microglial cells in the pathogenesis of MS, the role of GA as a CR3 antagonist should not be overlooked as part of the PMA.

#### Simvastatin as Ligand Binding Kinetic-Dependent Antagonist of CR3

In pioneering studies by investigators from Novartis, it was demonstrated that lovastatin inhibited the function of LFA-1 (159, 160), thereby limiting T cell proliferation. Interestingly, the mechanism in this molecule involved a stabilization of the αLI in the closed conformation by the binding of the statin to the so-called L-site away from the MIDAS. This allosteric antagonism came as a surprise, since lovastatin in its activated form presents a carboxylate, which, from the findings mentioned earlier on αMI (159), was expected to chelate in the MIDAS. Jensen et al. (83) investigated the interaction between the openconformation αMI as well as activated CR3 receptors. In both types of assays, simvastatin, a compound highly similar to lovastatin and also an antagonist of αLI (159), inhibited the CR3 binding to iC3b, in the cellular experiment with an IC<sup>50</sup> in the order of 10µm, similar to IC<sup>50</sup> for inhibition of LFA-1 to ICAM-1. The simvastatin carboxylate (**Figure 5A**) was firmly chelated in the αMI MIDAS (**Figure 5B**), clearly advocating that the statin in this case acted as a competitive antagonist. A more detailed analysis was provided in SPR studies. With iC3b as a ligand, the total inhibition was limited, but discernible. Surprisingly, no inhibition was found with ICAM-1 as a ligand for αMI, a result also supported by earlier, but unexplained, findings (160). Closer inspection showed that inhibition of iC3b binding quantitatively came from a relatively select elimination of interactions with slow association and dissociation rates, while other types of interactions were left unaffected. Accordingly, the slow-binding-kinetic type of interaction was not found in the binding to ICAM-1, explaining why this ligand was not affected. It is not uncommon to find larger antagonist involved in complex binding schemes as recently demonstrated by the allosteric mechanisms of natalizumab, a MS drug, in antagoniszing the T cell adhesion molecule very-late antigen (VLA)-4 (161). However, that small-molecule drugs also seem to be capable of participating in complex inhibition reactions is more surprising with only speculative explanations provided so far.

#### CR4 as a Drug Target and What It may Help

With the more mysterious role of CR4, one should think it is difficult to identify clinically relevant inhibitors or other compounds influencing CR4 functions. Surprisingly, three examples can be extracted from the literature, one involving highly sulfated heparin fragments, another focusing on a food additive, OPN, which has been suggested to stimulate the immune system, and finally a potential relationship between adjuvants and CR4.

#### Heparin as a CR4 Ligand Binding Antagonist

Although both CR3 and CR4 were reported to support adhesion to heparin, a quantitative measurement with side-by-side comparison of the inhibitory potential of heparin fragments, clearly suggested that CR4 is the better receptor for heparin (112). As mentioned above, the affinity was strongly influenced by the level of sulfation, with higher sulfation strengthening the interaction. Likewise, the length of the heparin oligomers was important. Natural heparin has a degree-of-polymerization (dp) of ∼42 and was a strong inhibitor of CR4 with IC<sup>50</sup> at 0.30µm. Heparin with a dp21 (M<sup>r</sup> ∼ 6,000), similar to the low-molecular weight heparin used in the clinic (162), had a IC<sup>50</sup> of 0.1µm (0.6 mg/l) (88). This should be compared with the subcutaneously injected dosage of 1 mg/kg body weight (162). The simple calculation does not address the complex issue of distribution volume, but it seems not impossible that clinical injections of heparin may reach a concentration sufficient to impact the function of CR4. This should be compared with the effect of fondaparinux (ArixtraTM ), a pentameric, artificial heparin-like compound, which is capable of preventing coagulation (163). It showed no quantitative interaction with CR4 (112). This opens an interesting perspective on how to design artificial heparins. Accelerated by the so-called "Heparin crisis" in the early 2000s, where contaminated heparin provoked severe hypersensitivity responses in patients treated with contaminated heparin, a clinical unmet need exists in producing safe, synthetic formulations (164).

#### OPN, Immunostimulatory Food Additive

OPN is a highly phosphorylated protein, which serves roles both in the bone matrix and beyond. It is possible to purify the protein from both human and bovine milk. It is also found in human serum with some association of the concentration with human diseases, possibly suggesting a use of OPN as a biomarker. Its proposed role in immunology mainly stems from association with inflammatory diseases such as arthritis (135). Experimental studies with milk formula enriched in OPN suggested that OPN increases the number of circulating T lymphocytes in formulafed infants (165). From these and other data, an interleukinelike function seems likely (166). This opens the discussion on what receptors, expressed in leukocytes, are relevant. As already mentioned in Section Ligand binding selectivity of CR3 and CR4, CR4 binds OPN strongly, probably in consequence of the negative charge of this protein contributed by multiple glutamate side chains. CR4 is, however, not the only OPN receptor. It contains an Arg-Gly-Asp (RGD) motif, which has already been demonstrated to mediate interactions with β<sup>3</sup> integrins, and other integrins have been reported as receptors as well (167). With the high expression of CD11c/CD18 on dendritic cells and the central role of these cells in regulating intestinal cytokine levels and leukocyte proliferation, notably T lymphocytes, it seems a straight forward proposal that the strong CR4 binding of this molecule play a role in these observations.

#### CR4 as a Target in Vaccination

An emerging literature has pointed to CR4 as an important target in vaccination (168–170). As noted in Section Functions of CR3 and CR4 and the family of CD18 integrins, CR4 (CD11c/CD18) has long been established as a marker for dendritic cells with a particular high expression in murine dendritic cells (15). With its role as a complement receptor, CR4 is probably significant in the phagocytic uptake by these cells. In principle, this would enable the presentation of peptides from the phagocytozed antigen on MHC II molecules to CD4+ T lymphocytes. However, as shown by Castro et al. antigens conjugated to antibodies to CD11c are capable of raising a CD8+ T lymphocyte response (168). The therapeutic advantages of such a response includes T cell targeting to cancer cells or intracellular infections difficult to limit with an antibody response. Interestingly, the mechanism here seems to rely on cross presentation by dendritic cells, that is, presentation of phagocytozed proteins on MHC I molecules. Although promising, CD11c-targeted vaccines has not formally been tested in humans.

From what we now know of the protein binding properties of CR4, it is possible to ask the question if CR4 already is a part of vaccine responses. Jalilian et al. reviewed the use of adjuvants, mainly in influenza vaccination (171). In spite of some adjuvant formulations having been used for almost a 100 years, we know surprisingly little about their therapeutic effects. The particulate nature of these compounds, notably aluminum salts, seems to suggest that protein deposition on such surfaces could play a role in their interaction with antigen presenting cells visà-vis the expression of CR4. The deposition of complement and fibrinogen/fibrin would probably occur more efficiently on the particle-embedded antigen than for the free antigen. In addition, the complex processes leading to denaturation of particle surfaceadsorbed proteins apparently further aids the interaction with CR3 (172) and possibly CR4 from binding to such material. Finally, it is well-known that receptor-mediated phagocytosis requires a particle size of about 50 nm, which is a hard-to-reach dimension limit by applying soluble antigens. Taken together, this evidence suggests that a valuable direction of optimizing vaccine adjuvants would include a closer examination of CR3 and CR4 in this context (171).

#### CONCLUSION: CR3 AND CR4, SIGNIFICANT CONTRIBUTORS TO BOTH INNATE AND ADAPTIVE IMMUNITY

Are CR3 and CR4 simple double-ups in the leukocyte cell membranes? Three conclusions from the literature answer this question in the negative.

First, although both receptors unquestionably bind the C3 fragment iC3b, it is very clear that the principle of recognition and bindings sites in use are non-overlapping. Structural and functional analyses do not suggest any striking similarity as to what make CR3 and CR4 complement receptors. iC3b is a large, multidomain protein, which binds a plethora of different proteins, and accommodates all the critical features that enable binding by both receptors. Undoubtedly, this may support an altogether stronger affixing of complement-opsonized antigen to any particular myeloid cell surfaces.

Second, a striking property of both CR3 and CR4 is the large number of reported ligands, some distinct for each receptor, others shared. When comparing with LFA-1, which binds essentially only ICAMs, several sources of experimental evidence suggest that CR3 and CR4 indeed share a stronger ability to chelate carboxylate groups. Maybe for this reason, they also bind denatured or natively unfolded species better than their folded counter-parts. On the other hand, we know little about the strength of carboxylate chelation in the integrin family, as to our knowledge this has so far only been investigated in comparison of the αMI, αXI, and αLI. From the long known ability of the betachain I-like domain in β<sup>1</sup> and β<sup>3</sup> integrins to bind the minimal RGD motif, it is tempting to suggest that CR3 and CR4 are less unusual encounters in the integrin world than LFA-1. In other words, their ligand binding promiscuity is less of a unifying trait than otherwise could be thought.

Third, even if CR3 and CR4 share ligands, it is now possible to rationalize a different ligand binding selectivity. For more than a decade, it has been clear that polyanionic species, including both negatively charged carbohydrates and proteins are particularly strong ligands for CR4. By contrast, both studies over individual ligands as well as more systematic analyses, suggest that CR3 has preference for cationic species in so far as these may also offer a carboxylated moiety for chelation of the MIDAS. In this sense, CR3 and CR4 nicely fits with properties earlier attributed to scavenger receptors, where at least some also accommodate the binding of homogenously charged species, reflecting a decayed state of macromolecules, including charge exposure of denatured states of proteins.

With these properties of ligand recognition, one surprising observation embodies the clinical formulations in use, which may affect the function of CR3 and CR4. From simvastatin, one of the most often used drugs in the world, to heparin, a classic anticoagulant, in vitro, evidence suggests an impact of these drugs on the human immune system. Systematic studies over this impact are lacking, however, probably in part because we need a better understanding of the pharmaceutical benefits from drugs targeting CR3 and CR4. With the recent findings of the roles of complement, including both CR3 and CR4, in NK cell biology, as part of the use of therapy with monoclonal antibodies, these topics are likely to become of high significance.

CR3 and CR4 entered immunology almost a decade before the concept of innate and adaptive immunity was coined. Between them, they expand on both classic and more enigmatic ideas of the function of the immune system. With their functional similarity as complement receptors for C3 fragments, as well as their primary expression in myeloid leukocytes, it is justified to place their role in the innate immune system. However, at least in the case of CR3, its ability to bind C3d-opsonized antigens in conjunction with CR2 on B lymphocytes, highlights a well-established theme of complement as the primer of antibody formation. In addition, it is not without interest that this interaction seems to also highlight another recurrent phenomenon in lymphocyte biology, namely the importance of membrane-presentation of antigens to lymphocytes. The role

#### REFERENCES


of CR4 is less well understood, yet its abundant expression on dendritic cells places it in the center of modern immunology. From the review made above, a summary of the comparison between CR3 and CR4 points to a consort of receptors, which together embodies both a surprising versatility and complementarity in molecular recognition mechanisms.

#### AUTHOR CONTRIBUTIONS

TV-J and RJ wrote the paper and made the figures.

#### ACKNOWLEDGMENTS

We wish to thank Dr. Søren E. Degn for fruitful discussions and Dr. André Walter for proofreading the manuscript. RJ was supported by the Lundbeck Foundation Brainstruc Center. TV-J was supported by the Carlsberg Foundation (990760/20-1328 & 0122/20; 2005-1-711) and the Danish Multiple Sclerosis Association (R34-A255-B143; R62-A619-B143; R89-A1581- B143; R367-A25613-B143; R431-A29821-B143). This work was also supported in part by the French National Research Agency (ANR-15-CE18-0026-01).


(2001) 31:1173–80. doi: 10.1002/1521-4141(200104)31:4<1173::AID-IMMU1173>3.0.CO;2-9


CD21) with its ligands C3d, iC3b, and the EBV glycoprotein gp350/220. J Immunol. (2001) 167:1490–9.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Vorup-Jensen and Jensen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The emerging Link Between the Complement Cascade and Purinergic Signaling in Stress Hematopoiesis

*Mariusz Z. Ratajczak1,2\*, Mateusz Adamiak <sup>2</sup> , Magda Kucia1,2, William Tse1 , Janina Ratajczak1 and Wieslaw Wiktor-Jedrzejczak3*

*1Stem Cell Institute at James Graham Brown Cancer Center, University of Louisville, Louisville, KY, United States, 2Department of Regenerative Medicine, Center for Preclinical Research and Technology, Warsaw Medical University, Warsaw, Poland, 3Department of Hematology Warsaw Medical University, Warsaw, Poland*

Innate immunity plays an important role in orchestrating the immune response, and the complement cascade (ComC) is a major component of this ancient defense system, which is activated by the classical-, alternative-, or mannan-binding lectin (MBL) pathways. However, the MBL-dependent ComC-activation pathway has been somewhat underappreciated for many years; recent evidence indicates that it plays a crucial role in regulating the trafficking of hematopoietic stem/progenitor cells (HSPCs) by promoting their egress from bone marrow (BM) into peripheral blood (PB). This process is initiated by the release of danger-associated molecular patterns (DAMPs) from BM cells, including the most abundant member of this family, adenosine triphosphate (ATP). This nucleotide is well known as a ubiquitous intracellular molecular energy source, but when secreted becomes an important extracellular nucleotide signaling molecule and mediator of purinergic signaling. What is important for the topic of this review, ATP released from BM cells is recognized as a DAMP by MBL, and the MBL-dependent pathway of ComC activation induces a state of "sterile inflammation" in the BM microenvironment. This activation of the ComC by MBL leads to the release of several potent mediators, including the anaphylatoxins C5a and desArgC5a, which are crucial for egress of HSPCs into the circulation. In parallel, as a ligand for purinergic receptors, ATP affects mobilization of HSPCs by activating other pro-mobilizing pathways. This emerging link between the release of ATP, which on the one hand is an activator of the MBL pathway of the ComC and on the other hand is a purinergic signaling molecule, will be discussed in this review. This mechanism plays an important role in triggering defense mechanisms in response to tissue/organ injury but may also have a negative impact by triggering autoimmune disorders, aging of HSPCs, induction of myelodysplasia, and graft-versus-host disease after transplantation of histoincompatible hematopoietic cells.

#### Keywords: complement cascade, ATP, mannan-binding lectin, sterile inflammation, purinergic signaling

#### INTRODUCTION

The basic function of innate immunity is to alarm the organism of an infection or tissue/organ injury in order to launch an appropriate response. An important part of this response is the release or mobilization of effector cells, such as granulocytes, monocytes, and lymphocytes, from bone marrow (BM) and other hemato-lymphatic organs into peripheral blood (PB) and the lymphatics and will be involved in eliminating invading pathogens (1–4). In parallel, hematopoietic stem/progenitor cells

*Edited by:* 

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Peter A. Ward, University of Michigan, United States Janos Szebeni, Semmelweis University, Hungary*

> *\*Correspondence: Mariusz Z. Ratajczak mzrata01@louisville.edu*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 26 March 2018 Accepted: 24 May 2018 Published: 05 June 2018*

#### *Citation:*

*Ratajczak MZ, Adamiak M, Kucia M, Tse W, Ratajczak J and Wiktor-Jedrzejczak W (2018) The Emerging Link Between the Complement Cascade and Purinergic Signaling in Stress Hematopoiesis. Front. Immunol. 9:1295. doi: 10.3389/fimmu.2018.01295*

**69**

(HSPCs) are also released, which locally supply mature granulocytes or dendritic cells by clonal expansion of progenitors in the damaged tissues (3–6). Moreover, in addition to HSPCs, other types of stem cells are also released at a much slower pace into the circulation, including (i) mesenchymal stem cells (MSCs), (ii) endothelial progenitor cells (EPCs), and (iii) rare, primitive very small embryonic-like stem cells (VSELs). If needed, all of these stem cells may be involved in repair mechanisms in damaged tissues (4, 7–9).

Bone marrow is a semi-solid tissue spread within the spongy or cancellous portions of bones and contains hematopoietic "red marrow," which is the most important source of cells circulating in PB and in the lymphatics (1–6). The estimated total mass of BM tissue in an average human being is as much as 6 pounds. This dynamic organ daily produces approximately 5 × 1011 erythrocytes, leukocytes, monocytes, and platelets, which enter the systemic circulation by crossing the BM–PB barrier *via* a permeable vasculature of small-vessel sinusoids within the medullary cavity. As mentioned above, BM is also the birthplace of stem cells that circulate in PB (1–6). While stem cells reside in stem cell niches, which are located around small vessels (endothelial niches) and in contact with osteoblasts lining trabecular bones in BM (osteoblastic niches), granulocytes, monocytes, and other types of maturing hematopoietic cells (mostly erythroblasts) occupy the entire volume of the hematopoietic microenvironment (1, 10–13).

Under steady-state conditions, maturing erythrocytes, leukocytes, monocytes, and platelets enter the PB to replace blood cells that have a limited half-life along with stem cells that are patrolling peripheral tissues, keeping the stem cell pool at distant locations of the hematopoietic microenvironment in balance (1–6). This balance may rapidly change in response to inflammation and tissue/organ damage when more cells need to be released into the circulation. This requires intensification of hematopoiesis in the BM microenvironment to supply more blood cells, while at the same time more stem cells are released from their BM niches (1–4). Increased release of cells from BM occurs also in clinical settings after pharmacological mobilization of HSPCs in response to administration of certain pro-mobilizing drugs, such as granulocyte colony-stimulating factor (G-CSF), CXCR4 receptor antagonists, or some chemokines (growth-regulated protein beta, Gro-β) (14–16).

In this review, we will present the accumulated evidence that a major orchestrator in the release of cells from BM into PB is the complement cascade (ComC), which induces a "sterile inflammation" state in the hematopoietic microenvironment (17, 18). The ComC can be activated by the classical, alternative, or mannan-binding lectin (MBL) pathways. Recent evidence indicates that acute activation of the MBL pathway of ComC activation plays the most important role in the release of cells from BM in response to tissue/organ injury, pathogens, and certain pro-mobilizing drugs (17–19). On the other hand, chronic activation of the MBL pathway is most likely an important element in BM aging and myelodysplasia (20–23). This pathway also likely contributes based on some clinical observations to induction of graft-versus-host disease (GvHD) after histoincompatible hematopoietic transplantation (24–27).

What is important for the topic of this review is that the MBL pathway of ComC activation is triggered by danger-associated molecular patterns (DAMPs) (28–33). Adenosine triphosphate (ATP) is one of the most important members of this family of molecules. However, it is well known that this ubiquitous intracellular molecular energy source, when secreted from cells, becomes an important signaling molecule and mediator of purinergic signaling (34–36). The release of ATP from cells in the BM microenvironment provides a molecular basis, involving activation of the ComC, for the link between purinergic signaling and activation of the innate immune response.

In this review, we will focus on the role of this ATP-mediated link between purinergic signaling and innate immunity in BM stem cell homeostasis, mobilization, and aging as well as in certain pathological conditions, including myelodysplasia and GvHD. Because of space limitation, our short review will not discuss several pathologies related to (i) chronic activation of ComC seen in paroxysmal nocturnal hemoglobinuria or atypical hemolytic-uremic syndrome, (ii) coagulation consequences due to interaction between ComC and coagulation cascade (CoaC), and (iii) ComC activation related to some cases of leukopenia or thrombocytopenia.

#### RETENTION OF HSPCs IN BM AND THEIR RELEASE DUE TO ACTIVATION OF INNATE IMMUNITY

Hematopoietic stem/progenitor cells reside in BM niches, and some important mechanisms mediating their BM retention have already been identified (1, 10–13, 37). The most important mechanisms include (i) the interaction between the chemokine receptor CXCR4 expressed on the surface of HSPCs and its specific ligand, the α-chemokine stromal-derived factor 1 (SDF-1) expressed by cells in stem cell niches and (ii) the interaction between the integrin receptor known as very late antigen 4 (VLA-4), which is expressed by HSPCs, and its ligand in stem cell niches, vascular adhesion molecule 1 (VCAM-1) (1–4). What is important for the retention process is that both receptors, CXCR4 and VLA-4, are located in special cell membrane domains enriched for cholesterol and glycosyl phosphatidylinositol anchor protein (GPI-A) known as membrane lipid rafts (38, 39). Of note, the same membrane lipid rafts also contain the cell-surface proteins CD55 and CD59 that regulate complement activity (29, 30, 40). Accumulating evidence indicates that the structural integrity of membrane lipid rafts on the surface of HSPCs is important for their retention in BM niches (38, 39). A significant role in retention of HSPCs in BM niches is also played by the third protein component (C3) of the ComC, as its cleavage fragments, C3a and desArgC3a, promote incorporation of CXCR4 and VLA-4 into membrane lipid rafts (41). In addition, the interaction of C3a with C3aR, which is expressed on the surface of HSPCs, directly increases adhesion of HSPCs in the BM microenvironment (41).

Results from our group also indicate that the release of HSPCs from BM niches into PB in response to administration of pharmacological mobilizing agents, as well as to mediators released during tissue/organ injury, is triggered by activation of the ComC (4, 18, 19). The same mechanism plays a pivotal role in the release of other types of stem cells, including MSCs, EPCs, and VSELs. In support of the regulatory involvement of the ComC in the retention of HSPCs in BM niches, we have already demonstrated that, while blockage of C3aR on the surface of HSPCs promotes the mobilization process (42), cleavage of the fifth protein component (C5) and release of C5a and desArgC5a anaphylatoxins is crucial for egress of HSPCs into PB (43). Mice that were deficient in C5 and C5aR turned out to be poor mobilizers (43). We propose that the proximal and distal part of ComC regulates retention of HSPCs in the BM microenvironment in opposite manner (18, 42, 43). While activation of the proximal part of this cascade *via* C3 cleavage fragments promotes retention of cells in BM, activation and cleavage of C5 have the opposite effect, as C5 cleavage fragments promote their egress (4, 18, 42, 43). This demonstrates a fine-tuned ComC-mediated mechanism in auto-controlling this process.

At the beginning of our work on the role of the ComC in regulating trafficking of HSPCs, we posed the basic question of which of the ComC-activation pathway (classical, alternative, or MBL) plays a crucial role in triggering egress of cells from BM. Initially, we considered the involvement of the classical pathway. To our surprise, however, mice deficient in the C1q component of classical pathway activation turned out to be good mobilizers in response to administration of the most commonly used HSPC mobilizing agent, G-CSF (44). Therefore, we shifted our attention to the MBL pathway of ComC activation and performed mobilization studies in MBL-KO animals (19). In our experiments, MBL-KO or wild-type (WT) control mice were mobilized with G-CSF or the CXCR4 antagonist AMD3100. We found that MBL-KO animals displayed a significant decrease in the release of cells from BM into PB compared with control WT mice (19). This result provided evidence for the pivotal role of the MBL pathway in the mobilization process. However, despite a significant decrease in egress of HSPCs from BM to PB, this process was not completely inhibited, which suggests the presence of redundant pro-mobilizing mechanisms. Based on our finding that factor B deficiency in mice also impairs mobilization of HSPCs, the persistence of some level of mobilization in MBL-KO mice could be explained by parallel activation of an alternative pathway (45). This possibility is currently being investigated in more details in our laboratory.

#### THE PIVOTAL ROLE OF THE MBL PATHWAY OF ComC ACTIVATION IN TRIGGERING MOBILIZATION OF HSPCs

Recognition of the involvement of the MBL pathway in egress of cells from BM not only further supported a crucial role of innate immunity in triggering the mobilization process but also shed more light on the cellular and molecular events regulating this process. Our understanding of this phenomenon is supported by the experimental data depicted in **Figure 1**.

As indicated in **Figure 1**, pharmacological mobilizing agents, including recombinant G-CSF, synthetic AMD3100, and natural mediators of inflammation or tissue organ/injury, such as (i) endogenous G-CSF secreted by endothelium, macrophages and immune cells, (ii) C5a and C3a released in damaged tissues, (iii) interleukin 8 (IL-8), (iv) bacteria-derived N-formylmethionylleucyl-phenylalanine, or (v) leukotriene B4 (LTB4) secreted from activated granulocytes, are all able to initiate the sequence

FIGURE 1 | Interaction between elements of purinergic signaling and activation of the complement cascade (ComC) in the induction of sterile inflammation in bone marrow (BM). Stimulatory factors released during tissue/organ injury, systemic mediators of inflammation, and pharmacological inducers of hematopoietic stem/progenitor cells (HSPC) mobilization activate Gr-1+ leukocytes in BM to release danger-associated molecular patterns (DAMP) molecules, including adenosine triphosphate (ATP) and reactive oxygen species (ROS). As a DAMP molecule, ATP is recognized by MBL, which activates the ComC and CoaC in an MASP-dependent manner (*indicated on a graph as 1*). By contrast, ROS exposes neoepitopes, and neoepitope–IgM complexes are also recognized by mannan-binding lectin (MBL). This leads to activation of the ComC by the MBL-dependent pathway. Both classical C5 convertase, as a product of C3 cleavage, and C5-like convertase activity, provided by thrombin cleaving C5 to release cleavage fragments C5a and desArgC5a, are crucial in the egress of HSPCs from BM. In addition to serving as a DAMP (*indicated on a graph as 2*), ATP also activates purinergic receptors expressed on the surface of HSPCs, in which P2X7 plays an important role in promoting calcium influx into cells (*indicated on a graph as 2*). This facilitates intracellular actin fiber rearrangement that is crucial in cell migration and egress from BM.

of events leading to activation of the ComC (1–6). The important role of BM-residing leukocytes, which are crucial in the mobilization process, has been demonstrated by seminal papers showing that neutrophil depletion in BM negatively affects the efficiency of this process (46, 47).

Overall, activated leukocytes release several pro-inflammatory factors, including DAMPs and free radicals (ROS) (5, 48–50). DAMP molecules secreted by leukocytes include mainly ATP but also other members of this family, including high mobility group box 1 (HMGB-1) protein, heat shock proteins, and the S100 multigenic family of calcium-modulated proteins (28). What is highly relevant for the topic of this review is that ATP is the most important DAMP and is recognized by a soluble pattern-recognition receptor, MBL (19, 29, 40, 51). On the other hand, in addition to DAMPs, cells under stress release reactive oxygen species (ROS). When released from leukocytes, ROS expose neoepitope antigens on the surface of cells in the BM microenvironment that are recognized by naturally occurring antibodies, mainly from the IgM class (52). Of note in addition to ATP, neoepitope–IgM complexes are also recognized by the same MBL molecule (**Figure 1**).

In the next step, MBL activates mannan-binding serum proteases (MASPs) that cleave C3 and thereby trigger ComC activation in the MBL-dependent pathway (19, 33). As shown in **Figure 1**, MASP-1 activates the CoaC in parallel (33). Activation/cleavage of C3 creates C5 convertase, which cleaves C5 to the anaphylatoxins C5a, desArgC5a, and releases iC5b during the cleavage process. iC5b, in turn, is involved in generation of the membrane attack complex (C5bC9) (29, 43). Moreover, in parallel, cleavage of C5 is augmented by thrombin generated during activation of the CoaC, as thrombin is a proteolytic enzyme with C5 convertase-like activity (53). This activity explains why both the ComC and the CoaC are activated during the mobilization process (54, 55).

As mentioned above, activation of the distal part of the ComC is crucial for the egress of cells from BM. First, after the ComC is activated in the BM microenvironment, C5a and desArgC5a activate granulocytes that help to release HSPCs from their niches by (i) secretion of several proteolytic enzymes that disrupts the SDF-1–CXCR4 and VCAM-1–VLA-4 retention axes operating between HSPCs and the cells lining the BM niches and (ii) release of phospholipase Cβ2 that digests the GPI-A component of membrane lipid rafts, which is crucial in maintaining lipid raft integrity (38). Disruption of membrane lipid rafts negatively impacts the retention functions of the CXCR4 and VLA-4 receptors, which are membrane lipid raft-associated receptors (38). Next, the ComC activated in BM sinusoids directly chemoattracts granulocytes, which are the first cells to egress from BM into circulation. These cells are rich in proteolytic enzymes and help to disrupt the endothelial barrier and thus pave the way for HSPCs to follow behind (43). Finally, the HSPCs that are released from their niches follow a steep gradient of bioactive sphingolipids, such as sphingosine-1-phosphate (S1P) and ceramide-1-phosphate (C1P), which are present at high concentrations in BM sinusoids (5, 56–58). Both of these phosphosphingolipids are potent chemoattractants for HSPCs at the physiological concentrations present in PB (56). The gradients of both S1P and C1P are already very steep under steady-state conditions in PB and may additionally steepen due to the release of S1P form red blood cells in BM sinusoids exposed to MAC. As mentioned above, the egress of HSPCs that do not respond directly to a C5a chemotactic gradient is facilitated by granulocytes, which are the first cells to egress BM in a C5a gradient-dependent manner (43).

Besides activating MBL, ATP released from cells activates in parallel certain purinergic receptors on the cell surface that augment the mobilization process. The most important of these receptors seems to be a P2 family member, the P2X7 receptor ion channel (6, 17, 59, 60). As discussed below, P2X7 allows an influx of Ca2<sup>+</sup> ions into cells that activate changes in the cell cytoskeleton that are important for cell migration and adhesion (61).

#### PURINERGIC SIGNALING IN BM AND ITS LINK TO ComC ACTIVATION

As shown in **Figure 1**, ATP is an important DAMP and extracellular nucleotides (EXN) that is released from activated neutrophils, and as a DAMP, it activates the MBL pathway of the ComC, and as an EXN, it activates purinergic signaling pathways that additionally promote egress of HSPCs from BM into PB (17, 61–69).

Purinergic signaling is an ancient form of extracellular signaling mediated by EXNs, including most importantly the purine ATP and its metabolite nucleoside, adenosine (34). Purinergic signaling also involves certain rare extracellular pyrimidines, such as UTP and UDP. Purinergic receptors for EXNs are expressed on all cells in the body and are represented by several families of P1, P2X, and P2Y receptors, which are among the most abundant receptors in living organisms (34). HSPCs express several receptors that belong to two different purinergic receptor families, P1 and P2 (34). While the P1 receptor family consists of four G protein-coupled receptor subtypes, A1, A2A, A2B, and A3, which are activated by adenosine (62), the P2 family includes a total of eight receptors (P2Y1, 2, 4, 6, 11, 12, 13, and 14) identified so far, which are G protein-coupled receptors and respond to stimulation by ATP, ADP, UTP, and UDP. The P2X ionotropic channel receptor family consists of seven members (P2X1, 2, 3, 4, 5, 6, and 7), which are activated by ATP (34).

However, the main purpose of this review is to show the role of EXNs and purinergic signaling in inducing sterile inflammation in BM, which plays a role in the mobilization of cells into PB, and it is important to realize that EXNs also have pleiotropic effects in regulating hematopoiesis (59–64). For example, EXNs, particularly ATP and adenosine, have been reported to promote proliferation of HSPCs in zebra fish and murine embryos (63). By contrast, UTP has been reported to inhibit the proliferation and migration of leukemic cells. The overall role of purinergic signaling in maintaining BM homeostasis is discussed in excellent review elsewhere (64).

It has been postulated that in the induction of sterile inflammation in BM a crucial role is played by ATP, which is secreted from activated BM cells, mainly granulocytes, *via* pannexin channels as a DAMP molecule, and as we have demonstrated, pharmacological inhibition of pannexin by employing a drug (probenecid) or a specific anti-Panx1 blocking peptide decreases G-CSF- and AMD3100-induced mobilization of HSPCs (17). Connexin-43 is also involved in the release of ATP, and some ATP is also secreted in an extracellular microvesicle-dependent manner (31, 65). The involvement of connnexin 43 gap junction proteins in ATP secretion is supported by the fact that connexin-43-KO mice are poor mobilizers (65). This defect could be at least partially explained as we envision by impaired release of ATP from cells.

Based on this finding, ATP secreted by granulocytes and other BM cells is recognized as a DAMP by MBL, which triggers activation of the ComC (**Figure 1**). On the other hand, as depicted, ATP also interacts with P2 purinergic receptors, and the P2X7 receptor plays an especially pivotal role in mobilization. In support of this notion, we found that P2X7-KO mice are poor mobilizers (17). Moreover, studies in chimeric mice in which WT animals were reconstituted with P2X7 BM cells, and P2X7-KO mice were reconstituted with WT marrow cells revealed that this defect is due to a lack of P2X7 on the surface of hematopoietic and not non-hematopoietic cells in the hematopoietic microenvironment (17).

**Figure 2** shows that when ATP is released into the extracellular space, if it does not bind to MBL, it activates the P2X7 receptor to allow influx of Ca2<sup>+</sup> or, in parallel, is converted by cell-surface ectonucleotidase CD39 to AMP, which is then converted by ectonucleotidase CD73 to the nucleoside adenosine (34, 64). The importance of the purinergic signaling cascade is further supported by our recent observation that CD73-deficient mice, which because of ectonucleotidase deficiency have less free adenosine in the extracellular space, mobilize greater numbers of HSPCs, which indicates a negative regulatory role for adenosine in the mobilization process (17). The inhibitory mobilizing effect of adenosine has been confirmed by injecting mice with this nucleoside along with pro-mobilizing agents (17). These results demonstrate that, while ATP triggers and promotes the mobilization process, adenosine generated from ATP provides negative regulatory feedback and plays an opposing inhibitory role (**Figure 1**).

Based on these findings, ATP triggers, on the one hand, as a DAMP, activation of the ComC in an MBL-dependent manner and, on the other hand, regulates the mobilization process in a more complex way by activating P2 receptors and providing negative feedback control mechanisms for this process through its metabolite, adenosine, which engages P1 receptors on the surface of cells (17, 62, 64).

What is also important is to realize that purinergic receptors are expressed on the surface of several types of cells that comprise innate immunity cellular components, including granulocytes, basophils, eosinophils, monocytes, and dendritic cells as well as cells in the BM microenvironment, including stromal cells, osteoblasts, osteoclasts, pericytes, and endothelial cells (64). ATP is also released from the synapses of neural fibers innervating BM tissue (34). Furthermore, the C3 and C5 cleavage fragment receptors, C3aR and C5aR, respectively, are expressed by several hematopoietic and non-hematopoietic cells in the BM microenvironment (41–43). All these add significant complexity to the relationship between purinergic signaling and innate immunity in BM and requires further study.

Moreover, it is important to pin point that extracellular ATP also exerts strong pro-inflammatory effects that are independent from MBL activation (28). ATP may also activate NLRP3 inflammasome pathway in cells that controls in caspase-1-dependent manner maturation of two important pro-inflammatory members of interleukin (IL-1) family cytokines—namely IL-1β and IL-18 (67–69). It has been postulated that activation of the NLRP3 inflammasome is regulated at both the transcriptional and post-translational levels. While the transcription of inflammasome is induced by the signal mediated by toll-like receptor/

several ectonucelotidases, including cell-surface-expressed CD39 and CD73, and these are crucial to generating extracellular adenosine. While ATP promotes sterile inflammation in the BM microenvironment and mobilization of hematopoietic stem/progenitor cells (HSPCs), adenosine has the opposite effect. Inhibitors of ectonuclotidases facilitate sterile inflammation in BM and egress of HSPCs. By contrast, inhibitors of adenosine receptors are expected to inhibit this process.

nuclear factor-kB pathway, NRLP3 inflammasome activation at post-translational level is mediated e.g., by various DAMPs including ATP. Both IL-1β and IL-18 are released from cells in response to ATP-mediated activation of NLRP3 inflammasome and potentiate state of sterile inflammation in BM microenvironment that promotes mobilization process.

# PURINERGIC SIGNALING, INNATE IMMUNITY, BM STERILE INFLAMMATION, AND HEMATOPOIETIC AGING

Aging is an inevitable and complex process involving a sequence of pathological events (22, 23, 66). Several mechanisms are currently being proposed that accelerate this process at the cellular level, including shortening of the tips of chromosomes (telomeres); generation of ROS, which contribute to replication stress and oxidative DNA damage; impairment over time of the process of autophagy, a major degradation pathway essential for removing damaged organelles and macromolecules from the cytoplasm in eukaryotic cells, and which promotes recycling of amino acids during periods of starvation; the occurrence of pathologic lipid metabolism; and chronic inflammation (22, 66).

Aging also occurs in the hematopoietic system and is characterized by a myeloid bias, in which BM increases the number of myeloid progenitors along with impaired differentiation of these cells, gradually developing anemia and decreasing the number and fitness of B cell progenitors, which is accompanied by oligoclonal expansion of memory B and T cells (22, 23). Additional evidence indicates that this process is triggered by low-grade chronic inflammation that is a result of an increase in activation of innate immunity and impaired acquired immunity (22, 23). These changes may lead to the appearance of BM myelodysplasia and, in consequence, to clonal cell expansion and overt leukemia (22). These changes also lead to an increased incidence of autoimmunological diseases that are observed in patients with advancing age (22, 23, 66).

Since aged cells in the BM microenvironment are a rich source of DAMP molecules, including mostly ATP and HMGB-1, one can speculate that low-grade chronic inflammation in the BM microenvironment may be triggered by chronic activation of the ComC (6). A similar mechanism is also most likely involved in the aging of other organs, such as heart, brain, or kidney. This suggests that an effective countermeasure to ameliorate this unwanted effect would be anti-inflammatory treatment.

To shed more light on this phenomenon, there is a need for more long-term studies in ComC-deficient mice to see whether these animals are protected from age-related dysfunction of vital organs. The earlier study showing that C3-deficient mice fail to display age-related hippocampal decline lends support to undertaking more complex studies (70). In particular, it would be interesting to see whether MBL-KO mice or C5-KO mice are endowed with an extended life span and develop fewer agerelated pathologies in vital organs. Such experiments, of course, should prevent exposure of these animals to potential pathogens, as their susceptibility to infection may affect the final long-term experimental results.

# INNATE IMMUNITY AS A POTENTIAL TRIGGER OF GvHD

Graft-versus-host disease is a serious medical complication seen in patients who are recipients of transplanted tissue or cells from a genetically histoincompatible donor (25, 71). GvHD is commonly observed after hematopoietic stem cell transplants, when T cells present in the graft attack the tissues of the transplanted recipient. After perceiving host tissue antigens, among them the human leukocyte antigens, as antigenically foreign, T cells produce an excess of cytokines, including tumor necrosis factor alpha (TNFα) and interferon gamma (IFNγ) (25, 27, 71).

An important question remains: To what degree are innate immunity, in particular the ComC, and purinergic danger signaling involved in triggering this T-cell-mediated process? Unfortunately, conclusive experiments in animal models have not been performed. However, a very recent report indicates that patients with defects in activation of the MBL pathway of ComC activation are partially protected from this so often devastating transplant complication (25). It would be interesting to see whether MBL-KO mice are more resistant to GvHD after allogeneic BM transplants than their WT littermates.

#### THERAPEUTIC IMPLICATIONS FOR MODULATING STERILE INFLAMMATION IN BM

While activation of the ComC is important for optimal mobilization, its inhibition is highly relevant to ameliorating the chronic, sterile inflammation process seen in aging and myelodysplasia. Inhibition of the ComC may also be of importance in ameliorating the onset of GvHD, which occurs after infusion of histoincompatible hematopoietic cells.

Overall, since ATP-mediated activation of the MBL pathway of ComC activation leads to induction of sterile inflammation in the BM microenvironment, an anti-inflammatory treatment may have the opposite effect. However, this process, as depicted in **Figure 2**, tends to be somewhat self-limiting due to ATP conversion to adenosine. In fact, adenosine is known in immunology as an anti-inflammatory nucleoside (72). Therefore, appropriate activators of adenosine receptors would help to control sterile inflammation in the BM microenvironment. However, in proposing such a treatment, one would have to take into consideration the fact that adenosine is a powerful cardiovascular mediator, and a hyperphysiological dosage of adenosine mimetic may lead to cardiac complications (34). Moreover, taking into consideration the involvement of the P2X7 receptor in activating sterile inflammation in BM (6, 17), it would be important to test whether specific inhibitors of this receptor could be employed as potential anti-inflammatory drugs to dampen sterile inflammation in BM.

Another recently identified inhibitor of stem cell mobilization is heme oxygenase 1 (HO-1) (73–78). This anti-inflammatory enzyme, which is induced by oxidative stress in the BM microenvironment, counteracts the induction of sterile inflammation in BM. We provided evidence that HO-1 is a potent inhibitor of hematopoietic cell migration and the responsiveness of HSPCs to crucial chemoattractants, such as S1P, C1P, and SDF-1 (72, 78). Moreover, mice that lack HO-1 are easy mobilizers (72). The biological anti-inflammatory and ComC-activation properties of HO-1 have been demonstrated both in HO-1 deficient mice and in a case of rare human HO-1 deficiency in which the ComC became continuously hyperactivated (77). This hyperactivity of the ComC related to HO-1 deficiency leads to chronic inflammation in affected individuals. In our most recent work, we demonstrated that ATP and adenosine directly modulate expression of HO-1 in hematopoietic cells (17). While ATP inhibits HO-1 expression at the mRNA level, adenosine, by contrast, upregulates HO-1 mRNA expression. These results correspond with the opposing effects of ATP and adenosine on the mobilization process. Therefore, based on the results cited above, small-molecule activators of HO-1 could be employed to control sterile inflammation in the BM microenvironment.

By contrast, inhibition of adenosine generation in the BM extracellular space, for example, by employing inhibitors of the ectonucleotidase CD73 or downregulation of HO-1 expression in the BM microenvironment by employing small-molecule HO-1 inhibitors, should promote the onset of sterile inflammation. This would be beneficial for facilitating egress of hematopoietic cells into the circulation to harvest more HSPCs for transplantation (17, 72). In support of this possibility, our recent study showed that CD73-KO mice, which do not convert AMP to adenosine in the extracellular space, are in fact easy mobilizers of HSPCs (17). Other potential targets for facilitating this process are inhibitors of other ectonucleotidases, such as CD39, or even a P2X7 receptor mimetic.

#### CONCLUSION

In this review, we presented the concept that sterile inflammation in the BM microenvironment is involved in the egress of hematopoietic cells, including HSPCs, into the circulation (6, 17). We also presented a novel link between activation of purinergic signaling and the release of EXNs, mainly ATP, which is a crucial activator of the MBL pathway of ComC activation. We are aware that purinergic signaling and EXNs play pleiotropic roles in modulating the activity of the innate and acquired immune systems, but our recent results highlight the importance of ATP

#### REFERENCES


as a DAMP molecule in triggering the mobilization process. A similar mechanism regulating the egress of cells from BM into PB is probably also involved in the egress of cells into lymphatics (3, 6). Besides HSPCs, the interplay between purinergic signaling and innate immunity also plays a role in mobilization of lymphocytic progenitors and other types of stem cells, including MSCs, EPCs, and VSELs, and this is currently being investigated in our laboratory.

Functional P1 and P2 purinergic receptors are expressed on the surface of several types of non-hematopoietic cells in the BM microenvironment (64, 79), including cells forming stem cell niches, such as perivascular SDF-1<sup>+</sup> and KL<sup>+</sup> mesenchymal stromal cells and endothelial cells as well as cells in quiescent nestinbright NG2+ arteriolar and proliferative nestindimLepr+ sinusoidal niches. This wide distribution of these receptors opens up a new area of investigation to better understand the complexity of stem cell mobilization and to design optimal mobilization strategies. Purinergic receptors are also expressed by osteoblasts lining trabecular bones as well as osteoclasts (64). Finally, ATP may also be involved as a neurotransmitter, in addition to catecholamine, in neural fibers that innervate BM tissue in modulating β-adrenergic-mediated egress of HSPCs from BM niches (12, 34).

On the other hand, it is important to better understand the role of sterile inflammation in the aging of hematopoietic cells and its potential involvement in myelodysplasia and GvHD. Shedding more light on these phenomena will also allow us to develop more efficient treatment strategies. Purinergic signaling in both steady-state hematopoiesis and pathology has become an exciting field of investigation.

# AUTHOR CONTRIBUTIONS

This manuscript was written by MR in consultation with the rest of the authors. All authors approved the manuscript.

#### FUNDING

This work was supported by NIH grants 2R01 DK074720 and R01HL112788, the Stella and Henry Endowment, and the OPUS grant DEC-2016/23/B/NZ3/03157 to MR. We would like to thank Dr. Tomasz Jadczyk for his help to prepare final versions of the figures.

ROS signaling and SDF-1 release. *Blood* (2012) 119(11):2478–88. doi:10.1182/ blood-2011-06-358614


as a silent killer of stem cells and a culprit in aging. *Stem Cell Rev* (2017) 13(4):443–53. doi:10.1007/s12015-017-9728-2


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Ratajczak, Adamiak, Kucia, Tse, Ratajczak and Wiktor-Jedrzejczak. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Asparaginyl Endopeptidase (Legumain) Supports Human Th1 Induction via Cathepsin L-Mediated Intracellular C3 Activation

#### Simon Freeley 1†, John Cardone1†, Sira C. Günther 1,2, Erin E. West <sup>3</sup> , Thomas Reinheckel <sup>4</sup> , Colin Watts <sup>5</sup> , Claudia Kemper 1,3,6 \* ‡ and Martin V. Kolev <sup>3</sup> \* †‡

<sup>1</sup> School of Immunology and Microbial Sciences, King's College London, London, United Kingdom, <sup>2</sup> Institut für Medizinische Virologie, University of Zurich, Zurich, Switzerland, <sup>3</sup> Laboratory of Molecular Immunology and Immunology Center, National Heart, Lung and Blood Institute, Bethesda, MD, United States, <sup>4</sup> Faculty of Medicine, Institute of Molecular Medicine and Cell Research, Albert-Ludwigs University Freiburg, and German Cancer Consortium (DKTK), Freiburg, Germany, <sup>5</sup> Division of Cell Signaling & Immunology, School of Life Sciences, University of Dundee, Dundee, United Kingdom, <sup>6</sup> Institute for Systemic Inflammation Research, University of Lübeck, Lübeck, Germany

Autocrine activation of the complement receptors C3aR and CD46 by complement activation components C3a and C3b produced through C3 cleavage by the protease cathepsin L (CTSL) during T cell stimulation is a requirement for IFN-γ production and Th1 induction in human CD4<sup>+</sup> T cells. Thus, lack of autocrine CD46 activation, such as in CD46-deficient patients, is associated with defective Th1 responses and recurrent infections. We have identified LGMN [the gene coding for legumain, also known as asparaginyl endopeptidase (AEP)] as one of the key genes induced by CD46 co-stimulation during human CD4<sup>+</sup> T cell activation. AEP processes and activates a range of proteins, among those α1-thymosin and CTSL, which both drive intrinsically Th1 activity—but has so far not been described to be functionally active in human T cells. Here we found that pharmacological inhibition of AEP during activation of human CD4<sup>+</sup> T cells reduced CTSL activation and the CTSL-mediated generation of intracellular C3a. This translated into a specific reduction of IFN-γ production without affecting cell proliferation or survival. In line with these findings, CD4<sup>+</sup> T cells isolated from Lgmn−/<sup>−</sup> mice also displayed a specific defect in IFN-γ secretion and Th1 induction. Furthermore, we did not observe a role for AEP-driven autocrine α1-thymosin activation in T cell-derived IFN-γ production. These data suggest that AEP is an "upstream" activator of the CTSL-C3-IFN-γ axis in human CD4<sup>+</sup> T cells and hence an important supporter of human Th1 induction.

Keywords: complement, CD46, T cell, cathepsin L, AEP, legumain

#### INTRODUCTION

Liver-derived, serum-circulating complement is a critical part of innate immunity mediating protection against invading pathogens. Proteolytic activation of the complement key components C3 into C3a and C3b and C5 into C5a and C5b upon pathogen sensing in blood leads to opsonization and removal of invading microbes, mobilization of innate immune cells, and

#### Edited by:

Tom E. Mollnes, University of Oslo, Norway

#### Reviewed by:

Marina Noris, Istituto Di Ricerche Farmacologiche Mario Negri, Italy Peter F. Zipfel, Leibniz-Institut für Naturstoff-Forschung und Infektionsbiologie, Hans Knöll Institut, Germany

#### \*Correspondence:

Claudia Kemper claudia.kemper@nih.gov Martin V. Kolev martin.v.kolev@gsk.com

#### †Present Address:

Simon Freeley and Martin Kolev, GlaxoSmithKline (GSK), Stevenage, United Kingdom John Cardone, Oxford Immunotec Ltd., Oxford, United Kingdom ‡ These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 06 May 2018 Accepted: 04 October 2018 Published: 24 October 2018

#### Citation:

Freeley S, Cardone J, Günther SC, West EE, Reinheckel T, Watts C, Kemper C and Kolev MV (2018) Asparaginyl Endopeptidase (Legumain) Supports Human Th1 Induction via Cathepsin L-Mediated Intracellular C3 Activation. Front. Immunol. 9:2449. doi: 10.3389/fimmu.2018.02449

**78**

induction of the general inflammatory response (1). However, it is now also acknowledged that complement serves an equally central role in the direct regulation of human CD4<sup>+</sup> T cell responses: signals delivered by the human-specific C3b-binding complement regulator/receptor CD46 (membrane cofactor protein, MCP) and the C3a receptor (C3aR) are essential for the induction of IFN-γ in human CD4<sup>+</sup> T cells (2, 3). Furthermore, together with IL-2, CD46 also mediates IL-10 co-expression in expanded Th1 cells and, via this, the switch toward a (self) regulatory contraction phase (4) (**Figure S1A**). Unexpectedly, CD46 and C3aR-mediated activation of T cells is independent of liver-derived C3, but driven in an autocrine fashion by the C3 activation fragments C3b and C3a generated through cathepsin L (CTSL)-mediated cleavage of intracellular C3 that shunts to the cell surface upon T cell receptor (TCR) activation (4, 5). Thus, lack of autocrine CD46 activation, such as in CD46-deficient patients, results in reduced Th1 responses and recurrent infections (2), whilst uncontrolled autocrine C3 activation and dysregulated CD46 engagement contributes to hyperactive Th1 responses in autoimmunity (4–6).

Importantly, as increased intracellular CTSL-driven C3 activation can be pharmacologically targeted by a cell-permeable CTSL inhibitor leading to normalization of hyper-Th1 activity at least in vitro (5), we aimed at better understanding the modes of CTSL activation in T cells. When analyzing gene arrays derived from resting or TCR andCD46 activated human CD4<sup>+</sup> T cells (7), we noted that asparaginyl endopeptidase (AEP or legumain) was strongly expressed in T cells and further augmented upon CD46 co-stimulation. AEP is an asparagine-specific cysteine protease found in lysosomes and plays an important but non-exclusive role in the first step of invariant chain of major histocompatibility class II (MHC II) processing in antigen presenting cells (APC) (8). AEP also processes and activates a range of additional proteins. Among those are α1-thymosin and CTSL, which both drive intrinsically Th1 activity (5, 9), and AEP-deficient mice accordingly exhibit a defect in the maturation of catepsins B, H, and L in kidney cells (10). However, so far, AEP activity has not been described in human T cells.

Here we describe for the first time a role for AEP in human CD4<sup>+</sup> T cells and its specific requirement for normal Th1 induction.

#### MATERIALS AND METHODS

#### Healthy Donors

Blood samples were obtained with ethical approvals at King's College London (Wandsworth Research Ethics Committee, REC# 09/H0803/154). CD4<sup>+</sup> T cells were purified from buffy coats (NHSBT, Tooting, UK) or blood samples from healthy volunteers after informed consent.

#### Mice

Wild type and Lgmn−/−mice were generated by Drs. Thomas Reinheckel and Colin Watts as previously described (11). All animals were maintained in accredited BSL2 facilities at KCL and experiments performed in compliance with animal study proposals approved by KCL.

### Antibodies, Proteins, and Inhibitors

For a list of antibodies, proteins and inhibitors utilized, see **Supplementary Materials**.

# T Cell Isolation and Activation

For a protocol of human and mouse CD4<sup>+</sup> T cell isolation and activation, see **Supplementary Materials**.

# Cytokine Measurements

Cytokine secretion by human and mouse T cells was measured using the human and mouse Th1/Th2/Th17 Cytokine Bead Arrays (BD Biosciences, Oxford, UK) or the combined Secretion Assays for human IFN-γ and IL-10 purchased from Miltenyi Biotec (Bergisch Gladbach, Germany) according to the manufacturers' provided protocols.

# Confocal Microscopy

For a detailed protocol on the confocal microscopy performed here including the measurement of the Pearson's Correlation Coefficient, refer to **Supplementary Materials**.

#### Western Blotting

For a detailed protocol on the Western botting procedures and analyses utilized, see **Supplementary Materials**.

#### Statistical Analyses

Statistical analyses were performed on GraphPad Prism 7 (La Jolla, CA). Data are presented as mean ± SEM and compared using paired t-tests or one-way ANOVA with a Tukey multiple comparison post-hoc test, as appropriate. p < 0.05 denoted statistical significance throughout.

# RESULTS

#### AEP Is Required for Normal Th1 Induction in Human and Mouse CD4<sup>+</sup> T Cells

Gene expression analyses performed on resting and CD3+CD46 activated human CD4<sup>+</sup> T cells suggested the expression modulation of the LGMN gene, encoding the endopeptidase AEP (7). Indeed, resting CD4<sup>+</sup> T cells contained high levels of AEP protein in the cytoplasm and CD46-mediated co-stimulation during TCR activation further increased AEP protein levels but simultaneously induced the nuclear translocation of a proportion of AEP (**Figures 1A,B**). CD3+CD46-activation of T cells is a strong and specific inducer of human Th1 responses (2). The addition of increasing doses of a specific AEP inhibitor (12) during CD3+CD46 activation significantly reduced the percentage of actively IFN-γ-secreting cells as well as their switching into the IL-10-producing contracting phase in cultures in a dose-dependent manner (**Figure 1C** and **Figure S1B**). The observed reduction of IFN-γ and IL-10 secretion also in CD3 and CD3+CD28-activated T cells upon AEP inhibition was expected, as TCR stimulation and CD28-costimulation function upstream of CD46 and trigger increased intracellular CTSL-mediated C3b generation and "background" CD46 engagement (5). Of note, neither cell proliferation, viability nor production of Th2 cytokines such as IL-4 were affected by AEP inhibition and Th17

FIGURE 1 | Shown are one representative FACS and two Western blot experiments of n = 3 using a different donor each time. (C) AEP inhibition suppresses human Th1 induction. T cells were activated as described under "A" with or without 25 or 50µM of a specific AEP inhibitor and IFN-γ and IL-10 (co)secretion measured 36 h post activation. (Ci) shows FACS data derived from a representative donor whilst (Cii) summarizes the analyses for the shown activation conditions of n = 6 donors. (D) AEP inhibition does not affect cell proliferation. Cell trace violet-labeled CD4<sup>+</sup> T cells were CD3+CD46-activated in the presence or absence of 50µM AEP inhibitor and cell proliferation measured at 6 d post activation. (Di) Shows a representative FACS profile and (Dii) the accompanying statistical analysis from four different experiments (n = 4). (E) AEP is also required for normal Th1 induction in mice. Naïve CD4<sup>+</sup> T cells isolated from wild type (WT) or AEP-deficient (Lgmn−/−) mice (n = 5) were activated for 6 days under Th1, Th2, or Th17 skewing conditions and the total numbers of IFN-γ (Th1), IL-4 (Th2), or IL-17-positive (Th17) cells assessed by intracellular cytokine staining. The number of FoxP3-positive natural regulatory T cells (nTregs) was assessed in activated cell cultures without addition of skewing cytokines/antibodies. Error bar graphs represent mean ± SEM. \*P < 0.05, \*\*P < 0.01, \*\*\*P < 0.001; ns, not significant.

responses were only reduced significantly under the CD3+CD46 stimulation condition (**Figure 1D** and **Figures S1B,C**).

We confirmed a role for AEP in the production and secretion of IFN-γ by CD4<sup>+</sup> T cells also in mice. Importantly, mice do not express CD46 on somatic tissue (13) and a functional homologue that drives Th1 induction—as shown for CD46 in human T cells—has so far not been identified in mouse T cells (14). Thus, purified CD4<sup>+</sup> T cells from either wild type (WT) or AEP-deficient (Lgmn−/−) were activated under Th1, Th2, or Th17-skewing conditions and the percentage of IFN-γ, IL-4, and IL-17-positive cells, as well as secretion of these cytokines into culture media assessed. AEP deficiency was accompanied by a specific and significant reduction in Th1 induction (**Figure 1E**). We also observed an increase in the circulating natural FoxP3<sup>+</sup> regulatory T cell (Treg) pool in Lgmn−/<sup>−</sup> animals (**Figure 1E** and **Figure S1D**).

These data indicate that AEP activity is required for normal Th1 induction in mice and men.

#### AEP Is Required for CTSL-Mediated C3 Activation in Human CD4<sup>+</sup> T Cells

In human CD4<sup>+</sup> T cells, CTSL is key in activating C3 and thereby providing the ligands for autocrine CD46 and C3aR activation needed for Th1 induction (5) (**Figure S2A**). CTSL is usually synthesized as inactive pre-proenzyme and requires proteolytic activation (first removal of the pre-peptide and then processing into active single or double-chain forms) by other proteases (**Figure S2B**) (15). AEP has been shown to be involved in the processing of single-chain CTSL into the double-chain form in mice (10, 16). Inhibition of AEP slightly in resting cells but significantly during CD3+CD46 activation reduced the generation of active single and double chain CTSL forms as assessed by Western blotting and confocal microscopy in human T cells (**Figures 2A,B**). Importantly, the decrease in the generation of the CTSL active forms by AEP inhibition led also to a proportional and significant reduction in the activation of C3 as measured by the appearance of the C3a neo-epitope in CD3+CD46-stimulated T cells (**Figures 2B,C** and **Figure S2C**). A recent study found that AEP cleaves FoxP3 in mouse induced regulatory T cells and confirmed increased Tregs in Lgmn−/<sup>−</sup> mice (17). We did, however, not observe changes in Foxp3 levels in resting or activated human CD4<sup>+</sup> T cells upon AEP inhibition (**Figure 2D**).

AEP also processes pro-thymosin α into the active α1 thymosin form (18) and α1-thymosin has been implicated in human CD4<sup>+</sup> T cell activation and modulation of their cytokine production (19). However, the addition of recombinant α1-thymosin during T cell activation in the presence of the AEP inhibitor failed to rescue defective Th1 induction (**Figure 2E**). Human CD4<sup>+</sup> T cells upregulate major histocompatibility class II (MHC II) molecules (and their maturation require CTSL cleavage) upon activation (20)—however, this process also remained unaffected by AEP inhibition (**Figure S2D**).

Together these data suggest that AEP activity in human CD4<sup>+</sup> T cells drives Th1 induction via proteolytic activation of CTSL, which in turn increases C3 activation and autocrine CD46 and C3aR activation.

#### DISCUSSION

CTSL-mediated activation of intracellular C3 into C3a and C3b which then drive the autocrine engagement of the C3aR and CD46, respectively—is critical to normal Th1 cell responses. CD46 deficiency is accompanied hence by recurrent infections throughout life and patients with serum C3 deficiency have recurrent infections early in life. Of note, whilst C3 deficiency was long thought to be based solely on reduced pathogen lytic activity and impairment of innate immunity, it is now understood that C3-deficiency indeed also impairs adaptive immunity (21). Increased CTSL-mediated C3 activation, on the other hand, drives Th1 hyper-activity in rheumatoid arthritis and in systemic lupus erythematosus (5, 22). Pharmacological inhibition of intracellular CTSL activity can normalize dysregulated Th1 responses, demonstrating that this pathway is therapeutically amendable. Thus, understanding how CTSL activity is regulated adds to our better understanding of Th1 biology and may also provide additional needed therapeutic targets to control aberrant Th1 immunity.

Here we identified AEP as an enzyme that processes CTSL into its active forms in human CD4<sup>+</sup> T cells. Inhibition of AEP activity reduced the appearance of the CTSL active forms, the generation of C3a in T cells and the induction of IFN-γ-secreting cells by about 50%. This strongly suggests that AEP functions upstream of the CTSL-C3 axis and is an integral part of human Th1 initiation. AEP processes only the CTSL single chain into the double chain form in mouse kidney cells (10). It is currently unclear whether CTSL single chain-generating activity of AEP is human (T cell) specific and/or if there is a difference in the CTSL single vs. double chain forms' ability or potency to cleave C3. The

active single or double chains generated from cleavage of the proenzyme form was measured by (Ai) Western blotting with (Aii) appropriate densitometric analyses of (Continued)

FIGURE 2 | the percentage of active (<30 kDa size bands) vs. non-active (>30 kDa size bands) CTSL forms. (B) confocal microscopy (left panels). The generation of CTSL-dependent intracellular C3a was measured by (B) confocal microscopy (right panels) and (Ci) FACS with subsequent (Cii) statistical analysis. Data shown in "A–C" are derived from n = 4 healthy donors. (D) T cells were activated under the depicted conditions and Foxp3 protein expression measured at 36 h post activation via FACS analysis. Shown is one representative experiment of two similarly performed. (E) Autocrine AEP-mediated α1-thymosin activation is not required for Th1 induction. T cells were activated as under "A" with the addition of 1µM α1-thymosin as indicated and (Di) IFN-γ and IL-10-producing cells enumerated and (Dii) IFN-γ secretion measured 36 h post activation (n = 3). Error bar graphs represent mean ± SEM. \*P < 0.05, \*\*P < 0.01; ns, not significant.

latter is feasible since single chain and double chain CTSL can indeed have distinct substrate specificities (23).

This is the first description of a role for AEP in human Th1 biology. A previous study identified a role for AEP, together with the protease inhibitor SerpinB1, in the regulation of CTSL in mouse CD4<sup>+</sup> T cells (24). Interestingly though, AEP blockage with an inhibitor (the authors did not utilize Lgmn−/<sup>−</sup> animals) in presence of Serpinb1 deficiency negatively controlled CTSL-mediated Th17 induction without affecting Th1 responses. The difference in outcome may be due to additional, not yet defined, functions of SerpinB in Th17 induction. Here, using Lgmn−/<sup>−</sup> animals, we found that, similar to human CD4<sup>+</sup> T cells, mouse Th1 responses are reduced while Th17 responses remain unaffected. Of note, although mouse Th1 induction also requires intracellular C3 activation, CD4<sup>+</sup> T cells from Ctsl−/<sup>−</sup> mice produce normal IFN-γ levels (5), indicating that C3 activation in mice can occur in a CTSL-independent fashion. In line with this, C3a levels are unchanged in Lgmn−/<sup>−</sup> animals (**Figure S2E**). Given these differences, it seems unlikely that AEP regulates mouse Th1 induction via a similar CTSL-C3 axis.

Previous work also noted increased natural and induced Tregs in Lgmn−/<sup>−</sup> mice in line with the observation that AEP stabilizes regulatory T cells through nuclear Foxp3 cleavage (17, 25). Although CD46-activation indeed induced the nuclear translocation of both AEP and CTSL (**Figure 1B** and **Figure S2F**) we did not observe an effect of AEP inhibition on Foxp3 or other confirmed AEP/CTSL targets such as α1-thymosin or MHC II—in human CD4<sup>+</sup> T cells. It is possible that the AEP inhibitor used here only affects certain AEP activities. Patients with genetic AEP deficiency that could be of help to better define AEP's roles in humans have so far not been described and we would therefore not exclude that AEP (and CTSL) serves

#### REFERENCES


additional functions aside from C3 activation in supporting Th1 activation.

In sum, AEP activity is important to normal Th1 induction in mice and men—the exact modes of action of AEP in these cells, however, seem to be species-specific and require further investigation.

#### AUTHOR CONTRIBUTIONS

TR, CW, MVK, and CK conceived the study, helped designing key experiments, discussed the data, and drafted the manuscript. SF, JC, EEW, SCG, and MVK designed and performed experiments, analyzed and interpreted the data, and finalized the manuscript. CK and MVK share the last authorship. All authors approved the final version of the manuscript and agreed to be accountable to all aspects of this work.

#### FUNDING

This work was financed by the MRC Centre grant MR/J006742/1, a Wellcome Trust Investigator Award (CK), the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King's College London, and by the Division of Intramural Research, National Heart, Lung, and Blood Institute, NIH.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02449/full#supplementary-material


hepatitis B virus e antigen-positive chronic hepatitis B. J Int Med Res. (2010) 38:2053–62. doi: 10.1177/147323001003800620


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Freeley, Cardone, Günther, West, Reinheckel, Watts, Kemper and Kolev. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Interpretation of Serological Complement Biomarkers in Disease

Kristina N. Ekdahl 1,2 \*, Barbro Persson<sup>1</sup> , Camilla Mohlin<sup>2</sup> , Kerstin Sandholm<sup>2</sup> , Lillemor Skattum<sup>3</sup> and Bo Nilsson<sup>1</sup>

<sup>1</sup> Rudbeck Laboratory C5:3, Department of Immunology, Genetics and Pathology, Uppsala University, Uppsala, Sweden, <sup>2</sup> Centre of Biomaterials Chemistry, Linnaeus University, Kalmar, Sweden, <sup>3</sup> Section of Microbiology, Immunology and Glycobiology, Department of Laboratory Medicine, Clinical Immunology and Transfusion Medicine, Lund University, Lund, Sweden

Complement system aberrations have been identified as pathophysiological mechanisms in a number of diseases and pathological conditions either directly or indirectly. Examples of such conditions include infections, inflammation, autoimmune disease, as well as allogeneic and xenogenic transplantation. Both prospective and retrospective studies have demonstrated significant complement-related differences between patient groups and controls. However, due to the low degree of specificity and sensitivity of some of the assays used, it is not always possible to make predictions regarding the complement status of individual patients. Today, there are three main indications for determination of a patient's complement status: (1) complement deficiencies (acquired or inherited); (2) disorders with aberrant complement activation; and (3) C1 inhibitor deficiencies (acquired or inherited). An additional indication is to monitor patients on complement-regulating drugs, an indication which may be expected to increase in the near future since there is now a number of such drugs either under development, already in clinical trials or in clinical use. Available techniques to study complement include quantification of: (1) individual components; (2) activation products, (3) function, and (4) autoantibodies to complement proteins. In this review, we summarize the appropriate indications, techniques, and interpretations of basic serological complement analyses, exemplified by a number of clinical disorders.

Keywords: complement, deficiency, activation products, functional test, complement regulatory drugs

# THE COMPLEMENT SYSTEM

#### Activation of Complement

The complement system comprises approximately 50 proteins that are found in the fluid phase of the blood or bound to cells where they function as receptors or regulators of complement activation (**Figure 1A**). The system is organized in three activation pathways: the lectin pathway (LP), the classical pathway (CP), and the alternative pathway (AP), each with different recognition molecules. Complement activation leads to the formation of two proteolytic enzyme complexes, convertases which have C3, the central and most abundant complement component as their common substrate.

The LP is activated when one of several recognition molecules, mannan-binding lectin (MBL), collectins, or ficolins bind to carbohydrates, e.g., on a pathogen surface, and often on polymers. The CP is activated when C1q binds to IgM or IgG, which may be in the form of immune complexes or bound in an altered conformation to artificial surfaces, such as in medical devices. In addition, the CP can also become activated by pentraxins, e.g., C-reactive protein (CRP), and a variety

#### Edited by:

Maciej Cedzynski, Institute for Medical Biology (PAN), Poland

#### Reviewed by:

Marco Cicardi, Università degli Studi di Milano, Italy Peter F. Zipfel, Leibniz-Institut für Naturstoff-Forschung und Infektionsbiologie, Hans Knöll Institut, Germany

#### \*Correspondence:

Kristina N. Ekdahl Kristina.Nilsson\_Ekdahl@igp.uu.se

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 06 July 2018 Accepted: 10 September 2018 Published: 24 October 2018

#### Citation:

Ekdahl KN, Persson B, Mohlin C, Sandholm K, Skattum L and Nilsson B (2018) Interpretation of Serological Complement Biomarkers in Disease. Front. Immunol. 9:2237. doi: 10.3389/fimmu.2018.02237

**85**

of negatively charged molecules that includes DNA, LPS and heparin. This activation leads to activation of fluid phase proteolytic enzymes: mannan associated serine proteases (MASP)−1 and MASP-2 within the LP and C1r and C1s within the CP. These proteases mediate formation of the CP/LP C3 convertase C4bC2a.

Activation of the AP is accomplished by conformationally altered C3, either as a result of tick-over to C3b or C3(H2O) or by its binding to surfaces, but AP activation can also be facilitated through the binding of properdin to damageassociated molecular patterns (DAMPs) on pathogens (1).

Through these processes of activation, the formation of the AP C3 convertase, C3bBbP, is induced. The labile C3 convertases cleave C3 into the anaphylatoxin C3a and the larger C3b fragment. In the presence of an acceptor surface, e.g., a pathogen or antigen-antibody complex, C3b can form a covalent bond to amino acid or sugar residues. Then C3b can be cleaved in three steps by the plasma protease factor I, in cooperation with one of several co-factors. The two first cleavages generate iC3b, which promotes phagocytosis via interaction with different complement receptors (CR)-1 (CD35), CR3 (CD11b/CD18), CR4 (CD11c/CD18), and/or CRIg (complement receptor of the immunoglobulin family). The third factor I mediated cleavage separates the molecule into the target bound C3d,g fragment which is a ligand for CR2 (CD21), and C3c which is released from the activating surface. The same digestion also takes place in the fluid phase indicating that complement activation in vivo or in vitro can be monitored by measuring C3d,g, iC3b or C3a.

In addition to being a trigger of complement activation the AP also provides a potent amplification loop. Since each deposited C3b residue (regardless of the nature of the initial activation trigger) is the potential nucleus of a novel C3bBb C3 convertase, it has the potential capacity to activate numerous other C3 molecules. Deposition of additional C3b molecules to or in the vicinity of either of the C3 convertases alters their enzymatic specificity from C3 to C5. Cleavage of C5 yields the anaphylatoxin C5a and initiates the generation of the terminal pathway (TP) where the end product is the terminal complement complex, C5b-9, which may remain in the plasma as soluble C5b-9 (sC5b-9) or be inserted in the cell membrane as membrane attack complex (MAC). MAC may induce cell lysis (primarily in nonnucleated cells) and gram-negative bacteria or inflammation and upregulation of tissue factor, e.g., on endothelial cells, at sub-lytic concentrations (2, 3).

The anaphylatoxins C3a and C5a bind to their receptors C3aR and C5aRs, expressed on phagocytes: polymorphonuclear cells (PMNs), and monocytes, thereby attracting and activating them, thus further fuelling the inflammation.

#### Regulation of Complement

A number of regulators protect surfaces of autologous cells against complement attack. These regulators include (but are not restricted to) cell-bound molecules, such as CR1, decay acceleration factor (DAF; CD55), and membrane cofactor protein (MCP; CD46), all of which inactivate the C3 convertases in different ways. Additional regulators, C4b-binding protein (C4BP, which regulates the CP/LP convertase) and factor H (the main regulator of the AP), found in the plasma are recruited via glycoseaminoglycans and/or deposited C3 fragments to the cell surface, thus providing further down-regulation of complement.

Regulation at the level of the TP is accomplished by cell bound CD59, and clusterin and vitronectin in the fluid phase, which all inhibit MAC formation and its insertion into the membrane of autologous cells. Furthermore, C1 inhibitor (C1-INH) inhibits the proteases generated within the CP and LP; C1r/C1s and MASP-1/MASP-2, respectively, (**Figure 1A**). However, C1-INH is not specific for complement system-associated serine proteases but also inhibits proteases generated by the activation of the contact system like Factor (F)XIIa, FXIa, and kallikrein.

#### Pathology of Complement

The pathogenesis of many inflammatory diseases includes different complement deficiencies as well as excessive complement activation. Complement is engaged in a number of diseases exemplified in **Figure 2**. The pathologic effect may be caused either by an increased and persistent activation or an altered expression or function of various complement inhibitors resulting in defective control. Systemic lupus erythematosus (SLE), myasthenia gravis and other autoimmune disorders are examples of the former, where the presence of soluble or solid-phase antibody-antigen complexes induce excessive complement activation. C3 glomerulopathy (C3G), paroxysmal nocturnal haemoglobinuria (PNH), and atypical haemolytic uremic syndrome (aHUS), are diseases which are associated with insufficient complement inhibition/regulation, e.g., as discussed in (4, 5) and quoted in the references.

In many cases the complement activation is a part of reactions resulting from activation of all cascade systems of blood, and under conditions such as ischemia/reperfusion injury (IRI), there is a combination of excessive activation and insufficient control. IRI can occur under many pathological conditions but also during medical treatments. Cardiac infarction and stroke are associated with ischemia followed by reperfusion of an organ or blood vessel. Ischemia which is often complicated with IRI can also occur after transplantation (both allogeneic and xenogeneic) as well as during cardiovascular surgery facilitated by cardiopulmonary bypass. In IRI, excessive complement activation in combination with insufficient complement regulation play important roles and the resulting damage appears to be associated with all three pathways of complement. This complement activation leads to an inflammatory response which consists of generation of anaphylatoxins and other mediators which collectively induce activation of endothelial cells and phagocytes resulting in recrutiment and extravasation of PMNs, as described in (6) and cited references.

Despite the clear involvement of complement in a large number of conditions, there are only a limited number of diseases where serological complement biomarkers have been established as differential markers of disease. In the majority of other conditions, the biomarkers are able to distinguish between patients and normal individuals at group level. However, these markers can often be used to follow individual patients if the baseline values are known or can be anticipated e.g. as in trauma and shock (7), sepsis (8), in neurological diseases such

as myasthenia gravis (9), in ophthalmic diseases such as agerelated macular degeneration, (AMD) (10), bullous pemphigoid (11), antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis (12) etc. They can also be used for research purposes.

This review will describe the indications and specific methods that are used to determine the complement status of a patient and how the results of these assays are interpreted.

# ANALYTICAL METHODS

#### Activation in vivo vs. Activation in vitro

Complement system activation via different pathways in blood plasma is a feature of a large number of diseases. For example, in immune complex diseases, the CP and the TP components are mainly activated while in renal diseases the AP and the TP components are predominantly engaged. When a component is activated in vivo either by proteolytic cleavage and/or by induced conformational changes triggered by protein-protein interactions, the component is taken up by receptors of e.g., leukocytes and Kupffer cells. This results in consumption of complement components. If a whole pathway (CP+TP or AP+TP) is activated, all components are consumed and the function is reduced along this pathway and systemic activation products will be moderately increased. Poor function via either the LP/CP or the AP will also affect the other pathway if the activation is strong enough since the components of the common TP will be consumed. If on the other hand the pathway (CP+TP, LP+TP or AP+TP) is activated in vitro all components are inactivated along this path and the function of this activation pathway is also reduced, but unlike the in vivo situation the activated proteins remain in the sample and are not consumed and the activation products will stay at high levels in the tube. If EDTA-plasma is prepared, any further activation of complement is stopped, since EDTA chelates Ca2<sup>+</sup> and Mg2<sup>+</sup> and thereby blocks the function of the C1 complex and the two C3 convertases, respectively. If, by contrast, serum is prepared, further activation of the sample in vitro is possible and it can be used for functional testing, e.g., using haemolytic assays (**Figure 3**) (13).

# Preanalytical Factors

With this in mind, EDTA-plasma is suitable for analyses of individual components and for activation products, while serum is used for analysis of complement function. Serum can be replaced by plasma anticoagulated with the specific thrombin inhibitor lepirudin (i.e., recombinant hirudin) or any other thrombin or FXa inhibitor, which does not affect complement function (14). In order to maintain the function of the complement components and avoid further activation of individual components the samples must be kept cold until they

are frozen at −80◦C, which should be done preferably within 120 but no longer than 240 min. It is important not to freeze the samples at −20◦C, not even temporarily, since this creates a slow freezing rate and further activation/inactivation of individual components. During transportation dry ice must be used and the samples should be transferred directly from the to 80◦C freezer to the dry ice (13).

#### Analysis of Complement in Plasma/Serum Quantification of Individual Complement Components

Different types of immunoassays, most commonly immunoprecipitation assays, are used to determine the concentration of individual complement components. Previously, rocket immuno electrophoresis (RIE), radial immunoprecipitation or enzyme immune assays (EIAs) were most common, but have today to a great extent been replaced by nephelometry and turbidimetry. These techniques utilize polyclonal antibodies against the analyte, e.g., C1-INH, C4, C3, or factor B or activation fragments of these proteins (**Figure 1B**). These antibodies are added in excess to the sample and bind to their target, forming antigen-antibody complexes. Detection is performed by passing a light beam though the sample and which will be dispersed or absorbed by the formed immune complexes.

All techniques which use polyclonal antibodies for detection are relatively robust regarding the effect of preanalytical factors such as proteolytic cleavage or denaturation of the target protein induced during suboptimal sample handling. However, it is important to be aware that when polyclonal antibodies which are raised against C3c are being used, these assays will detect all forms of C3 which contain the C3c moiety, i.e, intact, non-activated C3 and its activated proteolytic fragments C3b, iC3b, and C3c. In analogy, anti-C4c antibodies will detect the corresponding forms of C4. Consequently, this type of assay is useful to determine the in vivo concentration of the protein (i.e., to monitor consumption or deficiency) but gives no information of the activation state or conformation of the protein.

More recently, multiplex assays for complement components have been developed and are now commercially available. The advantage of such assays is that they enable the simultaneous determination of several components, thereby saving both time and sample volume. So far, the analytes in the available kits are restricted to components with fairly high plasma concentrations, and to our knowledge, no LP-specific panels are yet on the market.

#### Quantification of Activation Products

The sequential proteolytic cleavage which occurs during complement activation generates activation products with different properties than those of the non-activated zymogen molecules (**Figure 1B**). In general, two principles are used in assays designed to determine the degree of complement activation: one is to use monoclonal antibodies (mAbs) which detect amino acid sequences

that are inaccessible in the native zymogen molecule but become exposed when the protein is activated (i.e., neoepitopes). Most commercially available assays for C3a, C3b/iC3b/C3c, C4a, C4b, C4d, Ba, Bb, C5a, and sC5b-9 are based on neo-epitope mAbs. The other option is to use polyclonal antibodies but this often requires fractionation of zymogen molecules and activation products according to size. One example is C3d,g which is detected by EIA or nephelometry/turbidimetry, but since the polyclonal antibodies (in this case raised against C3d,g) also recognize intact C3, C3b, and iC3b in addition to C3d,g, these larger molecules must be removed by precipitation before analysis (15).

In vivo there is a continuous physiological turnover of C3 which leads to generation of activation fragments including C3a and C3d,g. Consequently, in order to monitor ongoing complement activation it is mandatory to determine a ratio of C3a or C3d,g level to the total level of C3 (C3a or C3d,g/C3), e.g., during an exacerbation in SLE (15). Furthermore, in obese individuals, the levels of a number of complement components, including C3, are greatly increased, resulting in corresponding higher levels of C3a; this problem further underscores the importance of calculating a ratio as a measure of relative activation (16).

Formation of the lytic C5b-9 (MAC) complex is the last step of the complement cascade which causes cell damage or lysis as a result of its insertion into the cell membrane, or endothelial cell activation at sub-lytic concentrations (2, 3). Complement activation of the TP can be monitored by quantification of sC5b-9 in the fluid phase, with an EIA which uses a mAb specific for a neo-epitope in C9 for capture. The epitope for this mAb is exposed in conformationally changed complex-bound C9 but not in intact C9. After capture, the sC5b-9 complexes can be detected by using polyclonal antibodies against another protein present in the same macromolecular complex, e.g., C5 or C6 (17).

Most, if not all, complement activation markers can rapidly be produced by complement activation in vitro. Consequently, it is of utmost importance that samples intended for detection of complement activation are collected and handled properly. In particular, there is a great risk that C3a will be generated in vitro if samples are improperly handled (14). It should also be taken into account that different C3 activation products vary greatly with regard to their in vivo half-life: approximately 0.5 h for C3a (18) and 4 h for C3d,g (19). Since C3d,g is a more robust marker, it is more suitable for diagnostic use while the generation of C3a is the more common analysis in experimental settings (20).

#### Quantification of Complement Function

In order to maintain full function in an individual complement activation pathway it is necessary for each of the participating proteins to be active, i.e., a deficiency in one individual protein will stop the activity of the entire cascade. Functional tests, in particular different haemolytic assays that monitor a whole activation pathway from the recognition phase to MAC-formation (=lysis) can be used to detect both deficiencies in individual component as well as depression in complement function caused by consumption of intact complement components.

Activation of the CP of complement is monitored in haemolytic assays employing sheep erythrocytes coated with rabbit antibodies, preferably purified IgM (or mixed with IgG) to the Forssman antigen. Patient serum is added, C1q binds to the immunoglobulins which initiates formation of the CP C3 convertase and subsequent activation leads to assembly of the MAC which results in lysis of the erythrocytes, **Figure 4A** (21).

Activation of the AP of complement is monitored in haemolytic assays employing rabbit or guinea pig erythrocytes, which are spontaneous and potent activators of human AP. EGTA which chelates Ca2<sup>+</sup> and thereby inhibits activation via the CP and LP, is added to the patient serum prior to incubation. Under these conditions the AP C3 convertase is formed on the target erythrocytes, leading to C3 activation and subsequent lysis (22).

A variety of haemolytic assays have been developed using different serum dilutions and amounts of erythrocytes. In the original haemolytic assays called CH50 and AH50, a specified limiting amount of erythrocytes are incubated with serum in serial dilution to determine the dilution of serum needed to lyse 50% of the cells during a certain time interval (**Figure 4B**) (21, 22). In samples with low function it is often necessary to repeat the analysis with additional dilution steps.

Unlike in the CH50 and AH50 assays, where incubation of erythrocytes and serum takes place in the fluid phase, an alternative approach is to cast the target erythrocytes in an agarose gel. The patient serum is then added into wells punched in the agarose and diffuses in the gel causing cell lysis. This haemolysis-in-gel technique is quick and very useful to screen for complement deficiencies but does not enable quantification (23).

Instead of using erythrocytes, which may cause problems due to individual variation of the animals that have donated the blood, systems using artificial liposomes have been developed. Assays which are commercially available is performed in a CH50 like way (24).

An alternative to the CH50 and AH50 assays is the considerably less laborious, and much quicker one-tube assay. The sample is here incubated in one tube for 20 min. These assays give similar results as CH50 and AH50 (21, 25) and are based on the fact that the "dose" of complement is proportional to the number of cells lysed and the assay is therefore performed in an excess of erythrocytes (**Figure 4C**).

All haemolytic assays have problems to detect properdin deficiencies. Normally they give intermediate to normal values, never low function as is seen in other deficiencies. Therefore, special arrangements need to be made. One way is to make a kinetic analysis of the sample in the AP haemolytic assay. In the example shown in **Figure 5A** it is seen that the curves for a properdin deficient patient and a healthy control merge at the same level after 20 min in the one-tube assay. Therefore, it is necessary to also incubate for shorter time to detect this deficiency (**Figure 5A**). Alternatively, the concentration of properdin is determined separately.

More recently, a method has been reported that makes use of parallel EIAs to quantify the function of the three activation pathways of complement (26). Target molecules for each pathway are coated on wells of microtitre plates; IgM for the CP, mannan or acetylated bovine serum albumin for the LP, and LPS for the AP. Patient serum is incubated in the wells in the presence of additions which enable specific activation of only one pathway at the time, since the activity of the other pathways are inhibited. The readout for each EIA is formation of C5b-9 which is detected by a mAb specific for a neo-epitope in C9 which is exposed in complex-bound but not in native C9 (17). These assays are commercially available (**Figure 6**).

The techniques described here are valuable to identify complement deficiencies and for the haemolytic assays (except the haemolysis-in-gel) also to monitor levels in complement function, for example in patients with SLE during exacerbations. A suspected deficiency can be confirmed by determination of the protein, using relevant assays as described above. Furthermore, since most plasma complement components are commercially available, it is possible to verify the deficiency by reconstituting the patient sample with the protein in question and then repeat the functional assay, where the activity should be normalized. i.e., by combining these techniques it is possible to distinguish between functional deficiency and lack of a single complement component (13).

#### EXAMPLES OF INDICATIONS FOR COMPLEMENT DIAGNOSTICS AND THE INTERPRETATION OF COMPLEMENT STATUS

The complement status of a patient cannot be determined using only one assay. In order to get a complete status, assays from all four categories of analyses have to be used (**Table 1**). Two major basic indications exist: identification of complement component deficiencies and monitoring of complement activation. In order to screen for complement deficiencies functional assays (haemolytic or EIA) are used. Here, both CP and AP assays are compulsory. For the LP, a commercial EIA to monitor activation via MBL exists, but it only covers MBL, MASP-1 and MASP-2. Also, C9 deficiencies may be missed depending on the erythrocytes used. For monitoring of the degree of complement activation a minimum set up is to use a functional assay triggered via the CP, C3 and an assay for activation products (iC3b, Ba, C3d,g etc.). Examples of add-on assays are the concentrations

FIGURE 4 | Haemolytic assay for detection of complement classical pathway (CP) function. (A) CP haemolytic assay. Sheep erythrocytes coated with IgM antibodies are incubated with patient serum. The C1 complex binds and initiates formation of the CP convertase, leading to activation of C3, assembly of the C5b-9 complex, and subsequent erythrocyte lysis. (B) CH50 assay. Titration of the amount of serum needed to lyse 50% of a specified limited and fixed quantity of cells in the CH50 assay. The curves show three individuals with different levels of complement function. (C) One tube CP assay. Since the activity of complement is proportional to the quantity of cells that are lysed, this assay is performed in an excess of erythrocytes. The curves show three individuals with different levels of complement function. Assays for alternative pathway (AP) activation function similarly except that uncoated rabbit or guinea pig erythrocytes which spontaneously activate the AP, are used.

FIGURE 5 | Examples of specialized functional assays. (A) Time course alternative pathway (AP) haemolysis test. The left curve was obtained using serum form a complement sufficient individual and the right curve using serum form a properdin deficient patient. Since the curves merge at the same level after 20 min in the one tube assay it is necessary to also incubate for shorter time to locate this deficiency. (B) Correlation between functional test and individual analytes. Combination of haemolytic function via the classical pathway (CP) and an individual analyte (C1q) show a correlation in this material of systemic lupus erythematosus (SLE) patients. The exception, marked with (\*) is a patient with total C2 deficiency. (B) reproduced from (15) with permission from the publisher.

of, e.g., C1q, C4 and factor B, and a functional AP assay, e.g., a haemolytic assay (**Figure 5B**).

# Inherited and Acquired Complement Component Deficiency

#### Complement Factor Deficiencies (General)

In general, complement deficiencies are rare, but when diagnosed, they are generally associated with recurrent bacterial infections (this applies to all activation pathways) (**Table 2**).

In addition, individuals deficient in MBL or MASPs of the LP are also susceptible to viral and protozoan infections and deficiencies in CP components are generally associated with an increased incidence of SLE or SLE-like disease. Most susceptible to autoimmune disease are patients with C1q deficiency while individuals with C2 deficiency are less predisposed to this type of diseases. C2 deficiency is the most common CP specific deficiency with a frequency of 1/20,000 (27, 28).

A deficiency is detected by functional assays and gives a very low to non-existing function via one (AP-, CP-, LPspecific deficiency) or all pathways (TP specific). In order to confirm that the functional defect is due to a specific deficiency, cryoglobulinemia has to be ruled out. Cryoglobulins can totally inactivate the CP+TP in serum (or lepuridin anticoagulated plasma) after the sample has been drawn. Also, severe consumption due to complement activation in vivo has to be excluded by measuring complement activation products, e.g., C3d,g. Identification of which component is lacking is established by performing measurements of individual factor concentration, by Western blotting, genetic screening etc. Confirmation of the deficiency (see above) can be done by reconstitution of the deficient serum with the purified identified complement component.

#### Monitoring of Complement Regulatory Drugs

Currently, there are only two complement inhibitors available in the clinic: C1-INH and eculizumab. Purified or recombinant C1- INH inactivates the proteases generated by the CP and LP (C1r, C1s, MASP-1, and MASP-2) as well as FXIIa, FXIa and kallikrein of the contact system (29). Eculizumab is a humanized mAb that binds to C5, preventing its activation to the anaphylatoxin C5a and C5b which initiates C5b-9 formation. Treatment with eculizumab is approved for treatment of aHUS, PNH, and refractory myasthenia gravis, but it is also currently undergoing clinical trials for the prevention of antibody-mediated rejection (AMR) in allogeneic kidney transplantation (30). In addition to these two drugs, a large number of complement-modulatory compounds that act at different control points are under development for various indications. Examples of compounds which are in clinical trials include mAbs against C1s (31), which inhibit the CP, the peptide CP40 of the compstatin family, which blocks C3 activation by the convertases of all three pathways (32), and APT070 (33), which inhibits the C3 convertases thereby blocking down-stream complement activation. In this field, there is a pressing need to monitor the complement status in all patients receiving treatment with complement-regulatory drugs, a need that is only expected to increase in the future. In most


angioedema; C1-INH, C1 inhibitor; RA, rheumatoid arthritis; GPA, granulomatous polyangiitis; antibody mediated rejection; aHUS, atypical haemolytic uremic syndrome.

cases monitoring can be achieved using CP and AP functional tests (either EIA or haemolytic), since an acquired deficiency of specific complement components is created. If the inhibitor is an antibody, direct binding assays to the specific antigen can supplement these assays. A comprehensive overview of the field of therapeutic complement inhibition is found in (34).

#### Disorders With Complement Activation (Table 3)

#### SLE, Antiphospholipid Syndrome and Urticarial Vasculitis

SLE, antiphospholipid syndrome and urticarial vasculitis are autoimmune immune complex diseases (35, 36). Other members of this group include rheumatoid arthritis with vasculitis, and cryoglobulinemia, as well as very rare cases of Henoch Schönlein disease and granulomatous polyangitiis (GPA) (37, 38). Complement analyses, in particular determination of CP function and analysis of components within the CP: C1q, C3, and C4 (C2 in some laboratories) are useful markers to monitor disease activity and for differential diagnosis (**Figure 7**). Furthermore, the detection of autoantibodies against C1q and C3 can be used to verify diagnosis (39–41). Hypocomplementemic urticarial vasculitis syndrome (HUVS) features anti-C1q antibodies with distinctive specificity as well as severe complement consumption via the CP (36, 42).

#### Antibody Mediated Rejection in Transplantation (AMR)

AMR is the leading cause of long-term kidney graft loss (43, 44). The presence or formation of antibodies directed TABLE 2 | Hereditary complement deficiencies.


Bacterial infections, e.g., Neisseria

MBL, mannan-binding lectin; MASPs, MBL associated serine proteases; 3MC syndrome Malpuech-Michels-Mingarelli-Carnevale syndrome; SLE, systemic lupus erythematosus.

against the vascular endothelium in the graft is a major trigger of complement activation in transplantation, leading to microvascular inflammation and thrombosis followed by ischemia, apoptosis, or necrosis, and finally graft failure. AMR leads to a CP activation, which is presented in biopsies as C4d deposition. The majority of antibodies are either antiblood group ABO antibodies (so called natural antibodies) or antibodies against HLA, due to previous immunization. In blood in a severe AMR the typical signs of a CP activation are seen with low CP function, low levels of C1q, C4, and C3, and increased activation products (e.g., C3d,g/C3, sC5b-9 etc.). In less severe AMR only raised levels of activation products can be seen.

#### C3 Glomerulopathy (C3G)

(C9)

C3 glomerulonephritis (C3GN) and dense deposit disease (DDD) are subsets of C3Gs that present a predominant C3 deposition in the glomerulus and is associated with C3NeF, other autoantibodies to complement, and in some cases mutations in complement protein genes (45). Since C3NeF binds to and stabilizes the AP C3 convertase, a profound C3 consumption occurs, that may lead to a functional C3 deficiency. The consumed C3 gives rise to C3d,g which is an indicator of substantial activation of C3. The level of

factor B remains more or less unchanged. Since the stabilized convertase sometimes also cleaves C5, in some cases sC5b-9 can also be detected. Detection of C3NeF supports the diagnosis of C3G, in particular DDD. The severe C3 deficiency that results can, at least in theory increase the risk of bacterial infections.

#### Poststreptococcal Glomerulonephritis (PSGN)

Another type of glomerulopathy associated with AP activation is the post-streptococcal glomerulonephritis (PSGN) that may occur during the rehabilitation period in individuals that have suffered from Group A streptococcal disease. In particular C3, but also C5, is consumed and sC5b-9 is generated for typically 6–10 weeks following the infection (45). Since the levels of C3 and C5 can be depressed and, as is true for C3G with/without C3NeF, there is a theoretical risk of other bacterial infections. The levels of C3d,g are elevated resulting in a high ratio of C3d,g to C3. PSGN is associated with a concomitant consumption of properdin, which is the major complement-related diagnostic difference between these diseases (45).

#### Atypical Haemolytic Uremic Syndrome (aHUS) and Other Microangiopathies

aHUS is a disease that predominately appears in childhood and is characterized by thrombocytopenia, microangiopathic haemolytic anemia, and acute renal failure.

Affected cells in aHUS are endothelial cells, including those of the mesangium of the kidney as well as platelets and erythrocytes. aHUS is caused by uncontrolled complement activation due to combination of mutations of complement inhibitors, such as in factor H, but also factor I, factor H related proteins (FHR) 1, 3, 5, MCP, and thrombomodulin that impairs the function of these inhibitors (46). A deletion of the FHR-1/3 gene may lead to generation of anti-factor H antibodies which also is associated with aHUS. Gain of function mutations in C3 and factor B that lead to poorly controlled activation have also been reported (47, 48).

The most common cause of aHUS is mutations in the gene for factor H, and the majority of the mutations occur in the short consensus repeats (SCRs) 19 and 20 in the C-terminal of the molecule. These SCRs which interact with carbohydrates, e.g., heparan sulfate and sialic acid on the cell surface are important for binding factor H to the cell surface. Like in factor H dysfunction, other types of aHUS are also associated with AP activation, resulting in varying degrees of C3 consumption and the generation of C3d,g (and other C3 fragments) and sC5b-9. Factor H from aHUS patients may show different mobility compared to normal factor H when analyzed by SDS-PAGE, and western blotting. For a more precise diagnosis of AP components mutations, contact with a specialist laboratory is recommended.

In preeclampsia and other types of microangiopathy the levels of Bb and sC5b-9 are increased and various degrees of AP component consumption and low AP function can be seen (49).

#### Inherited and Acquired C1-INH Deficiency

C1-INH deficiency is the cause of the rare disorders hereditary angioedema (HAE) and acquired angioedema (AAE), but HAE may also, in some cases be caused by mutations in the gene coding for FXII (50). The hereditary form, HAE, is heterozygous autosomal dominant, whereas the acquired form, AAE, mainly occurs in patients with underlying disease but can also be idiopathic. Since the cause of these diseases is an unregulated generation of bradykinin by the contact system, these are not primarily complementsystem related diseases, but their diagnosis is based on complement analysis. Both HAE and AAE are associated with recurrent attacks of bradykinin-mediated, non-pitting, local angioedema that are not responsive to antihistamine or steroids.

There are different types of HAE, which can be distinguished only by laboratory analysis. Two types of C1-INH-related forms exist, one with low concentration and function of C1-INH (Type I), and one with normal concentration, but dysfunctional C1-INH (Type II). In contrast, HAE with normal C1-INH levels, which is not associated with low C1-INH function is a heterogenous group and therefore less characterized than the other types. In this group, certain patients have a gain of function form of FXII (Type III), due to a mutation in the coding gene leading to defective glycosylation (50).

Acquired deficiencies of C1-INH can occur in lymphoproliferative and autoimmune diseases as a result of formation of autoantibodies against C1-INH, or paraproteins e.g., M-components (51–53). C4 concentrations are typically low in both HAE and AAE (54). AAE occurs as the result of the hypercatabolism of C1-INH; in AAE, as opposed to HAE, the serum concentration of C1q is low in ∼70% of patients.

# AUTOANTIBODIES TO COMPLEMENT PROTEINS (TABLE 4)

#### Anti-C1q Autoantibodies

Autoantibodies against C1q (anti-C1q) were first identified as low molecular weight C1q precipitins (55). Their immunoglobulin (Ig) nature was later confirmed, and the SLE-associated anti-C1q was shown to be specific for the collagenous region of the C1q molecule (56). Anti-C1q occurs in ∼30% of unselected SLE patients but has a higher prevalence in lupus nephritis and is also associated with nephritis activity (57). In SLE, anti-C1q antibodies are often of the IgG2 subclass, but the reason for this selectivity is unknown. More than 95% of patients with HUVS are also positive for anti-C1q (58). However, anti-C1q antibodies are not specific for SLE and HUVS and may also be found in association with conditions such as primary glomerulonephritis and infectious diseases.

Anti-C1q is mainly analyzed by EIA. To avoid false-positive results in anti-C1q antibody analysis, it has been recommended that either only the collagenous part of C1q should be used as the antigen, or that high-salt buffer is used to abolish binding between the globular part of C1q and the Fc region of IgG in immune complexes (41). There are several commercially

TABLE 3 | Complement pathology.


CP, classical pathway; AP alternative pathway; C1-INH, C1 inhibitor; TP, termial pathway; SLE, systemic lupus erythematosus; MPGN, membranoproliferative glomerulonephritis; AMR, antibody mediated rejection; C3GN, C3 glomerulonephritis; aHUS, atypical haemolytic uremic syndrome; PSGN, post-streptococcal glomerulonephritis; HAE, hereditary angioedema, AAE, acquired angioedema.


Anti-C1q<sup>1</sup> , antibodies against native C1q; Anti-C1q<sup>2</sup> , antibodies against reduced chains of C1q; C3NeF, C3 nephritic factor; C4NeF, C4 nephritic factor; SLE, systemic lupus erythematosus; C3GN, C3 glomerulonephritis; aHUS, atypical haemolytic uremic syndrome; PSGN, post-streptococcal glomerulonephritis; HAE, hereditary angioedema, AAE, acquired angioedema.

available assays to detect anti-C1q. In addition, Western blot analysis of the separated C1q A, B, and C chains (after reduction) has been used to show that anti-C1q antibodies have different binding specificities in SLE than in (HUVS) (42, 59).

# C3 Nephritic Factors and Other Convertase Autoantibodies

C3NeF are autoantibodies that bind to and stabilize the AP C3 convertase (C3bBb) to prevent its extrinsic or intrinsic decay (60), prolonging the half-life of the convertase and resulting in increased consumption of C3. C3NeF are frequent in patients with (C3G; hence the name "nephritic factor"), where they are found in ∼80% of DDD cases and 40-50% of patients with C3GN (61). C3NeF can also be found in other conditions such as acquired partial lipodystrophy and an increased susceptibility to meningococcal infections secondary to persistently low C3 concentrations resulting from C3 consumption (62, 63).

Given the heterogeneity of C3NeF with regard to binding specificities and convertase-stabilizing effects, it is considered necessary to use more than one method for analysis (64, 65). Also, different methods are needed to demonstrate both the convertase-stabilizing capacity and the Ig nature of C3NeF. Most C3NeF have been shown to bind to the Bb part of the convertase (66). The heterogeneous nature of C3NeF was already described about 30 years ago in studies that showed differences regarding the dependency on properdin and the ability to activate the AP (67, 68). A recent study has also confirmed that some C3NeF are more efficient convertase stabilizers in the presence of properdin; these antibodies have been termed C5 nephritic factors (C5NeF) and shown to be associated with C5 consumption as well as low C3 (69).

C3NeF can be analyzed in several ways. The most widely used assays are (1) detection of fluid-phase C3 conversion after incubation of patient serum with normal serum at 37◦C (70, 71) and (2) a simple haemolytic assay that utilizes unsensitized sheep erythrocytes (72). However, these functional assays are nonspecific and do not establish that the C3NeF effects are caused by an antibody. Therefore, other types of EIA methods have been developed in which an AP C3 convertase is deposited in a microtiter plate (64, 73). Standardization of C3NeF analysis is as yet lacking, but recently an external quality assessment (EQA) program that includes C3NeF analysis has become available as the result of an international initiative to standardize complement analyses (74).

Rarely, patients will present autoantibodies against the CP convertase (C4b2a); these are known as C4 nephritic factors (C4NeF). Like C3NeF in the AP, these autoantibodies stabilize the CP convertase, leading to persistent C3 consumption. Also, like C3NeF, they are principally found in association with glomerulonephritis (75, 76). C4NeF can be identified via haemolytic assays in which the CP convertase is stabilized by C4NeF on sensitized sheep erythrocytes (75).

#### Anti-C1 Inhibitor Autoantibodies

Some patients with AAE have autoantibodies to C1-INH (anti-C1-INH), which can be of any Ig class. In AAE patients with monoclonal gammopathy, the Ig isotype of the M component is often identical to the anti-C1-INH isotype (77). Anti-C1- INH can increase the consumption of C1-INH or block its function, and AAE patients with anti-C1INH can benefit from treatment with B cell-inhibiting therapy (78). Although anti-C1- INH antibodies are much more common in AAE, they occur in a small fraction of patients with SLE, and anti-C1-INH IgM has also been reported in patients with HAE (79, 80). Anti-C1- INH antibodies are analyzed by EIA. Anti-C1-INH IgG, IgA, and IgM should all be determined, an unusual requirement in the context of autoantibody analysis. The assays may either detect the binding of anti-C1-INH to C1-INH or determine the capacity of the anti-C1-INH to block C1-INH function (81). Various in-house methods are used, since commercial methods are not available.

# Anti-factor H Autoantibodies

Autoantibodies against factor H are detected in 6–10% of patients with aHUS and are also found in some patients with C3G. In aHUS, anti- factor H antibodies are mainly specific for the C-terminal part of factor H, and thus they block the ability of factor H to bind negatively charged carbohydrate residues on autologous cells. In C3G, anti- factor H may be directed toward other parts of factor H and may also be monoclonal or light chain-restricted (82–84). In aHUS, positivity for antifactor H is strongly associated with a deletion of the genes for complement FHR-1/3 proteins, a deletion that is also common in the general population. aHUS with anti- factor H is considered a separate subgroup of aHUS for which the term "deficiency of CFHR plasma proteins and factor H autoantibody positive haemolytic uremic syndrome" (DEAP-HUS) has been proposed (85). Patients with anti- factor H -related aHUS will in most cases benefit from plasmapheresis and immunosuppressive treatment, unlike patients without anti- factor H, in whom C5-blocking therapy is mandatory (86). This difference implies that antifactor H is an important diagnostic marker that needs to be analyzed rapidly.

Anti- factor H antibodies are analyzed by EIA. A multilaboratory comparison of various in-house methods in 2014 established a recommended standard method, and a standard serum for calibration of arbitrary units is available (87). Commercial analysis kits are also available for anti- factor H determination.

# Other Complement Autoantibodies

Anti-complement autoantibodies with several other specificities have been described in association with various diseases. For example, autoantibodies against factor B have been detected in C3 glomerulopathy and membranoproliferative glomerulonephritis (MPGN) (88, 89), antibodies against factor I in aHUS (90), antibodies against MBL in rheumatoid arthritis (91), and antibodies against ficolins in SLE (92, 93). However, the clinical significance of most of these antibodies is not clearly defined, and their analysis has not been adopted in regular clinical practice. Autoantibodies with specificity for C3 and C4 fragments, termed immunoconglutinins, are found in different inflammatory conditions and in SLE, where they can influence C3-mediated functions (39, 94). More recently, antibodies recognizing different C3 fragments were investigated in patients with lupus nephritis and were found to be more common in patients with more severe disease (95).

#### CONCLUSIONS

In summary, complement biomarkers can be used to follow the activity of a huge number of diseases and disorders in individual patients if they are compared with baseline values of the same individual. However, independent evaluations of the complement status in individuals without previous analyses are applicable on a relatively limited number of conditions due to low sensitivity and specificity of the existing assays and to preanalytical problems. Example profiles for a typical case of each condition are presented in **Table 3**. It is likely that introduction of new complement modulatory drugs with novel indications

#### REFERENCES


will increase the demand for complement monitoring in the near future.

#### AUTHORS CONTRIBUTIONS

KE, LS, and BN wrote the article. CM prepared the figures. BP and KS edited, and all authors approved the final manuscript.

#### ACKNOWLEDGMENTS

We thank Dr. Deborah McClellan for excellent editorial assistance. This work was supported by grant 2016-2075-5.1 and 2016-04519 from the Swedish Research Council (VR), and by faculty funding for from the Linnaeus university. The study was also supported by grants development and validation of complement assays provided by the University hospitals in Uppsala and Lund.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Ekdahl, Persson, Mohlin, Sandholm, Skattum and Nilsson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Age and Sex-Associated Changes of Complement Activity and Complement Levels in a Healthy Caucasian Population

Mariana Gaya da Costa1†, Felix Poppelaars 1,2†, Cees van Kooten<sup>3</sup> , Tom E. Mollnes 4,5,6 , Francesco Tedesco<sup>7</sup> , Reinhard Würzner <sup>8</sup> , Leendert A. Trouw9,10, Lennart Truedsson<sup>11</sup> , Mohamed R. Daha1,3, Anja Roos 12† and Marc A. Seelen<sup>1</sup> \* †

<sup>1</sup> Division of Nephrology, Department of Internal Medicine, University of Groningen, University Medical Center Groningen, Groningen, Netherlands, <sup>2</sup> Department of Obstetrics and Gynecology, Martini Hospital, Groningen, Netherlands, <sup>3</sup> Department of Nephrology, University of Leiden, Leiden University Medical Center, Leiden, Netherlands, <sup>4</sup> Department of Immunology, Oslo University Hospital and University of Oslo, Oslo, Norway, <sup>5</sup> Research Laboratory, Bodø Hospital, and K.G. Jebsen TREC, University of Tromsø, Tromsø, Norway, <sup>6</sup> Centre of Molecular Inflammation Research, Norwegian University of Science and Technology, Trondheim, Norway, <sup>7</sup> Immunorheumatology Research Laboratory, Istituto Auxologico Italiano, IRCCS, Milan, Italy, <sup>8</sup> Department of Hygiene, Microbiology and Public Health, Medical University of Innsbruck, Innsbruck, Austria, <sup>9</sup> Department of Rheumatology, Leiden University Medical Center, Leiden, Netherlands, <sup>10</sup> Department of Immunohematology and Blood Transfusion, Leiden University Medical Center, Leiden, Netherlands, <sup>11</sup> Department of Laboratory Medicine, Section of Microbiology, Immunology and Glycobiology, Lund University, Lund, Sweden, <sup>12</sup> Department of Medical Microbiology and Immunology, St. Antonius Hospital, Nieuwegein, Netherlands

#### Edited by:

Uday Kishore, Brunel University London, United Kingdom

#### Reviewed by:

Teizo Fujita, Fukushima Medical University, Japan Robert Braidwood Sim, University of Oxford, United Kingdom Kenneth Reid, University of Oxford, United Kingdom

#### \*Correspondence:

Marc A. Seelen m.seelen@umcg.nl

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 30 June 2018 Accepted: 29 October 2018 Published: 20 November 2018

#### Citation:

Gaya da Costa M, Poppelaars F, van Kooten C, Mollnes TE, Tedesco F, Würzner R, Trouw LA, Truedsson L, Daha MR, Roos A and Seelen MA (2018) Age and Sex-Associated Changes of Complement Activity and Complement Levels in a Healthy Caucasian Population. Front. Immunol. 9:2664. doi: 10.3389/fimmu.2018.02664 Introduction: The complement system is essential for an adequate immune response. Much attention has been given to the role of complement in disease. However, to better understand complement in pathology, it is crucial to first analyze this system under different physiological conditions. The aim of the present study was therefore to investigate the inter-individual variation in complement activity and the influences of age and sex.

Methods: Complement levels and functional activity were determined in 120 healthy volunteers, 60 women, 60 men, age range 20–69 year. Serum functional activity of the classical pathway (CP), lectin pathway activated by mannan (MBL-LP) and alternative pathway (AP) was measured in sera, using deposition of C5b-9 as readout. In addition, levels of C1q, MBL, MASP-1, MASP-2, ficolin-2, ficolin-3, C2, C4, C3, C5, C6, C7, C8, C9, factor B, factor D, properdin, C1-inhibitor and C4b-binding protein, were determined. Age- and sex-related differences were evaluated.

Results: Significantly lower AP activity was found in females compared to males. Further analysis of the AP revealed lower C3 and properdin levels in females, while factor D concentrations were higher. MBL-LP activity was not influenced by sex, but MBL and ficolin-3 levels were significantly lower in females compared to males. There were no significant differences in CP activity or CP components between females and males, nevertheless females had significantly lower levels of the terminal components. The CP and AP activity was significantly higher in the elderly, in contrast to MBL-LP activity. Moreover, C1-inhibitor, C5, C8, and C9 increased with age in contrast to a decrease of factor D and C3 levels. In-depth analysis of the functional activity assays revealed

**100**

that MBL-LP activity was predominantly dependent on MBL and MASP-2 concentration, whereas CP activity relied on C2, C1-inhibitor and C5 levels. AP activity was strongly and directly associated with levels of C3, factor B and C5.

Conclusion: This study demonstrated significant sex and age-related differences in complement levels and functionality in the healthy population. Therefore, age and sex analysis should be taken into consideration when discussing complement-related pathologies and subsequent complement-targeted therapies.

Keywords: complement, health, sex and age, innate imunity, gender

#### INTRODUCTION

The complement system, a major component of innate immunity, plays a crucial role in the immune response. In health, maintaining a balance between activation and inhibition of the complement system is key to preserve tissue homeostasis and to enable immune surveillance (1, 2). However, an overactive system can cause autoimmune and inflammatory diseases, whereas an inactive complement system results in an increased risk for infection. Several elements can disrupt this delicate balance and with age these effects are exacerbated. Complement deficiencies or dysfunctions have been shown to be associated with diseases such as atypical hemolytic uremic syndrome (aHUS), age-related macular degeneration (AMD), paroxysmal nocturnal hemoglobinuria (PNH), systemic lupus erythematosus (SLE), C3 glomerulopathy (C3G), and other kidney diseases (3, 4). Furthermore, complement activation contributes to several diseases and pathological conditions such as renal replacement therapy (5–7), cancer (8), hypersensitivity reactions (9, 10), and neurological conditions (11).

The complement system is activated via an enzymatic cascade reaction and has three different activation pathways, namely the classical pathway (CP), the lectin pathway (LP), and the alternative pathway (AP). Complement activity for each of these pathways depends on the expression and function of a large number of complement proteins. The CP is mainly initiated by immune complexes binding to C1q leading to activation of C1r and C1s, but can also be activated in an antibody independent manner like C1q binding to C-reactive protein. The LP is initiated by the binding of mannose-binding lectin (MBL), ficolins, or collectins to sugars or acetylated compounds resulting in the activation of the MBL-associated serine protease (MASP)-1 and MASP-2. Recently it has also been shown that the LP can be activated by naturally occurring antibodies, underscoring the overlap and cross-talk between the pathways (12). LP and CP activation both lead to the cleavage of C4 and C2, and subsequently to the generation of C4b2a, the C3-convertase. Furthermore, C1-inhibitor (C1-INH) regulates the activity of the recognition complexes, while C4b-binding protein (C4bp) functions as a cofactor for factor I-mediated cleavage of C4b. Initiation of the AP occurs via spontaneous hydrolysis of C3 into C3(H2O) or by the binding of C3b to altered surfaces. Factor D cleaves factor B when the latter is complexed with C3b, creating the C3-convertase of the AP, C3bBb. In addition, this C3-convertase is stabilized by properdin, the only positive regulator of the complement system. Regardless of the initial pathway, complement activation can lead to the initiation of the terminal pathway (TP) and thereby the generation of C5 convertase, which cleaves C5 in C5a, a powerful anaphylatoxin, and C5b. Next, C5b binds C6 which then attaches to a surface and interacts with C7, C8, and C9 to form the membrane attack complex (MAC/C5b-9) (13, 14). If there is no surface present, C5b6 will bind to C7, C8 and C9 together with the control proteins vitronectin and clusterin in the fluid-phase and thereby the soluble form of the terminal complement complex, sC5b-9, is formed. To prevent unintended complement activation the system is kept under control by a variety of regulators. The major regulators of the AP are the plasma proteins factor H and factor I. Factor H inhibits complement activation by accelerating the decay of the C3bBb convertase of the AP and by providing cofactor activity for factor I-mediated cleavage of C3b.

Both age and sex are known to influence and significantly impact the immune system (15). Females and males have distinct innate and adaptive immune responses (16). Moreover, they also differ in their immunological responses to self and foreign-antigens (17). Sex-based immunological differences can be found in various species (18). In general, the immune response seems to be stronger in females than in males (15). These immunological sex differences are thought to arise from discrepancies in hormones, genetic factors and environmental mediators. Notably, these sex-related differences in the immune system could give insight into the epidemiology and etiology of autoimmune and infectious diseases (15). Likewise, aging is associated with a progressive decline in immunity. The impact of aging on adaptive immunity is well accepted (19). However, less certainty exists on the effect of aging on innate immunity (20). Impaired function of neutrophils and macrophages as well as reduced interaction between dendritic cells and T cells, suggest also a decline in innate immune function (21). As a result of impaired immune function, the ability of elderly to

**Abbreviations:** aHUS, Atypical haemolytic uremic syndrome; AMD, Age-related macular degeneration; AP, Alternative pathway; C1-INH, C1-inhibitor; C4bp, C4b-binding protein; C3G, C3 glomerulopathy; CP, Classical pathway; IQR, Interquartile range; LP, Lectin pathway; MAC, Membrane attack complex; MASP, MBL-associated serine protease; MBL, Mannan-binding lectin; MBL-LP, MBL-induced lectin pathway; PNH, Paroxysmal nocturnal hemoglobinuria; SLE, Systemic lupus erythematosus; SNP, Single-nucleotide polymorphism; TP, Terminal pathway.

respond to microorganisms is diminished and the number of infectious disease is increased (22). Also the increased incidence of autoimmune diseases might be related to altered immunity in elderly (23).

In the current era of personalized medicine combined with the recent success of complement inhibitors in clinical trials, it is essential to identify the influence of sex and age on the complement system. Given that complement therapeutics are an effective treatment for complement-mediated diseases, individuals will need different doses in order to efficiently block the complement system depending on the concentrations and functional activity of the complement components. Factors influencing the levels could be e.g., sex and age. Thus, the aim of the present study was to explore the effect of age and sex on the complement system in a healthy Caucasian population by performing functional and quantitative complement analyses. The current study also enables us to better understand the complement system in different physiological conditions.

### MATERIALS AND METHODS

#### Serum Samples

Serum samples were obtained from a population of 120 healthy Caucasian individuals from Norway, registered as blood donors. From each sex, 12 samples were obtained per age decade between 20 and 70. Sixty females with a mean age of 44.7 years (range: 20– 69 years) and 60 males with a mean age of 45.1 years (range: 20– 65 years) were included. Serum samples obtained were directly aliquoted and stored at −80 ◦ C.

FIGURE 1 | Complement pathway activity according to sex. The activity of the classical pathway (CP), alternative pathway (AP), and MBL-induced lectin pathway (MBL-LP) was measured in 120 Caucasian healthy subjects, of which 60 males and 60 females. The solid lines indicate the median values in each group. The differences between males and females was assessed by the Mann Whitney test (\*\*\*P < 0.001). CP/AP activity is referred to the left Y-axis in linear scale whereas MBL-LP activity is referred to the right Y-axis in a logarithm scale.

# Assessment of Pathway Activity in Normal Human Serum Samples

The complement kit (Complement System Screen Wieslab <sup>R</sup> , Eurodiagnostica, Malmö, Sweden) for assessment of CP, LP, and AP activity was used according to the manufacturer's instructions (24, 25). In brief, strips of wells for CP were coated with IgM, strips for MBL-LP were coated with mannan and strips for AP were coated with LPS. Sera were diluted 1/101 for the CP and MBL-LP assays and 1/18 for the AP assay in specific buffers to ensure that activation of only the actual pathway occurred, and were incubated for 1 h at 37◦C. After washing, alkaline phosphatase-conjugated anti-human C5b-9 was added before incubation at room temperature for 30 min. Additional washing was performed, substrate was added and the wells were incubated for 30 min. Finally, absorbance values were read at 450 nm. In each assay standard positive and negative control sera, provided in the kit as lyophilized material and reconstituted with distilled water, were assessed. The positive standard serum was a pool of 5 sera from healthy individuals and the negative control consisted of sera heat-inactivated at 56◦C for 20 min. Complement activity was calculated using the following formula: activity = 100% x (mean A450 (sample)–mean A450 (negative control) / (mean A450 (standard serum) - mean A450 (negative control). Samples as well as standard serum and negative control serum were tested in duplicate.

#### Serum Complement Concentrations

Assessment of MBL concentrations was performed using a commercial MBL ELISA obtained from BIOPORTO Diagnostics A/S (Hellerup, Denmark) and performed according to the manufacturer's instructions. C1q, C4, C3, and C1-inhibitor concentrations were assessed using nephelometry on a BN-Prospec, with validated diagnostic protocols provided by the manufacturer (Siemens Healthcare Diagnostics, Marburg, Germany), calibrated using a complement standard serum provided by the manufacturer. C2 and factor B were measured using rocket-immunoelectrophoresis as described by Sjöholm et al. (26–28). Properdin and Factor D were assessed by electroimmunoassay as previously described (29, 30). Concentrations of C2, factor B, properdin and factor D were expressed in arbitrary units (% of pooled human serum).

MASP-1, MASP-2, Ficolin-2, and Ficolin-3 levels were determined using in-house ELISAs. For all in house ELISA's (Leiden University Medical Center, Leiden, the Netherlands) a standard protocol was used. In brief, Nunc Maxisorb plates (Nunc, Roskilde, Denmark) were coated using coating buffer (100 mM Na2CO3/NaHCO3, pH 9.6), for 16 h at room temperature. After each step, plates were washed three times with PBS containing 0.05% Tween 20. Residual binding sites were blocked by incubation with PBS containing 1% BSA. Unless otherwise indicated, all subsequent steps were incubated in PBS containing 0.05% Tween 20 and 1% BSA, for 1 h at 37◦C. Detection antibodies were conjugated to digoxigenin (Dig) using Dig-3-O-methylcarbonyl-εaminocaproic acid-N-hydroxysuccinimide ester (from Boehringer Mannheim, Mannheim, Germany), followed

by detection using HRP-conjugated rabbit anti-Dig Abs (Fab, from Boehringer Mannheim). Enzyme activity of HRP was detected using 2,2′ -azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) (Sigma) and absorption was measured at 415 nm.

In more detail, for the detection of MASP-1 and MASP-2, polyclonal antibodies were raised in rabbits against the recombinant protease domain of MASP-1 and MASP-2, respectively, kindly provided by Dr. Peter Gál (31). However, based on the specificity of the anti-MASP-1 antibody it is likely that the results from this ELISA represent the sum of MASP-1 and MASP-3 levels. ELISA plates were coated with rabbit IgG anti-MASP-1 and rabbit IgG anti-MASP-2, respectively. Patient samples were incubated in GVB/NaCl/EDTA (Veronalbuffered saline containing 0.05 % Tween-20, 0.1 % gelatine, 0.5M NaCl, 10 mM EDTA; pH 7.5), followed by detection using the same antibodies, which were conjugated as described above. Results were expressed in arbitrary units per ml using a standard line of serially diluted pooled normal human serum for calibration. For detection of ficolin-2, a similar protocol was followed using mouse monoclonal antibody GN5. Results were expressed in pg/ml as described previously (32). Both GN5 and 4H5 were kindly provided by dr. T. Fujita, Department of Immunology, Fukushima Medical University, Fukushima, Japan). For detection of ficolin-3, the standard ELISA protocol was followed and the mouse monoclonal antibody

4H5 was used for both coating and detection. Results were expressed in arbitrary units/ml using pooled human serum for calibration.

C4bp, C5, C8, and C9 were measured by ELISA as previously described (33–35). C6 and C7 were measured using an in-house ELISA assay as previously described (36). In brief, wells were coated with an in house mouse monoclonal antibody against C6 (WU 6-4, Hycult, Uden, NL) and with an in house polyclonal antibody against C7. For the detection, respective biotinylated polyclonal antibodies were used. For the set-up of the ELISA, purified C6 or C7 (Cytotech, San Diego, CA, USA) were used to calibrate human serum as a standard.

The standards used in the different assays were not calibrated to the official European complement standard.

#### Statistics

Statistical analysis was performed using IBM SPSS 22.0 (IBM Corporation, Chicago, IL, USA). Concentrations of complement proteins and pathway activity were presented as median with interquartile range [IQR]. Differences between sexes were assessed with the Mann Whitney U test. Correlation between age and complement proteins or pathway activity was evaluated using the Spearman Rank correlation coefficient (r). Univariate and subsequent multivariate linear regression analysis were performed to identify independent determinants of complement pathway activity. Multivariate analysis models were constructed using backward selection (Pout > 0.05) including complement proteins that were associated significantly (P < 0.05) with complement pathway activity in univariate analysis. P < 0.05 were considered statistically significant.

#### RESULTS

#### Differences Between Males and Females in Complement Pathway Activity and Components

CP, MBL-LP, and AP activity were determined by deposition of C5b-9 in serum samples from 60 males and 60 females. CP and MBL-LP activity were similar in both sexes (**Figure 1**). However, AP activity was significantly lower in females compared to males. The median AP activity for females was 69.5% compared with 81.0% for males (P < 0.001, **Figure 1**).

In accordance, the levels of C3 and properdin were also significantly lower in women in comparison with men. Median C3 concentrations in females were 1.37 mg/mL [1.19–1.59] compared to 1.51 mg/mL [1.34–1.76] in males (P = 0.001, **Figure 2A**), whereas median levels of properdin in females were 107% [91–128] vs. 120% [98–138] in males (P = 0.03, **Figure 2B**). Conversely, serum concentration of factor D showed an opposite pattern (P < 0.001, **Figure 2C**), since levels were significantly higher in females (140%, IQR: 115%−200%) than in males (100%, IQR: 82%−137%). Males had higher levels of serum MBL in relation to female subjects. Median MBL concentrations were 533 ng/mL [142–1,076] and 843 ng/mL [289–1,646] in females and males, respectively (P = 0.03, **Figure 2D**). Likewise, female subjects also had significantly lower levels of ficolin-3 compared TABLE 1 | Levels of complement components that did not significantly differ between the sexes.


The current table displays levels of components that were not significantly different. Values are presented as median and interquartile range [IQR]. P-values represent the difference between males and females tested by Mann Whitney test. Abbreviations: C1- INH, C1-inhibitor; MASP, MBL-associated serine protease; C4bp, C4b-binding protein. The standards used in the different assays were not calibrated to the official European complement standard.

to male subjects, with a median of 830 AU/mL [676–1,034] and 1042 AU/mL [883–1,192] in females and males, respectively (P = 0.001, **Figure 2E**).

The rest of the components measured did not differ between females and males (**Table 1**). In line with CP activity, serum levels of C1q, C4, C2, C1-INH, and C4bp did not differ between the sexes (**Table 1**). Furthermore, despite the fact that MBL-LP activity was not significantly different between the sexes, a trend was still seen for lower activity in females compared to males. Additionally, 17% of the males were MBL deficient (0% MBL-LP activity) whereas in females this was 23%. The threshold of MBL concentration to result in zero activity was 130 ng/mL. The other components from the LP such as MASP-1, MASP-2 and Ficolin-2 did not differ between both sexes (**Table 1**).

Quantification of the terminal complex components revealed that all these components, except C6, were significantly lower in female subjects compared to male subjects (P < 0.05, **Figure 3**). The presented data show that in women, levels were 53, 15, 59, and 14% lower, for C5, C7, C8, and C9, respectively. For C8, concentrations above 50µg/ml were exclusively observed in males (65% of male subjects had concentrations of C8 above 50µg/ml).

#### Age-Related Changes in Complement Pathway Activity and Components

Next, we determined the effect of age on the complement system by correlating complement components and activity with age in 120 healthy individuals with a mean age of 45 years ± 13.5, ranging from 20 to 69 years old. Linear regression analysis demonstrated a significant age-related effect for CP and AP activity, but not for the MBL-LP activity (data not shown). In accordance, age significantly correlated with CP activity (r = 0.42, P < 0.001, **Figure 4A**) and AP activity (r = 0.30, P < 0.001, **Figure 4B**), but not with MBL-LP activity (**Figure 4C**).

However, when the twenty-five oldest individuals (mean age 62 ± 2.3 years) were compared to the twenty-five youngest individuals (mean age 26 ± 3.3 years) significant differences were observed in all pathways: higher CP activity (106 vs. 91%, P < 0.001), lower MBL-LP activity (35 vs. 60%, P = 0.01) and higher AP activity (83 vs. 74%, P = 0.01). Subsequently, we analyzed the effect of age on the concentration of the different complement components and factors. Regardless of statistics, a correlation coefficient below 0.3 was not perceived as clinically relevant (37). A significant effect of age was observed for C1- INH, factor D, C5, C8, and C9 levels (**Figure 5**). C1-INH showed an age-related rise (**Figure 5A**, r = 0.30, P = 0.001), whereas factor D (**Figure 5B**, r = −0.32, P = 0.001) levels decreased with age. Consistent with the age-related increase in CP and AP activity, levels of C5 (**Figure 5C**, r = 0.40, P < 0.001), C8 (**Figure 5D**, r = 0.43, P < 0.001) and C9 (**Figure 5E**, r = 0.31, P = 0.001) positively correlated with age. The comparison between the oldest and youngest twenty-five individuals showed that C5, C8, and C9 levels increased by 47% (P < 0.001), 60% (P < 0.001), and 24% (P = 0.001), respectively, with age. Finally, if we correct for sex, the impact of age on CP and AP activity remains significant. However, the age-related effect on AP is predominantly found in female subjects (females: r = 0.35, P = 0.05), whereas the effect of age on CP activity

was predominantly found in male subjects (males: r = 0.61, P < 0.001).

#### The Relationship Between Complement Components and Pathway Activity

To better understand the determinants of complement activity of the different pathways, we related individual complement components with their respective complement pathway (**Tables 2**–**4**). CP activity correlated significantly with C2 levels (**Table 2**, r = 0.51, P < 0.001), whereas MBL-LP activity significantly correlated with MBL concentrations (**Figure 6**, **Table 3**, r = 0.89, P < 0.001). Furthermore, for AP activity significant correlations were seen with properdin (**Table 4**, r = 0.35, P < 0.001), factor B (**Table 4**, r = 0.51 P < 0.001) and C3 (**Table 4**, r = 0.42, P < 0.001). In addition, we correlated terminal pathway components with complement activity of the different pathways (**Table 5**). CP activity correlated significantly with C5 (**Table 5**, r = 0.39, P < 0.001) and C9 (**Table 2**, r = 0.42, P < 0.001), while AP activity correlated with C5 (**Table 5**, r = 0.56, P < 0.001), C8 (**Table 5**, r = 0.43, P < 0.001) and C9 (**Table 5**, r = 0.38, P < 0.001). MBL-LP activity did not correlate with any of the terminal pathway components (**Table 5**). However, besides the correlation between individual complement components with their respective complement

FIGURE 4 | Correlations between complement pathway activity and age. In 120 healthy subjects age was correlated to the activity of the (A) classical pathway (CP), (B) alternative pathway (AP) and (C) MBL-induced lectin pathway (MBL-LP). These correlations were evaluated using the Spearman Rank correlation coefficient. P < 0.05 were considered to be statistically significant.

pathway, there were also correlations between different complement components with each other. The latter could form a possible confounder for the pathway analysis. Subsequently, we performed a multivariate linear regression analysis and found that levels of C2, C1-INH, and C5 were independent determinants of CP activity (model R <sup>2</sup> = 0.46, **Table 6**), whereas MBL and MASP-2 were independent determinants of MBL-LP activity (model R <sup>2</sup> = 0.66, **Table 6**). Moreover, the multivariate linear regression analysis demonstrated that only C3, factor B and C5 were independent determinants of AP activity (model R <sup>2</sup> = 0.46, **Table 6**).

#### DISCUSSION

In the current study, we demonstrate in a healthy Caucasian population that sex and age significantly impact the complement system. Sex-related analysis revealed that females have lower AP activity and lower AP and LP complement components, with an exception for factor D. In addition, females showed significantly lower levels of C3 and terminal pathway components. These results demonstrate that females have significantly lower complement activity and levels of complement components compared to males. Furthermore, age-related analysis showed that aging is associated with an enhanced functional activity of the CP and the AP. Correspondingly, terminal pathway components levels increased with age. Lastly, we analyzed the determinants of functional complement pathway activity, by relating individual complement components with their respective complement pathway. In healthy individuals, CP activity was determined by C2, C1-INH, and C5, MBL-LP activity by MBL and AP activity by C3, factor B and C5. These results support the relevance of age- and sexmatched control cohorts in studies related to the complement field, including the assessment of reference values used in clinical laboratory diagnostics. Moreover, these results suggest that age and sex should be taken into account in complement-related pathology as well as in complement-targeted therapies.

For the majority of complement components the liver is the predominant source of production, and production of complement proteins can be regulated by an acute phase response (38). Other main production sites includes peripheral blood mononuclear cells which are responsible for the production of C1q, properdin and C7, adipocytes which produce factor D and the lungs where ficolin-3 is mainly produced (39, 40). In fact, most tissues and inflammatory cells are able to produce various complement proteins, e.g., upon stimulation with cytokines (40). Additionally, genetic environmental and lifestyle factors such as obesity and smoking also influence complement levels (41– 43). Multiple complement deficiencies have been described and associated with pathology (44). In addition, single-nucleotide polymorphisms (SNP) can strongly affect the concentration and/or function of various complement proteins, as have been well-documented for e.g., MBL (45, 46). Yet, limited studies have investigated the influence of sex and age on the complement system (47–49).

The immune system varies between males and females and differences in innate and adaptive immunity have already been demonstrated. At least two factors are known to explain the influences of sexual dimorphism in immunity: genetics (e.g., the X chromosome) and hormonal differences (50). Remarkably, the X chromosome contains genes that encode for several proteins related to immune response such as toll-like receptors and interleukins. Moreover, properdin is encoded on the short arm



Spearman's correlation was performed and data are presented as correlation coefficient and corresponding P-value. Significant correlations are highlighted. C1-INH, C1-inhibitor; CP, classical pathway activity; C4bp, C4b-binding protein.

TABLE 3 | Correlations between MBL-lectin pathway functional activity, complement levels, and age.


Spearman's correlation was performed and data are presented as correlation coefficient and corresponding P-value. Significant correlations are highlighted.C1-INH, C1-inhibitor; C4bp, C4b-binding protein; MBL, mannose-binding lectin; MBL-P, MBL-induced pathway activity; MASP, MBL-associated serine protease.

of the X chromosome (51). However, in our study, females had lower properdin levels than males, demonstrating once more that genetic factors do not solely determine the concentration of the components. In addition, sex hormones are known to influence innate and adaptive immunity. The majority of the cells from innate and adaptive immunity express estrogen receptors (52). However, data on the influence of sex on the complement system remains limited. A study by Roach et al. demonstrated that in children (age range 1–19 years) there were differences in several complement components between boys and girls (49). However, the differences found in this study were age-dependent. Troldborg et al. studied complement proteins restricted to the lectin pathway and showed comparable results to ours, showing that female have lower complement levels than males (53). Previously, an animal study demonstrated the influence of sex hormones on the complement system by injecting estrogen and testosterone in healthy and castrated mice from both sexes (54). Mice treated with testosterone showed increased late acting


TABLE 4 | Correlations between alternative pathway functional activity, complement levels, and age.

Spearman's correlation was performed and data are presented as correlation coefficient and corresponding P-value. Significant correlations are highlighted. AP, alternative pathway activity.

complement activity while mice treated with estrogen showed diminished activity (48). In accordance, a recent animal study demonstrated lower complement functionality in female mice due to lower levels of terminal complement components (47). This has been a reason for our group to perform complementrelated animals experiments solely in male rodents (55, 56). Altogether, these results in rodents are in line with our findings that females have lower terminal complement components and lower functional activity. In our study, only Factor D was significantly higher in woman. A possible explanation could be a higher amount of adipose tissue in woman than in men, resulting in enhanced production. Unfortunately, factor H and factor I were not determined in our cohort. Previous studies did not find significant differences between sexes in factor H levels in adults (57). In addition, very recent work did not observe an effect of sex or age in factor H levels in children (54). However, since we observed a lower AP functional activity in females, a possible explanation for this difference could be higher levels of factor H and/or factor I. Nonetheless, the clinical consequences of these differences remain to be elucidated. Yet, some clues already indicate that sex could form a possible confounder in complement-mediated diseases. For instance, in a cohort of healthy people, low MBL levels were associated with cardiovascular disease, however this association was seen only in men and not in women (58). Moreover, in a study of AMD, distinct alterations were shown between the sexes in levels of AP components (59). Thus, more studies are needed to clarify the significance of sexual dimorphism on the complement system. In the future, these shortcomings could be addressed by studying the complement system in transsexual subjects undergoing hormonal replacement therapy. Nevertheless, an important question that remains is: should we treat women and men equally when it comes to complement therapeutics?

In the early phase of life, innate immunity plays a fundamental role since adaptive immunity is still under development, whereas during adolescence this is reversed (60). Immunity undergoes severe deterioration with age (15, 60). Previous studies showed that both for innate and adaptive immunity, the function reduces with age (22, 61). Whether this decline in innate

immune function is also true for the complement system was still unknown. In a Japanese healthy cohort, C3 levels varied according to different age ranges, however not in a continuous way (62). Certain LP components were also investigated in a large population and showed MBL levels are reduced in adults when compared to children (63) Previously, in a cohort of centenarians, MBL levels were reported to be reduced compared to the general population (64). However, this reduction was based on a higher prevalence of mbl2 gene mutations, suggesting a beneficial role of intermediate levels of MBL for longevity. In our study, we found that age had a minor effect on MBL-LP activity, but significantly enhanced the activity of CP and AP. Accordingly, in an earlier study in healthy subjects aged between 20 and 69 years old, increased CH50 activity during aging was observed


TABLE 5 | Correlations between terminal pathway components, functional pathway activity and age.

Spearman's correlation was performed and data are presented as correlation coefficient and corresponding P-value. Significant correlations are highlighted. AP, alternative pathway activity, CP, classical pathway activity; MBL-LP, Mannose-Binding Lectin-induced pathway activity.

TABLE 6 | Univariate and multivariate determinants of functional complement pathway activity.


Multivariate linear regression analysis was performed to identify independent determinants of complement pathway activity. All variables described in Tables 2–5 were tested, only variables with P < 0.05 in the univariate analysis are shown. Multivariate analysis models were constructed using backward selection (Pout > 0.05) including complement proteins that were associated significantly (P < 0.05) with complement pathway activity in univariate analysis. C1-INH, C1-inhibitor; C4bp, C4b-binding protein; MBL, Mannose-Binding Lectin; MASP-2, MBL-associated serine protease 2.

and associated with increased levels of individual components of the CP (65). Moreover, in line with the enhanced functional activity of CP and AP, the levels of terminal pathway components C5, C8, and C9 also raised with age. However, an important limitation of our results is that our study design is cross-sectional and not longitudinal. Accordingly, our study is unable to discriminate between true age-related changes or better survival due to evolutionary advantages. Possible explanations for the aging immune system could be epigenetic changes, metabolic changes, changes in protein production and degradation and accumulation of senescence cells. Finally, the increased levels of terminal pathway components with age could be a mechanism to compensate for the impaired clearance of pathogens and apoptotic cells due to lower cellular immunity (66).

The current findings help to become conscious of changes in complement function during physiological conditions. Furthermore, the data obtained provide insight in the complex teamwork of all the different complement proteins to achieve the end goal, i.e., the production of the membrane attack complex upon complement activation. To further study the contribution of each individual component to their respective pathway(s) we first correlated individual complement components with their complement pathway function. However, since complement components also correlated with each other, this could cause a potential confounder. Thus, we next performed multivariate regression analysis, to correct for other components and confounders. In this analysis, MBL and MASP-2 levels were independent determinants of MBL-P activity. For the CP activity, C2 and C5 levels were the strongest determinants. In accordance, C2 has previously been described as the rate-limiting factor in the CP activation (67). In addition, C1-INH was also an independent determinant of CP activity, however with a negative value, confirming that C1-INH acts as a negative regulator of CP activation (68, 69). Furthermore, AP activity was dependent on C3, factor B and C5. Factor D has previously been described as the rate-limiting step of the AP, however the current study does not confirm these results (67, 70). Our studies confirm the key role of C5 as a rate limiting step for activation of the terminal pathway of complement, both via the CP and via the AP.

Expression of complement genes and complement function is at least partially genetically determined. In the present study, the function of the MBL pathway was strongly determined by the concentration of MBL, of which the expression of functional molecules is largely genetically controlled by a number of SNP in the promoter and coding region (45, 71). Our data also show that the concentration of factor B is positively correlated to the concentrations of C4 (r = 0.39) and C2 (r = 0.42), suggesting co-regulation of the genes involved, which are located in close proximity to each other in the MHC class III region at chromosome 6 (72). Similarly, expression of C5 and C8 is strongly correlated, suggesting co-regulation of expression of genes for C5 and C8-gamma, located at chromosome 9q33.

The present study focused on complement function and individual component concentration. It was out of the scope to investigate complement activation products reflecting basic ongoing in vivo complement activation. In contrast to differences found in native component concentrations, we have previously shown that a number of complement activation products did not differ between female and male blood donors indicating that the degree of physiologic activation not necessarily is reflected by the different individual component concentrations (73). Furthermore, the concentration of activation products did not differ between the age groups divided in decades from 20 to 70 years (unpublished data). Others have shown that there is an individual diurnal variation of complement components and anaphylatoxins dependent of sleep (74).

We acknowledge that the present study has limitations. Although our cohort includes a variety of subjects in different age ranges, one limitation is the absence of age below and above blood donor acceptance (i.e., 20 and 69 years). Additionally, we did not have genetic and lifestyle background from the subjects of the study, which could add valuable information. However, all donors were accepted as blood donors according to the health criteria for donating blood at the hospital, implying that they represent the health part of the population. Furthermore, the complement system comprises over more than 50 proteins and we were therefore unable to determine all complement proteins.

#### REFERENCES


Concerning LP, the current study only measured functional activity of MBL and not from the other initiators such as ficolins and collectins. Finally, the standards used in the different assays were not calibrated to the official European complement standard. However, this would not change our conclusions, just the absolute values. On the other hand, strengths include the number of complement proteins determined, the combination of quantitative and functional analysis, the population size and the in-depth performed statistical analysis.

In conclusion, there are important sex- and age-related differences in the complement system. These changes should be taken into account when studying complement-related diseases. Furthermore, complement therapies that are now reaching the clinical phase should be tested for the different sexes and age ranges since different complement profiles might affect the efficiency of a therapy.

#### ETHICS STATEMENT

The study was approved by the local ethical committee of the North-Norwegian Health Authority. The Blood donors were recruited from the Bloodbank and signed a written informed consent before blood donation.

#### AUTHOR CONTRIBUTIONS

Research idea and study design by MS, AR, MD, CvK; data acquisition by MS, AR, TM, FT, RW, LAT and LeT; data analysis/interpretation by MG, FP, MS, and AR. Statistical analysis by FP, MG; FP, MG, and MS wrote the manuscript. All authors were involved in editing the final manuscript. All authors read and approved the final manuscript.

#### ACKNOWLEDGMENTS

The authors like to thank the members of the EU consortium Complement in disease for helpful comments and discussions, Dr. M.W. Turner, Dr. J. Wieslander, Dr. T. Fujita, and Dr. Peter Gál for valuable reagents used in the present studies. The authors thank the department of Clinical Chemistry, LUMC, Leiden, The Netherlands for their contribution in nephelometric measurements. Part of this work was supported by a grant from the European Union (QLGT-CT2001-01039) and by the Dutch Kidney Foundation (PC 95, C98-1763).


MBL-associated serine protease-2 (MASP-2). Pediatr Allergy Immunol. (2011) 22:424–30. doi: 10.1111/j.1399-3038.2010.01104.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Gaya da Costa, Poppelaars, van Kooten, Mollnes, Tedesco, Würzner, Trouw, Truedsson, Daha, Roos and Seelen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A New Tool for Complement Research: *In vitro* Reconstituted Human Classical Complement Pathway

Michele Mutti\*, Katharina Ramoni, Gábor Nagy † , Eszter Nagy † and Valéria Szijártó†

Arsanis Biosciences, Vienna, Austria

#### *Edited by:*

Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France

#### *Reviewed by:*

Ashley Frazer-Abel, Exsera BioLabs, United States Cees Van Kooten, Leiden University, Netherlands Stephen Reece, Kymab Ltd, United Kingdom

*\*Correspondence:*

Michele Mutti michele.mutti@arsanis.com

*†Present Address:*

Gábor Nagy, Eszter Nagy, and Valéria Szijártó, Independent Researchers, Vienna, Austria

#### *Specialty section:*

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

*Received:* 23 June 2018 *Accepted:* 12 November 2018 *Published:* 04 December 2018

#### *Citation:*

Mutti M, Ramoni K, Nagy G, Nagy E and Szijártó V (2018) A New Tool for Complement Research: In vitro Reconstituted Human Classical Complement Pathway. Front. Immunol. 9:2770. doi: 10.3389/fimmu.2018.02770 The complement, as part of the innate immune system, represents the first line of defense against Gram-negative bacteria invading the bloodstream. The complement system is a zymogen cascade that ultimately assemble into the so-called membrane attack complex (MAC), which lyses Gram-negative bacteria upon insertion into the outer membrane. Traditionally, serum has been used as complement source, for example to study the bactericidal activity of monoclonal antibodies or antibodies raised upon vaccination. Due to the significant donor to donor variability, as well as susceptibility of complement factors to handling and storage conditions, assay reproducibility using human serum is low. Moreover, the presence of pre-existing antibodies and antimicrobial compounds are confounding factors. To remove antibodies from human serum, we applied κ/λ-light chain specific affinity chromatography, however the method severely reduced the complement activity due to the depletion of complement components. Therefore, we attempted to reconstitute human complement—namely the alternative (rAP) and the classical (rCP) pathways—from purified complement factors. We found that adding C1-inhibitor to the mixture was essential to maintain a stable and functional C1 and thus to generate an active rCP. We further confirmed the functionality of the rCP by testing the complement-dependent bactericidal activity of a human monoclonal antibody, A1124 against an E. coli clinical isolate belonging to the ST131 clonal complex, and that of a polyclonal IVIg against a laboratory E. coli strain (MG1655) not expressing LPS O-antigen and capsule. Although the alternative pathway did not have any bactericidal activity by itself, it enhanced MAC deposition induced by rCP and increased the overall bactericidal activity against the ST131 E. coli strain. In conclusion, we report for the first time the successful in vitro reconstitution of the classical pathway of the human complement to establish a serum-free, complement dependent bactericidal assay. This system offers high level of standardization and could support the study of the complement in different research fields.

Keywords: C1-INH, bactericidal activity, monoclonal Ab, classical pathway, alternative pathway, *E. coli* ST131, MG1655, complement system

# INTRODUCTION

As part of the innate immunity, the complement system is one of the first lines of defense against Gram-negative pathogens. Besides its direct bactericidal activity, the activation of the complement system stimulates phagocytosis and triggers pro-inflammatory signaling. The three different pathways alternative, classical and lectin—converge into the terminal pathway (TP) that leads to the assembly of the C5b-9 complex, also called Membrane Attack Complex (MAC). The alternative pathway (AP) is spontaneously activated on the pathogen surface. The classical (CP) and lectin pathways (LP), are induced by an initial receptor-ligand recognition on the target surface. In case of the CP, the antigen-antibody complex (or immune complex, IC) activates the complement cascade by binding to C1, a complex formed by C1q, C1r, and C1s. In the LP, various lectins recognize specific sugar structures on the microbial surface and activate the Mannose Binding Protein-Associated Serine Proteases (MASPs) (1). Both CP and LP are negatively regulated by the C1 inhibitor (C1-INH), which sequesters and inhibits C1r, C1s, and MASPs (2). In the absence of C1-INH, C1 undergoes spontaneous activation (3, 4). Cleavage and activation of the various complement factors leads to the formation of the C5 convertase cleaving C5 to C5a and C5b, initiating the terminal pathway. C5b, attached to the target surface, associates with C6, C7, and C8 and induces the polymerization of C9 monomers, forming the MAC. MAC is a pore forming complex able to insert into lipid membranes, including the outer membrane of Gram-negative pathogens to induce membrane damage and cell lysis.

The different complement pathways interact with each other establishing a complex network in the serum (1, 5, 6). This complexity is further increased by the presence of additional serum bactericidal factors that may act in concert with the complement, or work independently against the invading pathogens (7–9). Therefore, it is difficult to dissect the contribution of the different complement pathways to the bactericidal action of the serum. It is especially relevant for studies using human serum as complement source due to the heterogeneity of the pre-existing antibody repertoires against human pathogens (10) and complement activity (11). While previous studies reported the successful reconstitution in vitro of the alternative pathway from individual components (12, 13), up to our knowledge, there have been no reports of a functionally active reconstituted classical pathway, particularly not to study its bactericidal activity.

We aimed to establish an in vitro model system with purified human complement components that allows studying complement without confounding factors.

#### METHODS AND MATERIALS

#### Bacterial Strains and Media

E. coli strains 81009 (14, 15) and MG1655 (16) were grown in LB or on Trypticase soy agar (TSA) plates (bioMérieux). To prepare overnight cultures, a single colony was inoculated in LB and incubated at 37 ◦C with agitation at 200 RPM overnight. Mid-logphase bacterial cultures were obtained by diluting the overnight cultures 1:100 in LB and incubating at 37◦C at 200 RPM until the cultures reached an OD<sup>600</sup> of ∼0.5.

#### Complement Factors, Sera, and Reagents

Rabbit erythrocytes, Gelatin Veronal Buffer with or without EDTA, as well as the complement factors C3 (A113c), H (A137), I (A138), B (A135), P (A139), C5 (A120), C6 (A123), C7 (A124), C8 (A125), C9 (A126), C1 (A098), C2 (A112), C4 (A105c), and C1-INH (A140) were purchased from CompTech. Dullbecco's phosphate-buffer saline (DPBS) without (14190144) or with calcium and magnesium (14040133) was obtained from ThermoFisher Scientific. Human serum albumin (HSA) (Albiomin, 200 g/L) was from Biotest. The monoclonal antibody A1124, a humanized IgG1 targeting the LPS O25b-antigen of E. coli ST131-H30 strains, as well as an isotype-matched control mAb were expressed in CHO cells as described by Guachalla et al. (17). Blood was taken from healthy volunteers into Vacutainer <sup>R</sup> clot activator tube (367896, Becton Dickinson and Company), and the off-the-clot normal human serum (NHS) was separated by centrifugation and aliquots were stored at −80◦C until use. ClairYG <sup>R</sup> (50 mg/mL) was obtained from LFB-Biomedicaments.

#### Determination of Complement Activity

The complement activity was measured by commercial ELISAbased kits from Eurodiagnostica detecting C5b-9 deposition specific to classical (Wieslab COMPL CP310 RUO), alternative (Wieslab COMPL AP330 RUO) and MBP (Wieslab COMPL MP320 RUO) pathways according to manufacturer's instructions. Hemolytic activity of the complement was determined using rabbit erythrocytes according to the manufacturer's instructions using NHS as positive control.

#### Removal of Antibodies From Human Serum

To remove antibodies from NHS, κ/λ-light chain specific affinity chromatography was employed. The matrix was prepared by mixing three parts of κ-light chain specific beads with one part of λ-light chain specific beads (083310 and 084910, respectively, Life Technology). To avoid clogging of the resin, NHS was spin-filtered before incubating with double volume of resin at 4◦C for 10 min. The supernatant was separated from the matrix by centrifugation (Ig depleted serum, Ig-dep). As control, equivalent volume of serum was treated identically but without the chromatography matrix (mock). Both Ig-depleted and mock sera were concentrated with Vivaspin 500 (10,000 MWCO PES, Sartorius). The dilution factor was determined based on absorbance measurement at 405, 490, and 650 nm, compared to that of untreated serum. To remove beads, fibers and precipitates, the samples were spin-filtered. Bound antibodies were recovered by washing the resin with DPBS with Ca/Mg, eluting with 0.1 M glycine pH 2.0 buffer and neutralizing with 1 M Tris pH 8.0. This immunoglobulin fraction was concentrated with Vivaspin 20 (10,000 MWCO PES, Sartorius) followed by Vivaspin 500 and finally spin-filtered. In all cases UltrafreeTM-MC Centrifugal Filter Devices 0.22µm (EMD Millipore) was employed for spinfiltration. Samples were stored at −80◦C.

# Preparation of Fab Fragment of Human IVIg

Fab fragments of human IVIg (ClairYg <sup>R</sup> ) were prepared by digestion with agarose-immobilized papain (20341, ThermoFisher) according to the manufacturer's instructions. DPBS with 10 mM EDTA pH 7.0 was used as buffer. The uncleaved Igs and Fc fragments were removed by using MabSelect Sure <sup>R</sup> (17543801, GE Healthcare) according to the manufacturer's instructions. The flow-through, containing the Fabs, was buffer exchanged in DPBS, concentrated (Vivaspin 500, Sartorius) and spin filtered (UltrafreeTM-MC Centrifugal Filter Devices 0.22µm; EMD Millipore) to ensure sterility. An untreated aliquot of ClairYg <sup>R</sup> was buffer exchanged and treated the same way to be used as control. The concertation of both Fab and full IgG in DPBS was determined by bicinchoninic acid (BCA) assay (23225, ThermoFisher).

# Surface Staining of *E. Coli*

To detect surface binding antibodies, 10<sup>8</sup> CFU/ml bacteria were incubated with various concentration of ClairYg <sup>R</sup> or its Fab fragments, followed by incubation with secondary detection reagent Alexa Fluor 488 goat anti-human IgG Fcγ or Alexa Fluor 488 goat anti-human Fab<sup>2</sup> (109-546-170 and 109-546-097 respectively, Jackson Immuno) and stained with 5µM SYTO 62 nucleic acid stain (S11344, ThermoFisher). Samples were measured in a CytoFLEX flow cytometer (Beckman Coulter) and data were analyzed using the FCS Express Flow 5 (De Novo Software).

#### Immunoblotting

Samples diluted 1:400 in Laemmli buffer were separated on Mini-PROTEAN TGX 4-20% gels (4561096 or 4561094, Bio Rad) and blotted onto 0.22µm nitrocellulose membrane using the high MW program of the TransBlot Turbo system (Bio Rad). The membrane was blocked in 5 % skim milk in Tris Buffered Saline (TBS, Fischer scientific), incubated with 1:50,000 diluted horseradish peroxidase (HRP) conjugated goat anti-human IgG (H+L) (109-035-088, Jackson ImmunoResearch) and 1:50,000 HRP conjugated goat anti-human albumin (PA1-28334, Pierce). To detect C4 cleavage, the membrane was incubated with goat anti-human C4 serum (1:2,000, A205, CompTech) as primary antibody and with donkey IgG-HRP anti-goat-IgG (1:40,000, 6420-05, Southern Biotech) as secondary antibody. Blots were developed with Amersham ECL Prime solution reagent (GE Healthcare).

#### Determination of C1q and MBL2 Levels in Human Sera

Concentrations of C1q and MBL2 were determined by ELISA using a commercial kit according to the manufacturer's instructions (HK356 and HK323, respectively, HycultBiotech). The serum samples were diluted 1:200 for C1q and 1:50 for MBL2 determination. The Ig depleted sera and the eluates were diluted according to the estimated protein concentrations.

# Preparation of C1/C1-INH Mixture

C1-INH was concentrated to about 2 mg/mL (Vivaspin 500, Sartorius), mixed with about 0.2–0.4 mg/mL C1 and 25 mg/mL HSA, and incubated for 10 min on ice. This mixture or C1 alone was dialyzed twice against DPBS with Ca/Mg at 4◦C for 3 h followed by concentration to ∼1.3 mg/mL of C1 (Vivaspin 500, Sartorius). In the C1/C1-INH mixture, the C1-INH to C1 ratio was 1.8 times the physiological ratio. The samples were spin filtered to ensure sterility (UltrafreeTM-MC Centrifugal Filter Devices 0.22µm; EMD Millipore) and then aliquoted and stored at −80◦C until use. At least of three different batches of C1/C1- INH were prepared and used.

# Complementation Assay With C1q-Depleted Serum

C1 alone or premixed with C1-INH was incubated on ice or at 37◦C for 60 min, and added to C1q-depleted serum (A300, CompTech). The activity of the classical pathway was measured with the ELISA-based kit described above.

# C4 Cleavage Assay

Heat-aggregated IgG (HAGG) was generated by incubating ClairYg <sup>R</sup> in DPBS with Ca/Mg at 63◦C for 30 min as described previously (18). C4 (200µg/mL), HSA (20 mg/mL) and HAGG (750µg/mL) were mixed on ice (final concentration indicated). C1 (67.5µg/mL) combined with C1-INH at various final concentrations ranging between 68.5 and 685µg/mL was added to the mixture. The concentration of Ca and Mg were kept constant at 0.45 mM and 0.25 mM, respectively. After 5 min on ice, the reaction was started by incubation at 37◦C for 5 min. The reaction was stopped by boiling in reducing Laemmli buffer (1610737, BIORAD for 5 min. C4 cleavage was detected by immunoblotting as described above.

#### Reconstitution of the Human Complement

The alternative pathway (rAP) was reconstituted as reported by Schreiber et al. (12, 13). Briefly, C3, factor H and factor I were incubated for 30 min at 37◦C. After cooling the mixture on ice, factor B, factor P, and the terminal pathway components C5, C6, C7, C8, and C9 were added. For the reconstitution of the rCP, C3 was incubated for 30 min at 37◦C (analogous to treatment in rAP), cooled on ice, then TP components, C2, C4, and C1/C1-INH mixture were added, and then the solution was supplemented with DPBS and HSA. For the combination of the alternative and classical pathways (rAP+rCP), the alternative pathway was reconstituted as above, then TP components, C2, C4, and C1/C1-INH mixture were added, and finally the mixture was supplemented with DPBS and HSA. In all experiments the final concentration of Ca, Mg and HSA was 0.17 mM, 0.09 mM, and 25 mg/mL respectively. The proteins were added in the order described above. The final concentration of each factor matched the physiological concentration of 50% diluted human serum (3, 12): C3 at 600µg/mL, C5 at 36µg/mL, C6 at 32µg/mL, C7 at 27µg/mL, C8 at 27µg/mL, C9 at 29.5µg/mL, factor B at 100µg/mL, factor D at 1µg/mL, factor H at 235µg/mL, factor I at 17µg/mL, factor P at 10µg/mL, C2 at 12.5µg/mL, C4 at 200µg/mL, C1 at 67.5µg/mL and C1-INH at 120µg/mL.

#### Activity of rCP With or Without C1-INH

When the rCP with or without C1-INH was tested, the preparations were treated as follows. Two mixtures of rCP with or without C1-INH were prepared simultaneously as described above, on ice and aliquoted into 5 tubes. One aliquot was snap frozen in liquid nitrogen and stored at −80◦C. The remaining four aliquots were incubated at 37◦C or on ice for 10 min or 60 min and then snap frozen. For measuring complement activity, the aliquots were thawed on ice at once and the assay was started immediately.

#### Bactericidal Assay

Survival of bacteria was assayed as described previously (19). Briefly, bacteria at about ∼5 × 10<sup>4</sup> CFU/mL (collected from midlog phase culture) were added to the reconstituted complement and incubated at 37◦C with agitation. After 3 h, surviving bacterial count was enumerated by plating. In case of E. coli strain MG1655, the final concentration of each factor matched the physiological concentration in 25% diluted human serum, while in assays with the E. coli strain 81009, the factors were set to 50% of the physiological concentration.

#### Stability of rCP

rCP mixture was divided into four aliquots; one was immediately frozen at −80◦C (untreated control), one underwent 3 cycles of freeze and thaw, one was stored at 4◦C for 48 h, and one was incubated at 37◦C for 24 h before freezing at −80◦C. On the day of the testing, the samples were thawed at once and their complement activity was compared with the ELISA based kit as described above.

#### Statistical Analysis

Data were analyzed with paired two-tailed t-test using GraphPad Prism version 6. Difference was considered statistically significant if P < 0.05.

# RESULTS

#### Removal of Antibodies From Human Serum Impairs the Complement

First, we tested the possibility of removing antibodies from normal human serum (NHS) to generate a complement source devoid of pre-existing antibodies for studying complement-mediated bactericidal activity. We employed affinity chromatography with Ig light chain (κ and λ) specific resin to remove all isotypes of antibodies. Based on immunoblot analysis, no IgGs could be detected in the depleted sera (**Figure 1A**). Next, we determined whether Ig depletion affected the complement activity of the human serum samples. The four NHS tested had comparable classical and alternative pathway activities prior to depletion, however, all had reduced or no activity after incubation with the resin (**Figure 1B**). We detected both MBL2 and C1q in the elution fractions together with the antibodies (**Supplementary Figure 1**).

Given that depletion of antibodies from NHS was not possible without compromising the complement activity, we next attempted to reconstitute the classical and the alternative complement pathways in vitro.

### *In vitro* Reconstitution of the Alternative Complement Pathway

First, we wanted to recapitulate the results of Schreiber et al. (12) reported four decades ago demonstrating the functional activity of reconstituted alternative pathway (rAP). We combined purified complement components in the following order: C3, factor H and factor I incubated at 37◦C, then adding factor B, factor P, and the terminal pathway components C5, C6, C7, C8, and C9. As a measure of complement activity, we detected C5b-9 assembly by ELISA (**Figure 2**) or by the rabbit erythrocyte lysis assay (**Supplementary Figure 2**). The activity of the rAP measured by ELISA was in the range of the AP activity of untreated NHS samples and the rabbit RBC lytic activity of rAP was also comparable to that of NHS (**Supplementary Figure 2**). The activity in both assays was specific, as it was completely lost upon the exclusion of the C5 component from to the rAP mixture.

### C1-INH Is Required to Preserve the Functionality of the Classical Pathway

Analogous to the generation of rAP, we attempted to reconstitute the classical pathway (rCP) by combining all factors from C1 to C9. First, C3 was incubated at 37◦C, and cooled on ice before the terminal pathway components C5, 6, 7, 8 9, then C2, C4, and finally C1 were added at concentrations equivalent to those in 50% human serum. To test the stability of the reconstituted classical pathway, the mixture was preincubated for up to 60 min at 37◦C, before being assayed for the C5b-9 deposition. To resemble the physiologic activation, we measured the activity of the classical pathway with an ELISA kit, with wells coated with IgM, mimicking the immunocomplexes. Surprisingly, we detected only a very low complement activity (approximately 20% that of human serum) that was even further decreased when the mixture was incubated on ice or was completely lost after incubation at 37◦C (**Figure 3A**).

Since the rAP, that shares many components with the CP, was functionally active and stable, we focused our investigation on the components specific for the classical pathway, particularly C1. To assess the activity of C1, we employed a C1q depleted (C1q-dep) serum preparation that lacked any CP activity and supplemented this with C1. We found that C1 was able to rescue the classical pathway activity of the C1q depleted serum, but with decreased efficiency after pre-incubation of C1 at 37◦C (**Figure 3B**). These data suggested that the purified C1 was functional, but it lost its ability to induce the C5b-9 deposition upon incubation at 37◦C.

We speculated that the loss of C1 activity was due to the spontaneous activation of C1 itself. Therefore, we assessed whether the C1 inhibitor (C1-INH), a serine protease inhibitor naturally present in human serum, would be able to stabilize C1. Indeed, we observed that the supplementation of C1qdepleted serum with C1 pre-incubated at 37◦C in the presence of C1-INH, fully restored the CP activity (**Figure 3B**). Hence, we added C1-INH to the rCP, and the experiment was repeated as

antibody. (B) Classical (CP), alternative (AP), mannose-binding lectin (MP) pathway activity was measured by ELISA in sera depleted with light chain specific resin (Ig-dep), or in mock treated (mock) samples and compared to a positive control provided in the kit. Graph shows mean values obtained with sera from 4 donors ±SEM. C5b-9 deposition in Ig-dep was compared to the mock treated sample by two-tailed paired t-test (\*P < 0.05; \*\*P < 0.01). The circle, square, triangle and diamond represent donor #3, #5, #12, and #25, respectively.

described above. When adding C1-INH to the rCP, we observed a full classical pathway activity, in all the conditions tested (**Figure 3A**).

To monitor the C1 activity at the molecular level during the incubation of rCP, we detected C4 and C3 cleavage products by immunoblotting. We observed the proteolytic conversion of both C4 and C3 α chains to α'-chain in the absence, but not in the presence of C1-INH (**Figure 3A**). These data corroborated that spontaneous activation of C1 led to activation of the CP, which ultimately consumed the native complement factors in the mixture. On the other hand, in presence of C1-INH we did not observe any cleavage of C4 and C3, even after incubation for 1 h at 37◦C (**Figure 3A**).

To examine whether the C1-INH concentration used in the assay still allows the specific activation of C1 initiated by immune complexes (IC), we monitored C4 cleavage triggered by heataggregated Ig. We found that even large excess of C1-INH (up to 10-fold the physiological C1-INH/C1 ratio) did not impair the specific activation of C1 (**Supplementary Figure 3**).

We also tested the stability of rCP upon freeze-thaw cycles or incubation at 37◦C or at 4◦C. As shown in **Supplementary Figure 4**, the rCP is resistant to three cycles of freeze and thaw, and stable at 4◦C for 48 h, but incubation at 37◦C for 24 h significantly impairs the complement activity.

### rCP, Alone or in Combination With rAP, Is Functional and Induces a Strong Bactericidal Action

Next, we reconstituted the classical complement pathway in combination with rAP. Both rAP and rCP were functionally active when combined (rAP+rCP) (**Figure 4A**). However, when measuring the classical pathway activity in the combined system, it was lower than in the rCP alone (**Figure 4A**). These results suggested that components present in the rAP (partially) inhibit the classical pathway activity of rCP.

As proof of practical application, we tested the bactericidal activity of the rAP, rCP or their combination (rAP+rCP) against two different Escherichia coli strains in the presence of specific antibodies.

ST131 strain 81009 expressing the O25b LPS O-antigen and mAb A1124 specific for O25b (17) were co-incubated with the reconstituted complement. In presence of the rCP, we observed significant antibody-dependent bactericidal activity, while rAP alone did not support bacterial killing (**Figure 4B**). Interestingly, the mixture of rCP and rAP displayed enhanced bactericidal effect (**Figure 4B**).

Another E. coli, the K-12 strain MG1655, not expressing LPS O-antigen (rough), was incubated with a human IVIg preparation (ClairYg <sup>R</sup> ) that was shown to contain antibodies binding to the surface of this strain (**Supplementary Figure 5**). In the presence of IVIg (500µg/ml), the MG1655 strain was completely killed in rCP or rAP+rCP, while it survived in the rAP (**Figure 4C**). MG1655 also survived and even grew in rCP and rAP+rCP when incubated with ClairYg <sup>R</sup> -derived Fab,

FIGURE 3 | Preserving C1 activity by C1-INH. (A) The activity of the reconstituted classical pathway (rCP) was measured by ELISA. rCP with or without C1-INH was pre-incubated (p.i.) for 10 min or 60 min, at 37◦C or on ice, before testing for C5b-9 deposition. Graphs show mean ± SEM from 3 experiments, compared with two-tailed paired t-test (\*P < 0.05; \*\*P < 0.01; \*\*\*P < 0.001). The same samples were assayed for C4 and C3 cleavage by immunoblotting that is indicated by the appearance of the α'-chain. (B) C1 incubated for 1 h at 37◦C or on ice was added to C1q-depleted serum. The classical pathway activity was measured by ELISA detecting C5b-9 deposition. The results are expressed as % of signal compared to the positive control. Graph shows mean ± SEM from 4 independent experiments compared with two-tailed paired t-test (\*P < 0.05; \*\*\*P < 0.001).

FIGURE 4 | Complement and bactericidal activity of in vitro reconstituted complement pathways. (A) ELISA detecting the alternative (AP) or classical (CP) pathway-dependent C5b-9 deposition by the in vitro reconstituted alternative (rAP) and classical pathway (rCP), alone or in combination (rAP+rCP). The results are expressed relative to positive control. Graphs show mean values ±SEM of 6 independent experiments analyzed with two-tailed paired t-test (\*\*P < 0.01; \*\*\*P < 0.001). (B) Survival of the E. coli strain 81009 in the presence of the reconstituted complement pathways with mAb A1124 after 180 min incubation. The survival is expressed relative to the survival with an unrelated control mAb. (C) Survival of the E. coli strain MG1655 in the presence of the reconstituted complement pathways with ClairYg® after 180 min incubation. The survival is expressed relative to the survival in presence of ClairYg®-derived Fab. Graphs (B,C) show mean ± SEM of 3 and 2 independent experiments, respectively, analyzed with two-tailed paired t-test (\*P < 0.05; \*\*P < 0.01).

corroborating that the bactericidal activity was not intrinsic to the rCP.

Low amounts of human IgG contamination was detected in all three mixtures of reconstituted complement (**Supplementary Figure 6**). Contaminants were—at least partially—introduced with the human albumin. Despite the presence of contaminating antibodies, we did not detect any specific, surface binding antibodies against the two E. coli strains (data not shown).

#### DISCUSSION

Serum bactericidal assay (SBA) is the gold standard in vitro assay to assess complement-dependent bactericidal activity of antibodies raised upon active immunization (10, 20) or applied as passive immunization (17). In SBA, human or rabbit serum is used as complement source and target bacteria are incubated in presence of specific antibodies or immune serum. In such assays, bacterial survival is

compared to those with a negative control antibody or non-immune serum as well as an inactivated complement source. Most often rabbit serum is preferred over human serum, because baby rabbit sera lack pre-existing antibodies (10, 20). Previous reports showed the possibility of depleting human serum from pre-existing antibodies by using Protein G affinity purification, with the drawback that C1q and C5 were significantly depleted during the process (21). Moreover, the method did not deplete IgM or IgA, two isotypes also able to activate the complement system (1, 22).

In this study, we aimed to establish a reproducible method with a complement source free of antibodies. First, we depleted human serum from all isotypes of immunoglobulins with a light chain specific resin. While we were able to remove the majority of IgGs, this treatment abolished activity of the classical and lectin pathways due to the non-specific depletion of C1q and MBL2, significantly impaired activity of the alternative pathway, and potentially depleted other complement factors by the resin.

In addition to antibodies, non-complement serum factors can also significantly contribute to the killing of bacteria in human serum (7, 8, 23). Therefore, we decided to test a bottom-up approach and reconstituted the complement system from its individual components. While reconstituted cytolytic alternative pathway was reported by two papers of Schreiber et al. ((12), (13)), up to our knowledge there are no reports of successful reconstitution of the complete classical pathway from C1 to C9, despite the significant amount of work put into reconstituting parts of it (24–31). We aimed to establish a method to reconstitute the classical pathway also in combination with the alternative pathway, to be used in studying complement mediated bactericidal activity. To make it available for the broader scientific community, we used commercially available factors purified from human serum.

First, we reproduced previous work of reconstitution of the alternative pathway (12). This was primarily done to confirm that the purified proteins were functionally active. The protocol was slightly modified with the addition of human serum albumin as carrier protein. The mixture of the AP components was active based on the detection of the C5b-9 complex measured by ELISA and the lysis of rabbit erythrocytes, and the activity was found comparable to that of human serum (32).

When we mixed the factors of the classical pathway, we detected very low CP activity. Furthermore, short incubation at 37◦C (10 min) completely abolished this activity.

Apart from C1, C2 and C4, all components were already used and proved to be active in rAP, therefore we tested the activity of CP-specific factors. Importantly, the commercially available C1 is supplied with EDTA and protease inhibitors, which inhibit the activity of the complement factors, therefore, we had to remove these inhibitors by buffer exchange. This treatment could have affected activity or stability of the C1 or its components, like C1q. In line with this, C1 pre-incubated on ice was able to rescue the classical pathway in C1q-depleted serum, but not when pre-incubated at 37◦C. This suggested that the buffer exchange per se was not responsible for the impairment of C1 activity, but rather affected its stability. Furthermore, we speculated that the lack of C5b-9 deposition in both rCP and C1q depleted serum was due to the spontaneous activation of C1r and C1s at 37◦C. Indeed, in normal serum, the spontaneous activation of C1 is prevented by C1-INH that has a dual role; it inhibits irreversibly the zymogen form of C1s and C1r and in a reversible way the pro-zymogen form of C1r and C1s. In the presence of ICs, C1- INH disengages the pro-zymogens C1r and C1s, allowing their activation (4) (**Figure 5**). Thus, we tested whether C1-INH was able to maintain a functional C1 in an in vitro system as well without affecting the specific activity of C1. A previous study by Nielsen (33) showed that even large excess of C1-INH had no effect on the activation of C1 by immune complexes. However, the C1-INH preparation used in that work differed from the C1- INH available for us (34, 35). Therefore, we set up a method to detect the specific activation of C1 induced by heat-aggregated gamma globulins. We observed that even a 10-fold excess of the C1-INH did not impair the specific activation by IC.

Next, we showed that the classical pathway was fully rescued when C1q-depleted serum was supplemented with C1 coincubated with C1-INH at 37◦C. Likewise, the presence of C1- INH in rCP allowed full reconstitution of the classical pathway that was unchanged even after 1 h of incubation at 37◦C. We confirmed that in the absence of C1-INH, C4 was promptly cleaved, even without specific activation. The cleavage of C4 led to the cleavage of C3 as well, but with a slower kinetics. The very low activity of the classical pathway in the rCP correlated well with the amount of native C3 in the mixture. On the other hand, when C1-INH was present, both C4 and C3 remained uncleaved (in native state). In conclusion, the very low level of C5b-9 deposition by rCP in the absence of C1-INH is due to the consumption of the downstream factors, rather than to C1 instability. An interesting indirect proof for the requirement of C1-INH in CP is the phenotype of hereditary angioedema (HAE) type I and II. In HAE low activity of C1-INH leads to the activation of the contact system (which induces edema) and the collateral spontaneous activation of C1 and consumption of C4 and C2 (36).

While rCP appeared to remain active after three freezethaw cycles and at 4◦C for 48 h, incubation at 37◦C impaired the complement activity, in agreement with the observation by Mollnes et al. (37).

When we combined alternative and classical pathways, we found both to be active alone and in combination in the ELISAbased complement activity assay. The lower CP activity of rAP+rCP compared to rCP alone could be explained by the presence of Factor H and I. Factor H acts as cofactor for factor I to cleave C3b and C4b, and it can block the induction of the downstream cascade (5). In addition, factor H alone could also displace C1 from the ligand, as shown by Kishore and Sim (38).

After confirming the activity of the reconstituted in vitro complement, we proved that this experimental system is suitable for studying the complement mediated bactericidal activity of monoclonal or polyclonal antibodies. Using an extra-intestinal pathogenic E. coli strain 81009, we found that the rCP was able to induce a strong and specific bactericidal activity in combination with the specific mAb, A1124. The bactericidal activity was further increased when alternative and classical pathways were combined. This confirms that although the alternative pathway alone does not induce bactericidal effect, it enhances the activity of the classical pathway through the amplification loop. Of note, in the ELISA based activity assay we did not observe an increase in the classical pathway activity when rCP was combined with rAP, however this was due to the use of different complement dilutions in the two assays: in the ELISA the complement mixture was diluted 1:101 (vs. the 1:2 dilution used in the bactericidal assay) and at such a high dilution the contribution of the alternative pathway could not be detected.

We found that IVIg ClairYg <sup>R</sup> contained a significant amount of antibodies against MG1655, as also reported by others (39); therefore it was used as a source of Abs to target this strain. MG1655 is a rough and non-encapsulated K-12 archetype strain, extremely sensitive to the bactericidal activity of the serum. In

#### REFERENCES


agreement with its sensitivity, the strain was completely killed in both rCP and rAP+rCP in the presence of ClairYg <sup>R</sup> . Despite of the low content of contaminating antibodies in the reconstituted complement pathways, we did not observe any killing when no ClairYg <sup>R</sup> was added to the mixture. In contrast with our data, the K-12 derivative strain, W1485, was previously shown to be killed by a reconstituted alternative pathway alone (12). However, the growth conditions used in the two studies were different, and previously also no carrier protein was added to the reconstituted mixture. Moreover, it cannot be excluded that the complement factors used previously differed in purity and potentially contained bactericidal proteins.

To our knowledge, this is the first report of the successful reconstitution of the classical pathway from individual components. In this model system, it is possible to strictly control the concentration of each component or to substitute a factor with a mutated version, as well as to supplement the assay with additional purified components, as demonstrated by adding a specific antibody. By adding lectins and MASPs to the C1/C1-INH mixture, we speculate that the system can be further complemented with the lectin pathway.

We propose that the method could have a broad applicability beyond the testing of complement mediated bactericidal activity of antibodies. It was proposed that new methods are needed to study the role of the complement in the activity of the anti-CD20 mAB, rituximab in non-Hodgkin lymphomas (40). As serum wasreported to display great donor to donor heterogeneity in the killing of cancerous cells (41), the in vitro reconstituted complement may offer a simple method to be used in cancer studies. Similarly, this tool could be applied to study the role of the complement in viral pathogenesis (42), transplant rejection (43) and autoimmune diseases (44, 45).

#### AUTHOR CONTRIBUTIONS

MM performed the experiments. KR provided technical assistance. MM and VS designed the experiments. MM, VS, GN, and EN wrote the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02770/full#supplementary-material


C1-inhibitor, and a monoclonal antibody directed against the Neisserial H.8 antigen. J. Clin Invest. (1989) 83:397–403. doi: 10.1172/JCI113897


**Conflict of Interest Statement:** MM and KR are employees of Arsanis Biosciences. KR, GN, EN, and VS hold shares in Arsanis Inc, the parent company of Arsanis Biosciences.

Copyright © 2018 Mutti, Ramoni, Nagy, Nagy and Szijártó. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Development of a Quantitative Assay for the Characterization of Human Collectin-11 (CL-11, CL-K1)

Rafael Bayarri-Olmos <sup>1</sup> \*, Nikolaj Kirketerp-Moller <sup>1</sup> , Laura Pérez-Alós <sup>1</sup> , Karsten Skjodt <sup>2</sup> , Mikkel-Ole Skjoedt <sup>1</sup> and Peter Garred<sup>1</sup>

<sup>1</sup> Laboratory of Molecular Medicine, Department of Clinical Immunology, Faculty of Health and Medical Sciences, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark, <sup>2</sup> Department of Cancer and Inflammation Research, University of Southern Denmark, Odense, Denmark

#### Edited by:

Maciej Cedzynski, Institute for Medical Biology (PAN), Poland

#### Reviewed by:

Robert Braidwood Sim, University of Oxford, United Kingdom Marcin Okrój, Intercollegiate Faculty of Biotechnology of University of Gdansk ´ and Medical University of Gdansk, ´ Poland Kazue Takahashi, Harvard Medical School, United States

> \*Correspondence: Rafael Bayarri-Olmos rafael.bayarri.olmos@regionh.dk

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 03 July 2018 Accepted: 10 September 2018 Published: 28 September 2018

#### Citation:

Bayarri-Olmos R, Kirketerp-Moller N, Pérez-Alós L, Skjodt K, Skjoedt M-O and Garred P (2018) Development of a Quantitative Assay for the Characterization of Human Collectin-11 (CL-11, CL-K1). Front. Immunol. 9:2238. doi: 10.3389/fimmu.2018.02238 Collectin-11 (CL-11) is a pattern recognition molecule of the lectin pathway of complement with diverse functions spanning from host defense to embryonic development. CL-11 is found in the circulation in heterocomplexes with the homologous collectin-10 (CL-10). Abnormal CL-11 plasma levels are associated with the presence of disseminated intravascular coagulation, urinary schistosomiasis, and congenital disorders. Although there has been a marked development in the characterization of CL-11 there is still a scarcity of clinical tools for its analysis. Thus, we generated monoclonal antibodies and developed a quantitative ELISA to measure CL-11 in the circulation. The antibodies were screened against recombinant CL-11 and validated by ELISA and immunoprecipitation of serum and plasma. The best candidates were pairwise compared to develop a quantitative ELISA. The assay was validated regarding its sensitivity, reproducibility, and dilution linearity, demonstrating a satisfactory variability over a working range of 0.29–18.75 ng/ml. The mean plasma concentration of CL-11 in healthy controls was determined to be 289.4 ng/ml (range 143.2–459.4 ng/ml), highly correlated to the levels of CL/10/11 complexes (r = 0.729). Plasma CL-11 and CL-10/11 co-migrated in size exclusion chromatography as two major complexes of ∼400 and >600 kDa. Furthermore, we observed a significant decrease at admission in CL-11 plasma levels in patients admitted to intensive care with systemic inflammatory response syndrome. By using the in-house antibodies and recombinant CL-11, we found that CL-11 can bind to zymosan independently of calcium by a separate site from the carbohydrate-binding region. Finally, we showed that CL-11/MASP-2 complexes trigger C4b deposition on zymosan. In conclusion, we have developed a specific and sensitive ELISA to investigate the ever-expanding roles of CL-11 in health and disease and shown a novel interaction between CL-11 and zymosan.

Keywords: CL-11, collectin, complement system, lectin pathway, zymosan

# INTRODUCTION

CL-11 (alias collectin-11 and CL-K1) is a pattern recognition molecule (PRM) of the lectin pathway of complement with diverse functions such as pathogen recognition, tissue homeostasis, and embryonic development (1–4). It is the most recently discovered member of the collectin family, mammalian C-type lectins involved in the innate immune response of the host. Collectins recognize conserved pathogen-specific structures and altered-self molecules and mediate their removal by, among others, agglutination, complement activation, and modulation of inflammatory and adaptive immune responses (5). Collectins share a common bouquet-like structure, with monomeric subunits that assemble into oligomers of trimers. Each monomeric subunit is, in turn, composed of a cysteinerich N-terminal region, a collagen-like domain, an α-helical neck domain, and a C-terminal carbohydrate recognition domain (CRD). The biological functions of the collectins rely on the highavidity binding resulting from the oligomerization and clustering of the CRDs.

CL-11 is a secreted collectin found in the circulation (6–8). Elevated CL-11 plasma levels are associated with a decreased risk of contracting urinary schistosomiasis (9) and the presence of disseminated intravascular coagulation (8), characterized by simultaneous clotting and bleeding that frequently leads to multiple organ failure and death. CL-11 deficiency is one of the causes of the 3MC (Malpuech, Michels, Mingarelli, and Carnevale) syndrome (3, 10), a congenital disorder associated with craniofacial dysmorphism, mental retardation, growth deficiency, and physical abnormalities. CL-11 is expressed ubiquitously in the body and it appears to play an important role in maintaining homeostasis in the tissue of expression (1, 11). During embryonic development, CL-11 is highly expressed in the craniofacial cartilage and vertebral bodies, where together with MASP-3 and CL-10, may act as a guidance cue for neural crest cell migration (12). In the mature kidney, locally-produced CL-11 appears to be critical for the development of renal injury after hypoxic and hypothermic stress (13), while in the retina local CL-11 facilitates phagocytosis of photoreceptor outer segments and modulates cytokine production in the retina (4).

Recently, CL-11 has been observed to circulate in complex with another member of the collectin family, CL-10 (collectin-10, CL-L1), forming high molecular weight heterocomplexes (i.e., CL-10/11) that interact strongly with the MASPs (14). Upon binding of CL-11 or CL-10/11 heterocomplexes to carbohydrate structures on the surface of microorganisms, the MASPs activate complement by cleaving C4 and C2 and generating the lectin pathway C3 convertase (C4b2a) (2, 15). Besides their role as the central lectin pathway enzymes, the MASPs are also involved in a variety of biological processes (16–18). MASP-1 has a particularly broad substrate-specificity and it is known to participate in the coagulation cascade, inflammatory reactions, and cell activation; MASP-2 is associated with the development of renal ischemia/reperfusion injury; while MASP-3 might be involved in the modulation of the alternative pathway by activating pro-FD in resting blood.

The multiplicity of biological functions attributed to CL-11 results from its ability to both associate with the MASPs and to interact with a broad range of self and non-self structures, such as yeast, bacteria, viruses, and apoptotic cells (1, 6, 19). Glycan array studies revealed that CL-11 has a preference toward highmannose oligosaccharides and a subset of fucosylated glycans (10). This unusual binding specificity, different from the other collectins, can be explained by the presence of an extended binding site on the CRD that recognizes both the terminal and the penultimate sugar. Moreover, CL-11 has been observed to bind to certain ligands—e.g., nucleic acids (19), Aspergillus (2), and 9-O-acetylneuraminic acid (10)—in the absence of calcium via a binding site distinct from the carbohydrate-binding region.

Though the biological importance of CL-11 is evident, there is still a scarcity of clinical tools for its analysis. In this study, we generated monoclonal antibodies (mAbs) for the study of human CL-11. The mAbs were screened against serum and plasma using different immunological methods. Furthermore, we developed a quantitative ELISA based on two mAbs to measure the concentration of CL-11 in the circulation. The assay was thoroughly validated and used to determine the levels of CL-11 in plasma from healthy individuals. Our measurements of CL-11 are in good agreement with the literature and are tightly associated with those of CL-10/11. By size exclusion chromatography of plasma, we show that CL-11 forms complexes of 300–400 kDa and >600 kDa with a distinct composition of CL-10 glycosylation states and CL-11 isoforms. Finally, we demonstrate that CL-11 recognizes zymosan in a calcium-independent manner, and triggers C4b deposition in the presence of MASP-2.

# MATERIALS AND METHODS

#### Buffers

The following buffers were used: PBS (10.1 mM disodium phosphate, 1.5 mM monopotassium phosphate, 137 mM NaCl, 2.7 mM KCl), PBS-Tw (PBS, 0.5% Tween-20), TBS-Tw ± Ca/EDTA/Mg-EGTA (10 mM Tris, 150 mM NaCl, 0.05% Tween-20, ±2.5 mM CaCl2/10 mM EDTA/2 mM MgCl2, 10 mM EGTA), sample buffer (TBS-Tw-EDTA, 0.1% v/v bovine serum), barbital-Tw ± EDTA (4 mM sodium barbitone, 145 mM NaCl, 2.6 mM CaCl2, 2.1 mM MgCl2, with 0.05% Tween-20, ± 5 mM EDTA).

# Generation of Anti-CL-11 Monoclonal Antibodies (mAbs)

Two outbred NMRI mice were immunized subcutaneously three times with 25 µg of purified recombinant CL-11 (rCL-11) (R&D, USA) adsorbed to Al(OH)<sup>3</sup> and diluted 1:1 with Freund's incomplete adjuvant (Sigma-Aldrich, USA). Three days prior to the fusion, the mice received a final intravenous injection of 25 µg of antigen and adrenalin. The fusion and subsequent selection were done according to the principles of the hybridoma technology (20). Positive clones were screened in MaxiSorp microtiter plates (Thermo Fisher Scientific, USA) coated with 0.6µg/ml of polyclonal antibody (pAb) rabbit anti-CL-11 (15269-1-AP, Proteintech, USA) and in-house rCL-11 as antigen. Selected mAbs were purified using a HiTrap Protein G column in an Äkta Pure system (both from GE Healthcare, UK). Briefly, clarified hybridoma supernatant was applied onto the column and washed until a stable UV reading was observed. Bound antibodies were eluted with 0.5% citric acid and neutralized with 0.1 M Tris pH 9. Elution fractions were pooled and dialyzed against PBS. MAbs were biotinylated with 20% w/w (+)-Biotin N-hydroxysuccinimide ester (Sigma, USA) for 3 h at room temperature (RT), followed by dialysis against PBS to remove unreacted biotin.

### Production of Recombinant Proteins

Recombinant CL-11 (rCL-11) and CL-10 (rCL-10) were produced in Flp-InTM-CHO cells (Invitrogen, USA). The coding sequences (NM\_024027.4 and NM\_006438.4, respectively) were retrieved from the RefSeq NCBI nucleotide database and cloned into pcDNATM5/FRT vectors flanked by two Flp recombinase target (FRT) sites. Stable transfectants were generated by co-transfecting the coding vectors with a pOG44 plasmid encoding the Flp recombinase. Positive clones were selected for Hygromycin resistance and grown in Ham's F-12 Nutrient Mix supplemented with 0.5% FBS, 0.1 mg/ml of Penicillin/Streptomycin (Sigma) and 2 mM of L-Glutamine (all from Gibco, Thermo Fisher Scientific unless otherwise stated) according to the manufacturer's recommendations. The concentration of rCL-11 in the supernatant was calculated using commercially-available rCL-11 (R&D) as calibrator.

#### Immunoprecipitation

Anti-CL-11 mAbs Hyb-15 and Hyb-17 (described in section Generation of Anti-CL-11 Monoclonal Antibodies and Development of a CL-11 Specific Sandwich ELISA), or mouse IgG1κ isotype control (BD Biosciences, USA) (2 µg each) were coupled to 25 µl of sheep anti-mouse IgG Dynabeads (Invitrogen) end-over-end for 30 min at RT. The conjugated beads were washed with PBS-Tw and incubated end-over-end 1 h at RT with serum diluted 1:1 in TBS-Tw-EDTA. After washing with PBS-Tw, bound CL-11 was eluted with 0.5% citric acid and subjected to SDS-PAGE as described below.

### SDS-Page and Western Blot

Denatured samples were separated by electrophoresis in NuPAGE <sup>R</sup> Tris-Acetate 3–8% gels (Invitrogen, USA) according to the manufacturer's recommendation. Proteins were blotted onto a PVDF membrane (Amersham Bioscience, UK). Membranes were blocked with 5% skim milk (Merck Millipore, USA) and incubated overnight with 0.2µg/ml of pAb rabbit anti-CL-11 (15269-1-AP, Proteintech) followed by 1 h incubation with pAb swine anti-rabbit-HRP (P0399, Dako, Denmark) in a 1:5000 dilution. After thorough washing, the membranes were developed with SuperSignalTM West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) and imaged with Microchemi (Bio-imaging systems, Israel).

# CL-11 Specific Sandwich ELISA

MaxiSorp microtiter plates were coated overnight at 4◦C with 2µg/ml mAb Hyb-17 in PBS. Serial dilutions of the calibrator, serum and plasma samples in sample buffer were applied to the plate and incubated for 2 h, followed a 2 h incubation with 2µg/ml of biotinylated mAb Hyb-15 in PBS-Tw. Streptavidin-HRP conjugate (RPN1231V, Sigma) was added to the wells for 1 h in a 1:2000 dilution in PBS-Tw. TMB One (KemEnTec Diagnostics, Denmark) was used as substrate. The reaction was stopped with 0.2 M H2SO<sup>4</sup> and the optical density (OD) was recorded at 450 nm using an ELx80 absorbance reader (BioTek, USA). Plates were washed with PBS-Tw between steps and unless otherwise stated all incubations took place at RT. The optimal combination and concentration of mAbs was determined using a two-dimensional of capture and detection antibodies against serial dilutions of serum and plasma and rated according to their signal intensity and signal-to-noise ratio. The optimal incubation times and sample buffer composition were determined based on the same criteria.

#### Assay Validation: Parallelism, Precision, Limit of Detection, Working Range, and Variation

Supernatant from Flp-In CHO cells expressing rCL-11 was used as calibrator. The calibrator was applied in a two-fold dilution to generate an eight-point curve with a concentration ranging from 37.5 to 0.29 ng/ml. OD values from serial dilutions of the calibrator, commercial rCL-11 (R&D), serum and EDTA plasma pools (25 and 10 donors, respectively) were logisticallytransformed and a log(agonist) vs. response equation was applied to evaluate the parallelism between the best-fit hill slopes. The calibrator was stored at −20◦C in single-use aliquots. The limit of detection of the assay was expressed as the background absorbance plus two times the standard deviation (SD). The precision of the assay was evaluated by calculating the ratio between the OD and the mean OD of the triplicates of five donors over a 10-point dilution curve. The dilution linearity was calculated as the ratio between the interpolated concentration for each dilution and the mean concentration calculated from the linear part of the curve. The working range was determined as the concentrations for which both the ratio OD/mean OD and ratio concentration/mean concentration had a coefficient of variation (CV) <20%. Intra-assay variation was calculated as the CV of a plasma pool in 40 wells of a single microtiter plate. To calculate the inter-assay variation, serum and plasma pools were run in triplicates on four separate plates over six different days.

#### Stability of CL-11 in Serum and Plasma Samples

Serum and plasma samples from three healthy individuals were stored at 4◦C, RT, or 37◦C for up to 72 h. To study the effect of repeated freeze-thaw cycles, serum and plasma samples from a healthy donor were frozen at −80◦C and thawed at RT five times. The levels of CL-11 were compared to a control sample subjected to a single freeze-thaw cycle.

#### Measurement of CL-10/11 Complexes in Plasma

MaxiSorp microtiter plates were coated overnight at 4◦C with 2µg/ml mAb anti-CL-10 (HM2356, Hycult) in PBS. Serial dilutions of the calibrator and EDTA plasma 1:40 in sample buffer were incubated for 2 h. Detection and development were performed as described in section CL-11 Specific Sandwich ELISA.

#### Serum and Plasma Samples

The concentration of CL-11 and CL-10/11 complexes in the circulation was determined from blood of Danish healthy volunteers (n = 126). In addition, CL-11 levels were measured in plasma from patients at admission to an Intensive Care Unit at Rigshospitalet diagnosed with systemic inflammatory syndrome (n = 65). The patients have been described elsewhere (21). Informed consent was obtained from the patients or their relatives in accordance with the Helsinki Protocol. The study was approved by the local ethics committee for Copenhagen County (record no. KA 96097). Serum and plasma were processed within 1 h of blood collection and stored at −80◦C. Plasma was obtained by incubating fresh blood with EDTA, citrate, or hirudin a naturally occurring anticoagulant that binds to and inhibits activated thrombin while maintaining complement activity in blood.

#### Size Exclusion Chromatography

Serum (100 µl) from a pool of blood donors was applied into a Superdex 200 HR 10/30 column (GE Healthcare) at a flow rate of 0.5 ml/min at RT with PBS 2 mM EDTA. The relative molecular weight of CL-11 and CL-10/11 complexes was estimated using a gel filtration marker kit (29–700 kDa, Sigma). CL-11 and CL-10/CL-11 levels in the elution fractions were measured by ELISA as described previously. Results were confirmed by western blotting following Trichloroacetic acid (TCA, Sigma) precipitation. Briefly, elution fractions (500 µl) were incubated 30 min on ice with 10% v/v TCA and centrifuged at 13,000 × g for 30 min at 4◦C. Supernatants were discarded and the pellet was washed with ice-cold acetone (Sigma) at 13,000 × g for another 30 min at 4◦C. Dry precipitates were resuspended in SDS loading buffer and separated by SDS-PAGE as described above. Anti-CL-10 (0.05µg/ml, HM2356, Hycult) and pAb rabbit anti-CL-11 (0.05µg/ml, 15269-1-AP, Proteintech) were used as detection antibodies with rCL-10 and rCL-11 (in-house) as controls.

# Binding of rCL-11 to Ligands

Ligand interaction studies were performed in the fluid and solid phase. Unless otherwise stated, all reagents were purchased from Sigma. Fluid phase binding was analyzed by incubating 100 µl of zymosan particles (100 µg, Z4250), mannan-agarose (M9917), D-mannose-agarose (M6400), and Sepharose CL-2B (CL2B300) with 2.75 µg of rCL-11 in 500 µl of TBS-Tw-Ca or TBS-Tw-EDTA for 1.5 h at RT in an end-over-end shaker. Samples were washed thrice with their respective buffer and centrifuged at 3000 × g for 30 min. Pellets were resuspended in SDS loading buffer and subjected to SDS-PAGE using anti-CL-11 Hyb-6 (2µg/ml) and pAb anti-CL-11 (15269-1-AP, Proteintech) as detection antibodies under non-reducing and reducing conditions, respectively.

Solid-phase binding was determined on microtiter plates coated with zymosan (Z4250), mannan (M7504), BSA (A2153), and D-mannose-BSA (NGP1108, Dextra, UK) in two-fold dilutions starting at 10µg/ml in PBS overnight at 4◦C. The plates washed between steps with TBS-Tw, and unless otherwise stated all incubations took place at RT. rCL-11 (1µg/ml) was diluted in TBS-Tw-Ca, TBS-Tw-EDTA or TBS-Tw-Mg-EGTA and incubated for 2 h. Hyb-15 (2µg/ml) in matching buffer was used as detection antibody for 2 h, followed by 1 h incubation with Streptavidin-HRP conjugate (RPN1231V) in a 1:2000 dilution. Development was performed as described previously.

To study the effect of soluble inhibitors on the binding to zymosan and mannan, rCL-11 was incubated with two-fold dilutions of L-fucose (F2252), N-acetyl-glucosamine (A3286), or human genomic DNA (isolated in-house) in TBS-Tw-Ca or TBS-Tw-EDTA for 1 h. Protein/ligand mixes were applied to Maxisorp plates pre-coated with zymosan or mannan (10µg/ml) and were allowed to bind for 2 h. Quantification of bound rCL-11 was carried out as described before.

# CL-11 Mediated C4 Deposition on Zymosan

Microtiter plates were coated as for the solid-phase binding studies. rCL-11 (2µg/ml) in barbital-Tw or barbital-Tw-EDTA was applied to the wells and incubated for 2 h at RT, followed by rMASP-2 produced in CHO DG44 cells (0.5µg/ml, in-house) in barbital-Tw or barbital-Tw-EDTA for 4 h at 37◦C. Subsequently, purified C4 (0.5µg/ml, CompTech, USA) in barbital-Tw was added to the plate for 1 h 37◦C. Polyclonal rabbit anti-C4c (2µg/ml, Dako) was used as a detection antibody for 1.5 h at RT, followed by swine anti-rabbit-HRP (0.15µg/ml, P0399, Dako). The plates were washed between steps with barbital-Tw. Development and data recoding were carried out as before.

# Statistical Analyses

Statistical analyses were performed using GraphPad Prism version 7.02 (GraphPad Software, USA). Measurements of samples and the calibrator were performed in duplicates unless otherwise stated. The concentration of CL-11 in serum and plasma was interpolated by regression analysis using a fourparameter logistic curve fitting. Effect of storage of CL-11 concentration and differences between measured CL-11 in serum and plasma samples was analyzed using two-way ANOVA with Dunnett's multiple comparisons tests. Two-tailed person correlation coefficient was used to determine the correlation between CL-11 and CL-10/CL-11 levels. Differences between CL-11 levels in controls and SIRS patients were analyzed with the Mann-Whitney test. The half maximal inhibitory concentration (IC50) was calculated using the inhibitor concentration vs. response equation on a nonlinear curve fitting constraining the top and bottom parameter to equal 100 and 0, respectively. The significance of observed differences on C4 deposition on zymosan was assessed via one-way ANOVA with Tukey's corrections for multiple comparisons. Data are represented as mean ± SEM of three independent experiments.

# RESULTS

# Generation of Anti-CL-11 Monoclonal Antibodies and Development of A CL-11 Specific Sandwich ELISA

Stable monoclonal hybridoma populations were generated by several rounds of cloning by limiting dilution. Each cloning round, culture supernatants were screened by indirect ELISA with rCL-11. Selected positive clones were characterized further by immunoprecipitation of native CL-11 from serum followed by western blotting. As seen in **Figure 1**, under reducing conditions both mAbs Hyb-15 and Hyb-17 precipitated a protein with an apparent molecular mass of ∼34 kDa corresponding to the CL-11 monomer. Under non-reducing conditions, a characteristic

expressing rCL-11 was used as positive control. The blots were developed

ladder-like banding pattern is visible with a major protein band around 180 kDa and several others larger than 250 kDa indicating that CL-11 assembles into high molecular weight oligomers held together by interchain disulfide bridges. rCL-11 used as positive control appeared to form predominantly lower oligomeric forms. An equivalent banding pattern was also seen with the other tested mAbs (data not shown). No bands were apparent when using a mouse IgG1κ control as precipitating antibody.

The best candidates were pairwise screened by sandwich ELISA against a serum and plasma pool. The optimal combination and concentration of mAbs was determined using a two-dimensional of capture and detection antibodies and rated according to their signal intensity and signal-to-noise ratio. Based on their properties, Hyb-17 and Hyb-15 were selected as capture and detection antibody, respectively. No signals were obtained when an irrelevant mouse antibody was used as capture or detection antibody (data not shown).

#### Assay Validation

with pAb rabbit anti-CL-11.

Supernatant from CHO cells expressing rCL-11 was used as calibrator. A four-parameter fit model using the equation log(agonist) vs. response was applied to estimate the concentration of rCL-11 in the supernatant (R <sup>2</sup> > 0.996 for all curves). Parallelism was observed between the slopes of the calibrator and two batches of commercial rCL-11 (**Figure 2A**). Log-transformed values fitted a linear regression with R <sup>2</sup> > 0.98 and the slopes (ranging from 0.56 to 0.58) did not differ significantly (p > 0.05). Parallelism was also observed between the slopes of serial dilutions of the calibrator and serum and plasma pools in the range 0.29–37.5 ng/ml, with slopes between −0.7 and −0.77 that were not significantly different (p > 0.05) (**Figure 2B**). The limit of detection, defined as the mean background absorbance plus two times the SD, was 0.17 and 0.19 ng/ml in serum and plasma respectively. The precision of the assay, evaluated using the ratio between individual OD values and the mean OD values of triplicate measurements for five different donors (**Figure 2C**), was found to be acceptable (CV < 20%) for concentrations above 0.29 ng/ml. To determine the dilution linearity the ratio between the interpolated concentrations and the mean concentration of a plasma pool was calculated for each dilution of a 10-point dilution curve (**Figure 2D**). The dilution linearity was satisfactory in the range 0.15–18.75 ng/ml with a deviation under 20% from the mean concentration. The working range of the assay—defined based on the parallelism, limit of detection, precision, and dilution linearity—was 0.29–18.75 ng/ml.

Intra- and inter-assay variation were calculated to estimate the variation of the assay. The intra-assay CV, calculated by measuring one sample forty times on the same plate, was 3.96 % (**Table 1**). The inter-assay CV was 10.5%, measured in triplicates on four plates analyzed on six different occasions.

#### Effect of Storage and Freeze-Thawing

The effect of temperature on the stability of CL-11 in blood samples was evaluated on matched serum and plasma from three different donors stored at 4◦C, RT, or 37◦C for up to three days, whereas the effect of freezing and thawing was studied on serum and plasma samples from a single donor over five consecutive cycles. No significant trend on CL-11 levels was observed when the samples were stored at 4◦C (**Figure 3A**) or at RT (**Figure 3B**), nor after repeated freeze-thawing cycles (**Figure 3D**). On the other hand, storage at 37◦C degrees caused a highly significant drop in CL-11 concentration in citrate and EDTA plasma (**Figure 3C**). CL-11 levels in serum and plasma did not differ significantly.

### CL-11 Levels in Plasma

CL-11 concentration in plasma was calculated from EDTA plasma samples of 126 Danish blood donors (mean 289.4 ng/ml, range 143.2–459.4 ng/ml) using rCL-11 as calibrator (**Figure 4A**). Since it has been reported that a major fraction of CL-11 exists in complex with the homologous CL-10, we measured the same samples by sandwich ELISA using an anti-CL-10 as capture antibody and anti-CL-11 Hyb-15 as detection. The estimated mean concentration of CL-10/11 complexes in plasma was 256.2µg/ml (range 37.89–622.7) and there was a strong correlation between CL-11 and CL-10/11 levels (r = 0.729, **Figure 4B**. A pool of EDTA plasma from 10 healthy individuals was subjected to size exclusion chromatography followed by sandwich ELISA in order to define the relationship of CL-11 and CL-10/11 across their molecular weight distribution (**Figure 4C**). CL-11 and CL-10/11 migrated as two major oligomerization forms of approximately 400 and >600 kDa. ELISA results were confirmed by western blot using anti-CL-11 (**Figure 4D** above) and anti-CL-10 antibodies (**Figure 4D** below), where most of the protein is observed in fractions 7–8 and 13–14. Under reducing conditions, CL-10 migrates as a triple band of ∼40 kDa. CL-11 present in the high molecular weight complexes migrates as a single band of ∼34 kDa, whereas in the lower oligomers it appears as a double band—hinting at the occurrence of two different isoforms.

In light of the tight association between CL-11 and CL-10, we conclude that the ELISA developed in the current work is

0.146–18.75 ng/ml).


The intra-assay variation was calculated by measuring a plasma pool sample on the same microtiter plate. The inter-assay variation was determined by comparing the mean of triplicates of a serum and plasma pool on four microtiter plates on six different days.

accurate and relevant to describe the natural situation of CL-11 in the circulation.

#### CL-11 Levels in Patients With Systemic Inflammatory Response Syndrome (SIRS)

Since CL-11 is expected to participate in the response against invading pathogens by activating the lectin pathway of complement, we measured the levels of CL-11 in 65 randomly selected plasma samples from patients admitted to the intensive care unit with SIRS [reported elsewhere(21)] (**Figure 5A**). Compared to healthy controls (n = 69), CL-11 was significantly decreased in patients with SIRS and spanned a wider range of concentrations (mean 276.4 ng/ml, range 136.2–511.7 vs. 240 ng/ml, range 73.76–876.4 ng/ml for controls and SIRS patients, respectively). As observed in healthy individuals, the concentration of CL-11 in SIRS plasma was strongly correlated with that of CL-10/11 complexes (r = 0.915, **Figure 5B**).

#### Ligand Characterization

Using our validated antibodies, we studied the ligand binding characteristics of CL-11 to identify a suitable ligand for in vitro complement activation assays. Binding was performed in the presence of calcium or EDTA (as cationic chelator) in the fluid phase using zymosan particles, mannan-agarose, D-mannoseagarose, and Sepharose as negative control. CL-11 displayed a characteristic C-type lectin activity and bound to D-mannose and mannan in the presence of calcium (1, 6) (**Figures 6A,B**). Remarkably, CL-11 was also able to bind to zymosan in a calcium-independent manner, suggesting an alternative ligand recognition mechanism. We confirmed the results by coating two-fold dilutions of mannan, zymosan, D-mannose-BSA, and BSA as negative control (**Figures 6C–F**). Binding of rCL-11 (1µg/ml) was performed in the presence of calcium, EDTA, and magnesium-EGTA (a commonly used buffer for alternative pathway activation). As seen in the fluid-phase assay, calcium was required for binding to mannan and D-mannose (**Figures 6C,E**) but not to zymosan (**Figure 6D**). No binding to BSA was observed regardless of the buffer (**Figure 6F**).

In order to determine whether the observed binding profiles are a result of the involvement of different binding sites for mannan and zymosan, we compared the effect of L-fucose (a strong ligand), N-Acetyl glucosamine (GlcNAc, a weak ligand), and genomic DNA (a calcium-independent ligand) on the binding to mannan and zymosan (**Figure 7**). Briefly, CL-11 was co-incubated with serial dilutions of L-fucose or GlcNAc starting in a 125 mM, or human genomic DNA starting at 30µg/ml prior to addition to coated plates. L-fucose was an effective inhibitor of the binding to mannan in the presence of calcium with an IC50 of 36.48 mM, whereas N-acetyl-glucosamine had just a minor effect (IC50 > 125 mM) (**Figure 7A**). On zymosan, L-fucose was only a moderate inhibitor in the presence of calcium, forming an apparent plateau around 50% inhibition, while GlcNAc had no effect (**Figure 7B**). The limited effect of L-fucose suggests the presence of two separate binding sites for zymosan and carbohydrates.

CL-11 has been reported to bind to DNA in a calciumindependent fashion via an alternative binding-site. Coincubation with DNA in the presence of calcium increased the binding to mannan, suggesting a cooperative effect between both molecules (see Discussion) (**Figure 7C**). DNA did not affect zymosan binding under calcium conditions, whereas in EDTA it showed a pronounced inhibition with an IC50 of 2.3µg/ml (**Figure 7D**). These results indicate the presence of a shared binding site for nucleic acids and zymosan. Bearing all of the above in mind, we have identified a novel interaction with zymosan independent of divalent cations that we hypothesize may be relevant in environments where the binding to calcium is impaired, such as the gastrointestinal tract.

#### CL-11-Mediated C4 Deposition on Zymosan

To investigate if CL-11 bound to zymosan could trigger the lectin complement pathway, we developed an in vitro assay using rMASP-2 and purified C4 (**Figure 8**). CL-11/MASP-2 complexes on zymosan caused a significant C4b deposition under calcium conditions in a level comparable to mannan. In the presence of EDTA, C4 activation was only seen on the zymosan ligand. No C4b deposition was observed on BSA, nor in the absence of CL-11 and MASP-2.

#### DISCUSSION

We have generated mAbs against human CL-11 and established a quantitative sandwich ELISA. The antibody-producing hybridomas underwent successive selection rounds using first rCL-11, and later serum and plasma by ELISA and western immunoblotting. While all clones readily bound rCL-11, just a few reacted strongly with native CL-11. This is probably due to the existence of complexes in the circulation that hide epitopes otherwise exposed in the homomeric rCL-11 used as immunogen. In fact, immunoprecipitation of serum demonstrates that native CL-11 has a different oligomeric distribution than our recombinant protein (**Figure 1**). The best candidates were pairwise compared to develop a sandwich

samples.

ELISA. Two antibodies were selected as capture and detection antibody in the basis of their signal-to-noise ratio. The assay was validated in terms of linearity, variability, and precision, demonstrating a satisfactory variability, and parallelism between calibrator and serum and plasma across a working range of 0.29– 18.75 ng/ml (**Figure 2**). We observed no significant differences between CL-11 concentration in serum and four different widely used plasma preparations (**Figure 3**). Similarly, common storage conditions and freeze-thawing procedures did not affect the measurements of CL-11, highlighting the suitability of this assay for routine measurements in the clinic. Of interest, some of the mAbs cross-reacted with CL-11 from other species, granting the possibility of forthcoming studies into CL-11 in animal models of disease.

When optimizing the sandwich ELISA we noticed that EDTA was required to detect native CL-11 in serum and plasma (but not

Sepharose in the presence of Ca2<sup>+</sup> or EDTA under non-reducing (A) and reducing conditions (B). Binding of rCL-11 (1µg/ml) to serial dilutions of coated D-mannose-BSA (C), mannan (D), zymosan (E), and BSA (F) in the presence of Ca2+, EDTA, and Mg2+-EGTA. Data is reported as mean ± SEM of three independent experiments.

rCL-11) (**Supplementary Figures 1A–C**). Moreover, Western blot results following immunoprecipitation of serum CL-11 in a calcium-containing buffer vs. EDTA buffer resulted in a different banding pattern (**Supplementary Figure 1D**). Calcium stabilizes the tertiary structure of the CRD of CL-11 (10), and is required for the association of the MASPs with the collagen stalks of the PRMs (22, 23). To determine whether the EDTAdependent detection of native CL-11 was due to a change in its three-dimensional structure or masking of the epitope by bound MASPs, we incubated rCL-11 in the presence of calcium, EDTA, and MASP-3. Calcium and MASP-3 had no effect on the detection of rCL-11. It has been frequently observed that serum can inhibit the binding of spiked collectins and ficolins to their ligands (14). This inhibition is greatly reduced when using EDTA in the dilution buffer. At the same time others have reported the use of a calcium-dependent antibody for the purification of native CL-10/11 complexes (24). Our hypothesis is that CL-11 (or CL-10) exists in a calcium-dependent complex with a hitherto unknown protein that both interferes with detection of native CL-10/11 complexes at low serum dilutions and inhibits the binding of spiked CL-11.

Since it appears that CL-11 is found in the circulation associated with the homologous CL-10 (14), we used our assay and a semiquantitative ELISA to compare the levels of CL-11 and CL-10/11 complexes in plasma from Danish blood donors (**Figure 4**). The mean plasma concentration of CL-11

DNA were incubated with rCL-11 (1µg/ml) in the presence of Ca2<sup>+</sup> or EDTA. Data is reported as percent binding compared to the well without inhibitor. Bound rCL-11 was measured with biotinylated mAb anti-CL-11 Hyb-15. Connecting lines represent nonlinear regression fits using the equation inhibitor concentration vs. response with variable slope. Data is reported as mean ± SEM of three independent experiments.

was 289.4 ng/ml, range 143.2–459.4 ng/ml, in agreement with two previously published Danish and Japanese cohorts (284 ng/ml and 340 ng/ml, respectively) (7, 25), and equivalent to the estimated concentration of CL-10/11 (256.2 ng/ml, range 37.89– 622.7 ng/ml). However, we want to remark that lacking a carefully validated ELISA for CL-10/11, the later concentration is merely an educated guess. There was a strong correlation between CL-11 and CL-10/11 levels in plasma (r = 0.729) apparent across a wide range of molecular masses after subjecting plasma to size exclusion chromatography and measuring the levels of CL-11 and CL-10/11 in the elution fractions by ELISA and western blot. This tight association previously reported for CL-11 and CL-10 (26–28), conforms with quantitative mass spectrometry estimates suggesting that the CL-10/11 subunit is composed of one CL-11 and two CL-10 polypeptide chains (14). Plasma CL-11 and CL-10/11 eluted as two distinct peaks at 300–400 kDa and >600 kDa that may correspond with tetramers of trimeric subunits and higher forms, respectively. Although direct area-under-the-curve comparisons suggest that most of CL-10 and CL-11 exist as high molecular weight complexes in the circulation, the lack of resolution beyond 700 kDa makes it impossible to differentiate hexamers of trimers from higher oligomeric forms, that could otherwise elute in distinct peaks with a similar area-under-the-curve as the ∼400 kDa peak. Western blot of the elution fractions under-reducing conditions revealed multiple monomer bands for CL-10 and CL-11 in agreement with the observations of Henriksen et al. that CL-10 is found in the circulation as three glycosylated states and CL-11 as two differentially-spliced isoforms (14). Detection of CL-11 demonstrated the presence of a second band in the low molecular weight complexes, possibly corresponding to the shorter isoform D. Of the five isoforms predicted to be secreted by in silico analyses (29), only isoform A and D have been detected in blood (1, 14, 25). Isoform D has a shorter collagen-like domain (48 amino acids as opposed to 72 in isoform A) and lacks the putative MASP-binding motif HGKIGP (**Figure 9**). Though shorter, isoform D maintains intact the Gly-X-Y repeats of the collagen-like domain required for the formation of interchain

hydrogen bonds and stabilization of the collagen triple helix (30, 31). Because the C-terminus of isoforms A and D are identical, we hypothesize that the neck domain could direct the assembly of a trimer with a ragged N-terminal region with cysteines incapable of stabilizing high order oligomeric structures (32). Detection of CL-10 in the elution fractions revealed that non- or singly-glycosylated CL-10 was more abundant in the ∼600 kDa oligomers, while the doubly-glycosylated form was more abundant in the ∼400 kDa complexes. It has been suggested that CL-11 stabilizes and facilitates secretion of CL-10 (14), in part due to the difficulties of expressing the CL-10 recombinantly (33, 34). In the current work we describe the production of rCL-10 using stable transfected CHO cells. The protein was secreted into the supernatant and appeared to form oligomers of trimers (data not shown). Further studies in the characterization of homomeric CL-10 and the influence of alternative spliced variants and glycosylation states in the oligomerization of CL-10/11 are guaranteed, as are studies on the effect (if any) of the isoform D lacking the MASP binding motif in the weak interaction of low molecular weight CL-10/11 with the MASPs.

Based on the putative role of CL-11 in the modulation of the immune response against pathogens (2, 9), we performed a pilot study on randomly-selected plasma samples from patients diagnosed with SIRS (**Figure 5**). Notably, the mean CL-11 plasma concentration was significantly decreased in SIRS patients upon admission compared to healthy controls (p < 0.0001). Nonetheless, we choose to remain cautious and refrain from any biological interpretation until a thorough analysis of new ongoing projects can be performed. The levels of homomeric and heteromeric CL-11 were strongly associated in SIRS patients (r = 0.915), as seen for healthy controls, reinforcing the current opinion that CL-10 and CL-11 are co-regulated and exist as a hybrid molecule.

CL-11 has been shown to bind to several microorganisms (6), with a preference for complex sugars as opposed to other collectins (10). Using the developed mAbs we confirmed that our in-house recombinant protein displays the characteristics of native CL-11 and binds to mannan and D-mannose in a calcium-dependent manner (**Figure 6**). As expected, L-fucose but not GlcNAc—was an effective inhibitor of the binding to mannan (IC50 = 36.48 mM). Interestingly, rCL-11 also bound to zymosan in a calcium-independent manner that could be only partially inhibited by L-fucose in the presence of calcium. Zymosan is a major constituent of the cell wall of the yeast Saccharomyces cerevisiae. It is composed mainly of cross-linked β-glucan and α-mannan, and has been widely used to elicit immune and inflammatory responses (35–38). Several calciumindependent ligands, such as nucleic acids, have been reported for CL-11 reflecting complex interactions beyond its lectin activity (2, 10, 19). Thus, we incubated rCL-11 with DNA and measured the amount of bound CL-11 to coated mannan and zymosan (**Figure 7**). On mannan DNA appeared to increase the binding to the coated ligand. The same phenomenon has been reported by Henriksen et al., where the authors concluded that DNA may increase the avidity or align the CRDs in a more favorable conformation (19). On zymosan, DNA had a minor effect in calcium, while in EDTA buffer it was a moderately effective inhibitor (IC50 = 2.3µg/ml). We propose that CL-11 recognizes zymosan via two distinct mechanisms: a classical calcium-dependent lectin activity, and an alternative calciumindependent interaction shared with nucleic acids. Other C-type lectins, such as dectin-1 and Reg family proteins III and IV, are also capable of recognizing fungal and bacterial ligands in a metal ion-independent fashion (39–41). Interestingly, it has been proposed that the Reg family proteins evolved to bind sugars without calcium to carry their antimicrobial function in the gastrointestinal tract, where the low pH may impair the ability of classical C-type lectins to bind to carbohydrates. Similar low pH, low calcium conditions result from infection-induced local inflammation (42). Immunohistochemical analyses have revealed that CL-11 is expressed (among others) in the stomach and the intestines (11). At this stage we can only conjecture whether the calcium-independent zymosan interaction may be of biological relevance in acidic environments with low calcium availability.

Finally, we measured the ability of rCL-11 to trigger lectin pathway activation on zymosan (**Figure 8**). It has been reported that serum inhibits the binding of purified and recombinant collectins and ficolins to their ligands (14). At the same time, the relatively low concentration of CL-11 in circulation requires using serum dilutions that would trigger complement activation via the alternative pathway. To circumvent these limitations, we employed a step-wise activation assay using recombinant and purified proteins in a buffer that allows the cleavage of C4 by MASP-2—the first step of the activation of the lectin pathway. Binding of rCL-11 followed by incubation with rMASP-2 and purified C4 resulted in the deposition of C4b on coated zymosan, comparable to that on mannan. Calcium was not necessary for the binding of rCL-11 to zymosan, but for the complex formation with rMASP-2. We consider that not requiring calcium is an asset when designing a specific assay to discriminate the contribution of CL-11 from the other PRMs of the lectin pathway.

In conclusion, we have developed monoclonal antibodies and a robust and sensitive sandwich ELISA for the quantification of CL-11 in the circulation and used them to study native CL-11 in plasma and the binding characteristics of rCL-11 to different ligands.

#### AUTHOR CONTRIBUTIONS

RB-O, NK-M, LP-A, and KS performed the experiments. RB-O, KS, M-OS, and PG designed the study. RB-O, PG wrote the manuscript. All authors critically reviewed the manuscript.

#### FUNDING

The Danish Research Foundation of Independent Research (DFF-6110-00489), the Sven Andersen Research Foundation, and Novo Nordisk Research Foundation.

#### ACKNOWLEDGMENTS

The authors would like to thank Ms Jytte Bryde Clausen, Mr Lars Vitved, and Mr Jesper Andresen for their excellent technical assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02238/full#supplementary-material

Supplementary Figure 1 | Effect of calcium in the detection of native CL-11. The concentration of CL-11 was measured in three different donors in a three-fold dilution of serum (A) and plasma (B) or supernatant from CHO cells

#### REFERENCES


(C) with and without EDTA the sample buffer. (D) Western blot of immunoprecipitation (IP) of CL-11 from serum in a calcium or EDTA-containing buffer under non-reducing conditions. Lanes 1 and 2, purified rCL-11 and supernatant from CHO cells expressing rCL-11. Lane 3, serum before IP. Lanes 4 to 9, serum after IP using CL-11 specific mAbs Hyb-15 and Hyb-17 or a mouse IgG1κ isotype control antibody. The blots were developed with pAb rabbit anti-CL-11. ELISA, IP and Western blotting were performed according to sections CL-11 Specific Sandwich ELISA, Immunoprecipitation, SDS-page and western blot, respectively.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Bayarri-Olmos, Kirketerp-Moller, Pérez-Alós, Skjodt, Skjoedt and Garred. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# CL-L1 and CL-K1 Exhibit Widespread Tissue Distribution With High and Co-Localized Expression in Secretory Epithelia and Mucosa

*Soren W. K. Hansen1 \*, Josephine B. Aagaard1 , Karen B. Bjerrum1 , Eva K. Hejbøl <sup>2</sup> , Ole Nielsen2 , Henrik D. Schrøder <sup>2</sup> , Karsten Skjoedt <sup>1</sup> , Anna L. Sørensen1 , Jonas H. Graversen1 and Maiken L. Henriksen1*

*<sup>1</sup> Institute of Cancer and Inflammation Research, University of Southern Denmark, Odense, Denmark, 2Department of Pathology, Odense University Hospital, Odense, Denmark*

#### *Edited by:*

*Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France*

#### *Reviewed by:*

*Uday Kishore, Brunel University London, United Kingdom Kenneth Reid, University of Oxford, United Kingdom*

*\*Correspondence: Soren W. K. Hansen shansen@health.sdu.dk*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

> *Received: 02 April 2018 Accepted: 16 July 2018 Published: 31 July 2018*

#### *Citation:*

*Hansen SWK, Aagaard JB, Bjerrum KB, Hejbøl EK, Nielsen O, Schrøder HD, Skjoedt K, Sørensen AL, Graversen JH and Henriksen ML (2018) CL-L1 and CL-K1 Exhibit Widespread Tissue Distribution With High and Co-Localized Expression in Secretory Epithelia and Mucosa. Front. Immunol. 9:1757. doi: 10.3389/fimmu.2018.01757*

Collectin liver 1 (CL-L1, alias collectin 10) and collectin kidney 1 (CL-K1, alias collectin 11) are oligomeric pattern recognition molecules associated with the complement system, and mutations in either of their genes may lead to deficiency and developmental defects. The two collectins are reportedly localized and synthesized in the liver, kidneys, and adrenals, and can be found in the circulation as heteromeric complexes (CL-LK), which upon binding to microbial high mannose-like glycoconjugates activates the complement system *via* the lectin activation pathway. The tissue distribution of homo- vs. heteromeric CL-L1 and -K1 complexes, the mechanism of heteromeric complex formation and in which tissues this occurs, is hitherto incompletely described. We have by immunohistochemistry using monoclonal antibodies addressed the precise cellular localization of the two collectins in the main human tissues. We find that the two collectins have widespread and almost identical tissue distribution with a high expression in epithelial cells in endo-/ exocrine secretory tissues and mucosa. There is also accordance between localization of mRNA transcripts and detection of proteins, showing that local synthesis likely is responsible for peripheral localization and eventual formation of the CL-LK complexes. The functional implications of the high expression in endo-/exocrine secretory tissue and mucosa is unknown but might be associated with the activity of MASP-3, which has a similar pattern of expression and is known to potentiate the activity of the alternative complement activation pathway.

Keywords: collectin, complement system, 3MC syndrome, mucosal immunology, innate immunity

#### INTRODUCTION

The innate immune system is a first line of defense and plays major roles in preventing microorganisms in settling and becoming pathogens, and also in the mounting of suitable immune responses against eventual pathogens. It often relies on recognition and binding to pathogen-associated molecular patterns by host pattern recognition molecules (PRMs). Based on an often observed association between microbial evasion of the complement system and pathogenicity (1), the complement system appears to play substantial roles in the innate immune system. The overall antimicrobial functions of the complement system are to tag microorganisms for elimination by phagocytes, to initiate inflammation and to lyse the microorganisms. The complement system is a "self-amplificative" cascade system, mainly found in the blood, and includes several PRMs; among which some activate a pathway known as the lectin activation pathway, *via* activation of serine proteases known as MBL-associated serine protease (MASP-1, -2, and -3) (2, 3). Mannan-binding lectin (MBL) is probably the most studied PRM of the lectin activation pathway and binds to mannose-rich glycoconjugates on the surface of microorganisms, initiating complement activation. In humans, MBL deficiency may increase susceptibility toward infections in certain situations but not in general (4), most likely contributed by coincidental activation of different complement activation pathways by a given microorganism.

Collectin liver 1 (CL-L1) and collectin kidney 1 (CL-K1) are "MBL-like" proteins that also are found in the circulation in association with MASPs and with specificity toward mannose-rich glycoconjugates and negatively charged molecules (5–10). In the circulation they can be found as heteromeric molecules, referred to as CL-LK, with superior complement abilities *via* MASP-2 in comparison with their respective homomers (11). They are both synthesized in the liver by hepatocytes, in the adrenal glands and in the tubules of the kidney, in addition to other tissues as well (5–7). On the protein level, human CL-K1 has been associated with the same tissues, while human CL-L1 in the original study only was associated with hepatocytes, however, without examination or exclusion of other tissues and cells (6, 7). Among normal healthy populations of different origins, they both constitute average serum concentrations of 250–450 ng/ml, with a clear correlation between levels of CL-L1 and CL-K1, supportive of heteromeric complexes between the two or similar regulation (7, 12–15). The liver and adrenals are due to their endocrine nature and a relative high synthesis of mRNA believed to be the major organs contributing to the CL-L1 and CL-K1 found in the circulation. Partly due to the recently described characterization and association with complement, little is known about their roles *in vivo*. In a recent work using mice deficient of CL-K1, Wakamiya and colleagues showed that CL-K1 protected mice against *Streptococcus pneumonia* infections induced *via* nasal inoculation (16). However, in another work there was no protective effect of CL-K1/ CL-LK in a mouse model for infection by *Mycobacterium tuberculosis* (17). CL-K1 has been shown to bind with relative high affinity to the disaccharide Man(α1-2) and to negatively charged molecules, including nucleic acid ligands, and may also play a role in the opsonization of apoptotic cells by recognizing a combination of carbohydrate and nucleic ligands (10, 18).

Recently, it was demonstrated that CL-K1-deficient mice partly were protected against destructive complement-mediated inflammatory responses in post ischemic kidneys and that CL-K1 further promoted development of renal fibrosis in the tubules (19, 20).

CL-K1 and CL-L1 are not regulated significantly by inflammatory stimuli. Their plasma/serum levels do not correlate with increased levels of traditional inflammatory mediators, including CRP and TNF-alpha (8, 9).

The two collectins play apparently also important roles for embryogenesis. Deficiency of CL-K1 or CL-L1 leads in humans to a rare congenital developmental syndrome known as 3MC (alias Malpuech facial clefting syndrome), an effect that the two collectins share with MASP-3. It has been shown that CL-K1 and CL-L1 may act as attractants and guidance cue for neural crest cells, although the precise mechanism for embryonic involvement remains to be elucidated (21, 22).

A functional and complete activation of the complement system involves many complement factors and is in general only associated with the effect in the circulation or at inflamed sites, where blood components gain access. Most tissue expresses certain complement components, e.g., the lung and the intestinal system (23, 24). At such sites the complement system is believed to be partially functional or to mediate activation when inflammation progresses. In the quest of elucidating the function of CL-L1 and CL-K1, we characterized their localization in human tissues. It appears that the previously characterized sites of localization in the circulation, adrenals, liver, and kidney, may have disparaged a compelling localization in especially exocrine/endocrine tissues and mucosa, suggestive of that CL-L1, CL-K1, and CL-LK may play roles in the periphery as well.

# MATERIALS AND METHODS

#### General Reagents and Buffers

Unless otherwise stated, reagents were obtained from Sigma-Aldrich, Denmark. Phosphate-buffered saline: 10 mM Na2HPO4, 140 mM NaCl, and 2.7 mM KCl pH 7.4. Tris-buffered saline (TBS): 10 mM Tris and 145 mM NaCl. TNT buffer: 0.10 M Tris, 0.15 M NaCl, and 0.05% Tween 20 pH 7.5.

### Generation of MAbs Against CL-L1 and CL-K1

CL-LK was purified from outdated plasma by calcium sensitive immunoaffinity chromatography as previously described (25). Purified CL-LK (50 µg) was used for *s.c.* immunizations of outbred NMRI female mice using Gerbu as adjuvant. Three days before the fusion, mice were boosted (i.p.) with the same amount of CL-LK. The fusion between spleen cells obtained from the CL-LK immunized mice and myeloma cells (Sp2) was performed using polyethylene glycol essentially as described previously in Ref. (26). Positive clones were identified by ELISA using microtiter plates coated with either purified recombinant CL-K1 or CL-L1 expressed in CHO cells as full-length molecules without any tags. Cells from the positive wells were recloned at least thrice by the limiting dilution method. For antibody production and subsequent purification, hybridomas were grown and allowed to express the MAb in Hybridoma-SFM (Invitrogen). Monoclonal antibodies were purified by means of affinity chromatography using a HiTrap Protein G HP column (GE Healthcare) under previously described conditions and elution with 50 mM glycine, pH 2.3 (27). The two antibodies, MAb 11-1 (anti-CL-K1) and MAb 16-13 (CL-L1), which were superior in specificity and IHC sensitivity and applied in the following studies, were both of the isotype IgG1kappa.

# SDS-PAGE and Western Blotting

SDS-PAGE was performed according to the method of Laemmli (1970) using pre-casted NuPAGE 4–12% Bis-Tris gels (Invitrogen) and MES or MOPS SDS running buffer (Invitrogen) (28). Proteins were transferred to the Hybond-P polyvinylidene fluoride membrane (GE Healthcare) (29). The membrane was blocked in 5% non-fat dried milk and 0.1% HSA, and incubated with primary monoclonal antibodies (0.5 µg/ml). Subsequently, the membrane was washed and incubated with HRP-conjugated rabbit anti-mouse antibody diluted (1:20,000) accordingly to the manufacturer's recommendation (Dako, Denmark) and developed by means of the ECL plus Western blotting detection kit (GE Healthcare). For specificity testing of applied antibodies, 1 µl of serum was applied to the gel per 4 mm well width.

#### Surface Plasmon Resonance

Binding characteristics of the monoclonal antibodies, 11-1 and 16-13, was investigated by SPR on a Biacore 3000 instrument (Biacore, Sweden) with immobilized CL-K1, CL-L1, and CL-LK on a CM5 chip. The proteins were immobilized on EDC/ NHS-activated flow cells by injecting the collectin (10 µg/ml) in 10 mM acetate pH 5.0 to a surface density ranging from 0.028 to 0.042 pmol/mm2 . One M of ethanolamine, pH 8.5 was used for capping. For binding experiments antibodies were diluted in running buffer (10 mM HEPES, 150 mM NaCl2, 5 mM EDTA, and 0.005% Tween-20 pH 7.4) in a range from 2.08 to 166 nM. Each sample (40 µl) was injected with a flow rate of 5 µl/min. Regeneration was obtained for antibody 11-1 by injection of two cycles of 10 µl: 10 mM glycine, 20 mM EDTA, 500 mM NaCl, and 0.005% Tween-20 pH 4.0. Regeneration was obtained for antibody 16-13 by injection of two cycles of 10 µl: 100 mM glycine, 5 mM EDTA, 500 mM NaCl, and 0.05% Tween-20 pH 3.0. BIAevaluation software 4.0.1 was used for analysis of the data. The apparent dissociation constants were found by fitting the curves to a 1:1 binding model. All experiments were as a minimum conducted as triplicates of duplicates.

# mAbs' Impact on Ligand Binding and Complement Activation

MaxiSorpTM 96-well plate were coated with DNA (2 µg/ml, cat. no. D2001, Sigma-Aldrich) or mannan (20 µg/ml, cat. no. M3640, Sigma-Aldrich) in 1 M NaCl. Purified native CL-LK (0.25 µg/ml) was pre-incubated in non-adsorbent wells for 1 h at RT in the presence of serum, mannose, or mAbs, diluted twofold from 10%, 200 mM, or 2 µg/ml respectively, and subsequently incubated on ligand-coated plates for 4 h. Bound CL-LK was detected using biotin-labeled anti-CL-K1 mAb (0.5 µg/ml, Hyb 14-29, N-terminal specific) and streptavidin-HRP (0.1 µg/ml). Samples were diluted in TBS (20 mM Tris, 125 mM NaCl, pH 7.4) with 2 mM CaCl2, 0.05% Tween 20, and 0.1% BSA. For assessing CL-LK-mediated complement activation plates were prepared

Figure 1 | Structure of CL-LK and antibody specificity. (A) Schematic illustration of the domain organization of CL-L1 and CL-K1 polypeptide chains. The symbol "\*"on the CL-L1 polypeptide chain represent two N-linked glycosylation sites in the carbohydrate recognition domain. CL-K1 is found in the circulation in the form of two isoforms: CL-K1a represents full-length and CL-K1d represents an alternative spliced form devoid of a part of the collagen-like region. (B) Subunit of CL-LK and oligomeric structures. A total of three polypeptide chains of CL-K1 and CL-L1 join to form a heteromeric subunit, which may further oligomerize into structures ranging from dimers to hexamers of subunits, here illustrated by a tetramer. (C) Analysis of purified CL-LK by SDS-PAGE and visualization by silver staining. The three bands of CL-L1 represent non-, partially, and fully glycosylated forms of CL-L1. (D,E) Specificity of monoclonal antibodies by Western blotting of serum under. reducing conditions and visualization by ECL. (F–H) SPR analyses of monoclonal antibodies with immobilized CL-K1, CL-L1, and CL-LK as antigen, respectively. MAbs 16-13 (--) and 11-1 (---) were analyzed for binding to immobilized purified CL-L1 (F), CL-K1 (G), or CL-KL (H) MAb using concentrations ranging from 0.1 to 10 µg/ml.

as above. Purified native CL-LK (2 µg/ml) was pre-incubated in non-adsorbent wells with mAbs (0.5 or 4 µg/ml) or mannose (50 or 200 mM) at 37°C for 1 h, and subsequently incubated on ligand-coated plates at 37°C overnight. Wells were washed and incubated with recombinant MASP-2 (0.25 µg/ml for 3 h) followed by incubation with purified C4 (4 µg/ml, CompTech, Tyler, TX, USA, for 1 h) at 37°C. C4b deposition was detected using biotinylated anti-C4 mAb (0.5 µg/ml, HYB 162-02, BioPorto, Gentofte, Denmark). Samples were diluted in VBS (5 mM barbital, 142 mM NaCl, pH 7.4) with 2 mM CaCl2, 1 mM MgCl2, 0.05% Tween 20, and 0.1% BSA.

#### Human Tissue Samples

Human tissue samples were obtained from the tissue bank at the Department of Pathology, Odense University Hospital (Odense, Denmark) and derived from surgically removed specimens fixed 4% phosphate buffered formaldehyde for 12–48 h. Samples were conventionally dehydrated, and subsequently embedded in paraffin before sectioning (4–5 µm) and mounting on slides.

#### Immunohistochemistry

Paraffin-embedded, formalin-fixed human tissue sections were deparaffinized and rehydrated through serial wash in xylene and decreasing concentrations of ethanol. Endogen peroxidase activity was blocked by incubation with 1.5% H2O2 for 10 min. Antigen retrieval was performed by microwave boiling in 10 mM Tris base, 0.5 mM EDTA, and pH 9.0 buffer for 15 min. The tissue sections were washed in TNT buffer and incubated with primary antibodies MAb 16-13 (2 µg/ml), 11-1 (0.5 µg/ml), and mouse MAb anti-chicken IgY (1 µg/ml) for 1 h. The tissue sections were washed in TNT buffer and incubated with EnVision + System HRP-labeled polymer (Dako) for 30 min. After wash in TNT buffer, the tissue sections were incubated with DAB+ (Dako) for 10 min followed by staining with hematoxylin. The final immunohistochemical analysis was carried out using "multi block" sections comprising the following normal tissues: the cerebellum, esophagus, fetal and adult liver, gall bladder, kidney, large intestine, lung, skeletal muscle, pancreas, parotid gland, placenta, prostate, pylorus, spleen, tonsils, thymus, thyroid gland, rectum, small intestine, testis, and urinary bladder. The adrenal gland was derived from a patient diagnosed with pheochromocytoma.

#### Image Acquisition

Histology slides were scanned at 20× (controls) or 40× magnification using a NanoZoomer-XR (Hamamatsu Photonics, Japan). Image sections were acquired using NDP.view2 software (NanoZoomer Digital Pathology; Hamamatsu Photonics) and final JPG images were all uniformly adjusted for color saturation (+25) and light (−1) in Adobe Photoshop.

# RESULTS

#### Antibody Specificity, Affinities, and Impact on Complement Activation

The reactivity of the applied MAbs was demonstrated by Western blotting using serum as source of antigens (**Figure 1**). This analysis showed that the applied MAbs 16-13 (anti-CL-L1) and 11-1 (anti-K1) only reacted with protein bands correlating with the molecular weight of CL-L1 and CL-K1, respectively (**Figure 1**) (7). There was no cross-reactivity of the two antibodies, and both MAbs recognized all isoforms, ensuring detection

complement activity mediated by CL-LK. (A) Impact on ligand-binding. CL-LK was incubated on ligand-coated plates (mannan or DNA) in the presence or absence of the mAbs used for IHC (mAb 11-1 and mAb 16-13), mannose, or serum (known to interfere with the binding of CL-LK to ligands). mAb 131-1 (anti-MBL) was included as control. Bound CL-LK was detected with biotinylated mAb-anti-CL-K1 (compatible with the applied mAbs) and HRP-streptavidin. (B) Impact on CL-LK-mediated complement activation. Prior to incubation with MASP-2 and C4, plates were prepared with CL-LK and coated ligands (mannan or DNA) as above. Deposition of C4b was detected with biotin-labeled anti-C4b mAb and HRP-streptavidin. The results shown are representative of three independent experiments. Error bars refer to max and min of triplicate measurements. None of the tested mAbs interfered with ligand binding or complement activation. CL-LK binding to mannan and DNA occurs *via* two separate binding site, and the latter is not inhibited by mannose, whereas uncharacterized blood components inhibit both types of interaction (10, 11).

of all forms of CL-K1 and CL-L1 in the tissue sections. To further validate the specificity and reactivity, the two monoclonal antibodies were analyzed by SPR using immobilized purified collectins and antibodies in fluid phase. MAb16-13 bound to CL-L1 (*KD* = 0.16 ± 0.007 nM, means ± SD) and to CL-LK (*KD* = 0.14 ± 0.003 nM) but not to CL-K1. MAb 11-1 bound to CL-K1 (K1 *KD* = 5.4 ± 1.8 nM) and to CL-LK (*KD* = 4.6 ± 1.9 nM) but not to CL-L1. Again, cross-reactivity was undetectable and binding affinities/avidities were of satisfactory strengths. In the characterization of the two applied MAbs, we found that they neither interfered with the binding activity of CL-LK to suitable ligands nor did they modulate or inhibit the CL-LK-mediated complement activation *via* MASP-2 (**Figure 2**).

#### Immunohistochemical Localization of CL-L1 and CL-K1

In the majority of the tested tissues we observed identical localization of CL-K1 and CL-L1, both in terms of tissue and cell types. Unless the difference in immunoreactivity between the two was striking, the co-localization is not commented further, neither is the absence of staining of the tissues incubated with isotype matched control antibody. Frequently, the immunoreactivity of the CL-K1 MAb (11-1) was stronger than that of the CL-L1 MAb (16-13). This may not necessarily reflect an increase in CL-K1 quantity in comparison with CL-L1 but may originate from the nature of the antibodies (also discussed further below).

In the liver, immunoreactivity for CL-K1 and CL-L1 was associated with hepatocytes with absent staining of Kupffer cells. Staining intensities of CL-L1 was pronounced in the centrilobular hepatocytes (**Figure 3**).

In the kidney, immunoreactivity for CL-K1 and CL-L1 was especially associated with the tubular system, with the most pronounced staining of the distal tubules (**Figure 3**), in comparison with proximal tubules. CL-L1 immunoreactivity was for some distal tubules distinctly associated with the brush border. Immunoreactivity for both collectins was also associated with the epithelial cells lining the Bowman's capsules, whereas immunoreactivity in the glomerulus itself mainly was associated with CL-K1 and only minimally with CL-L1. CL-K1 immunoreactivity in the glomerulus was associated morphologically appeared to include both podocytes and mesangial cells.

In the lung, CL-K1 immunoreactivity was associated with alveolar macrophages, type I and II pneumocytes (**Figure 3**). CL-L1 immunoreactivity appeared only to be associated with alveolar macrophages.

In the thyroid gland, cuboidal epithelial cells lining the base membrane of thyroid follicles and parafollicular cells (C-cells) were associated with immunoreactivity for both CL-K1 and CL-L1 (**Figure 4**). Most pronounced staining was observed for CL-K1.

In the pancreas CL-K1 and CL-L1 immunoreactivity was associated with the islets of Langerhans and the pancreatic epithelial acinar cells and ducts (**Figure 4**). Within the islets, the vast majority of cells stained positive, indicating for sure that insulin-producing cells (beta cells) were associated with immunoreactivity and also most likely glucagon-producing cells (alpha cells) as well.

In the adrenal tissue section (**Figure 4**), derived from a patient diagnosed with pheochromocytoma, the histology was slightly unclear. However, as immunoreactivity for both CL-K1 and CL-L1 was associated with nearly all cells, it was deducted that the majority of adrenal cells, including both medullary and cortical cells, are associated with the two collectins, similar with previous findings for the localization of CL-K1 (7).

In the gall bladder, immunoreactivity for both CL-K1 and CL-L1 was associated with columnar epithelial cells of the mucosal folds, with increasing intensity toward the luminal side of the folds (**Figure 5**). Various cell types in the lamina propria stained weakly positive. We observed only scattered staining of cells in the muscularis and serosa layers.

In the duodenum, immunoreactivity for CL-K1 and CL-L1 was associated with epithelial cells in both the mucosa and submucosa. In the mucosal luminal membrane of the villi, especially columnar cells (enterocytes) stained positive (**Figure 5**). Further and intense immunoreactivity of the mucosa was associated with the crypts of Lieberkuhn, whereas the muscularis externa was only weakly positive for staining. In the submucosa, immunoreactivity was associated with cells of the Brunner's glands.

In the colon, CL-K1 and CL-L1 immunoreactivity was dominantly associated with mucosa and especially with columnar epithelial cells in the crypts of Lieberkuhn (tubular glands) (**Figure 5**). In the lamina propria, the staining was scattered and associated with various cells types. Within the layers of the muscularis externa and submucosa, staining was also associated with endothelial cells, best illustrated for the localization of CL-K1.

In the testis, CL-K1 immunoreactivity was associated with germinal epithelial cells lining the tunica propria of the seminiferous tubules, spermatogonia (type A and B), and spermatocytes (primary and secondary) (**Figure 6**). These cells were embedded in the less immunoreactive Sertoli cells. In the interstitium between seminiferous tubules, Leydig cells and endothelial cells of capillaries stained weakly positive. CL-L1 immunoreactivity was weak in comparison with that of CL-K1, but the pattern of the two collectins followed each other.

In the prostate, CL-K1 and CL-L1 immunoreactivity was associated with epithelial cells of the prostatic glands, with staining of both acini and ducts (**Figure 6**). The staining was associated with both columnar pseudostratified and involuted luminal epithelial cells; however, with most intense staining of the basal epithelial cells. In the stroma, scattered staining was observed in various cells types, including staining of endothelial cells. In the ducts, secretory vesicles and concretized material stained weakly positive, especially for CL-K1.

In the corpus uterus, CL-K1 and CL-L1 immunoreactivity was localized to the epithelial cells in the endometrial glands, glandular ducts, and at luminal surface, with comparable staining of glandular structures in both the stratum functionalis (compactum and spongiosum) and stratum basale (**Figure 6**). Both stratified columnar and ciliated cells in the glands stained intensely. In the stroma, the immunoreactivity was moderate but associated with the majority of cells, with a pronounced staining of endothelial cells of the capillaries.

In the skin, CL-K1 and CL-L1 immunoreactivity was associated with the sweat glands and ducts (**Figure 7**). CL-K1 staining was further associated with the basal layer of the epidermis. In the sweat glands and ducts, especially epithelial cells, stained positive, while myoepithelial cells only stained weakly positive. Within the inner duct, staining was associated with the luminal part and eventual content in the duct. Staining of sporadically distributed and non-identifiable cells in the dermis was also observed.

In the partoid salivary glands, immunoreactivity for CL-L1 and CL-K1 was associated with both epithelial glandular (acini) and epithelial ductal cells (**Figure 7**). Immunoreactivity was localized dominantly to the epithelial cells constituting the salivary ducts and less with the secretory acini. However, the majority of serous-producing epithelial cells were associated with immunoreactivity. Basal epithelial cells of mucin producing stained weakly positive. All three kinds of ducts: intercalated (minor), intralobulated (striated), and major showed equal and dominant immunoreactivity. The immunoreactivity of CL-L1 in the salivary serous glands was superior to that of CL-K1.

In the full-term (mature) placenta, CL-K1 and CL-L1 immunoreactivity was mainly associated with the syncytiotrophoblast layer, at the border of maternal and fetal circulation, and weakly with the underlying cytotrophoblasts associated with a villus (**Figure 7**).

Various levels of CL-K1 and CL-L1 immunoreactivity were also found to be associated with the following tissues: the thymus, spleen, tonsil (Figure S1 in Supplementary Material), esophagus,

ciliated epithelial cells. Scale bars in large sections and in isotype control sections (C,F,G) correspond to 500 µm and in small sections to 50 µm.

stomach (Figure S2 in Supplementary Material), cervix uterus, portio uterus, abortus (Figure S3 in Supplementary Material), skeletal muscle, cerebellum, and urinary bladder (Figure S4 in Supplementary Material). A description of their detailed localizations is provided in the associated figure legends.

#### CL-K1 and CL-L1 mRNA Abundancies

To compare protein localization by IHC with site of synthesis for the two collectins, data were retrieved from the three major RNA expression databases HPA, GTEx, and FANTOM5 RNA. For comparison, expression levels were normalized and graduated into four categories based on a logarithmic division (Tables S1 and S2 in Supplementary Material). Levels of immunoreactivity were visually validated by three independent persons and categorized similarly (**Table 1**).

mRNA transcripts encoding CL-K1 was detectable in all tested tissues and the major sites of synthesis, grouped in the "high" category, comprised the adrenals, gallbladder, and liver. In all the tested tissues, there was only minimal variance, in terms of a single category shift, i.e., "high" to "medium," between CL-K1 mRNA levels and immunoreactivity, therefore we considered there to be an excellent accordance between site of CL-K1 synthesis and protein localization. mRNA transcript encoding CL-L1 was not detected in as many tissues as CL-K1. Some tissues were categorized with "extremely low/absent" number of CL-L1 transcripts. However, by IHC quite a lot of these tissues were found to be associated with immunoreactivity, albeit in a "low" degree. The major site of CL-L1 mRNA synthesis was the liver and placenta and in these tissues the protein was also readily detected. Similar to the observations for CL-K1 and using the same criteria, there appeared in general to be accordance between site of CL-L1 synthesis and protein localization (discussed further below).

To associate the protein localization with an eventual local functionality of the two collectins, mediated *via* the presence of MASPs, localization of MASPs (and MAps), and synthesis of their respective mRNAs were evaluated by the same approach. As the major RNA expression databases currently do not take alternatively splicing of the MASP genes into consideration, data were retrieved and gathered from the previous work by Thiel and colleagues and Garred and colleagues (30–33). MASP-3 expression appeared to both overlap and being as widely distributed as the two collectins, whereas the other MASPs and MAp had a restricted pattern of tissue localization, with the liver being a tissue of major synthesis and/or detection: an observation, which also applied for MBL.

stratified cuboidal epithelia cells; partoid salivary gland *d*: duct, *ed*: epithelial ductal cells, *sa*: serous acini, *ma*: mucous acini, *mj*: major duct, *il*: intralobulated duct, and *ic*: intercalated duct; placenta *v*: villus, st: syncytiotrophoblast, *ct*: cytotrophoblast, and *md*: mesoderm. Scale bars in large sections and in isotype control sections (C,F,G) correspond to 500 µm and in small sections to 50 µm.

# DISCUSSION

The present work describes the localization of CL-L1 and CL-K1 in human tissues as determined by immunohistochemistry and summarizes further their mRNA tissue profiles derived from transcriptome databases. Both CL-L1 and CL-K1 were demonstrated in epithelial cells in a variety of tissue throughout the human body.

Of all the tested MAbs, MAbs 16-13 (anti-CL-L1) and 11-1 (anti-CL-K1) had the best sensitivity and specificity. The two MAbs demonstrated excellent immunoreactivity in the three tissues, the liver, kidney, and adrenals, wherein human CL-K1 and CL-L1 localization previously have been demonstrated (6, 7). The major positive cell types comprised, hepatocytes, renal epithelial cells of tubules, and medullary and cortical cells of the adrenals. In addition to the adrenals, tissues from other endocrine glands, i.e., pancreas, demonstrated a similar convincing excellent immunoreactivity for both collectins, derived from cells in the islets of Langerhans and epithelial cells of the ducts. Exocrine tissues of the digestive system comprising the gallbladder, duodenum, colon, and also partly the stomach and esophagus were also associated with epithelial and mucosal immunoreactivity for generally both collectins. Other exocrine tissues, comprising sex-specific organs, such as the testis, prostate, and uterus, had also excellent to moderate immunoreactivity for both collectins. Among all the analyzed tissues, the testis and uterus appeared to be the two tissues with the relative highest CL-K1 immunoreactivity. Again, epithelial cells and mucosa in the uterus were the major source of immunoreactivity, whereas immunoreactivity in testis was associated with germinal epithelial cells of the tubules. Among the tissues analyzed, the highest CL-L1 immunoreactivity was observed in the liver, followed by the kidney and parotid gland, wherein immunoreactivity was also associated with epithelial cells. Our observation of CL-K1 synthesis in various tissues falls in line with previous work by Wakamiya and colleagues, who by immunofluorescence techniques demonstrated partly overlapping localization of CL-K1 in murine tissues, using a polyclonal anti-mouse-CL-K1 antibody (34). In general, we observed a stronger staining of CL-K1 than of CL-L1. This may reflect that CL-K1 is more abundant than CL-L1, although their levels in the circulation are approximately the same (12), or it may simply be a matter of affinities of the applied MAbs in combination with availability of antigen epitopes on the fixed and embedded tissue sections.

Retrieval and normalization of mRNA levels from three transcriptome databases demonstrated accordance between site Table 1 | Levels of RNA expression: The symbols "+++," "++," "+," and "– "denote high, medium, low, and absent/extremely low expression, respectively, based on the criteria established in Figures S1 and S2 in Supplementary Material.


*IHC immunoreactivity levels were judged arbitrarily into the same categories by three independent validations (persons). Missing symbols reflect that the tissue has not been* 

*analyzed. MASP-1, MASP-3, and Map44 derive all from the MASP1 gene. MASP-2 and MAp19 derive both from the MASP2 gene.*

*a Relative and normalized data from HPA, GTEx, and FANTOM5 RNA expression databases (Tables S1–S3 in Supplementary Material).*

*bRelative and normalized data from Seyfarth et al. (30) and Degn et al. (31) (Tables S4 and S5 in Supplementary Material).*

*c Skjoedt et al. (32).*

*dRelative and normalized data from Degn et al. (31) and Skjoedt et al. (32) (Table S6 in Supplementary Material).*

*e Degn et al. (33) (Table S7 in Supplementary Material).*

*f Relative and normalized data from Seyfarth et al. (30) and Degn et al. (33) (Table S8 in Supplementary Material).*

*g The Human Protein Atlas Project (antibody CAB016782).*

of synthesis and protein localization. The only tissue, wherein there appeared to be a notable difference, was for CL-L1 in the salivary gland. By IHC CL-L1 localization was judged to be medium, whereas CL-L1 mRNA appeared to be absent. Other CL-L1-specific antibodies showed varying staining of particular the salivary gland (data not shown), making us hypothesize that the observed disagreement could reflect some sort of uncharacterized alternative splicing of CL-L1 in this tissue.

Throughout the IHC staining it was evident that within the majority of the tissues, the localization pattern of two collectins was identical; meaning that exactly the same cells in a given tissue demonstrated immunoreactivity for both collectins. This is best exemplified when two neighbor sections were mounted and processed, as illustrated, e.g., with the corpus uterus (**Figure 6**). Thus, in the majority of tissues there is opportunity for the making of CL-LK heteromeric complexes, and as previously described, this structure also appears to be the most thermodynamic stabile conformation (11). In some of the tissues, comprising the thyroid gland, skeletal muscle, skin, urinary bladder, and partly the testis and esophagus, CL-K1 appeared to be present in large excess in comparison with CL-L1, as judged by immunoreactivity; it is likely that CL-K1 homomers will be the dominating form in these tissues. The precise distribution of homomers vs. heteromers in different tissues should not be judged by immunoreactivity and remains thus to be characterized in detail. Although the heteromers are eminent in their association with MASP-2 and C4b deposition, in comparison with the homomers, it is worth emphasizing that both types of homomers interact well with MASP-1/-3, and may mediate downstream complement activation *via* those alone (8, 11).

To further illustrate the presence of heteromers in different tissues we tried to establish a proximity ligation assay using antibodies usable on formalin fixed sections but without convincing results. By using purified and fixed CL-LK it appeared that even the best combination of antibodies partly shadowed for each other in proximity ligation assays and were only capable of detecting the very high oligomers of CL-LK, with a sensitivity of only 0.2 µg of purified CL-LK per ml immobilized onto polylysine-treated object glasses (data not shown).

The overlapping localization of the two collectins in the same cells justifies, with a few exceptions, possible assembly and presence of the heteromeric CL-LK in most tissues. Based on a combination of previous observations and unpublished results by our laboratory, it appears that the oligomeric state of CL-LK depends on the relative content of CL-L1 and the ratio of the CL-K1a/d isotypes (11). As all of our anti-CL-K1 MAbs recognize the two isotypes equally well, the immunohistochemical results does not *per se* allow us to deduct any final conclusions on the variability of oligomers in different tissues. However, tissues with a relative large expression of CL-L1 could potentially favor assembly of CL-LK into large oligomers, ranging from 2 to 6 subunits.

We have previously demonstrated that the binding activity of CL-K1 and CL-LK, and hence also their complement activating ability, in serum/plasma is inhibited by unknown factors (11). This has made it difficult to comprehend the role of the two collectins as *bona fida* activators of complement. In the light of the widespread presence of CL-K1 and CL-L1 in various tissues, it is possible that binding activity in the periphery, in the absence of inhibitory blood components, may be more efficient.

As the hitherto described biological functions of CL-K1 and CL-L1, in terms of complement activation or involvement in embryogenesis, appear to rely on MASPs it is relevant to investigate co-localization with MASPs in the periphery. However, there is a lack of suitable antibodies specific for the three products of the MASP-1 gene, MASP-1/-3, and MAp44, but the summarized mRNA profile presented in **Table 1** shows that MASP-3 synthesis, in contrast with all other MASPs and MAps, appears to overlap greatly with the localization of CL-K1 and CL-L1. Thus, it is likely that the role of CL-K1 and CL-L1 in the periphery is mainly mediated *via* MASP-3, which was recently shown to activate profactor D to factor D, and thereby potentiate the alternative pathway and amplification loop (35–37). Although (pro) factor D mainly is synthesized in adipose tissue (hence the alias "adipsin") various tissues synthesize minor amounts of profactor D, which could be a target for MASP-3 in complex with CL-K1/-L1/-LK, and thereby potentiate the complement amplification loop in the periphery, upon encounter and binding of collectins to suitable (microbial) ligands. Alternatively, the two collectins may in the periphery, and in parallel with MBL and C1q, exert some of their functions by interacting with the metalloproteases bone morphogenic protein 1 and tolloid-like proteases, involved in

# REFERENCES


extracellular matrix assembly and growth factor signaling (38). Interactions between CL-K1, -L1, or -LK with these metalloproteases remain to be investigated.

In the light of our (co-)localization of CL-K1 and CL-L1 to peripheral tissues it appears that the previously focus on their roles in the circulation, liver, and kidney, may have disparaged a compelling localization in especially exocrine/endocrine tissues and mucosa, suggestive of that CL-L1, CL-K1, and CL-LK may play roles on epithelial surfaces in general and in tissue characterized by a high degree of exocytosis. The localization of CL-K1 and CL-L1 reminds also in many ways of the localization of the collectin surfactant protein D (39).

# AUTHOR CONTRIBUTIONS

SH and MH designed the study and carried out: antibody development, characterization, immunohistochemistry, data analysis, computational bioinformatics, and wrote the paper, on which all authors commented. JA and KB carried out antibody development, characterization, and immunohistochemistry. EH, ON, and HS participated in designing and performing the immunohistochemistry and analyzing data. KS carried out development of antibodies. AS and JG carried out SPR analysis and analyzed data.

# ACKNOWLEDGMENTS

The authors thank Anette Holck Draborg, Department of Cancer and Inflammation Research, University of Southern Denmark for critical reading of the manuscript and for giving valuable comments. The authors thank Lisbeth Mortensen, Department of Pathology, Odense University Hospital for technical advice relating to immunohistochemistry.

# FUNDING

This work was in part supported by the A. P. Møller Foundation, Augustinus Foundation, and Dagmar Marshall Foundation.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01757/ full#supplementary-material.


microbial-binding activity. *J Immunol* (2010) 185(10):6096–104. doi:10.4049/ jimmunol.1002185


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Hansen, Aagaard, Bjerrum, Hejbøl, Nielsen, Schrøder, Skjoedt, Sørensen, Graversen and Henriksen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Reference Intervals of Factor H and Factor H-Related Proteins in Healthy Children

*Anna E. van Beek 1,2, Angela Kamp1 , Simone Kruithof <sup>1</sup> , Ed J. Nieuwenhuys <sup>3</sup> , Diana Wouters1†, Ilse Jongerius1 , Theo Rispens1 , Taco W. Kuijpers 2,4 and Kyra A. Gelderman3 \**

*1Department of Immunopathology, Sanquin Research and Landsteiner Laboratory of the Academic Medical Centre, University of Amsterdam, Amsterdam, Netherlands, 2Department of Pediatric Hematology, Immunology and Infectious Diseases, Emma Children's Hospital, Academic Medical Centre, Amsterdam, Netherlands, 3Sanquin Diagnostic Services, Amsterdam, Netherlands, 4Department of Blood Cell Research, Sanquin Research and Landsteiner Laboratory of the Academic Medical Centre, University of Amsterdam, Amsterdam, Netherlands*

#### *Edited by:*

*Thomas Vorup-Jensen, Aarhus University, Denmark*

#### *Reviewed by:*

*Michael Kirschfink, Universität Heidelberg, Germany Gunnar Houen, State Serum Institute (SSI), Denmark*

> *\*Correspondence: Kyra A. Gelderman k.gelderman@sanquin.nl*

#### *†Present address:*

*Diana Wouters, Centre for Infectious Disease Control, National Institute for Public Health and the Environment, Bilthoven, Netherlands*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 28 May 2018 Accepted: 12 July 2018 Published: 02 August 2018*

#### *Citation:*

*van Beek AE, Kamp A, Kruithof S, Nieuwenhuys EJ, Wouters D, Jongerius I, Rispens T, Kuijpers TW and Gelderman KA (2018) Reference Intervals of Factor H and Factor H-Related Proteins in Healthy Children. Front. Immunol. 9:1727. doi: 10.3389/fimmu.2018.01727*

Complement is activated as part of the innate immune defense against invading pathogens. Also, it helps to remove apoptotic debris and immune complexes from the circulation. Impaired complement function due to aberrant plasma levels of complement proteins may be indicative for complement-mediated diseases or can be involved in susceptibility for infections. To determine whether plasma levels are abnormal, reference intervals (RIs) are used from adult healthy donors. Since many complement-mediated diseases have an onset during childhood, it is important to know whether these RIs can be extrapolated to children. RIs of Factor H (FH), the crucial fluid-phase regulator, and the FH-related proteins (FHRs), its homologous counterparts, are unknown in healthy children. While FH is measured to diagnose and monitor therapy of patients with atypical hemolytic uremic syndrome, recent studies also implicated increased plasma levels of FHRs in disease. Here, we investigated the levels of FH and FHRs in healthy children using recently developed specific ELISAs. We found that levels of FH, FHR-2, and FHR-3 were equal to those found in healthy adults. Levels of FHR-4A and FHR-5 were lower in children than in adults. However, only the FHR-5 levels associated with age. The RIs of these FH family proteins now serve to support the interpretation of plasma levels in prospective and retrospective studies that can be used for routine diagnostic and monitoring purposes including pediatric patient samples.

Keywords: normal ranges, complement, complement factor H, factor H-related proteins, pediatrics, diagnostics, reference intervals

# INTRODUCTION

Complement is part of innate immunity, comprising a powerful cascade of proteins able to eradicate invading pathogens and is important for removal of apoptotic debris and immune complexes from the circulation. Complement activation is tightly controlled and regulator proteins make sure that bystander damage to healthy host cells is kept to a minimum. Within the population, there is variation in the expression levels of these proteins and other complement components, leading to different steady-state complement activities in healthy individuals (1). Assessment of abnormal circulating levels can help to diagnose complement-mediated diseases such as atypical hemolytic uremic syndrome (aHUS) and differences in expression levels can help in understanding the susceptibility for infectious diseases as described in retrospective studies (2).

To discriminate between normal and abnormal levels, and to interpret retrospective studies, clinical laboratory reference intervals (RIs) are needed. As many complement-mediated diseases can have their onset during childhood, it is important to know whether adult levels can be extrapolated toward pediatric patients. For proteins such as C3 and C4, it has been determined that the normal ranges can be different in childhood compared to adults and between different ethnicities, and as such, adjusted RIs may be used (3–6). No pediatric RIs are known of Factor H (FH) and the FH-related proteins (FHRs), of which their plasma levels associate with various diseases.

Factor H is a crucial regulator of the alternative complement pathway and protects human host cells from unwanted complement activation. Genetic variants in complement regulator FH are associated with multiple diseases. Such variants can either alter protein functionality or induce variation in levels of expression. Many have been described to associate with aHUS or age-related macular degeneration, affecting the regulating function of FH (7, 8). However, some genetic variants result in lower (insufficient) circulating levels of FH (9–11). Differences in steady-state FH protein levels are associated with susceptibility for meningococcal disease and have recently been implicated as a marker of cardiovascular risk in chronic Chagas disease (12, 13). In general, low expression of complement regulators, such as FH, would make an individual more prone for chronic inflammation but more protected against infectious diseases, while high expression rather associates with risk of infectious diseases but less chronic inflammation (14).

Apart from FH, the FH protein family also includes the short splice variant of FH, FH-like-1 (FHL-1), and the FH-related (FHR) proteins, named FHR-1, FHR-2, FHR-3, FHR-4, and FHR-5, all of which are encoded by their own gene. FHR-4A and FHR-4B are the two splice variants of *CFHR4*, but FHR-4A is the only circulating variant found in human serum (15). While FHRs share homology with FH in its surface binding domains, they lack domains similar to SCR1-4 in FH and FHL-1, and for that reason are believed to have no complement-regulatory activity (16). Although limited data are available on the *in vivo* function of FHRs, many have shown associations of complement-mediated diseases with these FHR genes due to their copy number variations (17, 18), internal duplications (19–21), fusion proteins (22–26), or polymorphisms (27–30).

In addition, recent developments in the determination of circulating FHR levels in adults have led to the discovery of new associations with disease. FHR-1 levels were shown to be increased during IgA nephropathy (31, 32), although the authors report much higher levels than we and others have published (33, 34). FHR-3 levels were shown to be elevated during sepsis (35) and in systemic lupus erythematosus, rheumatoid arthritis, and polymyalgia rheumatica (36). Although Schäfer et al. did not find increased FHR-3 levels in aHUS patients, a recent study demonstrated increased levels in a larger, well-characterized cohort (36, 37). FHR-2 and FHR-4A levels have, so far, not been studied except in healthy donors, although FHR-2 and FHR-4A are implicated in the acute phase of bacterial infections (van Beek et al., manuscript in preparation) (15, 33, 38). FHR-5 levels were shown to be decreased in patients with C3 glomerulonephropathy (C3G) (39) and was recently identified as an independent risk factor for IgA nephropathy (32, 40). In summary, assessment of FHR protein levels contributes to the understanding of various diseases.

To investigate whether different RIs should be used for FH and the FHRs in children, we assessed the circulating levels in a cohort of healthy Dutch children and adolescents (all referred to as children), covering various age categories. These RIs now serve to support the interpretation of plasma levels in retrospective studies that include children. Moreover, they can be used for routine diagnostic and monitoring purposes in pediatric patient samples.

#### MATERIALS AND METHODS

#### Samples

Serum samples were obtained from anonymous, healthy children from a previous study, in accordance with Dutch regulations and approved by the Sanquin Ethical Advisory Board in accordance with the Declaration of Helsinki (41). Samples from adult healthy donors (*n* = 124 for FH and FHR-3, *n* = 120 for FHR-1, 2, 4A, and 5) were collected and measured during previous studies (15, 33, 35).

#### ELISAs

All ELISAs were performed as previously described for an adult healthy donor cohort (15, 33, 35). Briefly, the FH ELISA uses anti-FH.16, a monospecific mAb directed against SCR16-17, as a coat and goat anti-human-FH antiserum as detection. FHR-1/1 homodimers were measured using anti-FH.02 (directed against SCR20 of FH and cross-reactive to SCR5 of FHR-1) both as catching and detecting mAb. FHR-1/2 homodimers were also caught by anti-FH.02, but detected with a commercially available anti-FHR-2 (R&D Systems). FHR-2/2 homodimer levels, as well as total levels of FHR-1 and FHR-2 were calculated based on the observed levels of FHR-1/1 and FHR-1/2 dimers. The FHR-3 ELISA uses anti-FHR-3.1 (cross-reactive to FHR-4A) as a coating mAb and anti-FHR-3.4 (cross-reactive to FH) as a detecting mAb. FHR-4A was measured by catching with the monospecific mAb anti-FHR-4A.04 and detecting with rabbit anti-FHR-3 antiserum. FHR-5 homodimers were measured using two monospecific mAbs, anti-FHR-5.1 and anti-FHR-5.4. Two control sera were included in each plate to ensure limited inter assay variation.

#### Statistics

GraphPad Prism software v7 was used to analyze data and perform statistics (GraphPad Software, La Jolla, CA, USA). Significant differences were assessed by unpaired *t*-test. Correlations were assessed with a parametric Pearson's correlation test.

# RESULTS

With this study, we obtained more insight in the normal ranges of FH and the FHRs in children. For this, we used a cohort of 110 healthy children, of which 53% were females (**Table 1**). The subjects were evenly distributed across the age categories, aged 7 months up to 251 months (20.9 years) (41). We compared the levels in children to the levels that we previously found in adult healthy Dutch donors (15, 33, 35).


We investigated the plasma levels of FH and the FHRs in these healthy children (**Figure 1**). We observed that the levels of FH and FHR-3 were similar between the two genders and independent of age (**Figures 1A,G**). Indeed, the levels are equal to those previously found in adult healthy Dutch donors (**Table 2**) (35).

Next, FHR-1 and FHR-2 were assessed using dimer-specific ELISAs (**Figures 1B,C**). The levels of FHR-1/1 homodimers were independent of age and gender, although we did find a minor, but significant, difference when comparing the FHR-1/1 levels to adults (**Table 2**, difference between means = 1.2 µg/mL) (33). FHR-1/2 heterodimers and FHR-2/2 homodimers were also found to be independent of age and gender but were similar to the adult healthy donors (**Figure 1D**; **Table 2**). This implied that only the FHR-1 plasma levels differed from the adults. Indeed, when we calculated the concentrations of total FHR-1 and FHR-2 monomers, only FHR-1 levels were significantly lower [difference between means = 1.3 µg/mL (33 nM)] than the healthy adults (**Figures 1E,F**; **Table 2**).


#### Table 2 | Factor H (FH) family normal ranges characteristics.

*a Values of FHR-2/2 homodimers, and total levels of FHR-1 and FHR-2 monomers are calculated based on measured levels of FHR-1/1 homodimers and FHR-1/2 heterodimers. bDonors lacking FHR-1, FHR-2 (in the adult donor cohort) or FHR-3 were excluded from correlations and unpaired t-tests*

We recently demonstrated that FHR-4A is the only circulating form of FHR-4 and that no FHR-4B could be observed in serum (15). Therefore, we measured only FHR-4A in the children and found that FHR-4A levels were lower than expected based on levels found in adult healthy donors (**Figure 1H**; **Table 2**). It would, therefore, be expected that the levels showed an association with age. Surprisingly, the FHR-4A levels did neither show an association with age, nor with gender, in the children.

Last, we assessed the levels of FHR-5/5 homodimers. Similar to the other FHR proteins, FHR-5 levels were independent of gender. However, the levels did increase with age (**Figure 1I**, **Table 2**), being approximately 0.5 µg/mL lower in the youngest children than in the oldest children. While the younger children indeed showed significantly lower levels, the older children presented with levels equal to the adult healthy donors.

As the *CFHR* genes originated as part of segmental duplications of the *CFH* gene, it would be possible that protein expression is similarly regulated (42). Therefore, we investigated whether FH plasma levels associated with plasma levels of the FHRs. We saw an association between FH and FHR-1/1 homodimer levels in adult donors, when they carry two copies of *CFHR1* (*r* = 0.62, *P* < 0.0001), in contrast to those who carry only 1 copy of *CFHR1* (*r* = 0.09, *P* = 0.67) (33, 35). Children who most likely carry two copies of *CFHR1* [expressing > 10.1 µg/ mL FHR-1/1 homodimers, as determined by ROC analysis (area under the curve = 0.97)] showed a similar association (*r* = 0.49, *P* < 0.0001) (33). No association between FH and other FHR levels was noted. As a general conclusion, we observed no remarkable differences compared to adult circulating levels of FH family proteins.

#### DISCUSSION

We have determined RIs for FH and FHR-1 to 5 in Dutch healthy children. We were able to interpret the circulating levels of these FH family proteins in relation to adult healthy donors, which we have previously assessed (15, 33, 35). We found differences in some but not all of these proteins in the healthy children when compared with adults.

In contrast to FHR-1, FHR-4A, and FHR-5, no remarkable observations were made when analyzing the circulating levels of FH, FHR-2, and FHR-3. The three proteins were independent of age and gender, confirming a previous study on FH in Brazilian children (43). FH levels were previously found to be low in neonates, suggesting that plasma levels reach adult ranges within the first 6 months after birth (44, 45). Unfortunately, no sera were available from children below the age of 6 months. Future studies should test cord blood and plasma of neonates for the presence of FHRs at birth and early infancy to investigate these protein levels in more detail.

FHR-1 levels were independent of age and gender. We did observe lower FHR-1 levels than previously seen in adults, although the biological relevance may be disputed. FHR-3 levels were also trending toward significance, indicating that a minor difference in the copy number variation in *CFHR3/ CFHR1* between the two cohorts might be affecting the results (33, 35).

We found lower FHR-4A levels in children than in adults, even though FHR-4A did not associate with age of the children. Our group demonstrated previously that FHR-4A is stable up to at least 10 freeze-thaw cycles (15). However, we cannot exclude the possibility that long-term storage of these samples may have suffered from breakdown of FHR-4A when kept at −30°C (15). New studies on more recent samples are needed to confirm or disprove this possible explanation.

For FHR-5, we observed an increase with age, indicating that normal ranges for FHR-5 are low in the youngest children and that RIs may need to be adjusted accordingly. As FHR-5 levels positively associated with severity of IgA nephropathy in adults (32), and as IgA nephropathy is the main nephropathy in children (46), measurements of FHR-5 in a pediatric cohort will be highly informative to further study the role of FHR-5 in this nephropathy.

This study represents the most complete assessment of FH family proteins to date in a cohort of healthy children providing RIs. These RIs can now be used to interpret serum levels in prospective and retrospective studies that include children and used for routine diagnostic and monitoring purposes in pediatric patient samples. Ideally, each laboratory should adapt these RIs for their own assays.

### ETHICS STATEMENT

Serum samples were obtained from anonymous, healthy children from a previous study, in accordance with Dutch regulations and approved by the Sanquin Ethical Advisory Board in accordance with the Declaration of Helsinki.

#### REFERENCES


# AUTHOR CONTRIBUTIONS

AB, DW, TK, and KG designed research. AB, AK, SK, and EN performed research. AB, IJ, TR, TK, and KG analyzed data and wrote the paper. All authors critically reviewed the manuscript, gave final approval of the version to be published, and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

### FUNDING

Research leading to these results has received funding from the European Union's seventh Framework program under EC-GA no. 279185 (EUCLIDS; www.euclids-project.eu). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


the complement factor H-related 5 gene. *J Hum Genet* (2012) 57:459–64. doi:10.1038/jhg.2012.57


via its interaction with C3b protein. *J Biol Chem* (2012) 287:19528–36. doi:10.1074/jbc.M112.364471


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 van Beek, Kamp, Kruithof, Nieuwenhuys, Wouters, Jongerius, Rispens, Kuijpers and Gelderman. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Richard B. Pouw1,2\*, Mieke C. Brouwer1 , Anna E. van Beek1,2, Mihály Józsi3 , Diana Wouters1† and Taco W. Kuijpers2,4†*

*1Department of Immunopathology, Sanquin Research and Landsteiner Laboratory of the Academic Medical Center, University of Amsterdam, Amsterdam, Netherlands, 2Department of Pediatric Hematology, Immunology and Infectious Diseases, Emma Children's hospital, Academic Medical Center, Amsterdam, Netherlands, 3MTA-ELTE "Lendület" Complement Research Group, Department of Immunology, ELTE Eötvös Loránd University, Budapest, Hungary, 4Department of Blood Cell Research, Sanquin Research and Landsteiner Laboratory of the Academic Medical Center, University of Amsterdam, Amsterdam, Netherlands*

#### *Edited by:*

*Maciej Cedzynski, Institute for Medical Biology (PAN), Poland*

#### *Reviewed by:*

*Paul Nigel Barlow, University of Edinburgh, United Kingdom Robert Braidwood Sim, University of Oxford, United Kingdom*

> *\*Correspondence: Richard B. Pouw r.pouw@sanquin.nl*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 17 January 2018 Accepted: 23 March 2018 Published: 17 April 2018*

#### *Citation:*

*Pouw RB, Brouwer MC, van Beek AE, Józsi M, Wouters D and Kuijpers TW (2018) Complement Factor H-Related Protein 4A Is the Dominant Circulating Splice Variant of CFHR4. Front. Immunol. 9:729. doi: 10.3389/fimmu.2018.00729*

Recent research has elucidated circulating levels of almost all factor H-related (FHR) proteins. Some of these proteins are hypothesized to act as antagonists of the important complement regulator factor H (FH), fine-tuning complement regulation on human surfaces. For the *CFHR4* splice variants FHR-4A and FHR-4B, the individual circulating levels are unknown, with only total levels being described. Specific reagents for FHR-4A or FHR-4B are lacking due to the fact that the unique domains in FHR-4A show high sequence similarity with FHR-4B, making it challenging to distinguish them. We developed an assay that specifically measures FHR-4A using novel, well-characterized monoclonal antibodies (mAbs) that target unique domains in FHR-4A only. Using various FHR-4A/FHR-4B-specific mAbs, no FHR-4B was identified in any of the serum samples tested. The results demonstrate that FHR-4A is the dominant splice variant of *CFHR4* in the circulation, while casting doubt on the presence of FHR-4B. FHR-4A levels (avg. 2.55 ± 1.46 µg/mL) were within the range of most of the previously reported levels for all other FHRs. FHR-4A was found to be highly variable among the population, suggesting a strong genetic regulation. These results shed light on the physiological relevance of the previously proposed role of FHR-4A and FHR-4B as antagonists of FH in the circulation.

Keywords: factor H-related-4A, factor H-related-4B, factor H, factor H-related proteins, CFHR4 gene, the complement system

#### INTRODUCTION

The complement system is an evolutionarily ancient protein cascade which, through a series of events, recognizes, attacks, and kills foreign cells like bacteria, but can also target host cells [reviewed by Ricklin et al. (1)]. In order to prevent damage of healthy host cells, humans possess several complement regulators. Some are membrane bound and expressed on the cell surface; however, one of the most important regulators, complement factor H (FH) circulates freely in plasma. FH is a 155 kDa glycoprotein that circulates in blood with a reported average concentration ranging from 233 up to 400 µg/mL or 1.5–2.6 µM (2–6). FH is a crucial regulator of the alternative activation pathway of the complement system, orchestrating complement activation toward foreign cells by specifically binding to and inhibiting complement on human cells. FH belongs to the FH protein family, which consists of eight proteins derived from six genes, encoded in tandem in the *CFH* locus. The *CFH* gene itself encodes two proteins, FH and its splice-variant FH-like 1 (FHL-1). Next to FH and FHL-1, six factor H-related (FHR) proteins are found in plasma. Like FH and FHL-1, the FHRs are completely comprised of domains called short consensus repeat (SCR) domains. FHR-1, FHR-2, FHR-3, and FHR-5 are encoded separately by corresponding genes (*CFHR1, CFHR2, CFHR3,* and *CFHR5*), while FHR-4A and FHR-4B are splice variants encoded by *CFHR4* (7, 8). FHR-4A is the largest FHR (86 kDa) and comprises nine SCR domains, while FHR-4B has been predicted to consist of five SCR domains (43 kDa) (7, 8). The five SCR domains of FHR-4B are completely identical to SCR 1 and SCRs 6–9 of FHR-4A (**Figure 1B**). The unique four SCR domains between SCRs 1 and 6 of FHR-4A appear to have arisen from an internal duplication in *CFHR4* and are highly similar to the other SCR domains in FHR-4A and, therefore, also to FHR-4B (8). SCR 1 is highly similar to SCR 5 (85%), 2–6 (90%), 3–7 (93%), and SCR 4 to SCR 8 (87%) in FHR-4A. This makes it challenging to specifically distinguish FHR-4A from FHR-4B in immunoassays.

It is still unclear what the role of FHR-4A and FHR-4B is within the complement system. Increasing evidence indicates that FHRs act as antagonists of FH, competing with FH for the binding to complement C3b and human cell surfaces [reviewed by Józsi et al. (9)]. FHR-4A and -4B seem to lack physiologically relevant complement inhibitory activities on their own. FHR-4A has been reported to enhance the co-factor activity of FH at supraphysiological concentrations (10, 11). Furthermore, binding of

Figure 1 | Characterization of anti-FHR-4A monoclonal antibodies (mAbs). (A) Cross-reactivity of the anti-FHR-4A mAbs to biotinylated rhFHR proteins and biotinylated plasma-derived FH was determined by ELISA. (B) Schematic representation of FHR-4A, FHR-4B, and the recombinant fragments of FHR-4A used for epitope mapping, with the duplicated SCR domains in FHR-4A depicted in black. Corresponding percentage of sequence identity between the domains in FHR-4A is indicated. Domains 1 and 6–9 of FHR-4A are completely identical to the domains of FHR-4B (indicated by gray shading). (C) Epitope mapping of anti-FHR-4A antibodies using the fragments of rhFHR-4A as depicted in (B), determined by ELISA. As a control, polyclonal anti-FHR-3 (poly), which cross-reacts with all FHR proteins was used. (D) Schematic representation of the epitope location of each of the anti-FHR-4A mAbs. Single epitope mAbs are listed above FHR-4A, mAbs with multiple epitopes below. Note that most cross-reactive mAbs have two epitopes in FHR-4A, due to the high sequence similarity of the SCR domains. (E) Competition ELISA with monospecific mAbs (anti-FHR-4A.02 and 4A.04) and cross-reactive mAbs (anti-FHR-4A.08 and 4A.11). Binding of biotinylated rhFHR-4A is expressed as relative to the binding of biotinylated rhFHR-4A without any competing mAb. Bars represent mean of independent replicates with error bars indicating SD. All graphs are representative of multiple independent experiments.

FHR-4A and FHR-4B instead of FH allows complement activation to occur on the surface (11) and FHR-4A and FHR-4B might act as a competitor of FH. This form of regulation might serve to fine-tune complement inhibition by FH on surfaces where balanced complement activation is required for clearance, such as on necrotic and apoptotic cells. This hypothesis would also explain associations between genetic variation in *CFHR* genes and various diseases. For instance, the lack of *CFHR3* and *CFHR1* due to copy number variation (CNV) has been reported to be protective against age-related macular degeneration (12, 13). However, no such link with disease has yet been found for *CFHR4*. Like other FHRs, FHR-4A and FHR-4B bind to C3b and C-reactive protein, allowing complement activation to occur (11, 14, 15). Furthermore, FHR-4A has been reported to recruit CRP to necrotic cells *in vitro*, allowing more complement activation to occur, and is found to be accumulated in necrotic tissue (14). However, in contrast to all other members of the FH protein family, FHR-4A and FHR-4B seem to lack any affinity toward heparin (16, 17).

Highly relevant for the proposed antagonistic properties of FHR proteins is their physiological concentration, especially in relation to FH. Recently, we have reported normal protein concentrations for all FHRs in >100 healthy individuals, except for FHR-4A and FHR-4B. FHR-1/1 homodimers, FHR-2/2 homodimers, and FHR-1/2 heterodimers circulate at average concentrations of 14.6 (±3.0), 0.7 (±0.4), and 5.8 (±2.4) μg/mL, respectively, in healthy donors with two *CFHR1* copies, whereas FHR-5/5 homodimers circulate at 1.7 (±0.4) μg/mL in healthy donors (18). Of note, the FHR-1 total levels have recently also been reported to be 122 (±26) μg/mL, without distinguishing between homo- or heterodimers, nor fully describing the calibration used (19). However, a third, independent, group recently reported combined, total FHR-1, FHR-2, and FHR-5 levels, measured in an immunoassay, to be 10.7 (±5.4) μg/mL, thus supporting the lower levels for FHR-1 and FHR-2 (20). FHR-3 circulates at average concentrations of 0.7–1.1 µg/mL (6, 21). For FHR-4A and FHR-4B, only total FHR-4 levels ranging from 6.5 to 53.9 µg/mL with an average concentration of 25.4 µg/mL have been described, measured in only 11 healthy individuals (11). These levels were determined without distinguishing between FHR-4A and FHR-4B. Therefore, in order to specifically measure FHR-4A and FHR-4B, as well as establishing normal levels in a larger cohort of healthy individuals, we have developed specific reagents for FHR-4A and FHR-4B. Monoclonal antibodies (mAbs) were used to specifically detect FHR-4A, but we were unable to detect any freely circulating FHR-4B in human serum, despite the use of various cross-reactive mAbs. Finally, we demonstrate that FHR-4A levels vary greatly among healthy individuals.

#### MATERIALS AND METHODS

#### Samples

Healthy donor serum samples were collected as part of a previous study from anonymous, healthy volunteers with informed, written consent in accordance with Dutch regulations and this study was approved by the Sanquin Ethical Advisory Board in accordance with the Declaration of Helsinki (6). Normal human serum (NHS) pool comprises serum from 400 healthy donors. FHR-3-deficient serum pool comprises serum of four healthy donors previously genotyped by multiplex ligation-dependent probe amplification to carry no *CFHR3* gene copies (6). Samples of patients with confirmed bacterial infections were collected as part of the EUCLIDS project (van Beek et al., manuscript in preparation). CRP levels were determined in these samples as part of routine testing.

#### Proteins and Reagents

Rat anti-mouse kappa (RM-19) mAb was from Sanquin Reagents (Sanquin, Amsterdam, the Netherlands). High-performance ELISA buffer (HPE) was provided by Sanquin. Proteins were biotinylated according to the manufacturer's instructions using EZ-Link Sulfo-NHS-LC-Biotin, No-Weigh Format (Thermo Scientific, Waltham, MA, USA), when indicated. Polyclonal anti-FHR-3 antibodies and mAb clone anti-FHR-3.3 were obtained and characterized as part of a previous study (6). Recombinant human (rh) FHR proteins, containing a C-terminal 6×-histidine tag, were produced and purified as previously described (6). In short, proteins were expressed by transient transfection of pcDNA3.1 expression vectors in HEK293F cells, after which proteins were purified from the supernatant by Ni2+ affinity chromatography using HisTrap™ High Performance 1 mL columns (GE Healthcare Life Sciences, Freiburg, Germany). rhFHRs were filtered and concentrated using Amicon® Ultra Centrifugal Filter Devices (Merck Millipore, Darmstadt, Germany) according to the manufacturer's instructions using appropriate molecular weight cut-offs. For rhFHR-4A, a 100 kDa cut-off Amicon® Filter was used to further purify rhFHR-4A from any high molecular weight aggregates observed after Ni2<sup>+</sup> affinity chromatography purification (Figures S1A,B in Supplemental Material).

#### Immunization, mAb Generation, and Characterization

Anti-FHR-4A mAbs were generated, screened, and purified as previously described, using rhFHR-4A as immunogen (6). Isotypes of mAbs were determined by ELISA or with the use of the Mouse mAb Isotyping Kit (Hycult Biotech, Uden, the Netherlands) according to the manufacturer's instructions. Cross-reactivity against other FH protein family members was determined as previously described (6).

#### Epitope Mapping

The epitope location of each of the anti-FHR-4A mAbs was determined using rhFHR-4A, rhFHR-4B, and rhFHR-4A fragments consisting of SCR domains 1–3, 2–4, 4–9, 5–7, or 8–9 as previously described (15). Proteins and fragments were coated (4 µg/mL in PBS) onto Nunc Maxisorp 96-well microtiter plates (Invitrogen, Life Technologies, Carlsbad, CA, USA) by overnight (O/N) incubation at 4°C. Plates were washed three times with PBS + 0.02% (w/v) Tween-20 (PT) and blocked with 4% (w/v) BSA in PBS by incubation for 1 h. After another wash with PT, wells were incubated with the anti-FHR-4A mAbs diluted in PT for 1 h. Unbound anti-FHR-4A mAbs were washed away by washing the wells five times with PT, followed by incubating with 0.05% (v/v) HRP-conjugated rabbit anti-mouse IgG (Dako Agilent, Santa Clara, CA, USA) for 1 h and subsequently washing the plates five times. The ELISA was developed using 3,5,3′,5′-tetramethylbenzidine (TMB) solution and the reaction was stopped using 50 µL 2 M H2SO4. As a positive control, FHR-4A fragments were detected using polyclonal rabbit anti-FHR-4A, followed by HRP-conjugated goat anti-rabbit IgG (Dako Agilent). All ELISA steps were performed with a final volume of 50 µL per well and incubated at room temperature while shaking unless stated otherwise. Absorption was measured at 450 nm and corrected for background absorbance at 620 nm.

#### Competition ELISA

To determine whether the mAbs competed for the binding of rhFHR-4A, Nunc Maxisorp 96-well microtiter plates (Invitrogen) were coated with 100 µL of 2 µg/mL anti-FHR-4A mAbs, in PBS, by incubating O/N at room temperature. Next, the plates were washed with PT. Biotinylated rhFHR-4A (0.1 µg/mL, in HPE) was incubated with 10 µg/mL of each mAb for 20 min, followed by incubation on the washed plate for 1 h. Next, unbound biotinylated rhFHR-4A was washed away and the wells were incubated with 0.01% (v/v) strep-poly-HRP (Sanquin), diluted in HPE, for 20 min. After washing, the assay was developed by addition of 100 µL of 100 µg/mL TMB in 0.11 M sodium acetate containing 0.003% (v/v) H2O2, pH 5.5. Substrate conversion was stopped after approximately 10 min by addition of 100 µL 2 M H2SO4. Absorbance was measured at 450 nm and corrected for the absorbance at 540 nm with a Synergy 2 Multi-Mode plate reader (BioTek Instruments, Winooski, VT, USA). All ELISA steps were performed with a volume of 100 µL per well and incubated at room temperature while shaking unless stated otherwise.

#### Sucrose Gradient

Sucrose gradients were described previously (18). In short, NHS or FHR-3-deficient serum (150 µL of 50%, v/v, pooled serum, diluted in PBS) were loaded on 5–32.9% (w/v) sucrose (Merck 1.07654) gradients. Gradients were centrifuged for 20 h at 160,000 × *g* after which they were fractionated in 24 fractions of 500 µL. Proteins were immunoprecipitated and visualized on Western Blots as described below.

#### Immunoprecipitation (IP)

Factor H-related-4A and FHR-4B were immunoprecipitated from human healthy donor serum or sucrose gradient fractions using indicated mAbs. IP was performed by incubating 200 µL serum or 250 µL sucrose gradient fraction with 500 µL of 5 mg/mL CNBr-activated sepharose (GE Healthcare, Little Chalfont, UK) to which RM-19 was coupled (25 mg mAb per 1 g sepharose), and 50 µL of 100 µg/mL mAb, diluted in PBS supplemented with 0.1% Tween-20, 0.1% BSA, and 10 mM EDTA. Following overnight incubation at 4°C, while rotating, the sepharose was washed three times with 1 mL PT and two times with 1 mL PBS. Precipitated proteins were eluted by addition of 50 µL 1× NuPAGE Sample buffer solution (Invitrogen) and incubation at 70°C for 10 min. After spinning down the sepharose, SDS-PAGE under non-reducing conditions was performed using a Novex NuPAGE 10 or 4–12% Bis–Tris gel followed by Western Blot onto a nitrocellulose membrane (Novex iBlot Gel Transfer kit, Invitrogen). Membranes were blocked with 1% (v/v) Western Blocking Reagent (WBR) (Roche, Basel, Switzerland) in PBS for 30 min and incubated with 1 µg/mL biotinylated polyclonal rabbit anti-FHR-3 in PBS + 0.5% (v/v) WBR, O/N. After washing three times with PT, membranes were incubated with 0.1% (v/v) Strep-HRP in PBS + 0.5% (v/v) WBR. After 1 h, the membranes were washed three times with PT followed by two washes with PBS. Western Blots were developed with the Pierce ECL 2 Western Blotting substrate kit (Thermo Scientific) according to the manufacturer's instructions and analyzed using the ChemiDoc™ MP System (BioRad, Hercules, CA, USA).

#### FHR-4A ELISA

To measure FHR-4A in serum, RM-19 was coated (3 µg/mL in PBS) onto Nunc Maxisorp 96-well microtiter plates (Invitrogen) by O/N incubation at room temperature. After coating, plates were washed five times with PT, followed by incubation with 1 µg/mL anti-FHR-4A.04 in HPE, for 1 h. After washing, samples, diluted in HPE, were added and incubated for 1 h. Following washing with PT, 0.5 µg/mL biotinylated polyclonal anti-FHR-3 (in HPE) was incubated on the plate for 1 h. Next, unbound conjugate was washed away and the wells were incubated with 0.01% (v/v) strep-poly-HRP (Sanquin), diluted in HPE, for 20 min. After washing, the assay was developed as described above using TMB and measuring absorbance at 450 nm and correcting for the absorbance at 540 nm. All ELISA steps were performed with a volume of 100 µL per well and incubated at room temperature while shaking unless stated otherwise. For the calibration of the FHR-4A ELISA, highly pure rhFHR-4A was used (Figure S1B in Supplemental Material) of which the concentration was determined by measuring the absorbance at 280 nm and using an absorbance coefficient of 2.134 (0.1% w/v solution).

# Statistical Analysis

Analysis and statistical tests were performed using GraphPad Prism, version 7.03 (GraphPad Software, La Jolla, CA, USA).

# RESULTS

#### Characterization of mAbs Against FHR-4A

We characterized 13 mouse mAbs raised against rhFHR-4A. All mAbs were first tested for cross-reactivity against all other members of the FH protein family (**Figure 1A**). The mAbs were named in order of increasing cross-reactivity: anti-FHR-4A.01 to 4A.06 being mono-specific for rhFHR-4A, anti-FHR-4A.07, and 4A.08 recognizing both rhFHR-4A and -4B, anti-FHR-4A.09 to 4A.11 recognizing rhFHR-3, 4A, and -4B, anti-FHR-4A.12 binding rhFHR-2, -3, -4A, and -4B, whereas anti-FHR-4A.13 recognizes all rhFHRs except rhFHR-5. None of the anti-FHR-4 mAbs showed cross-reactivity with either FHR-5 or FH.

With the use of recombinant fragments comprising different FHR-4A domains (**Figure 1B**), the epitope location of almost all anti-FHR-4A mAbs were mapped. As expected, all FHR-4Aspecific mAbs (anti-FHR-4A.01 to 4A.06) bound to an epitope located in domains 2–5, which are unique for rhFHR-4A and not present in rhFHR-4B (**Figures 1C,D**). Of these FHR-4A specific mAbs, only anti-FHR-4A.04 recognized an epitope located in domain 5, whereas all other mAbs bound to an epitope in domain 2 or 3. Anti-FHR-4A.01, 4A.02, 4A.03, 4A.05, and 4A.06, competed with each other for the binding of rhFHR-4A, indicating identical or partially overlapping epitopes (data not shown).

Of the mAbs that cross-reacted with rhFHR-4B and other FHR protein family members, most appeared to bind to either of two epitopes in rhFHR-4A, reflecting the high degree of similarity between these SCR domains 2–5 in FHR-4A and the domains 1, 6, 7, and 8 in FHR-4A and FHR-4B.

Only anti-FHR-4A.09 appeared to have one epitope in domain 8 or 9, recognizing both rhFHR-4A and rhFHR-4B, while also cross-reacting with FHR-3.

The cross-reactive mAbs anti-FHR-4A.08 and anti-FHR-4A.11 were able to block binding of rhFHR-4A to the monospecific FHR-4A mAbs; however, in the reverse setting, this cross-blocking was not achieved (**Figure 1E**). This is in line with the presence of two epitopes for anti-FHR-4A.08 and 4A.11 in rhFHR-4A. It indicates that one of the two epitopes for anti-FHR-4A.08 and anti-FHR-4A.11 is partially overlapping or sterically hindering the binding site of the monospecific mAbs. The epitope location of anti-FHR-4A.13 could not be mapped due to a very low and ambiguous binding signal when the fragments were used. However, binding of rhFHR-4A to anti-FHR-4A.13 could be blocked with anti-FHR-4A.09, suggesting an epitope location in either domain 8 or 9 for anti-FHR-4A.13. An overview of the mAbs characteristics, the mapped epitope location, and the cross-reactivity is given in **Table 1**.

### Only FHR-4A Is Detected in NHS

Next, we tested whether the anti-FHR-4A mAbs capture FHR-4A and FHR-4B from NHS. To this end, an IP followed by Western Blot was performed, which was developed with cross-reactive biotinylated polyclonal anti-FHR-3. Of the specific anti-FHR-4A mAbs, only anti-FHR-4A.04 seemed to efficiently capture FHR-4A from NHS, resulting in a clear protein band corresponding to the molecular weight of FHR-4A (86 kDa) (**Figure 2A**). Only a very faint FHR-4A band was visible in the precipitates of anti-FHR-4A.01, 4A.02, 4A.03, 4A.05, and 4A.06. In addition, both anti-FHR-4A.07 and anti-FHR-4A.13 were neither able to efficiently immunoprecipitate FHR-4A from serum nor any of the other FHR proteins with which these mAbs cross-react. A band corresponding to FHR-4B (43 kDa) in the IP with

Table 1 | Anti-factor H-related (FHR)-4A monoclonal antibodies characterized in this study.


Figure 2 | Immunoprecipitation (IP) of plasma-derived factor H-related (FHR) proteins by the anti-FHR-4A monoclonal antibodies (mAbs). (A) Western blot following IP from pooled normal human serum using the indicated mAbs. For comparison, 100 ng rhFHR-4A and rhFHR-4B were loaded on the left side of the gel. Precipitated proteins are indicated with arrowheads on the right side of the blot. (B) As in (A), but with pooled serum of healthy donors who are deficient for *CFHR3* and *CFHR1* and using only the cross-reactive anti-FHR-4A mAbs for IP. (C) IP using polyclonal anti-FHR-3 and serum of two healthy donors with either two *CFHR3* (*CFHR3* suf.) or no *CFHR3* (*CFHR3* def.) gene copies. All Western blots were stained with cross-reactive biotinylated polyclonal anti-FHR-3. Precipitated proteins are indicated with arrowheads on the right side of each blot. Results are representative of multiple independent experiments.

anti-FHR-4A.08 also appeared to be missing, while the mAb was found to be cross-reactive with rhFHR-4B in the ELISA. A protein band corresponding to FHR-4B could not be distinguished in the IP of anti-FHR-4A.09 to anti-FHR-4A.12, as these mAbs also precipitated FHR-3 (43–50 kDa), which migrates at the same expected molecular weight and is also recognized by the polyclonal antibody. Therefore, to better visualize the IP of FHR-4B, an IP from a FHR-3-deficient serum pool was performed, using the cross-reactive mAbs. However, also in these conditions, no band corresponding to FHR-4B was detected in the IP of anti-FHR-4A.08 to anti-FHR-4A.12, which was unexpected given the cross-reactivity we observed with rhFHR-4B in ELISA (**Figure 2B**). On the other hand, anti-FHR-4A.12 did immunoprecipitate FHR-2 from NHS, confirming the cross-reactivity results obtained with rhFHR-2. As a control, the IP was repeated with the cross-reactive polyclonal anti-FHR-3 using serum of two healthy donors, one with two *CFHR3* gene copies, and one with no *CFHR3* gene copies. This resulted in the precipitation of FH, FHL-1, FHR-3, and FHR-4A (**Figure 2C**). However, also with this set-up, no FHR-4B was detected.

Serum fractionated by sucrose gradients was previously used to investigate the molecular size of FHR proteins in their native state (18). Using a recently characterized anti-FHR-3 mAb that cross-reacts with FH, rhFHR-4A, rhFHR-4B, and FHL-1 (clone anti-FHR-3.3) (6), we investigated the circulating molecular size of FHR-4A relative to FH and FHR-3, by IP of these proteins from sucrose gradient fractions. As expected, FH was precipitated from the fractions also containing IgG, while FHR-3 was mainly present in the fractions containing albumin, corresponding with their respective molecular weights (155 and 50 kDa) (**Figure 3A**). FHR-4A was found in the fractions corresponding with its molecular weight (86 kDa), between the main fractions containing FH and FHR-3. Again, no bands corresponding with FHR-4B were found in the sucrose gradient fractions of neither the FHR-3 sufficient nor the FHR-3-deficient serum pool (**Figure 3B**).

# FHR-4A Levels Vary Greatly Between Individuals

As we could not detect FHR-4B in human serum, we next focused on specifically measuring FHR-4A by ELISA. FHR-4A was measured using the monospecific mAb anti-FHR-4A.04, captured on immobilized RM-19, and using biotinylated polyclonal anti-FHR-3 antibody as a conjugate. The specificity

of the FHR-4A ELISA was confirmed using rhFHR proteins and plasma purified FH, giving a clear signal for rhFHR-4A (**Figure 4A**). Both rhFHR-3 and rhFHR-4B were detected at the highest concentrations we tested, but the FHR-4A ELISA was found to be 843-fold and 1,490-fold more sensitive for rhFHR-4A compared to rhFHR-4B or rhFHR-3, respectively. Next, using purified rhFHR-4A, we calibrated a NHS pool to contain 2.33 µg/ mL FHR-4A, which was subsequently used as a calibration curve in the FHR-4A ELISA (**Figure 4B**). FHR-4A levels in serum and plasma did not differ (**Figure 4C**, *P* > 0.05), nor were the levels affected by repeated freeze/thaw cycles (**Figure 4D**). Next, we measured the levels of FHR-4A in serum from 129 healthy donors. The average concentration of FHR-4A was found to be close to the concentration found in the NHS pool, being 2.55 µg/mL or 29.65 nM. Moreover, FHR-4A levels were found to be highly variable with a SD of 1.46 µg/mL and ranging from 0.26 up to 6.20 µg/mL (**Figure 4E**). We subsequently investigated whether FHR-4A behaves as an acute phase protein. To this end, we measured FHR-4A levels in 78 patients during an acute bacterial infection. No correlation was found for FHR-4A with CRP levels (*r* = 0.024, *P* = 0.834, **Figure 4F**). Of note, no total deficiency for FHR-4A was found in any healthy donor or patient measured during this study.

#### DISCUSSION

The FHR proteins are hypothesized to act as antagonists of complement regulator FH, competing for binding to ligands and allowing, instead of inhibiting, complement activation on surfaces (9). Several reports have shown that recombinant FHR-4A and -4B indeed bind to known FH ligands and allow complement activation to occur *in vitro* (11, 14, 15). To date, normal levels of FHR-1, FHR-2, FHR-3, and FHR-5 have been reported and were found to be circulating at much lower concentration as compared to FH (6, 18, 21, 22). In this report, we describe the circulating levels of FHR-4A in healthy individuals, using novel, well-characterized anti-FHR-4A mAbs. FHR-4B is apparently absent in blood, implying that FHR-4A is the dominant splice variant of *CFHR4*.

We obtained thirteen mAbs that were raised against rhFHR-4A, of which, six were found to be monospecific for FHR-4A. Their epitopes were mapped to be located within the four SCR domains unique for FHR-4A (and absent in FHR-4B). Of the seven cross-reactive mAbs that could also recognize the rhFHR-4B protein, six mAbs were found to bind to two epitopes within rhFHR-4A, corresponding to SCR domains that share high sequence similarity within rhFHR-4A because of the known internal gene duplication. The presence of two epitopes for

Figure 4 | Development of the factor H-related (FHR)-4A specific ELISA. (A) Representative result on the reactivity of the FHR-4A ELISA against rhFHR proteins and plasma-derived FH. Anti-FHR-4A.04 was used as monospecific catching mAb, and binding of antigen was detected using cross-reactive polyclonal anti-FHR-3. (B) Calibration of pooled normal human serum using rhFHR-4A (10 µg/mL) in the FHR-4A ELISA. (C) Comparison of FHR-4A levels in seven paired serum and EDTA plasma samples using the FHR-4A ELISA. Each point represents the mean of three independent measurements per sample. (D) Effect of multiple freeze/ thaw cycles. Each point represents the mean of three independent measurements per sample with error bars indicating SD. Dashed lines indicate 90–110% range. (E) Concentration of FHR-4A in 129 healthy donor sera, measured by ELISA. Each point represents the mean of three independent measurements per serum sample. Line indicates mean with SD. (F) Scatter plot of FHR-4A levels versus CRP levels determined in 78 patients with acute bacterial infection. Correlation was assessed using Spearman's correlation test.

these mAbs in rhFHR-4A was further supported by competition experiments. The rhFHR-4A-specific mAbs (one epitope) were not able to prevent binding of rhFHR-4A to cross-reactive mAbs (two epitopes), whereas the cross-reactive mAbs blocked binding to the specific mAbs, indicating that one of the two epitopes of the cross-reactive mAbs is overlapping with the epitope of the specific anti-FHR-4A mAbs.

Seven mAbs (anti-FHR-4A.01, -4A.02, -4A.03, -4A.05, -4A.06, -4A.07, and -4A.13) seemed unable or very inefficient in immunoprecipitating FHR-4A from NHS. The lack thereof is indicative for a much lower binding affinity for plasma-derived FHR-4A compared to rhFHR-4A. Strikingly, the epitopes of anti-FHR-4A.01, -4A.02, -4A.03, -4A.05, -4A.06, and -4A.07 are all located in SCR domains of FHR-4A that contain (multiple) putative *N*-linked glycosylation sites. While rhFHR-4A is produced by HEK293F cells and thus possesses glycans from human origin, it is possible that differences in the exact glycan composition from those found *in vivo* are involved in the apparent lack of binding affinity toward serum-derived FHR-4A. This might also be the case for anti-FHR-4A.13, for which we could not precisely map the epitope location.

We were unable to identify a band corresponding to FHR-4B in the IP by five different cross-reactive mAbs against rhFHR-4A and rhFHR-4B. These mAbs were able to precipitate FHR-4A from serum. Because anti-FHR-4A.08, -4A.10, -4A.11, and -4A.12 all recognized two epitopes in rhFHR-4A, it is possible that the lack of FHR-4B detection following IP was caused by a difference in binding avidity, favoring FHR-4A over FHR-4B. However, anti-FHR-4A.09, which also did not precipitate FHR-4B from human serum, only recognized one epitope in domain 8/9 or 4/5 of FHR-4A and FHR-4B, respectively. Thus far, only one report has suggested the presence of FHR-4B in human serum, being distinguished from FHR-3 by Western blotting following 2D electrophoresis separation (23). These blots were stained using cross-reactive polyclonal antibodies, making it unclear whether it is truly FHR-4B that was originally detected or not. Even if present, the levels of freely circulating FHR-4B must be extremely low compared to FHR-4A, since it was undetectable in any of our assays. Our current results are supported by previous reports, in which Western Blot analysis of human plasma also seems to indicate that FHR-4A is the dominant splice variant (8, 24). Hence, we developed an ELISA capable of specifically measuring FHR-4A and determined the normal average concentration of FHR-4A at 2.55 µg/mL (i.e., 29.65 nM). This is a 67.5-fold molar difference compared to the ~2 μM concentration of FH in human serum. The FHR-4A levels reported here are in line with previous reported levels determined by mass spectrometry (25) but stand in great contrast with the FHR-4 concentration of 25.4 µg/mL, as determined previously by an immunoassay (11). As discussed above, it seems unlikely that FHR-4B, which in theory was also measured in the previous report, can account for this difference. Both ELISAs catch FHR-4A with a monoclonal antibody while detecting it with a polyclonal antibody. Hence, it is more likely that there is a difference in the Sf9-derived rhFHR-4A previously used (14) and the HEK293F-derived rhFHR-4A used here for the calibration of the assays. Considering the recently reported normal values for all other FHR proteins (6, 18, 21, 22), ranging from 0.7 to 12 µg/mL (with FHR-1 being the most abundant FHR protein), it seems the values that we report for FHR-4A are likely to be correct.

We found considerable variation in FHR-4A concentration among healthy individuals, which is in line with previous observations by others (8, 17). This high degree of variation is also found in FHR-3 (6). However, the variation in FHR-3 levels can, to a major extent, be explained by the *CFHR3/CFHR1* deletion, which is relatively common with an allele frequency of about 20% in the Western population (12, 13, 26–30). The second known deletion in the *CFH-CFHR* locus, encompassing *CFHR1* and *CFHR4*, is far less common with an allele frequency of about 0.9% in the Western population (31). Thus, it is unlikely that CNV of *CFHR4* explains the variation seen in FHR-4A levels, suggesting that other as yet unknown genetic variations determine the concentrations of FHR-4A, which is independent of inflammation since FHR-4A did not behave as an early acute phase protein. Studies investigating the genetic variation(s) resulting in altered FHR-4A expression are currently ongoing.

In summary, we characterized novel anti-FHR-4A mAbs, which were employed to develop a highly specific FHR-4A ELISA. FHR-4A was found to be the dominant splice variant of *CFHR4* and circulates in healthy individuals at an average concentration of 2.55 (±1.46) μg/mL. Similar to the other FHR proteins, the circulating levels of FHR-4A are much lower as compared to FH (~0.03 versus ~2 μM, respectively).

#### ETHICS STATEMENT

Healthy donor serum samples were collected as part of a previous study from anonymous, healthy volunteers with informed, written consent in accordance with Dutch regulations, and this study was approved by the Sanquin Ethical Advisory Board in accordance with the Declaration of Helsinki.

# AUTHOR CONTRIBUTIONS

RP, DW, and TK designed the study. RP, MJ, DW, and TK designed experiments. RP, MB, and AvB performed experiments. All authors analyzed and discussed data. RP, DW, and TK wrote the first draft of the manuscript. All authors revised the data and contributed to the final version of the manuscript.

# ACKNOWLEDGMENTS

The authors would like to express thanks to the blood donors and patients for their contribution. Research leading to these results has received funding from the European Union's seventh Framework program under EC-GA no. 279185 (EUCLIDS; www. euclids-project.eu).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00729/ full#supplementary-material.

# REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Pouw, Brouwer, van Beek, Józsi, Wouters and Kuijpers. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

#### *Mischa P. Keizer 1,2\*, Angela Kamp1 , Gerard van Mierlo1 , Taco W. Kuijpers 2,3 and Diana Wouters <sup>1</sup>*

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Teizo Fujita, Fukushima Medical University, Japan Péter Gál, Institute of Enzymology (MTA), Hungary*

*\*Correspondence: Mischa P. Keizer m.keizer@sanquin.nl, smpd.keizer@gmail.com*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 24 December 2017 Accepted: 06 June 2018 Published: 27 June 2018*

#### *Citation:*

*Keizer MP, Kamp A, van Mierlo G, Kuijpers TW and Wouters D (2018) Substitution of Mannan-Binding Lectin (MBL)-Deficient Serum With Recombinant MBL Results in the Formation of New MBL/MBL-Associated Serine Protease Complexes. Front. Immunol. 9:1406. doi: 10.3389/fimmu.2018.01406*

*1Department of Immunopathology, Sanquin Blood Supply, Division Research and Landsteiner Laboratory of the Academic Medical Center, University of Amsterdam, Amsterdam, Netherlands, 2Department of Pediatric Hematology, Immunology and Infectious Diseases, Emma Children's Hospital, AMC, University of Amsterdam, Amsterdam, Netherlands, 3Department of Blood Cell Research, Sanquin Blood Supply, Division Research and Landsteiner Laboratory of the AMC, University of Amsterdam, Amsterdam, Netherlands*

The lectin pathway (LP) of complement activation depends on the activation of the MBL-associated serine proteases (MASPs) circulating in complex with mannan-binding lectin (MBL). MBL deficiency is the most common complement deficiency and has been associated with several pathological conditions. As we had previously shown, plasma-derived MBL (pdMBL) contains pre-activated MASPs that upon *in vivo* pdMBL substitution results in restoration of MBL concentrations but no LP functionality due to immediate inactivation of pdMBL–MASP complexes upon infusion. In this study, we analyzed MBL-sufficient and -deficient serum by size-exclusion chromatography for complexes of LP activation. In both sera, we identified non-bound free forms of MASP-2 and to lesser extent MASP-1/3. After addition of recombinant MBL (rMBL) to MBL-deficient serum, these free MASPs were much less abundantly present, which is highly suggestive for the formation of high-molecular complexes that could still become activated upon subsequent ligand binding as shown by a restoration of C4-deposition of MBL-deficient serum. Ficolin (FCN)-associated MASPs have been described to redistribute to ligand-bound MBL, hereby forming new MBL/MASP complexes. However, reconstitution of MBL-deficient serum with rMBL did not change the relative size of the FCN molecules suggestive for a limited redistribution in fluid phase of already formed complexes. Our findings demonstrate that rMBL can associate with free non-bound MASPs in fluid phase while preserving full restoration of LP functionality. In contrast to pdMBL products containing pre-activated MASPs which become inactivated almost immediately, these current data provide a rationale for substitution studies using rMBL instead.

Keywords: recombinant MBL, MBL-associated serine protease, redistribution, mannan-binding lectin–MBLassociated serine protease complexes, size-exclusion chromatography

# INTRODUCTION

The complement system is an intricate and subtle cascade comprising more than 50 soluble and cell-bound proteins to defend against a wide range of bacterial and fungal pathogens. The complement system has an array of different functions, which includes opsonization and lysis of pathogens, elimination of immune complexes, and stimulation and chemotaxis of leukocytes (1, 2).

In general, the complement system is divided into three activating pathways, which all converge into the so-called terminal pathway. The antibody-dependent classical pathway (CP) is activated upon the binding of C1q to immune complexes. Upon binding of mannan-binding lectin (MBL) (or other collectins such as CL-L1 or CL-K1), or one of the three ficolins (FCNs), to pathogen-specific carbohydrates, the MBL-associated serine proteases (MASPs) are activated resulting in activation of the lectin pathway (LP). The alternative pathway is spontaneously activated on surfaces that lack complement regulatory proteins and can act as an amplification loop for both the CP and the LP. All three activating pathways converge at the level of C3 and can subsequently activate the terminal pathway to form the membrane attack complex for complement-mediated lysis of target cells.

Several LP pattern-recognition molecules (PRMs) have been identified. Most important are MBL and the so-called FCNs (3). In serum, the basic unit of MBL (a trimer of 32 kDa polypeptide chain) oligomerizes to form higher-order molecules ranging from dimers to hexamers (650 kDa) and higher (4, 5). MBL circulates in complex with the inactive zymogens of MASPs (4). The different higher-order oligomeric forms of MBL may have a different composition of MASPs, possibly combining different MASPs in one complex (6, 7). There are two major MASPs: i.e., MASP-2 and MASP-1. The 74-kDa MASP-2 is auto-activated upon ligand binding of the PRMs, and the activation is greatly increased in the presence of 77-kDa MASP-1 (8, 9). MASP-3 is a splice variant of the same *MASP1/3* gene with catalytic activity and has an important role in the activation of factor D (10). Different MASPs can be present within a single oligomeric MBL–MASP complex (6, 8). The role of MASP-1 in the activation of MASP-2 suggests that a heterocomplex of these MASPs with MBL is required for LP activation (8, 9). Alternatively, smaller MBL–MASP complexes could bind in close vicinity to allow cross-activation of MASP-2 by MASP-1 (6).

Mannan-binding lectin deficiency is a common complement deficiency. Depending on the definition up to 20% of the human population is affected, and MBL deficiency has been associated with several diseases (11–13). MBL deficiency in different vulnerable patient groups, including pediatric oncology patients (12), neonates (14), and patients with cystic fibrosis (15), has shown an increased severity and frequency of infections. Earlier MBLsubstitution trials with plasma-derived MBL (pdMBL) in MBLdeficient individuals have shown to be safe, without any major adverse event reported (11). Although MBL levels were restored, LP functionality remained unexpectedly low following *in vivo* MBL substitution (16). This was found to be due to immediate interaction of inhibitors, such as C1-inhibitor (C1-INH), in the recipient blood with the pre-activated MBL–MASP complexes in the pdMBL product (17). Substitution with rMBL would restore the C4-deposition of MBL-deficient serum upon substitution (18), without the rapid inactivation seen in pdMBL. For this reason, there is renewed interest in the use of rMBL.

In this study, we analyzed the composition and size of LP-activating complexes in MBL-sufficient, MBL-deficient, and rMBL-reconstituted deficient serum in detail to determine whether rMBL associates with MASPs to act as effective LP-activating complexes upon MBL substitution.

#### MATERIALS AND METHODS

#### Reagents

Mouse monoclonal antibodies (mAbs) against MBL (e.g., αMBL-1; murine IgG1) were generated at our department, and the properties have been described before (19–21). mAbs against MASP-1/3 (4H2A9; murine IgG1), MASP-2 (8B5; murine IgG1), FCN-2 (GN4; murine IgG1), and FCN-3 (4H5; murine IgG1) were obtained from Hycult Biotech (Uden, the Netherlands). A second mAb against FCN-3 (334; murine IgG1) was obtained from Enzo Life Sciences (Bruxelles, Belgium). Affinity-purified recombinant human MBL [rMBL at 3.1 mg/ml in 10 mM Tris–HCl with 140 mM NaCl, pH 7.4, and 5 mM EDTA (TBS-E)] (22), recombinant human MASP-1 (rMASP-1 at 0.28 mg/ml), recombinant MASP-2 (rMASP-2) at 3.1 µg/ml (23), and recombinant MASP-3 (0.425 mg/ml), all expressed in HEK cells, and mAbs against MASP-1/3 (4H2A9) and MASP-3 (38:12.3) were a kind gift of Professor J. C. Jensenius (Aarhus University, Aarhus, Denmark). High-performance ELISA buffer (HPE) and poly-HRP-labeled streptavidin (poly-HRP) were obtained from Sanquin (Amsterdam, the Netherlands). 3,5,3′,5′-Tetramethylbenzidine (TMB) was obtained from Merck (Darmstadt, Germany). Mannan and acetylated BSA were obtained from Sigma-Aldrich (St. Louis, MO, USA). Serum from healthy volunteers was obtained with informed consent and prepared as described elsewhere (24). Ethical review and approval were obtained for this study in accordance with the local legislation and institutional requirements (Sanquin Research Medical Ethical Committee). MBL status was determined by genotype and MBL serum levels. All serum aliquots were stored at −80°C until tested.

#### Monoclonal Anti-MASP-2 Antibody (mAb 12D12) and MASP-Specific ELISAs

Monoclonal antibodies against pdMBL were obtained by a fusion of spleen cells from a mouse immunized with MBL purified from Cohn fraction III (20). We and others have described that during purification of MBL several other proteins are co-purified (17, 25). During selection of the mAbs, several mAbs appeared to be non-responsive to recombinant MBL (rMBL). These mAbs were further characterized for their reactivity toward MASPs. Briefly, 96-well Nunc Maxisorp microtiter plates were coated overnight with 0.5 µg/ml of different recombinant MASPs in PBS (MASP-1 and MASP-2). All following steps were conducted in HPE. The reactivity of biotinylated mAbs obtained from the original immunization, and control (biotinylated) antibodies (αMASP-1 4H2A9 and αMASP-3 38:12.3) were assessed by titration on the plate. After washing, plates were incubated with 0.01% poly-HRP and developed using 0.1 mg/ml TMB in 0.1 M sodium acetate containing 0.03% (v/v) H2O2, pH 5.5. Absorbance was measured at 450 nm. In total, 42 non-MBL-binding mAbs were tested for their specificity. mAb 12D12 detected only MASP-2, and not MASP-1 and MASP-3 (Figure S1A in Supplementary Material). As a control for coating the MASPs, we also developed an assay with mAbs against MASP-1/3 (Figure S1B in Supplementary Material).

# MBL Serum Levels and Genotyping

Mannan-binding lectin concentration was measured by ELISA as previously described by Brouwer et al. (16). Briefly, mannan was coated to the solid phase, and biotinylated αMBL-1 was used as detection mAb. After washing, plates were incubated with 0.01% poly-HRP and developed as described earlier. Samples were compared with the mean MBL serum level found in a pool of 3,000 healthy control sera (1.5 µg/ml MBL). MBL status was confirmed using genotyping, as described previously by Frakking et al. (26) In short, a TaqMan assay with specific primers and minor groove binding for each SNP were used, hereby directly amplifying, using both forward and reverse allele-specific primers, the coding polymorphisms.

# Size-Exclusion Chromatography (SEC)

Size-exclusion chromatography was performed in running buffer [veronal-buffered (VB) saline with added 140 mM NaCl]. MBLdeficient serum was reconstituted with rMBL [final concentration 50 µg/ml and incubated at room temperature (RT) for 60 min and, like MBL-sufficient and MBL-deficient sera, double filtered with a 0.2 µM Whatman filter (GE Healthcare, Buckinghamshire, UK)], before 500 µl sample was applied to 23.5 ml, 10 mm × 300 mm Superdex™ 200 10/300 GL (GE Healthcare, Buckinghamshire, UK) prepacked column equilibrated and eluted with running buffer at a constant 0.5 ml/min flow rate. 500 µl fractions were collected in 96-deep well plates and stored at 4°C until tested. The column was calibrated with GE Filtration Calibration Kits (GE Healthcare, Buckinghamshire, UK) according to the manufacturer's guidelines.

# LP Components Detected by ELISA

All incubations were performed in 100 µl volume at RT, all washes in between incubation steps were done with water. Different fractions, obtained from SEC, were incubated at mannan-coated plates, as described earlier, and incubated for 60 min at RT. After washing, different components (MBL, MASP-1/3, and MASP-2) were detected using specific biotinylated mAbs (αMBL-1, αMASP-1/3 4H2A9, and αMASP-2 12D12) by incubation for 1 h at RT. Finally, poly-HRP [0.01% in VB with added 10 mM CaCl2, 2 mM MgCl2, 0.3% (v/v) BSA, and 0.02% (v/v) Tween-20 (VB2<sup>+</sup> B/T)] was added to each well and incubated for 20 min at RT. Plates were developed as described previously.

To circumvent binding of already preformed MBL/MASP complexes and to detect the presence of MASP-1/3 and MASP-2 in MBL-deficient serum, mannan-coated plates were incubated with a fixed amount of rMBL (0.2 µg/ml). Briefly, different fractions were diluted (1:5) in VB saline with added 10 mM CaCl2, 2 mM MgCl2, 0.3% (v/v), and 0.02% (v/v) Tween-20 (VB2<sup>+</sup> B/T) and incubated for 1 h at RT. Plates were developed and detected as described earlier. Absorbance was measured at 450 nm and compared with a normal serum pool which was set to 100%.

The functional binding of FCN-1, FCN-2, and FCN-3 was determined by overnight coating with acetylated BSA (5 µg/ml in 0.1 M carbonate buffer, pH 9.6) in 96-well Nunc Maxisorp microtiter plates (Invitrogen, Breda, the Netherlands). After washing, different fractions were diluted 1:50 in VB2<sup>+</sup> B/T and incubated for 60 min. The binding of FCN was determined by incubation for 60 min at RT with specific biotinylated mAbs: αFCN-1 (AF4209), αFCN-2 (GN4), and αFCN-3 (FCN334) in VB2<sup>+</sup> B/T. Finally, plates were developed with poly-HRP (0.01% in VB2<sup>+</sup> B/T) and developed by addition of 0.1 mg/ml TMB I 0.11 M sodium acetate, containing 0.003% (v/v) H2O2, pH 5.5. Absorbance was measured at 450 nm and compared with a normal serum pool.

# RESULTS

### Elution Profile of MBL-Deficient and MBL-Sufficient Serum

Serum of either MBL-sufficient or MBL-deficient donors was fractionated by SEC to determine the elution profile of MBL and the MASPs. MBL serum levels of MBL-deficient donor were below 0.1 µg/ml and confirmed by genotype (HYPD/LXPA). The MBL-sufficient donor had MBL serum levels of 6.4 µg/ml (genotype HYPA/LYQA). MBL-deficient serum was completely restored in LP functionality following *in vitro* substitution with rMBL, indicating the presence of normal levels of LP components (data not shown).

Before fractionation of serum, the column used in the SEC was equilibrated using running buffer and calibrated with low- and high-weight markers (**Figure 1A**). A direct correlation (*r*<sup>2</sup> = 0.994) between elution volume and marker size was calculated (**Figure 1B**), allowing us to correlate the relative size of the protein based on the elution profile.

First, all fractions were incubated on mannan-coated plates, and bound MBL and complexed MASPs were detected. As expected, after fractionation of MBL-deficient serum, no LP components were detected on mannan-coated plates (**Figure 2A**). Fractionation of MBL-sufficient serum showed the presence of MBL and MBL/MASP-2 complexes binding to the mannan-coated plate (**Figure 2B**). Mannan-bound MBL was mainly observed at 550 kDa and around 80 kDa, corresponding to a polymeric and a trimer of the 32-kDa MBL protein. A single peak of complexed MBL/MASP-2, around 430 kDa, was observed, indicating that MASP-2 only associated with higher-order oligomers. We did not find any MASP-1/3 in the higher polymeric forms of MBL being not detected on mannan-coated plates in these experimental conditions.

# Redistribution of MASPs

To investigate whether MASP-1/3 and MASP-2 were able to redistribute from serum and associate to mannan-bound rMBL, all fractions were incubated on the mannan coat, which was first saturated with rMBL. Detection with mAb against MASP-1/3 and MASP-2 revealed that both MASP-1/3 and MASP-2 from the MBL-deficient serum were able to associate with mannanbound rMBL (**Figure 2C**). MASP-1/3 that associated with rMBL on the plate was derived from fractions containing proteins

with different molecular sizes, i.e., at around 100–130 and 210–260 kDa. MASP-2 showed a similar elution profile with a peak around 440 kDa. Both MASPs were apparently also present in their monomeric free form, not bound to any collectin including MBL, as indicated by the presence of the protein in fractions corresponding to a molecular size of 70 kDa.

MBL-associated serine protease-2 from the MBL-sufficient serum that associated with rMBL on the plate was derived from fractions containing high-molecular weight proteins (peak around 540 kDa) in which the polymeric MBL was present, and from fractions containing lower molecular weight proteins (65–84 kDa), corresponding to non-complexed monomers of MASP-2. In this serum, MASP-1/3 showed a more diffuse elution pattern with a peak around 430 kDa, which was at a higher molecular weight compared with that in MBL-deficient serum and could correspond to serum fractions containing FCNs. Like MASP-2, MASP-1/3 was also found in a fraction of MBL-sufficient serum with a relative size corresponding to a non-complexed, free monomeric form (**Figure 2D**).

# Restoration of MBL-Deficient Serum by rMBL

The addition of rMBL to MBL-deficient serum can completely restore the LP functionality *in vitro* as shown by a recovery of C4-deposition (18). To investigate whether rMBL associates with free MASPs in fluid phase or MASPs that are in complex with other collectins (for instance, FCNs) may redistribute to rMBL, MBLdeficient serum was reconstituted with rMBL and subsequently fractionated. rMBL contains primarily trimeric (225 kDa), tetrameric (300 kDa), and pentameric (375 kDa) oligomers (16, 27). All fractions were incubated on a mannan-coated plate and tested for the presence of MBL–MASP complexes.

Both MBL/MASP-1/3 and MBL/MASP-2 were detected in fractions containing high-molecular weight proteins (**Figure 3A**), indicating that in fluid-phase rMBL associated with MASP-1/3 and MASP-2. MBL/MASP-1/3 complexes were found in a diffuse pattern, but the most prominent peaks were around 350 and 710 kDa. MBL/MASP-2 complexes showed a similar pattern with a main peak around 710 kDa.

Figure 2 | Size-exclusion chromatography (SEC) of mannan-binding lectin (MBL)-deficient and MBL-sufficient serum. SEC of MBL-deficient serum (A,C) or MBL-sufficient serum (B,D). Serum was fractionated on a Superdex™ 200 10/300 GL column (A,C). Fractions were tested for the presence of MBL, or MBL–MBL-associated serine protease (MASP)-1/3, MBL–MASP-2 complexes on mannan-coated plates and compared with a pool of normal human sera (NHS). (B,D) Mannan-coated plates were pre-incubated with fixed amount of recombinant MBL, and followed by the different fractions of MBL-deficient or MBL-sufficient serum, and the formation of new MBL–MASP-1/3 or MBL– MASP-2 complexes was detected and compared with a pool of NHS.

To determine the origin of the MASPs that associated with rMBL in fluid phase, all fractions were also analyzed on mannancoated plates saturated with rMBL (**Figure 3B**). Both MASP-1/3

and MASP-2 were found in a distinct peak around a relative size of 440–560 kDa for MASP-1/3 and around 560–710 kDa for MASP-2. In contrast to the previously observed double peak for MASP-1/3, we now observed a single peak at a higher molecular weight with a noted absence of the non-complexed form of MASP, suggesting that free soluble MASP-1/3 and MASP-2 associated with rMBL once added. However, our experimental setup was not able to distinguish between heterocomplexes of MASP-1/3 and MASP-2 within a single MBL protein or separate complexes of either MBL-associated MASP-1/3 or MBLassociated MASP-2.

#### FCN/MASP Complexes

We hypothesized that formation of MBL/MASP complexes more readily happens when MBL is bound to mannan, hereby attracting MASPs from FCN/MASP complexes (**Figure 2**). We therefore investigated the presence of different FCNs in the fractions of MBL-deficient and reconstituted MBL-deficient serum. FCN-1 was not detected in any of the fractions.

The fractions of MBL-deficient serum and reconstituted MBL-deficient serum showed the functional binding of FCN-2 to acetylated BSA as a peak around 540 kDa (**Figure 4A**), which did not alter after rMBL-reconstitution. Similar observations were made for FCN-3 (**Figure 4B**). Neither FCN-2 nor FCN-3 showed a different elution profile after reconstitution with rMBL, suggesting that the detected MASPs, in the mannan-bound MBL assay, are not derived from FCN/MASP complexes but were unbound MASPs (low molecular fraction) that reacted with the ligand-bound MBL.

#### DISCUSSION

In this study, we investigated the association of rMBL with MASPs in MBL-deficient serum. Our data demonstrated the unexpected presence of free non-complexed MASP-1/3 and MASP-2 that could associate to fluid-phase rMBL when added to MBL-deficient serum.

MBL-associated serine protease-1/3 and MASP-2 can associate with different LP PRMs. It has been hypothesized that MASPs are in equilibrium and can freely associate with the different available PRMs which act as carrier proteins (23, 28). We hypothesized that upon binding of these PRMs to their ligands the equilibrium of MASP binding would shift in favor of the ligand-bound protein, resulting in the formation of PRM/MASP complexes to initiate the complement activation cascade. *In vitro* substitution of MBL-deficient serum with rMBL restored the C4-deposition on mannan-coated plates, indicating the formation of functional rMBL/MASP complexes. Our findings showed the ability of mannan-bound rMBL to form new rMBL/MASP complexes in MBL-deficient serum and MBL-sufficient serum. Both MASP-1/3 and MASP-2 were found to associate with mannan-bound rMBL. The relative size of the newly formed complexes suggested that

MASPs are present as free non-complexed proteins as well as in a bound form, being associated with different LP complexes. A likely source of these LP complexes would be FCN-2 and FCN-3 in MBL-deficient serum. This is different form previous published reports showing the tendency of MASPs to form a dimeric molecule following analysis of recombinant proteins or complexes (29–31), which could be related to the origin of the proteins analyzed.

The elution profile of MBL-sufficient serum showed immediate similarity with the previously published results of Dahl et al. (4) These authors described different polymeric forms of MBL, ranging from relative sizes of 275 kDa and smaller (MBL-I), to 345–580 kDa (MBL-II), 580–900 kDa (MBL-III), and 900 kDa and above (MBL-IV). Although we were able to detect MASP-1 in unfractionated MBL-sufficient serum, the different fractions obtained from SEC after fractionation showed no detectable signal on mannan-coated plates. We confirmed that MASP-2 primarily associated with MBL-II in MBL-sufficient serum. After reconstitution of MBL-deficient serum, we were able to detect both MASP-1/3 and MASP-2 on mannan and mannan-bound MBL plates primarily in fractions corresponding to higher sizes, between 200 and 900 kDa. This suggests that an association takes place between other LP complexes and rMBL in fluid phase. We were unable to detect the presence of FCN-1 in our fractions, which was to be expected, given the low serum concentration of FCN-1 and further dilution during fractionation (32, 33). In contrast to FCN-1, higher-order oligomers of FCN-2 were detected in fractions corresponding to a relative size of 710 kDa, and FCN-3 oligomers were detected in a broad range of sizes 400–900 kDa. Although Ohashi and Erickson (34) and Munthe-Fog et al. (35) have reported even larger sizes around 800–870 kDa, our column has a lower discriminatory sensitivity for the very high-molecular weights. The distribution patterns of FCN-2 and FCN-3 are similarly between MBL-deficient serum and reconstituted MBLdeficient serum, supporting the hypothesis of MASP-2 preferentially binding to a ligand-bound protein. This is also supported by Megyeri et al. (36) who showed an increased yield of MBL/ MASP complexes following addition of recombinant MBL and subsequent purification using mannan-sepharose. The pattern of FCN-2 and FCN-3 appear to correspond to the observed pattern of newly formed MASP complexes on mannan-bound MBL, we were unable to observe a relative change in size.

The presence of a non-complexed form of MASP-2 and MASP-1/3 provides a new mechanism of restoration of the LP activity following MBL reconstitution. This non-complexed free form of MASP-2 and MASP-1/3 could no longer be detected following addition of rMBL to MBL-deficient serum, leading to the formation of a new complex with a larger molecular size indicating a redistribution of MASPs, which was suggested before by Megyeri et al. (36). Whether these different MASPs consist of homodimers or heterodimers before association with oligomeric MBL is still debated (6, 28, 37, 38).

Early MBL substitution studies in MBL-deficient patients have shown the safety of infusion of pdMBL (11), but a limited restoration of the LP functionality was observed despite reaching high plasma MBL levels (16). Analysis of pdMBL showed high levels of associated pre-activated MASP-2 (27), which is rapidly inactivated upon *in vivo* infusion by natural inhibitors such as C1-inhibitor (17). The MBL-associated MASPs in the pdMBL

ligand binding of FCN-3 before and after reconstitution.

product are rapidly cleared within 24 h after substitution with limited restoration of the LP (16). By contrast, rMBL has been shown to efficiently restore the LP functionality as indicated by C4-deposition both *in vitro* (18) and *in vivo* (36). The potential therapeutic value of MBL replacement depends on the potential to form activating complexes with MASPs that are functional upon ligand binding. Our study showed the presence of a natural non-bound form of MASP-2 in serum to explain the mechanism by which rMBL can restore LP functionality in MBL-deficient patients. These rMBL–MASP complexes are not yet activated, in contrast to the pre-activated MASPs bound to MBL due to affinity chromatography purification (17, 25). Our results indicate that rMBL—upon its formation of functional complexes after infusion—will not be promptly inhibited and cleared as we had previously observed for pdMBL (17). Both the fact that rMBL substitution was well-tolerated in a phase I trial (39) and the presence of non-bound forms of MASPs able to associate with rMBL and restore LP functionality provide a rationale to consider new clinical rMBL substitution studies in carefully selected patients.

# AUTHOR CONTRIBUTIONS

Study concept and design; drafting of manuscript: MK, DW, and TK. Acquisition of data: MK, AK, and GM. Statistical analysis of data: MK, AK, and DW. Interpretation of data: MK, AK, DW, and

#### REFERENCES


TK. Critical revision of the manuscript for important intellectual content: all the authors.

#### FUNDING

This study was supported by a grant from KiKa (stichting Kinderen Kankervrij) (project 41).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01406/ full#supplementary-material.

Figure S1 | Specific monoclonal antibody (mAb) against MBL-associated serine protease (MASP)-2. Plates were directly coated with recombinant proteins, respectively, rMASP-1 (○), recombinant MASP-2 (rMASP-2) (■), and rMASP-3 (□). Coated proteins were detected using a titration of specific biotinylated mAbs against MASP-1/3 (4H2A9) (36), against MASP-2 (12D12) or against MASP-3 (38:12-3) (36). (A) mAb 12D12 binds only to MASP-2 and does not show cross-reactivity toward the other serine protease. (B) mAb 4H2A9 is able to detect both MASP-1 and MASP-3 bound to the plate, by binding to a shared epitope (40).

Figure S2 | Reconstitution with recombinant MBL (rMBL) restores lectin pathway functionality. Plates were directly coated with mannan and a serum dilution of reconstituted mannan-binding lectin (MBL)-deficient serum with 50 μg/ml rMBL (●), MBL-deficient serum (○), or MBL-sufficient serum was tested for their C4-converting function.


recognition molecules in systemic lupus erythematosus. *Clin Exp Immunol* (2015) 182:132–8. doi:10.1111/cei.12678


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Keizer, Kamp, van Mierlo, Kuijpers and Wouters. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Be on Target: Strategies of Targeting Alternative and Lectin Pathway Components in Complement-Mediated Diseases

*József Dobó, Andrea Kocsis and Péter Gál\**

*Institute of Enzymology, Research Centre for Natural Sciences, Hungarian Academy of Sciences, Budapest, Hungary*

The complement system has moved into the focus of drug development efforts in the last decade, since its inappropriate or uncontrolled activation has been recognized in many diseases. Some of them are primarily complement-mediated rare diseases, such as paroxysmal nocturnal hemoglobinuria, C3 glomerulonephritis, and atypical hemolytic uremic syndrome. Complement also plays a role in various multifactorial diseases that affect millions of people worldwide, such as ischemia reperfusion injury (myocardial infarction, stroke), age-related macular degeneration, and several neurodegenerative disorders. In this review, we summarize the potential advantages of targeting various complement proteins with special emphasis on the components of the lectin (LP) and the alternative pathways (AP). The serine proteases (MASP-1/2/3, factor D, factor B), which are responsible for the activation of the cascade, are straightforward targets of inhibition, but the pattern recognition molecules (mannose-binding lectin, other collectins, and ficolins), the regulatory components (factor H, factor I, properdin), and C3 are also subjects of drug development. Recent discoveries about cross-talks between the LP and AP offer new approaches for clinical intervention. Mannan-binding lectin-associated serine proteases (MASPs) are not just responsible for LP activation, but they are also indispensable for efficient AP activation. Activated MASP-3 has recently been shown to be the enzyme that continuously supplies factor D (FD) for the AP by cleaving pro-factor D (pro-FD). In this aspect, MASP-3 emerges as a novel feasible target for the regulation of AP activity. MASP-1 was shown to be required for AP activity on various surfaces, first of all on LPS of Gram-negative bacteria.

Keywords: complement system, lectin pathway, alternative pathway, complement inhibitors, complement-related diseases

#### *Edited by:*

*Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France*

#### *Reviewed by:*

*Cordula M. Stover, University of Leicester, United Kingdom Maciej Cedzynski, Institute for Medical Biology (PAN), Poland Christian Drouet, Université Grenoble Alpes, France*

> *\*Correspondence: Péter Gál gal.peter@ttk.mta.hu*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 29 May 2018 Accepted: 26 July 2018 Published: 08 August 2018*

#### *Citation:*

*Dobó J, Kocsis A and Gál P (2018) Be on Target: Strategies of Targeting Alternative and Lectin Pathway Components in Complement-Mediated Diseases. Front. Immunol. 9:1851. doi: 10.3389/fimmu.2018.01851*

**172**

**Abbreviations:** 3MC syndrome, Malpuech–Michels–Mingarelli–Carnevale syndrome; aHUS, atypical hemolytic uremic syndrome; AMD, age-related macular degeneration; AP, alternative pathway; C4BP, C4b-binding protein; CARPA, complement activation-related pseudoallergy; CCP, complement control protein; CP, classical pathway; CNS, central nervous system; CR1, complement receptor 1; CR2, complement receptor 2; CRP, C-reactive protein; CUB, C1r/C1s, the sea urchin protein Uegf, and the human bone morphogenetic protein 1; CVF, cobra venom factor; EGF, epidermal growth factor; DAF, decay-accelerating factor; DAMP, damage-associated molecular pattern; FHL, FH-like protein; FHR, FH-related protein; FIMAC, FI membrane attack complex; GPI, glycosyl phosphatidylinositol; HAE, hereditary angioedema; HUVEC, human umbilical vein endothelial cell; LDLr, low-density lipoprotein receptor; LP, lectin pathway; MAC, membrane attack complex; MBL, mannose-binding lectin; MASP, MBL-associated serine protease; MCP, membrane cofactor protein; IGFBP-5, insulin-like growth factor-binding protein 5; IRI, ischemia–reperfusion injury; LPS, lipopolysaccharide; PAMP, pathogen-associated molecular pattern; PNH, paroxysmal nocturnal hemoglobinuria; PRM, pattern recognition molecule; SIRS, systemic inflammatory response syndrome; SP domain, serine protease domain; SLE, systemic lupus erythematosus; TAFI, thrombin-activatable fibrinolysis inhibitor; TSR, thrombospondin type repeat; TCC, terminal complement complex; VWA, Von Willebrand factor type-A.

### BRIEF OVERVIEW OF THE COMPLEMENT SYSTEM

#### Initiation Phase

The complement system is a sophisticated network of serum proteins (recognition molecules, proteases, modulators, inhibitors) as well as cell-surface regulators and receptors that constitute a key part of the host defense machinery. The complement system is a powerful effector component of the innate immunity and a vital modulator of the adaptive immune response (1–3). The complement system recognizes, tags, and eliminates microbial intruders and other dangerous particles such as immune complexes, damaged, and altered self cells. The complement system is inactive (or at least shows a very low basic activity: "tickover") until it is activated by various danger signals. There are three canonical pathways through which the complement system can be activated: the classical pathway (CP), the lectin (LP), and the alternative pathways (AP) (**Figure 1**). CP and LP have several features in common. In both cases, pattern recognition molecules (PRMs) bind to the danger-associated structures. The PRMs, like the other complement components, are modular proteins, consisting of multiple structural domains (**Figure 2**). C1q, the single PRM of the CP binds primarily to immune complexes containing IgG or IgM, and to C-reactive protein (CRP) *via* its C-terminal globular domains (4). These globular domains are fused to N-terminal collagen-like arms forming the characteristic "bunch-of-six-tulips" structure. The structure of the PRMs of the LP resembles that of C1q; globular heads and collagen-like arms. However, the recognition domains of mannose-binding lectin (MBL), other collectins, and ficolins bind to different structures. The C-type lectin domains of MBL recognize the carbohydrate pattern of the bacterial surfaces. Ficolins (ficolin 1, 2, and 3) bind to acetylated compounds, typically to acetylated sugars of bacteria, *via* their fibrinogen-like domains (5). Collectins (CL-K1 and CL-L1) also recognize sugars and other potential danger signals. Unlike C1q, which has the well-established hexamer structure, MBL, ficolins, CL-K1, and CL-L1 exist in different oligomerization states, from dimer to hexamer; the tetramer being the dominant form at least for MBL. These PRMs circulate in complex with serine protease (SP) zymogens and monitor

continuously for dangerous particles the bloodstream. When the PRMs bind to the target surface, the associated SPs become activated and initiate a proteolytic cascade system, which amplifies the initial signal tremendously. C1q is associated with two C1r and two C1s proteases (the so-called "tetramer") to form the C1 complex of the CP (6). MBL/ficolin-associated serine protease 1 and 2 (MASP-1 and MASP-2) are the initial proteases of the LP (7, 8). These SPs, together with the third MBL/ficolinassociated SP (MASP-3) form a protease family with the same domain structure (**Figure 2**) and similar function. The activation of the CP and LP results in the formation of the same enzyme complex, a C3 convertase (C4b2a) that cleaves C3, the central component of the complement system. The first enzymatic step in the CP activation is the autoactivation of C1r. Activated C1r then cleaves zymogen C1s, which in turn cleaves C4 and C2. In the LP, MASP-1 autoactivates first and then cleaves MASP-2 (9). MASP-2 is the enzyme of the LP that cleaves C4 (10, 11), while C2 is cleaved by both MASP-1 and MASP-2. C3 and C4 are closely related thioester-containing proteins that form the basis of the convertase complexes (12, 13). Their function is to covalently attach the convertase to the activation surface and to capture the SP components of the enzyme complex. C2 is the SP component of the C3 convertase of the CP and LP. Activation of the AP is quite different from that of the CP and LP (14). When the CP/LP C3 convertase (C4b2a) cleaves C3, a smaller fragment is released (C3a). The larger fragment (C3b) covalently binds to the activation surface preferably through an ester or, less likely, through an amide bond due to the reaction of the exposed thioester bond (15, 16). The nascent C3b component binds factor B (FB), the SP component of the AP C3 convertase. FB is cleaved by FD, a SP which circulates predominantly in its cleaved form in the blood. The resulting C3bBb is the AP C3 convertase, which converts more C3 into C3b. The new C3b molecules serve as platforms for new C3 convertase complexes. In this way, a positive feedback loop amplifies the initial signal tremendously generated either by the CP or the LP (17). According to the C3 tickover hypothesis, the AP can also initiate on its own without involvement of CP or LP (18). The circulating C3 molecules hydrolyze slowly and spontaneously in the bloodstream. The resulting C3(H2O) is a C3b-like molecule; it can bind FB and then form an "initiation" C3 convertase (C3(H2O)Bb). If this fluid-phase C3 convertase emerges near a surface, the nascent C3b molecules can bind to the surface and initiate the positive feedback loop. In this way, the AP continuously monitors the different surfaces and if it finds an activator surface, it launches efficient complement activation. The self-tissues are protected from AP-mediated damage by cell-bound and fluid-phase inhibitors (**Figure 1**). These inhibitors dissociate the C3bBb complex and serve as cofactors for the serine protease factor I (FI) in the degradation of C3b (19). Decay-accelerating factor (DAF, CD55), membrane cofactor protein (MCP) (CD46), complement receptor 1 (CR1, CD35) are cell-surface-bound while the master regulator of the AP is the fluid-phase protein, factor H (FH). FH binds to cell-surfacedeposited C3b and facilitates its degradation to iC3b, C3c, and C3dg by FI. On endogenous cell membranes, which expose sialic acid, binding of FH is tight and the degradation of C3b is rapid, while on the so-called "protected surfaces" (e.g., bacteria, fungi) binding is weak and the amplification loop of AP can build up. There is a positive regulator of the AP, properdin, which increases the half-life of the C3bBb complex. Originally, at its discovery, properdin was regarded as a pattern recognition-like initiator molecule of the AP (20). Later, it was considered as a positive regulator (21); however, recently, the pattern recognition function of properdin has been reconsidered (22, 23).

# From the Central Phase to the Terminal Pathway

The cleavage of C3 by the C3 convertases is the turning point of complement activation. At this point, the three activation pathways (CP, LP, and AP) merge into a unified terminal pathway (**Figures 1** and **3**). When the density of surface-deposited C3b reaches a certain level, the substrate specificity of the C3 convertases switches to cleave the C5 component. The C4b2a(C3b)n and C3bBb(C3b)n convertases cleave the C5 component into a smaller (C5a) and larger (C5b) fragments. The C5a fragment, like the C3a fragment, is an anaphylatoxin. The anaphylatoxins bind to their receptors (C3aR and C5aR1/2) on leukocytes and endothelial cells and initiate inflammatory reactions (24). Structurally, C5 is similar to C3 and C4 (although it does not contain thioester bond) (25). The cleavage of C5 is the last enzymatic step in the complement cascade. From this point, conformational changes drive the formation of a self-organizing protein complex that damages the membrane of the attacked cells [membrane attack complex (MAC)]. After cleavage, the nascent C5b undergoes a conformational change that enables it to bind the C6 and C7 components (26). The resulting C5b–C7 complex binds to the cell membrane and captures C8. After conformational changes, C8 integrates into the membrane and pave the way to the integration of multiple copies of C9 molecules. The C9 molecules form a pore in the membrane, which results in the disintegration and destruction of the cell (27).

# COMPLEMENT-MEDIATED DISEASES

The complement system is an extremely effective cell-killing and inflammation provoking machinery. To prevent excessive activation, the complement system is kept under strict control by the different inhibitory mechanisms. A delicate equilibrium between activation and inhibition is necessary to maintain the inflammatory homeostasis in the human body. When this equilibrium is disrupted by any reason, the self-tissues can be damaged and severe disease conditions can occur. There are many clinical disorders in which uncontrolled (or sometimes the insufficient) complement activation is involved. Usually, the etiology of these diseases is complex, and the unwanted complement activation is only one of the pathological factors. However, evidences obtained by using various disease models suggest that preventing or inhibiting the pathological complement activation can be a promising therapeutic approach.

#### Insufficient Complement Activation

Since the complement system provides a first line of defense against invading pathogen microorganisms, deficiency of a complement component can lead to severe infections. The consequences could be more severe during childhood, when the adaptive immune system is not developed enough. Deficiency of the initial SPs of CP and LP (C1r, C1s, MASP-2) can result in pyogenic infections (28). Deficiency of MBL is the most common immunodeficiency in humans, affecting approximately 30% of the human population (29). It predisposes to recurrent infections in infancy; however, it is not a major risk factor in the adult population. Deficiency of the alternative and the terminal pathway components can severely compromise the defense against Gram-negative bacterial infections (30). Deficiency of properdin or deficiencies in the components of MAC are associated with infections of *Neisseria* species causing meningococcal meningitis or sepsis (31–33). A very important function of the CP is the continuous removal of immune complexes and apoptotic cells. If this pathway is compromised in systemic lupus erythematosus (SLE) due to deficiency of C1q, or C4, or C1r/s, severe autoimmune reactions occur resulting in tissue injury in the kidneys.

# Excess Complement Activation

The majority of complement-related diseases are associated with overwhelming complement activation due to inappropriate control. The kidney is especially vulnerable for complement-mediated attacks. In the case of C3 glomerulopathy, C3 deposition occurs in the glomeruli without immunoglobulin deposition (34, 35). C3 deposition in this case is likely the consequence of uncontrolled AP activation. In contrast to that, in the case of membranoproliferative glomerulonephritis, CP activation elicits C3 deposition, since immunoglobulins and C1q are also deposited (36, 37). The IgA nephropathy is characterized by deposition of polymeric IgA1, which triggers complement activation through the AP and the LP (38, 39). Atypical hemolytic uremic syndrome (aHUS) is also a complement-related disease, which can lead to end-stage renal failure (37, 40). The driving force behind aHUS is the inappropriate AP activation often due to variants of (41) or autoantibodies against (42) FH, the master regulator of the AP. aHUS is a form of thrombotic

conversion processes, while red arrows stand for enzymatic reactions pointing from the enzyme toward its substrate. The circle-shaped red arrow symbolizes the autoactivation of MASP-1.

microangiopathy accompanied with thrombocytopenia, hemolytic anemia, vascular damage, and thrombosis.

Another rare clinical condition associated with uncontrolled AP activation is paroxysmal nocturnal hemoglobinuria (PNH) (43, 44). In PNH patients, red blood cells are particularly prone to complement-mediated lysis due to the lack of two membranebound regulator proteins: DAF (CD55) and CD59. This is the consequence of the defect in glycosyl phosphatidylinositol (GPI) synthesis in the cells. GPI is responsible for anchoring various proteins to the cell membrane including these inhibitors that regulate the activation of the AP (CD55) and the formation of MAC (CD59).

Age-related macular degeneration (AMD) is a complementrelated disease, which affects a large population (about 100 million AMD cases) worldwide. It is the leading cause of blindness among the elderly in the developed world (45). Genetic analyses strongly suggest that uncontrolled complement activation, especially that of the AP, plays a major role in the pathogenesis of AMD. Genetic variants of FH, C3, FB, FI, and C9 have been associated with AMD (46). In the center of the retina of AMD patients, the photoreceptor cells are gradually degraded due to a chronic inflammation, which manifests in the accumulation of immune deposits called drusen underneath the retinal pigment epithelium. The drusens (that contain activated complement components) compromise the transport of oxygen and nutrients to the photoreceptors facilitating their degeneration. Numerous attempts have been made to curb the unwanted AP activation in the eye with limited success (47). In order to efficiently influence complement activation in the eye, we have to reveal its exact mechanism, which could be different in the periphery than in the bloodstream.

Ischemia–reperfusion injury (IRI) can be considered as a severe autoimmune reaction (48), which plays a major role in a number of clinical conditions. When the blood flow in an organ stops temporarily for any reason, the deprivation of oxygen (hypoxia) induces changes in the tissues, which predisposes them for complement-mediated attack after reperfusion. The affected cells and tissues are recognized by the immune system as damaged self [damage-associated molecular pattern (DAMP)], and a complex inflammatory reaction is launched, in which the complement system plays a decisive role. IRI significantly contributes to the tissue damage in the case of myocardial infarction, stroke, transplant-induced inflammation, and it can cause a complication during coronary artery bypass graft surgery. Although the exact mechanism of complement activation in case of IRI is not fully clarified yet, a number of evidences suggest that inhibition of the LP could be therapeutically advantageous (49–53).

Artificial materials used in modern medicine, such as polymer plastics and metal alloys, can also activate the complement system and trigger inflammation (54). Nanoparticles used as contrast agents or drug carriers can also activate the complement system, sometimes causing a severe adverse reaction, called complement activation-related pseudoallergy (55, 56). In this type of hypersensitivity reaction, IgE is not involved. Liposomal drugs directly activate the complement system liberating C3a and C5a anaphylatoxins, which trigger mast cells and basophils.

If the immune system is exposed to an overwhelming amount of danger signals [pathogen-associated molecular patterns (PAMPs) or DAMPs], a systemic inflammatory reaction can occur, which could be more devastating than the original danger source. In the case of systemic inflammatory response syndrome, such as sepsis or polytrauma, the massive and systemic complement activation fuels a vicious cycle of hyperinflammatory events that can results in fatal tissue damage (57).

A growing number of evidences indicate that the complement system plays an important role in fundamental developmental processes. The lack of functional LP components (MASP-3, CL-K1, CL-L1) during embryogenesis results in the Malpuech– Michels–Mingarelli–Carnevale (3MC) syndrome that manifest in characteristic craniofacial dysmorphism and multiple other anomalies (58, 59). It was shown that an intact CP is essential for postnatal brain development. Contribution of C1, C4, and C3 was demonstrated to synaptic pruning essential for proper neuron circuit formation (60). These complement components tag synapses and mediate their elimination during a discrete window of postnatal brain development. C1q or C3 deficiency in mice results in improper central nervous system (CNS) synapse elimination. If these processes, essential during normal brain development, are pathologically upregulated during adulthood, they can contribute to the development of neurodegenerative diseases, such as Alzheimer disease and frontal temporal dementia (61). In addition to that, uncontrolled CP and LP activation in the CNS can also contribute to psychiatric disorders such as schizophrenia (62, 63).

### CROSS-TALK BETWEEN THE AP AND THE LP

As described above, there are three canonical activation routes of the complement system. It is also obvious that the CP and the LP would not work efficiently without the amplification loop provided by the AP, hence, the three pathways are naturally interconnected. It is also possible that homologous proteins C4 and C3, or C2 and FB can substitute each other to a certain degree; at least *in vitro* experiments indicate a weak cross-reactivity between CP/LP an AP C3 convertase components (64). MASP-1 (65) and MASP-2 (66) have both been implicated to be able to directly cleave C3; however, the physiological relevance of these reactions is uncertain. MBL was also shown to be involved in AP activation without the requirement of C2, C4, and MASPs (67), but in the light of our recent results, the observed effect could be mediated by MASP-1 (68). In summary, the involvement of the LP or LP components in AP activation has been demonstrated in the literature before; however, some results still remain controversial. In the subsequent two sections, recent discoveries are presented regarding the role of MASP-3 during the very early stage of AP activation, and the requirement of MASP-1 for AP activation on various surfaces.

#### Active MASP-3 Is the Professional Pro-FD Maturase in Blood

The first evidence that MASP-1 or MASP-3 might have an essential role in AP function came from the group of T. Fujita (69). They created *MASP1* knock-out mice by replacing the second exon (8). Since this region encodes a common part of both MASP-1 and MASP-3, the final homozygous mouse strain lacked both proteins. Surprisingly, these mice had pro-FD in their sera and had no AP activity (69). They suggested that MASP-1 acts as an essential enzyme for pro-FD maturation. At the time, it seemed like a logical assumption to favor MASP-1 over MASP-3 since MASP-1 is a more active enzyme in general with a relatively broad substrate specificity (70). Later, the same group suggested that MASP-3 might be more important than MASP-1 in pro-FD activation and suggested that even the proenzyme form of MASP-3 can act as the activator (71).

Subsequent publications questioned the requirement of either MASP-1 or MASP-3 for AP activity. In the serum of a 3MC syndrome patient lacking both proteins, functional AP was observed (72), and in mice deficient for MASP-1, MASP-3, and FH extensive, AP activation was observed, just like in mice deficient for FH only (73).

To clarify the roles of MASPs in pro-FD activation, we set up a series of experiments. *In vitro* all active MASPs were shown to be able to cleave pro-FD efficiently to produce FD, whereas the MASP zymogens lacked such activity (74). We prepared fluorescently labeled pro-FD, added it to different types of human plasma and serum preparations and followed the conversion of pro-FD to FD. Pro-FD was efficiently cleaved in all types of blood preparations, even in citrated and EDTA plasma, where neither the complement nor the coagulation cascade is expected to be activated. This experiment established that at least one protease is present in normal human blood capable of converting pro-D to FD without the prior activation of the abovementioned proteolytic cascades. Using a MASP-1-specific and a MASP-2-selective inhibitor, these two enzymes could be excluded. After adding recombinant active catalytic fragments of MASPs to normal human plasma samples, the half-life of labeled pro-FD was markedly reduced upon the addition MASP-3, whereas the other two enzymes had no effect (74).

The final "killer" experiment that established MASP-3 as the professional (near exclusive) activator of pro-FD came using a MASP-3 specific inhibitor, TFMI-3 (75). TFMI-3 blocked the conversion of labeled pro-FD to FD in citrated plasma, EDTA plasma, and hirudin plasma completely, while in serum, the halflife was markedly increased. Another conclusion of our studies was that active MASP-3 must be present in the blood, since only the activated form of MASP-3 can convert pro-FD to FD. Later, we provided direct evidence for the extensive basal-level activation of MASP-3 in human blood by an unknown mechanism (76). Finally, the debate seems to have settled. A recent paper showed that in 3MC syndrome patients lacking MASP-3, predominantly pro-FD is present in their sera, and moreover, in healthy individuals, some pro-FD is also present beside the dominant active form (77).

The picture is now clear. Under normal circumstances, active MASP-3 is present in the blood, which activates pro-FD, therefore, continuously supplying active FD for the AP (**Figure 3**). However, MASP-3-deficient individuals are not completely defenseless. At least one coagulation enzyme can probably also provide low levels of FD for the AP, or when the LP is activated, MASP-1 or MASP-2 might also contribute. These backup mechanisms also need some consideration when targeting MASP-3 to control AP activity, as it will be discussed later.

# MASP-1 Is Required for AP Initiation on Certain Activating Surfaces

Despite the fact that molecular mechanisms of the complement system have been thoroughly examined in the past decades, still many questions remained about the early steps of the activation. MASP-1 as a promiscuous enzyme with broad substrate specificity (70) has the potential to replace other SPs or amplify enzymatic reactions. It has been reported that MASP-1 is indeed involved in biological processes beyond LP or even the complement system.

Recently, we have found a novel function of MASP-1 in AP activation apart from its role in the LP (**Figure 3**). This function suggests an unexpected linkage between the two pathways and also highlights the differences between various activation surfaces (68). Previously, specific and highly selective smallprotein inhibitors against all MASPs were developed from canonical inhibitor scaffolds using the phage-display technique. SGMI-1 is a specific MASP-1 inhibitor, whereas SGMI-2 inhibits MASP-2, and TFMI-3 is a specific inhibitor of MASP-3 (9, 75). Activation of the AP *in vitro* can be carried out on ELISA microtiter plates coated by bacterial lipopolysaccharide (LPS) or yeast zymosan. These materials serve as models of the pathogenic surfaces. In the absence of Ca2+ using Mg2<sup>+</sup>-EGTA buffer, AP activation can be initiated without the involvement of CP and LP. Administration of the specific MASP-1 inhibitor, SGMI-1 in this system lead to surprising results. The activity of the AP was attenuated significantly through the inhibition of MASP-1. However, this effect was only seen on the bacterial surface represented by LPS, while zymosan-induced AP activation was not compromised. To rule out the possibility that SGMI-1 may impede other SPs, inhibitors of MASP-1 possessing different mode of action were also tested. Anti-MASP-1-SP antibody, N-terminal domains of MASP-1 (M1\_D1-3) (78), and serpin domain of C1 inhibitor resulted in the same, considerable reduction of AP activity but only on LPS. The activity of AP in MASP-1-depleted serum remarkably decreased on LPS-coated surface while on zymosan-coated surface, it was only moderately affected. The mechanism of C3b deposition, which is followed in our assay, can be divided to initiation and amplification phases. Time-course measurement of C3b deposition in the presence of subsequently added SGMI-1 inhibitor indicated that MASP-1 contributes to both phases of C3b generation. Although we proved unambiguously that MASP-1 has an effect on AP activation, there are still many puzzling details to be solved. First, we tested the known components of AP as possible reaction partners of MASP-1. It was clarified earlier that MASP-1 cleaves C3 only at a very low rate (11) and now we confirmed that MASP-1 does not react significantly with C3b-bound FB. The contribution of FD was also excluded since SGMI-1 does not inhibit FD, and MASP-1 is not the physiological activator of pro-FD (74). Our results lead to the conclusion that the key player of MASP-1-driven AP activation is probably not among the core components of the AP.

Differences between the activating surfaces also need further investigations since it seems to be likely that AP initiation occurs by various mechanisms. Using specific antibodies in the ELISA system, neither MASP-1 nor MBL could be detected on LPS surface in Mg2<sup>+</sup>-EGTA buffer (68). MASP-1 may be presented by some other PRMs, which do not necessarily require Ca2<sup>+</sup> for binding (possibly ficolins). Another possible scenario is that MASP-1 forms a labile and transient complex with its reaction partners. One clue arises from the literature that properdin, which stabilizes C3bBb complex, is crucial for LPS-induced but not for zymosan-induced AP activation (79). Another coincidence is that LPS, rather than zymosan, binds FH with high affinity, which enhances the decay of C3 by FI activity. The ratio of these factors, which play a role in the regulation of AP, may be influenced by MASP-1 through a yet unknown mechanism.

These new findings draw attention to MASP-1 in the promotion of LPS-induced AP and, therefore, its role in the defense against Gram-negative bacteria.

#### COMPONENTS OF AP AND LP AS POTENTIAL DRUG TARGETS

#### Pattern Recognition Molecules

The PRM of the CP and LP recognize danger signals (PAMPs and DAMPs) and provide the framework of the initiation multimolecular complexes (the C1 complex, MBL–MASP complexes, ficolin–MASP complexes). They have similar overall domain architecture: N-terminal collagen-like domains and C-terminal globular domains. The structure of C1q is different from that of the other PRMs, since the basic trimeric subunit of C1q is composed of three different polypeptide chains (A, B, and C chains), while that of MBL and ficolins consist of only one kind of polypeptide chain. In the case of collectins, collectin kidney 1 (CL-K1) and collectin liver 1 (CL-L1), a heterotrimeric subunit was observed in human blood composed of one CL-L1 and two CL-K1 polypeptide chains (called CL-LK) (80). Another structural difference between C1q and the proteins of the collectin/ficolin family is that the latter contain a short N-terminal cysteine-rich region and an α-helical coiled-coil neck region between the collagen-like and the globular domains facilitating trimerization of the polypeptide chains. The collagen sequences (Gly-Xaa-Yaa repeats) are interrupted at one point in C1q and MBL generating a flexible kink region that may play an important role in binding to the danger patterns and activating the associated SPs. The PRMs bind the associated SPs in a Ca2<sup>+</sup>-dependent manner. C1q binds the C1s-C1r-C1r-C1s tetramer, while MBL, ficolins, and CL-LK bind MASP dimers. It is probable that low oligomeric MBL and ficolins bind a single MASP dimer, while higher oligomers (pentamers, hexamers) can bind two MASP dimers simultaneously (81). There is a cross-interaction between the components of the CP and LP; MBL can bind the C1r2s2 tetramer and C1q can bind the MASP dimers, although with reduced affinity compared to the cognate pairs (82). It is unlikely that these interactions have a physiological relevance (except maybe in deficiencies); however, the existence of these crossbindings proves that the interactions are analogous between the PRMs and the associated SPs among the components of the CP and LP. The binding of MBL and ficolins to their targets is Ca2<sup>+</sup>-dependent. The affinity of a single carbohydrate-binding domain to its target sugar is low (Kd in the millimolar range), whereas the avidity of the whole oligomeric molecule is high with a Kd in the low nanomolar range. It is very likely that the PRMs of higher oligomeric state activate the LP more efficiently than the low oligomeric PRMs due to the stronger binding to both the target surface and to the MASPs (83). The mechanism of activation of the C1 and MBL–MASP complexes is not fully clarified yet.

Theoretically, if we want to prevent or inhibit improper complement activation in a pathological situation PRMs of the initiation complexes are ideal targets; since by inhibiting the PRMs, we can shut off the entire amplification machinery of the complement system at the very first step. There are three possibilities to inhibit the function of the PRMs: (1) to prevent the binding of the PRMs to their target; (2) to prevent the binding of the associated SPs to the PRMs; (3) to prevent the conformational changes of the PRMs that are necessary for the activation of the SPs. Monoclonal antibodies that bind to the globular domains of the PRMs can efficiently interfere with the ligand binding. Anti-C1q and anti-MBL antibodies were successfully used to block the CP and LP activation, respectively (84). An anti-MBL monoclonal antibody (3F8) attenuated myocardial IRI in mouse expressing human MBL (85).

Anti-C1q antibodies have been recently reported to greatly reduce the inflammatory demyelinating lesions in a mouse model of neuromyelitis optica (86) and also to attenuate injury with a consequent neuroprotective effect in acute Guillain–Barré syndrome mouse models (87). A peptide agent (called 2J) was selected from a peptide library on the basis of C1q binding (88). This peptide was shown to bind to the globular domain of C1q and prevented the binding of C1q to IgG. The 2J peptide efficiently inhibited CP-mediated C4 and C3 deposition and MAC formation *in vitro*. Although this peptide was a promising candidate for therapeutic complement inhibition, no further studies were reported about its *in vivo* application.

Another possibility for inhibiting the CP and the LP is to disassemble the initiation complexes. In this respect, it is worth noting that in *in vitro* experiments the C1 complex dissociates in the absence of Ca2<sup>+</sup> (in the presence of EDTA), or at high ionic strength (1 M NaCl), whereas in the case of the MBL–MASP complexes, both conditions should apply at the same time (89). Moreover, C1 inhibitor, which makes covalent complexes with the SPs, dislodges C1r and C1s from C1q, while it cannot disassemble the MBL–MASP complexes (90). Nevertheless, it was shown that there is a dynamic equilibrium between the different MBL/ficolin–MASP complexes in human serum, in other words, MASPs can migrate between the complexes (91). Recently, it was shown that asparaginase, which is used in oncological treatments, inhibits the LP by reducing the amount of MBL–MASP complexes, very likely through dissociating the complexes (92). In this case, it is an adverse effect of the oncological treatment, but it indicates that a similar approach can be feasible in anticomplement therapy. An anti-MASP-2 monoclonal antibody (OMS721, Omeros), which binds to a non-catalytic complement control protein (CCP) domain of MASP-2, successfully inhibited the LP in *in vivo* experiments, and also it could disassemble the MBL-MASP-2 complexes.

There is a report about a viral-derived peptide (PIC1), which inhibits the classical pathway through binding to the collagenlike region of C1q in the C1 complex (93). This peptide might lock the conformation of C1q and/or displace the tetramer. There is no other report in the literature about an agent, which can block the conformational change necessary for the proper function of the initiation complexes. A deeper understanding of the activation mechanism of the C1 and MBL/ficolin– MASP complexes is needed to harness this possibility in the therapy.

#### MASP-1

MASP-1 is the most abundant protease of the LP, and it plays a central role in complement activation. Its average serum concentration is 143 nM (11 µg/mL), which is 24-times higher than the serum concentration of MASP-2 (6 nM, 0.4 µg/mL) (94). The members of the C1r/C1s/MASP protease family share the same domain organization (**Figure 2**). At the N-terminus, there is a CUB domain (initially recognized in *C*1r/C1s, sea urchin protein *U*egf, and human *b*one morphogenetic protein 1), followed by an epidermal growth factor-like (EGF) module and a second CUB domain. The MASPs are present as dimers in the circulation, and the N-terminal CUB1-EGF-CUB2 region is responsible for the dimerization. Another important function of this region is that it mediates the binding to the PRMs. The CUB and the EGF domains bind Ca2<sup>+</sup>, and both the dimerization and the PRM binding are Ca2+-dependent. The C-terminal region, which possesses the catalytic activity, consists of two CCP domains and a SP domain. The SP domain belongs to the chymotrypsin family (Family S1, MEROPS) and shows trypsin-like specificity cleaving after basic amino acids (Arg, Lys) in the polypeptide chain. The two CCP domains have at least two functions: they serve as spacers between the CUB1-EGF-CUB2 region and the SP domain and they provide additional binding sites (exosites) for the substrates (13). Both functions have essential roles in the activation of the PRM–MASP complexes and in the cleavage of the subsequent components (C2 and C4).

MASP-1 has multiple roles in the innate immune response. Zymogen MASP-1 has a high autoactivation capacity, which plays a key role in the activation of the lectin pathway (95). When PRM–MASP complexes bind to the activation surface, zymogen MASP-1 autoactivates and the active MASP-1 activates zymogen MASP-2 (9, 72). In this way, MASP-1 is the initiator protease of the LP. Recently, it has been demonstrated that MASP-1 significantly contributes to AP activation on LPS surface through an unknown mechanism (68). MASP-1 is also capable of activating endothelial cells by cleaving protease-activating receptor 4 (91, 96). The activated endothelial cells secrete cytokines (IL-6 and IL-8), and these cytokines promote the chemotaxis of neutrophil granulocytes (97). Moreover, MASP-1 treatment increased adhesion between neutrophils and endothelial cells by upregulating E-selectin expression in human umbilical vein endothelial cells (HUVECs) (98). A genome-wide gene expression profiling study on HUVECs corroborated the role of MASP-1 in triggering inflammation (99). The analysis showed that MASP-1 up- and downregulated numerous inflammation-related genes bridging complement activation and endothelial-cell-related inflammatory processes. It was also demonstrated that MASP-1 is able to cleave high-molecular-weight kininogen and liberate bradykinin (100). Bradykinin is a potent vasoactive, pro-inflammatory peptide, which is responsible for the swelling attacks in hereditary angioedema (HAE), a disease associated with C1 inhibitor deficiency (101). Uncontrolled activation of MASP-1 may contribute to the development of HAE attacks and worsening the symptoms of HAE patients. It was also recognized that MASP-1 serves as a link between the complement and the coagulation cascades. MASP-1 promotes coagulation by activating prothrombin, fXIII, and thrombin-activatable fibrinolysis inhibitor (102–104). The effect of MASP-1 on blood coagulation was confirmed by using a microvascular whole-blood-flow model (105). The physiological relevance of this phenomenon is not quite clear; however, it is very likely that the proteolytic activity of MASP-1 contributes to pro-inflammatory and pro-thrombotic events facilitating the development of thrombotic complications under pathological conditions (106). As the above examples highlight, MASP-1 has a relatively broad substrate specificity (it has about 10 known substrates), which is quite unusual among complement proteases. It should be noted, however, that all the known substrates of MASP-1 are related to the innate immune response. Evolutionary considerations indicate that MASP-1 is an ancient enzyme of the complement system compared to the other members of the MASP/C1r/C1s family (107). The relaxed substrate specificity of MASP-1 is reflected in its 3D structure (70). The substrate-binding groove of MASP-1 is broad and accessible resembling that of trypsin, rather than those of other early complement proteases. The physiological inhibitors of MASP-1 are serpins. C1 inhibitor, and in the presence of heparin antithrombin, attenuate very efficiently the activity of MASP-1 (108). Alpha2-macroglubulin, a pan-specific protease inhibitor in the blood was suggested to inhibit MASP-1 and consequently the LP (109), but this issue is controversial (108, 110). Another potential physiological inhibitor of the LP is MAp44 (aka MAP-1), an alternative splice product of the *MASP1* gene (111, 112). MAp44 contains the CUB1-EGF-CUB2-CCP1 domains of MASP-1/3 plus a 17 amino-acid-long C-terminal peptide. Since MAp44 lacks the SP domain, it does not have proteolytic activity to initiate the LP, but it can dimerize and bind to the PRMs like the MASPs. MAp44 attenuates LP activity by competing with MASP-1 and MASP-2 for the PRMs and displacing them from the complexes. Recombinant MAp44 was shown to protect against myocardial IRI in mouse models, preserving cardiac function, decreasing the infarct size, and preventing thrombogenesis (113). Recombinant chimeric inhibitors were also designed and constructed by fusing MAp44 and the complement regulatory domains (1–5) of FH (114). One of these inhibitors showed simultaneous inhibition of the LP and AP.

Theoretically, the SPs are the most druggable targets in the complement system (115). The active sites of these enzymes can be easily targeted by small-molecule protease inhibitors. The main problem with this approach is the lack of specificity, since all the complement proteases and also the proteases of the other plasma cascade systems contain chymotrypsin-like SP domains (**Figure 4**). A small-molecule SP inhibitor, which blocks the activity of a particular complement SP, very likely will inhibit other complement proteases, as well as proteases of the coagulation, fibrinolysis, and kallikrein–kinin systems to some extent. For example, nafamostat mesilate (FUT-175 or Futhan) is a powerful inhibitor of the complement cascade, but it has a broad specificity. It was shown to attenuate renal and myocardial IRI (116, 117), but it also attenuates pancreatitis by inhibiting trypsin and other pancreatic enzymes (118), and also coagulation by inhibiting thrombin and other clotting enzymes (119). To enhance the specificity, the number of interactions should be increased between the SP and the inhibitor. A promising approach could be the fragmentbased drug discovery, which generates highly specific molecules *via* linking small chemical fragments (Mw < 300 Da) together that

bind only weakly on their own to the target. This approach was successfully used to develop specific small-molecule inhibitors against FD (120) (**Figure 5**), but there is no report about similar molecules against MASPs. Monoclonal antibodies and other biologics can also meet the specificity criterion. Highly selective MASP inhibitors were developed by the *in vitro* evolution of the interacting loop of canonical SP inhibitors. Sunflower trypsin inhibitor (SFTI) is a 14-amino acid-long cyclic peptide, which mimics the protease-interacting loop of the inhibitor scaffold of the Bowman–Birk inhibitor family. SFMI-1, an LP-selective peptide inhibitor was developed by phage-display selection of SFTI variants using MASP-1 as target (121). SFMI-1 proved to be a strong MASP-1 inhibitor (*K*<sup>i</sup> = 65 nM), and a weak MASP-2 inhibitor (*K*<sup>i</sup> = 1,030 nM). In order to further increase the specificity, a larger inhibitor scaffold was used in the phage-display selection. SGPI-2 (*S. gregaria* protease inhibitor-2) is a single domain small-protein inhibitor (35-amino acid-long) belonging to the Pacifastin family of canonical inhibitors. After randomizing six positions in the protease-interacting loop (P4, P2, P1, P1′, P2′, and P4′), a highly specific MASP-1 inhibitor (SGMI-1) was selected (122) (**Figure 4**). SGMI-1 inhibits MASP-1 very effectively (*K*<sup>i</sup> = 7 nM), and very selectively. This inhibitor was used to clarify the function of MASP-1 in the innate immune response using numerous *in vitro* and *ex vivo* assays. Although MASP-1 is a tempting target to halt unwanted LP activation and to prevent various pro-inflammatory processes, no pharmaceutical development of a MASP-1 inhibitor has been reported to date.

#### MASP-2

MASP-2 is the only protease in the LP that can cleave C4. Its serum concentration is rather low (6 nM, 0.4 µg/mL), compared to the other complement proteases. These characteristics make MASP-2 an ideal target to inhibit pathological LP activation.

MASP-2 has identical domain organization with MASP-1 and MASP-3 (**Figure 2**). Isolated MASP-2 has a tendency to autoactivate in a concentrated solution (11, 124). This autoactivation capacity, however, cannot manifest in normal human serum, where the MASP-2 concentration is low, and each MASP-2 molecule is surrounded by MASP-1 molecules on the target surface. Under these circumstances, MASP-1 is the exclusive activator of MASP-2. The autoactivation ability of MASP-2 might be important in situations, where there is no MASP-1 present (e.g., MASP-1 deficiency). It should be noted, however, that in the serum of a 3MC syndrome patient, where there was neither MASP-1 nor MASP-3 present due to a mutation in the *MASP1* gene, no LP activity could be detected (72). On the other hand, birds lack MASP-1, but have functional LP, suggesting that MASP-2 can independently drive LP activation (125). In the sera of these animals, however, the autoactivation capacity of MASP-2 must be much higher than that of human MASP-2 in normal human serum. A recent publication shows that MASP-2 can directly cleave C3 in the absence of C4 and/or C2 on LP-activating surfaces (66). MASP-2 was also suggested to promote fibrin polymerization by cleaving prothrombin (126). The *MASP2* gene, like the *MASP1* gene, has an alternative splice product MAp19 (aka sMAP, MAP-2) (127, 128). This truncated gene product contains only the CUB1 and EGF domains plus 4 unique C-terminal residues. Since MAp19 can bind to the PRMs, it may regulate LP activity through displacing the MASPs from the complexes. Theoretically, the recombinant form of MAp19 could be suitable to attenuate LP activity, in practice, the larger MAp44 was used for this purpose since it binds to the PRMs with higher affinity.

The pathological relevance of MASP-2 was demonstrated in MASP-2 knock-out mice, where the animals were significantly protected against myocardial and gastrointestinal IRI (50). In the hearts of MASP-2-deficient mice, the infarct volume was significantly smaller than in those of the wild-type animals. Moreover, a recent study demonstrated that an inhibitory monoclonal anti-MASP-2 antibody successfully attenuated myocardial IRI in wild-type mice (129). An anti-MASP-2 antibody, OMS721, developed by Omeros Corporation, is under clinical trial for treating aHUS (130) and other thrombotic microangiopathies (131), IgA nephropathy, lupus nephritis, membranous nephropathy, and C3 glomerulopathy (132). The mechanism of the protecting effect of MASP-2 inhibition in these diseases is not clear, since AP activation is believed to be the main driver of these conditions. Selective canonical inhibitors against MASP-2 were also selected by phage-display using the SFTI and SGPI scaffolds (121, 122). Both inhibitors were highly specific: SFMI-2 (*K*<sup>i</sup> = 180 nM) and SGMI-2 (*K*<sup>i</sup> = 6 nM) prevented LP activation efficiently, while they did not compromise the activity of the other two pathways.

#### MASP-3

MASP-3 was discovered as the third SP component of the LP (133). It has the same domain organization (**Figure 2**) as MASP-1 and MASP-2, as described above; moreover, the amino acid sequence of its first five domains is identical with that of MASP-1. This feature is the consequence of the fact that MASP-1 and MASP-3 are the alternative splice products of the same *MASP1* gene, along with a third protein MAp44 (111, 112). Variants of the *MASP1* gene, resulting in the loss of the activity of MASP-3, cause the 3MC syndrome, characterized by serious craniofacial, genital, and often mental defects (58, 59, 134). The results indicate that MASP-3 is involved in neural crest cell migration in early embryonic development. Interestingly, the same phenotype is observed in patients carrying mutations in the *COLEC11* gene (58) or the *COLEC10* gene (135), both encoding LP components CL-K1 (aka collectin-11) and CL-L1 (aka collectin-10). It is possible that MASP-3 is in complex with CL-K1/L1 when it exerts its function during embryogenesis, and it is likely that the proteolytic activity of MASP-3 plays an important role.

MASP-3 is different from MASP-1 and MASP-2 in several ways. MASP-3 does not autoactivate, does not cleave downstream LP/CP components, C4 and C2, and the active form has very low activity on most synthetic substrates (136). *In vitro*, it was shown to cleave insulin-like growth factor-binding protein 5; however, the relevance of this reaction is uncertain (137). It has also no natural inhibitor in the blood; therefore, control of its activity is probably achieved simply by its very restricted substrate specificity. MASP-3 is present in the blood as the mixture of the proenzymic and the activated forms; moreover, the activated form seems to be the more dominant variant (76). In this aspect, it also differs from MASP-1 and MASP-2, which are proenzymic. On the other hand, in many regards, MASP-3 has similarities to FD, which circulates predominantly in the active form, has no natural inhibitor, and has very restricted substrate specificity.

The function of MASP-3 in the blood had been mysterious until recently. Initially, it was considered simply as a negative regulator of the LP since it competes with MASP-1 and MASP-2 for binding to PRMs (133). This function may still be valid; however, now, strong evidences exist that the active form of MASP-3 is the primary physiological activator of pro-FD, producing FD, a key enzyme of the AP. The story was detailed in a previous section; therefore, we jump to the functional consequences of this activity.

Dobó et al. Targeting AP and LP Components

It seems logical that the activity of the AP can be downregulated by the inhibition of MASP-3. Inhibition of MASP-3 would result in the accumulation of pro-FD in the blood with only very low levels of active FD, hence greatly attenuating AP activity. A study presented at 16th European Meeting on Complement in Human Disease (138) provided a strong evidence for this assumption. A single dose of a monoclonal antibody inhibiting the activity of MASP-3 suppressed the activity of the AP and shifted the active to zymogen ratio of FD toward the proenzyme, pro-FD, both in mice and in cynomolgus monkey. So far, two specific inhibitors against MASP-3 were developed. One of them is a canonical Kunitz-type recombinant protein, which is based on the second domain of tissue factor pathway inhibitor (TFPI) and developed by phage-display (75). The other is the abovementioned monoclonal antibody by Omeros Corporation (138).

What are the potential advantages of MASP-3 inhibition over FD inhibition? Inhibition of both proteins is expected to result in similar systemic effects in the blood. The plasma concentration of both proteins is similar, around 60 nM. This relatively low value is attractive for drug development. On the other hand, FD has a very high turnover rate. Its half-life in humans is less than 1 h (139). The turnover rate of MASP-3 is not yet known, but because of its size, it is most certainly lower compared to FD. This could mean that a lower daily dose of a drug candidate inhibitor of MASP-3 would be required compared to a FD inhibitor.

Deficiency in the AP can result in potentially life-threatening meningococcal infections, and AP inhibition carries the same risk. Another potential benefit of MASP-3 inhibition would be that in this case, a pro-FD pool is still available. In case of a bacterial infection, the LP can be activated, and the resulting active MASP-1 and MASP-2 molecules could locally convert pro-FD to FD, making the AP amplification possible. Nonetheless, this mechanism needs experimental validation.

In all, based on MASP-3's requirement for the maturation of pro-FD, MASP-3 presents itself as a good target to attenuate the complement system, with several potential benefits over FD inhibition.

#### Factor D

Factor D (FD) is a single domain SP, which circulates in the blood predominantly in the active form (77, 140). It is synthesized mainly by adipocytes, hence the alternative name adipsin. In the 1970s, it was debated whether it is produced as a proenzyme or secreted in the active form (140, 141). Since only the active form could be isolated from blood (142), it was assumed that it might be activated even before secretion (143). Nevertheless, at the DNA level, after the signal sequence, an additional 5 to 7 amino acid long propeptide is encoded. Now the consensus is that active MASP-3 converts the pro from of FD to the active form constitutively (74, 75, 77).

Although it is an active SP, FD has an extremely restricted substrate specificity. It has very low activity toward synthetic substrates, basically, it cleaves only certain thioester compounds; however, its natural substrate, FB in complex with C3b or C3blike molecules, is cleaved very efficiently (144). The free enzyme's very low activity is due to a unique self-inhibitory loop (145), which is displaced when FD binds to C3bB (146).

It has a relatively low mass concentration of 1–4 µg/mL in humans (147–149), which in combination with early reports showing that FD is the bottleneck of AP activity (140), led to the assumption that FD could be the best target to achieve AP inhibition. However, FD is a small protein of only 25 kDa, so, its molar concentration of 40–160 nM combined with its high turnover (139) suggest that high daily doses of a FD inhibitor would be required to achieve complete sustained inhibition. Recent results also suggest that FD may not even be the bottleneck of AP activity. In the serum of FD-deficient mice, the addition of FD corresponding to only about 1–2% of the normal FD level was sufficient for normal AP activity *in vitro* (147). In a 3MC syndrome patient, whose serum contained mostly pro-FD, some AP activity was still present (72), although lower than the normal level (134). These data together suggest that even if a potent and specific inhibitor is used, at least equimolar amount is required for AP inhibition, and even higher doses are necessary for sustained inhibition. A study with lampalizumab, a humanized monoclonal anti-FD IgG Fab fragment, showed similar observations (150).

One must also consider that, at least *in vitro*, plasma kallikrein was shown to be able to cleave the C3bB pro-convertase (151), hence a residual, low-level AP activity might still be present even during complete FD inhibition, or FD deficiency; however, the *in vivo* relevance of this cleavage needs further validation.

Nevertheless, FD remains a prime target within the complement system. Several FD inhibitor molecules are under development, or in the clinical trial phase (152, 153) for PNH, aHUS, and AMD. Achillion developed several small molecule FD inhibitors that may be orally administered. A dose of 200 mg/kg of ACH-4471 per every 12 h resulted in complete AP inhibition in primates (154). An example of the combination of structurebased and fragment-based drug development targeting FD was published recently. Modifying the structure of a small-molecule kallikrein inhibitor several compounds were developed that selectively inhibit FD (120). **Figure 5** shows FD in complex with one of the compounds as an example.

Near complete inhibition of FD is expected to have a similar outcome as inhibition of MASP-3. Neisserial infection or other bacterial infections constitute a possible threat, which requires prophylactic treatment or treatment with antibiotics. This is actually valid for nearly all kinds of anticomplement drugs.

#### Factor B

Factor B (FB), a five-domain, 90 kDa glycoprotein, is composed of three CCP modules, a short connecting segment, a von Willebrand factor type A (vWFA) domain, and an SP domain (**Figure 2**). It circulates as a proenzyme, and its activation site (Arg234-Lys235) is hindered in the free enzyme from the cleavage by FD. FB can form a complex with C3b, or C3b-like molecules, to generate the AP pro-convertase, C3bB. The pro-convertase probably exists in two, closed and open conformations, in the latter, FB being accessible for FD cleavage (155). The FD-C3bB interaction facilitates both a shift toward the open conformation of C3bB, and a structural rearrangement in FD displacing its self-inhibitory loop (146). FD cleaves FB in the pro-convertase to release the Ba fragment. The other fragment, the catalytic Bb itself is still just a marginally active enzyme (156), it has full activity only as part of labile C3bBb complex (157). Once Bb dissociates from the convertase complex, it cannot re-associate with C3b (157).

FB is absolutely essential for the AP; therefore, it is a prime target for AP inhibition, but because of its high concentration (about 200–250 µg/mL, or 2–3 µM), it might not seem to be ideal at first sight. On the other hand, in order to prevent AP activation, only the newly formed C3bBb complexes may have to be inhibited. Using a potent inhibitor with low Kd toward C3bBb could completely block the amplification phase, thereby halting the activation process. While C3bBb might be a difficult target for testing small-molecule inhibitors, because of the transient nature of this complex, the cobra venom factor (CVF)-Bb complex is more stable; therefore, it presents itself as a viable target for the development of such molecules. It is also notable that at high pH (proenzymic), FB alone has significant, easily detectable activity toward C3 and certain para-nitroanilide substrates (158). Several substrate-analog aldehyde FB inhibitors were developed along the way (158).

Inhibitory antibodies might be more easily obtained. They only need to prevent access to C3, the very large substrate of C3bBb, which is attainable by a bulky antibody molecule binding near the catalytic site. However, it is possible that such antibody would also bind to free FB; therefore, higher doses might be required. Optimally, a small-molecule inhibitor or an inhibitory antibody should only bind to Bb, or even better only to the C3bBb complex, so that a relatively low dose of the molecule be sufficient for complement inhibition. A blocking antibody, binding to free FB, which prevents the formation of the pro-convertase complex, is also a feasible option. In this case, also high doses would be required for optimal effect.

A set of small-molecule inhibitors are under development by Novartis against FB (CVF-Bb) for indications such as AMD and other complement-mediated diseases (159, 160). Neutralizing monoclonal antibodies against Bb by Novelmed Therapeutics are under development for various indications (161, 162). A monoclonal antibody to mouse FB has been shown to be protective in a mouse model of renal IRI (163). Other approaches using antisense oligonucleotides (164), or a phage-display selected cyclopeptides (165) are other feasible options to control AP activation through FB.

While FB is a promising target, so far, no therapeutic agent hit the market, or is in the advanced state in clinical trials. As with FD or MASP-3 inhibitors, bacterial infections manifest a potential risk when patients are treated with FB inhibitors.

#### C3 and CVF

C3 is the central molecule of complement; the three activation routes are merged at the generation of C3b and continue together as the terminal phase (**Figures 1** and **3**). C3 circulates in the serum at high concentration (4–7 µM; 0.75–1.35 mg/mL). Native C3 is a 185 kDa protein containing 13 domains (**Figure 2**). C3 is composed of two chains, α and β. The core is built up of 8 domains belonging to the α2-macroglobulin family. The thioester domain carries a buried thioester bond, which is prone to suffer hydrolysis or other nucleophilic attack.

Primary C3 deficiencies were described in a few families over the world. Mutation in C3 gene caused impaired C3 synthesis or secretion, which produced a low C3 level in the blood. These individuals are extremely susceptible to recurrent pyogenic bacterial infections, especially to Gram-negative but also to Gram-positive bacteria (166). Moreover, C3 deficiency impairs maturation of immune cells (dendritic cells, memory B cells, certain T cells) (166, 167). Furthermore, SLE and various renal diseases were also observed; however, their mechanism is not fully understood. Secondary C3 deficiency is due to malfunctioning of the complement regulatory proteins, typically FI and FH (168).

Complement activation can be blocked completely at the level of C3. On the other hand, C3 is the most abundant protein in the complement cascade; therefore, a large amount of an inhibitor would be needed to achieve a substantial effect. Compstatin, a promising complement-based therapeutic agent, was developed against C3 by phage-display using naïve library in 1996 (169). Compstatin is a cyclic peptide of 13 amino acids with a single disulfide bond. It blocks the access of the convertase to C3 through steric hindrance. Crystal structure with C3c showed that compstatin forms expansive H-bonds with its partner (170) (**Figure 6**). Neither complement regulator proteins nor other structurally related proteins (C4, C5) bind compstatin. In the past 20 years, the compstatin family has been constantly developed. New generations of compstatin analogs possess better

FIGURE 6 | C3c in complex with a compstatin analog. Compstatin, a cyclic peptide, was developed by phage-display. Since its discovery, several modified compstatin analogs have been developed. Compstatin and its analogs bind to C3, C3b, or C3c between the MG4 and MG5 domains. Compstatin sterically prevents the C3-convertase (C3bBb) to access its substrate C3. The depicted structure was determined using the Ac-V4W/ H9A-NH2 variant of the original peptide. The figure was prepared based on the structure by Janssen et al. (170) (PDB entry 2QKI). On the left, the whole structure is shown with C3c (brown) in surface representation and compstatin (magenta) with spheres. On the right, a close-up of the binding site is shown with compstatin represented by sticks. Hydrogen bonds are indicated by yellow dashed lines.

pharmacokinetic and pharmacodynamic features. Compstatin derivatives were investigated in many complement-related animal diseases models and showed promising results (171). Just to mention a few, compstatins are efficient in primates in inflammatory diseases induced by cardiac surgery, cardiopulmonary bypass, or *E. coli* infection, in treatment of organ transplantation to reduce the possibility of xenograft rejection, and in sepsis. One of the compstatin derivatives (APL-2) already completed a Phase II clinical trial in treatment of AMD by Apellis Pharmaceuticals, and another one (AMY-101) started in 2017, the "first-in-human" clinical study against PNH by Amyndas Pharmaceuticals. Certainly, compstatins, as a peptide drug candidates, have their limitations especially considering oral administration. The rapid proteolytic degradation and poor biocompatibility make drug formulation challenging. On the other hand, the high specificity, the relatively low cost of production, and high variability gives the compstatin family members great potential to become widely used, effective, and safe complement therapeutics (171).

There are some other approaches to target complement activation through C3. Since 1970s, CVF is widely applied to deplete complement and to gain knowledge about its role in diseases. CVF is a structural and functional analog of C3, forms an AP convertase with Bb; however, it is resistant to the activity of FI and FH. Since decay of the convertase is abolished, C3 and C5 are rapidly exhausted from the blood. Nevertheless, CVF is immunogenic; therefore, it can be used only once to avoid antibody response. In the last decade, interesting results have been published about a chimeric protein, humanized CVF (172). It is a C3 derivative obtained by simply replacing the C-terminal part with the homologous sequence from CVF. This protein is safe and proved to be efficient in various animal disease models (AMD, collagen-induced arthritis, PNH, myasthenia gravis, etc.), and furthermore, no neutralizing antibody effect was detected in mice after prolonged usage (173).

#### Properdin

The only known positive regulator of the AP is properdin, also referred to as factor P. Properdin circulates in the plasma at 20–125 nM (4–25 µg/mL) concentration as a cyclic polymeric glycoprotein. In contrast to most complement proteins, it is synthesized primarily by leukocytes and shows different activity depending on the type of producing cells. The properdin monomer comprises six complete and one truncated thrombospondin type 1 repeat (TSR) domain in tandem connection. The 53 kDa monomer is able to form dimers, trimers, and even tetramers in a head-to-tail arrangement. Physiologically, the most abundant form is the trimer; however, properdin shows tendency to selfaggregate into higher oligomers under conditions used for its preparation. It has an extremely high positive charge, hence it tends to bind *via* ionic interactions to polyanion structures.

Properdin has a significant and well-established role in the AP of complement by stabilizing the very labile C3bB and C3bBb complexes offering binding sites to C3b, and FB or Bb. Extending the half-life of the AP convertase by 5- to 10-fold is essential for the effective AP activity (174). Another role of properdin, serving as a PRM, was proposed about 10 years ago. A similar function was originally suggested by Pillemer, who discovered the "properdin pathway." However, findings of this subject are controversial. Caution must be taken since repeated freezing and thawing resulted in highly polymerized, therefore, non-physiological, aggregated properdin, which binds non-specifically to surfaces. Experiments using unfractionated properdin could have led to physiologically not relevant observations (175). The binding abilities of properdin are also influenced by the contact surface and the presence of specific ligands. Experiments in properdin knock-out mice demonstrated that in the absence of properdin, bacterial LPS- and lipooligosaccharide-induced AP activation was absent, while zymosan or CVF-induced activation was only partially affected (79). Using compstatin and anti-FP antibody in ELISA assays, it has been shown that properdin does not attach directly to zymosan or *Escherichia coli* surfaces, but it contributes only to the stabilization of C3bBb complex (176). Recent studies showed that the binding of properdin to activating surfaces is always preceded by deposition of C3b (23), concluding that properdin can act as an initiator of AP only in a C3b-dependent manner.

Properdin deficiency especially in combination with the lack of other complement components (MBL, C2, etc.) causes unequivocal susceptibility to bacterial infections. Disorder in properdin and one of the late complement components (C5–C9) increases the risk of *Neisseria meningitidis* infection by 1,000- to 10,000-fold (28). Interestingly, the lack of IgG G2m(n) allotype in properdindeficient persons also increased the susceptibility to meningococcal disease (177).

Therapeutic application of properdin emerged recently. In mouse model, it has been shown that a highly polymerized form of recombinant properdin gives protection against *N. meningitidis* and *Streptococcus pneumoniae*. A single low-dose treatment was enough to boost complement-activated lysis, which significantly reduced bacteremia and increased survival rates (178). We should note, however, that the recombinant properdin had histidine tag, which could influence its antimicrobial activity (179). Newly developed mouse monoclonal antibody against properdin was proved to be useful in sandwich-ELISA system to determine serum level of properdin in human samples (180). This antibody also successfully blocked AP activation in human sera.

The lack of properdin can efficiently abolish physiological or even unwanted AP activity. Inhibiting the AP through properdin can abate the amplification of deposited C3, hereby, the activity of the complement cascade. In contrast to C3, properdin is present at a relatively low concentration. Consequently, as an important regulator of the AP, it may turn out to be a promising therapeutic target to block complement activation. Novelmed is developing new drug candidates against the components of the AP, which do not interfere with the CP, as part of their effors, they evolved a new monoclonal antibody against the N-terminal fragment of properdin (181).

#### Factor I

Factor I has an outstanding role in the control of the complement system. Along with its cofactors, FI belongs to the regulators of complement activation. Although it possesses a low catalytic activity on its own, FI can downregulate all activation routes by dismantling the central component of the complement cascade: C3b (and also C4b). FI contributes to the self-defense of host tissues against complement damage through the acceleration of the decay of fluid-phase and surface-bound C3b to iC3b. Degradation products of C3b initiate the cellular immune response *via* their interaction with various receptors on immune cells.

FI is an 88 kDa glycoprotein synthesized as a single polypeptide chain by hepatocytes. It is a trypsin-like SP consisting of five domains; some of them are common in the components of the terminal pathway. The heavy chain contains the first four domains: FI membrane attack complex domain, CD5-like domain, low-density lipoprotein receptor 1 and 2 (LDLr 1 and 2) domains, and a small section called d-region with unknown homology (**Figure 2**). The light chain, which is attached to the heavy chain by a disulfide bond consists of the catalytically active SP domain.

FI has several unusual characteristics: it circulates as an "active" enzyme in the blood and does not have an inhibitor (19). These and some other features of FI resemble that of two other proteases of complement, Factor D (182) and MASP-3. FI has an extremely low catalytic activity toward synthetic substrates and also toward free C3b and C4b. In order to cleave C3b and C4b efficiently, FI needs cofactors. C4b-binding protein (C4BP) and FH are soluble cofactors of FI, while MCP (CD46) and CR1 (CD31) are membrane-bound cofactors. Since no natural inhibitor of FI is known, it is regulated by other mechanisms. First of all, the type of the substrate and also the type of its cofactor influences the activity of FI, and also the cleavage site on C3b or C4b and their degradation products. Second, structural data has proven recently that many crucial loops of the SP domain are disordered without the interacting partners (183). As the ternary complex of C3b-FH-FI is formed, ligand binding induces stabilization of the SP domain and, therefore, FI obtains full proteolytic activity. After the cleavage of the first bond in C3b (**Figure 7**), the substrates rearranges and the second or third cleavage site becomes accessible, while the SP domain of FI endures only minor movements (184).

According to its important role in complement regulation, the absence of FI causes dangerous, even life-threatening conditions. Due to the lack of decay acceleration, increased amounts of C3b lead to uncontrolled AP activation. The more C3b molecules are present, the more C3 convertases are generated, which results in the rapid exhaustion of C3 from the plasma. Individuals with FI deficiency are prone to suffer from recurring bacterial infections, severe kidney diseases, and most of all AMD (185, 186). Recent studies show that identifying rare CR1 variants in combination with low serum level of FI can enable therapist to find patients, who are the most likely candidates to develop AMD (187). FI deficiency is often associated with aHUS. Symptoms frequently appear in early childhood after a severe infection or in young females shortly after pregnancy. Large international cohorts have been established to characterize all genetic variants and clinical outcomes. The prognosis of FI-associated aHUS is quite poor, in half of the cases, end-stage renal failure developed rapidly. Treatment with eculizumab, which is the major therapeutic for aHUS, resulted in partial remission in patients having FI-associated aHUS (188).

Another aspect that makes FI a potential drug candidate is the phenomenon of multiple polymorphisms in complement components that can affect the delicate balance between activation and regulation of an individual's complement system. The inherited repertoire of the complement gene variants was dubbed complotype (189). Some variant alleles can result a more reactive complement, which usually appears in the increased activity of C3b feedback cycle. Hyperactive complotypes raise significantly the risk of complement-related diseases at later age. Lower serum level of FI compared to FH (along with FH-related proteins) is advantageous to produce an effect. Moreover, the administration of FI in the presence of cofactor CR1 also enhances the conversion of inflammatory product iC3b to C3dg (190). Experimental data prove that increasing the amount of FI in serum of different complotypes can convert higher-risk to lower-risk activity. The extra amount of FI needed is approximately 50% of normal level, which would be a useful therapy in such patients (191). Comprehensive characterization of the complement regulatory genes in patients already suffering from complement-related disorders would enhance developments of personal and successful therapies.

by ribbon representation.

### Factor H, FH-Like, and FH-Related Proteins

Factor H (FH) is the major regulator of the AP. It is a fluid-phase molecule; however, it can bind to surface-deposited C3b and regulate the AP C3 convertase by several ways. By binding to C3b, it can prevent the capture of FB, consequently, the formation of the pro-convertase (C3bB). It has also a convertase decayaccelerating activity by facilitating the irreversible dissociation of the C3bBb complex. Probably, the most important function of FH is the cofactor activity, which is necessary for the FI-mediated cleavage of C3b to iC3b, through which it prevents the build-up of the amplification feedback loop of the AP. FH is a glycoprotein of 115 kDa and it consists of 20 CCP (aka short consensus repeat or sushi) domains (**Figure 2**). These domains, which are about 60 residues in length and contain two highly conserved disulfide bonds, are widespread among the complement proteins. Many complement regulatory proteins, such as FH and the FH-related proteins, CR1, CR2, MCP, DAF, C4BP, are composed predominantly or exclusively of these repeating structural motifs. The four N-terminal CCP domains of FH are responsible for the convertase decay-accelerating and cofactor activity. The other CCP domains take place in the interaction with different ligands. The C-terminal CCP 19–20 domains are indispensable for binding to self-surface deposited C3b. According to the current knowledge, FH recognizes the juxtaposition of C3b and carbohydrates containing sialic acid or glycosaminoglycan on the surface and binds strongly through the CCP 19–20 domains. The other domains may contribute to the binding to several ligands (e.g., heparin/CCP 7), but they are not indispensable for the function of FH. Based on this knowledge, minimal-size FH molecules were designed by combining the N- and C-terminal regions. Two constructs contain only six domains (CCP1-4 and 19-20) (192, 193). These mini FH molecules (**Figure 7**) showed more effective complement inhibition in different assay systems than the full-length FH molecule (192). Another, slightly extended construct containing CCP1–5 and 18–20 domains effectively inhibited complement activation *in vivo* and reduced abnormal glomerular C3 deposition in a FH-deficient mouse model of C3 glomerulopathy (194). Recently, a monoclonal anti-FH antibody has been found that could inhibit AP activation by potentiating FH (195). This potentiating antibody increases the affinity of FH for C3b and facilitates the degradation of the convertase by FH. There is an alternative splice product of the FH gene, which consists of only the CCP1-7 domains plus a four-amino-acid long C-terminus (196). This FH-like protein 1 (FHL-1) has complement-inhibitory activity, and it may have important function in the periphery. It is supposed that FHL-1 is able to penetrate through the Bruch's membrane beneath the retinal pigment epithelial cells in the eye, while FH cannot. In this way, FHL-1 may have a crucial role in the protection of retinal cells against complement-mediated attack and prevention of the development of AMD (197). Besides FH and FHL-1, there are five FH-related proteins (FHR) in the human serum. These proteins are encoded by separate genes situated next to the FH gene, and these genes arouse very probably through partial gene duplications. These proteins are shorter than FH and usually consist of CCP domains homologs to CCP6-9 and CCP18-20 of FH. Since the FHRs lack the complement regulatory CCP1-4 domains, their physiological relevance was underestimated at the time of their discovery (198). Since then, increasing number of evidences have been accumulated demonstrating the physiological role of FHRs, although this area is still controversial. Since the FHRs contain domains sharing high sequence identity with CCP18-20 of FH, these proteins can bind to ligands of FH (e.g., C3b, heparin, CRP).


*NA, not applicable, because inhibition of negative regulators is generally not desirable; however, potentiation could be a feasible approach (195).*

However, these molecules cannot efficiently inhibit the AP since they lack the N-terminal regulatory domains of FH (CCP1-4).

FHR-1 was reported to enhance, rather than to inhibit complement activation through binding to CRP (199). This phenomenon could explain the protective effect of FHR-1 deficiency in AMD (200). FHR-4 was able to facilitate AP and CP activation by binding to C3b and CRP, respectively (201). FHR-5 was also shown to promote complement activation by binding to pentraxin 3 (PTX3) and extracellular matrix and by enhancing C1q deposition (202). It is also possible that FHRs compete with FH on the surface of bacteria, thereby compromising the ability of the microorganism to evade complement-mediated attack (203). It has been demonstrated that FHR-3 acts as a decoy, being captured by *N. meningitidis* cells instead of FH (204). The level of protection against *N. meningitidis* infection may depend on the FH/ FHR-3 ratio in the serum. In general, the serum concentration of FH and the FHRs, and their affinity to various ligands may be a key factor in the fine tuning of complement-mediated opsonization and inflammation. If the delicate balance between FH and FHRs is disturbed due to genetic variations, or the amount and the composition of the ligands changes in the course of a disease (infection, oxidative stress), improper complement activation can take place resulting in self-tissue damage.

#### CONCLUDING REMARKS

The complement system was an appealing drug target even in the 1970s. However, the early drug development efforts failed mainly because of two reasons. The first reason was the lack of specificity of the anticomplement compounds. At that time, there was no technology to design or select highly specific agents against the individual complement components. The advance of structurebased and fragment-based drug design approaches made possible to generate selective and efficient small-molecule drugs.

#### REFERENCES


In addition to that, the modern biotechnological methods have provided highly specific biologics [monoclonal antibodies, recombinant proteins, nucleic-acid aptamers (205, 206), etc.] developed for anticomplement therapy. The most successful anticomplement drug so far, eculizumab, is a monoclonal antibody, and many antibodies are in preclinical or clinical phase in the pipeline. The second reason, which hindered the introduction of anticomplement drugs in the clinical practice in the past, was the insufficient knowledge about the mechanism of action of complement in both health and disease. In the recent years, the mechanism of activation and regulation of the LP and AP has been revealed in more detail, and we got insight into the cross-talks between the individual pathways inside the complement system and also the cross-talks between the complement and other proteolytic cascade systems (e.g., coagulation). New discoveries have also been made about the role of complement in the regulation of the adaptive immune system. Based on all of the above mentioned scientific and technical advances, essentially, all components of the LP and AP became targets of drug development (**Table 1**). It is likely that new drugs with more efficiency and less adverse effect will be approved for treating complement-related disorders in the near future.

# AUTHOR CONTRIBUTIONS

All authors contributed equally to this article.

#### FUNDING

The study was supported by the National Research, Development and Innovation Office (NKTH) OTKA grants K108642 and K119374, and by the MedInProt program of the Hungarian Academy of Sciences.


availability of hydroxyl groups. *Biochemistry* (1984) 23:3267–72. doi:10.1021/ bi00309a022


glomerulonephritis, and C3 glomerulopathy: core curriculum 2015. *Am J Kidney Dis* (2015) 66:359–75. doi:10.1053/j.ajkd.2015.03.040


COLEC11 and MASP1 cause 3MC syndrome. *Nat Genet* (2011) 43:197–203. doi:10.1038/ng.757


protease-3: kinetic modeling of lectin pathway activation provides possible mechanism. *Front Immunol* (2017) 8:1821. doi:10.3389/fimmu.2017.01821


expressed in heart and skeletal muscle tissues and inhibits complement activation. *J Biol Chem* (2010) 285:8234–43. doi:10.1074/jbc.M109.065805


for the treatment of an advanced form of dry age-related macular degeneration. *J Pharmacol Exp Ther* (2014) 351:527–37. doi:10.1124/jpet.114.215921


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Dobó, Kocsis and Gál. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Chimeric Proteins Containing MAP-1 and Functional Domains of C4b-Binding Protein Reveal Strong Complement Inhibitory Capacities

Cecilie E. Hertz 1†, Rafael Bayarri-Olmos <sup>1</sup> \* † , Nikolaj Kirketerp-Møller <sup>1</sup> , Sander van Putten<sup>2</sup> , Katrine Pilely <sup>1</sup> , Mikkel-Ole Skjoedt <sup>1</sup> and Peter Garred<sup>1</sup>

<sup>1</sup> Laboratory of Molecular Medicine, Department of Clinical Immunology Section, Rigshospitalet, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark, <sup>2</sup> Finsen Laboratory, Rigshospitalet, Biotech Research and Innovation Center (BRIC), Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark

#### Edited by:

Thomas Vorup-Jensen, Aarhus University, Denmark

#### Reviewed by:

Lourdes Isaac, Universidade de São Paulo, Brazil Søren Egedal Degn, Aarhus University, Denmark

#### \*Correspondence:

Rafael Bayarri-Olmos rafael.bayarri.olmos@regionh.dk

†These authors have contributed equally to this work and are joint first authors

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 26 March 2018 Accepted: 07 August 2018 Published: 28 August 2018

#### Citation:

Hertz CE, Bayarri-Olmos R, Kirketerp-Møller N, van Putten S, Pilely K, Skjoedt M-O and Garred P (2018) Chimeric Proteins Containing MAP-1 and Functional Domains of C4b-Binding Protein Reveal Strong Complement Inhibitory Capacities. Front. Immunol. 9:1945. doi: 10.3389/fimmu.2018.01945 The complement system is a tightly regulated network of proteins involved in defense against pathogens, inflammatory processes, and coordination of the innate and adaptive immune responses. Dysregulation of the complement cascade is associated with many inflammatory disorders. Thus, inhibition of the complement system has emerged as an option for treatment of a range of different inflammatory diseases. MAP-1 is a pattern recognition molecule (PRM)-associated inhibitor of the lectin pathway of the complement system, whereas C4b-binding protein (C4BP) regulates both the classical and lectin pathways. In this study we generated chimeric proteins consisting of MAP-1 and the first five domains of human C4BP (C4BP1−<sup>5</sup> ) in order to develop a targeted inhibitor acting at different levels of the complement cascade. Two different constructs were designed and expressed in CHO cells where MAP-1 was fused with C4BP1−<sup>5</sup> in either the C- or N-terminus. The functionality of the chimeric proteins was assessed using different in vitro complement activation assays. Both chimeric proteins displayed the characteristic Ca2+-dependent dimerization and binding to PRMs of native MAP-1, as well as the co-factor activity of native C4BP. In ELISA-based complement activation assays they could effectively inhibit the lectin and classical pathways. Notably, MAP-1:C4BP1−<sup>5</sup> was five times more effective than rMAP-1 and rC4BP1−<sup>5</sup> applied at the same time, emphasizing the advantage of a single inhibitor containing both functional domains. The MAP-1/C4BP chimeras exert unique complement inhibitory properties and represent a novel therapeutic approach targeting both upstream and central complement activation.

Keywords: complement activation, lectin pathway, classical pathway, MAP-1, C4BP, complement inhibition, chimeric protein

# INTRODUCTION

The complement system constitutes a central effector arm of the vertebrate immune system occupying a pivotal position as an early danger sensor and mediator of immunological and inflammatory processes (1, 2). Complement is activated via three distinct pathways, i.e., the classical (CP) the lectin (LP) and the alternative pathway (AP) that converge on the generation and deposition of opsonins, release of anaphylatoxins, and generation of the terminal complement complex (TCC) which all together help coordinating the cellular and humoral immune responses (3–6).

Initiation of CP is mediated mainly by the binding of C1q to immune-complexes, but also to conserved pathogen-specific structures, altered self-antigens or interaction with the pentraxins (7) and leads to the activation of the serine proteases C1r and C1s (8, 9). The LP is initiated by two groups of pattern recognition molecules (PRM): c-type lectins, such as mannosebinding lectin (MBL), collectin-10 (CL-10, CL-L1), and collectin-11 (CL-11, CL-K1) and ficolins, i.e., ficolin-1 (M-ficolin), ficolin-2 (L-ficolin), and ficolin-3 (H-ficolin or Hakata antigen). Binding of PRMs to ligands activates the MBL-associated serine proteases (MASPs), causing cleavage and activation of C2 and C4 (10). The AP is constantly "probing" surfaces in a non-specific fashion by means of a constitutive low level of spontaneous hydrolysis of C3 into the C3b analog C3(H2O) (11). Maybe more importantly, C3b generated from any of the three pathways is amplified via the AP. In fact, AP amplification may account for up to 90% of the total complement activation independently of the triggering pathway (7, 12). Whether the AP can be activated via a similar PRM-dependent mechanism involving properdin remains controversial (13, 14).

Complement is much more than a pathogen-sensing system, with ties spanning into coagulation, inflammation and embryonic development (2). In order to carry out its diverse functions it relies on a delicate balance between activation and inhibition by means of a set of membrane-bound and soluble regulatory proteins (15), the latter including the LP inhibitor MBL/ficolin/CL-associated protein 1 (MAP-1, also named MAp44), and the LP and CP regulators C4b-binding protein (C4BP) and C1 inhibitor (a protease inhibitor of the serpin super family that inactivates C1r, C1s, MASP-1, and MASP-2) (16).

Two genes encode all five naturally-occurring MASP isoforms in mammals (17): three proenzymes (i.e., MASP-1, -2, and -3) and the two non-proteolytic proteins sMAP (MAp19) and MAP-1. The proteolytic MASPs consist of a heavy chain comprised of two CUB domains separated by an epidermal growth factor (EGF) domain, and two complement control protein/short consensus repeats (CCP/SCR) modules—and a light chain composed of the serine protease domain. MAP-1 differs from the proteolytic MASPs in that it lacks the second SCR and SP domain and thus it cannot activate complement. MAP-1 competes with the MASPs in binding to recognition molecules of the LP leading to decreased complement activation. This has been demonstrated in vitro (18–20), and in multiple in vivo disease models (21, 22).

C4BP is a soluble protein encoded in the regulator of complement activation (RCA) gene locus of chromosome 1 (23) and possesses a unique structure among the RCA proteins in being a polymer composed of several CCP containing polypeptides. The most abundant isoform in the circulation is composed of seven identical α-chains (75 kDa each) and one β-chain (45 kDa) linked together by a central core and found in a high affinity complex with the anticoagulant vitamin K-dependent protein S (24, 25). The complement regulatory functions of C4BP are located within the first CCP domains of the α-chains. C4BP binds to the negatively-charged surface of C4b via the first three CCP domains of the α-chain preventing the assembly of the classical and lectin pathways C3 convertases (26, 27). Additionally, C4BP acts as a cofactor in the complement factor I (fI)-mediated proteolytic inactivation of both soluble and membrane bound C4b (28–30). By binding to C3b via the first 4 CCP domains of the α-chain, C4BP also participates in the fI-dependent C3b degradation to iC3b in the fluid phase (31). Although it is difficult to speculate upon the genuine physiological role of the inhibitory function of C4BP since no C4BP deficiency has been diagnosed in humans (32), C4BP injected peritoneally has been shown to alleviate inflammation and tissue damage in collagen- and collagen antibody-induced arthritis mouse models (33).

Since the US Food and Drug Administration approval of the first complement-specific drug in 2007 (34), rational modulation of the complement cascade using complement inhibitors has gradually demonstrated its potential as a drug discovery strategy and therapeutic treatment (35). Especially recombinant chimeric proteins targeting different levels of the cascade are of great interest in complement-mediated therapy and have previously been created with success (36, 37). Here we aimed to create a dual inhibitor with the ability to target initial activation by both the lectin and classical pathways by combining full length MAP-1 with the first five N-terminal CCP domains of the α-chain of C4BP. This could provide a unique platform for a novel class of complement inhibitor and thus contribute to the emerging field of complement therapeutics.

# MATERIALS AND METHODS

#### Buffers

The following buffers were used: PBS (0.2 M Na2HPO4, 35 mM K2HPO4, 0.15 M NaCl, 15 mM KCl), PBS/NaCl (0.2 M Na2HPO4, 35 mM K2HPO4, 0.5 M NaCl, 15 mM KCl), TBS/Ca2<sup>+</sup> and TBS/Tw/Ca2<sup>+</sup> (20 mM Tris-HCl, 150 mM NaCl, 5 mM CaCl2, with/without 0.05% Tween-20), TBS/EDTA and TBS/Tw/EDTA (20 mM Tris-HCl, 150 mM NaCl, 10 mM EDTA, with/without 0.05% Tween), VBS/Tw and sample buffer (4 mM C8H11N2NaO3, 145 mM NaCl, 2.6 mM CaCl2, 2 mM MgCl2, with 0.05% Tween-20 or 0.5% BSA respectively).

#### Design of Chimeric Proteins and Transfection

The coding sequences for MAP-1 (NM\_001031849.2) and C4BP (NM\_000715.3) were optimized for expression in Chinese hamster ovarian (CHO) cells in terms of codon adaptation index, mRNA stability, GC content, removal of cryptic splice sites, and repeats, 5' UTR, and signal peptide. All DNA manipulations were performed in Visual Gene Developer (38). MAP-1:C4BP1−<sup>5</sup> comprises the coding sequence of MAP-1 followed by the first five CCP domains of the α-chain of C4BP (C4BP1−<sup>5</sup> ). The reverse construct C4BP1−<sup>5</sup> :MAP-1 was designed with the coding

**Abbreviations:** CP, LP, AP, Classical, lectin, and alternative pathway; PRM, Pattern recognition molecule; IC50, Half maximal inhibitory concentration; Kd, half maximal binding concentration; SEM, Standard error of the mean; PTEC, Proximal tubular epithelial cells.

sequence of C4BP1−<sup>5</sup> located in the N-terminus of MAP-1 and separated by a flexible glycine serine linker (G4S)3. A control construct consisting of C4BP1−<sup>5</sup> alone was designed with a Cterminal hexa histidine-tag. In all cases, the signal peptide of MAP-1 was used to ensure secretion. The chimeric constructs were cloned into the pcDNA5/FRT vector and transfected into Flp-in CHO cells (both from Invitrogen, Thermo Fisher Scientific, USA) according to the manufacturer's instructions. Positive transfectants were selected for Hygromycin resistance in complete medium with Ham's F-12 Nutrient Mix, 10% fetal calf serum, 2 mM L-Glutamine, 1% Penicillin-Streptomycin (Sigma-Aldrich, USA), and 500 mM of Hygromycin B (all from Gibco, Thermo Fisher Scientific unless otherwise stated), and maintained at 37◦C in a humidified atmosphere with 5% CO2.

#### Purification of Recombinant Proteins

The chimeric constructs were purified by immunoaffinity chromatography using the monoclonal antibody (mAb) anti-MAP-1 20C4 (39) coupled to cyanogen bromide-activatedsepharose 4B (Sigma-Aldrich). Briefly, filtered cell culture supernatants were applied to the column and washed in PBS/NaCl. Bound protein was eluted with 0.5% (w/v) citric acid (pH 2.5) and neutralized with 1 M Tris-HCl (pH 9). Elution fractions were applied on a NuPAGE bis-tris 4–12% polyacrylamide gel (Invitrogen), followed by a direct protein stain with Instant Blue (Expedeon LTD, UK) to assess the purity. The concentration of the chimeric proteins was interpolated by ELISA using rMAP-1 as standard. rC4BP1−<sup>5</sup> was purified with a HisTrap excel column (GE Healthcare, Denmark) according to the manufacturer's instructions.

#### SDS-Page and Western Immunoblotting

SDS-PAGE was performed using NuPAGE bis-tris 4–12% gels with or without reducing agent and blotted onto HybondTM-ECL nitrocellulose membranes (GE Healthcare) according to the manufacter's instructions. The membranes were probed with 0.5µg/ml anti-MAP-1 20C4 or 0.3µg/ml anti-human C4BP polyclonal antibody (pAb; Abnova, 722-D01P, Taiwan) overnight, followed by 1 h incubation with 0.4µg/ml rabbit anti-mouse-HRP pAb (Dako, P0260, Denmark) or a 1:4000 dilution of donkey anti-rabbit-HRP pAb (GE Healthcare, NA934V) for MAP-1 and C4BP, respectively. The membranes were developed with SuperSignal WestFemto (Thermo Fisher Scientific) and analyzed with Microchemi (Bio-imaging systems, Israel).

#### Fluid Phase C4BP Cofactor Assay

The cofactor activity of C4BP, MAP-1:C4BP1−<sup>5</sup> , and C4BP1−<sup>5</sup> :MAP-1 was assessed with an established fluid phase cofactor activity assay (40, 41) measuring the fI-cofactor activity of C4BP in the proteolytic inactivation of C4b and C3b. Briefly, C4b and C3b were incubated with fI, complement factor H (fH), and equimolar concentrations of full length C4BP (all from Complement Technology Inc., USA), recombinant MAP-1 (rMAP-1, produced in house), MAP-1:C4BP1−<sup>5</sup> , or C4BP1−<sup>5</sup> :MAP-1 for 2 h at 37◦C. The reactions were stopped with LDS sample buffer and reducing agent (both from Invitrogen). Samples were subjected to electrophoresis and immunoblotting as described above. Cleavage of C3b was visualized by incubation with 1.6µg/ml rabbit anti-human C3d pAb (Dako, A006302) followed by donkey anti rabbit-HRP (GE Healthcare, NA934V) in a 1:5000 dilution. C4b cleavage was assessed by incubation with 0.3µg/ml mouse anti-human C4d mAb (Thermo Fischer Scientific) and 0.4µg/ml rabbit anti mouse-HRP (Dako). Membranes were developed as described previously.

### Size Exclusion Chromatography (SEC)

The chimeric proteins (50 µg) were applied to a Superdex 200 HR 10/30 column (GE Healthcare) with a flow rate of 0.5 ml/min in TBS/Ca2<sup>+</sup> or TBS/EDTA. The eluate was collected in 1 ml fractions and analyzed by ELISA (see below).

# ELISA

#### Assessment of Ca2+-Dependent Dimerization by Sandwich ELISA Following SEC

Microtiter wells (MaxiSorp plate; Thermo Fisher Scientific) were coated with anti-C4BP pAb (2µg/ml) in PBS overnight at 4◦C. SEC fractions were applied in a two-fold dilution starting at 1:20 and incubated for 2 h. Biotinylated 8B3 mAb (2µg/ml), recognizing the common heavy chain of MASP-1, MASP-3 and MAP-1 (39), was used as detection antibody for 2 h. Streptavidin-HRP conjugate (Sigma Aldrich, RPN1231V) (1:2000) was added for 1 h and the plates were developed with substrate solution TMB One (Kem Em Tek, Denmark). The reaction was stopped with 0.2 M H2SO<sup>4</sup> and the absorbance was measured at 450 nm using an ELx80 absorbance reader (BioTek, USA).

#### Binding of Chimeric Proteins to Ligand-Bound MBL and CL-11

Microtiter wells were coated with 10 µg/ml mannan (Sigma Aldrich) in PBS overnight at 4◦C followed by incubation with recombinant MBL or recombinant CL-11 (rMBL and rCL-11 produced in house; 2µg/ml) (42). Serial dilutions of MAP-1:C4BP1−<sup>5</sup> , C4BP1−<sup>5</sup> :MAP-1, rMAP-1, and rC4BP1−<sup>5</sup> in TBS/Tween/Ca2<sup>+</sup> or TBS/Tween/EDTA starting at 13.5 nM were allowed to complex with rMBL for 2 h. Detection and development were done as described above. Binding of rMBL and rCL-11 to mannan was confirmed using biotinylated anti-MBL Hyb-131-1 mAb and anti-CL-11 Hyb-15 mAb (2µg/ml), respectively.

#### Complement Activation Assays

Inhibition of complement activation via the LP was determined by an established ELISA-based assay (42) using specific antibodies to detect the activated and deposited products of C4, C3, and TCC. The chimeric proteins ability to inhibit complement was tested in MBL defect serum (43) and normal human serum (NHS). Mannan (10µg/ml) was immobilized onto microtiter plates and incubated in PBS overnight at 4◦C followed by incubation with 2µg/ml rMBL for 1.5 h. Serial dilutions of rMAP-1, C4BP1−<sup>5</sup> :MAP-1, and MAP-1:C4BP1−<sup>5</sup> starting at 667 nM were incubated with mannan/rMBL complexes for 1 h prior to addition of 2% MBL defect serum. When using NHS as a complement source, serial dilutions of C4BP1−<sup>5</sup> :MAP-1, MAP-1:C4BP1−<sup>5</sup> , MAP-1, and purified C4BP, starting at 667 nM, were added to round bottom non-absorbent titration plates (Thermo Fisher Scientific) with 2% NHS. The serum-protein mix was incubated 30 min at 4◦C, transferred to mannan (10µg/ml) coated plates and incubated for 45 min at 37◦C. Detection of deposition fragments was quantified with 2µg/ml biotinylated mouse mAbs anti-human C4 (Anti-C4) (Bioporto Diagnostics, Denmark, Hyb 162-02), 2µg/ml anti-human C3bc BH6 (44), and 1µg/ml anti-human C5b-9 (anti-TCC; Bioporto, Dia 011– 01) incubated for 2 h. Finally, Streptavidin-HRP conjugate was added in a 1:2000 dilution for 1 h. Unless otherwise specified, all washes and incubation steps were done with VBS/Tw. Development was performed with substrate solution TMB One. Data are represented as inhibition (%), calculated as [(ODinhibitor – ODbackground)/(ODnoinhibitor – ODbackground)]100.

#### Total Complement Screen

The WieslabTM Complement System Screenkit COMPL 300 (Euro diagnostica, Sweden) (45), a standardized immunoassay for measuring TCC formation by all three complement pathways, was used to assess the inhibitory capabilities of the chimeric proteins beyond the LP. MAP-1:C4BP1−<sup>5</sup> , C4BP1−<sup>5</sup> :MAP-1, rMAP-1, rC4BP1−<sup>5</sup> , and rMAP-1 plus rC4BP1−<sup>5</sup> (starting at 667 nM for LP and CP, and at 2667 nM for AP) were applied in a two-fold dilution in non-absorbent titration plates in the respective buffers for each pathway provided with the kit. NHS in a 1:100 dilution for the CP and LP and a 1:20 for the AP was coincubated with the inhibitors for 30 min at 4◦C. Protein/serum mixes were transferred to pre-coated 96-well microtiter plates and incubated 1 h at 37◦C. Following incubation the samples were analyzed with the Wieslab Complement System Screenkit according to the instruction from the manufacturer. Data are represented as described in the previous paragraph without background subtraction.

#### Flow Cytometry

Complement activation on kidney cells was studied by flow cytometry using HK-2 cells (ATCC, CRL-2190), an immortalized human proximal tubular epithelial cell (PTEC) line reported to be a suitable alternative to freshly isolated PTCs (46). Cells were cultured in Keratinocyte Serum Free Medium supplemented with bovine pituitary extract, human recombinant epidermal growth factor (all from Gibco) and 0.5% Penicillin-Streptomycin (Sigma-Aldrich). On the day of the experiments, the cells were detached with TryPLE Express Enzyme, washed with prewarmed Hank's balanced salt solution (both from Gibco) with calcium and magnesium, and transferred to polystyrene round bottom tubes (Corning, USA). Data was collected using a Gallios flow cytometer and analyzed by the Kaluza software 1.2 (both from Beckman Coulter, USA). Values are reported as median fluorescence intensities (MFI). To ensure analysis of single cells, a forward scatter area vs. height gate was defined upon a uniform gated population on the forward vs. side scatter (see **Supplementary Figure 1**). A total of 10,000 events were recorded per experimental condition.

#### Binding of MBL to Kidney Cells

HK-2 cells (1.0 × 10<sup>5</sup> cells/test) were incubated with serial dilutions of rMBL for 30 min at 4◦C in sample buffer with or without EDTA. Prior to the addition of the detection antibody, the cells were fixed in a 1% solution of paraformaldehyde (Sigma Aldrich) for 10 min. Bound rMBL was detected using anti-human MBL Hyb 131-11 mAb (2µg/ml) for 30 min plus goat anti-mouse PE conjugate (Sigma, P9670) in a 1:20 dilution for 20 min in the dark. The cells were washed with sample buffer at 300 × g for 5 min in between steps and all manipulations were performed at 4 ◦C.

#### Complement Deposition on Kidney Cells

HK-2 cells (1.0 × 10<sup>5</sup> cells/test) were incubated with rMBL for 30 min, followed by 10% MBL defect serum for 1 h. The cells were fixed with 1% paraformaldehyde for 10 min. Surfacebound C4 was measured with biotinylated anti-human C4c pAb (2µg/ml) (Dako, Q036905) for 30 min followed by Streptavidin APC conjugate (1:400) (Invitrogen, SA1005) for 20 min in the dark. To study the impact of the chimeric inhibitors on C4 deposition, serial dilutions of MAP-1:C4BP1−<sup>5</sup> , and C4BP1−<sup>5</sup> :MAP-1, rMAP-1, and C4BP1−<sup>5</sup> (200 nM) were allowed to react with 10% MBL defect serum for 20 min prior to incubating with the rMBL-bound HK-2 cells for 1 h. After fixing the cells C4 deposition was measured as described previously. All washes and incubation steps were done in sample buffer at 4◦C.

### Statistics

All statistical analyses were performed in GraphPad Prism software 7.02 (GraphPad, USA). Kd values (i.e., ligand concentration required to achieve half of the maximum binding) were calculated using the equation specific binding with Hill slope after subtracting the background OD and constraining the Bmax (maximum number of binding sites) to a value shared between all inhibitors (**Figure 3**). The half maximal inhibitory concentration (IC50) was calculated using a global nonlinear regression with the equation inhibitor concentration vs. response constraining the top and bottom parameters to equal 100 and 0 respectively (**Figures 4**, **5**, **8**). The significance of inhibitory differences was analyzed using one-way ANOVA with Tukey's corrections for multiple comparisons on the best-fit IC50 values from data sets with an acceptable goodness-of-fit (adjusted R-squared > 0.8). Significance of the binding of rMBL and MBL-dependent C4 deposition on the surface of HK-2 cells was assessed using a two-tailed t-test, while inhibition was analyzed using multiple unpaired t-tests with the Holm-Sidak correction for multiple comparisons. Data are represented as mean ± SEM of three independent experiments.

# RESULTS

### Production and Purification of Chimeric Proteins

We designed two differently-oriented chimeric inhibitors: MAP-1:C4BP1−<sup>5</sup> , with the first five CCP domains of the α-chain of C4BP (alias C4BP1−<sup>5</sup> ) located in the C-terminus of MAP-1; and C4BP1−<sup>5</sup> :MAP-1, where C4BP1−<sup>5</sup> was placed in the N-terminus of MAP-1 via a flexible glycine-serine linker. A control protein, C4BP1−<sup>5</sup> without MAP-1, was designed with a terminal hexa His-tag (**Figure 1A**). All three constructs were cloned into a pcDNA5/FRT plasmid and transfected into Flp-In CHO cells. MAP-1-containing chimeric proteins were purified by antibody affinity chromatography and C4BP1-5 was purified via immobilized metal ion affinity chromatography on a HisTrap excel column. Under reducing conditions, the chimeric proteins

specific binding with hill slope. Results are representative of three independent experiments and error bars represent minimum and maximum values of triplicate measurements.

migrated as a single band with an apparent molecular weight of ∼80 kDa, equivalent to the sum of rMAP-1 and rC4BP1−<sup>5</sup> molecular weights (i.e., ∼44 kDa and ∼37 kDa respectively; **Figure 1B**). The identity of the protein bands was confirmed by Western blotting (**Figure 1C**).

#### MAP-1:C4BP1−<sup>5</sup> and C4BP1−<sup>5</sup> :MAP-1 Form Dimers and Associate With the Collectins of the LP in the Presence of Calcium

The MASPs are known to form head-to-tail dimers via the interaction of the CUB1 and EGF domains and to associate with the recognition molecules of the LP in a Ca2+-dependent manner (18, 47, 48). To study the effect of calcium in the dimerization of our recombinant inhibitors, we performed gel filtration with and without EDTA, a Ca2+-chelating agent. Both chimeric proteins eluted in a single overlapping peak in fractions 6–8 under physiological calcium concentrations. In the presence of EDTA, the previous peak was drastically reduced and a new one appeared in fractions 7–9 indicating that both are capable of assembling into dimers stabilized by Ca2<sup>+</sup> (**Figure 2**).

Both chimeric proteins and MAP-1 were able to bind to rMBL and rCL-11 in a dose-dependent manner with equivalent dose-response curves in the presence of Ca2<sup>+</sup> (**Figure 3**). On rMBL, MAP-1:C4BP1−<sup>5</sup> and C4BP1−<sup>5</sup> :MAP-1 exhibited a similar

complement. Detection of C4 (A), C3 (B), and TCC (C) deposition was quantified using anti-C4, anti-C3, and anti-TCC mouse mAbs. Connecting lines are four parameters nonlinear fitting using the equation inhibitor concentration vs. slope. Data are reported as [(ODinhibitor-ODbackground)/(ODnoinhibitor-ODbackground)]100. Error bars represent the SEM of three independent experiments, and the dashed line the

50% inhibition level. \*P < 0.05; \*\*\*P < 0.001; \*\*\*\*P < 0.0001.

Kd of 0.35 and 0.56 nM respectively, slightly lower affinity than rMAP-1 at the given conditions (0.14 nM; **Figure 3A**). The same tendency, albeit with weaker interactions, was observed with rCL-11 (**Figure 3B**): rMAP-1 displayed the lowest Kd (1.65 nM), followed by MAP-1:C4BP1−<sup>5</sup> (1.97 nM) and lastly C4BP1−<sup>5</sup> :MAP-1 (5.49 nM). As expected, no binding could be detected when the proteins were incubated in the presence of EDTA.

#### Inhibition of LP Activation With MBL Defect Serum and NHS

To study the effect of the chimeric constructs on LP-mediated complement activation we used mannan-bound rMBL as the activating PRM. MAP-1:C4BP1−<sup>5</sup> and C4BP1−<sup>5</sup> :MAP-1

lines are four parameters nonlinear fitting using the equation inhibitor concentration vs. slope. Dashed line Data are reported as [(ODinhibitor – ODbackground)/(ODnoinhibitor – ODbackground)]100. Error bars represent the SEM of three independent experiments, and the dashed line the 50% inhibition level. \*P < 0.05; \*\*\*P < 0.001.

exhibited a strong dose-dependent inhibition on C4, C3, and TCC level when they were allowed to form complexes with rMBL prior to the addition of 2% MBL defect serum as complement source (**Figure 4**). rMAP-1 was the most effective at the C4 level: IC50 MAP-1 (0.14 nM) vs. IC50 MAP-1:C4BP1−<sup>5</sup> (0.45 nM), non-significant; IC50 MAP-1 (0.14 nM) vs. IC50 C4BP1−<sup>5</sup> :MAP-1 (1.2 nM), P < 0.0001. Nonetheless, MAP-1:C4BP1−<sup>5</sup> outperformed the rest when measuring C3 (IC50 MAP-1:C4BP1−<sup>5</sup> = 0.36 nM vs. 7.72 nM C4BP1−<sup>5</sup> :MAP-1, P < 0.05, and 3.35 nM MAP-1, non-significant) and TCC deposition (IC50 MAP-1:C4BP1−5 = 0.17 nM vs. 0.92 nM C4BP1−<sup>5</sup> :MAP-1, P < 0.0001, and 0.46 nM MAP-1, P < 0.05).

Next, we tested the inhibitors by directly incubating them with 2% NHS in non-adsorbent titration plates before applying the inhibitor/serum mix to mannan (**Figure 5**). Under these conditions, both chimeric inhibitors demonstrated higher inhibition than rMAP-1 alone in the later steps of the complement cascade (i.e., C3 and TCC formation). MAP-1:C4BP1−<sup>5</sup> exhibited a more pronounced inhibition of C4 (nonsignificant tendency), C3 (P < 0.05), and TCC deposition (P < 0.001). The other construct appeared to be less active but still outperformed rMAP-1 at the C3 (non-significant trend) and TCC levels (P < 0.05). No complement deposition was observed when MBL defect serum was incubated without rMBL (**Figure 4**) or with rMBL but no mannan (**Figure 5**).

#### C4BP Cofactor Activity in FI-Mediated Cleavage of Soluble C4b/C3b

It has previously been reported that C4BP can act as a cofactor in the fI-mediated proteolytic inactivation of C3b and C4b (29, 31, 49). Hence we examined whether our chimeric proteins retained the fI cofactor activity of native C4BP.

When C4b was incubated with fI in combination with native full length C4BP or one of the two chimeric proteins, the α-chain of C4b was cleaved and a band corresponding to the smaller cleavage product C4d appeared (**Figure 6A**). As expected no full cofactor activity was observed for MAP-1 but the cleavage product iC4b was observed when MAP-1 was in combination with FI and C4b. Similarly, a low level proteolysis of C3b into iC3b could be observed in the presence of fI and MAP-1. Both are probably artifacts from the elevated fI concentrations used in the assays. No C4b proteolysis was apparent in the absence of cofactors or fI, with the notable exception of purified native C4BP. When C4b was incubated with C4BP alone a cleavage of the α-chain and generation of the iC4b was observed, hinting at a possible contamination of the purified C4BP. We performed a new immunoblotting with C4BP alone and in combination with fI or C4b. The iC4b band was visible in all wells, suggesting that iC4b had been co-purified with C4BP (data not shown).

A cleavage of the C3b α-chain band was observed together with the appearance of a band corresponding to iC3b when C3b was subjected to fI in combination with C4BP or one of the two chimeric proteins (**Figure 6B**). Full cleavage of C3b to C3dg could only be archived when fH was applied in combination with C3b and fI. As expected no cofactor activity was observed for MAP-1.

#### Reduction of MBL-Dependent C4 Deposition in Kidney Proximal Tubular Epithelial Cells

The complement system, and in particular MBL, has been documented as one of the key mediators of kidney injury following renal ischemia (50–53). Proximal tubular epithelial cells (PTECs), responsible for many regulatory and endocrine functions, are especially vulnerable to complement-mediated tissue damage (51). After demonstrating the efficacy of our chimeric inhibitors in the previous ELISA-based assays, we tested whether we could quench complement activation on PTECs. A dose-dependent MBL binding was observed when human PTECs were incubated with increasing concentrations of rMBL (**Figure 7A**). Binding could be neutralized with the

addition of EDTA in the sample buffer, suggesting a classical c-type lectin interaction. Addition of 10% MBL defect serum to MBL-bound PTECs led to C4 deposition on the cell surface (**Figure 7B**) that was significantly reduced when rMAP-1 or the MAP-1-containing chimeras were incubated in a concentration of 200 nM with serum prior to addition to the MBL-bound cells as compared to the no inhibitor control (**Figure 7C**). Equimolar concentrations of rC4BP1−<sup>5</sup> alone showed no significant reduction of C4 deposition.

#### Total Complement Screen

We further assessed the ability of the chimeric inhibitors to regulate all three complement pathways using the commercial total complement screen assay Wielisa (45) (**Figure 8**). In agreement with the above results MAP-1:C4BP1−<sup>5</sup> was the most effective construct, significantly outperforming all other inhibitors in the LP (**Figure 8A**) and CP (**Figure 8B**) activation assays, while C4BP1−<sup>5</sup> :MAP-1 demonstrated a more modest activity. The combination of rC4BP1−<sup>5</sup> and rMAP-1 displayed an inhibitory potential comparable to C4BP1−<sup>5</sup> :MAP-1, and 5 to 10 times lower than MAP-1:C4BP1−<sup>5</sup> . In the AP however (**Figure 8C**), MAP-1:C4BP1−<sup>5</sup> had only a minor effect, with an IC50 value comparable to rC4BP1−<sup>5</sup> (alone or in combination with rMAP-1). Inhibitors whose dose-response curves had a fitting with an adjusted R-square value < 0.8 were excluded for the analysis—such as rMAP-1 (**Figures 8A,B**) and all inhibitors besides MAP-1:C4BP1−<sup>5</sup> in the AP activation assay (**Figure 8C**).

#### DISCUSSION

The unique position of the complement system as an early danger sensor and conductor of downstream humoral and cellular responses makes complement modulation an attractive pharmacological target (36, 54). Rational engineering of existing complement regulators and novel recombinant inhibitors targeting complement on different levels has proven to be a successful strategy, such as mini-fH (41) and chimeric regulators composed of the N-terminal domains of fH merged with CRIg (55) or soluble CR2 (40). Complement factor H is the main soluble regulator of the AP and with CRIg being a C3 regulator

and CR2 an inhibitor of C3 deposition all these molecules affect exclusively the alternative amplification loop. However, increasing understanding of the involvement of complement in highly diverse clinical conditions suggests a different approach with complement-targeted drugs interfering at various stages of the cascade (35). Our group has previously tested this strategy by combining full length MAP-1 with the first five CCPs of fH, generating a chimeric protein with the potential to target initial recognition and at the same time downregulate AP amplification (37). We wanted to investigate this approach further and designed two differently-oriented proteins

by genetic engineering; MAP-1:C4BP1−<sup>5</sup> , with the sequence of human C4BP1−<sup>5</sup> fused directly downstream of full length human MAP-1, and C4BP1−<sup>5</sup> :MAP-1, where C4BP1−<sup>5</sup> was located in the N-terminus and connected to MAP-1 via a flexible glycine-serine linker. In theory, combining two existing complement inhibitors could result in a synergistic effect, giving rise to unique complement-regulatory and anti-inflammatory properties. Furthermore, we hypothesize that the ability of targeting both CP and LP activation may be especially beneficial in multifactorial pathologies where both pathways are involved

in driving tissue damage, such as ischemia/reperfusion injuries (56).

MAP-1 forms head-to-tail homodimers (18) stabilized by calcium, that upon complex formation with collectins and ficolins downregulate LP activation (18, 57, 58). MAP-1:C4BP1−<sup>5</sup> , C4BP1−<sup>5</sup> :MAP-1, and rMAP-1 displayed comparable Kd values to rMBL (0.35, 0.56, 0.14 nM respectively) and rCL-11 (1.97, 5.49, and 1.65 nM respectively) in a solid-phase ELISA setup. Hill slopes values above 1 indicate a multivalent binding to MBL. The lower hill slope values observed with rCL-11 may be a consequence of its limited oligomerization and reflect the reduced number of collagen-stalks available for complexation with the inhibitors. Addition of the calcium-chelating agent EDTA completely inhibited binding. This clearly indicates that the interaction of the chimeras with the PRMs is mediated through their MAP-1 part in a calcium-dependent manner. No binding was observed with rC4BP1−<sup>5</sup> (data not shown).

To investigate the dimerization in detail both chimeras were subjected to SEC in the presence of physiological calcium concentrations or EDTA. We found that both constructs eluted as a single peak in the presence of calcium, whereas when exposed to EDTA-containing buffers the chimeras were separated into monomers leading to a shift in their spectrophotogram profiles toward a lower estimated molecular size. The identity of the peaks was confirmed using a specific ELISA capturing C4BP and detecting MAP-1. These results demonstrate that the constructs maintain the calcium-dependent dimerization of native MAP-1 that is critical for its role as an LP inhibitor.

Our group and others have previously demonstrated that rMAP-1 alone and as part of chimeric constructs inhibits LP activation in vitro (18, 20, 21, 37, 39). Here we tested the ability of the chimeric proteins to regulate LP activation in different functional ELISAs. First we incubated the chimeric proteins, rMAP-1, and C4BP1−<sup>5</sup> with MBL bound to mannancoated plates, and subsequently added 2% MBL defect serum as complement source. While rMAP-1 alone seemed more effective at inhibiting C4 deposition, MAP-1: C4BP1−<sup>5</sup> clearly outperformed rMAP-1 on the C3 and TCC levels. C4BP1−<sup>5</sup> :MAP-1 was the least effective regulator of the three. If C4 regulation is solely dependent on MAP-1, then differences on C4 inhibition could be explained by their different binding affinities toward MBL. In a more physiologically-relevant ELISA setup the proteins were co-incubated with 2% NHS, thus allowing direct competition between the inhibitors and intrinsic MASPs for binding to MBL. Again, incubation with MAP-1:C4BP1−<sup>5</sup> caused a significant inhibition resulting in the almost complete absence of C3, and TCC deposition. While C4BP1−<sup>5</sup> :MAP-1 failed to provide the same inhibition as MAP-1:C4BP1−<sup>5</sup> , it was still more effective than rMAP-1 alone. The difference in inhibition levels of rMAP-1 observed in the two experimental setups is in agreement with other studies (18, 37) and could be attributed to competition with the MASPs on MBL binding. This phenomenon supports our explanation that rMAP-1 occupies MASPs-binding sites at the MBL molecules when pre-incubated (20), that would result in either a lower concentration of catalytically-active MASPs or/and a decreased intercomplex activation by spacing out MBL/MASPs complexes (59). When added simultaneously with serum, rMAP-1 may not be capable of displacing already formed MBL/MASPs complexes (at least under our experimental conditions).

It has been reported that the domains CCP1-4 of the α-chain are critical for the fI cofactor activity of C4BP in the fI-mediated C4b (28–30) and C3b proteolytic inactivation (31). Thereby we expected that the chimeric proteins will manifest intact fI cofactor activity. We showed that C4BP1−<sup>5</sup> -containing constructs were as efficient as full length native C4BP in a fluid phase C3b/C4b degradation assay. Moreover it appears that C4BP is a more effective cofactor in the cleavage of C4b compared to C3b in agreement with a previous study (49).

We next looked for a pathological condition where complement injury was triggered by an initial recognition by the LP. MBL has been documented to be directly involved in kidney damage following renal ischemia (53). We used immortalized human PTECs to test whether we could quench the exacerbated complement activation observed on the tubular epithelium, known to be especially vulnerable to complement attack after reperfusion (51). In agreement with the literature rMBL was capable of binding to the cell surface in a calciumdependent manner, suggesting a traditional c-type lectin interaction (60). Binding to PTECs led to the deposition of C4 activation fragments. As seen on ELISA, the chimeras and rMAP-1 presented comparable regulatory effects (even when using a five times lower serum dilution than in ELISA). This to a degree limited inhibition can be expected assuming that trace amounts of PRM/MASPs complexes are sufficient to cause detectable C4 deposition, which in turn is mainly mediated by the MAP-1 part while the effect on the late cascade (i.e., C3 and TCC) depends on the combination of MAP-1 and C4BP. No C3 or TCC deposition was observed (see **Supplementary Figure 2**). The assay was performed at 4◦C, that while it allowed the cleavage and deposition of C4, it did not result in the assembly of the downstream convertases [62]. Incubation of MBL-bound cells at 37◦C led to the disappearance of C4 and MBL surface stain (data not shown). It has been reported that binding of MBL to PTEC causes its internalization followed by complement-independent cell death (60). Thus, this assay may have some inherited limitations to study downstream complement deposition, but is suitable to study the initial inhibiting effect of PRM binding inhibitors.

Finally, we tested the regulatory effect of the chimeric proteins in all three complement pathways using Wielisa, a platform for standardized complement activity measurements. In agreement with our previous results, MAP-1:C4BP1−<sup>5</sup> demonstrated a strong dose-dependent inhibition of TCC deposition in the LP (IC50 = 13.15 nM). Remarkably, it was also an efficient CP inhibitor (IC50 = 4.83 nM), outperforming rMAP-1 and C4BP1−<sup>5</sup> alone or in combination. C4BP1−<sup>5</sup> :MAP-1 inhibited the LP and CP to a lesser extent than MAP-1:C4BP1−<sup>5</sup> , equivalent to rMAP-1 and rC4BP1−<sup>5</sup> added at the same time. The equivalent regulatory effects of the chimeras in the LP and CP, together with the moderate activity of rMAP-1 in the LP, suggests that the observed LP-downregulation is mainly driven by the C4BP part, reflecting a limitation of the assay as a model for LPmediated activation. Thus, in a physiological setting where both functional domains of the chimeras would be involved, we may

observe a more pronounced inhibition [as shown for rMAP-1 (21)]. It should be noted that even though MAP-1 only associates with the PRMs of the LP and not the CP, MAP-1 containing chimeras were more effective than C4BP1−<sup>5</sup> alone in both CP and LP. This implies that the combination of two regulators within a single molecule—in the right configuration enables the enhanced functionality of the constructs, and not just displays the function of "free" MAP-1 or C4BP. In the AP activation assay MAP-1:C4BP1−<sup>5</sup> showed only a minor effect around 100 to 400-fold lower than in the LP and CP—that was directly comparable to rC4BP1−<sup>5</sup> (alone or in combination with rMAP-1), suggesting that AP activity is solely mediated by the C4BP part. It is important to highlight that the above-described experiments were performed using diluted serum as a source of complement. Experiments using animal models would be necessary to demonstrate that the chimeric proteins are effective complement inhibitors under physiological conditions.

The mechanism behind the observed differences in inhibitory properties between MAP-1:C4BP1−<sup>5</sup> and C4BP1−<sup>5</sup> :MAP-1 remain unclear, although it could be explained by their quaternary structure. Dimerization of C4BP1−<sup>5</sup> :MAP-1 could result in MAP-1 bending away from C4BP1−<sup>5</sup> via the flexible region in the CUB2/EGF region and the artificial linker between C4BP1−<sup>5</sup> and MAP-1. This contorted folding could give rise to a dimer with impaired ability to reach C4b/C3b and possibly also hindering the complex formation with MBL. On the other hand, MAP-1:C4BP1−<sup>5</sup> may fold into an elongated dimer with C4BP1−<sup>5</sup> pointing outwards from a MAP-1 dimer core bound to the collagen-like stalks of the PRMs similar to how the serine domain in the MASPs is assumed to protrude when bound to MBL. This conformation would enable both C4BP1−<sup>5</sup> monomers to interact with deposited C4b/C3b in the vicinity of the PRM binding site. We propose that the MAP-1 fragment may "dock" the chimeric inhibitor to the PRMs directing the multifunctional soluble inhibitor C4BP to the activating surface, thus increasing the efficacy. While data presented by our group in the present and past reports (37) provide a solid support for our hypothesis, conclusive evidence will require precise structural determination techniques, such as circular dichroism, multi-angle light scattering, small-angle x-ray scattering, or x-ray crystallography.

#### REFERENCES


#### In conclusion, we have successfully developed a novel complement inhibitor that harnesses the multifaceted functionality of the fluid phase CP and LP regulator C4BP, and directs it to danger foci by the specific association to PRMs mediated by MAP-1. The combination of MAP-1 fused to C4BP1−<sup>5</sup> presents unique complement regulatory properties in vitro and represents a potential novel therapeutic approach by targeting both upstream and central complement activation.

#### AUTHOR CONTRIBUTIONS

CH, RB-O, NK-M, M-OS, and SvP performed the experiments. KP optimized the ELISA-based inhibition assays. RB-O, M-OS, and PG designed the study. CH, RB-O, PG wrote the manuscript. All authors critically reviewed the manuscript.

#### FUNDING

The Danish Research Foundation of Independent Research (DFF-6110-00489), the Sven Andersen Research Foundation, and Novo Nordisk Research Foundation.

#### ACKNOWLEDGMENTS

The authors would like to thank Ms Jytte Bryde Clausen and Mrs. Jesper Andresen for excellent technical assistance.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.01945/full#supplementary-material

Supplementary Figure 1 | Gating strategy. HK-2 cells were defined as a uniform population on the forward scatter (FS INT) vs. side scatter (SS INT) plot, and single cells were selected gating on the forward scatter area or integral values (FS INT) vs. forward scatter height or peak (FS PEAK).

Supplementary Figure 2 | MBL-dependent complement deposition on HK-2 cells. Cells were incubated with rMBL for 30 min at 4◦C prior to addition of 10% MBL defect serum for 1 h at 4◦C. Deposition of C4 (A), C3 (B), and TCC (C). Cells incubated with no MBL, or with MBL but no serum were used as controls. X-Med, median fluorescence intensity.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Hertz, Bayarri-Olmos, Kirketerp-Møller, van Putten, Pilely, Skjoedt and Garred. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Functional characterization of alternative and classical Pathway c3/c5 convertase activity and inhibition Using Purified Models

*Seline A. Zwarthoff1 , Evelien T. M. Berends1 , Sanne Mol1 , Maartje Ruyken1 , Piet C. Aerts1 , Mihály Józsi2 , Carla J. C. de Haas1 , Suzan H. M. Rooijakkers1 \*† and Ronald D. Gorham Jr.1†*

*1Department of Medical Microbiology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands, 2Department of Immunology, ELTE Eötvös Loránd University, Budapest, Hungary*

Complement is essential for the protection against infections; however, dysregulation of complement activation can cause onset and progression of numerous inflammatory diseases. Convertase enzymes play a central role in complement activation and produce the key mediators of complement: C3 convertases cleave C3 to generate chemoattractant C3a and label target cells with C3b, which promotes phagocytosis; C5 convertases cleave C5 into chemoattractant C5a, and C5b, which drives formation of the membrane attack complex. Since convertases mediate nearly all complement effector functions, they are ideal targets for therapeutic complement inhibition. A unique feature of convertases is their covalent attachment to target cells, which effectively confines complement activation to the cell surface. However, surface localization precludes detailed analysis of convertase activation and inhibition. In our previous work, we developed a model system to form purified alternative pathway (AP) C5 convertases on C3b-coated beads and quantify C5 conversion *via* functional analysis of released C5a. Here, we developed a C3aR cell reporter system that enables functional discrimination between C3 and C5 convertases. By regulating the C3b density on the bead surface, we observe that high C3b densities are important for conversion of C5, but not C3, by AP convertases. Screening of well-characterized complement-binding molecules revealed that differential inhibition of AP C3 convertases (C3bBb) and C5 convertases [C3bBb(C3b)n] is possible. Although both convertases contain C3b, the C3b-binding molecules Efb-C/Ecb and FHR5 specifically inhibit C5 conversion. Furthermore, using a new classical pathway convertase model, we show that these C3b-binding proteins not only block AP C3/C5 convertases but also inhibit formation of a functional classical pathway C5 convertase under well-defined conditions. Our models enable functional characterization of purified convertase enzymes and provide a platform for the identification and development of specific convertase inhibitors for treatment of complement-mediated disorders.

Keywords: innate immunity, inflammatory disease, convertase enzymes, complement, complement therapeutics, multi-molecular proteases

# INTRODUCTION

The human complement system comprises a family of proteins that are essential to the human immune response against infections (1). Complement recognizes microbes or damaged host cells and subsequently triggers an enzymatic cascade that mainly serves to (a) label target cells for phagocytosis by immune cells, (b) produce chemoattractants, and (c) directly kill target cells *via*

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Marcin Okrój, Intercollegiate Faculty of Biotechnology of University of Gdańsk and Medical University of Gdan´ sk, Poland Elena Volokhina, Radboud University Nijmegen Medical Centre, Netherlands*

#### *\*Correspondence:*

*Suzan H. M. Rooijakkers s.h.m.rooijakkers@umcutrecht.nl*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

> *Received: 23 May 2018 Accepted: 10 July 2018 Published: 23 July 2018*

#### *Citation:*

*Zwarthoff SA, Berends ETM, Mol S, Ruyken M, Aerts PC, Józsi M, de Haas CJC, Rooijakkers SHM and Gorham RD Jr. (2018) Functional Characterization of Alternative and Classical Pathway C3/C5 Convertase Activity and Inhibition Using Purified Models. Front. Immunol. 9:1691. doi: 10.3389/fimmu.2018.01691*

**207**

pore formation (2). Unwanted complement activation on the body's own cells is a key pathological driver in a wide spectrum of immune diseases including autoimmune, inflammatory, and degenerative diseases (3–5). For current and future development of therapeutic complement inhibitors, knowledge of complement activation and how it can be regulated is of great importance.

Convertase enzymes fulfill a central role in the complement cascade as they cleave C3 and C5, which mediate nearly all complement effector functions. C3 convertases cleave C3 into C3a, a chemoattractant molecule, and C3b, which covalently binds to target surfaces and triggers phagocytosis. C5 convertases cleave C5 into C5a, a potent mediator of leukocyte recruitment and inflammation, and C5b, the initiator of the membrane attack complex and cell lysis. The complement cascade begins *via* specific recognition of target cells in the classical (CP) and lectin (LP) pathways. In the CP, antibodies bind epitopes on the target cell and subsequently recruit the C1 complex (C1qr2s2). Upon binding to the antibody platforms (6), C1q-associated protease C1s converts C4 and C2 to generate a C3 convertase enzyme (C4b2a) on the cell surface (**Figure 1A**). Similarly, the lectin

Figure 1 | Complement convertases mediate C3 and C5 conversion. (A) Upon complement activation, C3 convertases consisting of either C4b2a (CP and LP) or (B) C3bBb [alternative pathway (AP)] form on the cell surface. Conversion of C3 results in deposition of C3b molecules *via* the thioester (red dot), which form the basis for new AP convertases (amplification loop) or associate with existing C3 convertases to form C5 convertases. These accessory C3b molecules (C3bn) enable efficient C5 conversion, however, the molecular mechanisms underlying this process are not clear. Shown are C4b (purple), C2a (blue), C3 and cleavage products C3b/C3a (gray), Factor B (orange), C5 and cleavage products C5b/C5a (green). (C) Structural models of the C3 convertase C3bBb with its substrate C3 before and after cleavage. Models are based on structures of the C3bBb-SCIN dimer (PDB: 2WIN). The convertase is shown in ribbon representation, with C3b in dark gray and Bb in orange. On the left, the substrate C3 (light gray surface) is shown before cleavage. On the right, the product C3b (light gray licorice) is shown after cleavage. The red dots highlight the thioester. (D) Structural model of the previously proposed AP C5 convertase with its substrate C5. At a high density of C3b molecules, C5 is recruited to the target surface and can be cleaved after binding of Bb to C3b. The exact molecular arrangement of the C5 convertases remains unknown. This structural model is based on the CVF-C5 crystal structure (PDB: 3PVM), with accessory C3b molecules added manually surrounding the convertase and C5. CVF (cobra venom factor) is a C3b homolog that lacks the thioester domain and forms stable C5 convertases when associated with Bb in solution. Structure is shown in ribbon representation with C3b (dark gray), Bb (orange), accessory C3b molecules (light gray licorice), and C5 (green surface).

pathway also forms C4b2a *via* activation of mannose-binding lectin-associated serine proteases. The resulting CP/LP C4b2a convertases cleave C3 into C3a and C3b. Following cleavage, a reactive thioester in C3b is exposed, which enables its covalent attachment to target cell surfaces, leading to recognition of the cells by phagocytes. The labeling of target cells with C3b is amplified by the alternative pathway (AP) in which surfacebound C3b binds factor B (FB). The proconvertase C3bB is then cleaved by factor D (FD) to form an active C3 convertase complex that consists of C3b and the protease fragment Bb (C3bBb) (**Figure 1B**). Since the resulting active AP C3 convertase (C3bBb) is comprised of C3b itself, substrate cleavage results in generation of additional convertases, further propagating C3b deposition (**Figure 1B**). When the density of C3b molecules on the cell surface becomes sufficiently high, the existing C3 convertases (C4b2a and C3bBb) gain the ability to cleave C5, leading to formation of C5a and C5b (**Figures 1A,B**) (7, 8).

Selective inhibition of C3 and C5 convertases is of great therapeutic interest. Most complement inhibitors currently used in the clinic or in clinical development target precursor (not yet activated) complement proteins, that circulate through the body and do not mediate complement effector functions (4). Due to high concentrations of these precursor proteins, effective therapeutic concentrations of complement inhibitors are often quite high, and clearance of these molecules is enhanced due to rapid turnover of complement proteins. Furthermore, saturation of precursor proteins is more likely to systemically suppress complement activation, leading to increased susceptibility to infection (4). Such therapies would be more effective if they specifically targeted active protein complexes like convertases that are primarily formed during complement activation on target cell surfaces. While some therapeutic molecules inhibit convertase function, these likely inhibit multiple convertase enzymes and block all effector functions of complement (4). In some cases, specific inhibition of C5 convertases is desirable for complement therapy, since blocking these would prevent unwanted formation of the major inflammatory trigger C5a but leave C3b deposition *via* C3 convertases intact and thus phagocytosis of bacteria. However, the molecular details of C5 convertase formation and C5 cleavage remain poorly understood, largely due to the transient nature of convertases and the fact that efficient C5 conversion is constrained to cell surfaces (7, 9). Several earlier studies successfully investigated individual convertases in purified or semi-purified environments (10–14), however, no single model can fully characterize activity and inhibition of both AP and CP C3 and C5 convertases in a purified and controlled environment. Herein, we extended our recently developed model system for AP C5 convertases (7) to also study surface-bound AP C3 convertases, in order to screen for specific inhibitors of convertase enzymes. Furthermore, we developed functional analyses to study CP C3 and C5 convertases using purified complement proteins. Using these models, we can evaluate how known complement inhibitors affect C3/C5 conversion. In these analyses, we included bacterial and therapeutic complement inhibitors, and human complement regulatory proteins that protect healthy tissue from complement attack (8, 15–18). The analyses reveal that several C3b-binding molecules can discriminate between C3 and C5 convertases, suggesting that it is possible to develop more specific convertase inhibitors in the future. Through comparison of our inhibitory data with previously reported structural and biochemical data, we further postulate molecular models of convertase formation.

#### MATERIALS AND METHODS

#### Complement Proteins

C3 and C5 were prepared from human plasma as previously described (7). For CP assays and inhibitor dose-response assays, recombinant C5 was used. C5 with a C-terminal His-tag was therefore cloned from gBlocks (Integrated DNA Technologies) using Gibson assembly (Gibco), expressed in HEK293 cells (U-Protein Express, Utrecht, The Netherlands) and purified on a HisTrap column (GE Healthcare). C3bH2O and C3b-PEG11 biotin were prepared as previously described using 180 µg/ml maleimide-PEG11-biotin for the latter (Thermo Scientific Pierce Protein Research, IK, USA) (7). Methylamine-treated C3 (C3MA) was prepared by mixing 2.7 µM C3 with 300 mM methylamine hydrochloride (Sigma Aldrich) in VBS++ buffer (Veronal Buffered Saline pH 7.4, 0.25 mM MgCl2, 0.5 mM CaCl2). This reaction was incubated at 37°C for 1 h and dialyzed overnight to VBS++ buffer at 4°C. FB with N-terminal His-tag and FD were expressed recombinantly as previously described (U-Protein Express, Utrecht, The Netherlands) (7). C4 was isolated from blood from a healthy individual that was anti-coagulated with 20 mM EDTA. Plasma was collected and protease inhibitors (10 mM benzamidine, 1 mM PMSF, 7.5 µM SBTI, EDTA 5 mM, 2.1 mM Pefabloc SC, 30 µM NPGB) were added quickly, while stirring the plasma at 4°C. To remove large complexes, plasma was precipitated with 4% PEG 6000, which was added slowly to the plasma for 45 min. After centrifugation, the supernatant was isolated from which plasminogen was removed by adding 20 mM EDTA and Lysine-Sepharose and incubation for 1 h at 4°C. From the supernatant, C4 was isolated by SourceQ anion exchange. Loading was performed in 50 mM Tris–HCl, 100 mM NaCl, pH 8.0 (containing 1 mM benzamidine, 1 mM PMSF, 30 mM EACA, and 5 mM EDTA) after which C4 was eluted in a gradient of 100–500 mM NaCl in 50 mM Tris–HCl, 100 mM NaCl, pH 8.0 (containing 1 mM benzamidine, 1 mM PMSF, 30 mM EACA, and 5 mM EDTA). Fractions were analyzed by 10% SDS-PAGE following Instant Blue (Roche) protein staining according to the manufacturer's instructions. C2 with a N-terminal His-tag was expressed in HEK293 cells stably expressing EBNA1 (HEK293E) as described (U-Protein Express, Utrecht, The Netherlands) (19). C2 was purified from expression medium *via* immobilized metal affinity chromatography using a HisTrap column (GE Healthcare). C1 was obtained from Complement Technology Inc. (TX, USA).

#### Complement-Binding Molecules and Proteins

FH and C4b-binding protein (C4BP) were ordered *via* Complement Technology Inc. (Tyler, TX, USA). FHR5 was purchased at R&D systems (Minneapolis, MN, USA). Eculizumab was obtained *via* Genmab (Utrecht, The Netherlands). Cp40 was kindly provided by John Lambris. CRIg was kindly provided by Genentech (South San Francisco, CA, USA). OmCI was produced in HEK293E cells and purified as described previously (20). Efb-C and Efb-C mutant (Efb-C-R131E/N138E) were prepared as previously described (21, 22), as well as Ecb, Ecb mutant (Ecb-N63E/R75E/N82E) (23), and SSL7 (24).

#### Human Monoclonal Antibodies

Monoclonal human anti-DNP-IgG1 was produced recombinantly in human Expi293F cells (Life Technologies). Therefore, the variable region of the heavy chain (>VH7007-DNP-G2a2: DVRLQESGPGLVKPSQSLSLTCSVTGYSITNSYYWNWIRQF PGNKLEWMVYIGYDGSNNYNP SLKNRISITRD T SKNQFFLKLNSVTTEDTATYYCARATYYGNYRGFAYWGQ GTLVTVSA) and light chain (>VL7007-DNP-G2a2: DIRMTQT TSSLSASLGDRVTISCRASQDISNYLNWYQQKPDGTVKLLIY YTSRLHSGVPSRFSGSGSGTDYSLTISNLEQEDIATYFCQQG NTLPWTFGGGTKLEIK) (25) were cloned in the pFUSE-CHIghG1 and pFUSE2-CLIg-hk vector, respectively, according to the manufacturer's description (Invivogen). A KOZAK sequence and the HAVT20 signal peptide (MACPGFLWALVISTCLEFSMA) were included upstream each variable region. Human codon optimized sequences were ordered as gBlocks (Integrated DNA Technologies) for Gibson assembly (Bioke). TOP10F′ *E. coli* were used for propagation of the plasmids. After sequence, verification plasmids were isolated using NucleoBond Xtra Midi plasmid DNA purification (Macherey-Nagel). Transfection of EXPI293F cells was performed using ExpiFectamine 293 reagent according to the manufacturer's description (Life Technologies). 1 µg DNA/ ml cells was used in a 3:2 (hk:hG1) ratio. Cell supernatant was collected after 4 days of transfection and antibodies were isolated using a HiTrap protein A column (GE Healthcare).

#### U937 Cell Lines

U937 human monocyte cells and 293 T human embryonic kidney cells were obtained from American Type Culture Collection and grown (37°C, 5% CO2) in RPMI (Lonza) supplemented with penicillin and streptomycin (Gibco) and 10% FCS (Gibco). For stable expression of human C3aR in U937 cells, a lentiviral expression system was used. The human C3aR cDNA was cloned in a dual promoter lentiviral vector, derived from no. 2025.pCCLsin.PPT.pA.CTE.4x-scrT.eGFP.mCMV.hPGK. NGFR.pre (kindly provided by Dr. Luigi Naldini, San Raffaele Scientific Institute, Milan, Italy) as previously described (26). This altered lentiviral vector (BIC-PGK-Zeo-T2a-mAmetrine; EF1A) uses the human EF1A promoter to facilitate potent expression in immune cells and expresses the fluorescent protein mAmetrine and selection marker ZeoR. Virus was produced in 24-well plates using standard lentiviral production protocols and the third-generation packaging vectors pMD2G-VSVg, pRSV-REV, and pMDL/RRE. Briefly, 0.25 µg lentiviral vector and 0.25 µg packaging vectors were co-transfected in 293 T cells by using 1.5 µl Mirus LT1 tranfection reagent (Sopachem, Ochten, The Netherlands). After 72 h, 100 µl viral supernatant adjusted to 8 mg/ml polybrene was used to infect ~50,000 U937 cells by spin infection at 1,000 g for 2 h at 33°C. U937-C5aR cells were a generous gift from Eric Prossnitz (University of New Mexico, Albuquerque, NM, USA).

### C3 and C5 Conversion in AP Model

Streptavidin-coated magnetic beads (Dynabeads M-270 Streptavidin, Invitrogen) were washed once in VBS-T/Mg [Veronal Buffered Saline pH 7.4, 2.5 mM MgCl2, 0.05% (v/v) Tween]. To prepare fully loaded C3b-beads, beads (4 µl/sample) were resuspended in 0.4 ml VBS-T/Mg per sample with C3b-PEG11 biotin (1 µg/ml) and incubated for 1 h at 4°C on roller. To load beads with different amounts of C3b, five different amounts of beads per sample (4, 8, 16, 32, or 64 µl beads) were incubated in 0.4 ml VBS-T/Mg with 0.6 µg/ml C3b-PEG11-biotin. After C3b-labeling, beads were washed three times and incubated in 100 µl VBS-T/Mg per sample with FB (50 µg/ml) for 30 min at room temperature on roller. After FB incubation, beads were washed three times and incubated in 100 µl VBS-T/Mg per sample with FD (5 µg/ml) and either C3 (20 µg/ml) or C5 (20 µg/ml) and with or without inhibitor at the desired concentration (1 µM or threefold dilution starting from 1 µM) for 1 h at 37°C on shaker. After incubation, supernatant of each sample was collected and kept at −20°C until measurement in calcium mobilization assay.

# C3 and C5 Conversion in CP Model

Streptavidin-coated magnetic beads (Dynabeads M-270 Streptavidin, Invitrogen) were washed once in PBS-TH [Phosphate Buffered Saline pH 7.4, 0.05% (v/v) Tween, 0.5% HSA]. Beads (4 µl/sample) were resuspended in 0.4 ml PBS-TH per sample with 1 µg/ml biotinylated 2,4-dinitrophenol [DNP-PEG2-GSGS GSGK(Biotin)-NH2; 1,186 Da; obtained from Pepscan Therapeutics B.V., The Netherlands] and incubated for 30 min at 4°C on roller. Beads were washed once in PBS-TH and incubated in 0.2 ml PBS-TH per sample with 10 nM human monoclonal anti-DNP-IgG1 for 30 min at 4°C on roller.

After one wash in PBS-TH, beads were incubated in 0.1 ml VBS++-TH [Veronal Buffered Saline pH 7.4, 0.25 mM MgCl2, 0.5 mM CaCl2, 0.05% (v/v) Tween, 0.5% HSA] per sample with 0.8 µg/ml C1 for 30 min at 37°C, shaking. Beads were washed three times in VBS++-TH and incubated in 0.1 ml VBS++-TH per sample with 10 µg/ml C4 for 30 min at 37°C, shaking. After three washes in VBS++-TH, beads were incubated in 0.1 ml VBS++-TH per sample with 10 µg/ml C2, 10 µg/ml C3, and 0.5 µg/ml C5 with or without 1 µM inhibitor for 5 min at 37°C on shaker. After incubation, the supernatant of each sample was collected and kept at −20°C until measurement in calcium mobilization assay.

In some CP experiments, the amount of deposited C3b was influenced by adding lower concentrations of C3 (threefold decrease starting from 10 µg/ml). In one condition, C3 and C5 conversion were separated by incubating beads first in 0.1 ml VBS++-TH per sample with 10 µg/ml C2 and 10 µg/ml C3 for 5 min at 37°C and subsequently, after washing, in 0.1 ml VBS++-TH with 10 µg/ml C2 and 0.5 µg/ml C5 for 5 min at 37°C. Supernatant of both C2 + C3 and C2 + C5 incubation were here collected and used for calcium mobilization in U937-C3aR and U937-C5aR cells, respectively. Controls were carried out with 10 µg/ml C3bH2O or 10 µg/ml C3MA.

# Calcium Mobilization Assay With U937 Cells

U937-C3aR and U937-C5aR cells were washed in RPMI/0.05% HSA and diluted to 5 × 106 cells/ml. Cells were incubated with 0.5 µM Fluo-3-AM (Invitrogen) on roller in dark at room temperature for 20 min, washed and resuspended in RPMI/1% HSA to a final concentration of 1 × 106 cells/ml. For calcium mobilization measurements, the labeled cells were stimulated with sample supernatant (ratio cells to supernatant is 9:1) while cell fluorescence is measured by flow cytometry (BD FACSVerse) from 10 s before until 40 s after addition of the sample. The absolute calcium mobilization was calculated by subtracting the cell mean fluorescence intensity (MFI) before cell activation (*t* = 5–15 s) from the MFI after stimulation (*t* = 30–50 s) using FlowJo software. Standard curves were obtained using 10-fold dilutions starting from 1 µM of C3a (Complement Technology Inc., TX, USA) or C5a (Bachem, Switzerland) as stimuli for the cells. As negative controls the calcium mobilization in U937- C3aR and U937-C5aR cells induced by 0.1 µM C3 or C5 was measured.

# C3b Binding to Beads

To determine the level of C3b-biotin bound to streptavidincoated beads in the AP model or the level of actively deposited C3b in the CP model, beads were washed three times in PBS-T (AP) or PBS-TH (CP) after the C3b-biotin or C3 incubation and incubated (4 µl beads/sample) in 100 µl PBS-TH per sample with 1:100 FITC-conjugated goat-anti-human C3 (Protos) for 30 min at 4°C. Subsequently, beads were washed three times in PBS-T (AP) or PBS-TH (CP) and C3b binding was analyzed by flow cytometry (BD FACSVerse).

# Statistical Analysis

Statistical analysis was performed with GraphPad Prism 6 software. All calcium flux data are presented as mean ± SD from three independent experiments. C3b binding data are presented as geometric mean ± SD from three independent experiments.

# RESULTS

#### Functional Analysis of C3 Conversion *via* Purified AP Convertases

Previously, we described the development of a model system to study C5 convertases of the AP using purified components (7). The AP C5 convertase is formed when C3 convertases (C3bBb) cleave C3 into nascent C3b that covalently binds to target surfaces *via* the thioester (structural model in **Figure 1C**) (27). At a critical density of C3b molecules, multimeric C3bn complexes arise that have a high affinity for C5. These C3bn complexes, together with FB and FD, generate C5 convertases that bind and cleave C5 (**Figure 1D**). We recently set out to mimic surface-bound high C3b-density using purified proteins. To establish this, we first labeled C3b with biotin *via* the thioester by activating plasmapurified C3 into C3b in the presence of a biotinylation agent that reacts with the cysteine residue of the C3b thioester (28). These biotinylated C3b molecules were subsequently loaded onto small magnetic streptavidin (SA) beads (2.8 µm diameter) and incubated with FB and FD to form surface-bound convertases. C5 conversion was examined by quantifying the release of C5a using a flow cytometry-based calcium mobilization assay (29, 30). In short, U937 cells transfected with the C5a receptor (U937-C5aR) were exposed to sample supernatants containing C5a. Binding of C5a to the C5a receptor mediates intracellular calcium release, which is detected by a fluorescent indicator. To get more insights into convertase formation and inhibition, we here extended this model to also study C3 conversion by purified AP convertases. First, we transfected U937 cells with the C3a receptor (U937-C3aR) and showed that purified C3a, but not C3, successfully triggers the mobilization of intracellular calcium in U937-C3aR cells (**Figure 2A**). Low-level activation by purified C5a is likely due to low levels of endogenous C5aR expression in these cells (**Figure 2A**). Second, C3b-coated beads were incubated with FB, FD and C3 and release of C3a was determined in supernatants *via* calcium mobilization in U937- C3aR (**Figure 2B**). C3a-dependent calcium flux specifically required all convertase components. In our previous study, we determined that high C3b-density is essential to effectively convert C5 (7). By maintaining a constant concentration of C3b in each sample while increasing the number of beads, we artificially lowered the local concentration of C3b on the surface (Figure S1 in Supplementary Material). Here, we find that while lowering C3b surface density reduces C5 conversion by AP convertases, it does not lead to less C3 conversion (**Figure 2C**). These results demonstrate key differences in the conditions required for C3 and C5 conversion on a surface in the AP.

#### Select C3b-Binding Molecules Inhibit AP C5 but Not C3 Conversion

Having established a system to study both C3 and C5 conversion by AP convertases, we could now dissect whether known convertase inhibitory molecules block cleavage of both substrates. We focused on studying C3b- or C5-binding molecules for which the binding sites to C3b/C5 have been determined. Each of these molecules has been extensively characterized in previous studies and is known to influence convertase activity in physiological environments (**Table 1**). The structures of the C3b-binding molecules in complex with the C3bBb convertase are shown in **Figure 3A**. The inhibitors include naturally occurring complement inhibitors derived from humans [C3b-binders CRIg (31, 32), FH (33, 34), and FHR5 (35)], bacteria [homologous C3b-binders Efb-C (36) and Ecb (16, 23) and C5-binder SSL7 (37)], and ticks [C5-binder OmCI (20)], or therapeutic inhibitors eculizumab [Soliris, a clinically approved antibody against C5 (38)] and Cp40 [a compstatin analog, strong C3- and C3b-binding molecule (39)]. As expected, we observed that the three C5-binding molecules SSL7, OmCI, and eculizumab specifically interfered with C5 conversion but not C3 conversion (**Figure 3B**). C4BP (40) was included as a negative control for inhibition, as it should not have an effect on AP C3 or C5 convertases, since these convertases lack C4b (**Figure 3B**). Next, we analyzed the activity of C3b-binding molecules on C3 versus

Figure 2 | Development of an alternative pathway (AP) C3 convertase model. (A) C3a specifically induces calcium mobilization in U937-C3aR cells, while C3 and C5 do not. (B) C3 conversion by AP convertases on beads was analyzed in a calcium mobilization assay with U937-C3aR cells. C3a could only be detected in the sample supernatant in the presence of all AP components. C5 conversion in the AP model did not induce calcium flux in the U937-C3aR cells. (C) AP C3 and C5 conversion were performed on beads coated with different densities of C3b and analyzed by calcium mobilization assay with U937-C3aR and -C5aR cells, respectively. A high density of C3b molecules on the target surface enhances C5 but not C3 conversion. (A–C) Data of three independent experiments, presented as mean ± SD.

#### Table 1 | Overview of complement inhibitors used in this study.


C5 conversion. While C3b-binding molecules Cp40, CRIg, and FH blocked both C3 and C5 conversion by AP convertases, host regulatory protein FHR5, and staphylococcal immune evasion proteins Efb-C and Ecb specifically blocked C5 conversion, while leaving C3 conversion unaffected (**Figure 3C**). Mutants of Efb-C/Ecb proteins that cannot bind C3b could not inhibit C5 conversion, confirming that the observed inhibition is mediated through interaction with C3b (**Figure 3C**) (21, 23). Furthermore, Efb-C, Ecb, and FHR5 all inhibit AP C5 conversion in a concentration-dependent manner (**Figure 3E**), but do not affect C3 conversion (**Figure 3D**). As a control, we observed that Cp40, which showed inhibition of both AP C3 and C5 conversion at 1 µM concentration, also inhibits both C3 and C5 conversion in a concentration-dependent manner (**Figures 3D,E**). Thus, these data suggest that by binding C3b, selective inhibition of AP C5 convertases is possible. AP C5 convertases are similar to AP C3 convertases, but contain accessory C3b molecules (C3bn) that enable efficient C5 conversion on surfaces. Since Efb-C, Ecb, and FHR5 only affect C5 (and not C3) conversion, they likely inhibit through affecting accessory C3b molecules on the bead surface. However, the fact that AP C3 and C5 convertases both contain C3b confounds the ability to independently study the role of accessory C3b molecules in C5 conversion.

Figure 3 | C3b-binding molecules FHR5 and Efb-C/Ecb selectively inhibit C5 conversion by alternative pathway (AP) convertases. (A) Structural models of the C3bBb convertase with C3b-binding molecules. The structure of C3bBb [from the C3bBb-SCIN dimer structure, PDB 2WIN (55)] is identical to the convertase shown in Figures 1C,D with C3b (gray) and Bb (orange) shown as ribbons, and in the same orientation, with the C3b–C3b dimerization site (as shown in Figure 1C, right) on the left side of the convertase. C3b-binding molecules are shown as molecular surfaces; Cp40 (magenta) is based on the C3c–compstatin complex structure [PDB 2QKI (41)], CRIg (dark red) is based on the C3b–CRIg complex structure [PDB 2ICF (31)], FH is based on the C3b–FH (1–4) structure [light blue, PDB 2WII (43)] and the C3d–FH (19–20) structure [dark blue, PDB 2XQW (44)], FHR5 (yellow) is modeled from the C3d–FH (19–20) structure [PDB 2XQW (44)], Efb-C (spring green) is based on the C3d–Efb-C structure [PDB 2GOX (21)], and Ecb (spring green) is based on the C3d–Ehp structure [PDB 2NOJ (23)]. For FH and FHR5 only some domains are structurally resolved. Additional domains are represented by colored ovals. (B–E) Conversion of C3 and C5 (0.1 µM) in the AP convertase model in the presence of complement-binding molecules measured by calcium mobilization in U937-C3aR and U937-C5aR cells, respectively. Conversion is shown as a percentage relative to the control without inhibitor. Data of three independent experiments, presented as mean ± SD. (B) C5-binding molecules SSL7, OmCI, and eculizumab (all 1 µM) inhibit AP C5 but not C3 conversion. C4b-binding protein has no effect in the AP model. (C) C3b-binding molecules Cp40, CRIg, and FH prevent both C3 and C5 conversion, while FHR5, Efb-C, and Ecb selectively inhibit C5 conversion. Mutant Efb-C and mutant Ecb are unable to bind C3b and thus do not exhibit inhibition. (D) FHR5, Efb-C, and Ecb do not affect C3 conversion, but (E) selectively inhibit C5 conversion in a dose-dependent manner. Cp40 was used as positive control, inhibiting both C3 and C5 conversion.

# Development of a Purified Classical Pathway C3/C5 Convertase Model

To more closely investigate convertase inhibitory mechanisms of complement inhibitors, we next developed a model to study C3/C5 conversion *via* purified CP convertases. In this pathway, modulation of accessory C3b molecules can be better analyzed since the CP C3 convertase (C4b2a) does not contain C3b. We have used a model that fully recapitulates the CP activation pathway using purified complement components. Streptavidin beads were labeled with biotinylated DNP antigen (**Figure 4A**) and sequentially incubated with recombinant human IgG1 recognizing DNP, C1, C4, C2 and substrates C3 and C5. Similar to the AP, we detected the release of C3a and C5a in the sample supernatant *via* calcium mobilization in U937-C3aR and U937-C5aR cells, respectively. The lack of calcium mobilization by the sample supernatants in the absence of DNP, IgG or the individual complement proteins demonstrates the necessity of functional CP convertases on the bead surface for C3/C5 cleavage (**Figure 4B**). Our results also showed that there was no cross-reactivity between mismatched ligands and receptors. The absence of calcium flux in U937-C5aR cells by samples lacking C5 (and thus C5a) showed that C3a generated in these samples does not interfere with C5a detection. As calcium mobilization levels in U937-C3aR cells were not affected by the presence of C5, interference of C5a with C3a-specific detection could be excluded, as well. Unlike our AP C5 convertase model where we artificially coupled C3b to the bead surface, deposition of C3b in the CP could only be established by natural C3 conversion *via* C4b2a. By adding different concentrations of C3 to beads with naturally formed C4b2a convertases, we influenced the level of C3 conversion and thus C3b deposition on the bead surface (**Figures 4C,D**). Indeed, CP C5 conversion was highly dependent on the level of deposited C3b (**Figure 4E**). As a control, we showed that C5 conversion specifically depends

but is not affected by C3bH2O or C3MA in solution. Uncoupling C3 and C5 conversion by introduction of an extra washing step does not alter C5 conversion, indicating that CP C5 conversion only depends on deposited C3b molecules around existing C3 convertases. (B–E) Data of three independent experiments,

on covalently deposited C3b since addition of non-reactive C3 variants [C3bH2O in which the thioester had already reacted with water or methylamine-treated C3 (C3MA) (7, 56)] did not induce the convertase to cleave C5 (**Figures 4C–E**). In addition, introducing a washing step in between C3b deposition and C5 conversion confirmed that only deposited C3b and not C3 or the active process of C3 cleavage is required for C5 conversion (**Figure 4E**). These results demonstrate the specificity of our CP convertase models in measuring the activity of CP C3 and C5 conversion.

### C3b-Binding Molecules Inhibit Classical Pathway C5 Conversion by Modulating Accessory C3b

Next, we examined the effect of the above-tested C3b- and C5-binding molecules in the CP convertase model. Beads with actively formed C4b2a convertases were incubated with C2, C3, and C5 in the presence of complement binding molecules at a concentration of 1 µM and the C3a/C5a generated in the sample supernatant was measured. The known CP convertase inhibitor C4BP effectively inhibited both CP C3 and C5 conversion, confirming the validity of our model (**Figure 5A**). Furthermore, C5-binding molecules OmCI and eculizumab potently inhibited CP C5 conversion but left C3 conversion unaffected (**Figure 5A**). SSL7 also inhibits CP C5 conversion, but to a lesser extent, which could arise from differences in the models or in the C5-binding sites involved in convertase recognition. Since C3b is not part of the CP C3 convertase, we hypothesized that C3b-binding molecules would not influence C3 conversion in the CP model. Accordingly, most C3b-binding molecules did not affect C3 conversion, with the exception of Cp40, which showed potent inhibition (**Figure 5B**). The lack of C3 conversion in the presence of Cp40 can be explained by its strong affinity for uncleaved C3 causing steric hindrance during C3 recognition by the convertase (57). Efb-C and Ecb, which can also bind C3, do not inhibit

presented as mean ± SD.

convertase activity in that manner, as further evidenced by lack of inhibition of AP C3 conversion. Interestingly, we found that all C3b-binding molecules can prevent C5 conversion in the CP model (**Figure 5B**). This establishes an important role for accessory C3b in the formation and activity of CP C5 convertases. Moreover, the fact that FHR5, Efb-C, and Ecb exhibited similar effects on C5 conversion by the AP and CP convertases, indicates a similar role for accessory C3b in C5 conversion in both pathways. The CP model provides more detail about their inhibitory mechanism by showing that they can act specifically *via* accessory C3b molecules. Similarly, the CP inhibition data show that CRIg and FH can inhibit C5 conversion specifically *via* accessory C3b. Overall, our data suggest that all C3b-binding molecules tested can inhibit C5 conversion (in both AP and CP) through interaction with accessory C3b molecules, but only some can inhibit the core convertase enzyme (C3bBb) itself.

#### DISCUSSION

Since C3 and C5 convertase enzymes play such a vital role in propagating the cascade but also driving unwanted complement effector functions, it is essential to better understand mechanisms of convertase activation and inhibition. Since C5 convertases are largely constrained to cell surfaces, it has been difficult to study these enzymes with highly purified complement components. Here, we developed bead-based models to functionally characterize both C3 and C5 convertases on a surface using purified components and in the absence of confounding factors from serum. These models serve several important purposes: (1) to understand the molecular biology of convertases, (2) to characterize the mode of action of known complement inhibitors, (3) to characterize the role of disease-associated deficiencies and mutations of complement proteins, and (4) to screen for novel and specific convertase inhibitors.

In the past, several models have been developed to study convertase activation and inhibition. One of the most common models employs erythrocytes to serve as a platform for complement activation, using either serum or stepwise addition of purified complement components. Other studies have employed non-cellular surfaces, including SPR chips, to examine stepwise assembly and dissociation of convertases. While each of these models differ in many aspects, each has inherent advantages and disadvantages in addressing various aspects of complement function, and no single model can capture all molecular and physiological details of convertases or inhibition thereof. Our models offer the ability to (1) quantitatively compare activity and inhibition of AP and CP C3 and C5 convertases independently and (2) enable controlled formation and distribution of convertases in a highly purified environment in the absence of complement regulators found in serum and on cells. In our model, we chose to quantify C3 and C5 cleavage through measurement of chemoattractants C3a and C5a, molecules that are released into solution and can be selectively and sensitively detected in a functional cell-based calcium mobilization assay using flow cytometry. Alternatively, C3a and C5a can be quantified by ELISA, however, this is not a direct functional readout, and one should exercise caution in selecting antibodies with high specificity for each chemoattractant molecule (58). Measurement of C3a *via* calcium responses allows a more accurate quantification of C3 convertase activity than antibody detection of deposited C3b molecules. During C3 cleavage, the thioester of newly formed C3b molecules becomes exposed and can react with molecules on the cell surface. Rapid amplification results in dense clusters of C3b, which may deposit on top of each other, making accurate quantification difficult (59). Furthermore, many newly formed C3b molecules never attach to the surface (60). Therefore, immunodetection of deposited C3b is not the best measure of C3 conversion. In addition, since C3a and C5a are hallmarks of complement-mediated inflammation, detection of these chemoattractants is a critical readout when screening for convertase inhibitors as potential therapeutic molecules and disease-associated mutants of complement factors *in vitro.* It is important to note that in more complex environments (i.e., serum or *in vivo*), measurement of functionally active C3a/C5a is challenging due to proteolytic cleavage and scavenging by receptors. In addition, our bead-based models enable additional readouts, including quantification of surface complement deposition and breakdown of complement opsonins through cofactor activity of inhibitory molecules (Figure S2 in Supplementary Material).

The models presented in this work may assist in obtaining better insights into the structural organization of convertase enzymes. While significant progress has been made in understanding the structural organization of C3 convertases and C3 cleavage (**Figure 1C**) (55), the molecular details of C5 convertase formation remain poorly understood. Molecular models of C5 convertase activation have been proposed (**Figure 1D**) (47, 49, 61) but the exact organization of this complex remains unknown. It is known that C5 convertases form when C3 convertases (C4b2a and C3bBb) deposit high densities of C3b molecules on the target surface (9). The non-catalytic subunits of C3 convertases (C4b or C3b) are thought to associate with extra C3b molecules and form multimeric C4b-C3bn or C3b-C3bn complexes that have an increased affinity for C5 (62, 63). In this study, we verified the requirement of high C3b densities for C5 conversion. In line with previous data, we also find that C3b density affects C5, but not C3 conversion by AP convertases (7). Interestingly, our data for Efb-C/Ecb and FHR5 also suggest that the orientation of C3b molecules on the surface is particularly important for conversion of C5, but not C3. We previously demonstrated that the (natural) surface attachment of C3b molecules *via* the thioester is crucial for efficient conversion of C5 (7). Here, we found that three molecules that interact with the C3b thioester domain (TED) (Efb-C, Ecb, and FHR5) selectively inhibit AP and CP C5 convertases, while leaving C3 conversion unaffected. Among examined C3b-binding molecules, this selective inhibition of C5 conversion in the AP was specific for molecules interacting with the TED of C3b. Several crystallographic structures of C3b revealed interdomain interactions between TED and the MG1 domain, which facilitate the prototypical "upright" conformation of C3b attached to surfaces *via* its thioester (31, 43, 55, 64–69). However, recent electron microscopy data reveal conformational flexibility of C3b under different conditions, and in particular, TED can exhibit markedly different positions (70–74). Hydrogen-deuterium exchange experiments demonstrated a conformational change in C3b upon Efb-C binding to TED, suggesting that Efb-C acts as a wedge to disrupt the TED–MG1 interaction and affects the orientation of C3b on the surface (71). Although the exact binding interface of C3b and FHR5 is unknown, it does interact with TED (75), and therefore it is possible that FHR5 acts through a similar mechanism. It is unclear whether FHR5 also interacts with other regions of C3b. Binding of C3b by FHR5 is different from that by FH, because FHR5 lacks domains homologous to the FH N-terminal C3b binding and complement regulatory domains. The results reported here confirm the previously reported lack of solid phase C3 convertase inhibition by FHR5 (45, 76), although inhibition of fluid phase C3 convertase was described (45). Why the C3b orientation is critical for C5 conversion but not C3 conversion remains to be determined. Potentially it supports the recently proposed model in which C5 needs to be "sandwiched" in between the C3 convertase and the accessory C3b molecule in order to be primed for convertase cleavage (49). One could envision that such "sandwiching" is affected by differently oriented C3b's. Alternatively, C3b orientation may determine how closely C3b molecules can pack together. If tightly packed and aligned accessory C3b molecules are required for efficient C5 conversion, altered orientation may result in decreased conversion of C5 (**Figure 6**). The fact that Efb/Ecb and FHR5 also inhibit C5 conversion in the CP suggests that the accessory C3b molecules have a similar function in the activation of both the CP and AP C5 convertases.

Functional analyses of well-defined complement inhibitors also reveal other important binding interfaces of C3/C5 convertases.

Figure 6 | Model for selective inhibition of C5 conversion. (A) Under normal conditions, high levels of accessory C3b (C3bn) deposition on a cell surface around existing C3 convertases enables binding and conversion of C5. All C3b molecules are tightly packed and aligned in the same vertical orientation, both of which are necessary for efficient C5 conversion. (B) Thioester domain (TED)-binding molecules (i.e., Efb-C and Ecb) act as a wedge to separate TED from the rest of C3b, which alters its surface conformation and prevents efficient C5 conversion. Colors of molecules correspond to Figures 1C,D and 3A, with Efb-C/Ecb shown in spring green. The red crosses indicate inhibition of C5 and C3bBb binding to surface-bound accessory C3b molecules (C3bn).

Our model demonstrates that the C3b–C3b dimerization site (**Figure 1C**, right) is important for activity of all C3b-containing convertases, including AP C3 and C5 convertases, as well as the CP C5 convertase. The C3b-binding molecules Cp40 and CRIg, which bind at this interface (**Figure 3A**), universally inhibit activity of C3b-containing convertases. These molecules likely interfere with convertase-substrate binding (as in **Figures 1C,D**) and/ or surface-bound accessory C3b molecules (as in **Figure 1D**). Unlike CRIg, Cp40 inhibits all convertases, including the CP C3 convertase, which lacks C3b. Since Cp40 can bind to both C3b and uncleaved C3, it could inhibit CP C3 conversion by binding to the substrate (C3) and preventing recognition by the CP C3 convertase (57, 77). Substrate C3 binding can also explain the difference in inhibition of AP C3 and C5 conversion by Cp40 (**Figures 3D,E**). In the AP C3 conversion assay, Cp40 can bind to both C3 and C3b, and therefore the concentration required to block C3 conversion is higher than for in the AP C5 conversion assay, where C3 is not present. Next to CRIg, we also found that FH inhibits CP C5 convertases. Although FH is known as an inhibitor of AP C3/C5 convertases because it dissociates Bb from C3b, the mechanism of FH-mediated inhibition of CP C5 conversion is not known. Since FH is a large molecule with several distant binding sites on C3b (43, 44, 78) it likely interferes with binding C5 (79). Thus, our data for FH demonstrate that not only the C3b–C3b dimerization site, but also other sites on C3b, are important for its interaction with C5. Overall, C3b-binding molecules illustrate several key properties of convertase assembly and inhibition. While much of the data here are in line with previous inhibitor studies (**Table 1**), more extensive and complementary studies are required to fully understand the physiological modes of inhibition of these molecules.

Finally, the tools developed in this study can be used for identification of effective therapeutic convertase inhibitors. The ability to examine each convertase separately affords the opportunity to identify selective convertase inhibitors. The complement therapeutics landscape is rapidly expanding, as new roles for complement in disease continue to be uncovered. It is clear that not all complement-mediated diseases are created equal, and it is necessary to design therapeutics that target different points in the complement cascade. For example, diseases mediated

#### REFERENCES


primarily by C5a or MAC may benefit from selective inhibition of C5 conversion. Blocking the complement cascade upstream of the terminal pathway (i.e., inhibition of C3 cleavage) may unnecessarily increase patient susceptibility to infections by effectively inhibiting all complement effector functions. Our work now demonstrates the characterization of inhibitors that selectively inhibit C5 conversion, which may prove useful in treatment of MAC-mediated and inflammatory disorders of the complement system. Thus, these models provide a platform for the identification of tailored next-generation complement therapeutics.

# AUTHOR CONTRIBUTIONS

SZ, EB, and SR designed and developed convertase models. SZ, EB, SR, and RG designed the study and experiments. SZ, SM, MR, and RG performed the experiments and analyzed data. CH developed cell lines. CH and PA cloned and produced monoclonal antibodies. SZ, SM, SR, and RG wrote the manuscript. SZ, MJ, SR, and RG contributed to critical analysis and discussion of the results. All authors read and reviewed the manuscript.

#### ACKNOWLEDGMENTS

We would like to acknowledge John Lambris for providing Cp40, Menno van Lookeren Campagne (Genentech) for providing CRIg, Genmab for providing eculizumab, Brian Geisbrecht for providing constructs for Efb-C and Ecb mutants, U-Protein Express BV (Utrecht, The Netherlands) for help with protein expression, and Kaila Bennett, Bart Bardoel, and Julia Kolata for cloning advice. The work was funded by Marie-Skłodowska Curie Fellowship (659633, to RG); ERC Starting grant (639209-ComBact, to SR), the EMBO Young Investigator Program (to SR); Kidneeds Foundation, Iowa, US (to MJ). MJ was also supported by the Institutional Excellence Program of the Ministry of Human Capacities of Hungary.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01691/ full#supplementary-material.


protein 1 reveal functional determinants of complement regulation. *J Immunol* (2016) 196:866–76. doi:10.4049/jimmunol.1501919


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Zwarthoff, Berends, Mol, Ruyken, Aerts, Józsi, de Haas, Rooijakkers and Gorham. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Yun Song1†, Kun-Yi Wu1†, Weiju Wu <sup>2</sup> , Zhao-Yang Duan3 , Ya-Feng Gao1 , Liang-Dong Zhang4 , Tie Chong <sup>4</sup> , Malgorzata A. Garstka1 , Wuding Zhou2 and Ke Li <sup>1</sup> \**

*1Core Research Laboratory, The Second Affiliated Hospital, School of Medicine, Xi'an Jiaotong University, Xi'an, China, 2Medical Research Council (MRC) Centre for Transplantation, King's College London, Guy's Hospital, London, United Kingdom, 3Department of Nephrology, The Second Affiliated Hospital, School of Medicine, Xi'an Jiaotong University, Xi'an, China, 4Department of Urology, The Second Affiliated Hospital, School of Medicine, Xi'an Jiaotong University, Xi'an, China*

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Peter Monk, University of Sheffield, United Kingdom Lubka T. Roumenina, INSERM UMRS1138 Centre de Recherche des Cordeliers, France*

#### *\*Correspondence:*

*Ke Li ke.li@mail.xjtu.edu.cn These authors have contributed* 

*† equally to this work.*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 04 March 2018 Accepted: 17 April 2018 Published: 01 May 2018*

#### *Citation:*

*Song Y, Wu K-Y, Wu W, Duan Z-Y, Gao Y-F, Zhang L-D, Chong T, Garstka MA, Zhou W and Li K (2018) Epithelial C5aR1 Signaling Enhances Uropathogenic Escherichia coli Adhesion to Human Renal Tubular Epithelial Cells. Front. Immunol. 9:949. doi: 10.3389/fimmu.2018.00949*

Recent work in a murine model of ascending urinary tract infection has suggested that C5a/C5aR1 interactions play a pathogenic role in the development of renal infection through enhancement of bacterial adhesion/colonization to renal tubular epithelial cells (RTECs). In the present study, we extended these observations to human. We show that renal tubular epithelial C5aR1 signaling is involved in promoting uropathogenic *Escherichia coli* (UPEC) adhesion/invasion of host cells. Stimulation of primary cultures of RTEC with C5a resulted in significant increases in UPEC adhesion/invasion of the RTEC. This was associated with enhanced expression of terminal α-mannosyl residues (Man) (a ligand for type 1 fimbriae of *E. coli*) in the RTEC following C5a stimulation. Mechanism studies revealed that C5aR1-mediated activation of ERK1/2/NF-κB and upregulation of proinflammatory cytokine production (i.e., TNF-α) is at least partly responsible for the upregulation of Man expression and bacterial adhesion. Clinical sample studies showed that C5aR1 and Man were clearly detected in the renal tubular epithelium of normal human kidney biopsies, and UPEC bound to the epithelium in a d-mannose-dependent manner. Additionally, C5a levels were significantly increased in urine of urinary tract infection patients compared with healthy controls. Our data therefore demonstrate that, in agreement with observations in mice, human renal tubular epithelial C5aR1 signaling can upregulate Man expression in RTEC, which enhances UPEC adhesion to and invasion of RTEC. It also suggests the *in vivo* relevance of upregulation of Man expression in renal tubular epithelium by C5a/C5aR1 interactions and its potential impact on renal infection.

Keywords: C5aR1, uropathogenic *Escherichia coli*, renal tubular epithelial cell, bacterial adhesion/invasion, mannosyl residues

#### INTRODUCTION

Urinary tract infections (UTIs) remain among the most common infectious diseases worldwide (1). Globally, there have been estimated over 250 million cases of UTI each year (2). It is frequent in women and children and causes a particular problem for patients with diabetes and renal transplants as well as catheterized patients (3–5). The most common causative organism is uropathogenic *Escherichia coli* (UPEC), which is responsible for around 80% of all cases. UPEC express a variety of fimbrial adhesins (e.g., type 1, P, and Dr fimbriae) that enable them to bind to glycoproteins or glycolipids on urinary tract epithelial cells (the critical initial step in early colonization). Upon contact with the epithelial cells, UPEC liberate toxins (e.g., α-hemolysin and cytotoxic necrotizing factor-1) which mediate direct injury to the cells, disrupting the epithelial barrier, and opening access to the underlying tissue. In addition, UPEC can be uptaken by the epithelial cells (6–8), which could evade host defenses and establish reservoir that may act as a potential source for recurrent infections (9).

Recent research has suggested that the epithelial cell lining of urinary tract is not simply a passive target for infection, oxidative stress, and toxical drug, but may actively participate in the innate immune response (10). For example, in response to the invasion of pathogen or other injuries, tubular epithelial cells produce a number of proinflammatory mediators, which can activate and recruit immune cells to the site of infection and injury, contributing to innate immunity; however, excessive inflammatory responses can also cause renal tubule cell injury (10–12). Human renal tubular epithelial cells (RTECs) produce various components of the complement system (e.g., C3, fB, and fH) (13–15), local production, and activation of complement have been implicated in the pathogenesis of many renal tubular disorders (16, 17).

Complement is an important element in innate immunity against pathogens, mainly through direct killing, opsonization of pathogen, and induction of local inflammation, which are mediated by C5b-9, C3b, and C5a, respectively. However, most UPEC strains are resistant to complement-mediated killing (18). Previous studies by us and others have suggested that complement activation in UTI has harmful effects on the host (7, 8, 19). It has been shown that depletion of complement in non-human primate decreases the degree of tissue damage during renal infection (19) and mice deficient in C3, a pivotal component of the complement cascade, are protected from the development of renal infection (7). C3b-opsonized *E. coli* enhances the invasion of human tubular epithelial cells through interaction with a complement regulatory protein CD46 expressed on the epithelial cell membrane (8).

Further studies in murine models of acute and chronic pyelonephritis have found that C5aR1 is required for the development of renal infection (20, 21). It has been shown that kidney infection is significantly reduced in mice with genetic deletion or through pharmacologic inhibition of C5aR1 following bladder inoculation with UPEC. This is associated with reduced expression of terminal α-mannosyl residues (Man; a ligand for type 1 fimbriae of *E. coli*) on the luminal surface of renal tubular epithelium and reduction of early UPEC colonization in these mice. These studies strongly suggest that C5a/C5aR1 interactions are an important pathogenic mechanism of ascending UTI. However, the relevance of these findings to human remains unknown.

In the present study, we extended our observations made in murine models to human. We investigated whether renal tubular epithelial C5aR1 signaling has impact on UPEC adhesion to and invasion of RTEC using primary cultures of human RTEC. We also examined the potential mechanisms by which epithelial C5aR1 signaling influences bacterial adhesion. Furthermore, we performed several *ex vivo* experiments using human kidney biopsies and urine samples to assess the *in vivo* relevance of *in vitro* observations. Our data demonstrate that epithelial C5aR1 signaling promotes UPEC adhesion to and invasion of human RTEC through upregulation of MR expression in the RTEC. It also suggests the *in vivo* relevance of upregulation of Man expression in renal tubular epithelium by C5a/C5aR1 interactions and its potential impact on renal infection.

### MATERIALS AND METHODS

#### Materials

We used the following reagents and materials: mouse anti-human C5aR1 (P12/1, for immunochemistry) (AbD Serotec, Oxford, UK), PE-conjugated anti-human C5aR1 (S5/1, for flow cytometric analysis) (Biolegend, San Diego, CA, USA); polyclonal rabbit anti-human ZO-1 and 4',6-diamidino-2-phenylindole (DAPI) (Life Technologies Ltd., Beijing, China); goat anti-mouse alexa Fluor® 488 (Jackson ImmunoResearch Lab Inc., West Grove, PA, USA); fluorescein-labeled *Galanthus nivalis* lectin (GNL) (Vector Laboratories, Peterborough, UK); cytopainter F-actin staining kit (Abcam, Cambridge, UK). BD™ cytometric bead array human inflammatory cytokines kit (to measure TNF-α, IL-8, IL-1β, and IL-6) and human C5a ELISA kit II (BD OptEIA™) (BD Biosciences, San Jose, USA). Antibody reagents used in signaling pathway studies [i.e., anti-phospho-ERK1/2 (Thr202/Tyr204), -IκBα (Ser32), -SAPK/JNK (Thr183/Tyr185), -Akt (Ser473), and anti-ERK1/2, -IκBα, -JNK, and -Akt antibodies] and ERK1/2 inhibitor U0126 were purchased from Cell Signaling Technology (Danvers, MA, USA). Cell culture medium, fetal calf serum (FCS), insulintransferrin-selenium solution, gentamicin (Life Technologies Ltd.); recombinant human TNF-α (Peprotech, Suzhou, China); d-(+)- Mannose, hydrocortisone, tri-iodothyronine and tetramethylrhodamine (TRITC), protease inhibitor cocktail, LPS (from *E. coli* with serotype O55:B5, contains O antigen) (Sigma-Aldrich, Shanghai, China); collagenase type II (Worthington Bio. Co., USA); human recombinant C5a (with an endotoxin level of ≤0.1 EU per 1 µg of protein) (R&D systems). C5aR1 peptide antagonist [PMX53, Ac-Phe-cyclo (Orn-Pro-dCha-Trp-Arg)] and control peptide (random sequence) (synthesized by GenScript, Shanghai, China).

#### Bacterial Strains

Uropathogenic *E. coli* strain J96 (serotype O4; K6), isolated from a human pyelonephritis patient was kindly provided by Dr. R. Welch, University of Wisconsin, USA. This strain expresses type 1 and P fimbriae and secrets α-hemolysins and the cytotoxic necrotizing factor-1 (18).

# Human Kidney Specimens and Ethics

Normal human kidney specimens were prepared from the unaffected pole of nephrectomized kidneys of patients who had renal tumors (*n* = 4). Each patient gave an informed consent about the present study in accordance with the Declaration of Helsinki. The protocol was approved by the Hospital Research Ethics Committee.

#### Cell Cultures and Treatment

Human RTEC were isolated according to a previously described method (14). Briefly, renal cortical tissue obtained from normal pole of tumor nephrectomy specimens was cut into small fragments and digested with collagenase (type II, 750 U/mL) at 37°C for 15 min and passed through a series of mesh sieves. Tubular cells were isolated by centrifugation and grown in DMEM/F12 with 5% FCS supplemented with insulin (5 µg/mL), transferrin (5 µg/mL), selenium (5 ng/mL), hydrocortisone (40 ng/mL), and tri-iodothyronine (10<sup>−</sup>12 M). The cultured cells exhibited cuboidal morphology (by phase contrast microscopy) and were positive for the expression of alkaline phosphatase and cytokeratin (Figure S1 in Supplementary Material). Experiments were performed with cells up to the sixth passage. Results were obtained from RTEC cultured from the kidney of three different donors. For experiments, confluent layers of RTEC were incubated with C5a (0–50 nM), LPS (800 ng/mL), TNF-α (10 ng/mL), or C5aR1 peptide antagonist (PMX53, 5 µM) for 24 h at 37°C.

# Assessment of Bacterial Binding and Internalization

Bacterial plate count assay: The assay was performed as described previously (8). Briefly, RTEC grown in 24-well plates were pre-incubated with or without C5a or other stimuli for 24 h and then incubated with bacteria [J96, 2 × 106 colony forming units (c.f.u.)/well] for an additional 1 h at 37°C. For bacterial binding, cells were vigorously washed to remove unattached bacteria and lysed with sterile H2O. For bacterial internalization, after incubation with bacteria for 1 h, RTEC were washed three times and then incubated in culture medium containing 100 µg/mL gentamicin for 1 h to kill extra-cellular bacteria. Cells were then washed and lysed with sterile H2O, the lysate was plated onto cysteine lactose electrolyte deficient (CLED) plates (Oxoid). The agar plates were incubated at 37°C for overnight and the colonies were manually counted. RTEC protein concentrations were measured using the Coomassie (Bradford) protein assay kit according to the manufacturer's instructions. Results were expressed as colony c.f.u./mg of cell protein. In each experiment, assays were performed in quadruplicate.

Fluorescence microscopy analysis: RTEC monolayers grown on coverslips in 24-well plates were incubated with TRITCconjugated J96 (2 × 106 c.f.u./well) for 1 h at 37°C. Cells were vigorously washed to remove unattached bacteria and stained with DAPI, then viewed and imaged with the Leica SP8 system. Bound bacteria and RTEC numbers were counted at ×200 magnification, and results were expressed as number of bacteria per 103 tubule cells. To confirm the intracellular location of the bacteria, after incubation with labeled bacteria, the monolayers were washed then fixed in 4% formaldehyde and permeabilized with 0.1% TritonX-100 before incubating with green fluorescence -conjugated F-actin (Abcam) for 30 min. Confocal microscopy was carried out using cross-sectional analysis. To demonstrate the co-localization of bacteria with Man, after incubation with labeled bacteria, cells were vigorously washed to remove unattached bacteria and stained with DAPI and fluorescein-labeled GNL, then viewed and imaged with the Leica SP8 system. Two to three viewing fields, randomly selected from each coverslip at ×630 magnification, were examined.

# Assessment of Bacterial Migration in RTEC

Renal tubular epithelial cell were cultured on 12-well transwell filters (with a 3 µm transparent membrane insert) (Corning) and allowed to grow to confluent monolayers for 10 days. The tightness of the cell layers was verified by addition of 500 µL culture medium in the upper compartment followed by incubation at 37°C for 6 h. No medium was found in the lower compartment when cells confluent. Cells were stimulated with or without C5a (10 nM) for 24 h and then incubated with J96 (2 × 106 c.f.u./ well) for up to 3 h at 37°C. Bacterial migration was measured by plating out the culture medium collected from lower chamber on CLED plate and counting the colonies formed on the plate the next day. RTEC membrane integrity was assessed by immunochemistry and western blot analyses for tight junction marker ZO-1.

# Assessment of Man Expression in Cultured RTEC

Man expression in cultured RTEC was assessed by a fluorescence intensity-based microplate assay or fluorescence microscopic analysis as we described previously (20). For the microplate assay, RTEC grown in 24-well plates were incubated with or without C5a and/or LPS for 24 h at 37°C, then with fluorescein-labeled GNL (20 µg/mL in PBS) for 1 h at 37°C. Cells were then vigorously washed and lysed with sterile H2O. The fluorescence in the lysate was measured using a fluorescence plate reader (SpectraMax i3, Molecular Devices) with an excitation wavelength of 490 nm and results expressed as relative fluorescence units. Fluorescence microscopic analysis was used to detect surface Man on RTEC. RTEC grown on the coverslips were pre-treated with C5a and/or LPS and stained with fluorescein-labeled GNL and DAPI. Images were taken by the confocal microscope (Leica SP8) under ×100 magnification. The percentage of positive staining area in each image was calculated by using Image J software (Image J 1.41). Three to four viewing fields randomly selected from each coverslip were examined.

# **α**-Mannosidase Treatment

Renal tubular epithelial cell were pre-incubated with FCS-free medium for 2 h and then incubated with α-mannosidase or control enzyme (β-galactosidase) (5 mM) in glycol buffer at 37°C for 1 h. FCS-containing culture medium was added to neutralize the effect of mannosidase (22). Cells were then subjected to measuring changes in Man expression and bacterial binding.

# Western Blot

Human RTECs were primarily cultured on six-well plates and grown to confluence. Cells were incubated with C5a and lysed at indicated times. Equal amounts of protein (40 µg per lane) were subjected to SDS-PAGE electrophoresis, and transferred onto PVDF membranes. The membranes were incubated with primary antibody at 4°C overnight and followed by incubation with HRPconjugated secondary antibody. Protein bands were visualized by Amersham ECL Select™ detection reagent (GE Healthcare Life Sciences, USA).

# Detection of C5aR1 Expression in Kidney Tissue by Immunohistochemistry

Paraffin-embedded human kidney tissue sections (4 µm thick, from four different donors) were deparaffinized, rehydrated, and subjected to microwave-based antigen retrieval in citrate buffer. Endogenous peroxidases were blocked with 0.3% (v/v) H2O2 in PBS for 10 min. Non-specific immunoglobulin binding sites were blocked with normal goat serum. Sections were subsequently incubated with mouse anti-human C5aR1 primary antibody (P12/1, AbD Serotec) (1:300 dilution) at 4°C overnight. Nonspecific mouse IgG2a served as a negative control. Sections were then incubated for 1 h with HRP-conjugated goat anti-mouse secondary antibody (Jackson ImmunoResearch Lab Inc., 1:1,000 dilution) at room temperature. The antibody staining was visualized by DAB solution according to manufacturer's instructions. The stained sections were imaged with automatic slide scanner (Axio Scan Z1, Zeiss, Germany).

#### UPEC *In Situ* Binding to Kidney Tissue

Frozen sections (4 µm) of OCT-embedded human kidney samples (*n* = 4) were immobilized on Superfrost™ Plus microscope slides (Fisher Scientific) and were rehydrated in PBS for 5 min and incubated with PBS containing 0.1% BSA for 1 h, and then with TRITC-labeled J96 (2 × 107 c.f.u./mL, pre-incubated with 1 or 5% Glucose or d-mannose for 30 min and resuspended in PBS containing 0.1% BSA) for 1 h at 37°C. Sections were then gently washed five times to remove unattached bacteria and stained with DAPI and fluorescein-labeled GNL, then viewed and imaged with the Leica SP8 system. Bacterial colonies were manually counted at ×200 magnification and results were expressed as number of colonies per field. Five viewing fields at cortex area for each kidney section were examined.

# Statistical Analysis

Data are shown as mean ± SEM. Mann–Whitney test or Unpaired Student's *t*-test was used to compare the means of two groups. One-way or two-way ANOVA was used to compare the means of more than two independent groups. All the analyses were performed using Graphpad Prism 7 software. *P* < 0.05 was considered to be significant.

# RESULTS

#### C5a Enhances UPEC Adhesion and Invasion of RTEC by Stimulating the Epithelial Cells

We first examined the expression of C5aR1 in primary cultures of human RTEC by performing immunochemical staining, flow cytometry, and RT-PCR. C5aR1 expression was clearly detected in the RTEC by the three methods, further confirming that C5aR1 is expressed in human RTEC (Figures S2A–C in Supplementary Material). We also examined whether C5R2 (an additional receptor for C5a) is expressed in the RTEC by RT-PCR, which showed a negative result (Figure 2C in Supplementary Material). Therefore, any effects of C5a stimulation in our RTEC preparations likely through interaction with C5aR1.

Next we assessed the effects of C5a on bacterial adhesion and invasion in human RTEC by stimulation of the RTEC. RTEC were incubated with C5a, in the presence or absence of LPS, and then co-cultured with UPEC strain J96 and followed by a set of assays including bacterial binding, internalization, and transmigration. Bacterial binding to RTEC was assessed by plate count assay and fluorescence microscopy. C5a (alone) stimulation of RTEC significantly enhanced bacterial adhesion to RTEC in a dose-dependent manner (**Figure 1A**). Combined treatment with C5a (10 nM) and LPS (800 ng/mL) is more effective than treatment with LPS alone, the combined treatment also appears more effective than treatment with C5a alone, though there is no statistically significant difference (**Figures 1B–D**). Besides promoting adhesion, C5a stimulation of RTEC also significantly increased bacterial internalization into RTEC, regardless of in the presence or absence LPS, assessed by bacterial plate count assay and fluorescence microscopy (**Figures 2A–C**). Furthermore, C5a stimulation increased bacterial transmigration through the monolayer of RTEC (assessed by c.f.u. of transmigrated bacteria) and caused more severe epithelial barrier damage (assessed by ZO-1 expression) (**Figures 3A–D**).

Collectively, these data demonstrate C5a enhances UPEC adhesion to and invasion of human RTEC by stimulation of the RTEC.

#### C5a-Mediated Upregulation of Man Expression in RTEC Contributes to the Enhancement of UPEC Adhesion to and Invasion of RTEC

We next examined molecular mechanism by which C5a mediates the upregulation of bacterial adhesion/invasion in RTEC. Bacterial-host cell carbohydrate interactions are known to facilitate tissue invasion. Terminal α-Man are the carbohydrate ligand for type 1 fimbriae on UPEC (23, 24). Putative membrane proteins containing such ligands have been reported in the bladder in mouse and human (25). Our recent work has shown that terminal α-Man are also expressed in murine RTEC, the expression is upregulated by C5a stimulation. However, it is unknown whether this is true for human RTEC. We firstly confirmed the expression of Man in human RTEC by showing the cells were positively stained with fluorescence labeled GNL which specifically detects (α-1,3) Man (Figure S3 in Supplementary Material). We next assessed whether C5a influences the Man expression in human RTEC using two methods (i.e., fluorescence microscopy, fluorescence intensity-based microplate assay). Results obtained by both methods showed that C5a stimulation significantly increased Man expression in RTEC, the stimulatory effect was more profound in the presence of a small dose of LPS (800 ng/mL) (**Figures 4A–C**). Blockade of C5a/C5aR1 interactions with C5aR1 antagonist (PMX53), significantly reduced C5a-upregulated Man expression in RTEC and bacterial adhesion (**Figures 4D,E**), suggesting that C5a upregulates Man expression through interaction with C5aR1 in human RTEC.

We also verified adhesion of UPEC to the RTEC *via* binding to Man on cell surface. The confocal Z-stack images showed a spatial relation between Man and J96 on RTEC cell surface (**Figure 5A**).

Addition of d-mannose [which binds to lectin (Fim H of type 1 fimbriae)] in RTEC and UPEC co-culture led to a significant decrease in bacteria binding to RTEC (**Figure 5B**), demonstrating specific binding of bacteria to Man. Removal of surface Man from RTEC with α-mannosidase (which cleaves mannose containing glycoprotein and glycolipid) significantly reduced Man expression on RTEC and binding of J96 to RTEC (**Figures 5C,D**). Furthermore, removal of surface Man from RTEC inhibited the enhancement effect of C5a on Man expression and bacterial adhesion (**Figures 5E,F**), supporting the effect of C5a on bacterial adhesion is depended on upregulation of Man expression in RTEC.

Collectively, these data suggest that C5a/C5aR1 interactionmediated upregulation of Man expression in RTEC contribute to the enhancement of bacterial adhesion/invasion in RTEC.

### Investigate Intracellular Signaling/ Molecules Responsible for the Action of C5a on Man Expression in RTEC

Having demonstrated upregulation of Man expression in human RTEC by C5a/C5aR1 interactions, next we investigated which intracellular signaling pathways triggered by C5a could contribute

to the Man expression in human RTEC. C5aR1 is a G-proteincoupled receptor for C5a, engagement of C5aR1 mediates various intracellular signaling events (e.g., PI3K, MAPKs, and NF-κB). We therefore examined phosphorylation of IκB [an indicator of NF-(κB activation), AKT (a downstream effector of PI3K), and ERK1/2 (a member of MAPKs)] in RTEC in response to C5a stimulation. Following C5a (10 nM) stimulation, the amount of p-ERK and p-IκB was increased and peaked at 5 and 60 min, respectively (**Figures 6A,B**). C5a stimulation had no effect on

p-AKT (Figure S4 in Supplementary Material). We also assessed whether intracellular signaling of ERK1/2 responsible for the action of C5a on Man expression. Man expression in RTEC was significantly decreased by pre-treatment of the cells with specific inhibitor of ERK1/2 (U0126) (**Figure 6C**), indicating that intracellular signaling *per se* or through mediation of cellular inflammatory response could influence the Man expression in human RTEC. Further experiments were performed to examine cellular inflammatory responses to C5a stimulation in RTEC.

All results shown are representative of three independent experiments.

The results showed that C5a stimulation significantly increased gene and protein expression of a set of proinflammatory mediators in the RTEC (Figure S5 in Supplementary Material), among those, the effect of C5a on TNF-α was most prominent (**Figures 6D,E**). Together, these results suggest the possibility that TNF-α driven by intracellular signaling mediated by C5a modulates Man expression in human RTEC. To explore the possibility, we assessed the effect of TNF-α on Man expression in RTEC. TNF-α stimulation significantly increased Man expression (**Figure 6F**) and accordingly bacterial adhesion (**Figures 6G,H**), demonstrating an important role of TNF-α in modulating Man expression and bacterial adhesion in human RTEC. Thus, our data support that C5a-mediated intracellular signaling (e.g., ERK1/2, NF-κB) modulates Man expression in human RTEC through mediating inflammatory mediators (e.g., TNF-α).

#### Clinical Relevance of *In Vitro* Observations

To assess the *in vivo* relevance of our *in vitro* observations, we performed several *ex vivo* experiments using clinical samples. We examined the C5a levels in urine of 8 normal controls and 10 patients with active UTI (predominantly cystitis) by ELISA. C5a was detected in all urine samples including from the healthy

RTEC shown in (C). Data were analyzed by unpaired two-tailed Student's *t*-test [*n* = 4 individual images (×200 magnification) per group]. \**P* < 0.05, \*\*\*\**P* < 0.0001.

or with C5a, in the presence of C5aR1 antagonist (PMX53, 5 µM) or vehicle (Ctrl) for 24 h and subjected to assessment of Man expression and bacteria binding to RTEC. (A,B) Fluorescence microscopy analysis. (A) Representative fluorescence images of Man expression in RTEC (non-permeabilized). Man (green) detected by fluorescein-labeled *Galanthus nivalis* lectin (GNL) and 4',6-diamidino-2-phenylindole (blue) are shown. Scale bars, 25 µm. (B) Quantification of Man expression, shown as relative fluorescence intensity in GNL-positively stained area corresponding to the images shown in (A). (C,D) Fluorescence intensity-based microplate assay. (C) Man expression in RTEC that had been treated with C5a and/or LPS. (D) Man expression in RTEC that had been treated with C5a and PMX53. (E) Bacterial adhesion to RTEC, evaluated by bacterial plate count assay. (B–D) Data were analyzed by one-way ANOVA with Tukey's multiple comparisons [(B) *n* = 6 coverslips per group, under ×100 magnification, (C) *n* = 9 individual wells per group, (D,E) *n* = 6–8 individual wells per group]. All results shown are representative of three independent experiments. \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001, \*\*\*\**P* < 0.0001. Abbreviations: Ctrl, control; RFI, relative fluorescence intensity; RFU, relative fluorescence units.

controls and UTI patients. However, the urinary C5a levels were significantly higher in UTI patients than that in healthy controls (median concentration 228.5 vs. 15,230 pg/mL, controls vs. patients, *P* < 0.0001) (**Figure 7A**). Because all the samples are single spot urine samples, the detected C5a levels may not accurately reflect the C5a concentrations in the urine due to differences in urine flow rate over a day. Therefore, we examined the creatinine concentrations and determined the C5a/creatinine ratios in these urine samples, as urinary creatinine excretion in the presence of a table glomerular filtration rate is fairly constant in a given patient. Creatinine concentrations in the urine samples of UTI patients were compatible to that in healthy controls, but the C5a/creatinine ratios in the urine samples of UTI patients were significantly higher, compared with that in healthy controls (**Figures 7B,C**). Together, these results indicate there is increase in urinary C5a during UTI. In addition to detecting urinary C5a, we examined the expression of C5aR1 in (normal) human kidney biopsies. Immunohistochemistry showed that

microscopic images of bound bacteria in RTEC (non-permeabilized) that had been incubated with labeled J96 for 1 h. Bacteria (red), Man (green) detected by fluorescein-labeled *Galanthus nivalis* lectin and 4',6-diamidino-2-phenylindole (blue) are shown. Left image: compressed image. Scale bars, 10 µm. Right image: corresponding to the boxed region in the left image show the cross-sectional views in z stack (bottom and side panel) of RTEC, Man, and bacteria, demonstrating association of Man and J96 at cell surface of RTEC. (B) Bacteria binding to RTEC, evacuated by bacterial plate count assay. RTEC were incubated with d-mannose or glucose (1 or 5%) for 30 min before the addition of bacteria. Data were analyzed by two-way ANOVA with multiple comparisons (*n* = 8 individual wells per group). (C,E) Man expression in RTCE that had been pre-treated without C5a (C) or with C5a (E) for 24 h and then treated with buffer alone (PBS) or containing αmannosidase (5 mM) or control enzyme (β-galactosidase) for 1 h, evacuated by fluorescence intensity-based microplate assay. Data were analyzed by one-way ANOVA with Tukey's multiple comparisons (*n* = 6 individual wells per group). (D,F) Bacterial binding to RTEC that had undergone the same treatments as described in (C,E), evacuated by bacterial plate count assay. (D) Without C5a pre-treatment. (F) With C5a pre-treatment. Data were analyzed by one-way ANOVA with Tukey's multiple comparisons (*n* = 8 individual wells per group). \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001, \*\*\*\**P* < 0.0001. All results shown are representative of three independent experiments.

C5aR1 was clearly detected in the kidney tissue, mainly in renal tubules (**Figure 7D**). We also examined the Man expression in human kidney tissue and assessed the possibility that UPEC bind to renal tubular epithelium through Man by employing an *in vitro* model of kidney tissue infection. The human kidney was positively stained for GNL and positive signals were mainly found in the tubules of cortex and cortical medullary junction (**Figure 8A**). Fluorescence microscopy analysis showed that binding of UPEC to the tubular epithelium can be detected after 1 h infection (**Figure 8B**), blocking bacterial lectin by pre-incubation of UPEC with d-mannose markedly inhibited the binding of UPEC to tubular epithelium (**Figure 8C**), thus

phosphorylation in renal tubular epithelial cell (RTEC) after C5a (10 nM) stimulation for up to 60 min. In each set of blots, the top row of bands corresponds to incubating membrane with appropriate anti-phospho-antibody and the bottom row of bands corresponds to incubating membrane with appropriate total antibody. Relative amounts of protein phosphorylation are shown in the lower panel of each set of blots. Data were analyzed by one-way ANOVA (*n* = 3/group, resulting from three independent experiments). (C) Effect of inhibition of ERK1/2 pathway on Man expression in RTECs assessed by fluorescence intensity-based microplate assay. RTEC monolayers were pre-incubated with C5a for 24 h in the presence of ERK1/2 pathway inhibitor (U0126, 10 μM) and the vehicle control (DMSO) then were used for quantification of Man expression. Data were analyzed by unpaired two-tailed Student's *t*-test (*n* = 9–12 individual wells per group). A representative result of three experiments is shown. (D,E) Production of TNF-α in RTEC that had been treated with C5a (10 nM) and/or LPS (800 ng/mL) for 24 h and subjected to RT-qPCR (D) and ELISA (E). Data were analyzed by one-way ANOVA with Tukey's multiple comparisons (*n* = 3/group, resulting from three independent experiments). (F) Man expression in RTEC that had been pre-treated with recombinant human TNF-α (10 ng/mL) or vehicle control (BSA) for 24 h, evaluated by fluorescence intensity-based microplate assay. Data were analyzed by unpaired two-tailed Student's *t*-test (*n* = 9 replicate wells/group). A representative result of three experiments is shown. (G,H) Bacterial adhesion to RTEC. (G) Representative microscopic images of bacteria adhesion to RTEC that had been pre-treated with TNF-α for 24 h, then incubated with tri-iodothyronine and tetramethylrhodamine-labeled J96 for 1 h. Bacteria (red), Man (green) detected by fluorescein-labeled *Galanthus nivalis* lectin, and 4',6-diamidino-2-phenylindole (blue) are shown. Scale bars, 25 µm. (H) Quantification of bound bacteria corresponding to the images shown in (G). Data were analyzed by unpaired two-tailed Student's *t*-test [*n* = 6 individual images (×200 magnification) from two coverslips per group]. A representative result of three experiments is shown. \**P* < 0.05, \*\**P* < 0.005, \*\*\**P* < 0.001, \*\*\*\**P* < 0.0001.

supporting the involvement of Man in UPEC adhesion/colonization in renal tubularepithelium.

# DISCUSSION

Although it is well known that C5a/C5aR1 interactions mediate many cellular responses in different types of cells, its role in modulating surface carbohydrate expression of epithelial cells, thereby influencing bacterial adhesion/invasion is much less known. Our recent work in murine models of ascending UTI has suggested that C5aR1 signaling mediates upregulation of Man expression and subsequent bacteria adhesion contributes to the pathogenesis of renal infection (20, 21). In the present study, by using primary cultures of human RTEC and clinical samples (kidney biopsy and urine), we evaluated the relevance of observations on mice to human UTI. Our results provide supporting evidence for that human renal tubular epithelial C5aR1 signaling enhances UPEC adhesion to epithelial cells through upregulation of Man expression and its clinical implications.

As controversial observations have been reported on the expression of C5aR1 in human RTEC (26), we further examined the expression of C5aR1 in human kidney biopsies and primary cultures of RTEC in the present study. Our results of immunohistochemistry and immunocytochemistry are consistent with the reported positive expression of C5aR1 in kidney tissue and isolated tubular epithelial cells by most studies (27–31). Detection of C5aR1 expression in RTEC by RT-PCR (presented in this study) (Figure S2 in Supplementary Material) and in kidney tissues by *in situ* hybridization (reported previously) (31) further support for the expression of C5aR1 in human RTECs.

Consistent with observations in murine RTEC, in the present study we demonstrate that C5a/C5aR1 interactions upregulate Man expression and bacterial adhesion in human RTEC, suggesting the signaling/molecule mechanisms responsible for the action of C5a on Man expression and bacterial adhesion, namely C5a/C5aR1 interactions mediate the activation of ERK1/2/ NF-κB signaling pathway and generation of proinflammatory cytokines, such as TNF-α, which upregulates Man expression and consequent bacterial adhesion. Although the issue of how TNF-α regulates surface Man was not addressed in our study, it has been shown that proinflammatory cytokines such as TNFα can upregulate surface Man expression in endothelial cells through regulating α-mannosidase activity (32). The same would be expected in RTEC driven by C5aR1/ERK1/2/NF-κB signaling pathway.

Binding of bacteria to epithelial cells can mediate bacterial invasion of host cells through internalization by epithelial cells and/or transepithelial migration. Bacterial internalization by

epithelial cells is increasingly recognized as an important feature of UTI. It can enhance bacterial survival by providing protection from host immune defenses and allow pathogens greater access to deeper tissues. Recent studies have suggested that intracellular *E. coli* can form a reservoir within the uroepithelium that may serve as a source for recurrent acute infections, a well-recognized clinical problem (4, 9). While epithelial barrier damage caused bacteria transmigration provides a route by which bacteria can access deeper structures and establish tissue-invasive infection. In the present study, we not only demonstrate the effect of C5a on bacterial adhesion, but also the effects of C5a on bacterial invasion, as evidenced by C5a stimulation caused an increase in bacteria within epithelial cells and transmigrated through epithelial cells. The observed increase of bacteria within epithelial cells is most likely to be the consequence of C5a/C5aR1 signalingmediated upregulation of Man expression and bacterial binding. However, we cannot exclude the possibility of C5a/C5aR1 signaling-mediated enhancement of internalization through upregulating the expression of the receptors which have uptake functions, as we previously have shown that C3b-opsonized UPEC can interact with complement receptor (CD46) on the epithelial cell surface mediating UPEC internalization (8). In the case of bacteria transmigration, C5/C5aR1 signalingmediated enhancement of Man expression and bacterial binding could lead to bacteria transmigration. In addition, C5a/C5aR1 signaling-mediated production of inflammatory mediators could cause epithelial barrier damage, thus facilitating bacteria transmigration.

We recognized the limitations of using primary cultured RTEC in our study. RTEC that we prepared from cortex of kidney biopsies are not pure proximal tubular epithelial cells (which are thought to be a primary target of UPEC), which may contain some other types of tubule cells, such as distal tubule cells, as distal tubules are also located in renal cortex. Therefore, the variations in cell responses among different RTEC preparations might be expected. However, renal cortex presents much more volumes of proximal tubules than distal tubules. In addition, we have tried to overcome the limitations by staining the cultured RTEC (permeabilized) with fluorescein-labeled Lotus Tetragonolobus Lectin (LTL) (a proximal tubular marker), only RTEC preparation that shows >75% of cells were positively stained with LTL were used for experiments.

*In vitro* findings in human RTEC promoted us to test the *in vivo* relevance. In the analyses of human kidney biopsies and urine, we made several observations, including: (i) urinary C5a levels are increased in UTI patients, (ii) C5aR1 is predominantly expressed in renal tubules, (iii) Man is readily detected in renal tubules across cortex and medullar in normal kidney biopsies, predominantly expressed in the luminal surface, (iv) UPEC binding to renal tubules in a mannose-dependent manner, which provide evidence supporting the possibility of C5a/ C5aR1 interactions modulate Man-dependent UPEC adhesion *in vivo*.

In summary, our results demonstrate that human renal tubular epithelial C5aR1 signaling has a role in facilitating UPEC adhesion to and invasion of the epithelial cells through upregulation of Man expression and suggest that upregulation of Man is through C5aR1-mediated ERK1/2/NF-κB activation and release of proinflammatory cytokines, such as TNF-α. Our results from this study, together with our recent findings that C5aR1 has a pathogenic role in ascending UTI in mice (20, 21), suggest further studies to explore the therapeutic potential of targeting C5aR1 in UTI are warranted.

#### REFERENCES


#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the guidelines of the Hospital Research Ethics Committee with written informed consent from all patients. Each patient gave an informed consent about the present study in accordance with the Declaration of Helsinki. The protocol was approved by the Hospital Research Ethics Committee at Xi'an Jiaotong University.

#### AUTHOR CONTRIBUTIONS

KL and WZ conceived and designed the study. YS, K-YW, Z-YD, Y-FG, and L-DZ performed experiments. YS and KL analyzed data. YS, MG, WZ, and KL prepared manuscript. L-DZ and TC provided tissue samples and contributed to interpretation of results.

#### FUNDING

This work was supported by the National Natural Science Foundation of China (NSFC 81170644 and 81770696 to KL), the Natural Science Foundation of Shaanxi Province (2016JZ029 to KL), and the Medical Research Council of the UK (G1001141 and MR/L020254/1 to WZ).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00949/ full#supplementary-material.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Song, Wu, Wu, Duan, Gao, Zhang, Chong, Garstka, Zhou and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Decreased Ficolin-3-mediated Complement Lectin Pathway Activation and Alternative Pathway Amplification During Bacterial Infections in Patients With Type 2 Diabetes Mellitus

#### Edited by:

Maciej Cedzynski, Institute for Medical Biology (PAN), Poland

#### Reviewed by:

Anna Swierzko, Institute for Medical Biology (PAN), Poland Robert Braidwood Sim, University of Oxford, United Kingdom

#### \*Correspondence:

László József Barkai barkai.lj@gmail.com

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 30 July 2018 Accepted: 25 February 2019 Published: 20 March 2019

#### Citation:

Barkai LJ, Sipter E, Csuka D, Prohászka Z, Pilely K, Garred P and Hosszúfalusi N (2019) Decreased Ficolin-3-mediated Complement Lectin Pathway Activation and Alternative Pathway Amplification During Bacterial Infections in Patients With Type 2 Diabetes Mellitus. Front. Immunol. 10:509. doi: 10.3389/fimmu.2019.00509 László József Barkai <sup>1</sup> \* † , Emese Sipter 1†, Dorottya Csuka<sup>1</sup> , Zoltán Prohászka<sup>1</sup> , Katrine Pilely <sup>2</sup> , Peter Garred<sup>2</sup> and Nóra Hosszúfalusi <sup>1</sup>

<sup>1</sup> 3rd Department of Internal Medicine, Semmelweis University, Budapest, Hungary, <sup>2</sup> Laboratory of Molecular Medicine, Department of Clinical Immunology, Section 7631, Rigshospitalet, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark

Bacterial infections are frequent and severe in patients with diabetes mellitus. Whether diabetes per se induces functional alterations in the complement system hampering activation during infection is unknown. We investigated key elements of the complement system during bacterial infections in patients with type 2 diabetes mellitus (T2DM) and compared them to non-diabetic (ND) individuals. Using a prospective design, we included 197 T2DM, and 196 ND subjects, all with clinical diagnosis of acute community-acquired bacterial infections. Functional activities of the ficolin-3-mediated lectin (F3-LP), mannose binding lectin-mediated lectin- (MBL-LP), classical (CP), and alternative pathways (AP), as well as concentrations of complement activation products C4d and sC5b-9 were determined. Functional in vitro activities of F3-LP and AP were significantly higher in T2DM than in ND subjects, (median 64% vs. 45%, p = 0.0354 and 75 vs. 28%, p = 0.0013, respectively), indicating a decreased in vivo activation and lack of consumption of F3-LP and AP in T2DM patients, whereas no difference in functional capacities of CP and MBL-LP were observed between T2DM and ND subjects. Diminished F3-LP and AP activation was most pronounced in diabetic patients with urinary tract infections with positive microbiological culture results for Escherichia coli bacteria. In the T2DM group 3-months mortality significantly associated with diminished F3-LP and AP, but not with CP activation. Concentrations of C4d and sC5b-9 were significantly lower in the T2DM than in ND patients. In conclusion, we found impaired F3-LP activation and lack of AP amplification during bacterial infections in patients with type 2 diabetes, compared to non-diabetic subjects, suggesting a diminished complement mediated protection to bacterial infections in T2DM.

Keywords: ficolin-3, lectin pathway, alternative pathway, classical pathway, complement, type 2 diabetes, bacterial infection, Escherichia coli

# INTRODUCTION

Diabetes is considered as an independent risk factor for community acquired and nosocomial infections (1). It increases the susceptibility to infectious diseases, moreover, prolongs the infection-related hospitalization and may enhance its mortality (2, 3). The background of this acquired immunodeficiency is not completely clear (4, 5). Regarding innate immunity, in vitro studies identified polymorphonuclear neutrophil (PMN) dysfunction including impaired PMN transmigration through barriers (6), reduced PMN chemotaxis (7) and as the most convincing evidence, decreased microbial killing (7–9). Available data are controversial for the adaptive immunity. T-lymphocyte dysfunction seems to be dependent on glycemic control, as T cell proliferation was impaired in poorly controlled patients with type 1 diabetes (10). By contrast, patients with relatively good metabolic control showed a robust secondary immune response to standard antigens (11). With regard to humoral immunity, glycation may impair the biological function of antibodies (12).

Only a few data are known about the activation of the complement system in bacterial infections in diabetes. Nevertheless, complement activation has been shown to be a contributing factor to complications of diabetes (13). C3 as a central component of complement and its activation may contribute to diabetic nephropathy, retinopathy and neuropathy (13–17). Regarding the macrovascular complications, Hess and colleagues showed the possible role of C3 in diabetes related cardiovascular risk, by proposing a mechanism in which C3 participates in a hypofibrinolytic, and thus prothrombotic state (18). In a recent review, Ghosh and colleagues summarized the body of evidence supporting the role of the complement system and complement regulatory proteins in the pathogenesis of diabetic vascular complications, with specific emphasis on the role of the membrane attack complex (MAC) and of CD59, an extracellular cell membrane-anchored inhibitor of MAC formation that is inactivated by non-enzymatic glycation (19).

On the other hand a diminished complement-activating capacity through the classical pathway in type 2 diabetes mellitus was reported in the context of free sialic acid as a potential modulator of complement activation (20). Regarding the effect of high glucose on complement activation, in vitro assays showed that classical and alternative pathway activities were not affected by elevated glucose or other hexoses tested (21). However, high glucose concentrations inhibited the complement activation via the mannose binding lectin (MBL) mediated pathway (21). The role of the complement system in infectious diabetic complications has been studied scarcely.

Ficolins−1,−2,−3 and mannose binding lectin are pattern recognition molecules playing an important role in activating the lectin complement pathway (22–24). MBL binds directly to high mannose or fucose structures on microbial surfaces and drives activation of the lectin pathway (25). Ficolin-1 and ficolin-3 were shown to bind carbohydrate structures of bacteria, especially Nacetyl-galactosamine, and N-acetyl-D-glucosamine, additionally ficolin-3 can associate also with glucose and fucose (26). Ficolin-2 is the major 1,3-β-glucan-binding protein in human plasma and can bind to lipoteichoic acid, thus, ficolin-2 may bind to a wide variety of fungi and Gram-positive bacteria (27, 28). Two patients with congenital ficolin-3 deficiency were reported suffering from serious and life-threatening infections caused by Haemophilus influenzae and Pseduomonas aeruginosa (29), and necrotizing colitis (30). Despite these data, previously, ficolin-3 was identified to bind only a few common pathogenic bacteria (31). Moreover, its potential to activate the lectin pathway in vivo, and to augment phagocytosis, has been described only for the opportunistic bacteria, Hafnia alvei (32) until now. The role of ficolins in infections of patients with diabetes was not studied earlier, despite that MBL, and ficolins recognize specific carbohydrate patterns expressed by microorganisms, and elevated blood glucose and/or protein glycation seen in diabetes may alter such carbohydrate patterns.

Therefore, the aim of our study was to investigate complement activation and consumption via the alternative, classical and lectin pathways during bacterial infections in patients with type 2 diabetes (T2DM). Accordingly, the objective of our study was to provide complex observational data on the lectin pathway focusing on the ficolin-3-mediated pathway in diabetes mellitus in the context of community acquired bacterial infections.

#### MATERIALS AND METHODS

#### Patients

In this prospective, observational study patients with clinical diagnosis of bacterial infections were enrolled and divided into two groups according to the presence [T2DM group, based on WHO criteria (33) or absence of type 2 diabetes mellitus (non-diabetic, ND group)]. Inclusion criteria were a minimum age of 18 years and the need for hospitalization on a general medicine ward due to acute community-acquired bacterial infection. Patients with hematological, oncological or immunological illnesses were excluded. Sepsis was defined in case of at least two existing SIRS (systemic inflammatory response syndrome) criteria from the following: 1. Temperature >38◦C (100.4◦F) or <36◦C (96.8◦F), 2. Heart rate >90/min, 3. Respiratory rate >20/min or PaCO2 <32 mm Hg, 4. WBC >12.000/mm<sup>3</sup> , <4.000/mm<sup>3</sup> , or >10% bands (34, 35). To ascertain post infection mortality rates, subjects were followedup after a 3-months period through their social insurance identification number, and via interview on their-, or their family member's telephone number.

#### Data Collection

Clinical data of current symptoms and past medical history were assembled from the patients' medical records and a further thorough interview after admission to the hospital.

#### Blood Sampling

Samples (serum, EDTA-anticoagulated plasma and sodiumcitrate-anticoagulated plasma) of both groups were collected from antecubital venipuncture within the first 3 days of the hospitalization, between September 2013 and December 2016. Cells and supernatants were separated by centrifugation at 2000x g and the aliquoted serum and plasma samples were stored at −70◦C until analysis.

#### Determination of Complement Parameters

For functional assessment of either the alternative pathway (AP, reference range based on the values of healthy blood donors: 70–125%), or the MBL-mediated lectin pathway activation (MBL-LP, measured by MBL activation with mannan, reference range: 30–130%), as well as the ficolin-3-mediated lectin pathway activation (F3-LP, measured by ficolin-3 activation with acetylated BSA, reference range: 25–130%) in serum samples, commercial ELISA kits (Wieslab, Eurodiagnostica, Malmö, Sweden) detecting generation of terminal complement complex (C9 neo-epitope) were used, according to the manufacturer's instructions. Total classical pathway activity (CP, reference range: CH50 48–103 U/mL) was assessed with a home-made sheeperythrocyte hemolytic titration test, using serum samples. In these standardized assays, the decreased or absent in vitro functional activity of complement pathways may be related to in vivo consumption (36). The concentrations of ficolin-1 (F1, reference range: 10–1,890 ng/mL) (37), ficolin-2 (F2, reference range: 1.00–12.20µg/mL) (38), ficolin-3 (F3, reference range, 3–54µg/mL) (39) and MBL (reference range: 0–5,000 ng/mL) (40) were determined in serum by standard sandwich ELISA techniques: monoclonal antibodies specific for each protein were coated on the plates, and after the incubation of the samples, biotinylated antibodies were applied, and streptavidin/HRP complexes were used for detection.

Complement C3 (reference range: 0.90–1.80 g/L) and C4 (reference range: 0.15–0.55 g/L) levels were measured in serum by turbidimetry (Beckman Coulter, Brea, CA), whereas, radial immunodiffusion was performed to measure the antigenic concentration of C1-inhibitor (reference range: 0.15–0.30 g/L) using polyclonal goat anti-human C1-inhibitor (Quidel, San Diego, CA, USA). Commercial ELISA kits were used to quantify the levels of C4d (reference range: 0.70–6.30µg/mL) and sC5b-9, the soluble form of the terminal pathway activation complex (reference range: 110–252 ng/mL); both complement activation products were measured in EDTA-plasma (Quidel, San Diego, CA, USA).

#### Other Laboratory Measurements

C-reactive protein (hsCRP) levels were ascertained by turbidimetry (Beckman Coulter, Brea, CA), other clinical laboratory parameters were measured with Beckman Coulter (Brea, CA) or the Cell-Dyn 3,500 hematology analyzer. Blood glucose level was determined using the hexokinase assay. Fructosamine was measured with Roche Fructosamine colorimetric test kit (using nitrotetrazolium blue chloride on a Beckman Analyzer AU680, reference range: 205–280 µmol/L). HbA1c concentration was quantified with ion exchange high pressure liquid chromatography (HPLC) method (reference range: 4.0–6.0%). Advanced glycation end product (AGE) levels of both groups were measured in skin by non-invasive autofluorescence technique (AGE Reader mu, DiagnOptics); according to the manufacturer's instructions (41).

TABLE 1 | Clinical characteristics of the T2DM and the ND group.


Mortality rate: within 3 months after infection, BMI, Body mass index; AGEs, Advanced Glycation End products; AU, Arbitrary Unit; CRP, C-reactive protein; WBC, white blood cell count (total).

Sepsis was defined in case of at least two existing SIRS (systemic inflammatory response syndrome) criteria from the followings:

1. Temp >38◦C (100.4◦F) or < 36◦C (96.8◦F), 2. Heart rate > 90/min, 3. Respiratory rate >20/min or PaCO2< 32 mm Hg, 4. WBC > 12,000/mm<sup>3</sup> , < 4,000/mm<sup>3</sup> , or > 10% bands.

\*Values indicate medians [25–75% percentile] of the variables.

For comparison of the two groups Mann-Whitney test or chi-square test was performed. The values in bold indicate the medians and the significant differences.

#### Statistical Analysis

Statistical calculations were carried out using GraphPad Prism 5 (Graphpad Software, USA; www.graphpad.com). Results are presented as medians with 25–75% percentile, or numbers (percentage). To compare variables in two independent groups Mann-Whitney test or chi-square test, to analyze associations between variables Spearman correlation was used. All statistical analyses were two-tailed, significance threshold was set at p = 0.05.

#### RESULTS

#### Clinical Characteristics of the Two Cohorts

A total of 197 T2DM and 196 ND subjects were included; clinical characteristics of the two groups are presented in **Table 1**. Blood glucose, glycation parameters reflecting short-, middle-, and long-term blood glucose levels (fructosamine, HbA1c, AGEs) and BMI were significantly higher (all p < 0.001) in the T2DM group compared to the ND. No differences were found between the two groups regarding age and gender distribution. Infectionrelated laboratory markers (C-reactive protein, white blood


Values indicate medians [25–75% percentile] of the variables, conc.: concentration. Mann-Whitney test was performed to compare study groups.

The values in bold indicate the medians and the significant differences.

cell count) were equally, substantially elevated in both groups (average CRP concentration well-above 100 mg/L, and WBC above 12 G/L), with at least half (54–60%) of subjects with sepsis (**Table 1**).

### Complement Parameters in the Study Groups

In order to get a reasonably detailed profile on the complement system, concentrations of recognition molecules (ficolin-1,−2,−3, and MBL), levels of C1-inhibitor, C3, C4, C4d, sC5b-9, as well as functional activity of the ficolin-3- or MBL-mediated lectin pathway activation (referred to as F3-LP, or MBL-LP, respectively), and of the classical (CP) and alternative (AP) pathways were determined. As shown in **Table 2** patients with bacterial infections but without diabetes (ND) had marked, strong complement activation, as reflected by the very high terminal pathway activation marker sC5b-9 levels. Patients with type 2 diabetes (T2DM) during infection had also elevated but significantly lower sC5b-9 concentrations when compared to ND (p = 0.0022). Regarding the complement activity of various complement pathways, diabetic patients and the ND group had similar classical pathway consumption (**Table 2**). In contrast, in vitro activation of F3-LP and AP were more pronounced in samples of patients with bacterial infections and diabetes in comparison with ND group, as reflected by the difference of F3-LP and AP activity (in average by 19–47% higher values observed by the in vitro activation test in T2DM, p = 0.0354, p = 0.0013, respectively) (**Table 2**). Note, that the higher in vitro complement pathway activation reflects a lower in vivo activation and/or consumption. The diminished in vivo activation of F3-LP and AP in diabetic patients with bacterial infections was not due to lower levels of ficolins or C4 as antigenic concentrations of these recognition molecules and complement components were similar in the two groups (**Table 2**). Concentrations of C3 were only slightly elevated in the T2DM group (p = 0.0482) (**Table 2**). No difference concerning MBL concentration and functional capacity of MBL-LP was observed between the two groups (**Table 2**). Importantly, diminished activation of F3-LP in patients with bacterial infections and diabetes was also supported by the significantly lower C4d concentrations, when compared to ND (p = 0.0063). Similarly, the significantly higher concentration of C1-inhibitor in the T2DM group compared to ND (p < 0.0001) may reflect the impaired activation and consumption of F3-LP (**Table 2**).

To further explore the relationship between consumption of different pathways during bacterial infections in diabetic patients and non-diabetic individuals, analyses of parallel consumptions were done. Consumption of various complement pathways such as CP, AP and F3-LP were defined according to the lower limit of the reference range, as indicated in **Table 2** (<48 U/mL, <70%, <25%, respectively). "No consumption" was considered if the values were equal or above the lower limit of the reference. As shown on **Figure 1**, distribution of complement consumption was polarized in non-diabetic patients: a significant percentage of ND patients fell into subgroups without any consumption (38% for CP and AP, 41% for F3-LP and AP) however, another significant percentage showed a consumption of both pathways (45% for CP and AP, 44% for F3-LP and AP) (**Figures 1A,B**). In contrast, diabetic patients with bacterial infections showed different patterns: 42% fell into subgroups without any consumption for CP and AP and 59% for F3-LP and AP; while only 28% had a consumption of both pathways for CP and AP and 21% for F3-LP and AP. The T2DM group had less frequent parallel consumptions of F3-LP and AP in comparison with the ND group (p = 0.0007) (**Figure 1B**), and a similar finding was observed for CP and AP (p = 0.0079) (**Figure 1A**). In addition, when analyzing all three pathways together, 48% of ND patients, but only 27% of diabetic patients had parallel activation and consumption of all three pathways F3-LP, CP and AP (p = 0.037) (**Figure 1C**). This difference is mainly attributable to the diminished activation of F3-LP and AP in diabetes (**Figure 1**).

#### Activation and Consumption of F3-LP and AP Among Different Types of Infections

To obtain a more detailed view on the diminished F3-LP and AP activation in diabetic patients with bacterial infections, subjects were allocated into subgroups, according to the

anatomic location of infections: respiratory tract (Resp.), urinary tract (UTI), skin and soft tissue (SSTI), and infections in other locations.

The diminished in vivo activation (lack of consumption) of F3-LP and AP among diabetic patients was most pronounced in case of UTI (**Tables 3**, **4**). Non-diabetic subjects had in average 30% lower in vitro functional activity of F3- LP than patients with diabetes (37% [8–79] vs. 67% [21– 102], p = 0.0456). Slight difference alike was observed in patients with sepsis and UTI (27% [6–73] vs. 67% [20– 102], p = 0.038, **Table 3**). Note, that the observed lower in vitro complement pathway activity reflects a higher in vivo activation and/or consumption, and the higher in vitro complement pathway activation means a lower in vivo activation and/or consumption, as previously described. Similar results were seen in the subgroups with respiratory infection, though the difference between the groups did not reach statistical significance (**Table 3**). Regarding SSTIs, SSTIs with sepsis, and other localization of infections, no significant differences in F3- LP activity were seen between the groups (**Table 3**). Similarly, to the F3-LP, in vivo activation of AP was diminished among diabetic patients having respiratory or urinary tract infections (p = 0.0276 and p = 0.0092, respectively, **Table 4**). These two locations represented the majority of patients in both study groups.

Based on the pronounced difference in F3-LP and AP activity, we aimed to further analyze the subgroup with UTI. As shown on **Figure 2**, diabetic patients with positive culture results for Escherichia coli (E. coli) had diminished F3-LP activation when compared to those of non-diabetic subjects (70% [30–103] vs. 33% [16–91], p = 0.0286). Similar results were observed regarding AP (87% [77–99] vs. 6% [0–53], p = 0.0003). However, no difference concerning F3-LP and AP was observed for those, with non-E. coli UTIs. (**Figure 2**).

#### Association of Complement Activation and Consumption With Clinical Parameters

To determine the relationship of complement activation and consumption with clinical parameters including glycationrelated markers in T2DM, Spearman correlation analysis was done. No associations of F3-LP or AP with blood glucose, fructosamine, or HbA1c were found, however, an inverse weak correlation of F3-LP with the long-term glycation marker of AGEs (p < 0.05, r = −0.2765) was observed among T2DM subjects.



UTI, Urinary tract; SSTI, Skin and soft tissue infections. Due to low case number, sepsis subgroups for "Other" infections are not presented.

"n" indicates number of subjects. Values indicate medians [25–75% percentile] of F3-LP

(%). Mann-Whitney test was performed to compare study groups.

The values in bold indicate the medians and the significant differences.

Three-months mortality after bacterial infection was similar in the diabetic and non-diabetic groups (**Table 1**), however, lack of F3-LP activation and consumption with lack of AP amplification were associated with mortality in the diabetic group (p = 0.012 and p = 0.025, respectively) (**Figure 3**). Whereas, activation of CP, F3-LP and AP was present in 77, 62, and 76% of ND subjects who died, respectively, the same was not observed for T2DM patients. Although 60% of those diabetic patients who died had CP activation and consumption, this proportion was only 29 and 25% for F3-LP and AP, respectively (**Figure 3**). There was no difference in the occurrence of complement consumption among the groups with sepsis only (**Figure 3**).

#### DISCUSSION

In our prospective study, we found decreased in vivo activation of ficolin-3-mediated lectin and alternative pathways during bacterial infections in patients with type 2 diabetes in comparison with non-diabetic subjects. On the other hand, there was no difference between the two study groups regarding the classical pathway and the MBL-mediated lectin pathway activities. The functional assays we used in case of F3-LP, MBL-LP, CP, and AP detect the residual, inducible terminal complement complex generation in the given sample, as these tests are based on the in vitro activation of the samples using the specific activators of the F3-LP (acetylated bovine serum albumin), the MBL-LP (mannan), the CP (immune complexes) or the AP (lipopolysaccharide). Therefore, the observed higher in vitro activation capacity for F3-LP and AP activity found in the T2DM cohort indicate that these complement pathways have not been activated extensively in vivo. In accordance with the diminished in vivo activation in the T2DM patients, less increased sC5b-9 activation product concentrations were found TABLE 4 | Activity of alternative pathway (AP) in relation to anatomic location of bacterial infections.


UTI, Urinary tract; SSTI, Skin and soft tissue infections. Due to low case number, sepsis subgroups for "Other" infections are not presented.

"n" indicates number of subjects. Values indicate medians [25–75% percentile] of AP (%). Mann-Whitney test was performed to compare study groups.

The values in bold indicate the medians and the significant differences.

in T2DM in comparison with the ND group. In addition, the observed higher C1-inhibitor concentrations along with the decreased C4d levels support a diminished in vivo F3- LP activation.

When we analyzed the F3-LP, AP, and CP pathways together, only 27% of diabetic patients had parallel activation in contrast to 48% of ND cases. This difference was mainly attributable to the diminished activation of F3-LP and AP in T2DM. However, one limitation of our study is that we do not have "baseline" (noninfectious) data regarding complement profile of the study subjects.

Regarding the different infection localizations, impaired F3- LP and AP activations were found in T2DM with urinary tract infections in comparison with the ND group. In diabetic patients having respiratory tract infections diminished activation of AP was seen, with a similar tendency regarding F3-LP activity. However, no changes were found in case of skin or soft tissue, or other types of infection locations between the two groups. Whether the observed difference in the altered activation of F3- LP and AP among the different localizations is explained by the different body compartments and/or the diversity of the etiological bacterial agents, still remains to be determined.

Recently, growing number of data have been published about the complement system and diabetes mellitus however, most of these studies have targeted chronic diabetes complications. Based on these observations activation of the complement system plays a role in diabetic vascular complications. C3 as a central component of complement system and its activation may contribute to chronic diabetic micro- and macrovascular complications (14–18). In a recent review, Ghosh and colleagues showed that increased complement activation may have an impact on the pathogenesis of diabetic vascular complications

FIGURE 2 | F3-LP and AP activity in subjects with UTI and a positive microbiological culture. Ficolin-3-mediated lectin (A) and alternative (B) pathway activations (in vitro) of patients with UTIs and positive urine and/or blood cultures were chosen from both cohorts. Depending on the presence or absence of the E. coli bacterium they were each once more divided into two subgroups. T2DM subjects with E. coli positive cultures had higher in vitro F3-LP levels than those of the ND cohort (70% [30–103] vs. 33% [16–91], p = 0.0286). Similar results were found regarding AP (87% [77–99] vs. 6% [0–53], p = 0.0003). Difference between the groups was analyzed with the Mann-Whitney test, \*p < 0.05, and \*\*\*p < 0.001. Horizontal dashed lines indicate limits of reference ranges, horizontal red lines indicate median values of the variables.

(19). Remarkably, higher sC5b-9 values and renal deposition in patients with diabetes were observed and were shown to be related with diabetic nephropathy (42). On the other hand, diminished complement-activating capacity through the classical pathway in type 2 diabetes mellitus was reported in the context of free sialic acid as a potential modulator of complement activation (20). It has to be noted, that in these studies subjects were involved in an infection-free period. Limited data were reported about the complement system during infections in hyperglycemic conditions: a previous study demonstrated that sera taken from healthy donors in hyperglycemic conditions in vitro (glucose concentration: 10–17 mmol/L) altered the interaction of C3 and the pathogenic bacteria Staphylococcus aureus (43). To the best of our knowledge our prospective study is the first publication about a complex investigation of complement activation during bacterial infections in diabetes compared to non-diabetic status. Based on our detailed measurements, diminished activation of F3-LP and AP in diabetic patients with bacterial infections was not due to lower levels of ficolins, C4 or C3, as antigenic concentrations of these complement components were similar in the two groups, or even higher in case of C3 in the T2DM group, in accordance with previous data (42, 44). Average concentrations of components of the lectin pathway (LP) were described previously (45–47) and several studies reported their levels during ongoing infections, but not in T2DM (48–50). Furthermore, it has to be highlighted, that concentrations of these components may vary age-dependently (47), and different comorbidities may have influence on them (51, 52), therefore, it remains a subject of research, how different pathogens or severity of infections could affect their levels.

High glucose related alterations in glycation pattern of complement components, or changes in tertiary structure of complement components may represent a potential explanation for the diminished F3-LP activation and AP amplification during infections of diabetic patients. C3 mediated effector functions were found to be inhibited in hyperglycemic conditions and the elegant study of Hair et al. showed that the inhibition is related to glucose induced changes in the tertiary structure of C3 (43). In our study blood glucose levels, short- (fructosamine), middle- (HbA1c) or long- (AGE) term glycation markers per se were elevated in the T2DM group compared to ND group. However, none of them were associated with activation and consumption of F3-LP and AP, except an inverse correlation of the F3-LP activity with the long-term glycation marker of AGEs in diabetic subjects. One possible explanation for the missing association with the short- and middle-term glycemic parameters could be that despite their higher levels in the T2DM group, our diabetic patients had a relatively good glycemic control around the time of the infection, as shown by the acceptable levels of blood glucose, fructosamine, and HbA1c.

We also evaluated associations of complement activation with clinical parameters and observed that 3-months post infection mortalities were similar in both groups. However, lack of ficolin-3-mediated lectin pathway activation along with lack of alternative pathway amplification were associated with mortality only in the T2DM group. There was no difference in the occurrence of complement consumption among the groups with sepsis without fatal outcome.

Throughout the various types of bacterial infections, the most remarkable difference between the two groups, regarding F3-LP and AP activation, was found in patients with UTI, especially in those suffering from E. coli infection. Ficolin-3 is one of the most effective activator of the LP in vitro (53) although our knowledge of ficolin-3 interaction with human pathogens is limited. Based on previous studies some human pathogens, such as Hafnia alvei (32) can activate the ficolin-3 mediated lectin pathway and ficolin-3 is linked with growth inhibition of Aerococcus viridans (54). Additionally, ficolin-3 recognized pathogenic Pasteurella pneumotropica and two pathogenic E. coli (enteroaggregative E. coli O71 and enteropathogenic E. coli O111 ab:H2) in an interaction study (31), but Sorensen et al. did not find any binding capacity of four prototypic enteroaggregative E. coli strains to recombinant ficolin-3 (55). Complement evasion strategies of E. coli, and altered factor H could also contribute to the diminished F3-LP activation and AP amplification in T2DM patients having E. coli related UTIs.

# CONCLUSIONS

In summary, we found impaired ficolin-3-mediated lectin pathway activation and decreased alternative pathway

# REFERENCES

1. Muller LMAJ, Gorter KJ, Hak E, Goudzwaard WL, Schellevis FG, Hoepelman AIM, et al. Increased risk of common infections in patients with type 1 and type 2 diabetes mellitus. Clin Infect Dis. (2005) 41:281–8. doi: 10.1086/431587 amplification during bacterial infections in patients with T2DM in comparison with non-diabetic individuals. Lack of ficolin-3-mediated lectin pathway and alternative pathway consumption in the T2DM group were associated with 3-months mortality. Less diabetic patients had parallel consumption of F3-LP, CP and AP compared to the non-diabetic subjects. Our data suggest that ficolin-3-mediated lectin and alternative pathway activations may have an impact on the clinical outcome of patients with type 2 diabetes having bacterial infections.

# DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

# ETHICS STATEMENT

Study protocol was approved by the Hungarian Medical Research Council (TUKEB 396/2013 - 31584/2013/EKU) and the Institutional Review Board of the Semmelweis University, Budapest. Patients were involved into the study after informed, written consent in accordance with the Declaration of Helsinki.

# AUTHOR CONTRIBUTIONS

ZP, ES, and NH designed and supervised the study. LB, ES, NH, and DC collected the blood samples. LB, KP, and DC performed ELISA, functional and hemolysis assays. LB, ES, and NH collected clinical and laboratory data. LB and ZP performed the statistical analysis. LB, ES, NH, ZP and DC wrote the manuscript with the help of KP and PG, who revised it critically for important intellectual content. All authors revised and approved the manuscript.

# FUNDING

This study was funded by EFSD New Horizons Program, Dr. Koranyi Andras Foundation, The Svend Andersen Research Foundation, The Danish Research Foundation of Independent Research (DFF-6110-00489), The Danish Heart Foundation (16- R107-A6650-22966), and the Higher Education Institutional Excellence Programme of the Ministry of Human Capacities in Hungary, within the framework of the molecular biology thematic programme of the Semmelweis University.

# ACKNOWLEDGMENTS

The authors wish to express their gratitude toward Mr. Jesper Andresen for excellent technical assistance.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer AS and handling Editor declared their shared affiliation.

Copyright © 2019 Barkai, Sipter, Csuka, Prohászka, Pilely, Garred and Hosszúfalusi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The lectin complement Pathway is involved in Protection against enteroaggregative *Escherichia coli* infection

*Camilla Adler Sørensen1,2, Anne Rosbjerg1 , Betina Hebbelstrup Jensen2 , Karen Angeliki Krogfelt <sup>2</sup> and Peter Garred1 \**

*<sup>1</sup> Laboratory of Molecular Medicine, Department of Clinical Immunology, Section 7631, Rigshospitalet, University Hospital of Copenhagen, Copenhagen, Denmark, 2Department of Bacteria, Parasites and Fungi, Statens Serum Institut, Copenhagen, Denmark*

#### *Edited by:*

*Maciej Cedzynski, Institute for Medical Biology (PAN), Poland*

#### *Reviewed by:*

*Cordula M. Stover, University of Leicester, United Kingdom Peter F. Zipfel, Leibniz-Institut für Naturstoff-Forschung und Infektionsbiologie, Hans Knöll Institut, Germany*

> *\*Correspondence: Peter Garred peter.garred@regionh.dk*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 26 February 2018 Accepted: 08 May 2018 Published: 29 May 2018*

#### *Citation:*

*Adler Sørensen C, Rosbjerg A, Hebbelstrup Jensen B, Krogfelt KA and Garred P (2018) The Lectin Complement Pathway Is Involved in Protection Against Enteroaggregative Escherichia coli Infection. Front. Immunol. 9:1153. doi: 10.3389/fimmu.2018.01153*

Enteroaggregative *Escherichia coli* (EAEC) causes acute and persistent diarrhea worldwide. Still, the involvement of host factors in EAEC infections is unresolved. Binding of recognition molecules from the lectin pathway of complement to EAEC strains have been observed, but the importance is not known. Our aim was to uncover the involvement of these molecules in innate complement dependent immune protection toward EAEC. Binding of mannose-binding lectin, ficolin-1, -2, and -3 to four prototypic EAEC strains, and ficolin-2 binding to 56 clinical EAEC isolates were screened by a consumption-based ELISA method. Flow cytometry was used to determine deposition of C4b, C3b, and the bactericidal C5b-9 membrane attack complex (MAC) on the bacteria in combination with different complement inhibitors. In addition, the direct serum bactericidal effect was assessed. Screening of the prototypic EAEC strains revealed that ficolin-2 was the major binder among the lectin pathway recognition molecules. However, among the clinical EAEC isolates only a restricted number (*n* = 5) of the isolates bound ficolin-2. Using the ficolin-2 binding isolate C322-17 as a model, we found that incubation with normal human serum led to deposition of C4b, C3b, and to MAC formation. No inhibition of complement deposition was observed when a C1q inhibitor was added, while partial inhibition was observed when ficolin-2 or factor D inhibitors were used separately. Combining the inhibitors against ficolin-2 and factor D led to virtually complete inhibition of complement deposition and protection against direct bacterial killing. These results demonstrate that ficolin-2 may play an important role in innate immune protection against EAEC when an appropriate ligand is exposed, but many EAEC strains evade lectin pathway recognition and may, therefore, circumvent this strategy of innate host immune protection.

Keywords: enteroaggregative *Escherichia coli,* complement, lectin pathway, ficolin-2, serum resistance

# INTRODUCTION

Enteroaggregative *Escherichia coli* (EAEC) belongs to the group of diarrheagenic *E. coli* and is an increasingly recognized important cause of diarrhea. EAEC is known to cause watery and often persistent diarrhea in adults as well as children in both industrialized and developing countries. Though several virulence factors are reported, great heterogeneity among EAEC strains has made their molecular epidemiology unclear (1–3).

Enteroaggregative *Escherichia coli* infection is initiated by colonization of the small and large bowel mucosal surfaces by aggregative adherence. This is followed by biofilm formation, induction of an inflammatory response, and release of toxins (1). The precise mechanisms of pathogenesis are still not fully understood, but a combination of several factors such as adhesins and toxins are described to contribute to disease (4, 5). However, none of these factors are conserved in all EAEC strains and a number of similar factors are found in other *E. coli* pathotypes, suggesting that EAEC pathogenesis does not depend on one particular protein, but is probably based on a combination of several virulence factors (2, 4).

Enteroaggregative *Escherichia coli* strains can be recovered from stool samples of apparently healthy individuals and despite studies finding strains associated with diarrhea, some studies have failed to show significant association between EAEC and disease (6–8). This suggests that host factors are involved in manifestations of gastrointestinal disease and further investigations could be crucial for the understanding of EAEC pathogenesis.

The complement system is a complex surveillance system involved in innate immune protection against pathogens. It facilitates opsonophagocytosis of pathogens, induces inflammatory responses, and can lead to bacterial lysis upon activation. Activation can occur *via* three pathways: the lectin, the classical, and the alternative pathway. The complement system is primarily regarded to be of importance for systemic immune protection. But, also local production of complement components is recognized as being important as exudation of complement from the circulation during inflammation appears to be important for local innate immune protection (9).

In the lectin pathway, mannose-binding lectin (MBL) and ficolin-1, -2 and -3 are pattern-recognition molecules (PRMs) involved in initiation of complement activation (10). Recently, two other molecules collectin-10 (CL-10 or CL-L1) and collectin-11 (CL-11 or CL-K1) have to some degree been shown to mediate complement activation (11, 12). They interact with pathogen-associated molecular patterns on the surface of microbial pathogens and upon recognition activate the lectin pathway with help from lectin pathway-associated serine proteases termed MASPs (13). The MASPs cleave C4 and C2 leading to the formation of the C3 convertase (C4b2a). The C3 convertase cleaves C3 into anaphylatoxin C3a and the strong opsonizing factor C3b. Activation through the classical pathway depends on antibody–antigen recognition, which then binds the PRM C1q and leads to cleavage of C4 and C2 by associated proteases C1r/C1s and to deposition of C3b. The alternative pathway is activated spontaneously by hydrolysis of C3, this allows binding of the factor B, which is then cleaved by factor D, forming the C3 convertase of the alternative pathway (C3bBb). The alternative pathway works like an amplification loop for C3b formation and as C3b level rises the C5 convertase is formed (C4b2aC3b/ C3bBb3b) initiating formation of the terminal lytic C5b-9 membrane attack complex (MAC) (14).

The involvement of complement in EAEC pathogenesis is unresolved, and though it has previously been shown that ficolin-2 was able to recognize EAEC (15) the importance of the lectin pathway is yet unknown. Thus, we hypothesized that the lectin pathway molecules MBL, ficolin-1, -2, and -3 could be involved in recognition and thus complement dependent protection of EAEC bacteria.

# MATERIALS AND METHODS

#### Bacterial Strains

Four prototype EAEC strains, producing aggregative adherence fimbriae (AAF) I–IV, were investigated for binding of lectin pathway recognition molecules MBL, ficolin-1, ficolin-2, and ficolin-3. The strains have been described previously (16). In addition, 56 EAEC strains isolated from stool samples of Danish adults suffering from diarrhea, at the diagnostic laboratory at Statens Serum Institut, were randomly selected. Stock cultures were frozen at −80°C in Luria-Bertani broth (LB, Sigma-Aldrich) containing 10% (vol/vol) glycerol. Bacteria were cultivated in Dulbecco's modified eagle medium containing 4.5 g/l d-Glucose (DMEM-HG, Gibco™) overnight with shaking at 37°C until reaching an optical density (OD600 nm) of 1.8, corresponding to a bacterial concentration of approximately 5 × 108 cells/ml.

#### Proteins

Expression and purification of recombinant proteins was performed as previously described (17). Briefly, MBL and ficolin-1, -2, and -3 were expressed in CHO-DG44 cells cultivated in RPMI 1640 medium (Sigma-Aldrich) supplemented with 10% fetal calf serum (FCS), 100 U/ml penicillin, 0.1 mg/ml streptomycin, 2 mM l-glutamine, and 200 nM methotrexate.

# Consumption Assay

We screened four prototypic EAEC strains for binding of recombinant proteins, and 56 clinical EAEC isolates for binding of serum ficolin-2. The overnight bacterial cultures were centrifuged at 5,000 × *g* for 5 min and washed three times in phosphate-buffered saline (PBS, 137 mM NaCl, 2.7 mM KCl, 2 mM KH2PO4, and 8.1 mM Na2HPO4, pH 7.4). The cell pellet was resuspended in Barbital-Tween buffer (Barb-T, 5 mM barbital sodium, 145 mM NaCl, 2 mM CaCl2, 1 mM MgCl2, 0.05% Tween 20, pH 7.5) [barbital buffer previously used by Rosbjerg et al. (18); Hummelshøj et al. (19)] and incubated with 10% normal human serum (NHS) pool originating from four healthy donors, for 1 h at 4°C, or in-house recombinant proteins for 2 h at 37°C, end-over-end. Recombinant proteins were used in the following concentrations: recombinant ficolin-1 (rficolin-1) 2 µg/ml, rficolin-2 0.5 µg/ml, rficolin-3 0.5 µg/ml, and recombinant MBL (rMBL) 0.5 µg/ml. After centrifugation (5,000 × *g*, 5 min) the supernatant was transferred to quantification assays (described below) where the level of consumption was evaluated by comparing the amount of remaining protein in the supernatant with a control sample containing no bacteria. Serum ficolin-2 screenings were performed using *N*-Acetyl-d-glucosamine-Agarose (GlcNAc) beads (Sigma-Aldrich) as a positive binding control matrix.

# ELISA—Determination of Unbound Protein Fraction

Maxisorp polystyrene microtiter plates (Thermo Scientific) were coated with 5 µg/ml acetylated bovine serum albumin or 10 µg/ml mannan in PBS, overnight at 4°C. Plates were washed and blocked in Barb-T before adding the supernatants (from the consumption assay) in serial dilutions and incubating overnight at 4°C. Plates were washed in Barb-T, and detection was performed using the following primary monoclonal antibodies (mAb) in a concentration of 2 µg/ml at 20°C: anti-ficolin-1 mAb (HP9039, Hycult biotech), anti-ficolin-2 mAb clone FCN219 (20), anti-ficolin-3 clone FCN334 (21), and HYB-131-11 (Bioporto Diagnostics) for MBL detection. Plates were incubated 2 h at room temperature, shaking. Plates were washed and HRP-conjugated rabbit anti-mouse polyclonal antibody (P0260, Dako) (1:1,500) was added for 45 min at room temperature, shaking. Plates were thoroughly washed with Barb-T and subsequently developed for 20 min with tetramethylbenzidine One (TMB ONE, Kem-En-Tec Diagnostics). The reaction was stopped with 0.2 M sulfuric acid (H2SO4) and OD was measured at 450 nm.

#### Western Blot—Detection of Bound Proteins

The bacterial cell pellets (from the consumption assay) were washed thoroughly in Barb-T and analyzed by western blotting. Bacteria were lysed with LDS sample buffer (Invitrogen) and the content was run on a 4–12% bis-Tris polyacrylamide gel (Invitrogen). Rficolin-2 (0.25 µg) was used as a loading control. The separated proteins were blotted onto polyvinylidene difluoride membranes (GE Healthcare) and the membranes were probed with 0.5 µg/ml anti-ficolin-1 mAb FCN106 (cross reacting with ficolin-2) overnight at 4°C (22). After washing, the membranes were incubated with rabbit anti-mouse-HRP (1:10,000) (P0260, Dako) for 1 h at room temperature, shaking. Membranes were developed using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific).

# Flow Cytometry—Detection of Complement Activation

Prior to assessing binding and activation with flow cytometry, strain C322-17 was fixed in formalin. An overnight culture of C322-17 was washed in PBS and the concentration was determined by a colony-forming unit (CFU) count. Bacteria were fixed in 4% formalin for 40 min, washed in PBS, and resuspended in 50% ethanol. The stock was kept at −80°C until use.

10% NHS diluted in Barbital buffer containing 1% heatinactivated FCS (Barb-FCS) were combined with the following inhibitors in a concentration of 5 µg/ml: ficolin-2 inhibitory mAb FCN212 (18), anti-factor D inhibitory mAb clone 31A9 (anti-FD, a kind gift from Genentech), C1q inhibitory mAb clone 85 (C1q-85, Sanquin), and mock inhibitor mouse IgG1 (BD Biosciences). Besides, ethylene diamine tetraacetic acid (EDTA) 10 mM (324503, Calbiochem, Merck) was applied. 5 × 107 fixed cells/ml were added for 30 min at 37°C. Bacteria were washed in Barb-FCS and spun down at 5,000 × *g* for 5 min at 4°C. Deposition of C4b, C3b, and C5b-9 were measured using biotinylated anti-C4c pAb (A0065, Dako), biotinylated anti-C3c pAb (A0062, Dako) and biotinylated anti-MAC mAb clone aE11 (23). Anti-ficolin-2 mAb FCN219 (20) was used to detect ficolin-2. Primary antibodies were applied in a concentration of 5 µg/ml for 30 min at 4°C. The specificity of the primary antibodies were verified using a biotinylated rabbit IgG (Rb IgG, 10500C, Invitrogen) or biotinylated mouse IgG2 (mIgG2a, 553455, BD Pharmingen) isotype control. After washing, FITC-conjugated streptavidin (Strep-FITC, S3762, Sigma-Aldrich) were added for 30 min at 4°C. The samples were washed and the levels of bacterial-bound C3, C4, and MAC were measured by flow cytometry using Gallios Flow Cytometer (Beckman-Coulter) and data were analyzed using Kaluza 1.2 software (Beckman-Coulter).

#### Microscopy

Ficolin-2 binding in the presence and absence of the ficolin-2 inhibitor FCN212 was assessed by microscopy. The residual EAEC cells from flow cytometry was placed on slides by cytospin (centrifugation for 5 min at 300 × *g*) and mounted with ProLong Diamond Antifade Mountant (P36965, Life Technologies). Microscopy was performed using a Zeiss Axio Observer through a X63/1.40 oil DIC Plan-Apochromat objective. Imaging conditions were kept constant when acquiring images to be compared.

#### EAEC Serum Resistance

An overnight bacterial culture of C322-17 was diluted to an OD = 0.5 at 600 nm in PBS. 10% NHS were incubated in buffer containing the following inhibitors: FCN212 (5 µg/ml), anti-FD (5 µg/ml), C1q-85 (5 µg/ml) or mock inhibitor mAb Ciona (a mAb raised against an MBL homolog in *Ciona intestinalis*). In addition, we used a specific peptide inhibitor of C3 activation, compstatin (CP40, 6 µM, a kind gift from professor John Lambris, Philadelphia, PA, USA), and C5 inhibitor eculizumab (50 µg/ml, Soliris, Alexion Pharmaceuticals). Heat-inactivated NHS (IHS) (56°C, 30 min) was included as a negative control of complement activation. The baseline consisted of bacteria incubated only with buffer to assess the number of viable cells in the initial inoculum.

All tubes were preincubated at 37°C for 30 min to let the inhibitors work and to start the test approximately at human blood temperature. Then, 20 µl of the inoculum was added to each tube and the sample was homogenized. The samples were incubated at 37°C and 20 µl was spotted in a serial dilution of 10<sup>−</sup><sup>1</sup> to 10<sup>−</sup><sup>6</sup> on LB agar plates and incubated at 37°C overnight before CFU was assessed. The control sample with viable cells of the initial inoculum was diluted and plated at 0 min incubation.

#### Statistics

Statistical analyses were performed using GraphPad Prism 7 (GraphPad Software, USA). The results represent the means ± SD of three independent experiments. For comparisons, we used an unpaired Student's *t*-test (*p-*values: ns *p* > 0.05; \**p* ≤ 0.05; \*\**p* ≤ 0.01; \*\*\**p* ≤ 0.001; \*\*\*\**p* ≤ 0.0001).

# RESULTS

# Binding of rMBL and Ficolins to Prototypic EAEC Strains

Four prototypic EAEC strains were screened in a consumption assay for binding of rMBL, rficolin-1, -2, and -3. Neither rficolin-1 nor rficolin-3 appeared to bind to any of the strains (**Figures 1B,D**), whereas rMBL showed binding to two of the strains (JM221 and 042) (**Figure 1A**). The strongest binding, however, was observed for rficolin-2, which displayed high binding to two strains (55989 and C1010-00) and to a lesser degree to strain 042 (**Figure 1C**). We confirmed the binding by performing a Western blot on the eluates from the consumption assay (data not shown).

We found that rficolin-2 displayed highest binding to the four prototypic EAEC strains and, therefore, decided to examine ficolin-2 binding to 56 clinical EAEC isolates.

#### Binding of Serum Ficolin-2 to 56 Clinical EAEC Isolates

56 clinical EAEC isolates obtained from adult patients suffering from EAEC-related diarrhea were screened for binding of serum ficolin-2 and the unbound fraction of serum ficolin-2 was measured. Five (8.9%) of the 56 isolates appeared to bind serum ficolin-2 and especially one isolate, C322-17, showed very strong binding (**Figure 2**). We furthermore verified that serum ficolin-2 was bound to isolate C322-17 by performing a Western blot on the bacterial pellet for C322-17 and two isolates that were negative for ficolin-2 binding according to the ELISA. The Western blot showed strong monomeric and oligomeric ficolin-2 structures to be associated with C322-17. The two isolates, E3-1065539 and H57553, which were negative for ficolin-2 binding in the ELISA displayed low levels of oligomeric ficolin-2 binding, but no monomeric bands were observed (**Figure 3**).

Since isolate C322-17 displayed high binding of ficolin-2, we decided to use this isolate as a model for further investigation of EAEC-associated complement binding and activation.

#### Calcium Dependency of Ficolin-2 Binding

We examined whether ficolin-2 binding to isolate C322-17 was calcium dependent by pre-incubating NHS with 10 mM EDTA. **Figure 4** shows the mean fluorescence intensities (MFI) when detecting ficolin-2 binding in flow cytometry. Ficolin-2 binding was not reduced by EDTA providing evidence that that the binding was calcium independent. In fact, it seemed that EDTA

lectin (MBL), ficolin-1, -2, or -3.

enhanced binding of ficolin-2, perhaps due to elimination of other calcium-dependent binding partners.

#### Complement Activation by Ficolin-2 Through the Lectin Pathway

The contribution of ficolin-2 to lectin pathway complement activation was assessed by introducing FCN212, a ficolin-2 inhibitory mAb. **Figures 5A,B** shows inhibition of ficolin-2 binding to the bacterial strain C322-17 in flow cytometry and fluorescence microscopy, respectively, when applying the inhibitor. Both techniques showed that the binding was reduced by the employed ficolin-2 inhibitor. In flow cytometry, the inhibition was significantly reduced both when comparing to 10% NHS and to a mock ficolin-2 inhibitor (*p* < 0.0001).

The effect of ficolin-2 binding on complement activation was examined by looking at the deposition of C4b, C3b, and

MAC in the presence and absence of the ficolin-2 inhibitor. C4b deposition was significantly reduced by the inhibitor when comparing to both 10% NHS and the mock inhibitor (*p* = 0.0008 and 0.0026, respectively). The same tendencies were observed for MAC deposition when comparing to 10% NHS and the mock inhibitor (*p* = 0.0189 and 0.0134, respectively), suggesting an involvement of ficolin-2 and the lectin pathway in complement activation (**Figures 6A,C**). We observed a small significant reduction in C3b deposition when comparing to 10% NHS (*p* = 0.0495), but were unable to detect a significant reduction when comparing to the mock inhibitor (*p* = 0.0964), but overall the tendency followed that of C4b and MAC (**Figure 6B**). We also looked at MBL binding to see if this was involved in activation *via* the lectin pathway, but MBL did not bind to C322-17 and thus does not contribute to activation (data not shown).

#### The Classical Pathway Does Not Show Involvement in Complement Activation

Since ficolin-2 only seemed to be partially involved in complement activation, we investigated whether the remaining activity was initiated *via* the classical pathway. When introducing an inhibitor of C1q, we saw no significant reduction in the deposition of C4b, C3b, or MAC (*p* > 0.05) (**Figures 7A–C**, respectively) and, therefore, the classical pathway does not appear to be involved in the remaining complement activity.

#### The Alternative Pathway Contributes to Complement Activation

Next, we tested the involvement of the alternative pathway by applying an inhibitory antibody against factor D before detecting C3b and MAC deposition (**Figure 8**). We were unable to detect a significant reduction in C3b levels when introducing the factor D inhibitor alone. However, when combining the factor D inhibitor with the ficolin-2 inhibitor, we observed a significant reduction in C3b deposition when comparing both to 10% NHS and to a sample where the ficolin-2 inhibitor was exchanged with a mock inhibitor (*p* = 0.0057 and 0.0002, respectively) (**Figure 8A**). The deposition of MAC was highly reduced by the factor D inhibitor alone (10% NHS *p* = 0.0058 and mock inhibitor *p* < 0.0001) and adding the ficolin-2 inhibitor led to further reductions when comparing to the mock inhibitor (*p* = 0.0169) (**Figure 8B**).

These data suggest that both the lectin pathway and the alternative pathway are important in activation of complement on EAEC strain C322-17 and emphasize the importance of the alternative pathway in generation of MAC.

#### Ficolin-2 and Factor D Are Involved in Serum-Mediated EAEC Killing

To further investigate the involvement of ficolin-2 and the alternative pathway on bacterial clearance, we performed a serum resistance assay where bacteria were mixed with 10% NHS and the inhibitors of ficolin-2, factor D and C1q, as well as the C3-targeted complement inhibitor, compstatin, and the terminal complement inhibitor eculizumab that prevents MAC formation. We assessed the change in CFU/ml between 10% NHS and the different parameters. There was a significant reduction when bacteria were grown in 10% NHS compared to the baseline (No NHS) (*p* = 0.0046). Heat-inactivated human serum (IHS) was applied as a control of complement activity and did not only lead to rescue of bacterial growth, but could be interpreted to function as a growth medium for the bacteria. The C1q inhibitor did not rescue the bacteria (*p* = 0.6904), suggesting no involvement from the classical pathway in the observed serum-mediated killing. The ficolin-2 inhibitor and the factor D inhibitor each led to significant bacterial rescue (*p* = 0.0029 and

Figure 5 | Inhibitory effects of ficolin-2 inhibitor on ficolin-2 binding. 10% normal human serum (NHS) was combined with ficolin-2 inhibitor FCN212 (5 µg/ml) and incubated with isolate C322-17. Detection was performed with specific biotinylated monoclonal antibodies against ficolin-2, followed by FITC-conjugated Streptavidin. (A) Ficolin-2 deposition was measured by flow cytometry and expressed as mean fluorescence intensity (MFI). (B) Residual EAEC cells from flow cytometry was placed on slides by cytospin and examined by bright field (BF) and fluorescent microscopy (FITC). Results represent the means of three independent experiments ± SD, \*\*\*\**p* ≤ 0.0001 (unpaired Student's *t*-test).

0.0130, respectively), suggesting involvement of both the lectin and alternative pathway in serum-mediated killing. Combining the two inhibitors did not appear to further increase the rescue. Targeting C3 and MAC with inhibitors also led to significant rescue of bacterial growth (*p* = 0.0107 and 0.0038, respectively) (**Figure 9**).

# DISCUSSION

Enteroaggregative *Escherichia coli* (EAEC) is a well-known diarrheagenic pathogen, causing acute and persistent diarrhea in children and adults worldwide. However, the molecular epidemiology of EAEC still remains unclear and several studies have recovered EAEC from stool samples of apparently healthy individuals suggesting the involvement of host factors (2).

The lectin pathway of the complement system relies on PRMs to assist in the clearance of microbial intruders. Ficolins are a family of PRMs belonging to the lectin pathway. They bind structures such as *N*-acetyl-glucosamine (GlcNAc), *N*-acetyl-galactosamine (GalNAc) and acetylated compounds on target cells (24). Although complement proteins are generally considered in the systemic compartment, recent studies show

Figure 7 | Inhibitory effects of a C1q inhibitor on C4b, C3b, and membrane attack complex (MAC) deposition. 10% normal human serum (NHS) was combined with C1q inhibitor C1q-85 (5 µg/ml) and incubated with isolate C322-17. Detection was done with specific biotinylated monoclonal antibodies against (A) C4b, (B) C3b, and (C) MAC, followed by FITC-conjugated Streptavidin. Complement deposition was measured by flow cytometry and expressed as mean fluorescence intensity (MFI). Results represent the means of three independent experiments ± SD. No significant differences were detected between the C1q inhibitor and the mock inhibitor, ns *p* > 0.05 (unpaired Student's *t*-test).

production and secretion of complement components in human immune cells, as well as endothelial and epithelial cells (9). This makes the study of interactions between rare or non-systemic pathogens and complement highly relevant and could potentially give a better understanding of host defense systems, as well as bacterial pathogenesis.

In this study, we assessed the involvement of the lectin complement pathway on 56 EAEC strains isolated from patients suffering from EAEC-related diarrhea. First, we applied a consumption assay to screen the binding of rMBL and ficolins to prototypic EAEC strains in an ELISA setup. We found that rficolin-2 presented with the highest binding capacities for the prototype EAEC strains, showing strong binding to prototype strains 55989 and C1010-00, and to some degree to prototype 042. In a previous study no binding of ficolin-2 to prototype strain *E. coli* 042 was detected, but they were using NHS (serum protein) and in a different *E. coli* growth medium (15). Using a different growth medium for EAEC could potentially change the gene expression of the strain and lead to changes in surface presentation. Biofilm formation in some EAEC strains have shown to increase

significantly when using a high glucose containing medium as compared to regular LB (25). This emphasizes the importance of the methodology employed when studying bacterial interaction with host factors.

Based on our initial screenings, we focused on the PRM ficolin-2. Previous reports have described binding of ficolin-2 to Gram-positive bacteria, such as group B *streptococci*, *S. pneumoniae, S. pyogenes*, and capsulated *S. aureus* (26–28), but very few studies have described binding of ficolin-2 to Gram-negative bacteria. A study by Sahagún-Ruiz et al. explored the binding capacity of serum ficolin-2 and ficolin-3 to Gram-negative bacteria including four EAEC strains. They found that ficolin-2 and ficolin-3 recognized one EAEC strain (serotype O71), but not the other three. By testing binding to another EAEC serotype O71 they concluded that binding was not related to the bacterial LPS type (15), but did not determine the specific binding factor.

The binding capacity of the 56 clinical EAEC isolates was screened by incubating bacteria with 10% NHS. Ficolin-2 binding was only detected in five of the 56 EAEC isolates (8.9%). Complement evasion strategies are numerous and include mimicking or recruitment of complement regulators, inhibition of complement proteins, and enzymatic degradation leading to inactivation (29). Both Gram-positive and Gram-negative bacteria are capable of evading complement by recruitment of complement regulators, such as factor H and C4b-binding protein (C4BP) (30). The *E. coli* K1, causing neonatal meningitis, utilizes the outer membrane protein A (OmpA) as protection against complement-mediated killing, and serum resistance was correlated with the binding of C4bP to OmpA (31, 32). Immune evasion has also been reported for EAEC by cleavage of complement proteins by Pic, which is a serine protease present in approximately 50% of clinical EAEC isolates (4, 5, 33, 34). In a study by Abreu et al., they showed that Pic significantly reduced complement activation by cleavage of C3, C3b, C4, and C2, thereby affecting all three complement pathways (35). EAEC is known to cause persistent diarrhea, most likely due to the formation of a resilient biofilm. A study from 2013 showed that biofilm formation worked as an efficient way of evading complement detection (36) and could potentially be a way for EAEC to avoid complement. As part of the present study, we investigated whether ficolin-2 binding could be related to the five EAEC characteristic AAF (I–V). By PCR, we characterized the presence of AAFs in the 56 clinical EAEC strains, but we were unable to find a link between binding of ficolin-2 and the presence of a specific AAF, the strains that bound ficolin-2 harbored different AAFs (data not shown). We hypothesize that some bacterial strains are able to change their surface composition in order to avoid ficolin-2 binding, thus escaping the initiating effect of complement.

It is important to mention that we did see a difference in detection of ficolin-2 binding when comparing ELISA with western blot. Two of the strains reported negative for binding in the ELISA, were run on a western blot where we were able to detect low levels of oligomeric binding. This should be tested further to determine the sensitivity of the ELISA.

Before assessing the complement activation potential of EAEC, we determined whether ficolin-2 binding was calcium dependent. X-ray crystallographic analysis has revealed four different binding sites in the fibrinogen-like domain of ficolin-2, enabling binding to a wide variety of ligands (24). While some of these binding sites require the presence of calcium, others can bind calcium independently (10). Using isolate C322-17 as a model for high level ficolin-2 binding, we could show that ficolin-2 binding to the bacteria was calciumindependent.

We next showed that ficolin-2 binding to EAEC can lead to activation of complement. The activation could be partially inhibited by a ficolin-2-specific inhibitor. Furthermore, activation was independent of the classical pathway, thereby confirming the involvement of the lectin pathway of complement. We also show that the alternative pathway contributes to activation and that the amplification loop contributes to a high production of C3b, possibly explaining the need to introduce both a ficolin-2 inhibitor and a factor D inhibitor to see a decreasing effect on C3b levels. This observation is compatible with previous studies on fungi showing the importance of alternative pathway amplification on lectin pathway-initiated activation (18).

Finally, we show that ficolin-2 and factor D are involved in serum-mediated killing of EAEC and that this killing is completely dependent on the formation of the MAC complex. Although EAEC is not common in sepsis infections, there have been reports of EAEC-induced bacteremia (37, 38), and this could be due to lack of complement dependent systemic control of the infection.

#### CONCLUSION

This study shows for the first time a crucial influence of ficolin-2 in the control of some strains of EAEC bacteria, but it also shows that only a fraction of the strains indeed binds ficolin-2 by mechanisms that remain to be elucidated. This suggests that many EAEC strains may evade ficolin-2 and probably other innate immune recognition mechanisms in their pathogenic survival strategies.

#### AUTHOR CONTRIBUTIONS

CS: study design, experimental work, data interpretation, drafting the article, and final approval. AR, KK, and PG: study design, data interpretation, critical revision of the article, and final approval. BJ: collection of clinical strains used in the study and final approval.

#### REFERENCES


#### ACKNOWLEDGMENTS

The authors wish to thank their colleagues from the Department of Clinical Immunology at Rigshospitalet. The authors acknowledge the excellent technical assistance from Ms. Jytte Bryde Clausen. The authors also wish to thank Rie Jønsson and Susanne Jespersen from Statens Serum Institut for assistance in handling and characterizing the EAEC strains.

#### FUNDING

This work was supported by grants from the Danish Research Foundation of Independent Research (DFF-6110-00489), the Sven Andersen Research Foundation, the Novo Nordisk Research Foundation, and the Research Foundation at Rigshospitalet.

*Escherichia coli*. *Immunobiology* (2015) 220:1177–85. doi:10.1016/j.imbio. 2015.06.001


recognition molecules mannan-binding lectin, L-ficolin, and H-ficolin. *Infect Immun* (2005) 73:1052–60. doi:10.1128/IAI.73.2.1052-1060.2005


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be constructed as a potential conflict of interest.

*Copyright © 2018 Adler Sørensen, Rosbjerg, Hebbelstrup Jensen, Krogfelt and Garred. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Aleksandra Man-Kupisinska1 , Anna S. Swierzko2 , Anna Maciejewska1 , Monika Hoc1 , Antoni Rozalski <sup>3</sup> , Malgorzata Siwinska4 , Czeslaw Lugowski <sup>1</sup> , Maciej Cedzynski <sup>2</sup> and Jolanta Lukasiewicz <sup>1</sup> \**

*<sup>1</sup> Laboratory of Microbial Immunochemistry and Vaccines, Department of Immunochemistry, Ludwik Hirszfeld Institute of Immunology and Experimental Therapy, Polish Academy of Sciences, Wroclaw, Poland, 2 Laboratory of Immunobiology of Infections, Institute of Medical Biology, Polish Academy of Sciences, Lodz, Poland, 3Department of Biology of Bacteria, Faculty of Biology and Environmental Protection, Institute of Microbiology, Biotechnology and Immunology, University of Lodz, Lodz, Poland, 4 Laboratory of General Microbiology, Faculty of Biology and Environmental Protection, Institute of Microbiology, Biotechnology and Immunology, University of Lodz, Lodz, Poland*

#### *Edited by:*

*Robert Braidwood Sim, University of Oxford, United Kingdom*

#### *Reviewed by:*

*Teizo Fujita, Fukushima Medical University, Japan Péter Gál, Institute of Enzymology (MTA), Hungary*

*\*Correspondence: Jolanta Lukasiewicz jolanta.lukasiewicz@iitd.pan.wroc.pl*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 24 April 2018 Accepted: 15 June 2018 Published: 29 June 2018*

#### *Citation:*

*Man-Kupisinska A, Swierzko AS, Maciejewska A, Hoc M, Rozalski A, Siwinska M, Lugowski C, Cedzynski M and Lukasiewicz J (2018) Interaction of Mannose-Binding Lectin With Lipopolysaccharide Outer Core Region and Its Biological Consequences. Front. Immunol. 9:1498. doi: 10.3389/fimmu.2018.01498*

Lipopolysaccharide (LPS, endotoxin), the main surface antigen and virulence factor of Gram-negative bacteria, is composed of lipid A, core oligosaccharide, and O-specific polysaccharide (O-PS) regions. Each LPS region is capable of complement activation. We have demonstrated that LPS of *Hafnia alvei*, an opportunistic human pathogen, reacts strongly with human and murine mannose-binding lectins (MBLs). Moreover, MBL–LPS interactions were detected for the majority of other Gram-negative species investigated. *H. alvei* was used as a model pathogen to investigate the biological consequences of these interactions. The core oligosaccharide region of *H. alvei* LPS was identified as the main target for human and murine MBL, especially l- *glycero*<sup>d</sup>-*manno*-heptose (Hep) and *N*-acetyl-d-glucosamine (GlcNAc) residues within the outer core region. MBL-binding motifs of LPS are accessible to MBL on the surface of bacterial cells and LPS aggregates. Generally, the accessibility of outer core structures for interaction with MBL is highest during the lag phase of bacterial growth. The LPS core oligosaccharide–MBL interactions led to complement activation and also induced an anaphylactoid shock in mice. Unlike *Klebsiella pneumoniae* O3 LPS, robust lectin pathway activation of *H. alvei* LPS *in vivo* was mainly the result of outer core recognition by MBL; involvement of the O-PS is not necessary for anaphylactoid shock induction. Our results contribute to a better understanding of MBL–LPS interaction and may support development of therapeutic strategies against sepsis based on complement inhibition.

Keywords: lipopolysaccharide, endotoxin, anaphylactoid shock, complement, mannose-binding lectin, *Hafnia*

**Abbreviations:** Ab, antibody; AP, alternative pathway; CP, classical pathway; DS, disaccharide; LP, lectin pathway; LPS, lipopolysaccharide, endotoxin; MBL, mannose-binding lectin; MASP, MBL-associated serine proteases; MHP, mannose homopolymer; NHS; normal human serum; OS, oligosaccharide; O-PS, O-specific polysaccharide; PAMP, pathogen-associated molecular pattern; R-LPS, rough LPS; S-LPS; smooth LPS; TS, trisaccharide.

### INTRODUCTION

Mannose-binding lectin (MBL) is one of several pattern recognition molecules forming complexes with MBL-associated serine proteases (MASP) able to activate complement *via* the lectin pathway (LP). That process contributes to clearance of infection, but when excessive may be detrimental to the host (1).

Humans synthesize one type of MBL (hMBL), whereas mice (like the majority of mammals) synthesize two forms, MBL-A and -C, differing slightly in their specificity, serum concentration, activity, and local expression (2–5). Generally, hMBL recognizes carbohydrate patterns present on pathogens that are rich in d-mannose (d-Man), *N*-acetyl-d-glucosamine (d-GlcNAc), *N*-acetyl-d-mannosamine (d-ManNAc), or l-fucose (l-Fuc).

Lipopolysaccharide (LPS, endotoxin), the main surface antigen of Gram-negative bacteria, may be a ligand of MBL. LPS is composed of lipid A linked to a core oligosaccharide (OS) consisting of inner and outer regions that is further substituted with O-specific polysaccharide (O-PS) comprising oligosaccharide repeating units. O-PS is a very variable region that determines O-serotype, whereas core OS and lipid A are characterized by moderate structural variability. Smooth bacterial strains synthesize highly heterogeneous LPS being the mixture of S-LPS built of all three regions and short R-LPS (devoid of the O-PS) (**Figure 1**). Rough bacteria synthesize exclusively R-LPS. Such factors as bacterial growth phase and temperature influence LPS heterogeneity (6).

Each LPS region may induce synthesis of specific antibodies (Ab), able to activate the classical pathway (CP) of complement activation. However, in the absence of Ab, lipid A may activate CP *via* direct binding of C1, while core OS-LP (MBL-dependent) and O-PS may activate the alternative pathway (AP) and/or LP (involving MBL or ficolins) (7–10). Recently, MASP-1 (crucial for activating MASP-2 and therefore initiation of the LP cascade) was shown to participate in LPS-induced AP activation (11). Regarding core OS, l-*glycero*-d-*manno*-heptose (Hep) in the inner core region (characteristic for majority of LPS) and d-GlcNAc in the outer core region end (in *Salmonella enterica* serovar Minnesota) were reported as hMBL-binding motifs in R-LPS (12, 13). Although lipid A is considered the toxic principle of LPS, responsible for CD14–TLR-4–MD-2 complex-dependent immune cell response, the contribution of LPS polysaccharideinduced complement activation seems to be important for development of septic shock. Unlike lipid A-dependent endotoxic shock, polysaccharide-induced anaphylactoid reactions can be evoked in LPS-hyporesponsive mice (14, 15). Intravenous injection of certain S-LPS (but not isolated lipid A or R-type LPS) leads to rapid accumulation of platelets in the lungs and liver, followed by their degradation and release of serotonin, and death within 15–60 min, preceded by characteristic symptoms like convulsions and unconsciousness (16). Complement activated by LPS–MBL may be responsible for the degradation of platelets (16). LPS having mannose homopolymers (MHP) as O-PS (e.g., *Klebsiella pneumoniae* O3) (17) are potent inducers of anaphylaxis-like endotoxic shock in mice (16, 18). Some smooth bacteria (including *Proteus vulgaris* O25, *S. enterica* ser. Minnesota, and Abortusequi) have MBL-binding motifs within the core OS only and are capable of inducing a lethal early-phase shock (19, 20).

*Hafnia alvei* is an opportunistic human pathogen responsible for nosocomial mixed infections and sepsis (21). Most *H. alvei* LPS possesses smooth forms. So far, 40 O-serotypes (O-PS structures), and 4 types of core OS have been identified. *H. alvei* LPS is also an example of endotoxin having the *E. coli*-type structure of lipid A (22–24). A few strains of *H. alvei* synthesize LPS containing *E. coli* R4 [strains Polish Collection of Microorganisms (PCM) 23 or 1222] or *Salmonella* Ra (strain PCM 1212) core types (**Figure 1**) (25, 26). The OS1 hexasaccharide is the predominant core OS for this species, with Hep and Kdo residues in its inner core region like most Gram-negative bacteria (**Table 1**, footnote f) (24, 27).

A peculiarity of *H. alvei* LPS is the presence of Hep-Kdocontaining motifs also in the outer core region (24) (**Figure 1**). Branched trisaccharide (TS1), l-α-d-Hep*p*-(1→4)-[α-d-Gal*p*6 OAc-(1→7)]-α-Kdo*p*-(2→, was identified, for example, at the outer core region of *H. alvei* 32 and PCM 1192 LPS (OS1-TS1 core) (24, 31). Linear trisaccharide α-d-Gal*p*-(1→2)-l-α-d-Hep*p*-(1→4)-α-Kdo*p*-(2→ (TS2) is characteristic for *H. alvei* PCM 1196 (OS1-TS2 core) (32). The disaccharide (DS), l-α-d-Hep*p*3OAc-(1→4)-α-Kdo*p*-(2→ was identified in *H. alvei* PCM 1200 and 1209 (OS1-DS type core) (29). The presence of Hep-Kdo-containing motifs in the outer core region makes *H. alvei* LPS similar to *K. pneumoniae* and *P. vulgaris* O25 LPS (28, 33). This similarity prompted us to examine the ability of *H. alvei* LPS to bind MBL, activate human and murine complement systems and induce anaphylactoid reactions in mice.

Here, we explicate the structural basis of interactions between MBL and core OS of a variety of *H. alvei* LPS. These interactions lead to the activation of complement *via* the LP. Moreover, complexes of *H. alvei* LPS with MBL were able to induce anaphylactoid shock in BALB/c mice. LPS from 10 different species of opportunistic pathogens were tested to identify other examples of such interactions. We suggest that common interactions between core OS of LPS and MBL triggering LP activation might influence the course of Gram-negative infections, including nosocomial infections and sepsis. Therefore, consideration of surface antigen structure should be helpful in understanding pathogenicity and may influence development of new therapeutic strategies in Gram-negative sepsis.

#### MATERIALS AND METHODS

#### Animals

BALB/c mice (males, 7–8 weeks old) were purchased from the animal facility of the Polish Mother's Memorial Hospital, Research Institute, Lodz, Poland. The BALB/c mice were housed at the animal facility of the Ludwik Hirszfeld Institute of Immunology and Experimental Therapy (Wroclaw, Poland) and *in vivo* experiments were approved by the Local Ethical Commission for Animal Experimentation (Wroclaw, Poland).

#### Bacteria

*Hafnia alvei* strains PCM 537, 1188, 1190, 1191, 1192, 1195, 1196, 1200, 1203, 1204, 1205, 1206, 1207, 1208, 1209, 1210,

TABLE 1 | Structural characteristics of LPS and lectin blotting results of SDS-PAGE separated *Hafnia alvei* LPS with serum-derived hMBL.a


*a All LPS represent smooth type molecules. "*+*" or "*−*" indicate positive or negative interaction with hMBL, respectively. R4, Ra, and OS1 indicate core oligosaccharides present in LPS of E. coli R4, Salmonella spp. Ra, and typical H. alvei core OS, respectively (26, 27).*

*bTerminal residues present in outer core OS region (28).*

*c <sup>l</sup>-*α*-d-Hepp3OAc-(1*→*4)-*α*-Kdop (10, 23, 29).*

*<sup>d</sup>l-*α*-d-Hepp-(1*→*4)-[*α*-d-Galp6OAc-(1*→*7)]-*α*-Kdop (24).*

*e* α*-d-Galp-(1*→*2)-l-*α*-d-Hepp-(1*→*4)-*α*-Kdop (30).*

*f* α*-d-Glcp-(1*→*3)-*α*-d-Glcp-(1*→*3)-[l-*α*-d-Hepp-(1*→*7)]-l-*α*-d-Hepp4P-(1*→*3)-l-*α*-d-*

*Hepp4PPEtn-(1*→*5)-*α*-Kdo (27).*

*nd, not determined; LPS, lipopolysaccharide; hMBL, human MBL.*

*Schematic structures are presented in Figure 1.*

1211, 1212, 1213, 1214, 1218, 1220, 1221, 1222, 1224, and *E. coli* O55 were obtained from the PCM at the Ludwik Hirszfeld Institute of Immunology and Experimental Therapy (Wroclaw, Poland). *Proteus* spp. strains (*P. mirabilis*, *P. vulgaris*, *P. penneri*, *P. myxofaciens*, and *P. genomospecies*) came from the collection of the Laboratory of General Microbiology, University of Lodz (Poland). *K. pneumoniae* O3:K55<sup>−</sup> (strain 5505Δ*cps*) was kindly provided by Prof. S. Kaluzewski (National Institute of Hygiene, Warsaw, Poland). *H. alvei*, *E. coli*, and *K. pneumoniae* were grown till exponential phase (8 h) in Davis medium as described (34), and *Proteus* spp. strains were grown in liquid nutrient broth containing 1% glucose (35). They were stored in a glycerol mixture at −75°C.

#### Sera

Sera obtained from BALB/c mice were used as a source of murine MBL and ficolins. Pooled normal human serum (NHS) was used as a source of hMBL and came from the collection of the Laboratory of Immunobiology of Infections, Institute of Medical Biology, Polish Academy of Sciences. Polyclonal rabbit sera anti-*H. alvei* core OS (OS1) conjugated with tetanus toxoid (OS1-TT) came from Laboratory of Microbial Immunochemistry and Vaccines (Ludwik Hirszfeld Institute of Immunology and Experimental Therapy, Wroclaw, Poland). Polyclonal rabbit immunoglobulins specific for TS (Hep- [Gal]-Kdo) were isolated from antisera using an adsorption on bacterial mass as previously described (24). DS-specific Ab were isolated by two-step affinity chromatography of anti-*H. alvei* 1209 serum (immunization with killed bacteria, DS-positive strain) on: (i) *H. alvei* 1209 core OS1-Sepharose 4B gel and (ii) *H. alvei* 1209 O-PS-Sepharose 4B gel. Both resins were prepared as previously described (10, 36, 37). Eluates containing anti-DS Ab were collected in sterile vials and stored at −20°C.

#### Preparation of LPS

Lipopolysaccharide were extracted from bacterial cells by the hot phenol/water method (38) and purified by ultracentrifugation as previously described (34, 35). *Proteus* spp. LPS were extracted from dried bacterial cells, as previously described (39), by the phenol–water procedure according to the method of Westphal and Jann (38) and purified with aqueous 50% trichloroacetic acid. For analyses of growth phase dependence of hMBL–bacteria interactions (SDS-PAGE and lectin blotting), LPS was isolated from bacteria by Tri-Reagent method (40). LPS of *H. alvei* strains 1, 2, 17, 23, 31, 32, 37, 38, 39, 114/60, 481L, 600, 744, 981, *Edwardsiella anguillimortifera*, *Citrobacter* (kindly provided by Prof. E. Katzenellenbogen), and *E. coli* came from the collection of the Laboratory of Microbial Immunochemistry and Vaccines (Ludwik Hirszfeld Institute of Immunology and Experimental Therapy, Wroclaw, Poland).

# O-PSs, Core Oligosaccharides, and Lipid A Isolation

Polysaccharides, oligosaccharides, and lipids A were isolated by mild acidic hydrolysis of *H. alvei* PCM 1190, 1192, and 1200 LPS at 100°C for 45 min. Poly- and oligosaccharides were fractionated and purified as previously described using Bio-Gel P-10 (10). The Hep-Kdo-containing fraction was isolated from the heterogeneous core OS fraction and analyzed by the use of liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS) on SeQuant®ZIC®-HILIC column as previously described (41). Fractions 3 and 4 were pooled and used for surface plasmon resonance (SPR) analysis. Lipid A was isolated as a water-insoluble fraction of the LPS hydrolyzate. Prior to lectin blotting, lipids A were purified by extraction with 2:1:3 chloroform/methanol/water mixture (v/v/v) to remove membrane phospholipids and remains of LPS. Both water phase (w) and chloroform (ch) phase lipids A were collected (23).

#### SDS-PAGE

The LPS and lipids A were analyzed by SDS-PAGE. Briefly, 4.5 and 15.4% polyacrylamide-bisacrylamide gels were used as the stacking and resolving gels, respectively. Glycolipids (3 µg) were mixed with sample buffer (65 mM Tris, pH 6.8, 2% SDS, 35% glycerol, 0.6 M DTT, ~0.1% bromophenol blue) in ratio 1:1 (v/v). LPS/lipid A bands were visualized by the silver staining method (42).

#### Lectin Blotting

SDS-PAGE-separated LPS were transferred onto polyvinylidene fluoride membranes (Bio-Rad, USA). Membranes were blocked with SuperBlock® Blocking Buffer (Thermo Scientific, USA) for 2 h, followed by overnight incubation at 4°C, with 25-fold diluted human or murine serum as previously described (10). Bound proteins were detected by immunostaining with different primary Ab: (i) monoclonal mouse anti-hMBL Ab (clone HYB 131-01, BioPorto, Denmark), (ii) monoclonal rat anti-MBL-A (clone 2B4) and (iii) anti-MBL-C Ab (clone 16A8) (both from Hycult Biotech, The Netherlands), (iv) rabbit anti-ficolin-A kindly provided by Dr. Yuichi Endo (Fukushima Medical University, Fukushima, Japan), and (v) reactions were detected with HRP-conjugated rabbit anti-mouse, anti-rat secondary IgG Ab (Dako, Denmark) or anti-rabbit IgG secondary Ab, and visualized with Immun-Star HRP Chemiluminescent Substrate Kit (Bio-Rad, USA) and G:Box chemiluminescent imaging system (Syngene, UK). Nonspecific interactions of secondary Ab were excluded by controls without the primary Ab or the serum as a source of hMBL.

# ELISA for Lectin Binding

Interactions of hMBL, MBL-A, MBL-C, and murine ficolins A and B with *H. alvei* LPS were tested as previously described (10). Briefly, NUNC Maxisorp U96 plates were coated with 2 µg of LPS/well. After blocking with 0.1% BSA in TBS-Ca2<sup>+</sup> buffer (10 mM Tris, 120 mM NaCl, 1 mM CaCl2, pH 7.4), NHS or murine serum (prediluted in 0.1% BSA/20 mM Tris, 1 M NaCl, 10 mM CaCl2, pH 7.4) was added. After overnight incubation at 4°C, the bound proteins were detected by using specific primary antibodies (mentioned in the lectin blotting procedure) and HPRconjugated anti-mouse, anti-rat, or anti-rabbit corresponding anti-IgG antibodies (Dako, Denmark). As substrate for peroxidase, 2,2′-azino-bis(3-ethylbenz-thiazoline-6-sulfonic)acid (ABTS) (Sigma, USA) was employed. Absorbance values were measured at 405 nm using Benchmark Plus microplate spectrophotometer (Bio-Rad).

#### Determination of hMBL–MASP-2 and MBL–MASP-1 Complex Activity

Activity of lectin(s)-MASP-2 complexes was determined as previously described (43) with modification (44). LPS from various bacteria were used for coating of microtiter plates (Maxisorp U96, Nunc). The products of C4 activation were detected with rabbit anti-hC4c and HRP-conjugated anti-rabbit Ig (Dako). To test MBL-MASP-1 complex activity, VPR-AMC (Val-Pro-Argaminomethylcoumarin) peptide (Bachem, Switzerland), as the substrate for MASP-1 was used as previously described (45) and the fluorescence was read using a Varioskan Flash reader (Thermo Scientific, USA).

#### Flow Cytometry Analysis of Binding of hMBL to Formaldehyde-Inactivated Bacteria

Flow cytometry analysis was performed as previously described (10). Depending on experiment, bacteria were cultured and harvested at lag phase (3 h), log phase (6 h), or stationary phase (24 h) of growth. The growth phase of culture was determined on the basis of optical density measurement at 600 nm and the appropriate growth curve. Immediately before each experiment, bacterial cells were centrifuged, washed with PBS, and suspended in 10-fold diluted NHS (pool), used as a source of hMBL. Monoclonal anti-hMBL Ab (clone HYB 131-01) and fluorescein isothiocyanate-labeled anti-mouse IgG Ab (Dako) were used as detection system. The analysis of the FITC-labeled bacteria was performed using a Cytomics FC 500 MPL Beckman-Coulter (USA) flow cytometer. Bacteria were detected using log-forward and log-side scatter dot plot. Gating region was set to exclude debris and larger aggregates of bacteria. A total of 10,000 events were acquired.

#### Induction of Anaphylaxis-Like Endotoxic Shock in muramyldipeptide (MDP)-Primed Mice

The BALB/c mice were treated i.p. with 100 µg of MDP in PBS (20), and after 4 h, animals received i.v. 100 µg of LPS. *K. pneumoniae* O3 and *E. coli* O55 LPS were used as a positive and negative control, respectively. Incidence, severity, and scoring of the anaphylaxis-like shock were recorded within 30 min: 0, no signs of shock; 1, staggering; 2, crawling and prostration; 3, prostration and weak convulsions; 4, prostration and strong convulsions (16). Subsequent mortality was recorded within 1 h and after 24 h after LPS injection.

# Surface Plasmon Resonance

Surface plasmon resonance studies were assessed with a Biacore T200 system (GE Healthcare Bio-Science AB, Sweden). Carrier free recombinant hMBL (R&D Systems, USA) was immobilized in 10 mM sodium acetate, pH 4.0 on the CM5 series S sensor chip (GE Healthcare Bio-Science AB) at a flow rate 5 µl/min, to the level of 16,000 RU using the amine coupling chemistry. A flow cell with immobilized 240 RU of ethanolamine was used as a reference surface. HBS-P buffer (GE Healthcare Bio-Science AB) supplemented with Ca2+, Mg2<sup>+</sup> ions (5 mM MgCl2, 5 mM CaCl2) was used as a running buffer. *H. alvei* O-PS (PCM 1190, 1196, 1200, and 1209) and core hexasaccharides from *H. alvei* PCM 1200 LPS at various concentrations were injected at a flow rate 30 µl/min. 0.5% SDS injected for 30 s was used as a regenerator in all SPR experiments.

# RESULTS

# *H. alvei* LPS Core Oligosaccharide Is a Common MBL Target

Screening for the presence of structural motifs recognized by hMBL was performed by lectin blotting for LPS isolated from 39 different O-serotypes of *H. alvei* (**Figure 2**). LPS isolated from smooth strains gave a characteristic ladder-like multiband pattern after SDS-PAGE reflecting natural heterogeneity of LPS on the cell wall surface and facilitating core OS accessibility (**Figure 2A**). The fast-migrating fractions originate from lipid A substituted with core OS, whereas the slow-migrating fractions show the length distribution of the polymer built up of lipid A-core OS substituted with varying numbers of O-PS repeating units. LPS isolated from *K. pneumoniae* O3 strain 5505 (MHP as O-PS) was used as a positive control, where interactions were observed both in the core OS/lipid A and O-PS regions. hMBL recognized 34 out of 39 *H. alvei* LPS. Generally, reactivity was related to the fast-migrating fractions of the core OS-lipid A region. No interactions were

observed for *H. alvei* 23, PCM 744, 1204, 1212, and 1222 LPS (**Figure 2B**).

represents *Klebsiella pneumoniae* O3:K55− LPS used as a positive control.

Nine representative LPS, chosen on the basis of well-characterized structure (**Figure 1**) and different hMBL-binding patterns (**Figure 2B**; **Table 1**), were selected for further experiments to explore human and murine MBL specificity (**Figure 3**). For *H. alvei* PCM 1192, 1200, 1209, and 1212, hMBL bound within the core OS region only. For *H. alvei* PCM 1190 and 1196, hMBL bound within both the core OS and O-PS regions. For *H. alvei* 23 and PCM 1222, no binding was observed. These LPS–hMBL interactions were confirmed by ELISA (**Figure 4**). Eight *H. alvei* LPS (and *K. pneumoniae* O3 LPS as positive control) were used as solid-phase antigens. The strongest reactions of serum hMBL were observed for LPS *K. pneumoniae* O3 and *H. alvei* PCM 1190, 1196, and 1209 LPS. In contrast to the lectin blotting, no reaction with *H. alvei* PCM 1200 LPS was observed, what might be explained by competition between strong binding of O-PS-reactive ficolin-3 (10) and moderate binding of core OS-reactive hMBL. In addition, long O-PS chains of LPS 1200 might also hinder hMBL access to core OS.

Since murine model was chosen for further studies to test *in vivo* activity of LPS on complement-mediated anaphylaxislike endotoxic shock, the reactivity of LPS with murine MBL-A and MBL-C (as well as with ficolin A and ficolin B) was analyzed by lectin blotting (**Figure 3**). Murine MBL-C showed binding pattern very similar to hMBL within the core OS region. Strong reactions with *H. alvei* PCM 1190, 1196, 1200, 1209, 1212, and *K. pneumoniae* O3 LPS and negligible reactions with LPS 23, PCM 1192 and 1222 were noted. In addition, interactions with high molecular weight fractions (O-PS region) were easily visible for *H. alvei* 1190 and *K. pneumoniae* O3 LPS. For MBL-A, no reactivity was observed for PCM 1192 and 1222 and very weak reactivity for 23 (within O-PS). Binding of this lectin to LPS *K. pneumoniae* O3 and PCM 1209 strains was attributed to core OS only. Reactivity within both core OS and O-PS region

lipopolysaccharides (LPSs) and lectin blotting of their interactions with human MBL (hMBL) and murine mannose-binding lectin (MBL)-A, MBL-C, ficolin-A, and ficolin-B. Normal human serum and BALB/c mice sera were used as a source of hMBL or murine MBL-A, MBL-C, ficolin A, and B, respectively. Lanes are depicted with the strain number in Polish Collection of Microorganisms. Strain 5505 represents *Klebsiella pneumoniae* O3:K55− LPS used as a positive control.

were observed for *H. alvei* PCM 1190, 1196, 1212, and 1200. Especially strong recognition of PCM 1200 LPS was unique for MBL-A, whereas MBL-C (similarly to hMBL) were devoid of

such activity. Moreover, only traces of signal related to O-PS region were observed for interactions of ficolin A with PCM 1200 and ficolin B with PCM 1222 (**Figure 3**). Interactions of MBL-A and MBL-C with *H. alvei* lipid A was excluded by lectin blotting (**Figure 5B**).

#### Hep-Kdo Motifs in *H*. *alvei* LPS Inner and Outer Core Are Recognized by MBL

From lectin blotting (**Figures 2** and **3**), it was suggested that most *H. alvei* LPS were bound by hMBL *via* the core OS/lipid A region. SPR analyses confirmed interactions of hMBL with O-PS regions of PCM 1190 (46) and 1196 (47) and excluded O-PS of PCM 1209 and 1200 LPS as targets for the lectin (**Figure 5A**). Data from lectin blotting with the use of purified lipid A fractions of *H. alvei* PCM 1190 and 1192 confirmed the lack of hMBL reactivity with that part of LPS (**Figure 5B**).

Next, we identified the core OS regions involved. Immunostaining with the use of OS1, DS- and TS1-specific Ab revealed four bands of low molecular weight fractions of migrating LPS (**Figure 5C**) attributed to lipid A-OS1 (two bands), lipid A-OS1-TS1, and lipid A-OS1-DS molecules. Two bands marked by lipid A-OS1 reflected OS1 heterogeneity related to ethanolamine, phosphate groups, and glycine substituents and are common for all three studied LPS (PCM 1192, 1200, and 1209). The band assigned as lipid A-OS1-DS was present in DS-expressing LPS of *H. alvei* PCM 1200 and 1209, while the lipid A-OS1-TS1 band in LPS of *H. alvei* PCM 1192 LPS.

The ability of recombinant hMBL to bind different core OS fractions of *H. alvei* LPS was further investigated by SPR on Biacore T200 (**Figure 5D**). Core OS isolated from PCM 1200 LPS were used as analytes. Both isolated OS1 and low molecular

recombinant hMBL and O-specific polysaccharides (O-PSs) of *H. alvei* PCM 1190, 1196, 1200, and 1209 LPS. The bars represent binding response of immobilized recombinant hMBL (immobilization level: 16,000 RU) to O-PS (0.5 mg/ml) in running buffer HBS-P supplemented with MgCl2 and CaCl2. (B) Silver stained SDS-PAGE and lectin blotting analysis of the interaction between *H. alvei* lipids A [water phase (w) and chloroform phase (ch) preparations] and serum-derived murine MBL-A, MBL-C, and hMBL. Lanes are depicted with the strain PCM number. (C) Silver stained SDS-PAGE and lectin blotting analyses of the interactions between LPS of *H. alvei* PCM 1192 (TS1-positive), 1200 (DS-positive), 1209 (DS-positive) and serum-derived hMBL, DS-specific rabbit serum, TS1-specific rabbit serum, and OS1-specific rabbit serum. M—molecular weight markers. Marked bands corresponded to different forms of R-LPS: LA-OS1, LA-OS1-DS, LA-OS1- TS1, where LA stands for lipid A. (D) SPR sensograms for interactions of recombinant hMBL and OS1 and Hep-Kdo-containing OS isolated from *H. alvei* 1200 LPS. Sensor chip: CM5; Abbreviations: RU, response units; hMBL, human MBL; PCM, Polish Collection of Microorganisms; MBL, mannose-binding lectin.

weight fraction of Hep-Kdo interacted with immobilized recombinant hMBL in a concentration-dependent manner, with higher affinity observed for OS1. It was also confirmed by ELISA inhibition assay with the use of both analytes (data not shown).

### Bacterial Growth Phase Determines the Accessibility of LPS Core Region for MBL

Binding of hMBL to LPS on bacterial surface was further investigated by flow cytometry. Since bacterial growth phase may be associated with changes in LPS expression, accessibility of core OS regions for hMBL was examined using microbial cells collected at lag (3 h), log (6 h), and stationary phase (24 h) (**Figure 6**). Four strains were chosen for this study: (i) *H. alvei* PCM 1190 recognized by hMBL within DS-carrying core OS and O-PS regions, (ii) *H. alvei* PCM 1192 and 1209 with hMBL targets located in low molecular weight fraction of LPS (core OS1 decorated with TS1 or DS, respectively), and (iii) *H. alvei* PCM 1222 expressing LPS not recognized by hMBL. *K. pneumoniae* O3 grown to the stationary phase was used as a positive control. The percentage of hMBL-labeled cells was confronted with R-LPS and S-LPS distribution examined by SDS-PAGE analysis of LPS extracted from cells at lag, log, and stationary phases (**Figure 6**, inset).

For all hMBL-reacting strains (PCM 1190, 1192, and 1209), the highest values of labeled cells were recorded for lag phase, where R-LPS represented the prevailing LPS population. The low content of O-PS chains facilitated access of hMBL to outer core regions of R-LPS (OS1 and DS). For strains PCM 1190 and 1192, the proportions of labeled cells were inversely

FIGURE 6 | Growth phase-dependence of hMBL–*Hafnia alvei* interactions observed in FACS analyses. Binding of hMBL to inactivated smooth *H. alvei* PCM 1190, 1192, and 1209 bacterial cells at the lag (black bars), exponential (gray bars), and the stationary phases (white bars) of growth. Silver stained SDS-PAGE (inset) of LPS 1190, 1209, and 1192 isolated from bacterial mass collected at lag, log, and stationary phases. Data represent the mean values ± SD from three independent experiments. DS and TS1 stand for the disaccharide and trisaccharide outer core motif in LPS of selected *H. alvei* strains. Abbreviations: Nd, not determined; hMBL, human MBL; PCM, Polish Collection of Microorganisms; LPS, lipopolysaccharide.

associated with the expression of the S-LPS population. The highest values, at each growth phase, were recorded for the PCM 1190 strain. This was expected since its LPS has MBLbinding motifs not only in the core but also in O-PS region. PCM 1192 and 1209 LPS were recognized by hMBL within the core OS region only, whereby the most efficient binding was observed for bacteria at the lag phase (10.4 and 27.7% positive cells, respectively). In contrast to other strains, PCM 1209 bacteria showed the lowest accessibility for hMBL at log but not stationary phase. That might be explained by a higher content of R-LPS forms with accessible OS1 and DS motifs at stationary phase contrary to log phase (as evidenced by SDS-PAGE). Performed experiments demonstrated that observed relationships clearly resulted from LPS structure, i.e., the length of O-PS chains that hindered structural motifs recognized by hMBL (OS1 and DS).

### Interaction of hMBL With *H. alvei* LPS Leads to Complement Activation

The ability of selected LPS to initiate the complement cascade *via* the LP was tested by investigating activation of MASP-1 (cleavage of synthetic substrate, VPR-AMC) and MASP-2 (cleavage of C4) dependent on LPS recognition by LP molecules, especially hMBL. MBL–MASP-1 concentrationdependent activation was triggered by *H. alvei* 23, PCM 1190, 1192, 1196, 1200, 1209 LPS, as well as *K. pneumoniae* O3 (control) (**Figure 7**). The deposition of C4 activation products was additionally noted for PCM 1212 and 1222 LPS. It is worth mentioning that procedure employed does not exclude activation of LP by complexes of ficolin-3 with MASP (as described previously for 23 and PCM 1200 LPS) (10) or other than MBL collectins. Contribution of ficolin-1 and -2 was excluded (10). The influence of CP was excluded by high ionic strength of the buffer that inhibits the binding of C1q to immune complexes and disrupts the C1 complex, whereas MBL complexes integrity is maintained (48). The variations in reactivity profiles (**Figures 4** and **7**) may reflect differences in serum dilution used and sensitivity of assays.

#### Interaction of MBL With *H. alvei* LPS Core Oligosaccharide Induces Anaphylactoid Shock

The biological consequences of *in vivo* MBL interaction with *H. alvei* PCM 1190, 1192, 1200, 1209, and 1212 LPS were tested by ability to induce an anaphylactoid reaction in mice. *K. pneumoniae* O3 and *E. coli* O55 as well as *H. alvei* PCM 1222 LPS were used as positive and negative controls, respectively (14).

Intravenous injection of *H. alvei* PCM 1190, 1200, 1209, or 1212 LPS-induced rapid shock (within 30 min, score 3–4) leading to death of MDP-sensitized BALB/c mice (**Table 2**). The distinctive effect was observed for LPS of *H. alvei* PCM 1200 that was strongly recognized by MBL-A within the O-PS region and moderately within the OS1-DS core region (**Figure 3**). The reaction, comparable to that provoked by *K. pneumoniae* O3 LPS, was also induced by *H. alvei* PCM 1190 (OS1-DS core type) and 1212 (*Salmonella* Ra core type)

moderately bound by MBL-A and MBL-C within the O-PS and core OS regions. *H. alvei* PCM 1209 LPS (OS1-DS core type) that was reactive for murine MBL-A and MBL-C within the core OS region only (**Figure 3**), still had a powerful ability to induce an anaphylactoid shock similar to *K. pneumoniae* O3. By contrast, mice treated with *H. alvei* PCM 1192 (OS1-TS1 core type) or 1222 (*E. coli* R4 core type) LPS developed mild or no characteristic symptoms within the first hour and died in the late phase of endotoxic shock (lipid A-dependent) (**Table 2**), similar to animals injected with *E. coli* O55 LPS (negative control).

#### LPS Core Oligosaccharide Is a Common MBL Target in Many Gram-Negative Bacteria

Screening for LPS from a variety of opportunistic pathogens (Table S1 in Supplementary Material), recognized by serum hMBL was performed with the use of lectin blotting. False positive reactions of primary and secondary detecting Ab were excluded. We found interactions between hMBL and LPS core regions to be very common: 13 of 15 *K. pneumoniae*, 11/22 *P. vulgaris*, 10/33 *P. mirabilis*, 7/15 *P. penneri*, 1/1 *P. myxofaciens*, 5/10 *E. coli* (including R-LPS containing R2 and R3 core types), 1/5 *Citrobacter* spp., and all of 4/4 *Edwardsiella anguillimortifera* LPS gave positive results.

#### DISCUSSION

Lipopolysaccharide is a major pathogen-associated molecular pattern (PAMP) and virulence factor of Gram-negative bacteria, responsible for development of sepsis and septic shock. Whereas the role of lipid A in those life-threatening events is welldocumented (49), the influence of the polysaccharide region is poorly characterized. It is known that the carbohydrate moiety influences endotoxin clearance and biological activity of lipid A (50, 51). Recognition of LPS polysaccharide by a variety of pattern recognition molecules may lead to complement activation *via* CP, AP, and/or LP, all involved in sepsis development (52).

The core OS-lipid A region is a target for such plasma proteins as LPS-binding protein, BPI (bactericidal/permeability-increasing protein), CAP18 (cationic antimicrobial protein), and lysozyme. Consequently, bactericidal and inflammatory processes are induced by the host immune system. Due to the high structural heterogeneity of O-PS, the number of innate immunity factors interacting with that region is much lower. One example is ficolin-3, recognizing *H. alvei* PCM 1200 O-PS resulting in LP activation (10). Ficolin-3 was also demonstrated to enhance agglutination, phagocytosis, and killing of *H. alvei* PCM 1200 bacteria (53). Another example is pulmonary surfactant collectin SP-D binding mannose-rich O-PS of *K. pneumoniae* O3 and O5 (54).

This study provided well-documented evidence that core OS is the main target for human and murine MBL. Depending on the assay, the binding of recombinant (SPR) or NHS or murine serum MBL (lectin blotting, ELISA, flow cytometry) was detected in presented studies. Thus, it is worth noting that the oligomer distribution may vary for recombinant and NHS-derived MBL according to purification procedure (55), what may influences binding affinity between MBL and target ligand. Notwithstanding similar oligomer distribution patterns were reported for both forms (55, 56), including trimeric and tetrameric forms. Even though proposed oligomerization models indicated a polypeptide dimer as the basic unit in this process for MBL (57), higher oligomeric states are usually detected in rMBL and NHS-derived MBL preparations that ensure complement activation (56).

Performing screening analysis, we have shown that interactions of serum hMBL with different core OS regions were prevalent among LPS isolated from numerous opportunistic pathogens, such as *H. alvei*, *E. coli*, *K. pneumoniae*, *Proteus* spp., *Citrobacter*



*a Mice were sensitized with 100 µg of muramyldipetide (i.p.) and 4 h later injected with 100 µg of LPS (i.v.). Incidence, score, and mortality were recorded within 0.5 and 1 h after LPS injection.*

*bIncidence and mortality are shown as number/total.*

*c The scoring of anaphylaxis-like endotoxic shock was as follows: 1, staggering; 2, crawling and prostration; 3, prostration and weak convulsions; 4, prostration and strong convulsions (16).*

*dThe O-PS of this LPS is composed of mannose homopolymer.*

*e Determined only by inhibition of murine MBL–LPS O55 interaction in ELISA with LPS O55.*

*LPS, lipopolysaccharide; MBL, mannose-binding lectin.*

spp., and *E. anguillimortifera* representing different O-serotypes (O-PS structure) (Table S1 in Supplementary Material). Among 145 LPS tested, as much as approximately 57% were recognized by hMBL within fast-migrating fractions (corresponding to R-LPS built up of lipid A-core OS), whereas the reaction with slow-migrating fractions (S-LPS containing O-PS) was found in approximately 10% only. Accordingly, hMBL interacted with 34 (approximately 87%) out of 39 *H. alvei* LPS within the core OS region of LPS. Moreover, we also showed that highly purified *H. alvei* hexaacylated lipid A is recognized neither by human nor murine MBL. However, such interactions were previously suggested by Ono et al. (58) and Shiratsuchi et al. (59) for commercially available *E. coli* lipid A. This discrepancy may result from impurities of bacterial origin in the commercial preparations or the presence of MBL-binding motifs other than in the lipid A region of LPS (residuals of complete LPS after lipid A isolation by LPS hydrolysis).

Clinical isolates of Gram-negative bacteria are commonly of smooth type and therefore synthesize a highly heterogeneous pool of LPS, consisting of long-chain S-LPS, shorter S-LPS, and R-LPS unsubstituted by O-PS. We found that *H. alvei* PCM 1209 core OS within R-LPS forms exposed on the bacterial surface is accessible for hMBL (**Figure 6**). SPR analysis (**Figure 5A**) clearly demonstrated that PCM 1209 O-PS is not the MBL target. The core OS accessibility may depend on natural LPS heterogeneity (coexistence of R-LPS and S-LPS in smooth strains), and is hindered by core OS substitution with O-PS. It may be influenced by growth phase or environmental conditions. Generally, expression of R-LPS containing hMBL-binding motifs decreased with the culture progression (from lag to stationary phase) (**Figure 6**). Moreover, the immune response against O-PS may cause selective pressure on bacteria to lose the ability to express it (phase variation) (60, 61).

The LPS core OS region is relatively conservative and usually composed of an inner core and an outer core built up of Kdo and Hep residues and hexoses and hexosamines, respectively. For example, among *Salmonella* spp. strains one prevailing core type was described (Ra). Using mutants with defects in LPS core OS synthesis it was demonstrated that Hep residues in the inner core region are recognized by human and murine MBL due to their accessibility in truncated and incomplete core OS (12). Even though the inner core is common for the majority of enterobacterial LPS and represents MBL-binding motifs, our results indicated also outer core structures as natural MBL ligands. Hep and Hep-Kdo motifs were detected also in the latter region, for example, in *P. vulgaris* O25 and *K. pneumoniae* O3, O1, O2, O4, and O5 LPS (28, 33) as well as in numerous *H. alvei* strains (expressing DS, TS1, and TS2) (**Figure 1**). The lectin blotting (**Figure 5C**) and SPR analysis (**Figure 5D**) revealed recombinant MBL binding to DS-decorated *H. alvei* PCM 1200 OS1 and OS1 alone. Moreover, interaction of MBL with purified Hep-Kdo-containing motifs was also evidenced (**Figure 5D**), and determined by Hep residue (but not Kdo) according to the previous reports (12, 62). In spite of *manno* configuration, Kdo residues (even terminal) might be excluded as an MBL ligand, since deep rough mutants (Re) of *S. enterica* ser. Typhimurium or *Yersinia enterocolitica* O3, expressing LPS consisting of lipid A and one, two or three Kdo residues were not recognized (12, 63). Thus, Hep and D-GlcNAc present in outer core regions are the main MBL targets. Any steric obstacles within these motifs hinder MBL access, as was demonstrated for TS-OS1 core type of *H. alvei.* In TS1, the DS motif is substituted by terminal Gal residue that prevented hMBL binding to *H. alvei* PCM 1192 (**Figure 5C**).

Our results indicate a crucial role for MBL-binding motifs within the outer core OS in the recognition of *H. alvei* LPS by human and murine MBL, induction of an anaphylactoid reaction and rapid death in MDP-sensitized mice. Previously, it was demonstrated that such events were induced by several LPS of smooth bacteria with O-PS that were homopolymers of mannose (*K. pneumoniae* O3 and O5), able to activate complement *via*

the LP (16). Among six *H. alvei* LPS with different O-serotypes tested, the intravenous injection of DS-containing LPS (PCM 1200, 1209, and 1190) led to development of severe symptoms of an anaphylactoid reaction and resulted in the death of animals within 30 min. Furthermore, *H. alvei* PCM 1212 (synthesizing Ra core type, with d-GlcNAc residue in the outer core region) is as toxic for mice as the aforementioned LPS. By contrast, *H. alvei* PCM 1192 LPS with outer core OS1 substituted with TS1 (preventing MBL binding) was not active. Interestingly, in the case of PCM 1200, 1212, and 1190 LPS, MBL-A was able to recognize not only the core OS (Ra or OS1-DS type) but also the O-PS region. Moreover, MBL-A showed the highest affinity to O-PS of *H. alvei* PCM 1200 LPS, which was found to be the most toxic (**Table 2**). Our results demonstrated that MBL-binding motifs in outer core region are sufficient to induce an anaphylactoid reaction in mice; however, the presence of S-LPS in the heterogeneous LPS preparation was mandatory. It might be suggested that similar to SP-D exhibiting O-PS-stabilized reactivity with common core OS of *K. pneumoniae*, *E. coli*, and *S. enterica* ser. Minnesota LPS, the O-PS–MBL interaction may also stabilize residual interactions of the collectin with the core OS region (54, 64).

The data presented here have extended the repertoire of LPS recognized by MBL, including rough forms present in endotoxin preparations from smooth bacteria (**Figure 1**). Generally, d-GlcNAc or Hep residues in the outer core were common ligands for the lectin. Those structures may be accessible to MBL *in vivo* not only when LPS O-PS is relatively short but also when endotoxin is released due to bacterial cell damage (for example, after treatment of host with antibiotics). We demonstrated also that the O-PS structure might augment immune responses when recognized by MBL (the example of PCM 1200). We believe that clarifying MBL specificity/affinity may contribute to a better understanding of the role of the LP in Gram-negative infections in general, including those leading to sepsis or endotoxic shock. Species of the family *Enterobacteriaceae* are responsible for 40–50% of hospital-acquired infections leading to sepsis and septic shock. Over half of cases in the USA is connected with bacteria of the genera *Klebsiella*, *Escherichia*, *Proteus* or *Enterobacter*, and mortality is in the range of 20–50%. In some cases of invasive infections caused, for example, by *K. pneumoniae*, *E. coli*, or *Proteus* spp., MBL/ficolin-dependent complement activation by common core oligosaccharide regions or MHP might contribute to the severity of infections and sepsis. Although interaction of MBL (or other PAMPs) with LPS is generally beneficial for the host, it may be harmful under certain conditions. Upon antibiotic treatment, aggregates of endotoxin (mixed S- and R-LPS) are released into the bloodstream and activate a host immune response (49). Furthermore, R-LPS was reported to exhibit higher potency in cell activation through the TLR-4/MD-2 receptor (65). Although MBL deficiency has been associated with susceptibility to infections (especially in children or immunocompromised subjects), its contribution to life-threatening events (like

#### REFERENCES

1. De Pascale G, Cutuli SL, Pennisi MA, Antonelli M. The role of mannosebinding lectin in severe sepsis and septic shock. *Mediators Inflamm* (2013) 2013:625803. doi:10.1155/2013/625803

post-operative SIRS) has also been proven (66). Our results contribute to a better understanding of MBL–LPS interaction. They also support further development of therapeutic strategies against sepsis based on complement inhibition or complement-related replacement therapies.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Local Ethical Commission for Animal Experimentation with the headquarters in the Ludwik Hirszfeld Institute of Immunology and Experimental Therapy Polish Academy of Sciences (Wroclaw, Poland). The Local Ethical Commission for Animal Experimentation approved all *in vivo* protocols.

#### AUTHOR CONTRIBUTIONS

JL, AS, AM-K, MC, and CL conceived and planned the experiments. AM-K, AM, AS, and JL carried out the experiments with prevailing role of AM-K and AS. JL, MH, AM, and CL prepared OS1- or TS1-specific rabbit antisera. AM-K, AM, and JL isolated some *H. alvei* LPS. AR and MS isolated and provided the collection of *Proteus* spp. LPS. AM-K, JL, AS, and MC contributed to the interpretation of the results. JL, AS, MC, and AM-K took the lead in writing the manuscript. All the authors approved the manuscript and provided critical feedback.

#### ACKNOWLEDGMENTS

This work was supported by Wroclaw Research Centre EIT+ within the project "Biotechnologies and advanced medical technologies"—BioMed (POIG.01.01,02-02-003/08) co-financed by the European Regional Development Fund (Operational Programme Innovative Economy, 1.1.2) and the Polish Ministry of Science and Higher Education (project No. N40108432/1944). Publication of the results was supported by Wroclaw Centre of Biotechnology, programme The Leading National Research Centre (KNOW) for years 2014–2018. Wojciech Jachymek and Tomasz Niedziela from the Ludwik Hirszfeld Institute of Immunology and Experimental Therapy, Wroclaw, Poland are acknowledged for their assistance during anaphylaxis-like endotoxic shock induction *in vivo*. Dr. Yuichi Endo from Fukushima Medical University, Fukushima, Japan is kindly acknowledged for rabbit anti-murine ficolin-A Ab. We are very grateful to Dr. David C. Kilpatrick for critical reading of the manuscript and helpful discussion.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01498/ full#supplementary-material.


and MBL-C) and study of acute phase responses. *Scand J Immunol* (2001) 53: 489–97. doi:10.1046/j.1365-3083.2001.00908.x


flexibility studied by atomic force microscopy. *J Mol Biol* (2009) 391:246–59. doi:10.1016/j.jmb.2009.05.083


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Man-Kupisinska, Swierzko, Maciejewska, Hoc, Rozalski, Siwinska, Lugowski, Cedzynski and Lukasiewicz. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Virulence associated gene 8 of *Bordetella pertussis* enhances contact system activity by inhibiting the regulatory Function of complement regulator c1 inhibitor

*Elise S. Hovingh1,2, Steven de Maat3 , Alexandra P. M. Cloherty1 , Steven Johnson4 , Elena Pinelli2 , Coen Maas3 and Ilse Jongerius1,2\*†*

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Erik Waage Nielsen, Nord University, Norway Christian Drouet, Université Grenoble Alpes, France*

> *\*Correspondence: Ilse Jongerius i.jongerius@sanquin.nl*

#### *†Present address:*

*Ilse Jongerius, Department of Immunopathology, Sanquin Research and Landsteiner Laboratory of the Academic Medical Center, University of Amsterdam, Amsterdam, Netherlands*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 14 March 2018 Accepted: 11 May 2018 Published: 04 June 2018*

#### *Citation:*

*Hovingh ES, de Maat S, Cloherty APM, Johnson S, Pinelli E, Maas C and Jongerius I (2018) Virulence Associated Gene 8 of Bordetella pertussis Enhances Contact System Activity by Inhibiting the Regulatory Function of Complement Regulator C1 Inhibitor. Front. Immunol. 9:1172. doi: 10.3389/fimmu.2018.01172*

*1Department of Medical Microbiology, University Medical Centre Utrecht, Utrecht University, Utrecht, Netherlands, 2Centre for Infectious Disease Control, National Institute for Public Health and the Environment (RIVM), Bilthoven, Netherlands, 3Department of Clinical Chemistry and Haematology, University Medical Centre Utrecht, Utrecht University, Utrecht, Netherlands, 4Sir William Dunn School of Pathology, University of Oxford, Oxford, United Kingdom*

*Bordetella pertussis* is a Gram-negative bacterium and the causative agent of whooping cough. Whooping cough is currently re-emerging worldwide and, therefore, still poses a continuous global health threat*. B. pertussis* expresses several virulence factors that play a role in evading the human immune response. One of these virulence factors is virulence associated gene 8 (Vag8). Vag8 is a complement evasion molecule that mediates its effects by binding to the complement regulator C1 inhibitor (C1-INH). This regulatory protein is a fluid phase serine protease that controls proenzyme activation and enzyme activity of not only the complement system but also the contact system. Activation of the contact system results in the generation of bradykinin, a pro-inflammatory peptide. Here, the activation of the contact system by *B. pertussis* was explored. We demonstrate that recombinant as well as endogenous Vag8 enhanced contact system activity by binding C1-INH and attenuating its inhibitory function. Moreover, we show that *B. pertussis* itself is able to activate the contact system. This activation was dependent on Vag8 production as a Vag8 knockout *B. pertussis* strain was unable to activate the contact system. These findings show a previously overlooked interaction between the contact system and the respiratory pathogen *B. pertussis.* Activation of the contact system by *B. pertussis* may contribute to its pathogenicity and virulence*.*

Keywords: whooping cough, contact system, virulence associated gene 8, *Bordetella pertussis*, C1 inhibitor

# INTRODUCTION

*Bordetella pertussis* is the causative agent of whooping cough, also known as pertussis, a contagious disease of the respiratory tract that is re-emerging worldwide despite high vaccination coverage. To date, pertussis is still ranked in the top 10 most deadly childhood diseases posing a serious health problem (1). The acellular pertussis vaccine (ACV), used in many industrialized countries, protects against disease for up to 7 years while natural infection confers protection for up to 20 years (2, 3). Alarmingly, the ACV does not prevent transmission of the pathogen (4). For this reason, it is widely accepted that an improved pertussis vaccine is needed (5). In order to improve the pertussis vaccine, it is of great importance to better understand the interactions between the respiratory pathogen *B. pertussis* and the immune system.

The contact system is a key player in innate immunity and is part of the coagulation system (6, 7). The contact system consists of the two proenzymes factor XII (FXII) and plasma prekallikrein and the cofactor high-molecular weight kininogen (HK). *In vitro*, the contact system is activated when FXII binds to a negatively charged surface and is autocleaved forming FXIIa that is further processed to βFXIIa (8). Lessons from human pathology imply that analogous processes may take place on the surface of vascular endothelial cells (9) or platelets (10). FXIIa cleaves plasma prekallikrein forming active plasma kallikrein (PK). Activation of this protease subsequently mediates the cleavage of HK and formation of the pro-inflammatory peptide bradykinin (BK) (11). BK release triggers endothelial permeability resulting in vasodilation and infiltration of leukocytes (7). Activation and activity of the contact system is regulated by the 105 kDa complement regulator C1 inhibitor (C1-INH) (12), which inhibits the activity of β-FXIIa, α-FXIIa, and PK, by forming covalent complexes with its target proteases (13). C1-INH consists of a C-terminal protease inhibiting serpin domain and an N-terminal domain that is predicted to be heavily O- and N-linked glycosylated. Besides being involved in contact system regulation, C1-INH is also the main inhibitor of the classical and lectin pathways of the complement system where it inactivates the respective proteases necessary for activation of the complement cascade (14). Interestingly, the interplay between *B. pertussis* and the contact system remains unexplored even though *Escherichia coli* and *Salmonella* (15), *Streptococcus pyogenes*, *Bacillus stearothermophilus*, *Bacillus subtilis*, *Porphyromonas gingivalis, Pseudomonas aeruginosa, Serratia maracescens*, as well as several Vibrio species (16) have been shown to activate this system.

*B. pertussis* produces multiple virulence factors involved in immune evasion (17). It was recently shown that virulence associated gene 8 (Vag8) of *B. pertussis* binds to C1-INH (18, 19). Vag8 is a 95 kDa autotransporter. Autotransporters are typically processed into a channel and a passenger domain (20). The passenger domain will pass through the channel and can either remain attached to the bacterial membrane or be secreted into the bacterial surrounding (21). Autotransporter proteins, including Vag8, are also present on the surface of outer membrane vesicles (OMVs) that are secreted by Gram-negative bacteria (18, 22, 23). We have recently shown that secreted Vag8 binding to C1-INH away from the bacterial surface leads to complement evasion. This binding result in consumption of complement components C2 and C4 *via* uncontrolled cleavage by the proteases C1r, C1s, and MASP-2, away from the bacterial surface (18).

Since C1-INH controls both the complement and the contact system, we here investigated whether Vag8 influences contact system activity. We demonstrate that both recombinant and endogenously secreted Vag8 enhanced contact system activity by attenuation of the inhibitory function of C1-INH. Moreover, we show that *B. pertussis* effectively activated the contact system by producing Vag8.

# MATERIALS AND METHODS

# Bacterial Strains and Growth Conditions

*B. pertussis* wild type B1917 strain (isolated in 2000), the isogenic Vag8 knockout strain B1917ΔVag8 (18), the B0442 strain producing a mutated lipooligosaccharide (LOS) that was isolated in 1954 (24) and the pertactin-deficient B4418 and B4374 strains as well as the pertactin-producing B4430 and B4393 strains isolated in 2016 were grown at 35°C, 5% CO2 on Bordet Gengou plates containing glycerol and 15% defibrinated sheep blood (BD Biosciences, Franklin Lakes, NJ, USA). After 3–5 days of culture, the bacteria were collected in buffer containing 50 mM HEPES, 2 mM CaCl2, 50 µM ZnCl2, 0.02% NaN3, and 0.05% Tween-20 (pH 7.35) further referred to as buffer A, the optical density was measured at 600 nm and bacteria were washed in buffer A. OMVs of both strains were prepared by ultracentrifugation as described previously (18, 25, 26).

#### Recombinant Production of Histidine-Tagged (his-tag) Vag8 and the Negative Control *Bordetella* Resistance to Killing A (BrkA) Passenger Domain

Recombinant his-tag passenger domain of Vag8 was produced as previously described (18). BrkA was cloned using primers 5′-ATATGGATCCCAGGAAGGAGAGTTCGAC-3′ and 5′-ATATGCGGCCGCCTACTGCAAGCTCCAGACATG-3′ (restriction sites underlined) and ligated into a modified pRSET-B vector containing a non-cleavable six residue his-tag (MHHHHHHGS) at the N-terminus of the protein as described previously (18, 27). BrkA was expressed and purified as Vag8 (18).

#### Surface Plasmon Resonance (SPR)

SPR was performed using a Biacore T200 (GE Healthcare, Little Chalfont, UK). Recombinant passenger domain of Vag8 was dissolved in 50 mM sodium acetate pH 5.0 and immobilized using primary amine coupling onto a CM5 sensor chip (GE Healthcare). All binding experiments were performed at 25°C in 10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% (v/v) surfactant P20. Increasing concentrations (2.5–160 nM) of either full length C1-INH (C1-INHFL, Complement Technology) or C1-INH containing only the serpin domain (C1-INHNT98) (28) were injected over the flow channels at 30 µL/min. Dissociation was allowed for 300 s followed by surface regeneration with 10 mM glycine pH 2.5. BIAevaluation software (GE Healthcare) was used to analyze the data.

# Size Exclusion Chromatography With Multi-Angle Light Scattering Analysis (SEC-MALS)

100 µL of protein samples were injected onto an S200 increase 10/300 column (GE Healthcare) equilibrated in 50 mM Tris pH 7.5, 150 mM NaCl and eluted with a flow rate of 0.4 mL/min. Light scattering and refractive index changes were measured using a Dawn Heleos-II light scattering detector and an Optilab-TrEX refractive index monitor respectively. Analysis was carried out using ASTRA 6.1.1.175.3.4.14 software assuming a dn/dc value of 0.186 mL/g.

#### Ethics

The study was conducted using blood donation from ±50 healthy adults for plasma collection and according to the principles expressed in the Declaration of Helsinki. Written informed consent was obtained from all blood donors before collection and samples were used anonymously. Approval was obtained from the medical ethics committee of the University Medical Centre Utrecht.

#### Plasma

Blood was collected in blood tubes containing sodium citrate (Vacuette tube, Greiner Bio-one, Kremsmünster, Austria) from ±50 healthy volunteers after written informed consent. Following collection, samples were centrifuged twice at 2,000 *g* for 10 min and plasma of all donors was pooled. The pooled plasma was stored in aliquots at −80°C until use.

# Activation of the Contact System: Chromogenic Substrate Assay

To determine whether Vag8 could interfere with the inhibitory function of C1-INH on βFXIIa and PK, we made use of the chromogenic substrate H-D-Pro-Phe-Arg-pNA (L-2120, Sigma (Merck), Darmstadt, Germany) that can be cleaved by both proteases (29). All experiments were performed in 96-well PVC flat-bottom microplate (Corning GmbH, Wiesbaden, Germany). Activation of the contact system in the presence of Vag8 and BrkA was first analyzed in a purified system. βFXIIa (33.3 nM corresponding to 1 µg/mL, Enzyme Research Laboratories, South Bend, Ind, USA) or PK (5.81 nM corresponding to 0.5 µg/ mL, Enzyme Research Laboratories) was pre-incubated with or without serum-derived C1-INH (95.2 nM corresponding to 10 µg/mL, Complement Technologies, Tyler, TX, USA) that was pre-incubated for 10 min at 37°C with, Vag8, BrkA, buffer A (for βFXIIa), or phosphate buffered saline (PBS) (for PK). Activity was measured following the addition of 0.5 mM L-2120 (Bachem, Bubendorf, Switzerland) (29). Activation of the contact system in a more complex system was studied in citrated human plasma. For these experiments, 60% plasma was activated with βFXIIa (16.7 nM corresponding to 0.5 µg/mL) in the presence of PBS (buffer control), Vag8, BrkA (concentrations indicated in the figures), or OMVs obtained from the *B. pertussis* wild type strain B1917 or knockout strain B1917ΔVag8 (60 µg/mL) with 10 min pre-incubation at 37°C before addition of the chromogenic substrate L-2120. For maximum contact system activity, referred to as control, βFXIIa was added to the plasma at the same time as the addition of L-2120 ensuring that C1-INH did not have the chance to inhibit the contact proteases. The substrate conversion, referred to as kallikrein-like activity, was measured in a kinetic fashion up until 30 min (Figure S1 in Supplementary Material) and reported at the 10 min time point. Substrate conversion was measured with a microplate reader at 405 nm at 37°C over time (VersaMax microplate reader, Molecular Devices, Sunnyvale, CA, USA).

Activation of the contact system by *B. pertussis* was assessed by incubating *B. pertussis* wild type strain B1917 or knockout strain B1917ΔVag8 (3 × 107 CFU each) with 60% plasma or buffer A. Plasma incubated with βFXIIa (16.7 nM corresponding to 0.5 µg/mL, Enzyme Research Laboratories) or only buffer A was taken along as positive and negative controls, respectively. These experiments were performed with 5 min pre-incubation at 37°C before addition of the substrate L-2120. Substrate conversion was measured at 37°C every 30 s with a microplate reader at 405 nm over time (PowerWave XS Microplate Spectrophotometer, BioTek, Winooski, VT, USA).

# Cleavage of HK: Immunoblotting

To determine cleavage of HK in plasma in the presence of Vag8, βFXIIa (16.7 nM corresponding to 0.5 µg/mL) was added to 90% plasma and incubated with either PBS or Vag8 (1,017 nM corresponding to 60 µg/mL) for 10 min at 37°C. Samples were diluted 40 times in reducing sample buffer (15.5% glycerol, 96.8 mM Tris–HCl, 3.1% SDS, 0.003% bromophenol blue, and 25 mM DTT), incubated for 10 min at 100°C and separated on a 10% SDS-PAGE gel. Proteins were blotted onto PVDF membranes. Membranes were blocked with 4% skimmed milk in PBS containing 0.05% Tween (PBS-T) and washed with PBS-T three times for 10 min at 37°C between each incubation step. The immunoblot was subsequently incubated with a primary goat anti-human HK antibody (3 µg/mL final concentration, Affinity Biologicals, Ancaster, ON, Canada) overnight at 4°C and a secondary donkey-anti-goat-HRP antibody (0.5 µg/mL, Southern Biotech, Birmingham, AL, USA) mainly reacting with the HK light chain (30) for 2 h at 37°C. All antibodies were diluted in 1% skimmed milk in PBS-T. For detection, the Pierce ECL Western Blotting Substrate (Thermofisher Scientific, Waltham, MA, USA) was used and visualized using the ImageQuant (GE Life Sciences, Chicago, IL, USA).

For assessment of HK cleavage upon incubation of plasma with *B. pertussis,* several bacterial strains (2 × 109 CFU) were incubated with 50% plasma either alone or in the presence of the contact protease inhibitors aprotinin (31) (100 units/mL, Sigma), which inhibits PK and d-phenylalanyl-prolyl-arginyl chloromethyl ketone (32) (PPACK; 200 µM, Hematologic Technologies, Essex Junction, VT, USA), a multi-target serine protease inhibitor which restricts auto activation and self-digestion of FXIIa, and sampled after 30 min at 37°C shaking at 300 rpm. Plasma samples incubated with βFXIIa (16.7 nM corresponding to 0.5 µg/mL) or only buffer A were taken along as positive and negative controls, respectively. Samples were analyzed as described above.

# Statistical Analyses

Statistical analyses were performed using GraphPad Prism 6.02 and the differences between groups were analyzed for significance using the two-tailed Student's *t*-test. A *p*-value of ≤0.05 was considered statistically significant.

# RESULTS

# Vag8 Attenuates the Role of C1-INH as Inhibitor of **β**FXIIa and PK

It was recently shown that Vag8 binds to C1-INH using both ELISA and gel filtration chromatography (18, 19). In order to further characterize this interaction, we performed the combination of SEC-MALS on the individual proteins and the complex (**Figure 1A**). Both Vag8 and C1-INH behaved as monomers on the column. Addition of Vag8 to C1-INH shifted the elution position and calculation of the mass across the peak revealed that a 1:1 complex had been formed. We next performed SPR analysis to attempt to assess the affinity of the interaction (**Figures 1B,C**). Flowing increasing concentrations of C1-INHFL over Vag8 on the surface demonstrated clear binding. The almost non-existent off-rate of the interaction, combined with a slow enough on-rate that precluded reaching equilibrium, meant that we were unable to robustly calculate the KD of the interaction. However, attempts to fit the kinetics using a variety of binding models always produced KD values of 1 nM or lower, consistent with a tight interaction. In order to confirm that the interaction between Vag8 and C1-INH involved the serpin domain, we repeated the SPR with C1-INH lacking the N-terminal O-linked glycosylation domain. This construct interacted with Vag8 with very similar kinetics to the full length protein (**Figure 1C**).

We have demonstrated that the binding of secreted Vag8 to C1-INH results in attenuation of the inhibitory effect of C1-INH leading to the cleavage of essential complement proteins away from the bacterial surface (18). Since C1-INH is also one of the main inhibitors of the contact system, we hypothesized that the interaction between C1-INH and Vag8 would have a similar effect on the activation of the contact system. To investigate the effect of Vag8 on the contact system, we first studied this in a purified system. Addition of Vag8 to purified βFXIIa and C1-INH resulted in a dose-dependent enhanced conversion of the chromogenic substrate. Addition of 10 µg/mL Vag8 (169.5 nM) is sufficient to completely neutralize C1-INH activity as comparable levels of activation as that of βFXIIa alone were reached (**Figure 2A**). This dose-dependent attenuation of inhibition by C1-INH was also observed upon the addition of Vag8 to a purified system containing PK and C1-INH (**Figure 2B**). A concentration of 12.5 µg/mL Vag8 (203.4 nM) was sufficient to completely block the inhibitory capacity of C1-INH on PK activation. As a negative control we used BrkA. BrkA is another autotransporter protein of *B. pertussis* involved in complement evasion. BkrA has a similar structure and size as Vag8 and was produced in a similar way as Vag8 (33, 34). The inhibitory properties of C1-INH on βFXIIa and PK were not disturbed by the addition of BrkA at equimolar concentrations (Figure S2 in Supplementary Material). Taken together, Vag8 dose dependently attenuates the inhibitory effect of C1-INH on βFXIIa and PK in a purified system.

#### Vag8 Enhances Contact System Activity in Plasma

Since we have shown that Vag8 can interfere with the regulatory activity of C1-INH on βFXIIa and PK in a purified system, we next assessed the effect of Vag8 in a more physiological setting. To this end, we incubated plasma either with buffer (pre-incubated plasma) or increasing concentrations of Vag8 in the presence of the contact system activator βFXIIa for 10 min before addition of the chromogenic substrate. This pre-incubation step is needed for C1-INH to inhibit the contact proteases βFXIIa and PK. Moreover, the chromogenic substrate was added immediately following βFXIIa addition (control plasma) indicating the

maximum kallikrein-like activity. **Figure 3A** shows that Vag8 dose-dependently attenuates the inhibitory function of C1-INH thus enhancing the activation of the contact system, referred to as kallikrein-like activity. Comparable levels as control plasma are reached upon using 7.5 µg/mL of Vag8 (127.1 nM), with no significant difference between the control plasma and addition of 30 µg/mL of Vag8 (508.5 nM). The negative control, BrkA, did not show any effect on contact system activity (data not shown). To determine whether endogenously secreted Vag8 is also capable of mediating this effect, OMVs derived from *B. pertussis* wild type strain B1917 or the knockout strain B1917ΔVag8 (18) were incubated with plasma and βFXIIa. We show that the OMVs derived from *B. pertussis* wild type strain B1917 were capable of attenuating the inhibitory function of C1-INH resulting in enhanced contact system activity to levels observed in control plasma conditions, whereas the OMVs obtained from the knockout strain B1917ΔVag8 were not (**Figure 3B**).

As previously mentioned, activation of the contact system ultimately results in the cleavage of HK and the subsequent release of BK. Also clinically, cleavage of HK is related to BK production (30). To investigate the effect of recombinant Vag8 binding to C1-INH on HK cleavage in plasma, βFXIIa was incubated with plasma in the presence or absence of Vag8 and was assessed by immunoblotting to visualize HK cleavage. **Figure 3C** shows increased HK cleavage as indicated by the decreased intensity of the ~120 kDa full length HK band (30) in the presence of Vag8 compared to incubation with βFXIIa alone as well as by the detection of the ~46 kDa cLC-HK chain. This is the cleaved L-chain which is often used as a marker of extensive contact system activation in plasma (35). In conclusion, we show enhanced contact system activity in the presence of recombinant and endogenous Vag8, which is most likely due to Vag8 binding to C1-INH and hence hampering the inhibitory properties of C1-INH on βFXIIa and PK.

#### *B. pertussis* Activates the Contact System Predominantly Through Vag8 Production

Although several bacteria are known to activate the contact system (15, 36–42), the interaction between *B. pertussis* and the contact system has not been investigated. As shown above, Vag8 of *B. pertussis* hampers the inhibition of the contact proteases by binding C1-INH and hence enhances contact system activity (**Figures 2** and **3**). Next, we investigated whether *B. pertussis* itself can effectively activate the contact system. We show that *B. pertussis* wild type strain B1917 successfully activates the contact system in plasma as an increase in kallikrein-like activity was observed over time (**Figure 4A**). This activation was further examined by assessment of HK cleavage using immunoblot. **Figure 4B** shows a representative immunoblot in which degradation of the ~120 kDa full length HK and appearance of a ~46 kDa cLC-HK chain (30) was observed when *B. pertussis* strain B1917 was added to the plasma. To verify that the observed HK cleavage was the result of contact system activation, aprotinin (31) and PPACK (32) were added to the B1917 samples to prevent FXIIa and PK activity. **Figure 4B** shows a lack of HK cleavage upon addition of both these inhibitors to plasma incubated with *B. pertussis* wild type strain B1917 indicating that the HK cleavage observed in the presence of this bacterium can be attributed to the activation of the contact system. Next, we assessed whether Vag8 production was responsible for the activation of the contact system by *B. pertussis.* **Figure 4C** shows the lack of HK cleavage in the presence of the *B. pertussis* knockout strain B1917ΔVag8 by immunoblot. Even after 180 min of incubation, no activation of the contact system by the *B. pertussis* knockout strain B1917ΔVag8 was observed (data not shown). This is further supported by the decreased kallikrein-like activity shown upon incubating plasma with the *B. pertussis* knockout strain B1917ΔVag8 when compared to incubation with wild type strain B1917 (**Figure 4D**).

mean ± SEM of three separate experiments, while panels B and C are representative of three separate experiments. \**p* ≤ 0.05, \*\**p* ≤ 0.01, \*\*\**p* ≤ 0.001, \*\*\*\**p* ≤ 0.0001, ns, non-significant compared to the black bar.

Moreover, we show that contact system activation is not restricted to *B. pertussis* strain B1917. Incubation of plasma with the LOSmutant B0442 as well as clinical strains either producing (B4430 and B4393) or not producing pertactin (B4418 and B4374) also shows contact system activation as indicated by the detection of the ~46 kDa cLC-HK chain (**Figure 5**). In summary, *B. pertussis* is capable of activating the contact system as demonstrated by the observed kallikrein-like activity as well as by the cleavage of HK detected by immunoblot. We show that this is mainly dependent on the production of the autotransporter Vag8, which will bind to C1-INH and thus attenuate its inhibitory function, as the *B. pertussis* knockout strain B1917ΔVag8 showed no HK cleavage and decreased kallikrein-like activity compared to its isogenic wild type strain B1917.

# DISCUSSION

Recently, we unraveled the mechanism responsible for secreted Vag8-mediated complement evasion (18). We showed that binding of secreted Vag8 to C1-INH resulted in the release of the active proteases C1s, C1r, and MASP-2 since C1-INH could no longer bind and thus inhibit their proteolytic activity. The presence of active C1s, C1r, and MASP-2 proteases in the serum resulted in the degradation of the complement proteins C4 and C2 away from the bacterial surface (18). We suggest that *B. pertussis* uses this complement evasion strategy to aid in the prevention of opsonization and complement-mediated lysis. Vag8 is not only secreted but also present on the bacterial surface where it binds to C1-INH (19, 43). Binding of C1-INH to the bacterial surface could also result in the inhibition of complement activation which has been shown for other bacteria such as *Borrelia recurrentis* (44). The contact system is another innate immune component consisting of proteases (11). Here, we show that in addition to interacting with the complement system (18), Vag8 of *B. pertussis* induced enhanced activation of the contact system as demonstrated by increased contact system activity and HK cleavage. We propose that Vag8 mediates activation of the contact pathway by binding to C1-INH and attenuating its inhibitory function as we have previously shown for its effect on the complement system (18). C1-INH is the predominant fluid phase inactivator of FXIIa and PK. This serpin inactivates these proteases by irreversibly binding to them resulting in conformational changes that disrupt the active site of the target proteases (13). This process is hampered in the presence of Vag8, which we expect to bind to C1-INH, and consequently interfere with the protease inhibition allowing for enhanced activation of the contact system (illustrated in **Figure 6**). As activation of the contact system by bacteria is often induced by membrane-bound components (16), **Figure 6** depicts the activation of the contact system on the bacterial surface. However, we cannot exclude that activation occurs both on the surface as well as in fluid phase. Our results indicate that blockage of protease inhibition is essential for *B. pertussis* wild type strain B1917 induced activation of the contact system as in the absence of Vag8, C1-INH is free to inhibit the contact system proteases and, therefore, activation of the contact system is not induced (**Figure 3**). Whether *B. pertussis* can also interact with the PK inhibitor alpha-2-macroglobulin (45) remains to be investigated.

Here we show for the first time that *B. pertussis* can activate the contact system, which as a bacterium is not unique (15, 36–42). Other bacteria have been shown to activate this system *via* polyphosphates, which are present on *E. coli*, *Vibrio cholerae*,

Figure 4 | *Bordetella pertussis* activates the contact system mainly by Virulence associated gene 8 (Vag8) production. (A) *B. pertussis* wild type strain B1917 (3 × 107 CFU) induces similar kallikrein-like activity in 60% plasma compared to addition of 0.5 µg/mL βFXIIa. (B) Contact system activation by *B. pertussis* wild type strain B1917 was also shown by immunoblot. 50% plasma was incubated for 60 min with *B. pertussis* wild type strain B1917 (2 × 109 CFU) alone or in combination with the contact system inhibitors d-phenylalanyl-prolyl-arginyl chloromethyl ketone (PPACK) or aprotinin and analyzed using anti-high-molecular-weight-kininogen (HK). Incubation with *B. pertussis* wild type strain B1917 resulted in almost complete cleavage of the ~120 kDa full length HK as indicated by the appearance of the ~46 kDa cLC-HK chain. This feature was not observed in the presence of the inhibitors. (C) To determine whether Vag8 production was involved in contact system activation by this pathogen, 50% plasma was incubated either with *B. pertussis* wild type strain B1917 or with the knockout strain B1917ΔVag8 (2 × 109 CFU) and analyzed using anti-HK. The *B. pertussis* knockout strain B1917ΔVag8 was unable to activate the contact system as no cleavage of the ~120 kDa full length HK was detected. (D) This was also shown by the decreased kallikrein-like activity in 60% plasma upon incubation with the knockout strain B1917ΔVag8 (3 × 107 CFU) compared to the *B. pertussis* wild type strain B1917. Panels A–D are representative of three separate experiments.

*Corynebacterium diphtheria*, and *Haemophilus influenzae* but also on *B. pertussis* (46, 47). Moreover, LPS present on Gramnegative bacteria have been implicated in the activation of the contact system *in vitro* (48, 49). Contact system proteins were furthermore shown to assemble on the bacterial surfaces of *Salmonella typhimurium* and *E. coli* resulting in the release of BK (15). An increase in BK at the site of infection may cause leakage of plasma and be beneficial for the bacteria as this will provide bacteria with nutrients (16). Nonetheless, BK triggers endothelial permeability resulting in infiltration of leukocytes (11). Moreover, cleavage of HK results in the generation of the antimicrobial peptide NAT-26 which could drive bacterial killing (38). Hence, whether activation of the contact system is beneficial or detrimental for the bacteria remains unclear. Contact system protein assembly on *E. coli* occurs on curli pili (42). As we do not observe HK cleavage in the presence of the *B. pertussis* knockout strain B1917ΔVag8, we expect that although the *B. pertussis* membrane-associated polyphosphates or LOS may trigger the contact system, attenuation of C1-INH *via* sequestration by Vag8 is essential for full activation of this system. Alternatively, bacteria can express proteases that actively cleave contact system proteins such as staphopains of *Staphylococcus aureus* or streptokinase of *S. pyogenes,* which both cleave HK releasing BK (38, 50). *B. pertussis* is unique in enhancing the activation of the contact system by producing a protein, Vag8, which inhibits the inhibitory function of the contact system regulator C1-INH.

B4393, and B4374 was shown by immunoblot. 50% plasma was incubated for 60 min with the *B. pertussis* strains (2 × 109 CFU) and analyzed using anti-high-molecular-weight-kininogen (HK). Incubation with all the *B. pertussis* strains resulted in cleavage of the ~120 kDa full length HK as indicated by the appearance of the ~46 kDa cLC-HK chain.

mediated activation of the contact system. Vag8, either on the bacterial surface as part of an outer membrane vesicles (OMVs) or as the secreted passenger, binds to complement regulator C1 inhibitor (C1-INH) (left panel). This results in the lack of inhibition of the contact system proteases FXIIa and plasma kallikrein (PK) by C1-INH. The lipooligosaccharide and polyphosphates present on the outer membrane of *Bordetella pertussis* are most likely responsible for FXII activation on the bacterial membrane as has been shown for other bacteria (15, 36–42). This activation, which cannot be inhibited by C1-INH as it is bound to Vag8, will result in the release of bradykinin (BK). In the absence of Vag8, C1-INH will inhibit FXIIa and PK when formed and high-molecular-weight-kininogen (HK) will remain intact away from the bacterial surface (right panel).

Infection with *B. pertussis* results in the disease whooping cough, which is typically associated with fits of coughs (or paroxysms) followed by a typical high-pitched whoop. These coughing fits generally persist weeks after the bacterium has been cleared and contribute greatly to the morbidity caused by this disease (51). To date, it is not completely understood what causes this type of cough. The persistence of a chronic cough in the absence of a stimulus is not unique to pertussis but has also been observed in patients on angiotensin converting enzyme inhibitors (ACEi) that are being treated for hypertension (32). ACE, which is produced by lung endothelial cells, breaks down BK (52). The cough associated with ACEi often remains for several days or weeks after the patients have withdrawn from taking the drug. Although the mechanism of ACEi-induced cough remains unresolved, there are indications that BK, of which the levels are increased during ACEi treatment, might be involved (53). The contact system is not only present in plasma but also in the lungs (54) and respiratory administration of BK to guinea pigs but also humans evokes a paroxysmal cough much like the cough associated with pertussis (55). In guinea pigs, this cough was shown to be induced by BK activating the B2 receptors on the bronchopulmonary C-fibers (56). These receptors are also expressed in humans on epithelial cells, fibroblasts, and endothelial cells of the bronchial lamina propria (57). It is likely that *B. pertussis* infection results in increased levels of BK in the lungs as we have shown HK cleavage in the presence of *B. pertussis* wild type strain B1917 and BK is a cleavage product of HK. Consequently, we speculate that *B. pertussis*-induced activation of the contact system may be involved in the induction of pertussis-specific cough and thus transmission. Moreover, the activation of the contact system may also play a role in lung pathology. Lung lesions caused by an infection with *S. typhimurium* were shown to be prevented upon inhibition of the contact system in a rat model (40). Fatal *B. pertussis* infection is also characterized by lung lesions (58) and hence the increased BK levels in the lungs following an infection with *B. pertussis* may contribute to lung pathology. Further research needs to be conducted to really understand the role of the activation of the contact system on pertussis pathology.

In light of the re-emergence of pertussis, it has become evident that the development of a novel pertussis vaccine is necessary (5). One of the potential proteins that could be included in such a vaccine is Vag8. Vaccination with Vag8, which was previously only known as a complement evasion molecule of *B. pertussis*, has been shown to give rise to antibodies which protect mice from infection following a *B. pertussis* challenge (59). Moreover, Vag8-specific antibodies have been detected in pertussis patients indicating that Vag8 is produced by *B. pertussis* during human infection (60). Next to the proposed possibility of including Vag8 in a novel acellular pertussis vaccine, this protein is also a component of the OMV-based pertussis vaccine and the live attenuated BPZE1 vaccine that are currently being investigated (61, 62). Vag8 is highly present on OMVs of *B. pertussis* making up 34–50% of the total OMV proteins and can also be found on the bacterial membrane (18, 20, 23). Due to Vag8's high abundance on OMVs, presence on the outer membrane of *B. pertussis* and the protective effect of this protein as a potential vaccine antigen, the overactivation of the contact system described here, together with the degradation of essential complement protein described earlier (18) may have implications for the inclusion of Vag8 in novel pertussis vaccines. These side effects could include local C1-INH deficiency with consequences for complement and contact system mediated adverse reactions (63). Inhibition of C1-INH could result in increased BK formation which may mediate increased inflammation at the site of vaccination. Moreover, consumption of complement proteins resulting in decreased complement activation following vaccination could have a negative effect on the induction of memory B-cells which are normally induced *via* interactions between C3d-tagged microorganisms or immune-complex antigens and complement receptor 2 (64, 65). It may be advisable to modify Vag8 before the potential inclusion of this antigen in novel pertussis vaccines in order to avoid side effects that could be induced by binding of Vag8 to C1-INH.

In conclusion, we show that Vag8 enhances contact system activity and is mainly responsible for the observed activation of the contact system induced by *B. pertussis.* We propose that this is the result of C1-INH binding by Vag8. This potent C1-INH inhibitor secreted by *B. pertussis* not only mediates complement evasion but also an overlooked interaction between the contact system and the respiratory pathogen *B. pertussis* that may contribute to its pathogenicity and virulence*.*

#### ETHICS STATEMENT

The study was conducted using blood donation from healthy adults for plasma collection and according to the principles expressed in the Declaration of Helsinki. Written informed consent was obtained from all blood donors before collection and samples were used anonymously. Approval was obtained from the medical ethics committee of the University Medical Centre Utrecht.

#### AUTHOR CONTRIBUTIONS

EH, SM, AC, SJ, and IJ performed the experiments. EH and SJ drafted the figures. IJ, SM, and CM were involved in the study design. EP and IJ were responsible for funding. EH, SJ, and IJ wrote the manuscript. All authors approved the manuscript's final version.

#### REFERENCES


#### FUNDING

We thank Dorina Roem (Sanquin, The Netherlands) for C1-INHNT98. This work is part of the research program VENI with project number 016.156.051 awarded to IJ, which is (partly) financed by the Netherlands Organization for Scientific Research (NWO). In addition, this project was supported by the Dutch Ministry for Public Health and the Environment (RIVM) in the framework of the S/112001 grant. CM gratefully acknowledges financial support from the Landsteiner Foundation for Blood Transfusion Research and the Netherlands Thrombosis Foundation. Finally, SJ was supported by the Welcome Trust Investigator Award (100298) awarded to Prof. S.M. Lea.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01172/ full#supplementary-material.

Figure S1 | Substrate assay kinetics. Substrate conversion, referred to as kallikrein-like activity, was measured in a kinetic fashion up until 30 min. (A) 1 µg/mL βFXIIa was incubated alone or with 10 µg/mL complement regulator C1 inhibitor (C1-INH) alone or in combination with different concentrations of virulence-associated gene 8 (Vag8) (2.5, 5, 7.5, 10, 12.5, 15, and 30 µg/mL). (B) 0.5 µg/mL plasma kallikrein (PK) was incubated alone or with 10 µg/mL C1-INH alone or in combination with different concentrations of Vag8 (2.5, 5, 7.5, 10, 12.5, 15, and 30 µg/mL). (C) 60% plasma was incubated with Vag8 (5, 7.5, 15, and 30 µg/mL) following 10 min pre-incubation with 0.5 µg/mL βFXIIa. Incubation of plasma with βFXIIa alone either without or with pre-incubation served as a control. In the main text, substrate conversion is reported at the 10 min time point. Data represent the mean of three separate experiments.

Figure S2 | The negative control BrkA does not interfere with the inhibition of βFXIIa and plasma kallikrein (PK) by complement regulator C1 inhibitor (C1-INH). (A) Bordetella resistance to killing A (BrkA) (30 µg/mL) has no effect on the kallikrein-like activity of 1 µg/mL βFXIIa in the presence of 10 µg/mL C1-INH when compared to 1 µg/mL βFXIIa in the presence of 10 µg/mL C1-INH alone. (B) BrkA (30 µg/mL) has no effect on the kallikrein-like activity of 0.5 µg/mL PK in the presence of 10 µg/mL C1-INH when compared to 0.5 µg/mL PK in the presence of 10 µg/mL C1-INH alone. Data represent the mean ± SEM of three separate experiments.


vaccine – BPZE1; a single centre, double-blind, placebo-controlled, dose-escalating study of BPZE1 given intranasally to healthy adult male volunteers. *PLoS One* (2014) 9(1):e83449. doi:10.1371/journal.pone.0083449


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Hovingh, de Maat, Cloherty, Johnson, Pinelli, Maas and Jongerius. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Complement Component C1q as Serum Biomarker to Detect Active Tuberculosis

Rosalie Lubbers <sup>1</sup> , Jayne S. Sutherland<sup>2</sup> , Delia Goletti <sup>3</sup> , Roelof A. de Paus <sup>4</sup> , Coline H. M. van Moorsel <sup>5</sup> , Marcel Veltkamp<sup>5</sup> , Stefan M. T. Vestjens <sup>6</sup> , Willem J. W. Bos 6,7 , Linda Petrone<sup>3</sup> , Franca Del Nonno<sup>8</sup> , Ingeborg M. Bajema<sup>9</sup> , Karin Dijkman<sup>10</sup> , Frank A. W. Verreck <sup>10</sup>, Gerhard Walzl <sup>11</sup>, Kyra A. Gelderman<sup>12</sup>, Geert H. Groeneveld<sup>4</sup> , Annemieke Geluk <sup>4</sup> , Tom H. M. Ottenhoff <sup>4</sup> , Simone A. Joosten<sup>4</sup> \* † and Leendert A. Trouw13†

#### Edited by:

*Thomas Vorup-Jensen, Aarhus University, Denmark*

#### Reviewed by:

*Anthony George Tsolaki, Brunel University London, United Kingdom Uday Kishore, Brunel University London, United Kingdom*

#### \*Correspondence:

*Simone A. Joosten s.a.joosten@lumc.nl*

*†These authors have contributed equally to this work*

#### Specialty section:

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

Received: *31 May 2018* Accepted: *02 October 2018* Published: *23 October 2018*

#### Citation:

*Lubbers R, Sutherland JS, Goletti D, de Paus RA, van Moorsel CHM, Veltkamp M, Vestjens SMT, Bos WJW, Petrone L, Del Nonno F, Bajema IM, Dijkman K, Verreck FAW, Walzl G, Gelderman KA, Groeneveld GH, Geluk A, Ottenhoff THM, Joosten SA and Trouw LA (2018) Complement Component C1q as Serum Biomarker to Detect Active Tuberculosis. Front. Immunol. 9:2427. doi: 10.3389/fimmu.2018.02427* *<sup>1</sup> Department of Rheumatology, Leiden University Medical Center, Leiden, Netherlands, <sup>2</sup> Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, Banjul, Gambia, <sup>3</sup> Translational Research Unit, Department of Epidemiology and Preclinical Research, National Institute for Infectious Diseases, Rome, Italy, <sup>4</sup> Department of Infectious Diseases, Leiden University Medical Center, Leiden, Netherlands, <sup>5</sup> Department of Pulmonology, St. Antonius Hospital Nieuwegein, Nieuwegein, Netherlands, <sup>6</sup> Department of Internal Medicine, St. Antonius Hospital Nieuwegein, Nieuwegein, Netherlands, <sup>7</sup> Department of Nephrology, Leiden University Medical Center, Leiden, Netherlands, <sup>8</sup> Pathology Service, National Institute for Infectious Diseases, Rome, Italy, <sup>9</sup> Department of Pathology, Leiden University Medical Center, Leiden, Netherlands, <sup>10</sup> Section of TB Research & Immunology, Biomedical Primate Research Centre, Rijswijk, Netherlands, <sup>11</sup> Division of Molecular Biology and Human Genetics, Faculty of Medicine and Health Sciences, DST/NRF Centre of Excellence for Biomedical Tuberculosis Research, South African Medical Research Council Centre for Tuberculosis Research, Stellenbosch University, Cape Town, South Africa, <sup>12</sup> Sanquin Diagnostic Services, Amsterdam, Netherlands, <sup>13</sup> Department of Immunohematology and Blood Transfusion, Leiden University Medical Center, Leiden, Netherlands*

Background: Tuberculosis (TB) remains a major threat to global health. Currently, diagnosis of active TB is hampered by the lack of specific biomarkers that discriminate active TB disease from other (lung) diseases or latent TB infection (LTBI). Integrated human gene expression results have shown that genes encoding complement components, in particular different C1q chains, were expressed at higher levels in active TB compared to LTBI.

Methods: C1q protein levels were determined using ELISA in sera from patients, from geographically distinct populations, with active TB, LTBI as well as disease controls.

Results: Serum levels of C1q were increased in active TB compared to LTBI in four independent cohorts with an AUC of 0.77 [0.70; 0.83]. After 6 months of TB treatment, levels of C1q were similar to those of endemic controls, indicating an association with disease rather than individual genetic predisposition. Importantly, C1q levels in sera of TB patients were significantly higher as compared to patients with sarcoidosis or pneumonia, clinically important differential diagnoses. Moreover, exposure to other mycobacteria, such as *Mycobacterium leprae* (leprosy patients) or BCG (vaccinees) did not result in elevated levels of serum C1q. In agreement with the human data, in non-human primates challenged with *Mycobacterium tuberculosis*, increased serum C1q levels were detected in animals that developed progressive disease, not in those that controlled the infection.

**280**

Conclusions: In summary, C1q levels are elevated in patients with active TB compared to LTBI in four independent cohorts. Furthermore, C1q levels from patients with TB were also elevated compared to patients with sarcoidosis, leprosy and pneumonia. Additionally, also in NHP we observed increased C1q levels in animals with active progressive TB, both in serum and in broncho-alveolar lavage. Therefore, we propose that the addition of C1q to current biomarker panels may provide added value in the diagnosis of active TB.

Keywords: complement, tuberculosis, C1q, infection, innate immunity, blood, mycobacterium

#### INTRODUCTION

Tuberculosis (TB) is a major global health threat, which is caused by infection by Mycobacterium tuberculosis (M.tb) (1). Current estimations indicate that a quarter of the global population is infected with M.tb, with a life-long risk to develop active TB disease. Particular regions, such as South-East Asia, Western-Pacific, and Africa regions account for more than 80% of infected individuals (2). Annually over 6 million people are diagnosed with TB disease, a serious and highly contagious condition, resulting in 1.3 million deaths in 2016 only (1). While most infected individuals remain asymptomatic latently infected (LTBI), a minority (5–10%) of these individuals progress to active TB. Given the high rate of infections with M.tb in some regions it is important to discriminate infection from disease, which is difficult with the currently available tests. At present, only M.tb detection in sputum using smear, PCR or culture is definitive proof of TB disease. Early diagnosis and treatment of TB disease is important to reduce transmission of infection and prevent disease associated mortality (1).

Diagnosis of active TB is made by microbiological or genetic detection of M.tb in sputum (or other specimens in case of extrapulmonary TB), but this can be expensive and timeconsuming depending on bacterial burdens or requires complex methodology and infrastructure. Current immunological tests can detect infection with M.tb but often fail to discriminate active disease from latent infection (3). Therefore, there is an urgent need to identify biomarkers that can discriminate active and latent TB infection in order to promptly initiate treatment to prevent mortality and further spread of the pathogen, in particular in areas where M.tb is highly endemic. Ideally, such biomarkers should also be able to discriminate between TB and other respiratory infections that present with similar symptoms and abnormalities on chest X-rays.

Many studies have identified potential biomarkers that discriminate active TB from LTBI, or that are predictive of which individuals will progress to active TB (4–10). Differential gene expression profiles between patients with TB and LTBI or other (lung) diseases resulted in identification of an array of potential biomarkers, such as FCGR1A (11–13) and GBP5 (5, 6, 14). Recently, complement has been highlighted as candidate biomarker for active TB disease (15–19) also in the presence of HIV co-infection (20). Most currently identified biomarkers have been identified at the transcriptomic level; however, easy, robust markers that can be measured at the protein level would be more ideal candidates for application in the field. Therefore, validation of markers previously identified at the mRNA level at the protein level would provide important insights into the applicability of such markers in clinical practice.

Next to biomarker studies in humans, there are various experimental models to study the host-pathogen interaction. The best available model is the non-human primate model (NHP), infection of rhesus macaques with M.tb, resulted in TB disease which closely resembles human TB as they experience similar lesions and clinical courses as humans, suggesting a common pathophysiology (21). The NHP model adds important information on kinetics of disease development following M.tb infection and can be manipulated with e.g., different dosages of infection, different infecting strains, but also with vaccines.

The complement system is an important part of the innate immune system and functions as a proteolytic cascade. The classical complement pathway is initiated by binding of C1q to ligands, such as immune complexes. Following the binding of C1q to ligands, enzymatic processes lead to the release of inflammation stimulating peptides, C3a and C5a, formation of the opsonin C3b and formation of the membrane attack complex, resulting in target cell lysis (22). Furthermore, C1q can bind several receptors that contribute to other important functions, such as phagocytosis or myeloid cell modulation, outside traditional complement system activation (23, 24). For instance, C1q is involved in neovascularization during pregnancy, coagulation processes and neurological synapse function (25). C1q is a 480 kDa protein composed of six arms, each comprising one A, B and C peptide chain (26). These three chains are encoded by individual genes, C1QA, C1QB, and C1QC, located on chromosome 1p. In contrast to most complement proteins, C1q is not produced by hepatocytes but mainly by monocyte derived cells, such as macrophages and immature dendritic cells (27–29) and by mast cells (30). Increased expression of mRNA for C1q has been associated with TB disease (16, 19).

Here, we analyzed differential expression of complement genes in patients with TB. Since in publicly available datasets, C1q expression was most pronouncedly upregulated, we validated C1q at the protein level in samples from various patient groups as a biomarker for active TB. Patients with active TB disease were compared to latently infected individuals, vaccinnees and to patients with clinical conditions that are important differential diagnoses in clinical practice. Finally, to obtain more insight in the pathophysiology, kinetic analyses were performed samples obtained from NHP animal models of TB.

### MATERIALS AND METHODS

# Patients and Controls

Demographic data and classification of the cohorts are presented in **Table 1**. Below we have specified the specific inclusion criteria per cohort.

#### Tuberculosis

Smear, GeneXpert or sputum culture positive pulmonary TB patients, LTBI patients and treated TB patients as well as endemic controls (in different combinations) were included from various demographic locations: Italy (31), the Gambia, Korea and South Africa (**Table 1**). Patients with active pulmonary TB disease, referred to as "TB" in the manuscript, were diagnosed based on local, routine methodology. Active pulmonary TB was sputum-culture confirmed (BACTECTM, Becton-Dickinson, USA), or based on positive Xpert Mtb/RIF assay (Cephaid Inc., Sunnyvale, CA, USA), patients were included within 7 days of TB treatment initiation. Latent TB infection, LTBI, was determined by Quantiferon TB Gold-in tube positivity (Qiagen, The Netherlands).

In the cohort from South Africa people suspected for TB were used as controls, these people were presenting with symptoms


*In the table the different cohorts are described regarding the country of origin of the samples, the disease classification, and the total number of samples per group as well as the demographic info on age and sex.*

\**Nepal, Brasil, Ethiopia.*

*† Diagnosis made in the Netherlands.* compatible with active TB but had negative X-ray and negative sputum cultures for TB. These suspected TB patients were seen again after 2 months and had recovered spontaneously or after appropriate (non-TB related) treatment. All TB patients were HIV-negative, as were the endemic controls. Additionally, from Italy both LTBI (QuantiFERON TB Gold-In-tube-positive individuals) and successfully treated TB patients (2–72 months after end of therapy) were included. TB patients from the Gambia were followed over time (1, 2, and 6 months after diagnosis), until completion of successful treatment.

#### Other Mycobacterial Diseases and Vaccination

Leprosy patients (mostly immigrants with mixed ethnic backgrounds) at diagnosis of primary leprosy were included in the Netherlands. In addition, patients with type-1 reactions were enrolled in Brazil, Nepal and Ethiopia. Furthermore, we measured C1q in healthy Dutch individuals who were vaccinated with BCG Danish strain 1331 (Statens Serum Institut, Denmark) and followed over time (32).

#### Other Pulmonary Diseases

Patients with community acquired pneumonia were included in the Netherlands, one cohort comprised patients admitted to the intensive care unit in a tertiary care hospital in Leiden (one patient was HIV-infected with a normal CD4 count and one suffered from sarcoidosis) and the other cohort comprised patients admitted to a hospital ward of a non-academic teaching hospital in Nieuwegein. From both groups of pneumonia patients paired samples from the time of diagnosis and after recovery (10–124 days later) were available. Finally, samples were included prior to initiation of treatment from sarcoidosis patients in the Netherlands that had pulmonary involvement.

#### Additional Control Group

As a reference group we included a panel of Dutch healthy controls, not suffering from major infections or autoimmune disease.

#### Ethics Statement

Blood was obtained from individuals upon signing an informed consent. All studies comply with the Helsinki declaration. The use of the samples in this study was approved by local ethical committees. For Italy, Ethical Committee of the L. Spallanzani National Institute of Infectious diseases (02/2007 and 72/2015); The Gambia (SCC1333); Korea, Institutional Review Board for the Protection of Human Subjects at YUHS; South Africa, Health Research Ethics Committee of the Faculty of Medicine and Health Sciences at Stellenbosch University (N13/05/064); Brazil, National Council of Ethics in Research and UFU Research Ethics Committee (#499/2008); Nepal, Health Research Council (NHR#751); Ethiopia, Health Research Ethical Review committee Ethiopia (NERC#RDHE/127-83/08); The Netherlands leprosy patients, (MEC-2012-589); sarcoidosis patients, Medical research Ethics Committees United of the St Antonius (#R05.08A); BCG vaccinated individuals, Leiden University Medical Center Ethics Committee (P12.087); control group, (P237/94); pneumonia Leiden (P12.147); pneumonia Nieuwegein (C-04.03 and R07.12).

#### Gene Expression Analysis

Global transcriptomic analyses have been performed to compare patients with active TB disease with latently infected individuals. In addition, transcriptomes from patients with TB disease were compared with patients with other diseases, such as sarcoidosis, pneumonia or lung cancer (5–7, 33–37). Microarray data from these studies, publically available in Gene Expression Omnibus (GEO) (GSE37250, GSE19491, GSE39941, GSE28623, GSE73408, GSE34608, GSE42834, GSE83456), were retrieved from GEO and re-analyzed. Several of the studies described multiple independent cohorts, which we analyzed separately for each population [from Malawi (2), South Africa (3), United Kingdom (2), Kenya (only TB vs. LTBI), The Gambia (only TB vs. LTBI), USA and Germany (only TB vs. other diseases)]. All data were extracted from GEO and compared in the same way using GEO2R, thus not relying on the analysis performed in the original manuscript. GEO2R compared two or more groups of samples in order to identify genes that are differentially expressed across experimental or clinical conditions. Here, lists of differentially expressed genes between TB and LTBI or TB and other diseases (with a significance of p > 0.05 and a factorial change of >2 or <0.5) were generated. A list of complement genes and two reference genes was used to assess possible differential expression for each individual gene for each study population. All studies/populations with significant differential expression for a particular gene between patients with TB compared to LTBI/other diseases were enumerated and expressed as the percentage of the total number of comparisons investigated. E.g., differential expression of C1QA between TB and LTBI was observed in only 2/9 populations investigated (22%) whereas C1QB differed in 8/9 populations (88%). For all populations with significant differential expression, the factorial change (the difference between gene expression in TB patients compared to LTBI/other diseases) was calculated and plotted.

#### Animals

Non-human primate (NHP) serum was available from a biobank of samples collected from earlier TB studies in healthy, purposebred rhesus macaques (Macaca mulatta) for which ethical clearance was obtained from the independent ethical authority according to Dutch law. All housing and animal care procedures were in compliance with European directive 2010/63/EU, as well as the "Standard for Humane Care and Use of Laboratory Animals by Foreign Institutions" provided by the Department of Health and Human Services of the US National Institutes of Health (NIH, identification number A5539-01). Longitudinal banked serum samples were available for C1q analysis. Animals were non-vaccinated or vaccinated with BCG [BCG Danish 1331 (Statens Serum Institute, Denmark)] (n = 23) and experimentally infected via bronchoscopic instillation of 500 CFU of M. Erdman. Prior to infection and 3, 6, 12, 24, 36, and 52 weeks post-infection samples were collected and stored. Animals were sacrificed when reaching early humane endpoints (acute progressors) or after reaching the pre-defined end-point (52 weeks; non-progressors). Broncho-alveolar lavage (BAL) samples were available from six animals infected with 1–7 CFU of M. Erdman prior to infection and 6 or 12 weeks post-infection.

# Detection of C1q by ELISA

C1q levels in sera were measured using an in-house developed ELISA. Maxisorp plates (nunc) were coated overnight with mouse anti-human C1q (2204) (38), Nephrology department, LUMC) in coating buffer (0.1 M Na2CO3, 0.1 M NaHCO3, pH9.6). Plates were washed and blocked with PBS/1%BSA for 1 h at 37◦C. After washing, a serial dilution of a pool of normal human serum (NHS) was applied as a standard and samples were added in dilution to the plate, all in duplicate, and incubated for 1 h at 37◦C. Human sera were diluted 1:8,000 and NHP sera 1:4,000, the BAL fluids were diluted 1:1. After washing, plates were incubated with rabbit anti-human C1q (Dako cat#A0136) for 1 h at 37◦C and for detection a goat anti-rabbit HRP (Dako cat#P0448) was used which was also incubated for 1 h at 37◦C. All washing steps were performed with PBS/1%BSA/0.05%Tween. Plates were stained using ABTS and measured at absorbance of 415 nm, the measured C1q is expressed in µg/mL as compared to a C1q standard.

### Immunohistochemical Staining of Lung Tissue for C1q

Samples were collected at autopsy from patients with fatal TB disease (n = 3), fatal pneumonia patients (n = 4) and a control that died of vascular disease (n = 1) at the National Institute for Infectious Diseases, Rome, Italy under local ethical approval (72/2015). Paraffin sections of 4µm thickness were subjected to heat-induced antigen retrieval using tris/EDTA (pH 9.0) at 96◦C for 30 min, and then stained with rabbit anti-human C1q (1:1,000; Dako cat#A0136) in PBS/1%BSA for 1 h at room temperature, followed by an anti-rabbit Envision (Dako) HRP conjugated antibody also for 1 h at room temperature, with DAB+ as the chromogen. Negative control rabbit immunoglobulin fraction (Dako) was used as a negative control in the same concentration as the primary antibody. Sections were counterstained with Haematoxylin (Klinipath; 4085.9001).

#### Statistics

Statistical analyses were carried out using SPSS statistics version 23 (IBM) or Graphpad Prism version 7. To compare C1q levels the Mann-Whitney U test, Kruskall-Wallis and Dunn's multiple comparisons test were used. In all graphs the median is shown unless indicated otherwise. Receiver operating characteristic (ROC) analysis was performed to assess the sensitivity and specificity of C1q as biomarker and was expressed as Area Under the Curve (AUC).

# RESULTS

#### C1Q Expression Is Upregulated in Patients With Active TB Disease

Publically available microarray data from TB patients were retrieved from Gene Expression Omnibus, all data were ranked as differentially expressed between TB patients and either LTBI or other diseases (5, 7, 33–37). These studies contained information from diverse populations (Malawi, South Africa, United Kingdom, Kenya, The Gambia, USA, and Germany). A list of complement gene expression patterns was generated and the number of microarray studies that reported differential complement gene expression between patients with TB and LTBI (**Figure 1A**) or other diseases (**Figure 1B**) was enumerated. Complement genes C1QB, SERPING1 were expressed at higher level (more than 2-fold) in 8/9 (88%) of studies comparing patients with active TB to LTBI (**Figure 1A**). C1QC was expressed at a higher level in TB patients in 7/9 studies (78%), whereas C1QA expression was only increased in TB patients in 2/9 studies (22%; **Figure 1A**). The observed factorial changes in the expression of these complement genes between TB patients and LTBI individuals were comparable with the changes as seen for FCGR1A and GBP5, which were previously described as promising and highly consistent biomarkers of active TB (**Figure 1C**). A similar pattern, although less pronounced, was seen when comparing TB patients with patients having another lung-disease (**Figures 1B,D**). As C1q is abundantly present, easy to measure and stable, therefore we continued to analyse C1q protein levels.

# C1q Is Significantly Increased in Serum of TB Patients

C1q protein levels were measured in sera from TB patients and controls from independent and geographically distinct cohorts (**Figures 2A–D**). The levels of C1q were significantly higher in sera from patients with active TB as compared to their respective controls in all cohorts: Italy (**Figure 2A**), The Gambia (**Figure 2B**), Korea (**Figure 2C**) and South Africa (**Figure 2D**). LTBI individuals and successfully treated TB patients had serum C1q levels similar to controls (**Figure 2A**). Combined analysis of all TB patients, control groups, LTBI and the successfully treated TB patients from the different cohorts revealed that serum C1q protein levels are significantly (p < 0.001) increased in active TB (**Figure 2E**).

The Gambian TB patients were followed over time which allowed us to investigate C1q levels during successful treatment. At 1 month of treatment the median serum C1q level was still increased, however, the level of C1q began to decrease after 2 months (p = 0.0650), resulting in complete normalization compared to the TB contacts after 6 months of successful treatment (p < 0.001; **Figure 2F**). Thus, serum

transcriptome data was retrieved from Gene Expression Omnibus (5–7, 33–37) and analyzed using GEO2R. Data were available for nine populations comparing active TB with LTBI and for 8 populations comparing active TB with other diseases. For each population we determined if the complement family genes were differentially expressed between TB and LTBI or other diseases. Differential expression was defined as an adjusted *p*-value <0.05 and more than 2-fold change. Differential expression of a gene between TB and LTBI or other diseases was expressed as percentage of the total number of populations investigated. The differential expression of complement genes was scored (A,B) as well as the mean factorial change for the C1q genes, *C1QA, C1QB*, and *C1QC* (C,D) for the comparisons TB vs. LTBI (A,C) and TB vs. other diseases (B,D). As a reference two other highly upregulated potential diagnostic TB markers, *FCGR1A*, and *GBP5*, were included in the analyses.

C1q protein levels were significantly elevated in patients with active TB, and levels decreased to the level of the control population during successful treatment. This further indicates that increased C1q levels are associated with active TB disease and do not reflect genetic variation in C1q expression.

#### Vaccination With BCG Does Not Increase C1q Levels in Serum

To investigate if vaccination with M. bovis BCG, a live replicating mycobacterium, induced a similar increase in serum C1q levels, samples were taken before and after BCG vaccination of healthy Dutch volunteers and C1q levels were measured (**Figure 3**). Samples taken at screening and directly before vaccination showed minimal variation in C1q levels, reflecting normal variation within individuals. BCG vaccination did not induce fluctuations in C1q levels larger than this naturally observed variation. Thus, BCG vaccination did not increase C1q levels in contrast to what was observed in TB disease, despite the presence of live, replicating mycobacteria.

# C1q Levels Are Increased in Active TB Compared to Other Diseases

To investigate the specificity of increased C1q levels for active TB, sera from patients with other diseases with similar symptoms and radiological abnormalities (pneumonia, sarcoidosis) and sera from patients with other mycobacterial disease [leprosy, primary disease or type 1 (acute pro-inflammatory) reactions] were analyzed. C1q levels in sera from patients with TB were significantly higher compared to sera from patients with leprosy, pneumonia or sarcoidosis (**Figure 4A**, data from TB patients and controls same as used in **Figure 2E**). Some individual leprosy patients had C1q protein levels above the median of TB patients but this was not related to either primary disease or having type I reactions (**Supplementary Figure E1A**). Patients with sarcoidosis showed a slight increase in C1q levels compared to the controls. In contrast, patients with community acquired pneumonia showed a significant decrease in C1q levels compared to the control population. The reduced levels observed in patients with pneumonia at diagnosis was associated with the disease state since samples included from the same individuals at later time points had normal C1q levels (**Supplementary Figures E1B,C**).

To assess the value of C1q as possible TB biomarker, the sensitivity, specificity and the positive likelihood ratio (LR+) were calculated from C1q concentration cut-offs (10). With a cutoff at the 95th percentile of the control population (300.2µg/ml C1q) the sensitivity is 42% with a specificity of 91% resulting in a LR+ of 4.96. Application of a cut-off at the maximum of the control population (347.6µg/ml) resulted in a sensitivity of 29% and a specificity of 97% resulting in a LR+ of 8.99. The capacity of serum C1q to discriminate active TB from LTBI, pneumonia and sarcoidosis was also analyzed using ROC analyses and expressed as AUC (**Figures 4B,C**). The AUC of C1q levels for TB vs. LTBI was 0.77, for TB vs. sarcoidosis 0.69, and for TB vs. pneumonia even an AUC of 0.93 was achieved.

# C1q Is Locally Present in the Lungs of TB Patients

So far circulating RNA and protein levels of C1q have been analyzed, which reflect the systemic response to TB. Additionally, we analyzed the local C1q production or deposition in response to M.tb infection by staining lung tissue. Staining lung tissue from a control revealed scarce C1q staining with only few C1q positive macrophage-like cells in the lung parenchyma and in the intra-alveolar space (**Figure 5**). In contrast, lung tissue of fatal TB patients revealed, next to the intra-alveolar C1q positive cells also a pronounced C1q staining both in the necrotic centers of the granulomas and in the surrounding lung tissue with predominantly macrophage-like cells staining positive. Lungtissue from patients that succumbed from pneumonia showed C1q staining predominantly in the intra-alveolar space. Staining of consecutive sections with an isotype control did not reveal any staining, also not in the necrotic centers, confirming specific staining for C1q. Thus, C1q protein is locally detected at an increased level in the lungs of TB patients (n = 3) compared to tissue samples from a non-pulmonary disease control (n = 1) or pneumonia patients (n = 4).

# Non-human Primates With Active TB Disease Also Display Increased Serum C1q Levels

Non-human primate (NHP) M.tb infection models are widely used to study pathogen-host interactions and for pre-clinical evaluation of vaccine candidates (39). After infection, rhesus macaques develop TB disease which closely resembles human TB in most aspects. Sera banked over the course of a long-term follow-up study in rhesus macaques were used to determine C1q levels after experimental M.tb infection. 14 Out of 16 animals with active progressive disease, that had reached an early humane endpoint due to exacerbation of TB disease, had increased C1q levels compared to their baseline C1q levels before infection (**Figure 6A**). Such a rise in C1q levels was absent in six out of seven animals that did not develop overt disease, but controlled the infection over an extended period of time up to 1 year post-infection (**Figure 6A**). Additionally, in an separate cohort of M.tb infected NHPs we detected elevated C1q levels in five out of six broncho-alveolar lavage (BAL) fluid samples taken before necropsy, while no C1q could be detected in paired BAL fluids taken prior to infection (**Figure 6B**). The observed differences could not be explained by any differences in BAL volume recovery and thus these data reflect a true local increase in C1q after M.tb infection.

# DISCUSSION

The accurate and fast identification of patients with active TB remains challenging, largely because of limitations in the current diagnostic tools to differentiate active TB from other diseases, as well as LTBI. Extensive searches for biomarkers that can discriminate active TB from other diseases with similar clinical presentation as well as from LTBI have been recently reported (3, 40). Several studies reported genes encoding for complement components to be upregulated in TB (15–19). Here, we compiled available genome wide gene expression data for TB compared to LTBI and TB compared to other lung diseases (5–7, 33– 37) and observed that in particular C1q encoding genes were highly upregulated. Interestingly, these observations were made in studies using RNA from whole blood, indicating increased transcription of C1q genes in circulating blood cells, most likely monocytes/macrophages. Since C1q is not a typical acute-phase protein we were interested to confirm and validate these findings at the protein level. We have therefore measured C1q protein levels in serum and confirmed increased levels in patients with active TB, but not in other diseases with a similar clinical presentation, such as pneumonia or sarcoidosis, or mycobacterial exposure. C1q protein is also present in the lung tissue of deceased TB patients. The increased levels of local and circulating C1q in TB were replicated independently in a NHP TB infection model.

Literature suggests that expression of the three C1q genes C1QA, C1QB, and C1QC is regulated in a similar manner (41).

However, upon IFNγ-stimulation upregulation of C1QB is higher than C1QC, which is higher than C1QA (41). Similarly, our data also showed upregulated expression of C1QB and C1QC genes and C1QA was less frequently observed in active TB, also in the analysis from Cai et al. the extend of increase in the expression of C1QA was less pronounced as compared to the increase in expression of C1QB and C1QC. Since C1q protein production requires equal ratios of all three chains, the detected increase in C1q protein levels in TB patients indicates that all chains are expressed. The low detection of C1QA may thus be technical and reflect a poor capture of C1QA expression on the microarrays in general.

We measured C1q protein levels in four different geographical cohorts from Italy, The Gambia, Korea and South Africa. C1q levels were increased in patients with active TB compared to all relevant control populations. Importantly, treatment normalized serum C1q levels to those of endemic controls, indicating that the upregulation of C1q was associated with the disease and not intrinsic to the individuals. BCG vaccination, although being a live replicating mycobacterial vaccine, did not induce increased C1q levels. Although leucocytes of patients with leprosy reactions were reported to express increased levels of C1QA, B, and C (42), the cohorts analyzed here did not universal show increased serum C1q levels. Individual patients might have somewhat increased levels, both in this study (type 1 reaction) and a previous report (type 2 reaction) (43), which warrants more detailed analyses. We speculate that the different pathophysiologies of TB and leprosy, in particular the different levels of systemic inflammation and immune activation which are generally higher in active TB than in leprosy are responsible for the difference in C1q levels between the two diseases, even though both are caused by pathogenic mycobacteria.

To further evaluate the potential of C1q as a biomarker for active TB, C1q levels were compared to those of disease relevant controls as patients with untreated pulmonary sarcoidosis or pneumonia. Patients with TB had significantly higher circulating C1q levels as compared to patients with sarcoidosis or patients with pneumonia. Thus, the upregulation of C1q likely does not reflect a general response to inflammation. Patients with pneumonia rather had decreased C1q levels compared to

FIGURE 5 | C1q accumulates in lung tissue of patients with fatal pulmonary TB. Lung tissue, obtained at autopsy from a non-pulmonary disease control (*n* = 1), fatal active TB patients (*n* = 3), or patients with lethal pneumonia (*n* = 3) were stained for the presence of C1q. The left column shows the presence of C1q in a section of the control, two active TB patients and a pneumonia patient, scale bar at 200µm. The middle column shows the same samples stained for C1q, while the right column shows consecutive tissue slides stained with a matched control antibody, scale bars at 50µm. Necrotic areas that stain positive for C1q are highlighted with an asterisk (\*) and individual cells that stain positive for C1q are highlighted with arrows.

controls. Treatment of pneumonia normalized the levels of circulating C1q, suggesting that the observed decrease in C1q is non-genetic and likely associated with the disease process. ROC analysis indicated that C1q, even as a single marker, readily discriminated active TB from all other diseases investigated here. Furthermore, also in a setting of an experimental NHP model of TB disease we observed in both sera and BAL fluid increased C1q levels in animals with symptomatic TB disease compared to the level prior to infection with M.tb. This was not seen in animals that did not progress to TB disease, suggestion again an association with TB disease rather than infection only. These data fully agree with and support the observations described above in the human cohorts.

Longitudinal follow up of TB patients during treatment revealed that circulating C1q levels normalized to the level of endemic controls. However, here protein levels did not completely normalize until the 6 months time point. Previously published data showed a rapid decrease in C1q mRNA expression levels following treatment (16, 19). Cliff et al. show that already after 1 week of treatment the blood C1q gene expression decreased (16). Cai et al. reported a significant decrease in the expression of the C1q genes at 3 months of anti-TB chemotherapy whereas they also described a reduction in C1qc protein levels after 6 months (19). Our data presented here substantially expand the number of populations, including populations from different regions of the world. Transcriptomic analysis of genes encoding complement proteins were now assessed in 9 independent populations from different TB endemic as well as non-endemic regions, strongly supporting an increase in expression of C1Q as well as SERPING1 during active TB disease in cells present in peripheral blood. Moreover, 4 independent cohorts as well as the data obtained in the NHP model show increased circulating C1q plasma levels during active TB disease, but not infection. In addition to conforming the data previously published for a Chinese population (Cai et al.) in 4 different TB cohorts, we also showed the specificity for TB disease in comparison to clinically important differential diagnoses, such as sarcoidosis and pneumonia. As C1q is produced by cells of monocytic origin, it reflects another component of the immune space compared to most currently applied TB biomarkers, such as C-reactive protein and IP-10 (44–47). In addition, C1q levels are technically easy to measure and the C1q protein is not sensitive to degradation. Therefore, we hypothesize that addition of C1q to current biomarker panels or platforms, will have additive value in discrimination of TB patients.

Low levels of C1q have been reported in several inflammatory and autoimmune diseases, such as Systemic Lupus Erythematosus. This is largely the result of C1q consumption because of immune complex mediated disease and in some rare cases caused by genetic C1q deficiency (26). However, the increased levels of C1q, as occur in TB, are observed very rarely. So far the only other clinical condition in which increased levels of C1q have been reported is Kala Azar (48). The mechanism behind the increased C1q levels observed in TB patients, or the possible functional consequences for the host are unknown, and will need further investigation. The availability of the NHP model of tuberculosis for C1q research, as demonstrated for the first time in this study, should greatly accelerate and facilitate such work.

In conclusion, we show here that circulating C1q expression is increased in 9 different populations with TB disease, moreover, elevated C1q plasma levels were observed in 4 cohorts of TB patients compared to LTBI or endemic controls. Specifically, C1q levels in TB patients were significantly increased compared clinically relevant diseases, such as sarcoidosis, leprosy and pneumonia. Moreover, we show that increased C1q levels decreased to the level of the control population during successful treatment. In analogy with human TB, C1q also validated as a biomarker of TB disease in rhesus macaques, in both serum and BAL. Increased C1q levels were only observed in animals that progressed to active disease and not in those that controlled the infection, suggesting a direct association with disease rather than with infection. Therefore, we propose that the addition of C1q measurements to current biomarker panels may provide added value in the diagnosis of active TB.

# AUTHOR CONTRIBUTIONS

RL, FV, AG, TO, SJ, and LT designed the study. RL, RdP, KD, IB, and KG performed analyses. JS, DG, CvM, MV, SV, WB, LP, FD, GW, and GG oversaw recruitment and collection of specimens. RL, SJ, and LT interpreted the data. All authors critically revised and approved the manuscript.

# FUNDING

This study was supported by funding from the Italian Ministry of Health: Ricerca Corrente and a grant from the European Union EC HORIZON2020 TBVAC2020 (Grant Agreement No. 643381) and a grant from National Institutes of Health of USA (NIH 1R21AI127133-01). Additionally the work was supported by Zon-Mw TopZorg grant (842002001) (CvM and MV) and Zon-Mw Vidi grant (no. 91712334) (LT). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

# ACKNOWLEDGMENTS

The authors are grateful to all the patients, the physicians, nurses that helped to perform the study. In particular we thank Gilda Cuzzi and Valentina Vanini (INMI, Rome, Italy) for the recruitment organization and lab work of the Italian cohort. Additionally, we would like to thank Sang-Nae Cho (Yonsei University, Seoul, South-Korea) for making available TB and control sera and Malu Zandbergen (Dept of Pathology, LUMC, Leiden, The Netherlands) for excellent support with immunohistochemistry. The IDEAL consortium is acknowledged for providing sera of leprosy patients.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02427/full#supplementary-material

Supplementary Figure E1 | C1q levels in leprosy and community acquired pneumonia. From leprosy two different cohorts were measured, one consists out

#### REFERENCES


of patients that were included in the Netherlands at the moment of diagnosis, the other out of sera samples from patients included at the moment they presented with a leprosy reaction. Reference C1q levels both the control groups as the pooled data from the active TB patients are depicted from Figure 2E (A). For the community acquired pneumonia cohorts, samples were available from the moment the patientswere included and a follow up sample after recovery from both in Leiden (B) and in Nieuwegein (C).

modulation of the humoral immune response. J Infect Dis. (2013) 207:18–29. doi: 10.1093/infdis/jis499


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Lubbers, Sutherland, Goletti, de Paus, van Moorsel, Veltkamp, Vestjens, Bos, Petrone, Del Nonno, Bajema, Dijkman, Verreck, Walzl, Gelderman, Groeneveld, Geluk, Ottenhoff, Joosten and Trouw. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Active Human Complement Reduces the Zika Virus Load via Formation of the Membrane-Attack Complex

Britta Schiela<sup>1</sup> , Sarah Bernklau<sup>1</sup> , Zahra Malekshahi <sup>1</sup> , Daniela Deutschmann<sup>1</sup> , Iris Koske<sup>1</sup> , Zoltan Banki <sup>1</sup> , Nicole M. Thielens <sup>2</sup> , Reinhard Würzner <sup>3</sup> , Cornelia Speth<sup>3</sup> , Guenter Weiss <sup>4</sup> , Karin Stiasny <sup>5</sup> , Eike Steinmann<sup>6</sup> and Heribert Stoiber <sup>1</sup> \*

<sup>1</sup> Division of Virology, Medical University of Innsbruck, Innsbruck, Austria, <sup>2</sup> CNRS, CEA, IBS, University of Grenoble Alpes, Grenoble, France, <sup>3</sup> Division of Hygiene and Medical Microbiology, Medical University of Innsbruck, Innsbruck, Austria, <sup>4</sup> Department of Internal Medicine II, Medical University of Innsbruck, Innsbruck, Austria, <sup>5</sup> Center for Virology, Medical University of Vienna, Vienna, Austria, <sup>6</sup> Department of Molecular and Medical Virology, Ruhr University Bochum, Bochum, Germany

#### Edited by:

Zvi Fishelson, Tel Aviv University, Israel

#### Reviewed by:

Uday Kishore, Brunel University London, United Kingdom Péter Gál, Institute of Enzymology (MTA), Hungary

\*Correspondence:

Heribert Stoiber Heribert.stoiber@i-med.ac.at

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 08 June 2018 Accepted: 03 September 2018 Published: 17 October 2018

#### Citation:

Schiela B, Bernklau S, Malekshahi Z, Deutschmann D, Koske I, Banki Z, Thielens NM, Würzner R, Speth C, Weiss G, Stiasny K, Steinmann E and Stoiber H (2018) Active Human Complement Reduces the Zika Virus Load via Formation of the Membrane-Attack Complex. Front. Immunol. 9:2177. doi: 10.3389/fimmu.2018.02177 Although neglected in the past, the interest on Zika virus (ZIKV) raised dramatically in the last several years. The rapid spread of the virus in Latin America and the association of the infection with microcephaly in newborns or Guillain-Barré Syndrome in adults prompted the WHO to declare the ZIKV epidemic to be an international public health emergency in 2016. As the virus gained only limited attention in the past, investigations on interactions of ZIKV with human complement are limited. This prompted us to investigate the stability of the virus to human complement. At low serum concentrations (10%) which refers to complement concentrations found on mucosal surfaces, the virus was relatively stable at 37◦C, while at high complement levels (50% serum concentration) ZIKV titers were dramatically reduced, although the virus remained infectious for about 4–5 min under these conditions. The classical pathway was identified as the main actor of complement activation driven by IgM antibodies. In addition, direct binding of C1q to both envelope and NS1 proteins was observed. Formation of the MAC on the viral surface and thus complement-mediated lysis and not opsonization seems to be essential for the reduction of viral titers.

Keywords: complement, Zika virus, C1q, IgM, MAC, lysis

# INTRODUCTION

Within the genus Flavivirus (family Flaviviridae), several members were identified as human pathogens including dengue (DEN), Japanese encephalitis (JE), tick-born encephalitis (TBE), West-Nile, (WN), yellow fever (YF) and Zika (ZIK) viruses (1). The main route of infections is mediated by arthropods such as mosquitoes or ticks (2, 3). The viruses contain a single stranded RNA of positive polarity with a size of about 11 kb. The RNA is translated as a polyprotein and cleaved by viral and host-encoded proteases into seven non-structural (NS) and three structural proteins, including the capsid, membrane (prM/M) and envelope (E) protein (3). Interestingly, two flavivirus proteins are characterized as main participants in interactions with the immune system. The E protein binds to the cell surface and mediates fusion after endocytic virus uptake. The majority of the neutralizing antibody responses is directed against the E protein. NS1 functions as regulator of viral transcription and has been shown to antagonize the anti-viral immune response by interfering with the interferon pathway. In addition, NS1 interacts with several proteins of the complement system (4, 5).

As a first line of the defense mechanism of the innate immune system, complement activation triggers a proteolytic cascade leading to release of anaphylatoxins, chemokines and cytokines, opsonization and phagocytosis of invading pathogens and killing by the formation of the membrane-attack complex (MAC) (6). Upon viral entry, either the classical, the lectin or the alternative pathway is initiated. The classical pathway is activated by binding of C1q to immune complexes consisting of IgG or IgM bound to the viral surfaces or by direct interactions of viral proteins with C1q (7, 8). Carbohydrates expressed on the surface of pathogens or ficolins bound to viral surfaces may be recognized by the mannan-binding lectin (MBL) which triggers the lectin pathway. The alternative pathway is constitutively activated due to spontaneous hydrolysis of C3 and drives the amplification of the classical and the lectin pathways. The three activation pathways merge in cleavage of C3 into C3a and C3b by the C3-convertases and the formation of the C5-convertase. Cleaved and thus activated C5b initiates the lytic pathway. C5b becomes associated with C6, C7, C8 and several C9 proteins, referred to as membrane-attack complex (MAC) responsible for formation of the lytic pore resulting in destruction of viral pathogens (6).

The interest for ZIK virus (ZIKV) raised only recently, due to massive spread of the virus mainly in Latin America (9). Infection may cause severe neurological complications mainly in newborn children of infected mothers during pregnancy (10) and is associated with Guillain-Barré Syndrome in adults (11). As, in contrast to other members of the Flaviviridae family, investigations on the interaction of ZIKV with proteins of the complement cascade are limited, the aim of this study was to assess the potential complement activating capacity of ZIKV. A further aim was to identify proteins involved in this putative ZIKV-complement interaction and to investigate whether complement can reduce the viral titer.

# MATERIALS AND METHODS

#### Cells and Viruses

Aedes albopictus C6/36 mosquito cells were grown in Dulbecco's modified Eagle's medium (DMEM; Invitrogen, Carlsbad, USA) supplemented with 10% heat-inactivated fetal calf serum (FCS), antibiotic-antimycotic solution [10,000 units/mL penicillin, 10,000µg/mL streptomycin, and 25µg/mL Amphotericin B], L-glutamine, and non-essential amino acids (Gibco, Dublin Ireland) at 28◦C in 5% CO2.Alternatively, the human cell line A549 was used and cultivated under the same conditions. ZIKV strain MRS\_OPY\_Martinique\_PaRi\_2015 (GenBank: KU647676) was provided by European Virus Archive (Marseille, France). For propagation of ZIKV, cells were seeded in culture plates to get confluence of about 80% at the day of infection. Cells were washed with phosphate-buffered saline (PBS) and ZIKV was added with a multiplicity of infection (MOI) of 0.1. After 1 h at 37◦C, cells were washed twice with PBS and fresh medium was added. Depending on the growth kinetics of the cell line, the supernatants were harvested and filtered through a 0.45-µm filter to remove cell debris. To generate high titer viral stocks the supernatants were centrifuged overnight (Rotanta 460R Hettich; 4,600 rpm, 16 h, 4◦C). The concentrated supernatants were aliquoted and stored at −80◦C. All experiments were performed under bio-safety level-2 conditions.

#### Human Serum Samples

Normal human serum (NHS) was purchased from Dunn Labortechnik GmbH (Ansbach, Germany) and stored in aliquots at −80◦C. For experimental procedures, serum was thawed only once and kept on ice. Some aliquots from the serum pool were heat inactivated (hiNHS; 56◦C, 30 min) and served as controls. Complement C9-or C1q-depleted human serum and purified C9 or C1q were purchased from CompTech (Tylor, Texas USA).

#### Plaque Assay

To count plaque-forming units (PFU), the virus [total volume 100 µL/sample] was serially diluted 1:10 and added to Vero cells grown in 6-well or 12-well plates. ZIKV samples were incubated with the cells for 1 h at 37◦C and overlaid with plaque agarose (Biozym Scientific GmbH, Hessisch Oldendorf, Germany). Four days after incubation at 37◦C, viral plaques were visualized by crystal violet staining. The viral titers were expressed as PFU/mL, calculated as [(number of plaques per well) × (dilution)]/(inoculum volume).

#### Serum-Sensitivity Assay

ZIKV [1 x 10<sup>6</sup> PFU/mL] was incubated with 10%, 20% or 50% (final concentration) NHS, heat-inactivated NHS (hiNHS), or DMEM (supplemented with FCS) as controls. When indicated, complement-depleted sera were used instead of NHS. To block the lectin pathway, a mixture of sunflower MASP inhibitor (SFMI)-1 (3.2 mM) and SFMI-2 (3.2 mM) peptides was used (Metabion, Planegg, Germany). These peptides are known to selectively inhibit MBL-associated serine protease (MASP)-1 and -2 (12). For some experiments, putative IgM in human serum were blocked by an affinity purified goat IgG against human IgM as recommended by the manufacturer (Bethyl Laboratories, Montgomery, USA). After an incubation time of 1 h at 37◦C, all samples were serially diluted and titrated on Vero cells to determine the viral titer by plaque assay. When indicated, ZIKV [1 x 10<sup>6</sup> PFU/mL] derived from A549 cells was used.

#### Inhibition of Complement Activation

To prevent activation of all complement pathways, ZIKV [1 x 10<sup>6</sup> PFU/mL] was incubated with 50% NHS in the presence of EDTA [1, 2.5 or 5 mM] or Mg2+-EGTA [5 mM] for 1 h at 37◦C. Immediately after, the virus-containing samples were serially diluted and titrated on 12-well plates of overnight-plated Vero cells. One h after incubation at 37◦C, plaque agarose was overlaid. Four days post infection, the visualization and calculation of PFUs were performed as described for Plaque Assay.

#### Inhibition of MAC Formation

The contribution of the MAC was determined by using C9 deficient serum as follows: ZIKV (1 x 10<sup>6</sup> PFU/mL) was incubated for 1 h at 37◦C in the presence of active or heatinactivated C9-depleted human serum. As a control, the depleted serum was reconstituted with purified C9 protein (60 mg/mL). The titration and plaque visualization were performed as described above. In parallel, viral RNA was analyzed as described below.

#### C1q Binding ELISA

To investigate the interaction of the complement component C1q with ZIKV-derived proteins, 5µg/mL of recombinant ZIKV envelope or NS1 protein [MyBioSource, San Diego, USA] or 100 µL of ZIKV supernatant derived from C6/36 cells [containing 5 x 10<sup>6</sup> PFU/well] were coated in carbonate buffer (pH 9.6) on a 96-well flat-bottom plate (Maxisorp, Nunc, Roskilde, Denmark) overnight at 4◦C. After washing thrice with 200 µL PBS containing 0.05% Tween 20 (PBS-T), blocking solution (3% non-fat dry milk in PBS) was added and allowed to incubate for 30 min at room temperature. Subsequently, 100 µL of purified C1q protein was added as indicated in the figure and incubated with gentle shaking for 2 h at room temperature. The plates were washed five times with PBS-T, followed by the addition of 50 µL/well of homemade polyclonal rabbit antibody against the globular heads of human C1q (1:500). After 1 h at room temperature, the plates were washed again and 50 µL of a goat anti-rabbit IgG conjugated with horseradish peroxidase (HRP) was added (1:10,000). The plates were again incubated for 1 h at room temperature. For detection of bound antibodies, 200 µL 3,3′ ,5,5′ -tetramethylbenzidine (TMB) solution were added following the manufacturer's instructions. The absorbance was measured at 450 nm, using a Bio-Rad plate reader.

#### PCR to Determine the Relative Viral Titer

ZIKV [1 × 10<sup>8</sup> RNA copies/mL] was mixed with 10, 20, or 50% (final concentration) NHS or hiNHS and DMEM (supplemented with FCS) as controls. Additionally, 1 mg/mL RNase A [Macherey-Nagel 740505, Dueren, Germany] was added to digest the RNA of lysed viral particles. The mixture [total volume 100 µL/sample] was incubated for 3 h at 37◦C in a thermoshaker. Afterwards, the residual viral RNA of remaining intact virions was harvested using the NucliSENS easyMAG system (BioMérieux, Vienna, Austria) as recommended by the manufacturer. To measure the ZIKV-specific RNA, a reverse transcription-PCR (RT-PCR) was done, using the iScript One-Step RT-PCR kit (Quanta; Biorad, Munich, Germany). Primer and probe sequences as well as the thermal profile of the PCR were performed according to the protocol of Lanciotti et al. (13). The relative reduction of the viral RNA was calculated as follows: ct-value of the heat-inactivated serum controls was set at 100%. As a reduction of 3.2 ct-units in the real-time PCR equals 1 log (=90%), changes of the viral RNA in active serum were determined relative to the heat-inactivated serum controls.

#### Statistical Analysis

Statistical analysis was performed using GraphPad Prism 7.0 software. All experiments were repeated at least three times in duplicate. The difference between two groups was assessed by ttest. When comparing more than two groups, ANOVA followed by Bonferroni post-hoc tests was performed. A 95% significance

#### level (p < 0.05) was considered statistically significant and indicated in the figures as follows: <sup>∗</sup>p < 0.05 to p = 0.01, ∗∗p < 0.01 to p = 0.001, ∗∗∗p < 0.001 to p = 0.0001 and ∗∗∗∗p < 0.0001.

#### RESULTS

# Serum Neutralization of ZIKV Particles

To analyze the stability of ZIKV in human serum, insect (mosquito) cell line derived ZIKV was incubated with increasing amounts of pooled NHS and the number of infectious virions was quantified by plaque titration on Vero cells. NHS was tested for the presence of flavivirus antibodies (**Supplementary Figure 1**) to exclude antibody-mediated viral neutralization. The titer of ZIKV was reduced in a concentrationdependent manner in the presence of NHS, compared to heatinactivated serum (hiNHS), which was used at a concentration of 50% (**Figure 1**). The presence of 10% NHS did not affect ZIKV infectivity (**Figure 1)**, whereas 20% NHS yielded a titer reduction of more than 0.5 log. Increasing the serum concentration up to 50% resulted in a viral titer reduction of approximately 2 log (**Figure 1**). As the plaque counts of the hiNHS control were comparable to the initial viral load [1 x 10<sup>6</sup> PFU/mL] of the mock medium (data not shown), heat sensitive factors of human serum appear to be involved, implicating the complement system as anti-viral factor for this neutralization.


#### Time-Dependent Stability of ZIKV

The transmission of ZIKV by mosquito species like Aedes aegypti occurs at short time intervals (14). Consequently, the time frame between viral inoculation into the skin and receptor-mediated endocytosis by human cells will probably be shorter than the preselected incubation time of 60 min in the serum resistance assays described above. The infectivity of ZIKV remained unaffected for at least 4–5 min. A decrease in plaque titers was observed only after prolongation of the incubation time with about 1 log reduction after 30 min (**Figure 2**).

#### Antiviral Activity of Human Serum Is Mediated Mainly by the Classical Complement Pathway

The loss of anti-viral activity of human serum against ZIKV (shown in **Figure 1**), when hiNHS was used, suggests that a complement-dependent mechanism is involved. To confirm this hypothesis, complement activation of all three pathways was inhibited by adding the Mg2<sup>+</sup> and Ca2<sup>+</sup> chelator EDTA, which inhibited complement-mediated lysis of ZIKV in a dosedependent manner (**Figure 3A**).

While the initial plaque titer was reduced up to 2 orders of magnitude in 50% NHS, the addition of 5 mM EDTA fully inhibited the ZIKV neutralization activity of NHS. These results indicate that the sensitivity of ZIKV to human serum is mediated by complement activation. We further characterized the participation of the classical and lectin pathways. As both pathways are essentially linked to the availability of Ca<sup>+</sup> ions, the activation can be efficiently blocked with Mg2+-EGTA, whereas the alternative pathway remains active (15). ZIKV was incubated with NHS pretreated with 5 mM Mg2+-EGTA (**Figure 3A**). Like the 5 mM EDTA treatment, Mg2+-EGTA fully inhibited the ZIKV neutralization activity of NHS, excluding the participation of the alternative pathway. Therefore, it is likely that the anti-viral activity of human serum is mainly mediated by the classical/lectin pathway of complement. To further investigate the role of the classical/lectin pathway in the complement-mediated viral neutralization, complement activation was inhibited by C1 esterase inhibitor (C1-INH). The serine protease inhibitor C1-INH is a natural regulator of the activation of the classical and lectin pathways. The human serum was pre-incubated with increasing amounts of C1-INH prior to the serum resistance assay of ZIKV. The viral neutralization was inhibited by C1-INH in a concentration-dependent manner, further confirming the involvement of the classical and/or the lectin pathway (**Figure 3B**). Additionally, C1q-depleted serum was used to discriminate between activation of the classical and lectin pathways. In the absence of C1q, the viral titer was nearly unaffected. In contrast, addition of C1q at a concentration of 70µg/mL to the depleted serum restored the capacity of the serum to reduce the viral titer (**Figure 3C**), indicating that the classical pathway is the main trigger for the observed reduction of the viral titer. To further investigate a putative contribution of the lectin pathway, insect-derived ZIKV was incubated with NHS in the presence of the peptides SFMI-1/2. Surprisingly, these MASP-blocking peptides were unable to rescue the viral

FIGURE 2 | Time as a limiting factor on ZIKV infectivity. ZIKV [1 x 10<sup>6</sup> PFU/mL] was mixed with 50% active or heat-inactivated NHS and incubated for different time points (ranging from 1 min to 1 h) at 37◦C. Virus-containing samples were then serially diluted and titrated on Vero cells. After 1 h incubation at 37◦C, the cells were overlaid with agarose. Viral concentration was determined 4 days post infection using crystal violet staining. Mean data of 3 independent experiments are shown. The error bars represent the standard deviation.

titer (**Figure 3D**). By contrast, no lysis was observed when ZIKV was treated with SFMI-1/2 in C1q-depleted serum. These results underscored the dominant role of the classical pathway for the reduction of ZIKV load.

#### Neutralization of ZIKV Is Linked to Natural IgM Antibodies

As a putative binding of IgM to viral particles may trigger activation of the classical complement pathway (16), active NHS was pre-incubated with anti-IgM antibodies prior to ZIKV addition. As shown before (**Figure 1**), the titer of ZIKV was reduced at about 2-logs in the presence of 50% (nontreated) NHS, compared to hiNHS. Interestingly, the viral titer was equivalent to the hiNHS control in the presence of IgM-blocking antibodies (**Figure 4**), indicating that natural IgM antibodies in human serum are involved in complementmediated neutralization of ZIKV. The presence of insect-specific IgM antibodies in the serum pool was further confirmed by FACS using the insect cell line C6/36, which gave a clear positive signal, when incubated with the serum pool. In contrast, cell lines of human origin, such as A549, remained negative (**Supplementary Figure 2**).

#### C1q Binding to ZIKV Proteins

For WNV it has been reported that C1q directly interacts with the E protein. Further, the NS1 protein of DENV is known to bind C1q. These observations prompted us to test if the ZIKV E and/or NS1 proteins may interact with C1q. The viral recombinant proteins and entire viral particles were coated in an ELISA plate and incubated with increasing amounts of C1q. As shown in **Figure 5**, both recombinant viral proteins NS1 and the envelope protein bound C1q in a dose-dependent manner similarly to

lectin complement pathways, active or heat-inactivated human serum was pre-incubated with increasing amounts of C1 esterase inhibitor as indicated. (C) In the absence of C1q, most of the virus remained infectious whereas addition of purified C1q (70µg/mL) restored the lytic effect on the virus. HiNHS was set to 100%. (D) Blocking of the lectin pathway by synthetic peptides (SFMI-1/2) did not rescue the virus. Combination of C1q-depleted serum and SFMI-1/2 served as additional control showing that the peptides had no effect on the infectivity of the virus. In all experiments, the viral titer was determined by plaque assays using Vero cells. Data were analyzed with one-way-ANOVA with Bonferroni post-hoc comparison (\*\*p < 0.005). All virus lysis experiments were conducted in duplicates and repeated three times. The data represent mean values, and the error bars show standard deviations.

ZIKV particles, indicating that, beside IgM, a direct activation of the classical pathway is possible.

#### Lysis of ZIKV Derived From a Human Cell Line

As expected, NHS showed no reactivity against human A549 cells (**Supplementary Figure 2**). Thus, we concluded that a contribution of IgM against ZIKV derived from human cells was unlikely and provided a tool to investigate the significance of C1q binding to the envelope proteins, at least in vitro. In addition, the serum was tested negative for the presence of antibodies with cross-reactivity against flaviviruses (**Supplementary Figure 1**), which excluded a possible trigger of the classical pathway by putative "contaminating" antibodies. As shown in **Figure 6**, A549-derived ZIKV was sensitive to human serum, although the reduction of the viral titer was less pronounced compared to ZIKV harvested from insect cells. Again, the MASP-blocking peptides were unable to rescue the virus. By contrast, incubation of A549-derived ZIKV with C1q depleted serum had no effect on the viral titer. As NHS was unable to reduce the amount of A549 derived ZIKV in the presence of EGTA or Mg2+-EGTA (data not shown), we concluded that a direct binding of C1q to the virus significantly contributes to the control of ZIKV in NHS.

# Complement Induces Lysis of ZIKV

Complement-mediated neutralization of ZIKV may be due to deposition of proteins such as C3b fragments, which may hide viral epitopes important for infection or, alternatively, may occur by lysis of virions due to formation of the MAC. To analyze whether complement-mediated viral neutralization is linked to MAC assembly, we established an RNase digestion assay followed by PCR. ZIKV was incubated with NHS or hiNHS in the presence of RNases. The formation of the MAC enables the entry of RNases into the virions and digestion of viral RNA. By contrast, intact ZIKV, which are opsonized with C3b, are not affected. Quantification of the virus by real-time PCR (RT-PCR) showed a reduction of RNA copies only when MAC formation occurred. Similar to previous plaque titrations (**Figure 1**), the presence of human serum induced a concentration-dependent reduction of initial RNA levels (**Figure 7**). When serum was inactivated, no significant changes in RNA copy number were observed, compared to the virus incubated with mock medium (DMEM, 10% FCS; not shown). To further confirm the importance of MAC formation for the reduction of viral titers, C9-depleted serum was used. ZIKV was exposed to C9-depleted human serum in the presence of RNases. As a control, the C9-depleted serum was reconstituted with purified C9 protein adjusted to its natural concentration in serum (60µg/mL). As expected,

FIGURE 4 | Natural IgM blocking results in ZIKV rescue. Anti-human IgM blocking antibodies were incubated with 50% NHS or hiNHS for 30 min on ice before ZIKV was added. After incubation of 1 h at 37◦C, the virus-serum mixture was serially diluted and titrated on Vero cells. After 1 h r incubation at 37◦C, the cells were overlaid with agarose. Viral concentration was determined4 days post infection using crystal violet staining. All virus lysis experiments were conducted in triplicate, and the error bars show standard deviations.

FIGURE 5 | Binding of C1q to recombinant ZIKV envelope (E) and NS1 proteins. ZIKV proteins or viral particles were coated to ELISA plates and incubated with decreasing amounts of C1q as indicated. Bound C1q was detected using a polyclonal anti-C1q antibody followed by a HRP-labeled goat anti-rabbit IgG and visualized by TMB. The absorbance was measured at 450 nm, using a Bio-Rad plate reader. Data show the mean of two experiments performed in duplicate.

complement-induced lysis was inhibited in the absence of C9 (**Figure 8**). When the serum was reconstituted with purified C9 protein, a decrease of RNA was observed, indicating that the lytic function of complement was restored. Similarly, C9-depleted serum had no effect in the plaque assay (data not shown). In line with the results obtained by PCR, the neutralization capacity was restored in the plaque assay when purified C9 was re-added to the depleted serum (data not shown).

#### DISCUSSION

Our study focused on the anti-viral activity of NHS and identified the complement system as an immediate innate

FIGURE 6 | Dissecting the role of the lectin and classical pathways on ZIKV derived from the human cell line A549. NHS (50%) reduced the viral titer for about one order of magnitude. Inhibition of the lectin pathway by a peptide mix of SFMI-1 and 2 had no effect. By contrast, C1q depletion rescued the virus and most of the virus remained infectious. Viral titer was determined by plaque assays using Vero cells. Data were analyzed with one-way-ANOVA with Bonferroni post-hoc comparison (\*\*p < 0.005). All virus lysis experiments were conducted in duplicates and repeated two times. The data represent mean values, and the error bars show standard deviations.

FIGURE 7 | Reduction of viral infectivity is linked to decreased RNA levels. After incubating ZIKV with active or heat-inactivated human serum, the viral RNA was digested by addition of RNases. Three hours after incubation at 37◦C, the remaining genomic material was extracted and quantified by RT-PCR. The RNA copy number was calculated from the amount of RNA obtained by incubation of the virions with 50% hiNHS, which was set to 100%. Results are given as % of RNA loss. Data were analyzed with one-way-ANOVA with Bonferroni post-hoc comparison (\*\*\*\*p < 0.0001). All virus lysis experiments were conducted in triplicate. The data represent mean values, and the error bars show standard deviations.

immune response against ZIKV. To mimic the environmental conditions of infection in several compartments, different initial concentrations of complement have been chosen for the in vitro experiments. Once a blood-feeding female mosquito transmits the virus, ZIKV is confronted with high concentrations of complement. In contrast, sexual transmission exposes the virus to lower levels of complement at mucosal surfaces. Thus, 10% NHS followed by exposure to 20 and 50% NHS has been

FIGURE 8 | Blocking the assembly of MAC leads to virus rescue. ZIKV [1 x 10<sup>6</sup> PFU/mL] was exposed to either 50% C9-depleted (C9dep) human serum or 50% heat-inactivated C9-depleted (hiC9dep) serum for 1 h at 37◦C. As a control, the depleted serum was reconstituted with purified C9 protein, adjusted to its natural concentration in serum [60µg/mL]. During incubation, the RNA of lysed ZIKV was digested by external RNase addition. Subsequently, the amount of complement lysis-resistant virions was determined by RT-PCR. The RNA copy number was calculated by incubation of the virions with 50% hiNHS, which was set to 100%. Results are given as % of RNA loss. All assays were performed in triplicate. The data represent mean values, and the error bars show standard deviations.

used for the in-vitro studies. Experiments using Mg2+-EGTA showed that the alternative complement pathway is not involved in the reduction of the viral titer. The ZIKV neutralization assays with NHS in the presence of C1-INH indicated that the classical complement pathway is triggered. However, C1- INH may also interfere with the lectin pathway by inhibiting the MASP-1/MASP-2 proteases of the MBL-MASP complexes (17). Therefore, experiments with C1q-depleted serum were performed which clearly indicate that activation of the classical pathway is mainly involved in the reduction of the viral titer. This came by surprise as mosquito cell-derived virus exhibits a mix of high-mannose and paucimannose glycans which should favor binding of MBL (18). In addition, the participation of the lectin pathway to the neutralization of different members of the flaviviruses was shown in former studies. In line with these observations, Fuchs et al. demonstrated direct binding of MBL to WNV, which triggers the activation of MASP-2, followed by opsonization of the virion surface, which finally hampers the viral fusion with host membranes (19). Furthermore, Avirutnan et al. confirmed the involvement of MBL, since around 80% of the neutralizing capacity of all DENV serotypes was lost when mouse serum from MBL-A/C−/<sup>−</sup> mice was used (20). In addition, they showed that MBL, but not C1q or C5, was necessary to neutralize the virus, which states against the participation of the classical complement activation pathway and the formation of membrane attack complex. However, in our setting, blocking of the lectin pathway by synthetic peptides (i.e., SFMI-1 and 2) did not rescue the virus, indicating that the classical pathway is mainly responsible for the reduction of the viral titer.

In line with our observation that ZIKV E interacts with C1q, Douradinha et al. reported a direct binding of C1q to recombinant E protein of DENV and viral particles (21). This finding was confirmed by proteomic analysis demonstrating that samples from DENV-infected patients as well as purified domain III of DENV envelope protein bind several complement components among which C1q (22, 23). As in addition, binding of C1q to NS1 of DENV is described (24), we tested the possibility of a direct interaction of C1q with viral particles and purified recombinant NS1 and E proteins of ZIKV. Direct binding of C1q to ZIKV or its recombinant proteins was indeed observed, which prompted us to test if serum immunoglobulins generate immune complexes able to trigger the classical pathway. As we could exclude the presence of cross-reactive IgG from other flaviviruses, we investigated the involvement of natural antibodies. A first hint for the presence of anti-insect IgM was obtained by FACS analysis showing that IgM antibodies in the blood recognized insect cells, whereas cell lines of human origin gave no signal. In addition, ZIKV derived from monkey or human cell lines is more resistant to complement-mediated lysis and complement activation is independent of IgM, further indicating that insect-specific immunoglobulins are involved (manuscript in preparation). This allows the conclusion that neither flavivirus-specific IgG nor IgM is present in the serum pool. Although IgM affinity is low for antigens compared to antibodies of other isotypes (25, 26), it was previously shown that natural IgM antibodies are involved in the elimination of viruses like vesicular stomatitis virus (27, 28), lymphocytic choriomeningitis virus (29) or influenza virus (16, 30–32). In this regard, Beebe and Cooper reported that natural IgM antibodies form virus-immune complexes induced activation of the classical complement pathway. Once initiated at the viral surface, the activation resulted in a massive deposition of C3b, hampering the viral attachment to susceptible host cells (27). As the serum pool used in the present study was tested for the absence of flavivirus-specific antibodies, the natural IgM immunoglobulins are probably directed against insect-like structures such as glycosylation patterns or surface molecules. This is supported by the FACS analysis, which indicated that the serum contained IgM antibodies recognizing insect cells but not cells from human origin. Further, blocking of IgM in the serum pool resulted in complete rescue of the virus. However, these data should be interpreted with caution, as the generated immune complex between IgM and the goat anti-human IgM antibodies may consume complement, which is thus no more available for lysis of the virus. This may also explain, why blocking of IgM by the Ab resulted in a complete rescue of the virus. Our results are in contrast to in vivo WNV experiments by Diamond et al., who demonstrated that natural IgM were insufficient to provide protection against WNV (virus derived from C6/36 cells). This may be due to the experimental setting, as mice in animal facilities may have no contact with mosquitos and therefore probably lack IgM against insect antigens (33).

Our data provide evidence that the source of cells used for cultivation of the virus is an important parameter. For lysis induction of ZIKV derived from insect cells, probably both IgM and direct binding of C1q to the virus may account for activation of the classical pathway. By contrast, ZIKV derived from human cells may interact with C1q only. This would explain why the titer of the virus from insect cells is reduced in the order of two magnitudes, while ZIKV derived from human cells is more stable toward complement mediated neutralization.

Opsonization is discussed as a main factor of complementmediated flavivirus neutralization, as the efficiency of MAC formation may be limited due to the small surface size of the viral particle (5). These results contradict our findings of the effect of C9-depletion on RNA level and in plaque assays, which clearly indicated that the assembly of the MAC was crucial for anti-viral activity by complement-mediated virolysis. Our findings are supported by former in vivo studies of WNV suggesting a contribution of all activation pathways (34–36). The dependence on C9 for the reduction of the viral titer is in contrast to previous observations by Mehlhop et al., who highlighted that neither C5-depleted nor C5-deficient human or mouse sera significantly affected antibody-independent neutralization of WNV, indicating a C5-independent mechanism (37). Opsonization of the viral surface by complement proteins, which may cover viral proteins essential for infection, seems to be sufficient to inhibit infection (37). However, opsonization with purified C1q had no effect on the infectivity of the virus (data not shown). Similarly, covering HIV and parainfluenza virus by C3b seems to reduce infectivity of these viruses (38– 40). However, due to data obtained with C9-depleted sera, we suggest that opsonization itself is not the major mechanism for the reduction of the viral RNA level and infectious titers of ZIKV in our experiments.

#### REFERENCES


Nevertheless, our results indicate that natural IgM antibodies may have a protective function against ZIKV. Whether these natural IgM protect against infection by other mosquito-transmitted viruses in vivo remains to be determined. However, it would be interesting to test whether the mild disease progression observed in many ZIKV infected individuals is partly due to the presence of natural IgM against insect components, which trigger complement activation and thus reduce the viral titer.

#### AUTHOR CONTRIBUTIONS

BS, SB, ZM, DD, IK, and ZB experimental work, data interpretation, drafting the article, and final approval. NT, RW, CS, KS, ES, and GW study design, data interpretation, critical revision of the article, and final approval. HS study design, data interpretation, drafting the article, critical revision of the article, and final approval.

#### FUNDING

This study was supported by the FWF, Vienna Austria (HOROS doctoral Program, W1253-B24) to HS and RW. CS was supported by the FWF (Project Nr. P26117-B20).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02177/full#supplementary-material


glycol tetraacetic acid and MgCl2-ethylene glycol tetraacetic acid. Infect Immun. (1975) 11:1235–43.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Schiela, Bernklau, Malekshahi, Deutschmann, Koske, Banki, Thielens, Würzner, Speth, Weiss, Stiasny, Steinmann and Stoiber. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Ficolin-1 and Ficolin-3 Plasma Levels Are Altered in HIV and HIV/HCV Coinfected Patients From Southern Brazil

Maria Regina Tizzot <sup>1</sup> \*, Kárita Cláudia Freitas Lidani <sup>1</sup> , Fabiana Antunes Andrade<sup>1</sup> , Hellen Weinschutz Mendes <sup>1</sup> , Marcia Holsbach Beltrame<sup>1</sup> , Edna Reiche<sup>2</sup> , Steffen Thiel <sup>3</sup> , Jens C. Jensenius <sup>3</sup> and Iara J. de Messias-Reason<sup>1</sup>

<sup>1</sup> Laboratory of Molecular Immunopathology, Department of Medical Pathology, Federal University of Paraná, Curitiba, Brazil, <sup>2</sup> Clinic Hospital, Estate University of Londrina, Londrina, Brazil, <sup>3</sup> Department of Biomedicine, Aarhus University, Aarhus, Denmark

#### Edited by:

Robert Braidwood Sim, University of Oxford, United Kingdom

#### Reviewed by:

Alexander William Tarr, University of Nottingham, United Kingdom Christine Gaboriaud, UMR5075 Institut de Biologie Structurale (IBS), France

> \*Correspondence: Maria Regina Tizzot retizzot@gmail.com

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 22 May 2018 Accepted: 14 September 2018 Published: 08 October 2018

#### Citation:

Tizzot MR, Lidani KCF, Andrade FA, Mendes HW, Beltrame MH, Reiche E, Thiel S, Jensenius JC and Messias-Reason IJd (2018) Ficolin-1 and Ficolin-3 Plasma Levels Are Altered in HIV and HIV/HCV Coinfected Patients From Southern Brazil. Front. Immunol. 9:2292. doi: 10.3389/fimmu.2018.02292 The complement system is a key component of the innate immune system, participating in the surveillance against infectious agents. Once activated by one of the three different pathways, complement mediates cell lysis, opsonization, signalizes pathogens for phagocytosis and induces the adaptive immune response. The lectin pathway is constituted by several soluble and membrane bound proteins, called pattern recognition molecules (PRM), including mannose binding lectin (MBL), Ficolins-1, -2, and -3, and Collectin 11. These PRMs act on complement activation as recognition molecules of pathogen-associated molecular patterns (PAMPs) such as N-acetylated, found in glycoproteins of viral envelopes. In this study, Ficolin-1 and Ficolin-3 plasma levels were evaluated in 178 HIV patients (93 HIV; 85 HIV/HCV) and 85 controls from southern Brazil. Demographic and clinical-laboratory findings were obtained during medical interview and from medical records. All parameters were assessed by logistic regression, adjusted for age, ancestry, and sex. Significantly lower levels of Ficolin-1 were observed in HIV/HCV coinfected when compared to HIV patients (p = 0.005, median = 516 vs. 667 ng/ul, respectively) and to controls (p < 0.0001, 1186 ng/ul). Ficolin-1 levels were lower in males than in females among HIV patients (p = 0.03) and controls (p = 0.0003), but no association of Ficolin-1 levels with AIDS was observed. On the other hand, Ficolin-3 levels were significantly lower in controls when compared to HIV (p < 0.0001, medians 18,240 vs. 44,030 ng/ml, respectively) and HIV/HCV coinfected (p < 0.0001, 40,351 ng/ml) patients. There was no correlation between Ficolin-1 and Ficolin-3 levels and age, HIV viral load or opportunistic infections. However, Ficolin-3 showed a positive correlation with T CD4 cell counts in HIV monoinfected patients (p = 0.007). We provide here the first assessment of Ficolin-1 and−3 levels in HIV and HIV/HCV coinfected patients, which indicates a distinct role for these pattern recognition molecules in both viral infections.

Keywords: Ficolin-1, Ficolin-3, complement system, HIV infection, hepatitis C virus

# INTRODUCTION

Infection by the Human Immunodeficiency Virus (HIV) leads to a chronic disease and, when not treated, to Acquired Immune Deficiency Syndrome (AIDS). HIV/AIDS currently affects 36.9 million people worldwide, with a high rate of mortality and morbidity (1–3). The number of HIV cases has increased despite the decrease in the number of new infection cases, which was estimated to be 3.3 million in 2002 and 2.3 million people in 2012 (1, 2). The implementation of strategies which includes prevention campaigns, increased access to treatment, and the use of less toxic and more efficient anti-retroviral therapies (highly active antiretroviral therapy- HAART), have all contributed to increase the survival and decrease the infection rates in risk groups, thereby changing the HIV infection epidemiologic profile in the last decade (1, 2).

Coinfection with other pathogens such as hepatitis B virus (HBV) and hepatitis C virus (HCV), and related diseases are known to be aggravating factors in the clinical condition of HIV patients (4–6). In addition, mortality related to chronic liver diseases has increased significantly in the last decade in HIV patients, with about 24% of AIDS mortality being due to endstage liver diseases (ESLD), where 66% were attributed to HCV and 17% to HBV coinfection (7–11). It is important to note that these viruses share the same transmission routes, including sexual and blood, both contributing to the high prevalence of HIV/HCV and HIV/HBV coinfection worldwide (6).

The complement system is a key element of innate immunity which plays a crucial role in the host surveillance against pathogens, including HIV and HCV (12–19). Complement comprises a variety of membrane associated and soluble recognition molecules, known as pattern recognition molecules (PRM), and they include Ficolins (Ficolin-1, Ficolin-2, and Ficolin-3), mannose-binding lectin (MBL) and Collectin 11 (CL-K1) (14). These PRMs are able to recognize a variety of pathogen-associated molecular patterns (PAMPs), such as carbohydrates, N-glycan, LPS and sialic acid residues at the microorganism surface as well as endogenous altered cells and double strand RNAs (12, 14), leading to complement activation. This activation may occur through three different pathways: the classic, the alternative and the lectin pathways, which result in a proteolytic cascade that culminates in multiple biological processes including opsonization and phagocytosis of pathogenic agents and altered cells, production of cytokines, inflammation, and induction of adaptive immune response and homeostasis (14–16).

Ficolin-1 is a non-serum type Ficolin that is found as a membrane-associated protein expressed mainly by monocytes and granulocytes, being the less abundant Ficolin in plasma (average of 1.07µg/mL) (12, 14, 20). On the other hand, Ficolin-3 being the most abundant Ficolin in circulation (average of 26µg/mL) is a serum type protein expressed in alveolar macrophages type II, bronchial epithelial cells and hepatocytes (12, 21). Several studies showed that Ficolins have an important role in viral infections (17–19, 22–31). Denner et al. (31) reported that incubation of mononuclear cells with HIV-1 immunosuppressive gp41 peptides resulted in low Ficolin-1 mRNA concentrations. The authors suggested that Ficolin-1 may be downregulated by the isu domain of gp41 thereby preventing early local innate response, allowing infection and virus replication (31). A possible role for Ficolin-1 in the protection against HCV infection was proposed in a clinical study by Urban et al. (30), who observed the upregulated expression of Ficolin-1 in chronic HCV patients with IL28B rs12979860 CC genotype, which was associated to a favorable response to pegylated interferon-α and ribavirin treatment (30). In addition, Verma et al. (29), demonstrated in an experimental study the binding of Ficolin-3 to influenza A virus which inhibited viral infectivity thereby contributing to host defense against the virus (29). However, there are no prior studies evaluating the role of Ficolin- 1 and -3 in patients with HIV infection, with this now being the first study.

#### MATERIALS AND METHODS

#### Subjects and Samples

This study was approved by the Human Research Ethic Committee of the Clinical Hospital of the Federal University of Parana (1409.074/2007-04). All patients gave written informed consent in accordance with the Declaration of Helsinki. A total of 178 HIV-1 patients (positive for anti-HIV-1, negative for anti-HIV-2 according to Brazilian Ministry of Health guideline) (32) were attending at the Ambulatory of Infectious and Parasitic Diseases at the Clinical Hospital of the Federal University of Parana, in Curitiba, and the Infectology Ambulatory at the Clinical Hospital of the Londrina State University. Among the patients 85 presented coinfection with HCV (Anti-HCV antibody determination by immunoenzimatic micro assay with chemiluminescence QMA Architect—Abbott, USA). As control group, 85 HIV/HCV negative individuals without any clinical complaints were included. The clinical epidemiology data was obtained during the appointments with a questionnaire referring to HIV risk factors and by retrospective analyses of medical records. The following variables were analyzed: age, sex, date of first HIV positive result, possible forms of virus transmission. Same risk factors was also analyzed such as, injected drugs usage, sexual activity, and blood transfusion history as well as clinical progression, T CD4 counts, anti-retroviral treatments and opportunistic diseases (**Table 1**).

#### FICOLIN-1 AND FICOLIN-3 PLASMA LEVELS

Ficolin-1 plasma concentration was determined by the inhouse monoclonal antibody-based method of time-resolved immunofluorometric assay (TRIFMA) and were carried out at the Institute of Medical Microbiology and Immunology at University of Aarhus, Denmark (20) on 178 patients (93 HIV and 85 HIV/HCV) and 85 controls. Ficolin-3 plasma levels were determined by the enzyme-linked immune-sorbent (ELISA) assay HK 340 (Hycult Biotechnology, Uden, The Netherlands) in 79 patients (59 HIV and 20 HIV/HCV) and 85 controls. A total of 10 ml of peripheral whole blood was drawn and separated by



centrifugation for 10 min at 1,000–2,000 × g in plasma samples using vacutainer plastic blood collection tubes with K2EDTA (BD Vacutainer <sup>R</sup> Blood Collection Tubes, Curitiba, Brazil).

#### Data Analyses

Clinical and demographic data was analyzed with GraphPad Prism 3.0 Software (GraphPad Software, Inc., Califórnia, EUA). The distribution of all quantitative variables was evaluated with Kolmogorov-Smirnov and Shapiro-Wilk tests. When normal hypothesis was rejected, medians were compared using nonparametric Mann-Whitney and Kruskal-Wallis and Spearman correlation tests. The associations of Ficolin-1 and -3 levels with HIV and HIV/HCV infection were corrected by multiple logistic regression analysis using STATA 9.2 (StataCorp, Texas, USA). The age, sex, T CD4-cell count and viral load were included as variables in the regression model when the univariate analyses resulted in p < 0.2. Ficolin-1 and -3 levels descriptive statistics were presented with medians and percentiles. P-values lower than 0.05 were considered statistically significant.

#### RESULTS

Ficolin-1 plasma levels were significantly lower in HIV/HCV coinfected when compared to HIV patients (medians of 516 ng/ml vs. 667 ng/ml, respectively; p = 0.005) and to controls (medians of 516 ng/ml vs. 1,186 ng/ml, respectively; p < 0.0001). HIV infected patients also presented lower levels of Ficolin-1 when compared to controls (medians of 667 ng/ml vs. 1186 ng/ml, respectively; p < 0.0001) (**Figure 1A**). These values were corrected for age and sex with logistic regression and were still significant for the findings above.

The presence of AIDS, opportunistic infection and both T CD4 counts and viral load did not alter Ficolin-1 levels in HIV infected or coinfected patients (**Figures 1B–E**). There was also no significant correlation between Ficolin-1 levels and AIDS progression time, which was similar for both groups (medians of 8.8 years for HIV and 9.1 years for HIV/HCV patients; p = 0.4).

Lower Ficolin-1 levels were observed in males compared to females both in controls (medians of 1,079 vs. 1,457 ng/ml, respectively; p = 0.0003) and in HIV patients (medians of 500 vs. 859 ng/ml respectively; p = 0.03), but not in HIV/HCV coinfected subjects (p = 0.6) (**Figure 1F**).

Ficolin-3 levels were significantly increased in HIV patients (medians of 44030 vs. 18240 ng/ml, respectively; p < 0.0001) and HIV/HCV coinfected patients (40351 ng/ml; p < 0.0001) when compared to controls, with no difference between HIV and HIV/HCV coinfected patients (medians of 44,030 vs. 40,351 ng/ml, respectively; p = 0.6) (**Figure 2A**).

The presence of AIDS was associated with lower Ficolin-3 levels in HIV patients (medians of 20,658 vs. 28,306 ng/ml, respectively; p = 0.02) (**Figure 2B**). A statistically significant correlation was observed between Ficolin-3 levels and CD4 cell counts in HIV patients (p = 0.007, r = 0.33) (**Figure 2D**). There was no difference, however, for the presence of opportunistic diseases in HIV or in HIV/HCV patients (**Figure 2C**), HIV viral load (**Figure 2E**), and AIDS progression time in both patient groups.

Male HIV/HCV patients presented higher Ficolin-3 plasma levels compared to female (medians of 47,095 vs. 28,693 ng/ml, respectively; p = 0.01), with the same trend in HIV (medians of 39,110 vs. 28,784 ng/ml, respectively; p = 0.06) but no difference within the control group (p = 0.1) (**Figure 2F**). Ficolin-3 but not Ficolin-1 plasma levels correlated with the age in both groups of patients (p = 0.02, r = −0.27; HIV p = 0.009 r = −0.37, e HIV/HCV p = 0.65 r = −0.04).

# DISCUSSION

We presented here novel evidence that both Ficolin-1 and -3 plasma levels are altered in HIV and HIV/HCV coinfection. Whereas Ficolin-1 levels were found lower in HIV and HIV/HCV coinfected patients, Ficolin-3 were higher in these patients in comparison to controls. Our results indicate that Ficolin-1 and Ficolin-3 may operate as a PRM in a distinct manner facing HIV infection and HIV/HCV coinfection.

These findings are in fact in accordance with the notion that Ficolins—although presenting high structural and molecular homology and specification overlap to PAMPs binding—exhibit differences in functional activities and in the pattern of tissue expression, with consequences in the potential to activate complement as well having other roles in the immune response (14). Ficolins can interact with viral glycoproteins which are constituted of N-Acetylglucosamine (GlcNAc), such as gp120 and gp41, which are essential for cell binding and infection onset, but is also a target for Ficolin-1 and Ficolin-3 (17–19). Considering

that PRMs bind to these viral surface glycoproteins, they may have a role in the process of HIV infection. In fact, complement may play different activities in HIV pathogenesis, such as blocking virus entrance into the cell, signalizing phagocytosis, inflammation, as well as forming the lytic complex at the infected cell membranes (17–19, 22–31). On the other hand, viruses may bind to complement proteins impairing an effective adaptive immune response thereby facilitating its entrance into target cells through complement receptors (CR) (17–19, 22).

Studies relating the role of Ficolin-1 and -3 in viral infections are scarce (29–31) and to our knowledge, this is the first study evaluating the levels of Ficolin-1 and -3 in HIV and HIV/HCV coinfected patients. The low Ficolin-1 levels observed in HIV and HIV/HCV patients suggests possible protein consumption due to viral infection. It is known that neutrophil autocrine Ficolin-1 can bind to CD43 (a neutrophil membrane sialoprotein) inducing neutrophil adhesion at the beginning of an inflammatory response (14, 33). On the other hand, in the late phase Ficolin-1 was shown to have strong affinity to C-reactive protein resulting in downregulation of pro-inflammatory cytokines (34, 35). In addition, Ficolin-1 levels could be reduced due to downregulation effect of the viruses as described for HIV gp41 immunosuppressive (isu) domain (31). Interestingly, Ficolin-1 levels were even lower in HIV/HCV patients when compared to HIV. It is known that HCV coinfection is an aggravating factor in the clinical condition of HIV patient exacerbating the existing inflammatory process (4–6), and that complement can interact with glycoproteins of

both virus (17, 19, 25). Thus, our results suggest that low Ficolin-1 levels might be a consequence of the interaction with both HCV and HIV in addition to an immunomodulatory effect due to chronic inflammatory process.

Meanwhile, the elevated Ficolin-3 levels found in HIV and HIV/HCV patients compared to controls may derive from compensatory mechanisms of upregulation of this protein due to its interaction with viral glycoproteins as well as complement activation. Ficolin-3 binding to HIV or HCV has not yet been reported, however, it is known that HIV gp120 is rich in fucose, which is a particular ligand of Ficolin-3 (36). Still, high expression of Ficolin-3 found in both viral infections could be related to the inflammatory status seen in co-infected patients, contributing to the chronic process of these conditions. Elevated Ficolin-3 levels were also observed in ovarian tumor patients (37), related to shorter graftsurvival after kidney transplantation (38), in patients with Leprosy (39) and Systemic Lupus Erythematosus (40, 41), all conditions associated with inflammatory process.

Higher Ficolin-1 levels in female HIV patients and controls corroborates previous findings (42), but, such difference was not observed in HCV coinfected patients, what could be due to the higher male frequency in this group. Ficolin-3 levels were higher in males only in the co-infected group, a sex difference previously observed in healthy subjects (42).

This study has some limitations. First, polymorphisms shown association to Ficolin-1 (rs7857015, rs10120023, rs10117466) (43, 44) and Ficolin-3 (rs532781899, rs28362807, and rs4494157) (39) concentrations, that could have contributed to the variation of protein concentration in HIV and HIV/HCV patients, were not assessed. Future studies in this cohort of patients should include the investigation of these variants. Second, although Ficolins concentration showed significant results among the groups the relevance of these findings should be confirmed in a larger sample of patients, and, in experimental studies showing Ficolins 1 and 3 interaction with both HIV and HCV. Third, the status of HCV co-infection was based only on anti-HCV test, which is an evidence for prior exposure to HCV. Since all patients were attending the HIV ambulatory whose routine for HCV is based on serology, molecular tests were not available at the time of the study. In order to established direct evidence of current infection both the presence of HCV RNA and a longitudinal follow-up in HIV/HCV patients should be considered.

Nevertheless, our data represent a pioneer study on the role of Ficolin-1 and -3 in HIV/HCV infections. This novel finding

#### REFERENCES


suggests that these proteins contribute in a different manner to host defense against these viruses and may be helpful for future studies in understanding the function of Ficolins in HIV/HCV infection as well as in the development of new therapeutic targets.

#### AUTHOR CONTRIBUTIONS

IM-R, ST, JJ, and MT contributed with conception and design of the study. MT, HM, MB, and FA executed laboratory procedures. KL performed the statistical analysis. MT, KL, FA, and MB wrote the original draft of the manuscript. ER performed the recruitment of patients. All authors contributed to manuscript revision, read, and approved the submitted version.

#### ACKNOWLEDGMENTS

We gratefully acknowledge the subjects of this investigation for their consent in participating in the study, and to the staff of the Laboratório de Imunopatologia Molecular in Curitiba for DNA extraction. We also thank the Institute of Medical Microbiology and Immunology at University of Aarhus, Denmark for collaboration in plasma Ficolin-1 quantification. This work was supported by grants and scholarships from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Tizzot, Lidani, Andrade, Mendes, Beltrame, Reiche, Thiel, Jensenius and Messias-Reason. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Evasion of Classical Complement Pathway Activation on *Plasmodium falciparum*-Infected Erythrocytes Opsonized by PfEMP1-Specific IgG

#### *Edited by:*

Francesca Granucci, Università degli studi di Milano Bicocca, Italy

#### *Reviewed by:*

Alvaro Diaz, Universidad de la República, Uruguay Thomas Vorup-Jensen, Aarhus University, Denmark

*\*Correspondence:*

Lars Hviid lhviid@sund.ku.dk Peter Garred peter.garred@regionh.dk

#### *†Present Address:*

Mads Delbo Larsen, Department of Experimental Immunohematology, Sanquin Research and Landsteiner Laboratory, Amsterdam UMC, University of Amsterdam, Amsterdam, Netherlands

> ‡These authors shared senior authorship

#### *Specialty section:*

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

*Received:* 13 July 2018 *Accepted:* 13 December 2018 *Published:* 07 January 2019

#### *Citation:*

Larsen MD, Quintana MdP, Ditlev SB, Bayarri-Olmos R, Ofori MF, Hviid L and Garred P (2019) Evasion of Classical Complement Pathway Activation on Plasmodium falciparum-Infected Erythrocytes Opsonized by PfEMP1-Specific IgG. Front. Immunol. 9:3088. doi: 10.3389/fimmu.2018.03088 Mads Delbo Larsen1,2†, Maria del Pilar Quintana<sup>1</sup> , Sisse Bolm Ditlev <sup>1</sup> , Rafael Bayarri-Olmos <sup>2</sup> , Michael Fokuo Ofori <sup>3</sup> , Lars Hviid1,4 \* ‡ and Peter Garred<sup>2</sup> \* ‡

<sup>1</sup> Centre for Medical Parasitology, Department of Immunology and Microbiology, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark, <sup>2</sup> Laboratory of Molecular Medicine, Department of Clinical Immunology, Rigshospitalet, Copenhagen, Denmark, <sup>3</sup> Department of Immunology, Noguchi Memorial Institute of Medical Research, University of Ghana, Accra, Ghana, <sup>4</sup> Centre for Medical Parasitology, Department of Infectious Diseases, Rigshospitalet, Copenhagen, Denmark

Members of the PfEMP1 protein family are expressed on the surface of P. falciparum-infected erythrocytes (IEs), where they contribute to the pathogenesis of malaria and are important targets of acquired immunity. Although the PfEMP1-specific antibody response is dominated by the opsonizing and complement-fixing subclasses IgG1 and IgG3, activation of the classical complement pathway by antibody-opsonized IEs does not appear to be a major immune effector mechanism. To study the molecular background for this, we used ELISA and flow cytometry to assess activation of the classical component pathway by recombinant and native PfEMP1 antigen opsonized by polyclonal and monoclonal PfEMP1-specific human IgG. Polyclonal IgG specific for VAR2CSA-type PfEMP1 purified from a pool of human immune plasma efficiently activated the classical complement pathway when bound to recombinant PfEMP1 in ELISA. In contrast, no activation of complement could be detected by flow cytometry when the same IgG preparation was used to opsonize IEs expressing the corresponding native PfEMP1 antigen. After engineering of a VAR2CSA-specific monoclonal antibody to facilitate its on-target hexamerization, complement activation was detectable in an ELISA optimized for uniform orientation of the immobilized antigen. In contrast, the antibody remained unable to activate complement when bound to native VAR2CSA on IEs. Our data suggest that the display of PfEMP1 proteins on IEs is optimized to prevent activation of the classical complement pathway, and thus represents a hitherto unappreciated parasite strategy to evade acquired immunity to malaria.

Keywords: malaria, complement, evasion, PfEMP1, antibodies, knobs

# INTRODUCTION

Malaria remains a major health problem with an estimated 219 million cases and 435,000 deaths in 2017 alone (1). The human disease is caused by several protozoan parasites of the genus Plasmodium, but P. falciparum is responsible for most severe cases and essentially all malaria mortality (2).

The particular virulence of P. falciparum is related to the unique ability of this parasite to express members of a family of clonally variant surface antigens called P. falciparum erythrocyte membrane protein 1 (PfEMP1) on the surface of the infected erythrocytes (IEs) (3). This enables sequestration of IEs in the microvasculature, mediated by interaction of PfEMP1 with vascular host receptors such as CD36, endothelial protein C receptor (EPCR), and oncofetal chondroitin sulfate (4–7). The ensuing accumulation of IEs in tissues can lead to severe disease, precipitated by excessive inflammation and circulatory dysfunction.

The PfEMP1 antigens are important targets of naturally acquired immunity to P. falciparum malaria, and semi-immune individuals living in areas of stable parasite transmission possess a broad repertoire of PfEMP1-specific antibodies, dominated by the opsonizing and complement-fixing subclasses IgG1 and IgG3 (8, 9). It is therefore noteworthy that activation of the classical complement pathway does not appear to play a major role in acquired immunity to IEs, although acquired immunity leads to activation of complement by antibody-opsonized sporozoites and merozoites (10, 11). Indeed, P. falciparum has evolved several mechanisms to evade activation of the alternative and classical pathways by hijacking soluble complement regulators to these developmental stages, emphasizing the clinical importance of complement-mediated attack on malaria parasites (12–16). The display of PfEMP1 on the IE surface is normally restricted to electron-dense protrusions known as "knobs" (17). Although knob-less variants can express PfEMP1 and thrive in vitro, knob expression is generally thought to be required for P. falciparum survival in vivo, and the protection from falciparum malaria afforded by several hemoglobinopathies is thought related to abnormal knob formation on IEs (18). Nevertheless, the role of knobs in P. falciparum survival remains unclear. In this article, we investigate the hypothesis that the knob restriction of PfEMP1 on the IE surface may have evolved to prevent classical complement activation by preventing on-target hexamerization of IgG (19). We show that although polyclonal, and even monoclonal, IgG can activate the classical complement pathway when bound to surface-bound recombinant PfEMP1, such activation does not occur when the IgG is bound to native PfEMP1 expressed on the IE surface.

# MATERIALS AND METHODS

#### Recombinant Protein Production

The full-length ectodomain of the VAR2CSA-type PfEMP1 IT4VAR04 (FV2) was produced in ExpiCHO-S cells as a recombinant C-terminal histidine-tagged protein, as described elsewhere (20).

A recombinant version of the human monoclonal IgG1 antibody (mAb) PAM1.4 (21), which is specific for a conformational epitope in VAR2CSA-type PfEMP1, was produced as described elsewhere by cloning and inserting the variable domains of the antibody into plasmids encoding the constant regions of the γ1-chain and the κ-chain (22). This antibody has previously been shown to bind both plasticimmobilized FV2 in ELISA and the corresponding native antigen on the IE surface (20).

We also produced two variants of this antibody (PAM1.4-E345K and PAM1.4-E430G) by introducing single-nucleotide substitutions in the γ1-chain, using the QuickChange site-directed mutagenesis kit (Agilent) according to manufacturer's instructions. Briefly, mutations were introduced by a PCR reaction of the entire plasmid with a high-fidelity DNA-polymerase and a complementary primer pair with the desired mutation (E345K forward: 5′ - CCAAAGGGCAGCCCCGAAAACCACAGGTGTA-3′ ; E430G forward:5′ -CCGTGATGCATGGGCTCTGCACAACCACT-3′ ; substitutions are underlined). Plasmids were subsequently sequenced to confirm the introduction of the substitutions. All the recombinant antibodies were produced in human embryonic kidney cells (293-F; Gibco) according to manufacturer's instructions. Briefly, the cells were grown to ∼1.5 × 10<sup>6</sup> cells/mL, and adjusted to 1 × 10<sup>6</sup> cells/mL 1 h prior to transfection, and co-transfected with heavy- and light chain plasmids (0.5 µg DNA/plasmid/mL culture) and FreeStyle MAX reagent (1 µL/µg DNA). Cultures were incubated for seven days before harvesting the supernatant. The recombinant antibodies were purified using Protein G-coupled agarose beads (Pierce).

# IgG Purification From Human Plasma Samples

Total IgG was purified from pools of plasma from ten Ghanaian donors with natural exposure to P. falciparum parasites expressing VAR2CSA-type PfEMP1. The samples were collected in a previous study approved by the Institutional Review Board of Noguchi Memorial Institute for Medical Research, University of Ghana (study no. 038/10-11) (23). Samples were collected only after consent had been obtained in writing from each participant. One pool consisted of the ten available samples with the highest reactivities toward FV2, while the other consisted of the ten available samples with the lowest FV2 reactivities. IgG was purified from each pool of plasma using Gammabind G sepharose (GE Healthcare), using standard methodology. Affinity purification on FV2 was not employed, to allow direct comparison of the pools with high and low FV2 reactivity. However, previous studies have shown that PfEMP1 is the dominant IE target antigen of naturally acquired IgG (24).

#### Classical Complement Pathway Activation ELISA

Flat-bottomed 96-well plates (Nunc) were coated (4◦C, overnight in PBS) with FV2 (2µg/mL), mAbs (PAM1.4; PAM1.4 E345K; PAM1.4 E430G), purified human IgG (Invitrogen), or human serum albumin (HSA; 10µg/mL; Sigma Aldrich). Wells were blocked with PBS supplemented with TWEEN (0.05%) and BSA (1%) and subsequently incubated (1 h on shaker, as all subsequent

**Abbreviations:** AU, arbitrary units; BSA, bovine serum albumin; FV2, fulllength ectodomain of the VAR2CSA-type PfEMP1 IT4VAR04; HAS, human serum albumin; IE, infected erythrocyte; mAb, monoclonal antibody; NHS, non-immune human serum; PfEMP1, Plasmodium falciparum erythrocyte membrane protein-1; TCC, terminal complement complex.

steps) with the above-mentioned VAR2CSA-specific mAbs or with control reagents [mAb AB01, specific for a non-VAR2CSAtype PfEMP1 (25) or polyclonal rabbit anti-HSA (Dako)]. In some experiments, 96–well plates pre-coated with nickel, preblocked with BSA and coated with FV2 (5µg/mL) in PBS

standard deviations (error bars) of three individual experiment are shown.

supplemented with TWEEN (0.05%) were used (Pierce). After incubation with antibody reagents, the plates were washed four times in Barbital-Tween buffer [sodium barbital (4 mM), NaCl (145 mM), CaCl<sup>2</sup> (2.64 mM), MgCl<sup>2</sup> (2.12 mM), Tween (0.05%)] and incubated (1 h, 37◦C) in the same buffer supplemented with non-immune human serum (NHS; 1%) as source of complement components. After washing as above, bound complement components were detected with polyclonal rabbit anti-human C1q (2µg/mL; Dako), rabbit anti-human C4c (1µg/mL; Dako), rabbit anti-human C3c (1µg/mL; Dako), or monoclonal mouse-anti human-terminal complement complex (TCC; 1 µg/L; clone ae11). All the complement-specific antibody reagents were biotinylated prior to the experiments. After the last washing as above, bound antibody was detected by incubation with streptavidin-conjugated horse radish peroxidase (1:2,000; GE Healthcare), followed by TMB ONE (ECO-TEK). The color reaction was terminated with H2SO<sup>4</sup> (0.2 M), and quantified at 450 nm. Arbitrary units (AU) were calculated as (ODTest - ODBlank) / (ODControl - ODBlank).

#### *P. falciparum* Culture *in vitro* and Selection for Expression of PfEMP1 and Knobs

The P. falciparum laboratory isolate IT4/FCR3 was cultured in serum-free medium as described elsewhere (26).

Parasites were selected by antibody panning and density separation monthly to ensure expression of the VAR2CSA-type PfEMP1 IT4VAR04 and IE surface knobs as described previously (27). Briefly, cultures with primarily late trophozoite-stage IEs were incubated (20 min, 37◦C) in culture medium supplemented with gelatin (0.75%) to separate late trophozoite-stage knobby IEs from uninfected erythrocytes and ring-stage IEs. The late trophozoite-stage IEs were then incubated with DynaBeads A (DYNAL) that had been pre-incubated (30 min) with saturating amounts of PAM1.4 antibody. Bound IEs were isolated using a DynaMag (DYNAL).

#### Classical Complement Activation on IEs

Late trophozoite-stage IEs were purified by magnet-activated cell sorting (Miltenyi Biotec) in PBS supplemented with BSA (1%) (28). The purified IEs (2 × 10<sup>6</sup> cells/mL) were incubated (30 min, 4 ◦C) with mAbs (10µg/mL) or purified pooled human IgG (1 mg/mL), followed by incubation (1 h, 37◦C) with NHS (1%) and compstatin (6 nM; Tocris) (to inhibit cleavage of C3). As positive and negative controls, type A and 0 erythrocytes were incubated with type 0 NHS and compstatin.

Complement components were detected by incubation (4◦C, 30 min) with the same antibodies as above, but at different concentrations (anti-C1q: 30µg/mL; anti- C4c: 0.5µg/mL), followed by incubation with FITC-conjugated goat anti-rabbit IgG (1:150; Vector) and ethidium bromide (2µg/mL; to visualize parasite DNA). Binding of human IgG to the IEs was detected in a similar way, using FITC-conjugated goat anti-human IgG (1:150; Jackson Immuno Research) (28). All incubations, dilutions, and washes were done in PBS supplemented with BSA (1%), except for the incubation with NHS and compstatin, which was done in BSA-supplemented VBS++ buffer containing Ca2<sup>+</sup> and Mg2<sup>+</sup> (Complement Technology). Samples were run on a Cytomics FC500 flow cytometer (Beckman Coulter). Single ethidium bromide-positive cells were analyzed for complement components and antibody labeling by FlowLogic (Inivai Technologies).

# RESULTS

#### Human Specific IgG Bound to Immobilized Recombinant PfEMP1 Activates the Classical Complement Pathway

To our knowledge, the ability of PfEMP1-specific human IgG to activate the classical complement pathway has not been reported previously. Binding of C1q (**Figure 1A**), deposition of C4 (**Figure 1B**) and C3 (**Figure 1C**), and formation of TCC (**Figure 1D**) were detected in recombinant FV2-ELISA after incubation with human IgG purified from a plasma pool with known high FV2-reactivity, followed by NHS as a source of complement components. When IgG from the pool of lowreactive plasma was used instead of the highly FV2-reactive IgG, no complement deposition was detected (**Figure 1**). We conclude that naturally acquired PfEMP1-reactive IgG is able to activate the classical complement pathway when bound to recombinant FV2 that has been randomly immobilized on plastic.

#### Human Specific IgG Bound to Native PfEMP1 on Infected Erythrocytes Does Not Activate the Classical Complement Pathway

We next assessed the ability of human IgG purified from the same plasma pools to activate complement when bound to IEs

expressing the native PfEMP1 (IT4VAR04) represented by FV2. IgG purified from the highly FV2-reactive pool efficiently labeled the IEs, in contrast to the low FV2-reactivity IgG (**Figure 2A**). However, essentially no binding of C1q (**Figure 2B**) or deposition of C4 (**Figure 2C**) could be detected after incubation of IEs with either IgG preparation. The functionality of the assay was confirmed in experiments using uninfected type A and type 0 erythrocytes incubated with type 0 NHS (**Figure S1**). These findings suggest that the distribution of PfEMP1 on IEs inhibits classical complement activation, possibly due to the clustered, knob-restricted distribution of the antigen.

### Activation of Complement by Monoclonal IgG Bound to Immobilized Recombinant PfEMP1 Depends on Antigen Orientation

It was recently reported that complement activation by IgG requires on-target, Fc-dependent hexamerization of the antibody (19). We therefore proceeded to test whether the lack of complement activation by PfEMP1-specific IgG bound to IEs was due to an inability of the antibodies to form hexamers after binding to native PfEMP1 on the IE surface. To do so, we used IEs selected for surface expression of IT4VAR04, and a mAb (PAM1.4) with specificity for VAR2CSA-type PfEMP1 (including IT4VAR04). We also used two variants of this mAb (PAM1.4- E345K and PAM1.4-E430G), where we had introduced mutations in the Fc-region (E345K and E430G, respectively), known to enhance on-target hexamerization of IgG (19, 29).

All three mAbs had similar ability to activate complement when coated directly to ELISA plates, as binding of C1q (**Figure 3A**), as well as deposition of C4 (**Figure 3B**), C3 (**Figure 3C**), and TCC (**Figure 3D**) could be detected in a concentration-dependent manner. This agrees with the observation that enhanced complement activation by hexamerization-improved Fc mutants requires binding of the antibodies to their cognate antigen (29). To confirm this requirement directly, the mAbs were next applied in a setup where the ELISA plates were first coated with FV2 as in the experiments with purified immune IgG above. However, neither PAM1.4 nor the Fc-mutated variants of the mAb activated complement in this setup (**Figure S2**).

In the studies first identifying the complement activationenhancing mutations, the assays involved cell lines overexpressing the targeted antigen (19, 29). We therefore speculated that the lack of complement activation in our setup with

recombinant FV2 bound to plastic might be related to the random orientation of the immobilized antigen, in contrast to the usually uniform display of a given antigen on the surface of cells. To approximate such an ordered display, we exploited the fact that our FV2 protein has a C-terminal poly-histidine tag, and immobilized the recombinant protein on ELISA plates pre-coated with nickel. Because the nickel ions bind to the FV2 poly-histidine tags, this should facilitate homogeneous orientation of the antigen similar to the orientation of the native protein on IEs, although the distribution of the FV2 would remain more homogenous than the knob-restricted distribution of native PfEMP1 proteins on IEs. In this setup, binding of C1q (**Figure 4A**), and in particular deposition of C4 (**Figure 4B**) and C3 (**Figure 4C**), could be detected in a concentration-dependent manner for all three mAbs, whereas TCC formation was minimal (**Figure 4D**). Although this might theoretically be due to activation of the alternative pathway by binding of mannanbinding lectin to glycosylation determinants on the recombinant

antibodies, this appears unlikely. Firstly, the glycosylation pattern of the cells used to express the recombinant antibodies is very similar to native human IgG (30). Secondly, the hexamerizationenhanced Fc-region mutants, in particular PAM1.4-E430G, were superior to PAM1.4 (if the binding of C4 and C3 were due to activation of the lectin rather than the classical pathway, similar activation by wildtype PAM1.4 and the two hexamerizationenhanced mutants would be expected). When this improved assay was used to test complement activation by purified immune IgG, C4 deposition could be detected with both the high and the low FV2-reactive preparation, although the highly FV2-reactive IgG was superior (**Figure S3**).

Our data confirm that the E345K and E430G mutations in the Fc-region enhance the ability of IgG to activate complement, probably by facilitating on-target hexamerization. Furthermore, the results highlight antigen orientation as an important parameter in in vitro assays of classical complement pathway activation.

results of two independent experiments are shown.

#### Hexamerization-Enhancing Fc Mutations Do Not Lead to Activation of Complement by Monoclonal PfEMP1-Specific IgG Bound to Infected Erythrocytes

To investigate whether the lack of complement activation on IEs (**Figure 2**) could be overcome by enhancing the antibody capacity for on-target hexamerization, we tested the ability of PAM1.4 and the two Fc mutants to activate complement when bound to IEs. We did not detect C1q binding (**Figure 5A**) or C4 deposition (**Figure 5B**) with any of the mAbs. It thus appears that the distribution of native PfEMP1 prevents hexamerization of monoclonal IgG in a way that could not be overcome by Fc mutations enhancing the ability of PAM1.4 to form hexamers. Although we cannot exclude that other mAbs might be able to form hexamers when bound to PfEMP1 on the IE surface, or that on-target hexamerization might occur with Fc-mutated polyclonal IgG preparations, it seems most likely that PfEMP1 is distributed on the IE surface in a way that prevents the interactions among Fc regions of adjacent IgG molecules that would facilitate binding of C1q and activation of the classical complement pathway.

# DISCUSSION

IgG antibodies specific for the asexual parasites multiplying in the blood are a key element in naturally acquired protective immunity to P. falciparum malaria (31, 32). Antibodies to antigens on the surface of the IEs – in particular PfEMP1 – are of particular importance in this respect (3). This is probably due to their ability to block the vascular sequestration of IEs, which can otherwise cause inflammation and circulatory dysfunction (33). Unsurprisingly, P. falciparum has evolved a range of strategies to avoid PfEMP1-specific immunity, such as antigen polymorphism, clonal antigenic variation, acquisition of soluble host factors etc. (34).

The antibody response to P. falciparum asexual blood stage antigens, including the VAR2CSA-type PfEMP1 studied here, is dominated by IgG1 and IgG3 (35). It is therefore likely that phagocytosis of merozoites and IEs opsonized by antibody and complement also contribute significantly to parasite clearance. Although the role of the complement system in P. falciparum infections has been the focus of several recent studies, most have focused on the alternative pathway. The ability of IEs (13), merozoites (14), and gametocytes (12) to acquire the soluble complement regulator Factor H to their surface thus clearly suggests that this activation pathway is important in controlling parasitemia, and likely has forced the parasites to evolve strategies to evade this host defense. A very recent study indicates that this host-parasite tug-of-war is even more complicated, and that Factor H-related protein 1 may be involved in a host effort to overcome malaria parasite evasion of complement attack by acquisition of Factor H (36).

The clinical importance of classical complement pathway activation following opsonization by IgG has been less studied, although binding of C1q to IgG-coated merozoites and sporozoites has been associated with protection of malaria in patients from Oceania and Africa (10, 11). In addition, evasion of such immunity by hijacking of C1-inhibitor by merozoites was recently reported (15). The clinical relevance of antibodydependent complement attack on merozoites may be limited by the fact that this free-living stage is only exposed to antibody and complement for 1-2 min before invading a new erythrocyte. It is questionable whether this is long enough for antibody binding and classical complement attack, as in vitro experiments required >2.5 min. before C3 deposition could be demonstrated (15). By then, many merozoites would be expected to have safely reinvaded (37). In the case of sporozoites, the parasite is exposed to antibody and complement for longer (38), and it therefore seems more likely that complement plays a decisive role in the immune response against this developmental stage. However, the intra-erythrocytic parasites should theoretically be vulnerable to classical complement attack for the ∼30 of the 48-h asexual life cycle, where they express PfEMP1 on the IE surface to facilitate tissue sequestration and avoidance of destruction in the spleen (39). This notwithstanding, little is known about classical complement attack on IEs, let alone parasite strategies to evade this threat.

Here, we show that human IgG purified from plasma and having high reactivity to PfEMP1 can activate the classical complement pathway in ELISA. However, no complement activation was seen when the same IgG was bound to the corresponding native PfEMP1 on the surface of IEs. The molecular dimensions of IgG and PfEMP1 molecules, combined with estimates of the number of PfEMP1 molecules per IE (40–46) suggest that on-target hexamerization [required for efficient activation of the classical complement cascade by IgG (19)] would occur if the PfEMP1 molecules were evenly distributed over the IE surface. We therefore hypothesized that the lack of activation is related to the knob-restricted expression of PfEMP1 on the IE surface, which prevents ontarget hexamerization of IgG molecules bound to PfEMP1 molecules on neighboring knobs, as these are too far apart. The clustered distribution might thus represent a hitherto unidentified strategy by P. falciparum to evade acquired, IgG-mediated protective immunity. However, the molecular dimensions of knobs make on-target hexamerization of IgG bound to different PfEMP1 molecules within a given knob theoretically possible. We therefore produced two PfEMP1 specific mAbs with substitutions in the Fc-region that enhance their capacity for on-target hexamerization and complement activation (19, 29). Although we could demonstrate enhanced complement-activating capacity of Fc-mutated mAbs by ELISA, the mutations did not suffice to activate complement following binding of the mAbs to native PfEMP1 on the IE surface. Even with polyclonal immune IgG containing IgG with specificity for the many antibody epitopes that exist in VAR2CSA (21), we did not find evidence of activation of the classical complement cascade at the IE surface. Although the reason for the above observations is not known, the simplest explanation is that PfEMP1 molecules are not evenly distributed, even within the confines of a single knob, but are instead clustered together. Whether this is the case is not currently known, however.

Erythrocytes are inherently susceptible to complement attack, and they therefore possess endogenous membrane-bound complement regulators such as decay-accelerating factor (CD55) and protectin (CD59) to prevent inadvertent phagocytosis and lysis of complement-opsonized erythrocytes. Although CD59 has

#### REFERENCES


been reported as the factor preventing complement-mediated lysis of IEs (47), IE lysis is likely to be less important than opsonization for phagocytosis. In this study, we decided to focus on complement components upstream of the erythrocyte membrane-bound complement regulators' point of action, to avoid complications imposed by the need to enzymatically remove membrane-bound regulators.

To conclude, we report that although PfEMP1-specific IgG can activate the classical complement pathway in a system where the antigens are homogeneously distributed, this appears not to happen at the IE surface, where PfEMP1 display is restricted to well-defined knobs. The most parsimonious explanation for this discrepancy is that the focal display of native PfEMP1 interferes with the on-target hexamerization of IgG, which is a requirement for binding of C1q and activation of the classical complement cascade. The knob-restricted display may thus represent a hitherto unrealized strategy of P. falciparum to evade acquired protective immunity.

#### AUTHOR CONTRIBUTIONS

ML, LH, and PG formulated the hypothesis and designed the experiments and analyzed the data and wrote the paper. ML, MQ, SD, and RB-O produced the recombinant proteins. ML carried out all the experiments. MO was responsible for collection of the biological samples. All authors reviewed and edited the manuscript.

#### FUNDING

This work was supported by grants from the Danish Research Foundation of Independent Research (DFF-4183-00539 and DFF-6110-00489), the Sven Andersen Research Foundation, Novo Nordisk Research Foundation (NNF15OC0017654) and Rigshospitalet (R102-A4174). Sample collection in Ghana was supported by grants from the Consultative Committee for Development Research (DFC-12-081RH) and the Ghana Educational Trust Fund.

#### ACKNOWLEDGMENTS

We thank all plasma donors, health-care workers, and research staff involved in the field study in Ghana. We thank Maiken Visti and Anne Corfitz for excellent technical assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.03088/full#supplementary-material


receptor for adherence to CD36, thrombospondin, and intercellular adhesion molecule 1. Proc Natl Acad Sci USA. (1996) 93:3497–502.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Larsen, Quintana, Ditlev, Bayarri-Olmos, Ofori, Hviid and Garred. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Complement Activation as a Helping Hand for Inflammophilic Pathogens and Cancer

#### Marcin Okrój <sup>1</sup> \* and Jan Potempa2,3 \*

<sup>1</sup> Department of Medical Biotechnology, Intercollegiate Faculty of Biotechnology, University of Gdansk and Medical University ´ of Gdansk, Gda ´ nsk, Poland, ´ <sup>2</sup> Department of Oral Immunology and Infectious Diseases, University of Louisville School of Dentistry, Louisville, KY, United States, <sup>3</sup> Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Kraków, Poland

The complement system, an evolutionarily ancient component of innate immunity, is capable of protecting hosts from invading pathogens, either directly, by lysis of target cells, or indirectly, by mobilization of host immune mechanisms. However, this potentially cytotoxic cascade must be tightly regulated, since improperly controlled complement can damage healthy cells and tissues. The practical importance of this axis is highlighted when impairment of complement regulators or bacterial mechanisms of complement evasion result in pathogenic conditions. Recognition of complement as a "double-edged sword" is widely acknowledged, but another, currently underappreciated aspect of complement function has emerged as an important player in homeostatic balance—the dual outcome of complement-mediated inflammation. In most cases, the proinflammatory properties of complement are beneficial to the host. However, certain pathogens have developed the ability to utilize local inflammation as a source of nutrients and as a way to establish a niche for further colonization. Such a strategy can be illustrated in the example of periodontitis. Interestingly, certain tumors also seem to benefit from complement activation products, which promote a proangiogenic and immunosuppressive microenvironment.

Keywords: inflamation, periodontits, cancer, Porphyromonas gingivalis, complement activation

# INTRODUCTION

The term "inflammo-philic" (=loving or attracting inflammation) was introduced in 2014 by George Hajishengallis to describe dysbiotic microbiome on the tooth surface below the gum line, which thrive in the inflammatory environment of periodontal pockets (1). Remarkably, as described in details later in this review, bacteria responsible for initiation and progression of periodontitis (periodontopathogens) have the unique ability to manipulate the complement system to disengage bacterial clearance from inflammation.

In general, the local inflammatory response to bacterial and fungal pathogens triggered by complement activation is absolutely essential to eliminate invaders (2, 3). Therefore, all successful pathogens developed a large variety of means to interfere with complement activation and/or hinder complement-dependent bacterial clearance mechanisms (2–6) (**Table 1**). Unfortunately, if inflammatory reaction triggered by pathogens escape the control it becomes highly detrimental to the host as illustrated by invasive candidiasis [Candida albicans (7)], meningitis [Neisseria meningitidis (8)], and sepsis [N. meningitidis (8), Staphylococcus aureus (9), and Streptococcus pyogenes (10)]. It needs to be kept in mind that an overwhelming inflammatory response and a dysregulated immune response to these infections is by no means the manifestation of an

#### Edited by:

Maciej Cedzynski, Institute for Medical Biology (PAN), Poland

#### Reviewed by:

George Hajishengallis, University of Pennsylvania, United States Barbara Bottazzi, Humanitas Clinical and Research Center, Italy Peter Kraiczy, Goethe-Universität Frankfurt am Main, Germany

#### \*Correspondence:

Marcin Okrój marcin.okroj@gumed.edu.pl Jan Potempa jspote01@louisville.edu

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 21 August 2018 Accepted: 18 December 2018 Published: 10 January 2019

#### Citation:

Okrój M and Potempa J (2019) Complement Activation as a Helping Hand for Inflammophilic Pathogens and Cancer. Front. Immunol. 9:3125. doi: 10.3389/fimmu.2018.03125

**318**

TABLE 1 | Exemplary complement evasion strategies used by microbes.


#### TABLE 1 | Continued


(s), soluble protein/molecule.

(b), surface—bound protein/molecule. For original references see the review articles with reference numbers 2–6.

inflammophilic character of these pathogens since the controlled, local inflammation is protective against these pathogens (9). Therefore, the pathogenic strategy to endure inflammation but in the same time to take advantage of it, seems to be limited to periodontopathogens. It is fascinating that the apparently similar strategy is employed by cancer and in both cases exploitation of the complement system underlines pathology.

#### COMPLEMENT SYSTEM

The complement system is one of the oldest mechanisms of immunity. Its essential components, such as the C3 molecule, have existed through more than 500 million years of evolution (11). A primitive complement system probably appeared in the common ancestor of eumetazoa, and its original role was limited to opsonization and the induction of inflammation. Genetic events like the duplication-based appearance of pathway specific components (e.g., factor B and C2) and the gain of terminal pathway constituents (C5–C9) allowed the primodal complement system to evolve into a an advanced and complex defense system capable not only of promoting osmotic lysis of target cells, anaphylaxis, and phagocytosis, but also of crosstalk with other systems (e.g., coagulation) and signaling pathways (e.g., Toll-like receptors) involved in the maintenance of bodily homeostasis (12, 13). There are three independent complement cascades, the evolutionarily older alternative and lectin pathway (with basic elements like C3, MASPs, and factor B existing in invertebrates) and the relatively younger classical pathway developed in jawed vertebrates (11). The alternative complement pathway is constitutively active at a low level due to the spontaneous breakdown of C3 into anaphylatoxin C3a and the active C3b fragment, which activate downstream steps in the cascade. Therefore, propagation of the alternative pathway does not depend on specific activation but relies on the lack of inhibition by numerous endogenous regulators that differentiate self and non-self surfaces. This mechanism ensures constant monitoring of the body. In contrast to the alternative pathway, the classical and lectin pathways require specific stimuli, such as antibodies, C-reactive protein, phosphatydylserine, or certain sugar moieties, to be present on the surface of target cells (14– 16). The upstream components of both pathways, including C1q, mannan binding protein (MBL), and the ficolins, act as sensors and thus can be considered soluble pattern recognition molecules (PRMs) (17, 18). All pathways converge at the level of the central complement molecule C3. C3 is processed by enzymatic complexes called complement convertases into C3a anaphylatoxin and the C3b fragment, which in turn forms C5 convertases. C5 convertases cleave the C5 molecule into C5a anaphylatoxin and the C5b fragment, which initiates the common terminal pathway. Binding of C6, C7, C8, and C9 leads to formation of the membrane attack complex (MAC), which targets the cell membrane and causes osmotic lysis. A schematic representation of the complement system is shown in **Figure 1**.

The binding of antibodies as a stimulus for the initiation of the complement cascade bridges the innate and adaptive immune systems. Moreover, opsonins like C3b and their degradation products act as natural adjuvants, contributing to proper presentation of antigens to lymphocytes and providing a co-stimulatory and anti-apoptotic signal for B cells (19). Deficiencies of complement, while relatively rare, emphasize the importance of this multifunctional protein cascade (20). The exact symptoms that develop depend on precisely which complement components are lacking or impaired. Deficiencies in essential components of the alternative pathway and the terminal pathway result in higher susceptibility to recurrent bacterial infections, especially these caused by Neisseriae and incidence peak up in early childhood (20). The lack of early components of the classical pathway predisposes to systemic lupus erythematosus (SLE), a disease in which

the scavenging function of complement is impaired, and thus debris from dying cells persists and can act as a source of autoantigens (21). Autoimmune diseases that stem from direct damage of cells and tissues typically arise from deficiencies in complement inhibitors that normally protect the host from excessive or misguided complement attacks (22). These include C3 glomerulopathies, atypical hemolytic uremic syndrome (aHUS), age-related macular degeneration (AMD), paroxysmal nocturnal hemoglobinuria (PHN), and many more (20). Paradoxically, a deficiency of functional complement inhibitors such as factor H (the main soluble inhibitor of the alternative pathway) can also result in a deficiency of complement activation. Factor H is the main soluble inhibitor of alternative pathway, which prevents propagation of cascade beyond the spontaneous breakdown of C3 and formation of C3 convertase (**Figure 1**). The lack of such inhibitor fuels a positive feedback mechanism that unproductively depletes complement and leaves the host without an important line of defense (23). On the other hand, unwanted complement activation is an effector mechanism in many inflammatory diseases, including rheumatoid arthritis (24), diabetic nephropathy (25), and ischemia/reperfusion injury (26). All these examples support a perception of the complement system as a "double-edged sword," where a proper balance is pivotal for maintaining protection while avoiding autoimmunity. Both microbial infections and tumors influence this physiological equilibrium and employ two main strategies for survival in a complement-saturated microenvironment.

#### STRATEGY #1: TO COUNTERACT COMPLEMENT

Innate immunity relies on recognition of a spectrum of pathogen-associated molecular patterns (PAMPs), invariable molecular determinants typical for the most common invaders, including lipopolysaccharide (LPS), lipoteichoic acid, flagellin, double-stranded RNA, β1-3 glucan, N-formylmethionine peptides, and many more. The constant region of the antibody heavy chain (Fc) also falls into this category of molecules. Pathogen-associated molecular patterns (PAMPs) bind to specific receptors on innate immune cells and activate effector mechanisms such as complement or antibody-dependent cell cytotoxicity (ADCC). Molecules that sense PAMPs and trigger immune system activation are called pattern recognition receptors (PRRs) and include Toll-like receptors (TLRs), C-type lectins, and NOD-like receptors. Soluble molecules, which exert an analogical role are termed as PRMs and the most upstream components of the classical and lectin pathways (C1 complex, MBL, and ficolins) belong to this group (18). Complement is therefore a multispecific and powerful defense system against pathogens that is theoretically capable of eliminating every cell unless constrained by endogenous complement inhibitors. In practice, since complement co-evolved with pathogens over millions of years, pathogens have developed various mechanisms to evade complement attack. Pathogens employ a variety of tactics for this purpose, including proteolytic cleavage of complement components, mimicking and hijacking host complement inhibitors, inactivation of the C3 molecule, preventing of complement-mediated activation of immune cells, depletion of antibodies, and unproductive exhaustion of early complement components [reviewed in (2, 4, 27–29)]. Selected examples of abovementioned strategies are given in **Table 1**.

Similarly to bacterial, fungal, or viral pathogens expressing PAMPs, tumor cells are visible to the immune system due to changes in their mutational or metabolic status, which is reflected by changes in the expression of cell surface molecules. The presentation of epitopes derived from mutated proteins (so-called neoantigens) within MHC I molecules (30) as well as the peroxidation of membrane lipids or changed patterns of glycosylation distinguish tumor cells from normal cells.

Spontaneous fixation of complement onto the surface of tumor cells is of low physiological relevance due to the low titer of naturally occurring antitumor antibodies and the expression of complement inhibitors by tumor cells (31). The introduction of antitumor monoclonal antibodies, which is considered a breakthrough in tumor immunology, enabled researchers to use the cytotoxic potential of the complement system to combat cancer (32). Complement-activating therapeutics like the anti-CD20 antibodies rituximab and ofatumumab are firstline therapies in the treatment of B cell malignancies. However, certain patients fail to respond or only partially respond to antitumor antibodies, and one possible explanation is the unfavorable ratio of the molecular target (e.g., CD20) to membrane-bound complement inhibitors on the surface of tumor cells (33, 34). Successful experiments in which bispecific antibodies against CD20 and CD55 were used (35) or in which complement inhibitors were silenced (36) support the theory that inhibition of complement by tumor cells is an important mechanism of cancer resistance. Expression of membrane-bound

complement inhibitors like CD35 (Complement receptor 1, CR1), CD46, CD55, and CD59 is typical for nucleated cells, and the majority of cell types express at least one of these molecules. In contrast, the production of soluble complement inhibitors such as factor I, factor H, C4b-binding protein (C4BP) is usually the domain of liver hepatocytes, and there are only few extrahepatic sources of fluid-phase complement regulators (37). However, the expression of soluble complement inhibitors by tumor cells has been described, and it seems to provide an additional level of protection, as shown in an in vitro model of non-small lung cancer cell (NSCLC) lines expressing factor I, C4BP, and factor H (38). The tumor-supporting effect of endogenous factor H expressed by NSCLC cells was shown in vivo in a mouse xenograft model (39, 40). Further evidence for the pro-tumor effect of soluble complement inhibitors comes from analysis of tissue microarrays of breast cancer specimens. Expression of factor I was positively correlated with tumor size, de-differentiation score (Nottingham scale), and poor prognosis (cancer-specific survival and recurrence-free survival) (41). Other investigators reported a correlation between factor I expression and tumor aggressiveness in cutaneous squamous cell carcinoma (42). In addition to expressing soluble complement inhibitors, tumor cells can also hijack these proteins from the plasma. Horl et al. showed that blocking factor H binding to the surface of leukemia cells increased the cytotoxicity of rituximab (43) and ofatumumab (44). Although factor H is an inhibitor of the alternative complement pathway, it plays a role in enhancing the complement cascade when it is initiated via the classical pathway (such as by antitumor antibodies) at the level of C3b formation. C3b gives rise to an amplification loop (**Figure 1**) due to the formation of alternative convertases, which are targets for factor H. Points of action of particular complement inhibitors are indicated in **Figure 1**. Another possible way to increase tumor cell resistance to complement attack is removal of the MAC from the surface, a process dependent on endocytosis or active rearrangement of the cell membrane mediated by phosphorylation of essential signaling proteins (45).

The logical consequence of complement inhibition by microbes and tumor cells at various stages of the cascade is a more aggressive and more drug-resistant phenotype, as discussed above. However, the picture is not as simple as it may originally seem, and another strategy used by pathogens and tumor cells to evade complement has been described.

#### STRATEGY #2: TO EXPLOIT THE COMPLEMENT SYSTEM

Low oxygen concentration is a feature of rapidly growing solid tumors, which cannot develop the vasculature necessary for the efficient supply of nutrients to proliferating neoplastic cells. Therefore, the expanding tumor mass sooner or later develops hypoxic cores. Normal cells are equipped with a sensor of oxygen concentration that works at the transcriptional level. Hypoxia-inducible factor Iα (HIF-1α) can stabilize the p53 tumor suppressor, triggering either apoptotic signaling or metabolic reprogramming of the cell (46). Both of these processes lead to changes in the molecules expressed at the cell surface, which has the effect of making the tumor cell visible to the immune system. Previous studies with human umbilical vein endothelial cells (HUVECs) revealed that these cells activate the classical complement pathway in response to hypoxia and as well as during subsequent reoxygenation. At the same time, HUVECs increased their surface expression of two membrane-bound complement inhibitors, CD46 and CD55 (47), which induce the proteolytic cleavage of activated complement components C3b and C4b, respectively, and the dissociation of the corresponding complement convertases. Another study showed a 3.6-fold increase in HUVEC expression of complement receptor 1 (CR1 or CD35) after 48 h of hypoxia (48). These results suggest that endothelial cells actively counteract complement activation under hypoxic conditions, and therefore the expression of complement inhibitors in hypoxic NSCLC cells was studied (49). These cells not only expressed membrane-bound complement inhibitors but also produced soluble inhibitors of complement, including C4BP and factors I and H (38). In contrast to HUVECs, NSCLCs significantly downregulated the mRNA expression of all complement inhibitors tested except CD59 after 24 h of hypoxia, but a drop in the mRNA expression of soluble complement inhibitors was detected as early as 6 h after hypoxic challenge. Importantly, this rapid decrease did not correspond to the number of dying cells, which did not significantly increase in first 24 h (49). The conclusion is that unlike endothelial cells, NSCLCs do not utilize protection mechanisms that prevent the deposition of early complement components during hypoxia, but they do maintain expression of CD59, which protects from the terminal stages of complement attack (the insertion of the MAC into the membrane) (47–49).

From the research reviewed above, it has become apparent that lung cancer cells may benefit from the propagation of local inflammation mediated by C3a and C5a. Possible scenarios include the production of proangiogenic and growth factors by tumor-infiltrating lymphocytes and macrophages as well as the mobilization of immune suppressor cells that impair tumor antigen presentation (50–53). Indirect support for this hypothesis comes from studies done by Ajona et al. who reported elevated C4d deposition in lung tumors and its correlation with decreased survival (54). Moreover, high levels of soluble C4d in the plasma could discriminate between patients with benign pulmonary nodules and lung cancer (55), and were associated with reduced survival of individuals with early and advanced lung cancer. C4d levels in the plasma were also reduced after surgical removal of the tumor (54). C4d is an end degradation product of the activated C4b molecule, a hallmark of classical complement pathway activation. For that reason, one can assume that the survival and malignant potential of NSCLC cells is based on stimulation of complement. Indeed, anaphylatoxin C5a is one of the key players in complement-mediated support of lung cancer growth. Corrales et al. found that C5 deposition and subsequent C5a generation in NSCLC cells was much higher than in non-malignant bronchial epithelial cells in the presence of serum (56). Interestingly, tumor cells but not non-transformed cells produced endogenous C5, and C5a generation took place even in the absence of serum. C5a levels in the plasma of lung cancer patients were also found to be elevated, similarly to C4d levels. C5a also stimulated migration and tube formation by HUVECs in vitro. Finally, the impact of C5a was tested in a syngenic mouse model of 3LL lung cancer. Microvessel density was compared in 3LL tumors in mice treated with a C5a receptor (C5aR) antagonist. Tumors in the mice treated with the C5aR antagonist showed significantly fewer microvessels (56). Additionally, C5a signaling positively influenced the recruitment of myeloid-derived suppressor cells (MDSCs; CD11b+, Ly6c+), as blockade of C5aR reduced the number of MDSCs in tumorbearing mice. The authors also found decreased expression of molecules associated with an immunosuppressive state and silencing of the immune response (ARG1, CTLA-4, IL-10, LAG3, and PD-L1) in C5aR antagonist-treated mice (57–59).

Importantly, the first evidence for the impact of C5a on the mobilization of MDSCs into the tumor mass was shown by Markiewski et al. in the TC-1 tumor model, a lung epithelial cell line expressing human papilloma virus (HPV) E6 and E7 antigens (60). The authors found that C5aR-deficient mice developed smaller tumors than wild-type littermates, and the same effect was observed when a C5aR antagonist was administered. However, in this model, the slower rate of tumor growth in C5aR antagonist-treated animals was not dependent on tumor cell proliferation/apoptosis or angiogenesis, as evidenced by analysis of end-point tumor specimens. Conversely, there were differences in the infiltration of tumor tissue by cytotoxic T cells, the main effectors of the antitumor immune response. Profiling of MDSCs isolated from tumors and spleens of C5aR-deficient, tumor-inoculated animals confirmed that C5a contributes to the accumulation of MDSCs in peripheral lymphoid organs and their migration into tumors. Of note, MDSCs isolated from mice with disabled C5aR signaling were less able to suppress T cell proliferation in vitro. This deficiency was linked to lower production of reactive oxygen species (ROS) and reactive nitrogen species (RNS) in mononuclear MDSC from C5aRdeficient animals.

Similarly to lung cancer cells, endogenous C5a generation by pancreatic and colon cancer cells was later reported. These cells processed C5 with a cell surface-expressed serine protease and expressed C5aR, suggesting autocrine activation of complement (61). In ovarian cancer, endogenous production of complement components and autocrine stimulation of the anaphylatoxin receptors C3aR and C5aR was suggested to be an important mechanism supporting tumor growth (62). The observed effect was independent of infiltration by cytotoxic T cells, since experiments with silenced expression of C3 yielded the same result (i.e., reduced tumor growth) in CD8 T cell-sufficient and -deficient mice. A direct effect of C3aR and C5aR agonists on proliferation, migration, and invasion of tumor cells has also been reported. Finally, quantification of C3 mRNA in tumors from patients with ovarian cancer showed that overall survival in patients with low tumor expression of C3 was more than double that of patients with high expression of C3 in the tumor (62).

In recent years, there has also been growing evidence for the pro-tumor activity of anaphylatoxins and anaphylatoxin receptors in either tumor cells or the tumor stroma in multiple tumors types, including melanoma, breast, ovarian, cervical, colon, and intestinal cancer, as well as sarcoma [reviewed in (63)]. Interestingly, in addition to the larger body of work focusing on the role of C3a and C5a in promoting tumor growth, recent studies have described a pro-tumor effect of factor B silencing (64) as well as a complement-independent enhancement of tumor growth, adhesion, and angiogenesis by C1q produced by the tumor stroma (65). The concept of complement activation supporting tumor growth provided the rationale for combined inhibition of C5a and PD-1 (66, 67), suggesting that targeting complement may be an effective anticancer treatment. These novel discoveries may be perceived as contradictory to the acknowledged theory that complement inhibits tumor growth. For example, NSCLC cells, which have been shown to benefit from C5a generation (56), were previously shown to form smaller tumors in a mouse xenograft model when their endogenous expression of the complement inhibitor factor H was silenced (40). In addition, potent complement activators such as the anti-CD20 immunotherapeutics rituximab and ofatumumab are firstline therapies for treatment of B cell malignancies (33). Notably, solid and circulating tumors have different requirements for growth. While sold tumors are typically depend on angiogenesis, migration, degradation of extracellular matrix, liquid tumors originate in the bone marrow, peripheral blood, or lymph nodes, which are rich in both nutrients and complement. Even solid tumors of the same origin can differ one from another in their mutational status, basal expression of growth factors and metalloproteinases, and metabolic rate. All of these parameters can influence the overall effect of complement activation on tumorigenesis and/or tumor progression. Finally, tumor cells often produce both complement activators and complement inhibitors. Thus, it seems as though tumor cells actively regulate the complement system depending on microenvironmental conditions, rather than simply avoiding constitutive inhibition or activation of complement (68).

Despite the extremely long phylogenetic distance between eukaryotic cells and bacteria, some prokaryotes have acquired strategies similar to tumor cells, which utilize the host inflammatory status to create favorable survival conditions. Bacterial growth in the human body is less dependent on neovascularization than tumor growth, and in contrast to tumor cells, bacteria do not have to overcome internal mechanisms controlling proliferation. Moreover, most bacteria can stand much harsher conditions than eukaryotic cells in terms of pH, oxygen tension, temperature, concentration of metabolites, etc. Nevertheless, the common feature between bacteria and tumor cells is the demand for nutrients and certain microelements. While solid tumors induce angiogenesis to acquire a source of nutrients, bacteria can successfully utilize products from the breakdown of local tissue. Therefore, tissue-destructive processes linked to local inflammation form permissive conditions for prokaryotic pathogens, which can survive immune attack. An additional benefit of this strategy is the elimination of inflammation-sensitive bacterial species (human commensals or normal microflora) that normally compete within the same niche (69, 70).

### INFLAMMOPHILIC CHARACTER OF PORPHYROMONAS GINGIVALIS, WHICH PROPELS PERIODONTISIS

One of the most well-documented examples of bacteria hijacking host immunity to create an environmental niche occurs in periodontitis, a chronic inflammatory disease characterized by dysbiosis that results in degradation of the gingiva and tooth-supporting bone and ultimately leads to tooth loss (1). Periodontal disease begins from dental plaque, a microbial matrix colonizing the gum line usually as a result of inefficient oral hygiene (71). The next stage, gingivitis, is characterized by local inflammatory response to microbial plaque. The switch between non-destructive gingivitis and destructive periodontitis involve dysbiosis of the normal oral microbiome. The dysbiotic process results from an imbalance of homeostasis caused by so-called keystone pathogens (72). A keystone pathogen is usually a microorganism of low abundance that induces changes in the composition of the local microflora by introducing a new selective pressure, such as inflammation. In periodontitis, the keystone pathogen is Porphyromonas gingivalis. However, as shown by studies in mice, this Gram-negative bacteria cannot establish periodontitis by itself, but requires commensal microbes. These microbes are then converted from a symbiotic into a dysbiotic community. Pivotal experiments showed that bone loss was reduced when C3aR- or C5aR-deficient mice were inoculated with P. gingivalis and that no changes in the oral microbiota were observed in these knockout mice after P. gingivalis inoculation, in contrast to wild-type mice (69). As some tumor cells generate C5a through their surface enzymes, so does P. gingivalis. It is equipped with gingipains, outer membraneanchored bacterial surface arginine-specific proteases with C5 convertase-like activity (73, 74). Importantly, gingipains release C5a from C5, but at higher concentrations, they degrade the larger fragment (C5b), thus preventing MAC formation (75). C3 and C4 complement proteins are also degraded by high concentrations of gingipains. Thus, human serum pre-incubated with clinical strains of P. gingivalis but not mutants lacking gingipains is devoid of bactericidal activity (74, 76). Additionally, gingipains interact with the C1 complex and increase its deposition onto bacteria surfaces (74). Based on these findings, one can postulate a biphasic effect of P. gingivalis proteolytic enzymes. A low abundance of bacteria initiates the classical complement pathway, but increasing numbers of bacteria results in the degradation of crucial complement components, leading to osmotic lysis. C5 is present in gingival crevicular fluid at concentration corresponding to 70% of that in serum and the active C5a anaphylatoxin can be locally released by convertases and bacterial proteases (77). C5a is a strong inflammatory mediator that increases vascular permeability and attracts and modulates the function of neutrophils, monocytes, and mast cells. All these events are considered antimicrobial events. Paradoxically, P. gingivalis' strategy for immune subversion by proinflammatory C5a involves targeted immunosuppression of macrophages. C5a affects intracellular killing of engulfed P. gingivalis by RNS and corrupts the crosstalk between C5aR and TLR2, one of the most important PRMs in antibacterial innate immunity (73). At the same time, C5aR-TLR2 crosstalk results in release of proinflammatory cytokines such as IL-1β, IL-6, and TNF-α, which accelerate bone resorption and thus contributes to the pathological mechanism of periodontitis. Similarly, P. gingivalis spoils intracellular killing mechanism but not proinflammatory activity of neutrophils by degradation of TLR2 adaptor molecule MyD88 provoked by concomitant activation of TLR2 and C5aR (78).

Another functional similarity between P. gingivalis and tumor cells, which produce either complement inhibitors or complement components, is the fact that P. gingivalis not only possesses the proteolytic machinery to generate C5a but also expresses a unique enzyme, peptidyl arginine deiminase (PPAD), which can citrullinate the C-terminal arginine in C5a, a modification that results in substantial loss of anaphylatoxin chemotactic activity (79). This suggests that the evolutionary goal of pathogens like P. gingivalis is not constitutive activation or inhibition of the complement system, but rather the ability to actively control complement status depending on its current needs. As a keystone pathogen, P. gingivalis is a lowabundance species that plays a major role in remodeling the local microbiota community (69). Following establishment of P. gingivalis infection, a succession of other dysbiotic species proliferates in the periodontal plaque. Some express their own complement inhibitors, such as Tannerella forsythia, which produces karilysin and mirolysin (80, 81), Filifactor alocis, which produces FACIN (82), and Prevotella intermedia, which produces interpain A (83). Of note, interpain A works in concert with gingipains in the initial stages of infection, as both proteins activate the C1 complex and increase its deposition onto the cell surface. A dysbiotic bacterial community may promote a transcriptomic response that further improves bacterial fitness by regulation of nutrient acquisition and expression of virulence factors (84). This process resembles the crosstalk between tumor cells and the stroma, at least to a certain extent. Tumor cells can drive the polarization of infiltrating immune cells (e.g., into M2 macrophages), which in turn benefit the tumor cells by, for example, expressing angiogenic cytokines (85, 86). In addition, carcinoma-associated fibroblasts (CAF), which differentiate from normal fibroblasts upon stimulation by cancer-derived cytokines such as TGF-β, have emerged as important players in cancer progression and metastasis (87, 88).

# CONCLUDING REMARKS

The role of the complement system in combating bacteria and cancer is more complicated than was initially believed. Certain pathogens have evolved the ability not only to evade complement attack but also to use it as a tool for establishing their own niche, while remaining protected from complementmediated lysis. Such a strategy seems to be widespread in nature and has been adopted by both bacteria and tumor cells (**Figure 2**). It seems likely that additional pathogenic strategies remain to be discovered, and thus one must be careful when designing complement-based therapeutics. On the other hand, anti-complement approaches may be effective in the treatment of infections caused by inflammophilic microbes.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### REFERENCES


#### FUNDING

MO is supported by National Science Centre Poland, grant number 2014/14/E/NZ6/00182. JP is supported by National Science Centre Poland, grant number 2016/21/B/NZ1/00292 and NIH/NICDR grant number DE 022597.

protection from Fas-mediated apoptosis. J Immunol. (2003) 171:5244–54. doi: 10.4049/jimmunol.171.10.5244


inhibits complement by degrading complement factor C3. PLoS Pathog. (2009) 5:e1000316. doi: 10.1371/journal.ppat.1000316


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Okrój and Potempa. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Complementing Cancer Metastasis

*Dawn M. Kochanek† , Shanawaz M. Ghouse† , Magdalena M. Karbowniczek\* and Maciej M. Markiewski\**

*Department of Immunotherapeutics and Biotechnology, School of Pharmacy, Texas Tech University Health Sciences Center, Abilene, TX, United States*

Complement is an effector of innate immunity and a bridge connecting innate immunity and subsequent adaptive immune responses. It is essential for protection against infections and for orchestrating inflammatory responses. Recent studies have also demonstrated contribution of the complement system to several homeostatic processes that are traditionally not considered to be involved in immunity. Thus, complement regulates homeostasis and immunity. However, dysregulation of this system contributes to several pathologies including inflammatory and autoimmune diseases. Unexpectedly, studies of the last decade have also revealed that complement promotes cancer progression. Since the initial discovery of tumor promoting role of complement, numerous preclinical and clinical studies demonstrated contribution of several complement components to regulation of tumor growth through their direct interactions with the corresponding receptors on tumor cells or through suppression of antitumor immunity. Most of this work, however, focused on a role of complement in regulating growth of primary tumors. Only recently, a few studies showed that complement promotes cancer metastasis through its contribution to epithelial-to-mesenchymal transition and the premetastatic niche. This latter work has shown that complement activation and generation of complement effectors including C5a occur in organs that are target for metastasis prior to arrival of the very first tumor cells. C5a through its interactions with C5a receptor 1 inhibits antitumor immunity by activating and recruiting immunosuppressive cells from the bone marrow to the premetastatic niche and by regulating function and self-renewal of pulmonary tissue-resident alveolar macrophages. These new advancements provide additional evidence for multifaceted functions of complement in cancer.

*Edited by:* 

*Maciej Cedzynski, Institute for Medical Biology (PAN), Poland*

#### *Reviewed by:*

*Mariusz Z. Ratajczak, University of Louisville Physicians, United States Dimitrios C. Mastellos, National Centre of Scientific Research Demokritos, Greece*

#### *\*Correspondence:*

*Magdalena M. Karbowniczek magdalena.karbowniczek@ ttuhsc.edu; Maciej M. Markiewski maciej.markiewski@ttuhsc.edu*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 15 May 2018 Accepted: 02 July 2018 Published: 16 July 2018*

#### *Citation:*

*Kochanek DM, Ghouse SM, Karbowniczek MM and Markiewski MM (2018) Complementing Cancer Metastasis. Front. Immunol. 9:1629. doi: 10.3389/fimmu.2018.01629*

Keywords: complement system proteins, cancer, metastasis, alveolar macrophages, myeloid-derived suppressor cells, epithelial–mesenchymal transition

#### INTRODUCTION

In both mouse models of cancer and patients, the expression of several complement genes is increased, resulting in higher than normal concentrations of complement proteins in plasma or other body fluids (1, 2, 3). In addition, complement activation is thought to occur in cancers because activated complement fragments are deposited within tumors (4, 5). This deposition of complement cleavage products and complement protein complexes including the C5b-9 terminal complement complex was observed in breast cancer (6) and in papillary thyroid carcinoma (7, 8). Complement activation through the lectin pathway was shown in colorectal carcinoma (9, 10). Complement fragments were detected in ascites from ovarian carcinoma patients (11). Complement activation in cancer patients is also supported by detection of C5a circulating in plasma of non-small cell lung carcinoma patients (2). The early studies reporting upregulation and activation of the complement pathway led to a notion that complement, similar to lysing bacteria, may contribute to lysis of tumor cells and, consequently, participates in tumor immune surveillance. However, this is disputable because of the resistance of cancer cells to complement-mediated lysis, which, however, become obvious mainly in the context of use of monoclonal antibodies for cancer immunotherapy (12, 13). This resistance results from high expression of membrane complement regulatory proteins (CRPs) on tumor cells (14) and secretion of soluble complement regulators from these cells (15), especially in solid tumors (13, 16). In contrast, in hematologic malignancies, complement mediated killing can be relevant, at least in the therapeutic context. For example, rituximab, a chimeric CD20 monoclonal antibody used to treat B cell lymphomas utilizes complement-mediated cytotoxicity (CDC) to kill tumor cells. There is growing interest in targeting complement regulators to improve efficacy of monoclonal antibody therapy in cancer (13, 17). Another approach to improve complement-mediated killing of tumor cells is the use of the "hexabody" platform. This technology stems from a seminal discovery that IgGs form hexamers after binding to antigen on the activating surface. This process is mediated by noncovalent interactions between Fc fragments of IgGs (18). Engineering Fc segments can be utilized to enhance formation of hexamers and, consequently, improvement of CDC toward tumor cells (19). Additional example of antitumor complement functions is participation of the complement anaphylatoxins C3a and C5a in enhancing antitumor immunity after radiotherapy. Interestingly, dexamethasone, a drug often administrated during radiotherapy limited complement activation and, consequently, inhibited antitumor immunity (20).

In contrast to these beneficial outcomes of complement activity, it is conceivable that without the discussed here therapeutic interventions, complement enhances tumor growth through its proinflammatory properties (4, 5). This possibility is consistent with a well-established tumor promoting role of chronic inflammation (21). Indeed, the first work to demonstrate tumor promoting properties of complement showed that several complement deficiencies were associated with reduced tumor growth through mechanisms linked to improvement of antitumor immunity (22). Several follow-up studies demonstrated immunoregulatory properties of various complement proteins (23). In addition, complement enhances tumor growth through direct regulation of tumor cell proliferation and invasiveness through C3a and C5a receptors expressed on carcinoma cells (24). Interestingly, the receptors for anaphylatoxins are also expressed in several leukemia and lymphoma cell lines and the blasts from chronic myeloid leukemia and acute myeloid leukemia patients. These cells responded robustly to C3a and C5a stimulation *in vitro* through chemotaxis and this process is negatively regulated by heme oxygenase 1 (HO-1) (25). These findings indicate that trafficking and spread of tumor cells in hematologic malignancies is perhaps, at least partially, controlled by complement system, therefore, inhibiting complement or upregulating HO-1 offer a new therapeutic opportunity for hematologic malignancies.

Together, studies of the last decade provide compelling evidence for a pivotal role of the complement system in tumor growth and targeting complement for anticancer therapy. Interestingly, recent developments point to regulation of cancer metastasis by complement, which appears, in some studies, to be independent from complement functions in primary tumors. This work links complement to a phase of metastatic process that only recently has been proved experimentally and is termed the premetastatic niche (26). We focus our discussion here on these new advancements on complement in metastasis. We also discuss contributions of complement to epithelial-to-mesenchymal transition (EMT), which initiates metastasis in primary tumors.

#### COMPLEX COMPLEMENT

The complement system is an assembly of more than 50 proteins that work together to provide immunity from infections, regulate several homeostatic processes, and trigger responses to tissue damage or injury (23). Although the textbook definition places complement in the center of innate immunity, recent developments demonstrated that this versatile system functions beyond limits of the immune system, regulating, for example, synaptic pruning (27), tissue regeneration/repair (28, 29), and bone homeostasis (30). In addition to its key function in innate immunity, complement regulates adaptive immunity. The receptors for the complement activation fragments are expressed in B and T cells and their signaling is pivotal for maintaining efficient protection against infection (31, 32). The stimulation of the complement receptor 2 (CR2) through antigen coated with C3d reduces the threshold for B cell activation rendering costimulation for best antibody production (33, 34). The studies on a role of complement in regulating T cell responses has led to surprising discovery that complement proteins in the cytoplasm regulate several intracellular process, mainly of a metabolic nature, essential for T cell homeostasis. The intracellular complement, termed "complosome," interacts with other intracellular innate sensor systems to control processes that are fundamental for adaptive immune responses such as metabolic reprograming necessary for generation of effector T cells (35).

The complement system also includes soluble fluid-phase or membrane-bound proteins, cofactors, regulators, and receptors (36). Upon stimulation by either pathogen or danger-associated molecular patterns, or antibodies, a cascade of events occurs that leads to activation of complement through different complement pathways. The alternative pathway is initiated by bacterial surfaces or unconstrained fluid phase hydrolysis of the complement C3 thioester (37). The lectin pathway is triggered through binding of mannose binding lectin or the ficolins (termed ficolin-1, ficolin-2, ficolin-3) to particular carbohydrates or N-acteyl residues (38, 39). The classical pathway starts when C1q binds to at least two IgG molecules (or one IgM) in a complex with antigen (40). In addition, complement fragments can be cleaved and thus activated through proteolytic enzymes that are not traditionally linked to the complement system. We grouped these additional ways of complement activation under the "umbrella" of the "fourth extrinsic pathway" (41). All three traditional complement activation pathways lead to cleavage of a complement fragment C3, which results in generation of C3a anaphylatoxin (10 kDa) and a large component—C3b. The C3b is deposited on the bacterial or other activating surfaces (42, 43). Following cleavage of C3 by an enzymatic complex—C3 convertase, C5 is cleaved by C5 convertase and similar to C3 cleavage, small C5a and large C5b fragments are generated. C3b and C4b opsonize pathogens, e.g., flag them for phagocytosis by myeloid professional phagocytes that express receptors for C3 cleavage fragments. The large C5 cleavage product, C5b, binds to an activating surface and supports subsequent binding of C6, C7, C8, and finally C9 [membrane attack complex (MAC)]. The multiple C9 fragments polymerize and form a pore in the cell membrane resulting in cell lysis or cell activation in certain circumstances (44). The complement anaphylatoxins C3a, C4a, and C5a are potent mediators that orchestrate events of inflammation (41, 45).

Excessive complement activation can be deleterious; therefore, this process is tightly controlled by CRPs (46). There are both soluble and membrane bound CRPs that can be grouped into several functional categories: (i) CRPs with decay-acceleration activity that increases the rate of C3 convertase breakdown and (ii) with cofactor activity resulting in the cleavage of C3b and C4b, thus, stopping C3 convertase formation (47). Three additional important CRPs are factor H, C1 inhibitor (C1NH), and CD59. Factor H acts in the alternative pathway as a C3 convertase decay accelerator and as a cofactor for factor I-mediated cleavage of C3b. CD59 is the only CRP, which acts to prevent assembly of the MAC. C1NH acts in both the classical and lectin pathways by inactivating C1r, C1s, and mannose-binding lectin serine proteases (47, 48, 49).

#### COMPLEMENT AND CANCER

In 2008, complement C3, C4, and C5a receptor 1-deficient mice were shown to have slower tumor growth in a model of human papilloma virus-induced cancer (22). This paper was the first study to contradict a well-accepted, at that time, notion of complement participation in immune surveillance. Tumor promoting functions of complement, at least in this model, were linked to C5a/C5a receptor 1 (C5aR1)-mediated activation and recruitment of myeloid-derived suppressor cells (MDSC) to tumors and inhibition of antitumor immunity. At the time of this publication, concerns were raised that the observed phenotypes may be restricted to a single tumor model (4, 5, 50). However, multiple preclinical and clinical studies in the last decade supported tumor-promoting properties of different complement components (16, 49). For example, studies by Corrales and colleagues demonstrated that C5a regulates MDSC in a lung cancer model (2). The blockade of C5aR1 led to a reduction in expression of genes that suppress antitumor immunity including *Arg1*, *Il-6, Il-10, Ctla4, Lag3, and Cd234 (PDL1)* (2). Recently, it has been shown that C3aR and C5aR1 signaling have an important impact on the IL-10-mediated cytotoxic properties of CD8<sup>+</sup> T cells infiltrating tumors in models of melanoma and breast cancer (E0771) (51). In this manuscript, tumor infiltrating CD8<sup>+</sup> T cells were shown to produce C3, which in autocrine manner inhibited the expression of IL-10. This cytokine appears to be essential for the cytotoxic properties of these cells. Mechanistically, IL-10 was associated with C3aR and C5aR1 signaling in CD8<sup>+</sup> T cells. Complement's role in recruiting tumor-associated macrophages (TAMs) and controlling their proangiogenic characteristics was proposed in a work, exploring antitumor functions of pentraxin 3 (52).

In addition to research demonstrating contributions of complement to inhibition of antitumor immunity, several studies showed other mechanisms behind tumor promoting functions of complement. In ovarian carcinoma models, tumor cells were demonstrated to produce complement components. C3a and C5a generated through activation of complement fragments produced in tumor cells regulated proliferation and invasiveness of tumor cells in autocrine fashion (24). C1q deposited in several human malignancies and mouse tumors seems to accelerate tumor growth through its proangiogenic properties and direct regulation of tumor cell motility and proliferation (53). Of value, Ajona et. al recently demonstrated improved efficacy of programmed celldeath 1 (PD-1) blockade in the presence of complement inhibition in reducing progression of tumors in a model of lung cancer (54). These new findings divulge a feasible path for targeting the complement system with the use of immunotherapeutic agents along with T cell check inhibitors. The detailed and comprehensive descriptions of a role in regulating tumor growth can be found in recent reviews (16, 23, 49). Here, we focus the discussion on the role of complement in regulating metastasis, a role that seems to involve different mechanisms.

# METASTASIS A HALLMARK OF MALIGNANCY

Cancer metastasis is a process of relocation of tumor cells from a primary to a distant (disconnected from primary tumor) site, through lymph or blood. In fact, a metastatic potential determines the malignant character of primary growth (55). Cancer metastasis are responsible for approximately 90 percent of cancer-associated deaths, however, paradoxically, mechanisms regulating metastasis remain the most obscure aspect of cancer biology (56). The metastatic spread of cancer is a multistep and complex chain of alterations in tumor and host cells, and tumor stroma, known as the invasion-metastasis cascade (56, 57). This cascade involves processes in primary tumor sites, circulation, and metastasis-targeted organs. Some of the first steps in the metastatic cascade involve acquisition of the ability to migrate and invade and degrade the tumor stroma by tumor cells. This goal is achieved through triggering in tumor cells several cellular programs that are collectively termed EMT, which is also an essential process during embryogenesis and wound healing (58). The EMT occurs perhaps in several malignancies; however, the current understanding of these cellular adaptations stems from studies in the models of epithelial-origin neoplasms carcinomas (59).

Complement has been linked to EMT in two recent studies (**Figure 1**) (60, 61). In the first study, increased expression of C5aR1 was found in hepatocellular carcinoma and hepatocellular carcinoma-derived cell lines and positively correlated with stage and invasion of liver capsule by tumor cells. The stimulation of C5aR1 *via* C5a induced EMT, as demonstrated by downregulation of E-cadherin and Claudin-1 expression, and upregulation

of Snail. Mechanistically, C5aR1-mediated EMT was linked to ERK1/2 signaling (61). In another study, C3 expressed in ovarian carcinoma-derived cells reduced expression of E-cadherin through C3a and Krüppel-like factor 5. Interestingly, C3 expression in tumor cells is transcriptionally regulated by twist basic helix–loop–helix transcription factor 1 (TWIST1), which binds to the *C3* promoter and enhances its expression. TWIST1 and C3 colocalized at the invasive tumor edges, and in the neural crest and limb buds of mouse embryos. Therefore, this work identified TWIST1 as a transcription factor that regulates *C3* expression during pathologic and physiologic EMT (60). The phenotypes associated with EMT program resemble phenotypes of cancer stem cells (CSCs) that are essential for metastatic spread. The recent work showed that CD10<sup>+</sup> cancer-associated fibroblasts that express a second C5a receptor (C5L2) provide a survival niche for CSCs through C5L2-mediated NF-kβ activation (62). Through EMT, tumor cells reduce their attachment to neighboring tumor cells and surrounding stromal elements, increase motility, and acquire the ability to invade stroma, blood, or lymphatic vessels, thereby gaining access to the vasculature.

The invasion of blood or lymphatic vessels enables tumor cells to intravasate and enter the circulation. The histopathological identification of vasculature invasion is itself a poor prognostic factor and often correlates with advanced metastatic disease (63). The lymph node metastases are a critical factor in cancer staging and are independent prognostic factors in several malignancies (64). However, mortality in cancer patients results from hematogenous spread to the vital organs including lungs, liver, and ultimately brain. Although initially lymph node metastases were thought to precede the subsequent hematogenous spread of cancer, evidence that draining lymph nodes are just temporary "parking" sites for cancer cells, before their departure to blood, is rather limited. It seems that lymph nodes represent a final destination for some cancer cells while other tumor cells, for unclear reasons, spread through the blood vessels (59). Upon successful intravasation, tumor cells move with the bloodstream to distant sites. However, only a small fraction of tumor cells that enter circulation safely reach their destination in the capillary beds of lungs and liver or cross the blood–brain barrier. This low efficacy of metastatic spread in blood results from hemodynamic stress and elimination of circulating tumor cells by the innate immunity, mainly natural killer (NK) cells (65). In contrast to NK cells, interactions with platelets (66) and neutrophils (67, 68) appear to facilitate metastasis.

After reaching their final destination, tumor cells are trapped in the capillary beds of the vital organs because their size is usually larger than the diameter of a single capillary. The halting of tumor cells in narrow capillaries facilitates their interaction with endothelium that is required for adhesion to endothelial cells and subsequent crossing of this endothelial barrier by extravasating tumor cells (transendothelial migration). Several substances secreted by tumor and host cells in the capillary beds enhance adhesiveness of tumor and endothelial cells and increase vascular permeability (69, 70), thereby, facilitating tumor cell extravasation. In the liver and kidneys, the fenestrated endothelium seems to facilitate seeding of these organs by metastasizing tumor cells. Perhaps the mechanisms contributing to extravasation of tumor cells in different organs vary, depending on the location and intrinsic properties of metastasizing cells.

After successful seeding of distant sites, tumor cells usually persist in an indolent state as single disseminated tumor cells or subclinical microscopic metastases, sometimes for years. The reasons for tumor cells to remain in a dormant state are unclear; however, poor adaptation of tumor cells to new microenvironment of metastasis-targeted organs seems to play a significant role (59). In addition, transition to rapidly growing and clinically overt metastasis, known as metastatic colonization, requires robust angiogenesis and immune evasion that may not be evident during a dormant phase of metastatic progression (71). For breast, prostate, and kidney cancers, a dormant phase may last even for decades after initial therapy and eradication of a primary tumor (59). Therefore, dormant tumor cells need to find a microenvironment-niche that allows them to slowly self-renew, provides needed nutrients, and protects from anticancer drugs and elimination by the immune system (72). For example, prostate carcinoma cells often metastasize to bones where they compete for residence in the endosteal niche with hematopoietic stem cells (73). In multiple organs including lungs, bones, and brain, tumor cells reside in close proximity to blood vessels in a region known as the perivascular niche (72, 74).

Interestingly, as much as EMT is necessary to trigger the invasion-metastasis cascade in primary sites, the reversal of this process, called mesenchymal-to-epithelial transition (MET), contributes to metastatic colonization, which is a final stage of metastatic disease. In metastatic tumors, MET appears to be critical in restoring a complex and heterogonous structure resembling primary tumors (75). Metastatic colonization leads to development of clinically overt and rapidly growing metastatic lesions, which are the ultimate reason for cancer-associated mortality. The transition from dormant to rapidly growing metastases requires acquisition of specific cellular programs by tumor cells, such as, discussed already MET, but also complex and well-orchestrated changes in the microenvironment of metastasis-targeted organs that include angiogenesis (72, 76), inflammation (77), remodeling of extracellular matrix (78, 79, 80), and evasion of antitumor immunity (81).

# THE PREMETASTATIC NICHE

Surprisingly, in several mouse models of cancer, changes that appear to be essential for metastatic colonization, e.g., a final stage of the invasion-metastatic cascade, including vascular alterations, remodeling of extracellular matrix, inflammation, and immunosuppression are observed in certain organs that seem to be marked for metastasis even before the arrival of the tumor cells. These alterations, collectively known as the premetastatic niche, are thought to facilitate seeding of these organs by disseminated tumor cells and their survival after they arrive to distant sites. The establishment of the premetastatic niche is triggered by the primary tumors (82) because efficiency of seeding of metastasistargeted organs by intravenously (i.v.) injected tumor cells is greatly enhanced by the presence of these tumors (3). Tumor-free mice i.v. injected with murine cancer cells developed significantly less lung metastases-derived from these i.v. injected cells than breast tumor-bearing mice i.v. injected with the same amounts of cells, indicating that the presence of primary breast malignancy facilitated seeding of the lungs by circulating (i.v. injected) tumor cells (3). It also appears that different types of cancer selectively prepare the premetastatic niche in different organs. This reflects the tendency of some malignancies to metastasize preferentially to specific locations. This specificity, known also as organotropism, was initially noted by Stephan Paget in 1889 (82), however, mechanisms regulating organotropism remain unclear until now. These mechanisms perhaps involve complex interactions between tumor cells and metastasis targeted organs that were proposed by Paget in his "seeds (tumor cells) and soil (microenvironment of premetastatic sites) theory." However, until seminal studies of the last decade (83, 84), which indeed established the field of premetastatic niche, the experimental proof for Paget's theory was missing. It is increasingly accepted that the premetastatic niche is created by tumor-secreted factors and tumor-shed extracellular vesicles, mainly exosomes. These factors seem to collectively control the stepwise development of premetastatic niche that begins with vascular alterations and progresses through activation of resident cells, extracellular matrix remodeling, and recruitment of bone marrow-derived cells (85).

# Secreted Factors

The evidence that tumor-secreted factors contribute to the premetastatic niche and organotropism, was, perhaps, first provided by experiments showing that melanoma-conditioned medium injected into mice, directed the metastasis of Lewis lung carcinoma cells (which normally metastasize only to the lungs) to sites typical for experimental melanoma metastasis (84). Among several identified factors secreted by tumors, vascular endothelial growth factor A (VEGFA), placental growth factor (84), transforming growth factor β (TGF-β), and tumor necrosis factor (TNF) were first demonstrated to prepare "soil" for tumor cells (26, 83).

#### Exosomes

Exosomes, small extracellular vesicles formed on the cell surface through a budding mechanism, contain diverse cargo that facilitates cell-to-cell communication and homeostatic cell regulation (86). However, in patients and mouse models, formation of exosomes by tumor cells is increased compared to normal cells (87). Tumor-derived exosomes were isolated from plasma of cancer patients and mice with experimental tumors and found to carry tumor-derived cargo that promotes disease progression (87). This exosomal cargo, which includes tumor-derived miRNA and proteins, reprograms the target cells toward a prometastatic and pro-inflammatory phenotype, resulting in their contribution to the formation of the premetastatic niche. For example, melanoma B16-derived exosomes increase the expression of the receptor tyrosine kinase and MET in bone marrow progenitors, causing their exit from bone marrow and migration to the lungs, where they contributed to the premetastatic niche. Importantly, MET expression is also elevated in circulating CD45<sup>−</sup>C-KITlow/<sup>+</sup>TIE2<sup>+</sup> bone marrow progenitors from patients with metastatic melanoma (87). B16-derived exosomes increase vascular permeability and enhance expression of TNF, S100A8, and S100A9, contributing to recruitment of bone marrow cells to the lung premetastatic niche. Of note, the source of S100 proteins was not identified in this study (87).

#### Abundant Complement

In contrast to tumor-derived secreted factors and exosomes, complement proteins are present in abundance in plasma and body fluids (41, 88) and, therefore, are readily available to participate in the premetastatic niche in patients or mice even with very small tumors. Increased concentration of complement components in plasma and other bodily fluids has been observed in both cancer patients and mouse models of cancer (1, 2, 3) suggesting upregulation of the complement pathway. These higher amounts of complement proteins may be linked to enhanced expression and production of complement by the liver, however, local increases in expression of complement genes in tumors and organs targeted by metastasis contribute to augmented levels of complement fragments because endothelial and immune cells synthesize complement fragments (89, 90) and these cells are an integral component of the tumor microenvironment (81, 91) and the premetastatic niche (26). Thus, they are possible sites of origin for several complement proteins in tumors and metastatic sites. For example, in a mouse model of breast cancer with spontaneous metastatic spread mimicking human malignancy, increased concentrations of C3 were found in plasma and bronchoalveolar lavage indicating increased production of complement proteins (3). These higher levels of complement fragments correlated with increases in expression of C3 and C5 genes in the lungs (3, 92). Cytotoxic CD8<sup>+</sup> T cells were found to synthesize C3 in mouse models of melanoma and breast cancer (51). Importantly, tumor cells also produce complement proteins. Mouse ovarian carcinoma tumor cells and human ovarian carcinoma cell lines were demonstrated to produce C3 (24). In a squamous cell carcinoma model, expression of C3, factor B, and factor I were also demonstrated (93, 94). Boire and colleagues recently demonstrated that C3 produced and secreted from tumor cells has prosurvival functions and facilitates leptomeningeal metastasis (95).

Complement proteins are secreted from cells in their inactive forms, as zymogens. To exert their functions, these fragments are activated through a series of proteolytic cleavages that form a complement cascade, which ends with generation of complement effectors (41). Therefore, if complement plays a role in regulating metastatic progression, complement activation in metastasistargeted organs or tumors is anticipated. This activation can be revealed by detecting deposited complement fragments in tissue or secreted effectors such as complement anaphylatoxins, C3a, C4a, and C5a. The cleavage fragments of C3 were found to be deposited in the lungs prior to metastasis, indicating complement activation and participation of the complement system in the lung premetastatic (3). These data were obtained through a use of a syngeneic mouse model of metastatic breast cancer (4T1), in which tumor cells are injected into the mammary fat pad, and breast tumors formed there subsequently metastasize to distant sites, similar to human malignancy (96). The deposition of C3 cleavage fragments in the lungs correlated with increasing levels of C5a in plasma over time (**Figure 1**) (3).

#### Vasculature

Increased vascular permeability is one of the earliest changes observed in the premetastatic niche and is associated with increased metastatic burden (97, 98). The factors secreted from primary tumors, including epithelial growth factor receptor ligand epiregulin, metalloproteinases MMP1, and MMP2, are known to impact vascular permeability in primary tumors and distant sites, helping tumor cells to intravasate in a primary and then extravasate in a distant site, respectively (99). In a melanoma model, tumor cells secrete factors upregulating angiopoietin 2, MMP3, and MMP10 that synergistically destabilize vasculature in the premetastatic organs (97). Factors affecting vascular permeability can also be secreted from different cell populations recruited to the premetastatic niche. For example, myeloid cells were shown to produce MMP9 (100) and VEGFA (101). Endothelial cells, which are often targets for vasoactive substances, can themselves participate in vascular alterations in the premetastatic niche. VEGFA-dependent upregulation of E-selectin on the luminal surface of endothelium facilities adhesion of tumor cells to endothelium and subsequent extravasation (102).

The complement effectors, especially C5a, are powerful inflammatory mediators that are actively engaged in bringing leukocytes to sites of inflammation. C5a can achieve this goal, acting as a potent chemoattractant that causes cytoskeleton changes in leukocytes that are responsible for cell movement (103). However, it also enhances (directly and indirectly) vascular permeability, further adding to accumulation of leukocytes in inflammatory foci (103). The complement C3 cleavage fragments were found to be deposited in the premetastatic lungs as early as 4 days after injecting tumor cells into the mammary fat pad in a model of breast cancer (before any tumor cells are present in the lungs). Since C3 is a central component of complement cascade, on which all complement activation pathways converge, deposition of C3 cleavage fragments indicates complement activation and subsequent generation of C5a (88), which indeed was present in sera of these mice (3). It is, therefore, conceivable that complement contributes directly to increased vascular permeability in the premetastatic niche similar to its participation in inflammatory vascular alterations; however, a direct experimental evidence for these C5a functions in the premetastatic niche has yet to be provided. The indirect impact of complement on vascular changes can be attributed to recruitment of MDSC (3) because these cells can produce and release several vasoactive factors including MMP9, which is intimately involved in regulating vascular integrity in the premetastatic niche (83, 84). Genetic ablation of *Mmp9* was shown to normalize the aberrant vasculature in the premetastatic lungs and reduce metastatic burden (100). The seminal recent work has also demonstrated that cancer-cellderived C3/C3a through C3aR in the choroid plexus disrupts the blood–cerebrospinal fluid barrier. The increased permeability of this barrier facilities the entry of plasma proteins that are essential for tumor growth into the cerebrospinal fluid, thereby, facilitating leptomeningeal metastasis (95).

#### Resident Cells

Resident cells in metastasis-targeted organs are naturally suited to participate in the premetastatic niche because they are present before the arrival of tumor cells. Tumors can reach and potentially hijack these cells through several mechanisms including secreted tumor-derived factors, exosomes, and recruitment of bone marrow-derived cells that subsequently interact with resident components of the premetastatic niche. Recent work also demonstrated that complement activation regulates resident cells in the lungs (**Figure 1**) (92). As discussed already, endothelial cells are targets for several vasoactive substances whether derived from tumors, recruited cells, or generated locally. Fibroblasts contribute to remodeling of extracellular matrix through deposition of new extracellular matrix components or by secreting enzymes that affect preexisting components of the matrix (104). S100A4 expressing pulmonary fibroblasts incorporate exosomal cargo derived from breast cancer cells and through this mechanism upregulates S100 proteins (82). Exosomal cargo from pancreatic carcinoma induces similar changes in the liver-resident macrophages, Kupffer cells (105). S100 proteins were linked to recruitment of myeloid-origin cells to the premetastatic niche (106). Similar to Kupffer cells in the liver, another population of tissue-resident macrophages, pulmonary alveolar macrophages, were recently demonstrated to contribute to the premetastatic niche (92).

These recent developments on participation of tissue-resident macrophages to metastasis are of particular interest because roles of these cells in cancer remain unclear, in contrast to very well-studied TAMs or inflammatory monocytes/macrophages recruited to the lungs with metastases by CCL2 (101). Several early studies yielded conflicting results on how liver Kupffer cells and lung alveolar macrophages contribute to cancer progression (107). A role of these cells in cancer requires revision because recently published linagetracing data demonstrated that these cells have a different origin and biology than inflammatory macrophages (108). Unlike inflammatory macrophages and TAMs that are recruited from bone marrow to sites of inflammation or tumors, respectively, tissue-resident macrophages migrate to different organs during embryogenesis prior to hematopoiesis, and self-renew thereafter (109). Pulmonary alveolar macrophages are the resident macrophages of the lungs, and while they have well-known immunoregulatory and homeostatic roles in healthy lungs (110), they appear to be well-suited to partake in preconditioning the lungs for metastasis through their immunoregulatory properties. In support of this notion, alveolar macrophages were found to accumulate in premetastatic lungs and this accumulation was the result of cell proliferation rather than recruitment from bone marrow (92). The mechanisms controlling proliferation of these cells were linked to C5aR1 signaling because tumor-bearing C5aR1-deficient mice presented with a lower total number of these cells in the lungs compared to tumor-bearing wild-type controls and this reduced cell number associated with reduced Ki-67 expression. Immunoregulatory functions of these cells appeared to be related to skewing effector CD4<sup>+</sup> T cells responses toward Th2 phenotype, which plays limited role in antitumor immunity in contrast to Th1 responses (92). In addition, alveolar macrophages in tumor-bearing hosts reduced number and antigen-presenting capacity of lung dendritic cells through regulation of TGF-β1 in lung infiltrating leukocytes (**Figure 1**). The depletion of alveolar macrophages reversed immunosuppression and reduced lung metastatic burden (92).

#### Extracellular Matrix Remodeling

The continuous remodeling of extracellular matrix by tumorderived secreted factors, resident fibroblasts, and recruited bone marrow-derived cells is an integral part of the premetastatic niche (111). This remodeling is achieved through deposition of new extracellular matrix components or modification of existing components. For example, deposition of fibronectin produced by activated fibroblast provides a docking site for bone marrow-derived cells that express fibronectin receptor VLA-4 (84). These stromal fibroblasts, stimulated with tumorderived TGF-β, produce periostin in a mouse model of breast cancer (78). Periostin directly interacts with type I collagen, fibronectin, and Notch1 through its amino-terminal EMI domain and interacts with tenascin-C and BMP-1 through its fas I domains. These periostin interactions with mainly extracellular matrix molecules occur at first intracellularly. In addition, periostin serves as a ligand for integrins such as αvβ3 and αvβ5 and promotes cell motility by acting outside the cell (112). Periostin was demonstrated to facilitate melanoma metastasis to wounds (113) and to regulate immunosuppressive functions of MDSC during early stages of breast cancer metastasis (114). MDSC (mainly monocytic-MDSC), which accumulated in the premetastatic lung of MMTV-PyMT spontaneous breast tumor-bearing mice, secrete versican, an extracellular matrix proteoglycan. Versican contributed to MET and the formation of macrometastasis in the lungs (115). Enzymatic modulation of extracellular matrix proteins also occurs in the premetastatic niche and is mediated mainly by metalloproteinases produced by cells that are recruited to the premetastatic niche. In addition, the members of the LOX family crosslink collagen type I and IV and this crosslinking facilitates adhesion of bone marrow-derived cells to the extracellular matrix of the premetastatic niche. These cells produce more metalloproteinases contributing to further remodeling of extracellular matrix.

Although complement was not directly linked to extracellularmatrix remodeling in the premetastatic niche, studies in different model systems demonstrated that fibronectin can interact with several complement components including C1q (116) and C3 cleavage fragments (117). C3 cleavage fragments can also bind to different components of extracellular matrix including collagen (118). Interestingly, binding of C1q to fibronectin was not associated with complement activation but was connected to enhancement of phagocytosis of C1q coated particles through fibronectin. Therefore, functional significance of complement interactions with extracellular matrix proteins in the premetastatic niche remain to be elucidated. However, it is reasonable to theorize that complement C3 cleavage fragments, bound to extracellular matrix proteins, interact with its receptors broadly expressed on myeloid-origin cells that are recruited to the premetastatic niche. The receptors for C3 degradation products include CR1, which binds C3b and iC3b, CR2 (CD21), which binds the degradation products of C3b (iC3b, C3dg, C3d), CR3 (CD11b/CD18 or Mac-1), which binds iC3b, and CR4 (CD11c/ CD18), which binds iC3b, however, through a different domain than CR3 (119). C5a leads to upregulation of CR3 on MDSC, which may facilitate adhesion of these cells to endothelium and recruitment to tumors (22); however, it may also contribute to adhesion of MDSC to extracellular matrix in the premetastatic niches.

#### Recruited Cells

The identification of bone-marrow derived cells in the premetastatic niche and discovery of their roles in facilitating seeding of these niches by tumor cells, provided perhaps the first experimental evidence confirming the "seed and soil" theory (83, 84). A seminal work by Hiratsuka and colleagues defined Mac1 (CR3) positive macrophages as a source of MMP9 in the lungs in addition to endothelial cells. These macrophages were recruited to the lungs because resident alveolar macrophages do not express CD11b (https://www.immgen.org/). The study by Kaplan and colleagues demonstrated recruitment of VEGFR1 and VLA-4 expressing hematopoietic progenitors to the premetastatic lungs and their participation in the premetastatic niche (84). These studies opened an avenue for further investigations into discovery of other recruited components of the premetastatic niche. Less than a decade later, MDSC, which were long recognized as modulators of the primary tumor microenvironment (120), were identified as contributors to the premetastatic niche (100, 115). However, these studies reported on metastasis promoting properties of MDSC linked to increased vascular permeability (100) and remodeling of extracellular matrix (115) rather than to their well-established immunoregulatory roles.

The C5aR1/C5a signaling axis recruits MDSC to primary tumors (2, 22), therefore, it was explored whether similar mechanisms operate in the premetastatic niche. Utilizing a syngeneic mouse model of metastatic breast cancer, it has been demonstrated that C5aR1 knockout or wild-type mice administrated with a specific C5aR1 inhibitor (PMX-53) had decreased lung and liver metastatic burden compared to control mice. Interestingly, C5aR1 appear to regulate only metastasis in this model because lack of C5aR1in mice did not affect primary breast tumors. The differences in lung metastasis were associated with differences in a degree of infiltration of the lungs and livers by MDSC. The lung infiltrating MDSC were found mainly in interavleolar septa and due to intensity of this infiltration, the morphological picture resembled interstitial pneumonia. Therefore, the term the premetastatic pneumonia has been proposed to emphasize intensity of MDSC infiltration and specific localization of these cells in the lungs (3). These MDSC were recruited to the premetastatic sites through C5a since C5aR1 was expressed in blood MDSC and complement activation, leading to C5a generation, was observed in the premetastatic niche (**Figure 1**). To further investigate the role of C3 cleavage fragments and MDSC, tumor-draining lymph nodes from breast cancer patients were examined; they observed that C3 fragments' deposition and local C3 production were both intensified in lymph nodes with metastases (3). The decreased

#### REFERENCES


metastasis in mice lacking C5aR1 resulted from improved antitumor immunity due to escalated infiltration of the lungs by CD8<sup>+</sup> and CD4<sup>+</sup> T cells. In addition, the increases in these T cell subsets were found in peripheral blood and C5aR1-deficiency favored Th1 response. Also observed was a decrease in Tregs in both the blood and the lungs in C5aR1-knockout mice. The T cell subsets including CD4<sup>+</sup> and CD8<sup>+</sup> T cells isolated from the lungs of C5aR1 knockout mice produced increased amounts of IFN-γ. The elimination of cytotoxic CD8+ T cells by neutralizing antibody erased the inhibitory effect of C5aR1-deficiency on metastasis, supporting notion that this effect was caused by stimulating antitumor immunity. Importantly, these data also indicate immunoregulatory functions of MDSC in the premetastatic niche (3).

#### CONCLUDING REMARKS

The evidence supporting contributions of complement to cancer metastasis is scarce and limited to a few recent papers. However, it appears that complement affects key steps in the invasionmetastasis cascade including EMT and the premetastatic niche (**Figure 1**). Given ubiquitous presence of complement in body fluids and tissues, the potential contributions of complement to regulating metastasis are significant. Our recent work demonstrated that C5aR1 regulates resident (alveolar macrophages) (92) and recruited (MDSC) (3) cells in the premetastatic niche. The significance of this regulation was underscored by complete protection from lung metastasis in mice depleted of alveolar macrophages and treated with C5aR1 inhibitor (92). Thus, despite an early phase of studies on complement participation in metastasis, several complement fragments appear to be promising targets for therapies seeking to stop cancer metastasis.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This work was supported by the National Institute of Health (R01CA190209 to MM), the Cancer Prevention and Research Institute of Texas (RP120168 to MK), the U.S. Department of Defense (TS140010 to MK), Laura W. Bush Institute for Women's Health (seed grants to MM and MK). We thank the Development Corporation of Abilene for continued financial support.

T-cell responses in the metastatic niche. *Cancer Res* (2014) 74(13):3454–65. doi:10.1158/0008-5472.CAN-14-0157


(1994) 73(11):2808–17. doi:10.1002/1097-0142(19940601)73:11<2808:: AID-CNCR2820731125>3.0.CO;2-P


cell carcinoma. *J Invest Dermatol* (2015) 135(2):579–88. doi:10.1038/jid. 2014.376


proliferate at key stages during inflammation. *Nat Commun* (2013) 4:1886. doi:10.1038/ncomms2877


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Kochanek, Ghouse, Karbowniczek and Markiewski. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# MASP-1 and MASP-2 Serum Levels Are Associated With Worse Prognostic in Cervical Cancer Progression

Carlos Afonso Maestri 1,2, Renato Nisihara2,3 \*, Hellen Weinschutz Mendes <sup>3</sup> , Jens Jensenius <sup>4</sup> , Stephen Thiel <sup>4</sup> , Iara Messias-Reason<sup>3</sup> and Newton Sérgio de Carvalho<sup>5</sup>

<sup>1</sup> Liga Paranaense de Combate ao Câncer, Erasto Gaertner Hospital, Curitiba, Brazil, <sup>2</sup> Department of Medicine, Positivo University, Curitiba, Brazil, <sup>3</sup> Immunopathology Laboratory, Department of Clinical Pathology, Federal University of Parana, Curitiba, Brazil, <sup>4</sup> Department of Biomedicine, Aarhus University, Aarhus, Denmark, <sup>5</sup> Department of Gynecology, Clinical Hospital, Federal University of Parana, Curitiba, Brazil

#### Edited by:

Robert Braidwood Sim, University of Oxford, United Kingdom

#### Reviewed by:

Dan Anthony Mitchell, University of Warwick, United Kingdom Krishnan Hajela, Devi Ahilya Vishwavidyalaya, India

\*Correspondence:

Renato Nisihara renatonisihara@up.edu.br; renatonisihara@gmail.com

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 11 May 2018 Accepted: 07 November 2018 Published: 23 November 2018

#### Citation:

Maestri CA, Nisihara R, Mendes HW, Jensenius J, Thiel S, Messias-Reason I and de Carvalho NS (2018) MASP-1 and MASP-2 Serum Levels Are Associated With Worse Prognostic in Cervical Cancer Progression. Front. Immunol. 9:2742. doi: 10.3389/fimmu.2018.02742 Background: MBL-associated serine proteases (MASP-1, MASP-2, MASP-3, MAp-44, and MAp-19) are key factors in the activation of the lectin pathway of complement. Serum levels of these components have been associated with recurrence and poor survival of some types of cancer, such as colorectal and ovarian cancer. In this investigation, we determined the serum levels of MASP-1, MASP-2, MASP-3, MAp-44, and MAp-19 in patients with cervical cancer and cervical intraepithelial neoplasia (CIN).

#### Methods: A total of 351 women who underwent screening for cervical cancer or treatment at the Erasto Gaertner Cancer Hospital in Curitiba-Brazil, were enrolled in the study. Based on their latest cervical colposcopy-guided biopsy results, they were divided into four groups: CIN-I: n = 52; CIN-II: n = 73; CIN-III: n = 141; and invasive cancer: n = 78. All the serum protein levels were determined by time-resolved immunofluorometric assay (TRIFMA).

Results: Patients with invasive cancer presented significantly higher MASP-2, MASP-1, and MAp-19 serum levels than other groups (p < 0.0001; p = 0.012; p = 0.025 respectively). No statistically significant differences in MASP-3 and MAp-44 serum levels were found between the four studied groups. In addition, high MASP-2, MASP-1, and MAp-19 serum levels were significantly associated with poor survival in patients with invasive cancer and relapse (p = 0.002, p = 0.0035 and p = 0.025, respectively).

Conclusion: High MASP-2, MASP-1, and MAp-19 serum levels were associated with cervical cancer progression and worse disease prognosis. These novel findings demonstrate the involvement of the serine proteases of the lectin pathway in the pathogenesis of cervical cancer and future investigations should clarify their role in the disease process.

Keywords: complement, lectin pathway, HPV, cervical intraepithelial neoplasia, cervical cancer

# INTRODUCTION

Human papilloma virus (HPV) is amongst the most common worldwide viral infection transmitted sexually. And although most HPV infections generally clear up on its own, a small percentage of these infections caused by specific HPV types can persist and evolve to cancer. The highest prevalence rate of HPV is observed in women younger than 25 years and decreases at later ages, but in many populations, there is a secondary peak in peri-menopause (1, 2). Young infected patients can usually eliminate the virus without presenting clear clinical evidence, due to an effective cell-mediated immune response causing the lesions to regress. The inability to develop an effective immune response to clear or control infection results in persistent infection that, in the case of oncogenic HPVs, creates a greater likelihood of progression to high-grade cervical intraepithelial neoplasia (CIN-2, CIN-3) and carcinoma. On the other hand, about 50% of HPV infections in women with normal cytology will resolve in less than 1 year, and approximately 90% of women with either normal cytology or CIN-1 diagnoses will ultimately resolve on their own (2).

It is known that HPV has a variety of strategies to evade the immune response, and when this occurs, HPV replication continues leading to persistent infection (1, 2). Among the approaches used by HPV to evade the host immune response are suppression of important inflammatory/immunological pathways that enable virus escaping from host immune surveillance (3). Due to suppression of some danger signals, HPV infection does not produce cytolysis, cytopathic cell death or interferon (IFN) release, leading to lower grade inflammation (1, 4, 5).

Chronic inflammation in the tumor microenvironment and evasion of the antitumor effector immune response are two of the hallmarks required for oncogenesis and cancer progression. The innate immune system not only plays a critical role in perpetuating these tumor-promoting hallmarks but also in developing antitumor adaptive immune responses (6). Thus, understanding the dual role of the innate system in cancer immunology is required for the design of combined immunotherapy strategies able to tackle established tumors (6). The lectin pathway of the complement system is part of innate response, being triggered when pattern recognition molecules (PRMs), including the two collectins [mannan-binding lectin (MBL) and collectin-LK] and three ficolins (ficolin-1,-2, and-3) initiate the complement activation upon binding to carbohydrates present on the surfaces of microbes or altered tissues. When this occurs, three MBL-associated proteases (MASP-1,-2, and-3) provide activation of the complement system, and two MBL-associated proteins (MAp44 and MAp19) serve as natural endogenous competitive inhibitors (7). The proteins MASP-1, MASP-3, and Map44 arise from the MASP1 gene by mutually exclusive splicing (8). In a similar mode, MASP-2 and MAp19 arise from the MASP2 gene (7). All these components are key factors in the activation of the lectin pathway of complement. Serum levels of these components were associated with recurrence and poor survival of some types of cancer such as colorectal or ovarian cancer (9, 10). Increased serum levels of MBL and MASP-2 were found in patients with colorectal cancer, which were not explained by genetic profiles (9, 10). So far, no study has addressed the serine proteases of the lectin pathway proteins in relation to cervical intraepithelial neoplasia and carcinoma. Our research group had previously reported on the MBL concentrations in women presenting with HPV-associated cervical lesions, showing there was no statistically significant difference between the median serum MBL concentrations in women presenting with CIN-1, CIN-2, CIN-3 lesions or invasive cervical cancer (11).

The aim of the present study was to investigate whether MASP-1, MASP-2, MASP-3, MAp-44, and MAp-19 serum levels are involved in the pathogenesis and progression of cervical intraepithelial neoplasia by measuring their concentrations in women presenting moderate grade of cervical intraepithelial neoplasia and cervical invasive cancer.

#### SUBJECTS AND METHODS

This cross-sectional study was carried out at the Erasto Gaertner Cancer Hospital (HEG) in Curitiba, southern Brazil, which is a reference center for the treatment of gynecological malignancies. The Institutional Ethics Committee approved the study.

All subjects were followed at the out-patient clinic of the HEG and were consecutively included from January 2011 to March 2012. Inclusion criteria were cervical cancer screening and treatment at HEG. We also collected historical data on disease progression from patients' medical records. Exclusion criteria were pregnancy, HIV-positivity, systemic infection, autoimmune disease, and blood transfusion within the last 60 days. Written informed consent was obtained from all patients.

A group of 344 women was included in this study. Based on their latest cervical colposcopy-guided biopsy, the subjects were divided into four groups: low grade CIN-1: n = 52 (control group); moderate CIN-2: n = 73; CIN-3: n = 141; and invasive cancer (Ca): n = 78. The loop electrical excision procedure or cold-knife cone excision was used to confirm the previous biopsy results. A 3-ml sample of venous blood was collected from each subject and allowed to coagulate. The coagulated blood was centrifuged, and the serum was separated, aliquoted (500 µl), and stored at −80◦C, until use for MASP-1, MASP-2, MASP-3, Map-44, and Map-19 concentration determinations.

The concentrations of MASP-1, MASP-2, MASP-3, Map-44, and Map-19 were measured in according to published previously (8, 12, 13). The assays used for protein determinations were monoclonal antibody-based time-resolved immunofluorometric assays (TRIFMAs) and were carried out in partnership with the Institute of Medical Microbiology and Immunology, University of Aarhus, Aarhus. The human MASP-1 assay was based on competition from MASP-1 in serum with the interaction between anti-MASP-1 antibody and a fragment of MASP-1 coated onto microtitre wells. In the case of MASP-2 concentration ("sandwich ELISA"), rat anti-MASP-2 mAb (clone 8B5) was used for coating while biotinylated rat anti-MASP-2/MAp19 mAb (clone 6G12). MASP-3, Map-44, and Map-19 was determined by sandwich


IQR, Interquartile range; CIN, Cervical intraepithelial neoplasia; Ca, Cancer.

ELISA using specific antibodies. All the assays used Eu3+ labeled streptavidin (Perkin Elmer, USA)—for detection. All the antibodies were kindly provided by Prof. Jens C. Jensenius (Aarhus University, Denmark).

Data were analyzed as frequency and contingency tables. The Kolmogorov–Smirnov test was used to assess the data distribution. Central tendency was expressed as mean and standard deviation or median and interquartile range (IQR), and the Mann–Whitney test was used to compare numerical data. A receiver-operator characteristic curve (ROC) was also calculated for MASP-2. Results were considered statistically significant at p < 0.05. GraphPad Prism 7.0 software was used for statistical analysis.

#### RESULTS

The serum concentration of MASP-1, MASP-2, MASP-3, MAp-44, and MAp-19 proteins in the investigated groups is presented in **Table 1**. Significantly higher MASP-2, MASP-1, and MAp-19 serum were observed in patients with invasive cancer compared to other groups (p < 0.025). However, no significant differences were observed for MASP-3 and MAp-44 levels.

Regarding clinical evolution, 41 (11.9%; mean age 51.4 years) patients died within 24–48 months. Surviving patients had a mean age of 34.8 years. When serum concentrations were evaluated, median MASP-2 levels were higher among patients who died than among those who survived (439.4 ng/ml × 278.3 ng/mL; p < 0.0001). For MASP-1 (6988 ng/mL vs. 5110 ng/mL; p = 0.0035) and MAp-19 (366 ng/mL vs. 234 ng/mL; p = 0.012), the same trend was observed. Serum concentration of MASP-3 and MAp-44 were not significantly different between the two groups. When patients who died and those who experienced relapse of CIN or cancer during follow-up were considered, only MASP-2 and MAp-19 were significantly different (**Table 2**).

ROC curve was used to show the sensitivity and specificity for MASP-2 serum concentrations in relation to the prognosis of cervical cancer progression. High levels of MASP-2 (>291.9 ng/mL) showed a sensitivity of 71.3 and a specificity of 63.2 for worse prognosis of cervical lesions.

TABLE 2 | Association between MASP-1, MASP-2, MASP-3, MAp-44, and MAp-19 serum concentrations and clinical evolution in patients with cervical lesions.


\*Mann–Whitney test.

#### DISCUSSION

Our study demonstrates that MASP-2, MASP-1, and MAp-19 could have a role in the progression of cervical cancer. Patients diagnosed with invasive cancer showed a significant increase in the concentrations of these lectin pathway components. In another study conducted by our group, we failed to establish the relationship between serum concentration of MBL and the evolution of cervical intraepithelial neoplasias (10). These novel findings indicate possible involvement of the lectin pathway in the immunopathogenesis of cervical cancer, although it can result, in part, from activation of the complement system due to the presence of tumor cells. In fact, the participatory role of lectin pathway components in defense against pathogens such as viruses, bacteria, or fungi is much better known and studied: MASPs deficiencies are associated with increased infection susceptibility, and increased levels are associated with tissue injury (14). However, studies about its role in the development of cancer are scarce.

In a study of colorectal cancer, Ytting et al. (10) observed higher MASP-2 serum concentrations in patients with cancer, although no differences for MASP2 genotypes were detected between patients with colorectal cancer and healthy controls. Furthermore, serum levels of MASP-2 have been associated with post-operative infections, recurrent cancer, and poor survival in colorectal cancer patients (9). These findings suggest that increased MASP-2 serum concentrations may occur due to factors other than genetics and that MASP-2 may play a role in cancer development and disease outcome. Corroborating these observations, in our study high MASP-2 serum concentrations were related to worse prognosis in cervical cancer and interestingly, they were mainly altered in patients who died and/or had disease recurrence. MASP-2 is produced in hepatocytes, and its promoter is regulated by cytokines such as interleukin (IL)-1b and IL-6 or transcription factor STAT3 (15).It is possible that tumor tissue releases mediators that may increase serum levels of MASP-2, however this is a hypothesis that needs to be clarified. Studying genetic polymorphism and MASP-2 serum levels in patients with ovarian cancer, Swierzko et al. (9) concluded that the expression of MBL and MASP-2 is altered in ovarian cancer, possibly indicating involvement of the lectin pathway of complement in the disease. Similar results of increased MASP-2 serum levels were described in children with tumors of the central nervous system (16) and in patients with acute lymphoblastic leukemia and non-Hodgkin's lymphoma (16).

In our study, MASP-1 serum concentrations were significantly higher in cancer patients than in other studied groups. The normal MASP-1 concentration in serum/plasma is approximately 20-fold higher than that of MASP-2 (13). It is known that MASP-1 activates MASP-2 in heterocomplexes of large oligomeric MBL and produces 60% of the C2b responsible for C3 convertase formation (17). Additionally, MASP-1 was reported to be essential for the development of autoimmune-associated inflammatory tissue injury by activating the alternative complement pathway in an experimental model of inflammatory arthritis (18). There are no studies on MASP-1 concentrations in cancer.

Complement system and macrophages collaborate synergistically to maintain progression of angiogenesis, and create conditions of carcinogenesis such as dysregulation of mitogenic signaling pathways, cellular proliferation, angiogenesis, resistance to apoptosis, and escape from immunosurveillance. Macrophages release mediators that modulate inflammation and acquired immunity response. In addition, promote angiogenesis, tissue remodeling and tissue repair (19). The tumor growth is sustained by the infiltration of M2-tumor-associated macrophages, and high levels of C3a and C5a. Macrophages have receptors for both C3a and C5a on their cell surface, and this specific binding affects the functional modulation and angiogenic properties (19). High levels of MASP-2 and MASP-1 may increase activation of the complement system by the lectin pathway, thereby augmenting the release of C5a, a potent anaphylatoxin that activates cellular responses involved in tumor growth and progression (20). During cervical lesion progression, the number of M2 macrophages is significantly increased. The aggregation of M2 is a key event in the pathological process of carcinogenesis (21). M2 macrophages increase vascular endothelial growth factor (VEGF) and metalloprotease-9 secretion to aid in tissue repair. On the other hand, complement system activation by tumors may result in basement membrane disruption, tumor growth, and metastasis (22). In vivo, HPV does not bind directly to cells but requires contact with the basement membrane. This contact can be accomplished by microabrasions on the cervical surface, which reveal the basement membrane. The most plausible receptor of the major capsid protein (L1) of HPV appears to be the tissue-specific heparin sulfate proteoglycan, which belongs to the basement membrane (23, 24). It is possible that patients with higher serum levels of MASP-1 and/or MASP-2 present exacerbate activation of complement leading to the disturbance of the basement membrane, and ultimately to its damage or rupture and consequently, tumor invasion.

Interestingly, HPV protein (E2, E6, E7) action on IL10 gene leads to increased IL-10 levels which in turn increases HPV E6 and E7 expression, leading to a vicious cycle (3, 22). IL-10 is an anti-inflammatory cytokine that modulates cytokine synthesis and exerts effects on resident and circulating immune cells, suppressing the immune system; however, their role in cancer remains controversial (3). On the other hand, activation of complement system is clearly an inflammatory process that acts to protect the host. However, to date, the majority of studies on cancer and the lectin pathway have shown a relation with worse prognosis when serum levels of components of this pathway were increased (9, 16). Future studies should investigate how complement interacts with the basement membrane and tumor tissue along with the influence of cytokine network on this system.

The limitations of this study are due to its cross-sectional study design, which prevented the use of serial dosages of the lectin pathway components over time to evaluate the stability of component concentrations throughout disease progression. Thus, future investigations evaluating the role of the different components of the lectin pathway, including collectins, ficolins and MASPs in the evolution of cervical lesions should be encouraged. In addition, our study did not include a healthy control group, for obvious reason, since it would be very difficult to obtain a group of women without HPV infection, either for ethical reasons or lack of lab tests to prove this condition. Thus, we believe that CIN-1 was the appropriate comparison group for our study, especially considering that such patients show good immune response against HPV and generally (>90%) do not progress to cancer (2).

In conclusion, our study shows that high MASP-2, MASP-1, and MAp-19 serum levels are associated with cervical cancer progression and worse disease prognosis. Future investigations should clarify the role of these complement components in cervical cancer immunopathogenesis.

#### AUTHOR CONTRIBUTIONS

CM and NdC: protocol and project development, data collection or management, and manuscript writing and editing; RN: protocol and project development, data analysis, and manuscript writing and editing; HM: protocol and project development, laboratorial assays, and manuscript writing and editing; JJ, ST, and IM-R: protocol and project development, and manuscript writing and editing.

#### REFERENCES


diseases. Mol Immunol. (2015) 67:85–100. doi: 10.1016/j.molimm.2015. 03.245


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Maestri, Nisihara, Mendes, Jensenius, Thiel, Messias-Reason and de Carvalho. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Role of Complement Activating Collectins and Associated Serine Proteases in Patients With Hematological Malignancies, Receiving High-Dose Chemotherapy, and Autologous Hematopoietic Stem Cell Transplantations (Auto-HSCT)

#### Edited by:

Robert Braidwood Sim, University of Oxford, United Kingdom

#### Reviewed by:

Gunnar Houen, State Serum Institute (SSI), Denmark Zoltan Prohaszka, Semmelweis University, Hungary

\*Correspondence:

Maciej Cedzynski ´ mcedzynski@cbm.pan.pl

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 27 June 2018 Accepted: 31 August 2018 Published: 20 September 2018

#### Citation:

Swierzko AS, Michalski M, ´ Sokołowska A, Nowicki M, Eppa Ł, Szala-Pozdziej A, Mitrus I, ´ Szmigielska-Kapłon A, Sobczyk-Kruszelnicka M, Michalak K, Gołos A, Wierzbowska A, Giebel S, Jamroziak K, Kowalski ML, Brzezinska O, Thiel S, Jensenius JC, ´ Kasperkiewicz K and Cedzynski M ´ (2018) The Role of Complement Activating Collectins and Associated Serine Proteases in Patients With Hematological Malignancies, Receiving High-Dose Chemotherapy, and Autologous Hematopoietic Stem Cell Transplantations (Auto-HSCT). Front. Immunol. 9:2153. doi: 10.3389/fimmu.2018.02153 Anna S. Swierzko ´ <sup>1</sup> , Mateusz Michalski <sup>1</sup> , Anna Sokołowska<sup>1</sup> , Mateusz Nowicki <sup>2</sup> , Łukasz Eppa<sup>1</sup> , Agnieszka Szala-Pozdziej ´ 1 , Iwona Mitrus <sup>3</sup> , Anna Szmigielska-Kapłon<sup>4</sup> , Małgorzata Sobczyk-Kruszelnicka<sup>3</sup> , Katarzyna Michalak <sup>3</sup> , Aleksandra Gołos <sup>5</sup> , Agnieszka Wierzbowska<sup>4</sup> , Sebastian Giebel <sup>3</sup> , Krzysztof Jamroziak <sup>5</sup> , Marek L. Kowalski <sup>6</sup> , Olga Brzezinska ´ 6,7, Steffen Thiel <sup>8</sup> , Jens C. Jensenius <sup>8</sup> , Katarzyna Kasperkiewicz <sup>9</sup> and Maciej Cedzynski ´ 1 \*

<sup>1</sup> Laboratory of Immunobiology of Infections, Institute of Medical Biology, Polish Academy of Sciences, Łódz, Poland, ´ <sup>2</sup> Department of Hematology, Copernicus Memorial Hospital in Łódz Comprehensive Cancer Center and Traumatology, Łód ´ z,´ Poland, <sup>3</sup> Department of Bone Marrow Transplantation and Oncohematology, Cancer Center and Institute of Oncology, Gliwice, Poland, <sup>4</sup> Department of Hematology, Medical University of Łódz, Łódz, Poland, ´ <sup>5</sup> Department of Hematology, Institute of Hematology and Transfusion Medicine, Warsaw, Poland, <sup>6</sup> Department of Immunology and Allergy, Medical University of Łódz, Łódz, Poland, ´ <sup>7</sup> Department of Rheumatology, Medical University of Łódz, Łódz, Poland, ´ <sup>8</sup> Department of Biomedicine, Aarhus University, Aarhus, Denmark, <sup>9</sup> Department of Microbiology, University of Silesia, Katowice, Poland

We conducted a prospective study of 312 patients (194 with multiple myeloma, 118 with lymphomas) receiving high-dose conditioning chemotherapy and autologous hematopoietic stem cell transplantation (auto-HSCT). Polymorphisms of MBL2 and MASP2 genes were investigated and serial measurements of serum concentrations of mannose-binding lectin (MBL), CL-LK collectin and MASP-2 as well as activities of MBL-MASP-1 and MBL-MASP-2 complex were made. Serum samples were taken before conditioning chemotherapy, before HSCT and once weekly after (totally 4-5 samples); in minority of subjects also 1 and/or 3 months post transplantation. The results were compared with data from 267 healthy controls and analyzed in relation to clinical data to explore possible associations with cancer and with chemotherapy-induced medical complications. We found a higher frequency of MBL deficiency-associated genotypes (LXA/O or O/O) among multiple myeloma patients compared with controls. It was however not associated with hospital infections or post-HSCT recovery of leukocytes, but seemed to be associated with the most severe infections during follow-up. Paradoxically, high MBL serum levels were a risk factor for prolonged fever and some infections. The first possible association of MBL2 gene 3′ -untranslated region polymorphism with cancer (lymphoma) in Caucasians was noted. Heterozygosity for MASP2 gene +359 A>G mutation was relatively frequent in lymphoma patients who experienced bacteremia during hospital stay. The median concentration of CL-LK was higher in myeloma patients compared with healthy subjects. Chemotherapy induced marked increases in serum MBL and MASP-2 concentrations, prolonged for several weeks and relatively slighter decline in CL-LK level within 1 week. Conflicting findings on the influence of MBL on infections following chemotherapy of myeloma and lymphoma have been reported. Here we found no evidence for an association between MBL deficiency and infection during the short period of neutropenia following conditioning treatment before HSCT. However, we noted a possible protective effect of MBL during follow-up, and suspected that to be fully effective when able to act in combination with phagocytic cells after their recovery.

Keywords: CL-LK, collectin, complement, hematopoietic stem cell transplantation (HSCT), mannose-binding lectin (MBL), MASP, mutliple myeloma, lymphoma

#### INTRODUCTION

Hematological cancers derive from various cells of the immune system. Multiple myeloma (MM), a plasma cell malignancy, is relatively common. It more often affects males than females and is usually diagnosed at age >55 years. Although incurable, treatment strategies including chemotherapy followed by autologous hematopoietic stem cell transplantation (auto-HSCT) allow marked prolongation of patients' life expectancy (1–5). The term "lymphoma" includes a variety of highly heterogeneous lymphoid malignancies. Hodgkin's lymphoma (HL) accounts for approximately 1/10 of all lymphomas at diagnosis. It is characterized by giant, multinucleated Reed-Sternberg and large mononuclear Hodgkin's cells (of B cell lineage). Incidence of HL peaks during early adulthood (20– 34 years) and again at age >60 years; it is slightly commoner in males, and most newly diagnosed patients are successfully cured (6–9). Non-Hodgkin's lymphomas (NHL) are also more often diagnosed in men than in women (10–15). NHL usually arise from the B cells lineage (>40 subtypes; >80% of cases) although some T cell and NK cell lymphomas occur. Some NHL are histologically aggressive with a high mortality rate, but NHL are often curable when diagnosed and treated early.

Hematological malignancies obviously compromise the host immune response. Intensive chemotherapy and/or radiotherapy, adds further immunosuppression (mostly due to profound and prolonged neutropenia). High-dose chemotherapy and auto-HSCT is the standard first-line treatment for younger patients with multiple myeloma as well as standard second-line treatment in patients with Hodgkin's lymphoma and aggressive non-Hodgkin's lymphomas (1, 6, 10, 16).

Due to severe immunocompromise caused by both the disease and therapy, patients are at a high risk of morbidity and mortality due to infections. It is estimated that up to 80% of patients have fever, and more than one third suffer from well-defined infections (17–19). Septicemia is the commonest type of infection, and some patients are affected by pneumonia or alimentary tract infections (19). Post-HSCT reconstitution of bone marrow after conditioning therapy is usually attained in ∼3 weeks, but full recovery of functional leukocytes takes longer (20, 21). During the period of aplasia, patients have to be isolated, carefully monitored and receive prophylactic/supportive medication (4, 18, 22).

Complement is a complex system of numerous components that constitutes a crucial branch of the immune response. It protects from a variety of infections and takes part in clearance of host cells undergoing apoptosis, necrosis or neoplastic transformation. However, dysregulation of its activation may result in chronic or systemic inflammation and possibly promotion of tumor growth. Therefore, involvement of complement may have opposing effects in cancer. Furthermore, some of its components may be considered as candidate diagnostic or prognostic biomarkers of the disease (23, 24).

Pattern recognition molecules known to activate complement via the lectin pathway are two collectins [mannose-binding lectin (MBL) and CL-LK, composed of two peptides: collectin-10 (CL-10, CL-L1) and collectin-11 (CL-11, CL-K1)] and ficolins (ficolin-1,-2,-3). They form complexes with MASP (MBL-associated serine proteases: MASP-1,-2,-3) and non-enzymatic related proteins (MAp19, MAp44) (25–27).

The aim of our investigation was to estimate possible association of complement-activating collectins (MBL, CL-LK) and associated serine proteases with cancer (MM, LYMPH) itself and with hospital infections after chemotherapy followed by auto-HSCT.

Single-nucleotide polymorphisms, localized to codons 52 (A/D), 54 (A/B), and 57 (A/C) of the first exon of the MBL2 gene, influence both MBL serum level and its activity. The variant, dominant alleles (collectively designated O), are associated with lower MBL levels compared with the A (wildtype) one. Furthermore, promoter region polymorphisms (H/L and Y/X at positions−550 and−221, respectively) also determine MBL concentration to some extent, due to influence on gene expression. The O/O and LXA/O genotypes are considered to be associated with primary MBL deficiency [reviewed in (24)].

The role of MBL and clinical associations of its deficiency apparently depend on clinical context, patient's age, co-existence

**Abbreviations:** DLBCL, diffuse large B-cell lymphoma; HL, Hodgkin's lymphoma; HSCT, hematopoietic stem cell transplantation; LYMPH, lymphoma; MM, multiple myeloma; NHL, non-Hodgkin's lymphoma.

of other defects of the immune system, etc. Several reports and review papers suggested that MBL2 A/O heterozygosity may be advantageous for the host. Relative high frequency of such genotypes in a variety of populations has been suspected to result from an evolutionary, selective trend (28–31). It was observed that MBL2 O alleles (and especially O/O genotypes) are less common among centenarians than in general population. On the other hand, in this age group, a relatively low frequency of high MBL-producing haplotypes/genotypes was found. Similar trends were noted in octo- and nonagenarians (32, 33). Therefore, it was postulated that moderate MBL serum concentration may be favorable to healthy aging while both MBL deficiency and its hyper-reactivity may be deleterious for the host. The first was related to variety of infectious which, especially when severe or recurrent, may lead to the worsening of quality of life and therefore to shortening of life expectancy. On the other hand, high MBL may favor autoimmunity and/or promote tumorigenesis. The latter is often associated with cachexia and fatal outcome. That issue becomes even more complicated when some opposing effects are considered: MBL deficiency may be a risk factor for development of cancer (like childhood acute lymphoblastic leukemia) while high MBL may facilitate some intracellular pathogens to enter their target cells [reviewed in (24, 33)].

Originally, Peterslund et al. (34) found a strong relationship between MBL deficiency and serious infections in a relatively small group of patients suffering from hematological malignancies (n = 54, including 18 with MM and 13 with NHL). Many subsequent studies failed to reproduce the strength of this relationship although positive results were reported in some studies but not others. It is clear that MBL does not have a universal role in preventing infections following chemotherapy, but any influence or lack of it is still controversial. Further studies are therefore needed and should be carried out prospectively with substantial numbers of patients. We have investigated a large (n = 194) series of multiple myeloma patients and compared them with a more heterogeneous group of lymphoma patients (n = 118), as well as healthy controls (n = 267). In addition to MBL, we have also extended our investigations to include CL-LK and MASP-2 since those factors are also relevant to collectin-mediated complement activation.

#### MATERIALS AND METHODS

#### Patients and Controls

Three hundred and twelve patients suffering from hematological malignancies, undergoing auto-HSCT were recruited. This group included 194 persons diagnosed with multiple myeloma (MM; 95 females and 99 males; mean age: 58.9 ± 8.9 years), 27 with Hodgkin's lymphoma (HL; 6 females and 21 males; mean age: 40.8 ± 13.7) and 91 with non-Hodgkin's lymphomas (NHL; 40 females, 51 males; mean age: 51.6 ± 11.9). For analyses, HL and NHL groups were combined to form LYMPH (n = 118; mean age: 49.3 ± 13.1). The NHL group included patients with diffuse large B-cell lymphoma (DLBCL, n = 31), mantle cell lymphoma (MCL, n = 18), follicular lymphoma (FL, n = 13), Waldenström's macroglobulinemia (WM, n = TABLE 1 | Basic characterisation of patients and controls.


1), non-follicular (diffuse) lymphoma, unspecified (NFL, n = 21), marginal-zone lymphoma (mucosa-associated lymphoid tissue lymphoma, MALTL, n = 1), anaplastic large cell lymphoma, kinase negative (ALCL ALK-, n = 1), anaplastic large cell lymphoma, kinase positive (ALCL ALK+, n = 1), angioimmunoblastic T-cell lymphoma (AITL, n = 2), peripheral T-cell lymphoma (PTCL, n = 1), hepatosplenic T-cell lymphoma (HSTL, n = 1). In total, the group of patients included 141 females and 171 males; mean age was 55.2 ± 11.7 years; age range: 21–73). Basic demographic data are summarized in **Table 1**.

With the exception of 12 persons suffering from MM undergoing radiotherapy alone, patients were treated with conditioning high-dose chemotherapy before transplantation (**Table 1**). In 28 patients (24 LYMPH and 4 MM), both chemotherapy and radiotherapy were used. Standard treatment for multiple myeloma was MEL-200 (melphalan at dosage 200 mg/m<sup>2</sup> ) although in some instances lower doses (MEL-140, MEL-100) were administered. The majority of LYMPH patients were treated with BEAM: carmustine (300 mg/m<sup>2</sup> ), etoposide (800 mg/m<sup>2</sup> ), cytarabine (1600 mg/m<sup>2</sup> ), melphalan (140 mg/m<sup>2</sup> ), accompanied, or not by radiotherapy. For the remaining, BeAM (carmustine replaced with bendamustine), bendamustine (alone or accompanied by radiation), cyclophosphamide (CTX) and radiotherapy or TBC combination (thiotepa, busulfan, cyclophosphamide) were used (**Table 1**).

During hospital stay, crucial clinical parameters like white blood cell (WBC) count, absolute neutrophil count (ANC), platelet (PLT) count, inflammatory markers [C-reactive protein (CRP), fibrinogen (FBG), and procalcitonin (PCT)] levels and incidence of complications [infections (associated with bacteremia or not), febrile neutropenia (FN); duration of fever >38◦C] were recorded and used for analyses. Etiological agents of infections are listed in **Supplementary Table 1**.

Swierzko et ´ al. Collectins and MASP in Auto-HSCT

One hundred age-matched controls (unrelated volunteers with no history of cancer, autoimmune diseases, or recurrence of infections) were recruited. Furthermore, DNA/serum samples from 167 healthy adult subjects collected for our previous projects were used, therefore control group included 267 individuals (174 females and 93 males; mean age: 48 ± 13; age range: 18–84). Basic data are presented in **Table 1**. The study was approved by the Ethics Committee of the Medical University of Lodz. The written informed consent from patients/controls was obtained. This work conforms to the provisions of the Declaration of Helsinki.

#### Blood and Serum Samples

Blood samples for DNA extraction were taken into citrated tubes before chemotherapy and stored at −80◦C. DNA was extracted with the use of GeneMATRIX Quick Blood Purification Kit (EURx Ltd, Gdansk, Poland), according to the manufacturer's protocol. Samples for serum isolation were taken into tubes with clot activator before chemotherapy (sample 1), before HSCT (sample 2; usually 3–7 days later) and once weekly till hospital discharge (sample 3: usually 7 days after HSCT; mean 7.0 ± 0.8; sample 4: usually 14 days after HSCT; mean 13.6 ± 1.2; in some patients sample 5: 21 days after HSCT; mean 20.3 ± 2). Sixtythree patients were additionally sampled∼45 days (mean 45.3 ± 7.6; sample 45) while 59 were additionally sampled at ∼100 days (mean 104.1 ± 19.3; sample 100) after HSCT, during control examination. Among them, 40 persons were sampled at "45" and "100" points. Sera were stored at −80◦C until testing.

#### MBL2 Genotyping

Promoter (H/L, at position−550, rs11003125 and Y/X, at position−221, rs7096206) and exon 1 (A/D, codon 52, rs5030737; A/B, codon 54, rs1800450, and A/C, codon 57, rs1800451) single nucleotide polymorphisms of the MBL2 gene were investigated as previously described (35). SNPs of the same gene, located within the 3′ -untranslated region of the exon 4: Ex4-1483 T>C (rs10082466), Ex4-1260 C>T (rs56009657), Ex4- 1067 G>A (rs10824792), Ex4-1064 G>T (rs35327474), Ex4- 1063 G>T (rs35768126), Ex4-1047 T>G (rs12254577), Ex4- 939 T>C (rs774307463), Ex4-901 A>G (rs2120132), Ex4-879 A>C (rs2120131), Ex4-845 C>T (rs2165813), Ex4-718 G>T (rs2099903) and Ex4-710 A>G (rs2099902) were investigated by direct sequencing. Briefly, PCRs were run on a C1000 Thermal Cycler (Bio-Rad, Hercules, CA, USA), using appropriate spanning primers (designed with the use of PRIMER3 software, http://bioinfo.ut.ee/primer3/):

forward: 5′ -TTTCCCCATGGTTTTAATCTG-3′ , reverse: 5′ -TCACTAAAACCACCAAAACAAGA-3′ ,

under the following conditions: 96◦C for 5 min, then 35 cycles (96◦C for 40 s, 60◦C for 30 s, 72◦C for 45 s) and finally 72◦C for 5 min. The PCR products were purified with the help of EPPiC - Enzymatic Post-PCR Immediate Cleanup (A&A Biotechnology, Gdynia, Poland). Samples thus prepared were directly used as templates for sequencing, performed using the GeneAnalizer-3000 sequencer (Applied Biosystems, Foster City, CA, USA), reverse primer, BrightDyeTerminator Cycle Sequencing kit (NimaGen BV, Nijmegen, The Netherlands) and TABLE 2 | Frequency of MBL2 genotypes (SNP located within promoter and exon 1) within major groups.


Percentages are shown in parentheses. \*p = 0.034; OR 1.84; 95%CI (1.06-3.19) (vs. C).

BDX64 Sequencing Enhancement Buffer (MCLab, San Francisco, CA, USA) according to the manufacturer's instructions. Sequence electropherograms were visually inspected to confirm base calling at the SNP sites using novoSNP (36).

#### MASP2 Genotyping

The presence of missense mutation +359 A>G (Asp120Gly) located within exon 3 of the MASP2 gene, rs72550870) was investigated with the use of PCR-RFLP method, as described previously (37).

For an investigation of +1111 A>C (exon 10, Asp371Tyr, rs12711521) polymorphism, the PCR-RFLP procedure was employed. PCR reactions were run on a C1000 Thermal Cycler (Bio-Rad), using primers designed with the use of PRIMER3 software:

forward: 5′ -TTATTTTCAGACCATGGGGG-3′ , reverse: 5′ -TCTGGGTGGTTTCAAATTCC-3′ .

The following conditions were applied: 95◦C for 3 min, then 35 cycles (95◦C for 30 s, 61◦C for 20 s and 72◦C for 30 s), followed by a final elongation (72◦C for 5 min). After that, the PCR products were treated with MboI enzyme (Fermentas), (37◦C, 2 h). The digestion products were further analyzed on a 6% polyacrylamide gel. The PCR product (522 bp) corresponding to C variant has two digestion sites while that corresponding to A allele has only one. Consequently, the first underwent cleavage into three (286, 176, and 60 bp) while the second was cleaved into two (286 and 237 bp) fragments.

### MBL Quantification and Determination of MBL-MASP Activity

MBL concentration/activity was measured by ELISA based on binding to solid-phase mannan and detection using monoclonal anti-human MBL antibody (HYB131-1, BioPorto Diagnostics, Copenhagen, Denmark) as previously described (38). MBL-MASP-1 complex activity (39, 40) and MBL-MASP-2 activity (40, 41), were measured using a fluorescence method and an ELISA, respectively. VPR-AMC peptide (Bachem, Bubendorf, Switzerland) was used as the substrate for MASP-1 (39). Rabbit polyclonal anti-human C4c Abs (Dako, Glostrup, Denmark) and corresponding goat HRP-conjugated secondary antibodies (Dako) were used for detection of deposited C4 products after incubation of the serum sample on mannan-coated wells, for an estimation of MBL-MASP-2 complex activity (40). MBL deficiency was taken as 150 ng/ml (42, 43). "High" MBL concentration (>3000 ng/ml) was chosen based on the value


TABLE

3


Frequency

of

MBL2

genotypes

(SNP

located

within

exon

4

3′-UTR)

within

major

groups.

Percentages

 are shown in parentheses.

 Differences in comparison

 with gt1 are marked in red. \*<sup>p</sup> = 0.01; OR 0.52; 95%CI (0.32-0.86) (vs. C).



Percentages are shown in parentheses. \*Number of C allele carriers (A/C or C/C genotype) significantly lower in comparison with MM patients: p = 0.023; OR 1.86; 95% CI (1.11-3.13).

corresponding to the 95th percentile determined for the control group (2978 ng/ml). For MBL-MASP-1 or-2 complex activities, the value of 1 U/ml was arbitrarily assigned for the standard serum (38, 40).

#### Assay for CL-LK

CL-LK concentration was determined by sandwich TRIFMA, as described (44). Mouse anti-human CL-10 monoclonal antibody (clone 4F97D) was used for coating while biotinylated rabbit anti-human CL-10 mAb (clone P1), followed by Eu3+-labeled streptavidin (Perkin Elmer, Waltham, MA, USA) was used for detection. "Low" and "high" CL-LK levels were arbitrarily defined as <242 ng/ml (corresponding to the 10th percentile within control group) and >583 ng/ml (95th percentile determined for controls), respectively.

#### Assay for MASP-2

MASP-2 levels were determined by TRIFMA, as described by Moller-Kristensen et al. (45), with slight modification (37). Rat antibodies of clone 8B were used for coating, while those of clone 6G12 (biotinylated) were used for detection. "Low" and "high" MASP-2 levels (based on 10 and 95th percentiles within C group) were arbitrarily defined as <172 ng/ml and >627 ng/ml, respectively.

#### Statistical Analysis

The Statistica (version 13.1, Dell Poland) software package was used for data management and statistical calculations. The medians of protein concentrations were compared using the Mann-Whitney U-test. It was chosen because MBL, MASP-2, MBL-MASP-1, and MBL-MASP-2 values were not normally distributed (not shown). The same test was employed for CL-LK (normally distributed), to make the analyses comparable. Changes during hospital stay and after discharge were analyzed by Friedman's ANOVA test. The frequencies of low or high levels, as well as genotypes/alleles were compared by twosided Fischer's exact test (or χ <sup>2</sup> when appropriate). Correlations

conditioning chemotherapy) and controls. Blue bars present median values (given below group descriptions). (A) controls (C) vs. multiple myeloma (MM) patients; (B) controls vs. lymphoma (LYMPH) patients. MM-A, LYMPH-A: patients who experienced infections with proven bacteremia; MM-B, LYMPH-B: patients who experienced infections with no bacteremia; MM-C, LYMPH-C: patients who experienced febrile neutropenia; MM-D, LYMPH-D: patients who experienced none of afore-mentioned complications during hospital stay.

were determined by Spearman's test. P <0.05 were considered statistically significant.

# RESULTS

#### MBL2 Gene Polymorphisms Located Within Promoter Region and Exon 1

The frequency of A/A homozygotes was similar in MM, LYMPH and C groups. However, MBL deficiency-associated genotypes (LXA/O, O/O) were significantly more common in patients with multiple myeloma (**Table 2**). Those genotypes did not appear in the short term (2–3 weeks after HSCT) to be associated with higher susceptibility to infection as their frequency did not differ significantly between patients experiencing hospital infections/febrile neutropenia compared with patients free of those complications [MM: 15/114 (13.2%) vs. 18/80 (22.5%, p = 0.12); LYMPH: 10/82 (12.2%) vs. 4/36 (11.1%, p = 1)]. The high MBL-conferring genotypes (YA/YA) were associated with a trend toward more prolonged fever during the neutropenic period, with 47.1% of patients having fever of more than 4 days duration being YA/YA, compared with 24.1% having fever for less than 4 days and 30.1% who had no infection or FN. The LYMPH group differed from the MM patients in having a high frequency of the uncommon C allele (codon 57 SNP) (0.025 vs. 0.007 for C; p = 0.049), but this was not associated with hospital infections.

One hundred and two MM patients were followed-up for at least 6 months, of which six were hospitalized due to severe infections. One was O/O and 2 were LXA/O; two of them died. Other 3 persons (1 non-survivor) had A/A genotypes. Similarly, from 59 LYMPH patients that were followed-up, three had severe infections: two were MBL-deficient (O/O; one non-survivor) and one A/A (survivor).

#### MBL2 Gene Polymorphisms Located Within 3′ -UTR Region (Exon 4)

Twelve polymorphic sites located within 3′ -UTR region of the MBL2 gene were analyzed. That allowed identification of 27 various genotypes, of which the 7 most common accounted for nearly 95% of controls and cases (**Table 3**). The genotype 1 (gt1) was commonest in the C and MM groups, but for the LYMPH group, genotype 2 was the most frequent. The difference between LYMPH and C groups with regard to genotype 1 (23.7 vs. 37.2%; p = 0.01) was statistically significant, possibly the first association of a MBL2 gene 3′ -untranslated region polymorphism with cancer to be observed in Caucasians.

#### MASP2 Gene Polymorphisms

When the +359 A>G mutation was investigated, no variant homozygote (and therefore MASP-2/MAp19 deficient) was found among individuals recruited to the study. The number of heterozygotes (**Table 4**) did not differ among the major (C, MM, LYMPH) groups, although a trend toward higher incidence of heterozygosity was noted within LYMPH group (**Table 4**). That primarily reflected relatively high frequency of A/G genotype among LYMPH patients who had bacteremia during hospital stay [8/34 (23.5%); p < 0.04 vs. C group].

For another MASP2 SNP, +1111 A>C, no differences between patients and controls were found. However, this variant allele was commoner within the LYMPH patients than within the MM group (p = 0.023) (**Table 4**).

#### Serum MBL Concentration and Activity of Its Complexes With Serine Proteases

The median values determined for samples taken before chemotherapy did not differ significantly in either MM (1027 ng/ml; n = 187) or LYMPH (1072 ng/ml; n = 116) groups when compared with that found for healthy controls (789 ng/ml; n = 265) (**Figure 1**). However, the median was higher in MM patients who later experienced hospital infections (associated or not with bacteremia) and LYMPH patients who suffered from infections without bacteremia (**Figure 1A**).

As was expected from numerous other studies, the median serum MBL for males and females in the healthy controls were virtually identical (789 vs. 787 ng/ml), and the higher values

TABLE 5 | Serum MBL concentrations in controls and patients (sample 1), corresponding to most common MBL2 genotypes (basing on 3′ -UTR polymorphisms).

Table 3). Blue bars present median values (given below group descriptions).


found in the LYMPH group also showed no sex difference (1090 vs. 1054 ng/ml). However, male subjects in the MM group had significantly higher (1213 vs. 729 ng/ml; p = 0.002) values, and consequently male MM patients had higher average serum MBL than male controls (p = 0.032).

Despite well-known association between MBL2 SNP located within promoter and exon 1 with MBL concentration (not shown), an influence of exon 4 polymorphisms has been observed. When the four most common genotypes were compared within the control group, significant differences in corresponding median MBL levels were found (**Figure 2**). Comparison of patients and controls showed MBL concentrations significantly higher in patients with genotype 3 for both MM and LYMPH groups (**Table 5**).

In both MM and LYMPH groups, MBL concentrations underwent significant changes (p < 0.000001; Friedman's ANOVA test) in relation to treatment (**Figure 3A**). The changes were generally more evident in persons experiencing complications during hospital stay (with the exception of LYMPH patients with bacteremias). (**Figure 3A**). Data from samples taken during control examinations (although from limited number of patients) demonstrated that the concentration of this lectin ∼100 days post-HSCT practically returns to the initial (sample 1) value (**Figure 4A**). In the case of patients suffering from lymphomas, MBL increased more rapidly than in those with diagnosed MM, probably depending on chemotherapy used (BEAM vs. MEL) (**Figure 3A**). Again, a decrease was observed after discharge and data from sample 100 were comparable to those from sample 1 (**Figure 4A**).

Unexpectedly, very high MBL concentration (>3000 ng/ml) before treatment seemed to be associated with higher risk of hospital-acquired infection in MM patients. Such a high level of the protein was found in 9 of 71 (12.7%) MM patients who experienced infections, compared with 13 of 265 controls (4.9%; OR 2.81; p = 0.029). No such relationship was found for the LYMPH group (8/116 cases, 6.9%). No difference was found in the frequency of serum MBL deficiency (<150 ng/ml) between controls (17.7%) and patients (MM: 17.6%; LYMPH: 16.4%). independently of complications (not shown).

As expected, since correlation coefficients between serum MBL and activity of its complexes with MASP-1 or MASP-2 were ≥0.9, the latter activities gave similar results (**Supplementary Figure 1**). Moreover, observed MBL variations in MM patients, slightly but significantly, inversely correlated with variations in WBC, ANC, PLT counts (changes in leukocyte counts are demonstrated in **Figures 4D–F**). In contrast, relatively strong, positive correlations with concentrations of such markers of inflammation as CRP or PCT (but not FBG) were found

FIGURE 3 | Changes in concentrations of MBL (A), CL-LK (B), and MASP-2 (C) in patients, during treatment. 1–blood taken directly before conditioning chemotherapy; 2 – blood taken directly before autologous hematopoietic stem cell transplantation; 3–blood taken 1 week after HSCT; 4—blood taken 2 weeks after HSCT. Data from sample 5 (3 weeks after HSCT) are not shown due to relatively low number of cases within each subgroup. Median values for each time-point are presented. MM-A, LYMPH-A: patients who experienced infections with proven bacteremia; MM-B, LYMPH-B: patients who experienced infections with no bacteremia; MM-C, LYMPH-C: patients who experienced febrile neutropenia; MM-D, LYMPH-D: patients who experienced none of afore-mentioned complications during hospital stay. C—controls (sampled once). Statistics: given p-values regard to Friedman's ANOVA while asterisks (in colors corresponding to curves) mark significant differences in comparison with sample 1 (Wilcoxon's paired sample test). Graphs show data from complete sets (MBL levels measured in all 1-4 samples) only. Numbers of patients are given in parentheses.

100—blood taken 100 days after HSCT (mean 104.1 ± 19.3 days).

(**Table 6**). It should be mentioned that WBC/ANC/PLT were checked routinely and regularly (taking samples 1-5 was coordinated with that) while concentrations of inflammatory proteins (especially FBG) were tested occasionally, in the case of morbid indications. Therefore, even relatively high correlation coefficient not always was associated with statistical significance (**Table 6**).

#### Concentration of CL-LK

The median level of CL-LK in pre-treatment samples from MM patients (423 ng/ml) was significantly higher than that from the control group (385 ng/ml; p = 0.0002) (**Figure 5A**) or the corresponding samples from LYMPH patients (388 ng/ml; p = 0.009). This difference did not seem to be associated with post-HSCT complications. No difference (p = 0.56) was found between healthy subjects and patients suffering from lymphoma (**Figure 5B**).

No significant differences in median CL-LK concentration between males and females were apparent in MM patients (420 vs. 429 ng/ml) or healthy controls (368 vs. 400 ng/ml). However, this value was significantly higher in LYMPH group females (430 vs. 371 ng/ml) compared with males (p = 0.004).

The highest CL-LK serum levels (>583 ng/ml) occurred as frequently among MM patients (7%) and LYMPH patients (6%) as controls (5.6%). However, low (<242 ng/ml) values were under-represented within the MM group (2.1%) compared with the C (5.6%, p = 0.002) or LYMPH (6%, p = 0.037) groups. It should be stressed that low CL-LK was found in none of the MM patients experiencing infections or febrile neutropenia (**Table 7**).

CL-LK concentration during treatment varied, but the changes noted were not so striking as for MBL. They were however significant when data from MM patients who suffered from bacteremia or FN were analyzed (Friedman's ANOVA: p < 0.000001 and p = 0.008, respectively) (**Figure 3B**). Within LYMPH group, variations were significant for persons affected by bacteremia only (p = 0.035). The level of CL-LK was generally lower at point 2 (before HSCT) compared with preceding and following samples. Interestingly, data from minority of patients who were sampled during control examinations suggested slight but significant increase of CL-LK at 45 and 100th days after HSCT (443 ng/ml; n = 26, p = 0.016 and 457 ng/ml; n = 23, p = 0.048 compared with data from sample 1). In the case of LYMPH, the median corresponding to samples 100 (384 ng/ml) was similar to that from samples taken before chemotherapy (**Figure 4B**).

Described variations were not associated with changes of blood cell counts or inflammatory markers (**Table 6**).

An incidental finding was a lack of detectable CL-LK in a healthy 45-year old woman. A repeat sample 2 months later yielded only a trace (<40 ng/ml). Sequencing of the COLEC10 gene did not reveal homozygosity for any known variant allele


TABLE 6 | Correlations (Spearman's) between variations (samples 1–5) of concentrations/activities of tested proteins/complexes and selected clinical parameters (significant positive and inverse correlations marked with blue and red, respectively).

A – patients suffering from multiple myeloma; B – patients suffering from lymphomas.

or "new" mutation when compared to the reference gene sequence (NCBI Reference Sequence: NC\_000008.11, accessed from: https://www.ncbi.nlm.nih.gov/gene/10584) (not shown).

#### Concentration of MASP-2

No significant differences were found when pre-treatment median MASP-2 serum concentrations from MM (357 ng/ml) or LYMPH (359 ng/ml) were compared with that from healthy controls (344 ng/ml) (**Figure 6**). However, a higher value was observed in the subgroup of MM patients who had infections with no bacteremia (478 ng/ml, p = 0.017 vs. C group). Furthermore, median pre-treatment MASP-2 was higher in MM patients suffering from infections with Gram-positive bacteria (403 ng/ml) relative to those infected by Gram-negative bacteria (343 ng/ml; p = 0.008).

Within the control group, a significant sex difference in median MASP-2 level was found (372 vs. 330 ng/ml for males and females, respectively, p = 0,019). No such relationship was found in MM (360 vs. 352 ng/ml) or LYMPH (347 vs. 371 ng/ml) groups.

When the frequencies of high (>628 ng/ml) MASP-2 concentrations were compared, no significant difference between the groups was found (C: 5.1%, MM: 6.4%, LYMPH: 2.6%). Similarly, the incidence of low (<172 ng/ml) MASP-2 was similar among healthy subjects (10.2%), MM patients (13.4%), and LYMPH patients (13.8%).

MASP-2 concentrations underwent highly significant treatment-associated changes in both groups of patients, independently of complications (**Figure 3C**). There was again an increase after treatment, but the kinetics differed from those of MBL. One and three months after discharge, median MASP-2

FIGURE 5 | Collectin CL-LK serum concentrations in patients (before conditioning chemotherapy) and controls. Blue bars present median values (given below group descriptions). (A) controls (C) vs. multiple myeloma (MM) patients; (B) controls vs. lymphoma (LYMPH) patients. MM-A, LYMPH-A: patients who experienced infections with proven bacteremia; MM-B, LYMPH-B: patients who experienced infections with no bacteremia; MM-C, LYMPH-C: patients who experienced febrile neutropenia; MM-A, LYMPH-A: patients who experienced none of afore-mentioned complications during hospital stay.

concentration in LYMPH patients was still higher than at point 1 (386 ng/ml, n = 40, p = 0.0003 and 419 ng/ml, n = 36, p = 0.006) while for MM group, the difference was insignificant (**Figure 4C**).

TABLE 7 | Frequency of low serum concentrations of collectin CL-LK in investigated groups (for MM and LYMPH groups: data from sample 1, taken before chemotherapy).


\*MM-A, LYMPH-A: patients who experienced infections with proven bacteremia; MM-B, LYMPH-B: patients who experienced infections with no bacteremia; MM-C, LYMPH-C: patients who experienced febrile neutropenia; MM-D, LYMPH-D: patients who experienced none of afore-mentioned complications during hospital stay.

In both MM and LYMPH groups, MASP-2 correlated positively with CRP and inversely with WBC, ANC and PLT counts (**Table 6**). In contrast to MBL, serum MASP-2 concentrations did not differ between either group of patients and controls in relation to MASP-2 genotypes (**Table 8**).

#### DISCUSSION

Both cancer and its treatment impair immunity, but the role of MBL and the significance of its deficiency in this context are controversial since contrasting and apparently contradictory findings have been reported. Particular attention has focused on the effect of chemotherapy, both during the immediate neutropenic period and during longer periods of follow-up. Here we have demonstrated with a large series of multiple myeloma patients given similar therapy (and more heterogeneous lymphoma group), that MBL deficiency has no influence on the incidence of infections/febrile neutropenia. However, over a 6-month period of follow-up, MBL deficiency was overrepresented in the small number of patients experiencing very severe infections. Our findings with a substantially larger series are not inconsistent with the studies of Molle et al. (46, 47) with a broadly similar group of myeloma patients.

These results could be interpreted as meaning MBL has little influence during the period of chemotherapyinduced cytopenia (and does not affect post-HSCT recovery of leukocytes, **Supplementary Figure 2**), but could have a protective effect when able to act in combination with phagocytic cells. That interpretation could explain why for patients receiving chemotherapy in general, MBL is found

to be effective in studies with a medium to long follow-up period (48, 49) but ineffective in short-term follow-up studies

experienced none of afore-mentioned complications during hospital stay.

(50, 51). In this study we confirmed the observation of others that chemotherapy is generally followed by an increase in MBL concentration (46, 48, 52). We found that after several weeks it returns to the initial level. A new observation was an association between MBL deficiency and myeloma itself at the TABLE 8 | Serum MASP-2 concentrations in controls and patients (sample 1), corresponding to MASP2 genotypes (+1111 A>C SNP).


genetic level. Furthermore, our data suggest a protective effect of "gt1" from lymphoma. If and when confirmed, this would be the first association of a MBL2 gene 3′ -untranslated region polymorphism with cancer in Europeans [those SNPs were previously reported to have associations with breast or colon cancer in African-Americans but not in Caucasians (53, 54)]. The influence of 3 ′ -end variant on serum MBL concentration was originally reported by Bernig et al. (53, 55). They suggested that Ex4-1067 A haplotype corresponds to high MBL. Indeed, in our study A/A homozygotes ("gt1" genotype) had higher median than A/G heterozygotes ("gt3") while, consequently, heterozygotes ("gt2") had higher values than G/G homozygotes ("gt4") (**Figure 2**). The Ex4-710/901/1067/1483 SNP were found to be components of a haplotype block (28, 54, 55). Our data (**Table 3**) are in agreement with that finding. Therefore, in general (although with very few exceptions), analyzing of one polymorphic site may provide information about the others (including also Ex4-718/845/879) (**Table 3**). Interestingly, the C allele corresponding to Ex4-1483 (associated with Ex4- 1067 G) is responsible for interaction with miR-27a, a microRNA molecule which was suspected to predict lower MBL plasma concentration (54, 56). Again, it has been to some extent confirmed here: "gt1" and "gt3" carriers (T/T homozygotes) have generally higher MBL than "gt2" and "gt4" (T/C genotype) (**Figure 2**). Obviously, an interplay between 3′ block and 5′ one (promoter/5′ -UTR/exon 1) has to be taken into account in considering genotype-phenotype relationships (28, 54, 56). For example, strong linkage disequilibria of Ex4-1067 with promoter H/L and Y/X SNP were proven [LDmatrix (https://analysistools. nci.nih.gov) (57)]. It is therefore noteworthy that the majority of "gt1" genotype carriers (74/95 controls and 61 of 87 patients) had YA/YA ("high MBL-producing") genotypes. Interestingly, the most prevalent 5′ end variant associated with "gt3" was YA/XA (34/47 controls and 52/69 patients; majority of remaining individuals carried at least one LXA allele) (not shown). That altogether confirms apparent links between Ex4-1067 A and promoter (position−221) Y alleles as well as G and X variants, respectively.

MBL, naturally complexed with MASP has been evidenced (employing animal model) to contribute to mobilization of hematopoietic cells from bone marrow to peripheral blood). That complex is able to trigger both complement and coagulation systems activation, cross-talking in mobilization process (58– 60). It has been suggested that MBL-deficient patients may be poor HSPC (hematopoietic stem progenitor cells) mobilizers even upon stimulation for transplantation (58, 59). It however has not been confirmed in our study: most of MBL deficient individuals (confirmed by both genotyping and MBL serum level) responded normally (not shown). It may be speculated that MBL deficiency is, at least partially, compensated by other lectin pathway-associated pattern-recognition molecules when MASP function is normal. Furthermore, it seems not to affect markedly recovery of leukocytes (at least when their counts recorded at ∼14th day after HSCT, see **Supplementary Figure 2**).

Another collectin, CL-LK (as mentioned being a complex of CL-10 and CL-11), was found at higher concentrations in myeloma patients compared with controls. One and three months after discharge (samples 45 and 100) its level exceeded an initial one. However, no clinical significance of extremely high or low CL-LK was apparent. This is the first possible association of CL-LK with a hematological malignancy to have been noted. During hospital stay, CL-LK underwent marked changes in myeloma patients affected by bacteremia or febrile neutropenia suggesting involvement of this collectin in the immune response against some potentially life-threatening events.

The results of a retrospective study indicated that higher MASP-2 serum concentrations are associated with a longer event-free survival (EFS) in children with lymphoma (especially HL) (61). We have conducted a prospective study, and collected data from follow up longer than 3–6 months from a relatively low number of patients. Furthermore most of them survived that period and had no relapse so our results cannot easily be compared with the previous report. As with MBL, higher serum level of its associated serine protease-2 before chemotherapy seemed to be associated with hospital infections, at least in MM patients and Gram-positive bacterial agents. Earlier, MASP2 variant allele (SNP at position +1111) was suggested to be protective against DLBCL (62). Among 31 DLBCL patients studied here, 7 (22.6%) were heterozygotes (no difference in comparison to C group). It should however be stressed that relatively high incidence of MASP2 +359 A/G heterozygosity was noted among LYMPH patients who experienced bacteremia while relatively low - of +1111 A/C heterozygosity in LYMPH group in general. Both mentioned genotypes influence MASP-2 serum concentration (but not MBL-MASP-2 complex activity when measured as C4 cleaving potency; not shown). The MASP2 +359 G variant abolishes the formation of the MBL-MASP-2 complex and heterozygotes present approx. half MASP-2 concentration in serum. However, this level suffices for full activity of the MBL pathway (37).

Cancers studied here seem in general to affect concentrations of complement activating collectins and MASP-2 disparately in males and females. As previously demonstrated by Troldborg et al. (63), MBL, CL-10, and CL-11 average levels do not differ between men and women while MASP-2 is higher in men. That was fully confirmed here for controls only. Neither in MM nor LYMPH group the latter differed depending on sex while MBL was found significantly higher in MM males compared with females and opposite effect was noted for CL-LK.

To summarize, we have studied the complex cross-talk between numerous endogenous and exogenous factors in the process of treating patients suffering from MM and LYMPH. Monitoring was started immediately before conditioning chemotherapy and continued for several weeks. The investigated factors, involved in activation of the complement and coagulation cascades, are undoubtedly important players in the events investigated, influencing disease/treatment course and outcome and also being influenced by disease itself and therapeutic procedures. Our data indicate multiple MBL involvements. We found its primary deficiency to be associated with MM risk but not hospital infections during cytopenia. Moreover, high MBL-conferring genotypes and high serum MBL were to some extent associated with adverse effects. Later, during follow-up, MBL deficiency seemed to predict higher morbidity and mortality due to infections. Additionally, we reported the first possible clinical association of a MBL2 3 ′ -UTR polymorphism (protective effect of "gt1" from lymphoma) in Caucasians.

# AUTHOR CONTRIBUTIONS

MC, KJ, AW, ASS co-authored the project and designed the ´ study. ASS and MC planned and supervised experimental ´ work which was done by ASS, AS, MM, AS-P, ŁE, and KK. ´ AW, SG, and KJ supervised qualification and recruitment of patients and/or controls. MN, IM, AS-K, MS-K, KM, OB, and AG were responsible for patients' qualification, taking and collecting samples as well as collecting clinical data/follow-up. MLK and AG qualified controls, provided DNA/serum samples and corresponding clinical data. ST and JCJ produced anti-CL-10 and anti-MASP-2 monoclonal antibodies and discussed the data. MC and MM performed the statistical analysis. AS revised the data. MC wrote the original draft of the paper. All authors contributed to the manuscript revision/correction and approved the version to be submitted.

# FUNDING

This work was supported by National Science Centre, Poland, grant UMO-2013/11/B/NZ6/01739.

# ACKNOWLEDGMENTS

We are very grateful to Dr David C. Kilpatrick for critical reading of the manuscript and helpful discussion.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02153/full#supplementary-material

#### REFERENCES


pathway of complement activation. Structure (2015) 23:342–51. doi: 10.1016/j.str.2014.10.024


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Swierzko, Michalski, Sokołowska, Nowicki, Eppa, Szala-Po ´ zdziej, ´ Mitrus, Szmigielska-Kapłon, Sobczyk-Kruszelnicka, Michalak, Gołos, Wierzbowska, Giebel, Jamroziak, Kowalski, Brzezinska, Thiel, Jensenius, Kasperkiewicz and ´ Cedzynski. This is an open-access article distributed under the terms of the Creative ´ Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Spontaneous Remission in Paroxysmal Nocturnal Hemoglobinuria—Return to Health or Transition Into Malignancy?

*Eva-Stina Korkama1 , Anna-Elina Armstrong2 , Hanna Jarva1,3 and Seppo Meri1,3\**

*<sup>1</sup> Immunobiology Research Program, Department of Bacteriology and Immunology, University of Helsinki, Helsinki, Finland, 2Coagulation Disorder Unit, Helsinki University Hospital Comprehensive Cancer Center, Helsinki, Finland, 3Helsinki University Hospital Laboratory (HUSLAB), Helsinki, Finland*

Paroxysmal nocturnal hemoglobinuria (PNH) is an acquired syndrome characterized by intravascular hemolysis, thrombosis, and bone marrow failure. The disease is caused by a mutation in the *PIG-A* gene that leads to the lack of glycosylphosphatidylinositolanchored complement regulatory molecules CD55 and CD59 on affected blood cell surfaces. In previous studies, spontaneous clinical remissions have been described. The disease manifestations are very heterogeneous, and we wanted to examine if true remissions and disappearance of the clone occur. In a follow-up of a nation-wide cohort of 106 Finnish patients with a PNH clone, we found six cases, where the clone disappeared or was clearly diminished. Two of the patients subsequently developed leukemia, while the other four are healthy and in clinical remission. According to our data, spontaneous remissions are not as frequent as described earlier. Since the disappearance of the PNH cell clone may indicate either a favorable or a poor outcome—remission or malignancy careful clinical monitoring in PNH is mandatory. Nevertheless, true remissions occur, and further studies are needed to understand the immunological background of this phenomenon and to obtain a better understanding of the natural history of the disease.

#### *Edited by:*

*Maciej Cedzynski, Institute for Medical Biology (PAN), Poland*

#### *Reviewed by:*

*Yoshiko Murakami, Osaka University, Japan Zoltan Prohaszka, Semmelweis University, Hungary*

> *\*Correspondence: Seppo Meri seppo.meri@helsinki.fi*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 29 May 2018 Accepted: 16 July 2018 Published: 02 August 2018*

#### *Citation:*

*Korkama E-S, Armstrong A-E, Jarva H and Meri S (2018) Spontaneous Remission in Paroxysmal Nocturnal Hemoglobinuria—Return to Health or Transition Into Malignancy? Front. Immunol. 9:1749. doi: 10.3389/fimmu.2018.01749*

Keywords: paroxysmal nocturnal hemoglobinuria, aplastic anemia, spontaneous remission, leukemia

# INTRODUCTION

Paroxysmal nocturnal hemoglobinuria (PNH) is an acquired syndrome characterized by intravascular hemolysis, thrombosis, and bone marrow (BM) failure (1). The disease is caused by a mutation in the *PIG-A* gene that leads to the lack of glycosylphosphatidylinositol (GPI)-anchored molecules, including the complement regulatory proteins CD55 and CD59 from the surface of a clonal lineage of blood cells (2–4). In addition, rare forms of PNH with mutations in the CD59 gene have been described indicating the key role of the absence of CD59 in the disease (5, 6).

The diagnosis of PNH is made by analyzing GPI-anchored molecules or the anchor itself on blood cells by flow cytometry. The disease course is often unpredictable, and the therapeutic options are limited. While in previous studies spontaneous remissions have been reported to occur in up to 15–30% of cases (7, 8), there are only few detailed case reports published on these patients. Recently, four cases of spontaneous remission were described in two different studies (9, 10), but doubts exist whether remissions really occur in PNH patients. In particular, it would be important to distinguish true remissions from other developing disorders, especially from malignancy, because in both cases the PNH diagnostics may give a negative result.

Because of the variable clinical course of PNH, we wanted to explore whether true spontaneous remissions occur in PNH in the Finnish patient material that we have collected during a long period of time. A nation-wide study of all PNH patients diagnosed in Finland since 1995 and an extended follow-up time (up to 20 years) have provided a unique opportunity to perform a detailed analysis of the course of the disease in our patients.

#### PATIENTS AND METHODS

In a nation-wide project, we collected patients from all Health Care Districts in Finland (with a total population of 5.6 million). The sources of information included the Helsinki University Central Hospital Laboratory (HUSLAB) databases, flow cytometry analysis of red blood cells (RBCs) and leukocytes, patients' medical records, and a patient questionnaire. The patients were evaluated until September 2016, and the remission cases were followed until June 2017. In total, 106 patients with a CD59 deficient cell clone in flow cytometry were included in the study.

Besides the flow cytometry analysis the laboratory parameters included levels of hemoglobin (Hb), lactate dehydrogenase, haptoglobin, erythrocytes, leukocytes, platelets, inflammation markers (CRP, ESR), D-dimer, fibrinogen, coagulation factor VIII, creatinine, and ventricular natriuretic peptide. In addition, the glomerular filtration rates and BM analyses were performed. The clinical data included the symptoms and conditions attributable to PNH: hemoglobinuria, thrombotic events, infections, fatigue, renal failure, pulmonary hypertension, abdominal pain, dysphagia, dyspnea, erectile dysfunction, anemia, and the co-existence of PNH with aplastic anemia (AA), myelodysplastic syndrome (MDS), or leukemia.

An ethical permission for the study was provided by the coordinating ethical committee at the Helsinki University Hospital Health Care District. A research permission was also obtained from the National Institute of Health and Welfare (THL, Helsinki, Finland), and the patients gave written informed consents for the study.

#### RESULTS

In a cohort of 106 Finnish patients with a PNH clone, we found 6 cases, where the clone disappeared or clearly diminished (to a level equal or below 1.5%). Two of the six patients subsequently developed leukemia (acute myeloid leukemia and chronic myelomonocytic leukemia), while the other four are in clinical remission (**Table 1**).

Patient number 1 was diagnosed with AA in 1986 during a twin pregnancy. The patient had pre-eclampsia and a cesarean section was performed. The PNH diagnosis was made by Ham's test in 1990 at the age of 23 years. The first flow cytometry test was performed in 2001 (24% CD59-negative erythrocytes) (**Figure 1A**). The patient was first treated with steroids without any positive outcome and with androgens for 2 years. In 1994, the patient was diagnosed with multiple sclerosis. In the mid 90s, the Table 1 | Patients and treatments.


*PNH, paroxysmal nocturnal hemoglobinuria; AA, aplastic anemia; BM, bone marrow. a Diagnosis by Ham's test. First flow cytometry in 2001. For further clinical information, please see the text.*

patient got several blood transfusions and twice more in 2005 and 2006. No thrombotic complications occurred. The patient was given anticoagulation treatment. Low molecular-weight heparin was used postoperatively after a gynecologic operation in 2006, and since 2008 the patient used aspirin. In 2012, erythrocytes, monocytes, and neutrophils were analyzed by the flow cytometry test, and no PNH clone was detected. The patient had no longer signs of hemolysis. In the same year, however, the patient was diagnosed with MDS and later with acute myeloid leukemia. The patient died of leukemia in 2013 at the age of 45 years.

Case 2 (female) was diagnosed with PNH in 1998 at the age of 56 years (30% CD59-negative erythrocytes by flow cytometry). At the time of diagnosis, the Hb level was 70 g/L, and the numbers of leukocytes 2.5 × 109 /L and platelets 6 × 109 /L. The patient had a tendency for bruising, petechiae, and bleeding on probing. In 1999, the patient's BM sample showed hypoplastic features. The patient was treated with corticosteroids from 1998 until 2005. Androgen treatment was stopped because of elevated liver values. The CD59-negative erythrocyte clone peaked in 1998 after which it gradually declined. In 1999, the patient received cyclosporine treatment, which had a good effect on the anemia. In 2007, the patient recurrently developed anemia and thrombocytopenia. At that time, the RBCs had a normal CD59 expression level on flow cytometry (**Figure 1B**). The BM sample showed myelodysplastic changes and a systemic mastocytosis with c-*KIT* mutation was observed. In 2010, the patient was diagnosed with chronic myelomonocytic leukemia and azacytidine treatment was started. In 2011, the patient died because of acute ischemic heart disease during the leukemia treatment.

Case 3 (female) was diagnosed with AA in 1995 at the age of 21 years. At the time of diagnosis, the Hb level was 50 g/L, leukocytes 1.9 × 109 /L, and platelets 8 × 109 /L. The patient was treated with antithymocyte globulin, prednisolone, cyclosporine, and granulocyte growth factor. In 1996, 8% of the erythrocytes were CD59 deficient. When the flow cytometry was performed next time in 1998, 50% of the erythrocytes was type I CD59-deficient cells and 30% type II (partially CD59 deficient) cells (**Figure 1C**). No thrombotic complications or bleeding have occurred at any time of the disease history. The lactate dehydrogenase level was

elevated earlier, but has been normal since 2006 (**Figure 1C**). The patient had two deliveries, in 2003 and 2007. During the first delivery, the patient was given blood transfusions and suffered from postpartum endometritis. The second delivery was without complications. In the 1990s, the patient had periods of hemoglobinuria. In 2012, the cyclosporine treatment was stopped, and only 0.2% of the erythrocytes were CD59 negative. Since 2013 the expression of CD59 has been normal on erythrocytes and the GPI-negative clone in neutrophils and monocytes has been 0.3–0.6% and the patient is symptom free.

and 2 (A,B) suffered from leukemia and subsequently died. Patients 3–6 are in remission.

Patient 4 (male) was diagnosed with AA in 1992, and the PNH diagnosis was made in 1998 when the patient was 25 years old (10% CD59-negative erythrocytes in flow cytometry). Before that the patient had had a small CD59-negative RBC clone (1.7–7%) since 1996. In 1992, the patient got several RBC and platelet transfusions. Cyclosporine and androgen treatment were started in 1992 with a good response leading to recovery from the cytopenias. The androgen treatment was discontinued in 2003 and cyclosporine treatment in 2012. Since then the patient has had no symptoms, and the laboratory parameters have been in the normal range. In 2010, CD59 expression on the erythrocytes was normal. In later controls, the expression has been either fully normal or the CD59-negative erythrocyte cell clone has been small (between 0.2 and 1.4%) (**Figure 1D**). The GPI-negative clone in neutrophils and monocytes has been in the range 0.2–1.5%.

Case 5 (male) was diagnosed with AA and PNH in 2000 at the age of 43 years. Fifty-three percent of the erythrocytes were totally CD59 deficient. In addition, 10% had a decreased expression (type II cells). In the BM sample, the cellularity was 25–30% of normal. At the time of diagnosis, the Hb level was 54 g/L, neutrophils 0.3 × 109 /L, and platelets 26 × 109 /L. The patient was initially treated with prednisolone for a short time. Later, the patient was given antithymocyte globulin treatment and cyclosporine for 1 year. Initially, the response to the immunosuppressive treatment was not sufficient, and an allogeneic stem cell transplantation with register donor was already being planned. In 2002, the transplantation was canceled because of a late response to the treatment. At that point, the Hb level was 122 g/L, neutrophils 0.77 × 109 /L, and platelets 96 × 109 /L. The patient had hemolysis and hemoglobinuria, fatigue, and erectile dysfunction. No thrombotic complications occurred. The anemia was compensated, and there was no transfusion dependency. Since 2009 the patient has been symptom free and the lab values stable, Hb level 155 g/L. The PNH clone has been declining (**Figure 1E**). In 2017, it was below 1% in both RBCs and neutrophils.

Patient 6 (male) was diagnosed with AA at the age of 43 years in 1997. At that time, the patient had pancytopenia and 5.4% of RBCs were CD59 deficient. The patient was treated with cyclosporine with a good response. In 1998, the size of the PNH clone was 52%. The patient had tendency for bruising, hemoglobinuria, and erectile dysfunction, but no thrombosis. Since 2009 the LD values normalized, and the patient did not report hemoglobinuria any more. In 2010, the size of the PNH clone had already clearly diminished, and in 2013, it was 0.3% in RBCs and 0.8% in neutrophils (**Figure 1F**).

Among all the six patients, the median time from the diagnosis until the remission was 12.5 years (range 6–15 years). For the two patients who developed leukemia, the diagnosis was made 26 and 12 years after the initial PNH/AA diagnosis, respectively. In one patient, the CD59-deficient cell clone disappeared 3 years earlier, and for the other one in the same year leukemia was diagnosed. For all six patients, the median CD59-deficient RBC clone size was 64% at highest (range 30–91%) (**Figure 1**). Four of the patients had classical PNH with signs of hemolysis. None of the patients received eculizumab therapy since it was not available at the time of diagnosis. All patients had underlying AA or another type of BM failure. Four of the patients were treated with cyclosporine, three with androgens, and five with corticosteroids. Two of the remission cases also got antithymocyte globulin, and one was given granulocyte growth factor. None of the patients was heavily transfusion dependent.

#### DISCUSSION

Paroxysmal nocturnal hemoglobinuria is a potentially serious illness with variable manifestations and outcomes. Reliably defined complete remissions from the disease have only rarely been described, and doubt exists whether a permanent cure from the disease can occur. In Finland, it has been possible to conduct a nation-wide study because it has a relatively small population (5.6 million) with a comprehensive public healthcare system, where rare cases are centralized to specialized healthcare. Thus, for this study, we were able to collect all patients diagnosed with PNH from a period of over 20 years, altogether 106 PNH patients. Indeed, among these we found six cases, where the PNH cell clones disappeared. In four cases, this was accompanied with spontaneous clinical remission, as well. By contrast, in two other cases—showing apparently the disappearance of the PNH clone—the disease developed into malignancy.

Our data show that spontaneous remissions are not as frequent as previously described, and the disappearance of the PNH cell clone can be related to dramatically different outcomes. In a Spanish patient cohort, spontaneous remission was reported to occur in as many as 30% of the patients (17/56), but the cases were not characterized further, and the spontaneous remission was not defined (8). We considered that the patients were in remission when they had no clinical signs of disease (i.e., no cytopenias, signs of hemolysis, or thrombosis) and no flow cytometry evidence for PNH. For the latter, a threshold level of 1.5% for GPI-anchor negative RBCs was employed, because levels lower than this may be artifacts due to staining or transportation. Small PNH-like clones may also be seen in healthy individuals (11). In a British material from the year 1995, spontaneous clinical remissions were reported in 15% of the patients (12/80) (7). The patients had a negative Ham's test. In five cases, flow cytometry was later performed, and in all cases, the RBCs expressed normal levels of GPI-linked proteins. However, possible misdiagnoses with the earlier less specific Ham's test cannot be excluded.

Nevertheless, as our study shows, true remissions occur, but the patients need to be carefully followed up for a potential emergence of malignancy. Among the six patients, where the PNH clone disappeared, four are healthy and have no treatment. The actual mechanisms or the reasons for the remissions are unknown. They may be spontaneous, have an immunological background, or be related to therapy. A proposed explanation has been that the PNH clone has a finite lifespan and that normal stem cells are capable of repopulating the BM (7). Recently, a case was described, where a decrease in the PNH clone did not restore normal hematopoiesis but instead was associated with clonal replacement (9). Notably, however, in this case, the PNH clone did not disappear but remained at a 15% level. All of our six patients had an underlying BM failure. Patients received treatments (e.g., cyclosporine, androgens, steroids, CSF, or antithymocyte globulin) that may also have affected the PNH cell clone or the cell clone could have lost its growth advantage (12) after repopulating the BM.

Paroxysmal nocturnal hemoglobinuria has previously been considered to be a monoclonal disorder, and the expansion of the GPI-deficient clone has been explained by extrinsic factors (13). Recent studies have shown that multiple clones are present, and a number of additional somatic mutations in the PNH clones have been detected (14, 15). The latter may lead to clonal selection. Partially, CD59-deficient type II cells are thought to emerge because of only a partially defective or absent PIG-A enzyme. As PNH is currently regarded as a multiclonal disease with clonal hierarchy (14) and a large number of cells are involved, a repairing mutation can not be the explanation for remission. In our view, the most likely possibility is that the immune system is eliminating cells through complement attack and/or by the formation of antibodies or by a cell-mediated immune response. If antibodies against cells develop, as may occur after blood transfusions, the clearance of CD59-negative cells could be further boosted. Cells lacking CD59 and CD55 would, after all, be more susceptible to complement-mediated clearance. Some antibodies may even be targeted against unique epitopes on GPI-anchor deficient cells.

Paroxysmal nocturnal hemoglobinuria is known to be a very variable disease, and this is also typical among the remission cases. The fact that the clone disappears may be related to different outcomes, total recovery, or BM suppression and development of leukemia. This provides a challenge to the clinician. Careful follow-up is needed to determine the direction to which the disease is progressing. Further studies on the mechanisms of spontaneous remissions could also provide clues for finding a curative treatment for PNH in the future.

#### ETHICS STATEMENT

An ethical permission for the study was provided by the coordinating ethical committee at the Helsinki University Hospital Health Care District. A research permission was also obtained from the National Institute of Health and Welfare (THL, Helsinki,

#### REFERENCES


Finland), and the patients gave written informed consents for the study.

# AUTHOR CONTRIBUTIONS

E-SK performed the research, analyzed data, and wrote the initial draft of the manuscript. HJ and SM designed and supervised the research, revised, and approved the manuscript. A-EA provided clinical information, reviewed, and approved the manuscript.

# ACKNOWLEDGMENTS

This work was supported by grants from Sigrid Jusélius Foundation, Signe and Ane Gyllenberg Foundation, Helsinki University Hospital Funds (EVO), and Alexion Pharmaceuticals Inc. The sponsors support science in general and had no role in gathering, analyzing, or interpreting the data. E-SK is a PhD candidate at University of Helsinki, and this work is submitted in partial fulfillment of the requirement for the PhD degree.


**Conflict of Interest Statement:** E-SK and SM have received an investigatorinitiated research grant and E-SK, A-EA, HJ, and SM have received speaker honoraria from Alexion Pharmaceuticals, Inc.

*Copyright © 2018 Korkama, Armstrong, Jarva and Meri. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Overactivity of alternative Pathway convertases in Patients With complement-Mediated renal Diseases

*Marloes A. H. M. Michels1 , Nicole C. A. J. van de Kar1 , Marcin Okrój2 , Anna M. Blom3 , Sanne A. W. van Kraaij4 , Elena B. Volokhina1,4† and Lambertus P. W. J. van den Heuvel1,4,5\*† On Behalf of the COMBAT Consortium*

*1Department of Pediatric Nephrology, Radboud Institute for Molecular Life Sciences, Amalia Children's Hospital, Radboud University Medical Center, Nijmegen, Netherlands, 2Department of Medical Biotechnology, Intercollegiate Faculty of Biotechnology, University of Gdan´sk and Medical University of Gdan´sk, Gdan´sk, Poland, 3Medical Protein Chemistry, Department of Translational Medicine, Lund University, Malmö, Sweden, 4Department of Laboratory Medicine, Radboud University Medical Center, Nijmegen, Netherlands, 5Department of Pediatrics/Pediatric Nephrology and Department of Development and Regeneration, University Hospitals Leuven, Leuven, Belgium*

#### *Edited by:*

*Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France*

#### *Reviewed by:*

*Lubka T. Roumenina, INSERM UMRS 1138, France Lourdes Isaac, University of São Paulo, Brazil*

#### *\*Correspondence:*

*Lambertus P. W. J. van den Heuvel bert.vandenheuvel@radboudumc.nl*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 16 December 2017 Accepted: 12 March 2018 Published: 04 April 2018*

#### *Citation:*

*Michels MAHM, van de Kar NCAJ, Okrój M, Blom AM, van Kraaij SAW, Volokhina EB and van den Heuvel LPWJ (2018) Overactivity of Alternative Pathway Convertases in Patients With Complement-Mediated Renal Diseases. Front. Immunol. 9:612. doi: 10.3389/fimmu.2018.00612*

Overactivation of the alternative pathway of the complement system is associated with the renal diseases atypical hemolytic uremic syndrome (aHUS) and C3 glomerulopathy (C3G). C3 nephritic factors (C3NeF) play an important role in C3G pathogenesis by stabilizing the key enzymatic complex of complement, the C3 convertase. However, the reliability of assays detecting these autoantibodies is limited. Therefore, in this study, we validated and optimized a prototype hemolytic method for robust detection and characterization of factors causing convertase overactivity in large patient cohorts. The assay assesses convertase activity directly in the physiological milieu of serum and therefore is not restricted to detection of stabilizing autoantibodies such as C3NeF but may also reveal genetic variants resulting in prolonged convertase activity. We first defined clear cutoff values based on convertase activity in healthy controls. Next, we evaluated 27 C3G patient samples and found 16 positive for prolonged convertase activity, indicating the presence of factors influencing convertase stability. In three patients, the overactive convertase profile was persistent over disease course while in another patient the increased stability normalized in remission. In all these four patients, the convertase-stabilizing activity resided in the purified immunoglobulin (Ig) fraction, demonstrating the autoantibody nature. By contrast, the Igs of a familial aHUS patient carrying the complement factor B mutation p.Lys323Glu did not reveal convertase stabilization. However, in serum prolonged convertase activity was observed and segregated with the mutation in both affected and unaffected family members. In conclusion, we present a robust and reliable

**Abbreviations:** aHUS, atypical hemolytic uremic syndrome; AP, alternative pathway; C3G, C3 glomerulopathy; C3GN, C3 glomerulonephritis; C3NeF, C3 nephritic factor; CR1, complement receptor 1; DAF, decay-accelerating factor; DDD, dense deposit disease; EDTA, ethylenediaminetetraacetic acid; EDTA–GVB, EDTA–gelatin veronal buffer; FB, complement factor B; FD, complement factor D; FH, complement factor H; FHaAb, FH autoantibodies; FI, complement factor I; Ig, immunoglobulin; MAC, membrane attack complex; MCP, membrane cofactor protein; Mg–EGTA, magnesium–ethylene glycol tetraacetic acid; NHS, pooled normal human serum; NHP, pooled normal human plasma.

method for the detection, characterization, and evaluation over time of factors prolonging convertase activity (C3NeF or certain mutations) in patient cohorts. This assay may provide new insights in disease pathogenesis and may contribute to the development of more personalized treatment strategies.

Keywords: complement system, alternative pathway, convertase, C3 nephritic factor, C3 glomerulopathy, atypical hemolytic uremic syndrome, complement factor B mutation

#### INTRODUCTION

The complement system, a cornerstone of innate immunity, provides the body with protection against invading pathogens and dangerous host cells (1, 2). The system can be activated *via* three pathways—the classical, lectin and alternative pathway (AP)—depending on the initial trigger that is encountered. All pathways converge at the central event of complement activation: the cleavage of C3 by C3 convertases. This enzymatic reaction supports further activation of the complement cascade with the release of various anaphylatoxins and opsonins, and eventually the formation of the membrane attack complex (MAC). These events all contribute to inflammation, phagocytosis and the damage of susceptible targets (1–3).

In contrast to the classical and lectin pathway, which are only activated upon pattern recognition, the AP is continuously mildly activated and therefore specifically serves as a surveillance mechanism. C3 is subjected to spontaneous hydrolysis of its internal thioester bond at a very low rate, and this generates the active C3(H2O) molecule. Hydrolyzed C3 can bind to complement factor B (FB), which is subsequently cleaved by complement factor D (FD) to form an initial, fluid phase AP C3 convertase, C3(H2O)Bb, that is capable of converting native C3 into the active C3a (anaphylatoxin) and C3b (opsonin) fragments. This mechanism providing a persistent low level of active C3 in the blood is known as "tick-over" and allows constant responsiveness to potential danger. Besides the tick-over mechanism, the AP may also be initiated by C3 convertases of the classical and lectin pathway that generate active C3b (1, 4, 5). Furthermore, C3 may be cleaved into C3b by certain non-specific proteases, especially at sites of inflammation, coagulation, and infection (6).

Activated C3b molecules expose a reactive thioester moiety *via* which they can subsequently bind to hydroxyl or amino groups on target surfaces in close proximity to the activation site. This C3b binding, and thus initiation of AP activity, is preferentially triggered by certain carbohydrate structures on microbes and other foreign surfaces, e.g., LPS and zymosan. By interacting with FB and FD, target-bound C3b can subsequently form new surface-bound convertases (C3bBb) that can be further stabilized by properdin (C3bBbP). Properdin has also been proposed to act as a pattern recognition molecule able to initiate the AP on certain targets by recruiting C3b and FB. Once formed, AP C3 convertases can massively amplify the immune response by converting more C3 molecules into C3b, which in turn support new convertase formation. Incorporation of C3b into existing C3 convertases generates C5 convertases (C3bBbC3b), which cleave C5 into the potent anaphylatoxin C5a and the fragment C5b that initiates the assembly of the MAC (C5b-9) complex. The amplification loop of the AP can also be induced by C3b molecules generated by C3 convertases from the other two activation pathways. In this way, the AP can account for over 80% of total complement activity. Thus, besides being a surveillance system, the AP acts as an important amplifier of initiated complement responses (1, 4, 5).

The powerful action of the AP requires sophisticated regulation to prevent damage by excessive activation or self-attack. Therefore, human cells express membrane-bound complement inhibitors, including decay-accelerating factor (DAF), membrane cofactor protein (MCP), and complement receptor 1 (CR1). Together with the soluble regulatory proteins complement factor H (FH) and complement factor I (FI), which can act in fluid phase or can be recruited to surfaces, these complement regulators prevent the formation of potent C3 convertases on healthy host surfaces and keep AP activity in control. The inhibitors of the C3 convertase can act *via* two ways. The first is by accelerating the convertase decay and is fulfilled by DAF, CR1, and FH. These regulators can also compete with FB for binding to C3b and thereby inhibit new convertase formation. The second is by preventing convertase formation *via* acting as cofactors for FI-mediated inactivation of C3b. This action is supported by MCP, CR1, and FH (7–9). However, genetic aberrations in complement genes and/or autoantibodies against complement components may, in combination with triggering events, disturb this sophisticated regulation, causing overactivation of the system. A dysregulated AP has been particularly associated with the renal diseases the atypical hemolytic uremic syndrome (aHUS) and the disease entity C3 glomerulopathy (C3G) (10–15).

C3 glomerulopathy is a recently defined umbrella classification for severe renal diseases that are characterized by C3 accumulation in the glomeruli without or with sparse immunoglobulin (Ig) deposition. The main two diseases encompassed by C3G are dense deposit disease (DDD) and C3 glomerulonephritis (C3GN) (16). Up to 50% of patients progress to end-stage renal disease within 10 years after first presentation (17). The most important pathogenic factors in C3G are autoantibodies against the C3 convertase of the AP named C3 nephritic factors (C3NeF), although pathogenic variants in complement genes have also been reported (14, 15). By binding to neoepitopes of the formed C3bBb(P), C3NeF stabilizes the otherwise labile convertase and thereby prolongs its half-life (18). C3NeF has been detected in approximately 80–90% of DDD patients (15, 19, 20) and in 40–50% of patients with C3GN (15, 21).

Atypical HUS is a form of thrombotic microangiopathy and is not associated with C3NeF, but pathogenic mutations in complement genes or autoantibodies against FH are the most common cause of improper complement regulation (22–25). Blocking C5 and terminal pathway activation with the C5-directed antibody eculizumab is a very effective treatment for this disease. Eculizumab became the first approved complement-inhibiting drug for aHUS in 2012 (26, 27). By contrast, trials of eculizumab treatment in C3G patients have shown contradicting outcomes so far. Thus, the value of eculizumab as a therapy for C3G requires further investigation (17, 28, 29).

Detection of C3NeF is not only helpful in diagnosis of C3G but also in choice of therapy for the patient, especially in the light of upcoming complement-inhibiting strategies. Nevertheless, a robust and reproducible method for C3NeF detection is still lacking. A recent European quality assessment round reported a success rate of only 50% among the participating laboratories (30). This also hampers our understanding of the exact role of C3NeF in C3G pathogenesis. C3NeF are functionally heterogeneous (31) and conflicting associations have been found between its presence and the patient's disease state and progression (20, 32–34). Some reports even describe C3NeF in healthy individuals (35, 36).

Recently, we described a new hemolytic method for measurement of convertase activity in the physiological milieu of whole serum. This assay allows the robust detection of factors prolonging convertase stability in a direct way, such as C3NeF, and/or in an indirect way by impairing the regulation of the convertase (37). For convenience, we will use the term convertase-stabilizing factors for both these types of factors that are either directly or indirectly prolonging the convertase half-life. The assay uses C5-blocking agents [eculizumab or the tick protein *Ornithodoros moubata* complement inhibitor (OmCI) (37, 38)] to strictly separate the complement cascade into two steps: the formation of C3/C5 convertases by test sera in a first step and the formation of lytic MAC complexes in a standardized second step for readout. Dynamics of convertase assembly and decay in the first step are then monitored in time. In this work, we further modified this convertase activity assay to screen for convertase-stabilizing factors in patient cohorts by defining clear cutoff criteria based on healthy controls. In addition, we showed expanded analysis of clinical samples with different types of convertase-stabilizing factors and their presence over disease course.

#### MATERIALS AND METHODS

#### Sample Collection

Whole blood was obtained from patients and healthy controls and processed within 1 h after sampling. For serum, blood was allowed to clot at room temperature for 30–45 min followed by centrifugation (10 min, 2,500 *g*, 4°C). To obtain plasma, blood was drawn into tubes with the anticoagulant ethylenediaminetetraacetic acid (EDTA) that were immediately placed on ice and subsequently centrifuged according to the same procedure. Afterward, the serum and EDTA plasma fractions were collected, aliquoted, and stored at −80°C until use. For the healthy controls, following exclusion criteria were applicable: fever, bacterial/ viral infection in previous 2 weeks, chronic illness, inherited or acquired immune disorders, and immunosuppressive medication. In addition, pooled normal human serum (NHS) and pooled normal human plasma (NHP) were made by pooling 15 and 12 individual control samples, respectively. This study was carried out in accordance with the recommendations of the appropriate version of the Declaration of Helsinki. All subjects gave written informed consent, and the protocol was approved by the ethical committee of the Radboud University Medical Center (FH 06979).

#### Erythrocyte Working Suspensions

Rabbit erythrocytes (RbE) in Alsever's solution were obtained from Envigo (Venray, the Netherlands). The working suspensions of RbE were prepared by washing them with magnesium–ethylene glycol tetraacetic acid (Mg–EGTA) buffer (2.03 mM veronal buffer, pH 7.4, 10 mM EGTA, 7 mM MgCl2, 0.083% gelatin, 115 mM d-glucose, and 60 mM NaCl) until there was no visible hemoglobin left in the supernatant. To standardize the number of erythrocytes in each experiment, these working suspensions were calibrated so the absorbance measured at 405 nm of 10× diluted erythrocytes in water was between 0.8 and 1.2.

#### Ig Purification

The Ig fractions from NHP and from the EDTA plasma of patients P24, P25, P26, and P27 (**Table 1**) were isolated by affinity purification using NAb™ protein A/G 5 mL spin columns (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's protocol. In short, 0.85–1.0 mL of EDTA plasma was diluted with binding buffer (0.1 M phosphate, 0.15 M sodium chloride, and pH 7.2) to 10 mL and loaded on the columns. Unbound fractions were washed with binding buffer, and thereafter bound Igs were eluted using 0.1 M glycine, pH 2.5 followed by immediate neutralization with 1/10 volume 1 M Tris, pH 8.5. Subsequently, Ig fractions were dialyzed against phosphate buffered saline using SnakeSkin® dialysis tubes (Thermo Fisher Scientific, Waltham, MA, USA) with a 10 kD molecular weight cutoff and concentrated to the initial plasma sample volume using Amicon® Ultra-15 centrifugal filter devices (Merck Millipore, Billerica, MA, USA) with similar cutoff. Total protein concentrations in these samples were measured using NanoDrop Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and varied between 4.73 and 12.48 mg/mL, which are within the normal range for Igs.

# Convertase Activity Assay

Detection of AP convertase-stabilizing factors in human serum was achieved using a two-step hemolytic assay as previously described (37) and as further specified in **Figure 1**. Briefly, per experimental time point 10 µL of prepared RbE were mixed with 20 µL Mg–EGTA containing 150 nM of the C5 inhibitor eculizumab (Alexion Pharmaceuticals, Cheshire, CT, USA), after which at different time points an additional 20 µL Mg–EGTA containing test serum or Ig fractions mixed 1:1 with NHS was added for convertase assembly (5, 10, 15, 20, 30, 40, 50, or 60 min). Final concentrations of serum were optimized per batch of erythrocytes and were generally 3.75 or 5%. Subsequently, cells were washed with 150 µL 40 mM EDTA–gelatin veronal buffer (EDTA–GVB, 4.41 mM veronal buffer, 0.1% gelatin, 130 mM NaCl, pH 7.4) and collected by centrifugation (2 min, 1,000 *g*, room temperature).

#### Table 1 | Clinical and genetic data.


*a No distinction possible into C3GN or DDD, since no electron microscopy was performed or no data available.*

*bAll genetic variants found were heterozygous.*

*c To the best of our knowledge, not previously reported in context of C3G.*

*dPreviously reported in a DDD patient.*

*e Previously reported in C3GN patients.*

*f Previously reported in aHUS patients.*

*g C3 levels were measured by nephelometry with the normal range indicated between brackets. Data given are those available as closest to the time of convertase activity assessment. For samples obtained from peripheral centers and from which C3 data were accessible, results are presented as "normal" or "decreased" since normal ranges vary between centers.*

*aHUS, atypical hemolytic uremic syndrome; C3G, C3 glomerulopathy; C3GN, C3 glomerulonephritis; DDD, dense deposit disease; NA, not available.*

Thereafter, cells were incubated for 60 min with 50 µL of 2.5% guinea pig serum (Envigo, Venray, the Netherlands) in EDTA– GVB and another 50 µL of Mg–EGTA buffer to develop MAC and subsequent hemolysis. All incubation steps were performed at 37°C, with 600 rpm agitation in a VWR® Incubating Microplate Shaker with 3 mm orbit (VWR International, Radnor, PA, USA) in V-shaped 96-well plates (Greiner Bio-One, Kremsmünster, Austria). Finally, supernatant was collected by centrifugation and transferred to flat-bottom 96-well plates (Greiner Bio-One, Kremsmünster, Austria). Hemolysis was quantified as percentage of full lysis by an equal amount of erythrocytes in water: (*A*405 test sample − *A*405 blank)/(*A*405 full lysis − *A*405 blank) × 100.

#### Statistical Analysis

Where indicated, data were analyzed using two-way analysis of variance with Bonferroni's post test using GraphPad Prism 5.03 for Windows (GraphPad Software, San Diego, CA, USA).

#### RESULTS

#### Study Cohort

A group of 27 patients diagnosed with C3G and referred to the Radboud University Medical Center were included in the study. Diagnosis was suspected based on presence of clinical features such

sample. The C5 inhibitor eculizumab is added to halt complement activation at the level of the C3/C5 convertases. Convertase assembly and decay are followed over time using different incubation times (5–60 min). Convertase-stabilizing factors present in the sample, e.g., C3 nephritic factor (C3NeF), may interfere at this point with convertase decay. Before proceeding to step 2, erythrocytes are washed to remove remaining complement factors and C5 inhibitor. Then, convertasebearing erythrocytes are incubated for 60 min with guinea pig serum as a source of membrane attack complex (MAC) components. The presence of ethylenediaminetetraacetic acid (EDTA) disables *de novo* formation of convertases from guinea pig serum and assures that only preformed convertases of the first step may initiate MAC formation and subsequent hemolysis. The released hemoglobin is quantified by spectrophotometric measurement and reflects the activity of the preformed convertases in step 1 per experimental time point. These data are used to generate convertase activity profiles over time. \*If desired, immunoglobulin fractions may be added to NHS to dissect the nature of the stabilizing factor (see Materials and Methods).

as hypertension, proteinuria, and nephrotic/nephritic syndrome, possibly combined with low serum C3. Diagnosis was confirmed after pathological judgment of renal biopsy following the recommendations defined in the consensus report of the first C3G Meeting (16). Subdivision of C3G into C3GN and DDD was solely based on electron microscopy appearance. In total, 9 patients were diagnosed with DDD, 14 with C3GN, and in 4 patients no distinction between the C3G subforms could be made since data required for subdivision were not available. In this C3G cohort (age range: 5–66 years; 17 children, 10 adults), no FH autoantibodies (FHaAb) were detected, and 6 patients carried mutations in complement (regulating) genes (**Table 1a**). Besides samples from C3G patients, sera from a female patient with familial aHUS and six of her affected and non-affected family members were available for analysis. In total, three of these family members were carrier of the heterozygous FB p.Lys323Glu mutation and two of them were also affected with aHUS (**Table 1b**).

# Convertase Activity in Healthy Donors

To define normal convertase activity, we tested the sera of 15 healthy controls (**Figure 2**). Due to low serum C3 levels *in vivo*,

many C3G samples show low hemolytic activity. To compensate for this possible complement consumption, all samples in this study, including those from controls, were diluted 1:1 with NHS. We previously showed that this approach restored hemolytic activity while still allowing the detection of convertase-stabilizing factors in patient serum (37) (also see Figure S1 in Supplementary Material). The convertase activity profiles of healthy controls demonstrated a constant pattern, which was comparable to that of NHS. Convertase activity reached its maximum after 10–15 min of incubation, and at 30 min activity returned to background levels in all control samples.

The ratio between the highest achieved hemolysis (top) and the lysis at the point at which controls have returned to background levels, in this case at *t*= 30 min (*t*30), was 5.4 ± 2.7 (mean ± 2 SD) for healthy controls. Prolonged convertase activity, i.e., delayed decay, was defined as a top/*t*30 ratio below the mean − 2 SD of controls, here <2.7. Because samples with a low hemolytic activity overall are also likely to be positive for this criterion, only samples that have a total area under the curve equal to or higher than that of NHS are considered positive.

#### Convertase Activity in Patients

Next, we analyzed convertase activity in the cohort of 27 C3G patients. Samples were first screened at *t*30 in a single time point screening (**Figure 3A**). The 21 samples that at *t*30 showed lysis above that of NHS (18.7%) were selected for further analysis of convertase activity over time. All other samples were considered negative for the presence of convertase-stabilizing factors.

Of the 21 patients who were selected for analysis of the convertase activity profile, 16 showed a top/*t*30 ratio below 2.7 (**Figure 3B**). These samples were considered positive for convertase-stabilizing factors. The other five samples, selected in the initial screening as possibly positive, did not meet these criteria and were therefore considered to have normal convertase activity profiles (**Figure 3C**).

#### Convertase Activity Over Disease Course

Subsequently, we investigated whether convertase activity profiles could change over disease course. From four patients who showed increased convertase stability, P24–P27 (**Table 1**), serum samples were available derived over different periods of their disease, e.g., from active disease states and/or (partial) remission periods. The active disease state samples of P24 and the active disease and partial remission samples of P27 all displayed prolonged convertase activity (**Figures 4A,B**). Also P25 showed a persistent stabilized convertase profile in both active disease and in partial and complete remission (**Figure 4C**). By contrast, the prolonged convertase activity shown for P26 was only detected in the acute phase and normalized in remission (**Figure 4D**).

#### Prolonged Convertase Activity due to Autoantibodies

For these four patients, we then investigated whether the convertase-stabilizing effect was due to autoantibodies such as C3NeF (**Figure 5**). To this end, the Ig fractions from these patients were purified using protein A/G affinity chromatography. Samples were concentrated in phosphate buffered saline to the initial plasma sample volume. Addition of an equal volume of the purified Ig fractions to NHS revealed comparable stabilizing effects on the convertases as seen for these patient's serum samples. This stabilization could not be observed when healthy control Igs were added to NHS. Thus, in these patients prolonged convertase activity could be attributed to autoantibody activity.

#### Prolonged Convertase Activity due to FB Mutation in aHUS Family

Previously, we described prolonged convertase activity in the serum of an aHUS patient carrying the p.Lys323Glu mutation in FB (37). In this study, we analyzed convertase activity in available serum samples of affected and unaffected family members of this patient (**Figure 6A**). Stabilized convertase profiles segregated with the FB mutation in both affected and non-affected family members (**Figure 6B**). When purified total Igs of the index patient were added to NHS, no stabilization of the C3 convertase could be observed (**Figure 6C**), unlike Ig fractions from C3G patients (**Figure 5**). This indicates that in these individuals, prolonged convertase activity could be attributed to the FB mutation rather than to autoantibody activity.

#### DISCUSSION

Previously, we described proof-of-concept of a robust method to measure C3 convertase activity allowing detection of factors influencing convertase activity in whole serum. Here, we further standardized assay conditions for screening in patient cohorts and for critical evaluation of convertase-stabilizing factors in clinical samples.

We showed low variation among healthy controls in the assay (**Figure 2**) and based on this defined clear cutoff criteria for assessing the presence of convertase-stabilizing factors in patient samples. The samples were considered positive when they showed a top/*t*30 ratio lower than the mean − 2 SD of healthy

Figure 3 | Screening for convertase-stabilizing factors in patients with C3 glomerulopathy (C3G). (A) Convertase activity of 27 patients with C3G as measured in a single time point screening with 30 min of incubation time. Samples showing hemolysis above that of pooled normal human serum (NHS), represented by the dotted line, are indicated with colored symbols and were selected for further analysis of convertase activity over time. (B,C) Convertase activity profiles of the 21 patients selected from the *t*30 screening in panel (A). Samples positive for convertase-stabilizing factors (top/*t*30 ratio < 2.7) are given in panel (B); samples negative for the presence of convertase-stabilizing factors (top/*t*30 ratio > 2.7) in panel (C). Results given are from a single assay set representative for at least two performed on each sample. (A–C) Serum samples were all tested mixed 1:1 with NHS to a final concentration of 3.75%. Colors and symbols correspond to the same patients in all panels. Hemolysis levels are given as percentage of full lysis of erythrocytes in water.

controls in their convertase activity profile. Cutoff values have to be evaluated for each laboratory individually. When large cohorts of patients have to be analyzed, the samples may be first screened for high lysis at a single time point (e.g., *t*30) and compared with NHS (**Figure 3A**). This efficient approach allows to save both time and patient material, but we underline that it should supplement and not replace the evaluation of a complete convertase activity profile to confirm positivity. In this work, we only used serum samples to assess convertase activity, but from our unpublished data we know that EDTA plasma samples may be used as well; no complement-inhibiting effects of EDTA are observed under the described test conditions (Figure S2 in Supplementary Material). Furthermore, as an alternative for eculizumab, which is expensive and may be difficult to access, the tick protein OmCI [or its commercially available variant Coversin (43)] may be used for C5 blockage in the first step of the assay (37, 38). Future research may benefit from cheaper alternative C5-blocking agents.

The positive samples showed high variation in the degree of convertase stabilization: where some sera showed active convertases up to *t*60, other sera showed delayed decay with breakdown of convertases between 30 and 60 min (**Figure 3B**). First of all, this may be attributed to the heterogeneity of C3NeF autoantibodies, which recognize different neoepitopes of the convertase complex. All C3NeF prevent the natural intrinsic decay of the convertase (44), but they vary in the degree of resistance they provide against extrinsic decay mediated by complement regulators (31, 44–47). Two reports described C3NeF that were completely unable to interfere with complement regulators and the authors indicated them as likely less pathogenic *in vivo*, since patients with these C3NeF did not present with hypocomplementemia (31, 47). Besides, the interaction of C3NeF with the convertase can be dependent on properdin or be independent of this positive complement regulator. Properdin-dependent and -independent C3NeF types may differ in the degree of convertase stabilization they provide (31) and have been related to different degrees of terminal pathway activation observed in patients (48, 49). A recent study identified properdin-dependent C3NeF as C5NeF: nephritic factors stabilizing C5 convertases (50). Since in our assay whole serum conditions are used in which properdin is present, our assay should be able to detect those factors as well. Furthermore, few previous reports described autoantibodies against the individual components of the convertase, i.e., FB and

3.75% in experiments shown in panels (C,D) or 5% in the experiments shown in panels (A,B), since they were performed using different batches of erythrocytes. Abbreviations: Acu, acute phase at first clinical presentation of disease; Act, active disease; Part Rem, partial remission; Rem, remission. Heat-inactivated NHS (ΔNHS) and an NHS sample from which guinea pig serum were omitted during the second part of the assay (*no step 2*) served as negative controls for the first and second steps, respectively. Representative data are given. Hemolysis levels are given as percentage of full lysis of erythrocytes in water.

C3b, which may also increase the convertase half-life (19, 51–53). However, not much is known of the role of these autoantibodies and their occurrence in C3G patients, which tends to be lower than the occurrence of C3NeF (19, 52). In addition, our previous work indicated that samples of aHUS patients positive for FHaAb may also show C3NeF-like stabilizing activity in the assay (37). In this cohort, we also checked for presence of FHaAb, but no patients were found positive. In conclusion, this variety in factors influencing convertase activity *via* different mechanisms may reflect the heterogeneity of the stabilized convertase activity profiles in our cohort. To distinguish between these factors, additional methods using purified components, e.g., enzyme-linked immune sorbent assay (ELISA) techniques, are needed.

In total, 16 out of the 27 patients (59%) of the C3G cohort were positive in this assay (**Figure 3B**); 67% positive in DDD patients, 57% in the C3GN patients, and 50% in the C3G patient group that could not be subdivided into DDD or C3GN. Of these 16, 13 were children and 3 were adults. Considering that of all convertasestabilizing factors described earlier C3NeF is most common in C3G, our data are in line with the numbers of C3NeF positive patients reported by other studies (15, 19–21, 54).

Commonly used assays for C3NeF are often based on studying isolated interactions of C3NeF with convertases assembled out of purified proteins on artificial surfaces (ELISA) or on sheep erythrocytes (semiquantitative hemolytic assay). In contrast to RbE, these do not naturally activate the AP and therefore require complex and time-consuming steps to build up the enzyme. Furthermore, in some studies, different detection assays were used alongside each other. Patients can be variably positive in these different assays, again indicating the heterogeneous nature of C3NeF (19, 31).

An important advantage of our method is the direct assessment of convertase stability in whole, unhandled serum, which

patients or of pooled normal human plasma (NHP) were added to 5% NHS in an equal volume. Data were collected from three independent experiments; means are given with error bars showing SDs. Statistical analysis for test samples compared with NHP Ig from *t* = 20 to *t* = 60 as calculated using two-way analysis of variance is given for the patient Igs only: \*\**P* < 0.01, \*\*\**P* < 0.001, ns not significant. Heat-inactivated NHS (ΔNHS) and an NHS sample from which guinea pig serum were omitted during the second part of the assay (*no step 2*) served as negative controls for the first and second steps, respectively. Hemolysis levels are given as percentage of full lysis of erythrocytes in water.

allows easy screening for any factor directly or indirectly affecting convertase stability, not limited to C3NeF. Moreover, convertases are studied in their physiological surroundings where (negative and positive) complement regulators and other serum factors are present to interact with the convertases. For this reason, we believe our method is particularly suited to detect functionally relevant C3NeF and other stabilizing factors. The assay is also well suited to screen for convertase-influencing factors in other pathologies associated with complement dysregulation, such as age-related macular degeneration, immune complex-mediated membranoproliferative glomerulonephritis, postinfectious glomerulonephritis, systemic lupus erythematosus, and acquired partial lipodystrophy (30, 55, 56). However, future research may focus on further adaptation of the assay with human (renal) cells as platforms for convertase assembly, since they best reflect physiological circumstances regarding the expression of membrane-bound inhibitors and binding places for the soluble inhibitor FH.

A limitation of our assay's serum approach is that it does not directly give insight into the nature (autoantibody or genetic variant) and mechanism of the convertase-stabilizing factor present in the patient's serum. When clinically relevant, e.g., to evaluate whether B cell-depleting therapy may be indicated, this nature can be dissected by assessing stabilizing activity in purified Ig fractions, screening for FHaAb, and genetic analysis, as we described in this study.

We have shown that the convertase-stabilizing activity in patients P24–P27 resides in the Ig fraction (**Figure 5**), indicating the presence of stabilizing autoantibodies. For these patients, we also described the convertase activity during disease progression. In three of them (P24, P27, and P25; **Figures 4A–C**), convertasestabilizing factors were present in all samples taken during the different stages of disease. By contrast, the stabilizing Igs detected in the acute phase of P26 became undetectable in remission (**Figure 4D**). These data are in line with several previous reports on C3G/membranoproliferative glomerulonephritis cohorts that

showed a fluctuation in C3NeF activity over disease course in at least one-third of all positive patients (15, 20, 34). Whereas in some patients C3NeF were persistently present, in others they appeared or disappeared during follow-up. Above that, in these studies no correlation was found between C3NeF presence, plasma C3 levels, and clinical recovery. More longitudinal studies are needed to further examine the contribution of C3NeF to disease progression and to establish its value as a (prognostic) disease biomarker.

for the first and second steps, respectively. Hemolysis levels are given as percentage of full lysis of erythrocytes in water.

Even though we did not find patients positive for FHaAb in the described cohort, it is important to screen for these autoantibodies, since they may also indirectly increase convertase stability by impairing the convertase regulation by FH (37). FHaAb occur in around 10% of C3G cases and have different characteristics than those occurring in aHUS cases, for example, regarding their target epitope on FH and their pathological consequences (57, 58). Moreover, FHaAb have been previously reported in simultaneous presence of C3NeF, but the combined role of these two autoantibodies in mediating AP activation remains unclear and requires further investigation (58).

Increased stability of convertases cannot only be caused by autoantibodies but may also be due to genetic aberrations. We confirmed that the stabilizing C3NeF-like activity detected in the aHUS patient with known FB mutation (p.Lys323Glu) is due to its aberrant FB activity: increased convertase stability in serum segregated with the FB mutation in both affected and non-affected family members (**Figure 6B**), while the purified Igs from this patient did not support convertase stabilization (**Figure 6C**). This increased convertase half-life associated with the FB mutation is in line with previous functional assays showing that this particular variation causes resistance of the C3 convertase to decay by FH and DAF (41). The fact that a carrier of the mutation shows increased convertase stability in our assay while not being affected with aHUS is supportive for the theory of the incomplete penetrance of mutations in complement genes in aHUS. This theory postulates that multiple (environmental) triggers are required for the actual development of the disease (24). Thus, we show that our assay is also capable as a functional assay for identifying gain-of-function mutations in the building blocks of the AP convertase.

In conclusion, we present optimization of an efficient, robust, and reliable assay for detection and characterization of convertasestabilizing factors (C3NeF and some genetic changes) in patients with complement-mediated renal diseases and for monitoring these patients during treatment. This assay can be used for studying the role of these factors in the pathophysiology of C3G or related disorders and may contribute to the development of more personalized treatment strategies.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the appropriate version of the Declaration of Helsinki. All subjects gave written informed consent, and the protocol was approved by the ethical committee of the Radboud University Medical Center (FH 06979).

# AUTHOR CONTRIBUTIONS

Concept of the study: NK, MO, AB, EV, and LH. Design of the experiments: MM, MO, EV, and LH. Experimental work: MM and SK. Data analysis: MM, NK, EV, and LH. Collection and characterization of clinical samples (including genetic data): NK, EV, and LH. Manuscript writing: MM. All the authors approved the manuscript.

#### ACKNOWLEDGMENTS

This study was performed on behalf of the COMBAT Consortium. This is an interuniversity collaboration in the Netherlands that is formed to study basic mechanisms, assay development, and therapeutic translation of complement-mediated renal diseases. Principal investigators are (in alphabetical order): S. Berger (Department of Internal Medicine-Nephrology, University Medical Center Groningen, Groningen, Netherlands), J. van den Born (Department of Internal Medicine-Nephrology, University Medical Center Groningen, Groningen, Netherlands), P. Gros (Department of Chemistry, Utrecht University, Utrecht, Netherlands), L. van den Heuvel (Department of Pediatric Nephrology, Radboud University Medical Center, Nijmegen, Netherlands), N. van de Kar (Department of Pediatric Nephrology,

#### REFERENCES


Radboud University Medical Center, Nijmegen, Netherlands), C. van Kooten (Department of Internal Medicine-Nephrology, Leiden University Medical Center, Leiden, Netherlands), M. Seelen (Department of Internal Medicine-Nephrology, University Medical Center Groningen, Groningen, Netherlands), A. de Vries (Department of Internal Medicine-Nephrology, Leiden University Medical Center, Leiden, Netherlands). The authors thank all physicians for sharing patient samples and providing the clinical data needed: Dr. A. Bouts and Dr. M. Oosterveld, Academic Medical Center, Amsterdam, Netherlands; Dr. V. Gracchi, University Medical Center Groningen, Groningen, Netherlands; Prof. Dr. E. Levtchenko, University Hospitals Leuven, Leuven, Belgium; Dr. A. de Vries, Leiden University Medical Center, Leiden, Netherlands; Prof. Dr. J. Wetzels, Radboud University Medical Center, Nijmegen, Netherlands; Dr. Y. Konijnenberg, St Jansdal Hospital, Harderwijk, Netherlands; Dr. J. van der Deure, Deventer Hospital, Deventer, Netherlands; Dr. M. Keijzer-Veen and Dr. M. Lilien, Wilhelmina Children's Hospital, Utrecht, Netherlands; Dr. A. van Wijk and Dr. A. Bökenkamp, VU University Medical Center, Amsterdam, Netherlands; Dr. K. Cransberg and Dr. E. Dorresteijn, Erasmus Medical Center, Rotterdam, Netherlands. A special thanks to Dr. Roel Kurvers (Radboud University Medical Center, Nijmegen, Netherlands) for his contribution to the phenotyping of patients.

# FUNDING

This work was supported by grants from the Dutch Kidney Foundation (13OCA27 COMBAT Consortium, 13OI116, KFB 11.007, IP 10.22, 16OKK01), the European Renal Association— European Dialysis and Transplantation Association (grants ERA STF 138-2013, ERA LTF 203-2014), the National Science Centre Poland (grants 2015/18/M/NZ6/00334 and 2014/14/E/ NZ6/00182), and the European Society for Pediatric Nephrology (grant 2014.03).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00612/ full#supplementary-material.


58. Durey MA, Sinha A, Togarsimalemath SK, Bagga A. Anti-complement-factor H-associated glomerulopathies. *Nat Rev Nephrol* (2016) 12(9):563–78. doi:10.1038/nrneph.2016.99

**Conflict of Interest Statement:** NK is a member of the international aHUS Advisory Board of Alexion. The remaining co-authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Michels, van de Kar, Okrój, Blom, van Kraaij, Volokhina and van den Heuvel. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Alternative Pathway Is Essential for Glomerular Complement Activation and Proteinuria in a Mouse Model of Membranous Nephropathy

*Wentian Luo1,2, Florina Olaru1†, Jeffrey H. Miner <sup>3</sup> , Laurence H. Beck Jr <sup>4</sup> , Johan van der Vlag5 , Joshua M. Thurman6 and Dorin-Bogdan Borza2,7\**

*1Division of Nephrology, Department of Medicine, Vanderbilt Medical Center, Nashville, TN, United States, 2Vanderbilt Center for Kidney Disease, Vanderbilt Division of Nephrology, Nashville, TN, United States, 3Renal Division, Washington University School of Medicine, St. Louis, MO, United States, 4Division of Nephrology, Boston University Medical Center, Boston, MA, United States, 5Department of Nephrology, Radboud University Medical Center, Nijmegen, Netherlands, 6Department of Medicine, University of Colorado School of Medicine, Aurora, CO, United States, 7Department of Microbiology, Immunology and Physiology, Meharry Medical College, Nashville, TN, United States*

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Lubka T. Roumenina, INSERM UMRS 1138, Cordeliers Research Center, France Cordula M. Stover, University of Leicester, United Kingdom Jessy J. Alexander, University at Buffalo, United States*

#### *\*Correspondence:*

*Dorin-Bogdan Borza dborza@mmc.edu*

#### *†Present address:*

*Department of Dermatology, Heidelberg University Hospital, Heidelberg, Germany*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 04 May 2018 Accepted: 11 June 2018 Published: 22 June 2018*

#### *Citation:*

*Luo W, Olaru F, Miner JH, Beck LH Jr, van der Vlag J, Thurman JM and Borza D-B (2018) Alternative Pathway Is Essential for Glomerular Complement Activation and Proteinuria in a Mouse Model of Membranous Nephropathy. Front. Immunol. 9:1433. doi: 10.3389/fimmu.2018.01433*

Membranous nephropathy is an immune kidney disease caused by IgG antibodies that form glomerular subepithelial immune complexes. Proteinuria is mediated by complement activation, as a result of podocyte injury by C5b-9, but the role of specific complement pathways is not known. Autoantibodies-mediating primary membranous nephropathy are predominantly of IgG4 subclass, which cannot activate the classical pathway. Histologic evidence from kidney biopsies suggests that the lectin and the alternative pathways may be activated in membranous nephropathy, but the pathogenic relevance of these pathways remains unclear. In this study, we evaluated the role of the alternative pathway in a mouse model of membranous nephropathy. After inducing the formation of subepithelial immune complexes, we found similar glomerular IgG deposition in wild-type mice and in factor B-null mice, which lack a functional alternative pathway. Unlike wild-type mice, mice lacking factor B did not develop albuminuria nor exhibit glomerular deposition of C3c and C5b-9. Albuminuria was also reduced but not completely abolished in C5-deficient mice. Our results provide the first direct evidence that the alternative pathway is necessary for pathogenic complement activation by glomerular subepithelial immune complexes and is, therefore, a key mediator of proteinuria in experimental membranous nephropathy. This knowledge is important for the rational design of new therapies for membranous nephropathy.

Keywords: membranous nephropathy, glomerulonephritis, albuminuria, alternative pathway, membrane attack complex, factor B, complement C5, mouse models

#### INTRODUCTION

Membranous nephropathy (MN), one of the leading causes of nephrotic syndrome in adults, is an antibody-mediated kidney disease clinically characterized by proteinuria, often heavy and persistent. MN is a disease of heterogeneous etiology, defined histologically by immune complexes deposited on the subepithelial side of glomerular basement membrane (GBM), together with GBM thickening

**Abbreviations:** AP, alternative pathway; GBM, glomerular basement membrane; MN, membranous nephropathy; PLA2R, phospholipase A2 receptor; THSD7A, thrombospondin type-1 domain-containing 7A.

and podocyte foot process effacement, but little glomerular inflammation. The most prevalent form is primary MN, now understood as an autoimmune disease in which IgG autoantibodies (predominantly of IgG4 subclass) form subepithelial immune complexes with autoantigens expressed on podocyte cell surface (1, 2). About 70% of patients with primary MN have autoantibodies targeting phospholipase A2 receptor (PLA2R), and an additional 3–5% have autoantibodies targeting thrombospondin type-1 domain-containing 7A (THSD7A) (3, 4). Secondary MN can occur when circulating antibodies bind to antigens planted in the subepithelial space, such as cationic bovine serum albumin of dietary origin (5).

In the current paradigm for the pathogenesis of MN, complement activation is a major effector mechanism of subepithelial immune complexes (1, 6). Complement activation is initiated by three canonical pathways. The classical pathway is triggered by immune complexes and the lectin pathway—by certain danger patterns, while the alternative pathway (AP) is constitutively active and self-amplifies on foreign surfaces. The AP also amplifies activation that is initiated through the other two pathways. All three pathways converge toward the assembly of C3 and C5 convertases, which generate pro-inflammatory anaphylatoxins (C3a, C5a), opsonins that mediate immune adhesion (C3b, iC3b), and the membrane attack complex (C5b-9), which lyses cells. In human MN, C3 fragments and C5b-9 are present alongside IgG in subepithelial deposits, while urinary excretion of sC5b-9 associates with immune disease activity (7–9). Studies in passive Heymann nephritis, a faithful rat model of MN, have specifically implicated C5b-9 as a key mediator of podocyte injury and proteinuria (10, 11). However, the role of different complement pathways upstream of C5 activation in human and experimental MN remains largely unknown.

How immune complexes activate complement in MN remains a conundrum because the autoantibodies implicated in primary MN are predominantly of IgG4 subclass (6, 12–14). Although immune complexes typically bind C1q and activate the classical pathway, IgG4 does not bind C1q and is considered unable to activate complement—at least not *via* the classical pathway (15–17). In kidney biopsies from patients with primary MN, glomerular staining for C1q is almost always weak or absent, while staining for mannan-binding lectin (MBL) and C4d is usually positive (except in patients with MBL deficiency), consistent with the activation of the lectin pathway (18–20). Beck and Salant proposed that the lectin pathway may be activated by IgG4 glycoforms that lack terminal galactose and sialic acid, and, therefore, can bind MBL (1, 6). However, the occurrence of primary MN in patients with MBL deficiency shows that the lectin pathway is not absolutely required (21). Glomerular deposition of properdin and factor B—which is indicative of the AP activation—is also common in MN (18, 21), but the pathogenic relevance of this pathway is not known.

The goal of this study was to determine the contribution of the AP to glomerular injury and proteinuria mediated by subepithelial immune complexes. For this purpose, we used a mouse model that recapitulates clinical and morphologic features of human MN (22–25) and which was found to exhibit proteinuria exacerbated by C5 activation. Using Cfb<sup>−</sup>/<sup>−</sup> mice, which lack factor B (an essential component of the AP), we found that the absence of a functional AP prevented complement activation by subepithelial immune complexes (as assessed from the glomerular deposition of C3c and C5b-9) and abolished proteinuria. These findings provide the first direct evidence implicating the AP activation in the pathogenesis of MN. This knowledge may provide a framework for developing new therapeutic strategies for MN.

#### MATERIALS AND METHODS

#### Materials

The recombinant non-collagenous (NC1) domain of human α3(IV) collagen (rh-α3NC1) was expressed in HEK293 cells and purified as described (26).

#### Animal Experiments

DBA/1J (D1), DBA/2J (D2), and C57Bl/6J (B6) mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). Breeding pairs of Cfb<sup>−</sup>/<sup>−</sup> mice backcrossed on the B6 background for more than nine generations (B6.Cfb<sup>−</sup>/<sup>−</sup>) were obtained from Dr. Joshua Thurman and maintained by homozygous breeding. D1.Cfb<sup>−</sup>/<sup>−</sup> mice were generated by backcrossing onto the D1 background for four generations and then intercrossing F4 heterozygous mice. Mice were housed in a specific pathogen-free facility with free access to food and water. The study was carried out in accordance with the recommendations of the National Institutes of Health Guide for Care and Use of Laboratory Animals and the protocol was approved by the local Institutional Animal Care and Use Committee.

To induce experimental membranous nephropathy, mice (6–10 weeks old, both male and female) were immunized subcutaneously at two sites on the back with rh-α3NC1 antigen (30 µg in 50 µL sterile phosphate-buffered saline) emulsified in an equal volume of Complete Freund's Adjuvant (Sigma, St. Louis, MO, USA). D1 and D2 mice received one booster immunization with the rh-α3NC1 antigen in Incomplete Freund's Adjuvant (Sigma, St. Louis, MO, USA) on day 21 after the first immunization. B6 mice were boosted four times with rh-α3NC1 in Incomplete Freund's Adjuvant, at 10 days interval starting on day 14 after the first immunization. As negative controls, some mice were immunized with adjuvant alone in each experiment. Unless otherwise indicated, mice were euthanized at 8 weeks (for D1 and D2 mice) or 12 weeks (for B6 mice) after the initial immunization, and tissues and blood were collected for further evaluations.

#### Evaluation of Kidney Function

Spontaneously voided spot urine samples were collected by placing mice over 96-well microtiter plates and then analyzed as follows. Urinary albumin was measured by capture ELISA using a mouse albumin quantitation kit (Bethyl, Montgomery, TX, USA). Urine creatinine and blood urea nitrogen levels were measured using Infinity creatinine and urea liquid stable reagents (Thermo Fisher Scientific, Middletown, VA, USA), according to the manufacturer's protocols. To account for urine tonicity, albuminuria was expressed as urinary albumin-tocreatinine ratio (ACR).

# Evaluation of Mouse IgG Autoantibody Production

Sera were assayed for the presence of IgG antibodies to rh-α3NC1 by ELISA. Briefly, 96-well microtiter plates (Nunc MaxiSorp) were coated overnight with rh-α3NC1 (100 ng per well) in carbonatebicarbonate buffer, pH 9.6. After blocking with 1% bovine serum albumin, the wells were incubated for 1 h with mouse sera diluted 1/5,000 for detection of total IgG, 1/2,000 for detection of IgG1, or 1/500 for detection of IgG2a and IgG2b. Secondary antibodies were alkaline phosphatase-conjugated goat anti-mouse IgG (Rockland Immunochemicals, Gilbertsville, PA, USA) and horseradish peroxidase-conjugated goat anti-mouse IgG1, IgG2a, IgG2b, or IgG2c (Bethyl, Montgomery, TX, USA). Plates were developed with p-nitrophenol phosphate or tetramethylbenzidine (Sigma, St. Louis, MO, USA) as substrate, and absorbance was read at 405 nm with a SpectraMax ELISA plate reader (Molecular Devices, Sunnyvale, CA, USA).

# Renal Histopathology and Immunofluorescence Microscopy

For light microscopy, portions of mouse kidneys were fixed in 10% buffered formalin, dehydrated through a graded ethanol series, and embedded in paraffin. Paraffin sections (2 µm thick) were stained with periodic-acid Schiff reagent. Transmission electron microscopy was performed as described (27). For immunofluorescence, portions of mouse kidneys embedded in OCT were snap-frozen in isopentane and stored at −70°C. Cryosections cut at a thickness of 5 µm were fixed in cold acetone for 10 min. Mouse IgG was visualized using FITCconjugated goat anti-mouse IgG (BD Bioscience Pharmingen, San Jose, CA, USA). Complement C3 was visualized using FITCconjugated goat anti-mouse C3c (Nordic Immunology, Tilburg, Netherlands). Kidney deposition of rh-α3NC1 was visualized with rat IgG mAb RH34 (a gift from Dr. Yoshikazu Sado, Shigei Medical Research Institute, Japan), which is specific for human but not mouse α3NC1 (28). Mouse C5b-9, properdin, factor H, and C4 were visualized using rabbit anti-C5b9 (Abcam; Cambridge, MA, USA), rabbit anti-properdin (Santa Cruz, CA, USA), rat IgG anti-mouse C4d mAb (HyCult, Netherlands), and rat IgG anti-mouse factor H mAb (MAB4999, R&D Systems, Minneapolis, MN, USA). Nephrin was visualized using guinea pig anti-nephrin (Progen, Germany). Heparan sulfate chains were visualized with mouse IgM mAb JM403 (29). Agrin was visualized using a rabbit anti-agrin polyclonal antibody (kindly provided by Dr. Takako Sasaki, Oita University, Japan), as previously described (27). Secondary antibodies were Alexa Fluor 488-conjugated goat anti-rabbit IgG, goat anti-rat IgG, donkey anti-guinea pig IgG, goat anti-mouse IgM (Invitrogen, Carlsbad, CA, USA), and FITC-goat anti-rat IgG (BD Bioscience Pharmingen, San Jose, CA, USA). Sections were mounted with anti-fade reagent (Invitrogen, Carlsbad, CA, USA), then coverslipped and observed using a fluorescence microscope (Nikon Eclipse E800). Photomicrographs were captured with a digital camera attached to the microscope, using the same exposure settings for each primary antibody. For quantitative analyses, images were analyzed with Image J software, as described (30). All sections from one experiment were stained and analyzed at the same time.

# *In Vitro* Complement Activation

Fresh frozen normal mouse serum, collected from DBA/1 mice, and stored in aliquots at −70°C, was used as a source of complement. Cryosections of mouse kidneys fixed in cold acetone were incubated overnight at 37°C with normal mouse serum diluted 1:3 in veronal buffered saline (Sigma, St. Louis, MO, USA) containing 0.1% Tween 20 and supplemented with: (a) 2.5 mM calcium chloride and 0.7 mM magnesium chloride; or (b) 2.5 mM magnesium chloride and 6.2 mM EGTA; or (c) 25 mM EDTA. Complement activation was visualized by staining with FITC-conjugated goat anti-mouse C3c, as described above.

#### Statistical Analyses

Analyses were performed using GraphPad Prism 7.00 software (San Diego, CA, USA). ACRs were log transformed. The significance of differences was determined by unpaired *t* test or by oneway analysis of variance (ANOVA) with Dunnett's correction for multiple comparisons. A value of *P* < 0.05 was considered statistically significant. Values are presented as means ± SEM.

# RESULTS

# Complement C5 Deficiency Ameliorates Albuminuria Induced by Subepithelial Immune Complexes in Mice Immunized With **α**3NC1

DBA/1 (D1) mice immunized with α3NC1 develop kidney disease recapitulating clinical and morphologic features of human MN (22–25). This model exhibits proteinuria associated with subepithelial immune complexes and glomerular deposition of IgG, C3, and C5b-9, but the role of complement activation in proteinuria is not known. We reasoned that if C5b-9 is pathogenic in this mouse model (analogous to the rat Heymann nephritis model), then, proteinuria would be ameliorated by the genetic deficiency of complement C5, which occurs naturally in several inbred strains of mice (31). To test this conjecture, we induced experimental MN in C5-deficient DBA/2 (D2) mice, which are nearly 95% genetically identical to C5-sufficient D1 mice (32).

Compared to D1 mice, D2 mice immunized with α3NC1 developed much milder albuminuria (**Figure 1A**). At week 8 after the first immunization, the endpoint for this experiment, the urine ACR in α3NC1-immunized D2 mice (2.4 ± 0.96) was significantly lower (*P* < 0.001) than that in α3NC1-immunized D1 mice (77.1 ± 20.8), albeit greater than in control D2 mice immunized with adjuvant alone (ACR 0.16 ± 0.07) (**Figure 1B**). Blood urea nitrogen levels did not increase over the baseline (mean values at week 0 and at week 8 were 23.1 and 20.3 mg/dL for D1 mice, 27.8 and 25.4 mg/dL for D2 mice), indicating that the renal function did not decline during the duration of the experiment. Serum levels of mouse IgG anti-α3NC1 antibodies were similar in D1 and D2 mice (data not shown). Kidneys were collected at 8 weeks post-immunization for morphologic evaluation. By light microscopy, kidneys from α3NC1-immunized mice

appeared relatively normal, with little glomerular inflammation (**Figure 1C**); however, electron microscopy revealed areas of thickened GBM surrounding subepithelial deposits and podocyte foot process effacement (**Figure 1D**).

Immunofluorescence microscopy showed mouse IgG staining along the GBM in all α3NC1-immunized mice (**Figure 2A**). The IgG staining intensity was similar in D1 and D2 mice (**Figure 2B**). Staining with an antibody that binds to human but not mouse α3NC1 (mAb RH34) showed capillary loop deposition of exogenous antigen in α3NC1-immunized but not control mice

C3c deposition along the capillary loops in α3NC1-immunized D1 and D2 mice. In control mice, C3c staining is positive in the Bowman's capsule and kidney tubules. (E) Indirect immunofluorescence shows capillary loop staining for C5b-9 in α3NC1-immunized D1 mice. A weak background of non-specific immunofluorescence is observed in adjuvant-immunized and α3NC1-

immunized D2 mice. Original magnification 400×.

(**Figure 2C**). GBM staining for C3c, which indicates recent complement activation, was also found in all α3NC1-immunized mice (**Figure 2D**). However, glomerular deposition of C5b-9 was only found in α3NC1-immunized D1 mice (**Figure 2E**), as expected based on the C5 deficiency in D2 mice. A decrease in nephrin staining in α3NC1-immunized D1 mice compared to D2 mice confirmed an injured podocyte phenotype (not shown). These results imply that C5 activation exacerbates proteinuria induced by subepithelial immune complexes in α3NC1-immunized mice, presumably as a result of podocyte injury by C5b-9.

# Evidence of the AP Activation by Subepithelial Immune Complexes

To determine which complement pathways may be activated by subepithelial immune complexes, we assessed glomerular deposition of properdin, factor H, and C4d. Properdin and factor H are positive and negative regulators of the AP, respectively, while C4d is a marker of the classical and lectin pathway activation. Capillary loop staining for both properdin (**Figure 3A**) and factor H (**Figure 3B**) was found in α3NC1-immunized D1 mice but not in control mice. All mice had staining for C4d in a non-specific mesangial pattern (**Figure 3C**), also commonly seen in normal human glomeruli (33). In addition, α3NC1-immunized mice also had segmental staining for C4d along the capillary loops, suggesting that subepithelial immune complexes formed in these mice activate multiple complement pathways, similar to what is observed in human MN (18). While these results are indicative of the AP activation, the pathogenic relevance of the AP cannot be inferred from morphologic findings alone.

#### B6.Cfb**−**/**−** Mice Immunized With **α**3NC1 Develop Subepithelial Immune Complexes but Are Protected Against Albuminuria and Do Not Exhibit Glomerular Complement Deposition

The absence of a functional AP in Cfb<sup>−</sup>/<sup>−</sup> mice afforded a strategy to investigate the role of the AP in experimental MN. We initially obtained Cfb<sup>−</sup>/<sup>−</sup> mice backcrossed on the B6 background for nine generations (B6.Cfb<sup>−</sup>/<sup>−</sup>), which allowed us to compare the course

of albuminuria induced by α3NC1 immunization in B6.Cfb<sup>−</sup>/<sup>−</sup> mice to congenic wild-type B6 mice (B6.Cfb<sup>+</sup>/<sup>+</sup>) (**Figure 4A**). At the endpoint in this experiment (i.e., at 12 weeks after the initial immunization), the urinary ACR in α3NC1-immunized B6.Cfb−/− mice (0.11 ± 0.05) was similar to that in adjuvantimmunized B6 mice (0.057 ± 0.011); both were significantly lower (*P* < 0.0001) than the urine ACR in α3NC1-immunized B6.Cfb<sup>+</sup>/<sup>+</sup> mice (10.6 ± 5.9) (**Figure 4B**). Two mice from each group were observed for an additional 4 weeks; albuminuria further increased in Cfb<sup>+</sup>/<sup>+</sup> mice but not in Cfb<sup>−</sup>/<sup>−</sup> mice. In both groups of α3NC1-immunized B6 mice, kidney morphology mice appeared relatively normal by light microscopy (**Figure 4C**).

Figure 4 | B6.Cfb−/− mice immunized with α3NC1 are protected against albuminuria despite developing subepithelial immune complexes. (A) Time course of urinary albumin-to-creatinine ratio (ACR) in B6.Cfb+/+ mice (circles) and B6.Cfb−/− mice (squares) immunized with α3NC1 (*N* = 7 per group). Control B6 mice (triangles), including both genotypes, were immunized with adjuvant alone (*N* = 7). Shown are means and SEM. (B) Scatterplot depicts the ACR values at the endpoint of this experiment (week 12). The significance of differences among groups was analyzed by one-way ANOVA with Dunnett's correction for multiple comparisons. \*\*\**P* < 0.001, \*\*\*\**P* < 0.0001. n.s., not significant. (C) Morphology of kidneys from α3NC1-immunized B6.Cfb−/− and B6.Cfb+/+ mice appears normal by light microscopy (periodic acid–Schiff staining, original magnification 400×). (D) Transmission electron microscopy shows subepithelial electron dense deposits (arrowhead), expansion of the glomerular basement membrane (G), and podocyte (Po) foot process effacement. Original magnification 7,500×.

Electron microscopy showed features of MN including subepithelial deposits, GBM expansion, and foot process effacement (**Figure 4D**).

Humoral immune responses to α3NC1 were comparable in immunized B6.Cfb<sup>−</sup>/<sup>−</sup> and B6.Cfb<sup>+</sup>/<sup>+</sup> mice, as shown by similar levels of circulating mouse IgG, IgG1, IgG2b, and IgG2c anti-α3NC1 antibodies (**Figure 5**). Immune complexes and complement deposition were evaluated by immunofluorescence microscopy (**Figure 6**). GBM staining for mouse IgG was found in all α3NC1-immunized mice (**Figure 6A**), with similar intensity in B6.Cfb<sup>−</sup>/<sup>−</sup> and B6.Cfb<sup>+</sup>/<sup>+</sup> mice (**Figure 6B**). GBM deposition of exogenous antigen in α3NC1-immunized mice was indicated by positive staining with mAb RH34 (**Figure 6C**). Deposition of C3c along the GBM was found in α3NC1-immunized Cfb<sup>+</sup>/<sup>+</sup> mice but was completely absent in immunized Cfb<sup>−</sup>/<sup>−</sup> mice (**Figure 6D**); as also found for C5b-9 (**Figure 6E**). These results indicate that the absence of factor B does not affect the formation of subepithelial immune complexes but prevents glomerular complement activation and abolishes proteinuria induced by immunization with α3NC1.

# Absence of Proteinuria and Glomerular Complement Deposition in D1.Cfb**−**/**−** Mice Immunized With **α**3NC1 Corroborates the Pathogenic Role of the AP in Experimental MN

We sought to verify whether the ablation of the AP also prevents albuminuria in D1 mice, which are more susceptible to MN. Results in B6 mice cannot be directly compared to those in D1 and D2 mice because B6 mice are more resistant to experimental kidney disease and develop milder proteinuria requiring repeated immunizations with α3NC1. After backcrossing to D1 for four generations, we generated D1.Cfb<sup>−</sup>/<sup>−</sup> mice, which had about 94% D1 genetic background (comparable to almost 95% genetic similarity between D1 and D2 mice).

D1.Cfb−/− mice were resistant to development of albuminuria induced by α3NC1 immunization (**Figure 7A**). At the final time point (at week 8 post-immunization), urine ACR in α3NC1-immunized D1.Cfb<sup>−</sup>/<sup>−</sup> mice (0.36 ± 0.12) was similar to that in adjuvant-immunized D1 mice (0.078 ± 0.023); both were significantly lower (*P* < 0.0001) than urine ACR in α3NC1-immunized D1.Cfb<sup>+</sup>/<sup>+</sup> mice (78.8 ± 18.1) (**Figure 7B**). Morphologic analyses of the kidneys collected at week 8 largely recapitulated the findings from B6 mice. Light microscopy did not show major glomerular abnormalities (**Figure 7C**), while electron microscopy showed subepithelial deposits with GBM thickening and effaced podocyte foot processes (**Figure 7D**). GBM staining for mouse IgG was found in all α3NC1 immunized but not control mice (**Figure 8A**). The IgG

Figure 6 | Analysis of glomerular immune complexes and complement deposition in B6 mice immunized with α3NC1. Kidneys from adjuvantimmunized B6 mice (left), α3NC1-immunized B6.Cfb−/− mice (middle), and α3NC1-immunized B6.Cfb−/− mice (right) were collected at week 12 after the initial immunization. (A) Immunofluorescence staining for mouse IgG shows glomerular basement membrane (GBM) staining in α3NC1-immunized B6.Cfb−/− and B6.Cfb+/+ mice. Adjuvant-immunized mice show non-specific mesangial IgG deposition. (B) Mean fluorescence intensity (MFI) of IgG staining, expressed in arbitrary units (AU), was compared in α3NC1 immunized B6.Cfb−/− and B6.Cfb+/+ mice. Significance was evaluated by *t* test (n.s., not significant). (C) Staining with mAb RH34 shows GBM deposition of exogenous antigen in α3NC1 immunized B6.Cfb−/− and B6. Cfb+/+ mice, which is absent in control mice. (D) Direct immunofluorescence staining reveals C3c deposition in a capillary loop pattern in α3NC1 immunized B6.Cfb+/+ mice. C3c staining is absent in B6.Cfb−/− mice. (E) Indirect immunofluorescence shows C5b-9 deposition along the GBM in α3NC1-immunized B6.Cfb+/+ mice, but not B6.Cfb−/− mice. Original magnification 400×.

among groups was analyzed by one-way ANOVA with Dunnett's correction multiple comparisons. \*\*\**P* < 0.001, \*\*\*\**P* < 0.0001, n.s., not significant. (C) Morphology of kidneys from α3NC1-immunized D1.Cfb−/− (left) and D1. Cfb+/+ mice (right) appeared normal by light microscopy (periodic acid–Schiff staining, original magnification 400×). (D) Transmission electron microscopy shows subepithelial electron dense deposits (arrowhead), thickening of the glomerular basement membrane (G), and podocyte (Po) foot process effacement. Original magnification 15,000×.

staining intensity was similar in D1.Cfb<sup>−</sup>/<sup>−</sup> and D1.Cfb<sup>+</sup>/<sup>+</sup> mice (**Figure 8B**). Staining with mAb RH34 showed GBM deposition of exogenous antigen in these mice (**Figure 8C**). Capillary loop deposition of C3c (**Figure 8D**) and C5b-9 staining (**Figure 8E**) was observed in α3NC1-immunized D1.Cfb<sup>+</sup>/<sup>+</sup> mice but was absent from α3NC1 immunized D1.Cfb<sup>−</sup>/<sup>−</sup> mice. These results confirm that the AP is essential for pathogenic complement activation by subepithelial immune complexes and further show that the ablation of the AP abolishes proteinuria to a

deposition in D1 mice immunized with α3NC1. Kidneys were collected from adjuvant immunized D1 mice (left), α3NC1-immunized D1.Cfb−/− mice (middle), and α3NC1-immunized D1.Cfb−/− mice (right) at week 8 after the initial immunization. (A) Immunofluorescence staining for mouse IgG shows glomerular basement membrane (GBM) staining in α3NC1-immunized D1.Cfb−/− and D1.Cfb+/+ mice. Adjuvant-immunized control mice show non-specific mesangial IgG deposition. (B) Mean fluorescence intensity (MFI) of IgG staining, expressed in arbitrary units (AU), was compared in α3NC1-immunized D1.Cfb−/− and D2.Cfb+/+ mice. The difference was not significant (n.s.) by *t* test. (C) Staining with mAb RH34 shows the GBM deposition of exogenous antigen in both groups of α3NC1-immunized mice, which is absent in control mice. (D) Direct immunofluorescence staining reveals capillary loop deposition of C3c in α3NC1-immunized D1.Cfb+/<sup>+</sup> mice, while C3c staining is absent in D1.Cfb−/− mice. (E) Indirect immunofluorescence shows C5b-9 deposition along the GBM in α3NC1-immunized D1.Cfb+/+ mice but not D1.Cfb−/− mice. Original magnification 400×.

greater extent than C5 deficiency under comparable experimental conditions.

#### Subepithelial Immune Complexes Activate Complement *In Vitro via* the AP

Since α3NC1-immunized Cfb<sup>−</sup>/<sup>−</sup> mice had glomerular IgG but not C3c deposition, we investigated whether immune complexes formed in these mice can activate complement *in vitro*. To this end, kidney cryosections from D1.Cfb<sup>−</sup>/<sup>−</sup> mice were incubated with normal mouse serum as a source of complement. As shown in **Figure 9**, glomerular capillary loops deposition of C3c was found after incubation with normal mouse serum in the presence of Ca2+ and Mg2<sup>+</sup> (in which all three pathways are active), and also in the presence of Mg2<sup>+</sup> and EGTA (conditions under which the

alternative pathway is active, but the classical and lectin pathways are inhibited), but not in the presence EDTA (which inhibits all thee pathways). In control experiments using kidney cryosections from a non-immunized wild-type mouse (i.e., without glomerular IgG deposits), no C3c deposition along the capillary loops was observed under any conditions. Similar results were obtained using human serum as complement source (not shown). These results indicate that glomerular immune complexes formed *in vivo* in α3NC1-immunized Cfb−/− mice have the intrinsic ability to activate complement *via* the AP.

# The Loss of GBM Heparan Sulfate in Experimental MN May Affect the Local AP Regulation

Activation of the AP on extracellular matrices, which are not protected by intrinsic complement regulators, is largely controlled by factor H. Factor H selectively inhibits the AP on host surfaces by recognizing polyanions such as heparan sulfate as markers of self. The normal GBM contains heparan sulfate chains attached to the agrin core protein, but the GBM staining for heparan sulfate epitopes is markedly decreased in human MN and rat Heymann nephritis (34–36). We investigated whether similar changes occur after the induction of experimental MN in mice. Compared to adjuvantimmunized control mice, α3NC1-immunized B6 (**Figure 10**) and D1 mice (not shown) with proteinuria had almost complete

loss of GBM staining by anti-heparan sulfate mAb JM403, while staining for agrin core protein was not changed. These results demonstrate the loss of heparan sulfate epitopes from the GBM of α3NC1-immunized mice, which may locally dysregulate the AP by reducing the recruitment of factor H (37).

# DISCUSSION

Whereas podocyte injury by C5b-9 is a major effector mechanism of the subepithelial immune complexes that lead to proteinuria in MN, the role of specific complement pathways upstream of C5 activation remains poorly understood. Glomerular deposition of complement components specific to the AP in human MN raises the question of whether the AP activation is pathogenically relevant or just an epiphenomenon. We addressed this question by evaluating how the absence of a functional AP affects kidney disease in a mouse model of MN. We found that Cfb<sup>−</sup>/<sup>−</sup> and Cfb<sup>+</sup>/<sup>+</sup> mice immunized with α3NC1 developed subepithelial immune complexes, with similar IgG deposition. However, Cfb<sup>−</sup>/<sup>−</sup> mice were protected against proteinuria and did not exhibit glomerular deposition of C3c and C5b-9. These results imply that the AP is required for glomerular complement activation by subepithelial immune complexes, which in turn is necessary for proteinuria. We thus provide the first direct evidence implicating the AP in the pathogenesis of experimental MN.

Complement activation can contribute to kidney disease at several levels, by causing direct injury at the glomerular filtration barrier, and also by augmenting the production of pathogenic antibodies (38). The latter appears unlikely in our model, as the levels of circulating anti-α3NC1 antibodies and kidney-bound IgG were similar in Cfb<sup>−</sup>/<sup>−</sup> and wild-type mice. Hence, the more likely mechanism by which the absence of factor B protects against proteinuria is *via* the reduced complement activation at the level of C3 and C5, as implied by the absence of glomerular C3c and C5b-9 staining in Cfb<sup>−</sup>/<sup>−</sup> mice despite the presence of subepithelial immune deposits. While immune complexes typically activate complement *via* the classical pathway, the AP is secondarily activated and forms an amplification loop. In a quantitative assay of complement activation initiated *via* the classical pathway (by IgM or aggregated IgG), the AP accounted for more than 80% of the C5 activation products (C5a, sC5b-9) generated in the assay (39). This amplification explains the pathogenic role of the AP in various animal models of immune complex-mediated diseases, including lupus nephritis (40–42). An increased generation of C3b would exacerbate the formation of C5b-9, which is pathogenic in MN, because C5 binding to C3b is a prerequisite for C5 activation, and very high affinity C5 convertases only form at high surface density of C3b (43, 44).

How glomerular immune complexes activate complement depends on several factors, including their ultrastructural localization and immunoglobulin composition. In human anti-GBM disease, IgG1 auto-antibodies that bind to α345(IV) collagen in the human GBM (usually specific for α3NC1) cause severe glomerulonephritis, often featuring C3 deposition along the GBM (45). By contrast, in wild-type mice, human or murine IgG antibodies that bind to α345(IV) collagen in the mouse GBM are insufficient to cause local complement activation or glomerular injury (46, 47). It is, therefore, remarkable that subepithelial immune complexes formed by mouse IgG anti-α3NC1 antibodies with exogenous rh-α3NC1 antigen (*via* a planted antigen mechanism) are able to activate complement *in vivo* and *in vitro*, even though anti-α3NC1 antibodies are predominantly of mouse IgG1 subclass. Mouse IgG1 functionally resembles human IgG4 (the predominant IgG subclass in human primary MN), in that neither binds C1q to activate the classical pathway (48). However, mouse IgG1 can activate the AP (49), as do human IgG2 and IgA (50). Whether human IgG4 can activate the AP remains an important question yet to be solved. In preliminary studies, we found that the AP amplifies *in vitro* complement activation by immune complexes formed by human anti-PLA2R IgG autoantibodies (Dorin-Bogdan Borza and Laurence H. Beck, unpublished observations).

The extent to which the AP is activated is also modulated by the local microenvironment. Although the AP is constitutively active and self-amplifies on foreign surfaces, its activation on self-surfaces is normally restricted by host complement regulatory proteins that inactivate C3 and C5 convertases. In glomeruli, podocytes are protected by membrane-bound regulators such as CR1, while plasma factor H is recruited to inhibit the AP in the GBM (51). Studies in the nephrotoxic nephritis model, in which the AP contributes to chronic but not acute kidney injury, suggest that the mechanisms controlling the AP may be impaired over time by the persistence of antibodies or by glomerular injury (52). One mechanism that could locally dysregulate the AP is the loss of heparan sulfate chains, which factor H normally recognizes as markers of self (37). Decreased GBM staining by anti-heparan sulfate mAb JM403 occurs in α3NC1-immunized mice (this study) and also in human MN (53) and a rat model of MN, active Heymann nephritis (36). The loss of GBM heparan sulfate may be the result of enzymatic cleavage by heparanase, which is often upregulated in glomerular disease (54), including in Heymann nephritis (55). It may also be due to other mechanisms (56). An apparent loss of heparan sulfate occurs in lupus nephritis due to masking by immune complexes (57). Besides factor H, heparan sulfate can also bind properdin, thus serving as a platform for AP activation on some surfaces, such as the apical surface of kidney tubules or apoptotic T cells (58, 59). In our model, given the loss of GBM heparan sulfate, glomerular deposition of properdin is more likely due to properdin binding to C3b and stabilization of the C3bBb convertase (60). Elucidating the relationships among the loss of GBM heparan sulfate, local complement regulation and proteinuria is an area for future investigations.

The role of C5b-9 in proteinuria induced by subepithelial immune complexes has been demonstrated in passive Heymann nephritis (11), but not validated in other models of MN to date. In this study, we found that C5-deficient DBA/2 mice developed significantly less proteinuria than C5-sufficient DBA/1 mice, which further supports the paradigm that C5 activation mediates glomerular injury by subepithelial immune complexes. Cleavage of C5 generates C5b, a precursor for C5b-9, and also C5a, a pro-inflammatory anaphylatoxin. Although our results cannot formally exclude a pathogenic role for C5a in the MN model used in this study, there is currently no evidence to implicate C5a as meditator of glomerular injury in MN, and inflammatory cells are usually not detected in glomeruli in MN.

An unexpected finding in this study was that the absence of factor B abolished proteinuria to a greater extent than the absence of complement C5. Under similar experimental conditions, albuminuria in C5-deficient D2 mice (ACR 2.38 ± 0.96) was reduced by a factor of about 30 compared to C5-sufficient D1 mice (ACR 77.1 ± 20.8), while albuminuria in D1.Cfb<sup>−</sup>/<sup>−</sup> mice (ACR 0.36 ± 0.12) was reduced by a factor of about 200 compared to D1.Cfb<sup>+</sup>/<sup>+</sup> mice (ACR 78.8 ± 18.1)—a sixfold difference. This difference may be explained by the fact the absence of C5 only prevents events downstream of C5 activation (such as C5b-9 formation), while the absence of factor B also inhibits complement activation at the level of C3. Indeed, glomerular C3c deposition was absent in Cfb<sup>−</sup>/<sup>−</sup> mice but present in C5-null mice. These results suggest that complement activation at the level of C3 may also contribute to proteinuria in MN, independently of C5b-9 mediated glomerular injury.

Our study has some limitations. The role of the AP in experimental MN was only investigated using genetic approaches. Further corroboration by pharmacologic inhibition of the AP in a clinically relevant setting is desirable. This is evaluated in ongoing studies. Another potential caveat is that subepithelial immune complexes induced by immunization with α3NC1 are formed by a planted antigen mechanism, recapitulating secondary MN rather than primary MN. Nonetheless, glomerular deposition of factor B occurs not only in primary but also secondary MN (6), suggesting that the AP is involved regardless of how subepithelial immune complexes form. We expect that future studies will address the pathogenic role of the AP in emerging animal models of MN, which target autoantigens implicated in human disease. Of interest in this regard, mice injected with human anti-THSD7A antibodies develop mild proteinuria with histologic features of MN (61). However, efforts to develop models of MN targeting PLA2R have been hampered by the fact that rodents (unlike humans) do not express PLA2R on podocytes.

A better understanding of the complement-mediated mechanisms of injury in MN may help develop novel therapies for MN. MN is a common cause of nephrotic syndrome in adults, and up to 40% of patients eventually develop end-stage renal disease (62). Current therapies use non-specific immunosuppressive drugs, which have significant toxic side effects and are ineffective in about 25–30% of patients (2, 63, 64). Therapeutic inhibition of complement may be a viable approach for treating MN (65), especially in patients who do not respond to conventional therapy or have rapid deterioration of renal function. One anti-complement agent already in clinical use is eculizumab, a humanized IgG2/IgG4 anti-C5 monoclonal antibody that blocks C5 activation. An early, unpublished clinical trial of eculizumab in primary MN did not find a significant remission of proteinuria after 16 weeks of treatment, which may be explained by the insufficient dosage of the drug. Indeed, a recent study has found that C5 inhibition by eculizumab is incomplete at high C3b density (44), a setting relevant to MN. Compared to eculizumab, which does not prevent potentially harmful C3 activation, agents that inhibit complement at the level of both C3 and C5 may offer additional therapeutic benefit. The results of this study identify the AP as a novel target for therapy in MN. Inhibition of the AP has the advantage of leaving the classical and lectin pathways intact for defense against pathogens and other homeostatic functions.

In summary, our results suggest that the AP is necessary for sustained complement activation by subepithelial immune complexes, leading to glomerular deposition of C3c and C5b-9. Furthermore, the activation of the AP is essential for the development of proteinuria in experimental MN. These findings may provide a framework for the rational design of new therapies for MN.

# ETHICS STATEMENT

The study was carried out in accordance with the recommendations of the National Institutes of Health Guide for Care and Use

#### REFERENCES


of Laboratory Animals and the protocol was approved by the local Institutional Animal Care and Use Committee.

# AUTHOR CONTRIBUTIONS

D-BB conceived the idea, designed the experiments, performed data analysis, and wrote the manuscript. WL and FO performed experiments and collected, assembled, and interpreted data. JM performed experiments. JV provided reagents. JT provided mice and reagents and interpreted data. LB analyzed and interpreted data. All authors contributed to editing, reviewed and approved the final manuscript.

#### ACKNOWLEDGMENTS

We thank Linna Ge for excellent technical assistance. Polyclonal anti-agrin antibody was a gift from Dr. Takako Sasaki (Oita University, Japan). Rat mAb RH34 was a generous gift from Dr. Yoshikazu Sado (Shigei Medical Research Institute, Okayama, Japan).

#### FUNDING

This work was supported by Paul Teschan Research Fund grant from the Dialysis Clinic Inc. (DB-B), by Meharry Translational Research Center grant U54 MD007593 from the National Institute on Minority Health and Health Disparities of the National Institutes of Health, by the Norman S. Coplon Extramural Grant Program from Satellite Healthcare (DB-B), and by internal funds from the Meharry Medical College (DB-B). Support was also received from the National Institutes of Health grants DK113586 and DK076690 (JMT) and R01DK078314 (JM). Electron microscopy was performed by the Washington University O'Brien Center for Kidney Disease Research, supported by grant P30 DK079333.


protein and side chains in human glomerular diseases. *Kidney Int* (1993) 43(2):454–63. doi:10.1038/ki.1993.67


**Conflict of Interest Statement:** JT receives royalties from Alexion Pharmaceuticals, Inc. He is a consultant for AdMIRx, Inc., a company developing complement inhibitors. He also holds stock and will receive royalty income from AdMIRx. The other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Luo, Olaru, Miner, Beck, van der Vlag, Thurman and Borza. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Essential Roles for Mannose-Binding Lectin-Associated Serine Protease-1/3 in the Development of Lupus-Like Glomerulonephritis in MRL/*lpr* Mice

*Takeshi Machida1†, Natsumi Sakamoto1†, Yumi Ishida1 , Minoru Takahashi1 , Teizo Fujita2 and Hideharu Sekine1 \**

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Péter Gál, Institute of Enzymology (MTA), Hungary Anne Troldborg, Aarhus University Hospital, Denmark Peter Garred, University of Copenhagen, Denmark*

*\*Correspondence:*

*Hideharu Sekine sekine@fmu.ac.jp*

*† These authors have contributed equally to this work as first authors.*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 02 April 2018 Accepted: 14 May 2018 Published: 28 May 2018*

#### *Citation:*

*Machida T, Sakamoto N, Ishida Y, Takahashi M, Fujita T and Sekine H (2018) Essential Roles for Mannose-Binding Lectin-Associated Serine Protease-1/3 in the Development of Lupus-Like Glomerulonephritis in MRL/lpr Mice. Front. Immunol. 9:1191. doi: 10.3389/fimmu.2018.01191*

*1Department of Immunology, Fukushima Medical University, Fukushima, Japan, 2 Fukushima Prefectural General Hygiene Institute, Fukushima, Japan*

The complement system, composed of the three activation pathways, has both protective and pathogenic roles in the development of systemic lupus erythematosus (or lupus), a prototypic autoimmune disease. The classical pathway contributes to the clearance of immune complexes (ICs) and apoptotic cells, whereas the alternative pathway (AP) exacerbates renal inflammation. The role of the lectin pathway (LP) in lupus has remained largely unknown. Mannose-binding lectin (MBL)-associated serine proteases (MASPs), which are associated with humoral pattern recognition molecules (MBL or ficolins), are the enzymatic constituents of the LP and AP. MASP-1 encoded by the *Masp1* gene significantly contributes to the activation of the LP. After the binding of MBL/ficolins to pathogens or self-altered cells, MASP-1 autoactivates first, then activates MASP-2, and both participate in the formation of the LP C3 convertase C4b2a, whereas, MASP-3, the splice variant of the *Masp1* gene, is required for the activation of the zymogen of factor D (FD), and finally participates in the formation of the AP C3 convertase C3bBb. To investigate the roles of MASP-1 and MASP-3 in lupus, we generated *Masp1* gene knockout lupus-prone MRL/*lpr* mice (*Masp1/3−/−* MRL/*lpr* mice), lacking both MASP-1 and MASP-3, and analyzed their renal disease. As expected, sera from *Masp1/3−/−* MRL/*lpr* mice had no or markedly reduced activation of the LP and AP with zymogen forms of complement FD. Compared to their wild-type littermates, the *Masp1/3−/−* MRL/*lpr* mice had maintained serum C3 levels, little-to-no albuminuria, as well as significantly reduced glomerular C3 deposition levels and glomerular pathological score. On the other hand, there were no significant differences in the levels of serum anti-dsDNA antibody, circulating ICs, glomerular IgG and MBL/ficolins deposition, renal interstitial pathological score, urea nitrogen, and mortality between the wild-type and *Masp1/3−/−* MRL/*lpr* mice. Our data indicate that MASP-1/3 plays essential roles in the development of lupus-like glomerulonephritis in MRL/*lpr* mice, most likely *via* activation of the LP and/or AP.

Keywords: systemic lupus erythematosus, lupus nephritis, complement, MASP-1/3, lectin pathway, alternative pathway, murine models

# INTRODUCTION

The complement system, which consists of over 30 soluble and membrane-bound proteins, plays protective roles in host defense and a role in some immune regulatory functions *via* activation of the three different initial complement pathways: the classical pathway (CP), lectin pathway (LP), and alternative pathway (AP) (1). Each pathway follows a sequence of reactions to generate a C3 convertase (C4b2a in the CP and LP or C3bBb in the AP), and subsequently a C5 convertase (C4b2a3b or C3bBb3b). The terminal sequence of complement activation involves C5b, C6, C7, C8, and C9, which interact sequentially to form the membrane attack complex.

Activation of the CP is initiated by the binding of a C1 complex (C1q, C1r, and C1s), in which C1q recognizes IgM or IgG of antigen (Ag)–antibody complexes, followed by the activation of C1r and C1s, subsequently C4 and C2, resulting in the creation of C3 convertase C4b2a (2). However, activation of the LP is initiated by the binding of the LP pattern recognition molecules (PRMs), such as mannose-binding lectin (MBL), ficolins (-1, -2, -3 or M-, L-, H-, respectively), collectin (CL)-10, and CL-11 (3). The function of MBL or ficolins in opsonophagocytosis and in the complement pathway is similar to that of C1q. MBLassociated serine proteases-1 and -2 (MASP-1 and MASP-2), the enzymatic constituents of the LP, form a complex with the LP PRMs. After binding of the LP PRMs to carbohydrates typically found on the surface of microorganisms, MASP-1 autoactivates first, and subsequently activates MASP-2 (4). Activated MASP-2 cleaves both C4 and C2, resulting in the creation of C3 convertase C4b2a (5). However, activated MASP-1 cleaves MASP-2 and C2 but not C4 (6). Unlike the CP and LP, initiation of the AP does not require recognition molecules, and is thought to occur by a process termed "tickover," the spontaneous thioester hydrolysis of C3 (7). The product, C3(H2O), interacts with factor B (FB), and the subsequent cleavage of FB by the serine protease factor D (FD). This results in the creation of C3 convertase C3(H2O) Bb, which cleaves C3 generating metastable C3b. The thioester bond in metastable C3b mediates covalent attachment of C3b to the surface of self (i.e., host) or non-self (i.e., microorganisms) cell membrane. C3b bound to the host cells is subject to process inactivation by multiple complement-regulatory proteins, present in plasma and on host cell membranes. By contrast, C3b bound to microorganisms is subject to process a chain reaction-like amplification loop that can bind large numbers of C3b molecules on the cell surface after the initial C3b binding. Notably, uncontrolled activation of the AP is associated with multiple inflammatory diseases, such as systemic lupus erythematosus (SLE or lupus).

Our group previously provided a fundamental link between the LP and AP. To investigate the role for MASP-1 for complement activation, we generated C57BL/6 mice deficient for MASP-1 by targeting of the *Masp1* gene that transcribes two serine proteases, MASP-1 and MASP-3, and MAp44, which lacks a serine protease domain. In addition, MAp44 has been suggested to act as a competitive inhibitor of LP activation. Unexpectedly, our previous studies reported that *Masp1/3<sup>−</sup>/<sup>−</sup>* mice had little-to-no activation of both the LP and AP with an inactive form of FD (pro-FD) in their sera, indicating that MASP-1 and/or MASP-3 play essential roles in LP and AP activation (4, 8). We also reported that recombinant MASP-3 cleaved pro-FD in mouse serum (9), suggesting a role for MASP-3 in AP activation. Furthermore, a recent *in vitro* study using selective inhibitors against MASPs demonstrated that the monospecific MASP-1 inhibitor, SGMI-1, completely inhibited LP activation in human resting blood, and it was more potent than the monospecific MASP-2 inhibitor SGMI-2 (10). Another recent publication using SGMI-1 showed that MASP-1 is essential for LPS-induced but not for zymosan-induced AP activation under a Ca2<sup>+</sup>-chelating condition, suggesting the roles for MASP-1 in AP activation in addition to LP activation (11). On the other hand, the monospecific MASP-3 inhibitor TFMI-3 did not inhibit any of the three complement pathways in normal human serum, but completely blocked pro-FD activation in normal human plasma (12).

Systemic lupus erythematosus is a prototypic human systemic autoimmune disease involving aberrant complement activation that is initiated by immune complexes (ICs) formed by autoantibodies (auto-Abs) directed against a broad range of self Ags, including dsDNA and nuclear proteins. The kidney is a major site of IC formation and/or deposition, and lupus nephritis is a major cause of mortality in both human SLE and murine models of lupus. The roles of each complement pathway in the development of lupus nephritis have been investigated using serum and biopsy samples from lupus patients or murine models of lupus. CP activation is assumed to play a role in initial pathogenic complement activation in lupus nephritis since both lupus patients and lupus-prone mice exhibit significant glomerular IC deposition consisting of auto-Ag–auto-Ab complexes (e.g., dsDNA-antidsDNA Ab complexes) and C1q, a recognition molecule for the CP. Paradoxically, patients with homozygous deficiency of the CP components, including C1 (C1q, C1r, or C1s), C4, or C2, have a high risk of lupus or lupus-like disease (13, 14). Previous studies on murine models of lupus showed that mice deficient for C1q or C4 exhibited high titers of serum antinuclear Ab, thus increasing the incidence and prevalence of glomerular disease associated with multiple apoptotic bodies or lupus-like glomerulonephritis, as well as increased mortality (15–17). These results clearly indicate that the CP plays both pathogenic and protective roles against the development of lupus, including IC-mediated glomerulonephritis.

On the other hand, there is strong evidence that AP activation plays an exacerbating role in the development of lupus glomerulonephritis, either through direct initiation of the pathway or the magnifying effects of the amplification loop. Deficiency of the AP components in lupus-prone MRL/*lpr* mice, such as FB (18) and FD (19), exhibited significantly decreased glomerular C3 deposition levels, maintained serum C3 levels, and improved glomerular pathological score. Furthermore, Sekine et al. reported the benefit of the selective inhibition of the AP for renal disease in lupus-prone MRL/*lpr* and NZM2410 mice by therapeutic administration of a targeted and selective inhibitor of the AP CR2-fH, compared to CR2-Crry, which inhibits all complement pathways (20, 21).

In contrast to the extensive evidence on the CP and AP, little is known in regard to the involvement of the LP in lupus. Previous human genetic studies in lupus have shown that there are five polymorphic sites in MBL, a recognition molecule for the LP, and are associated with serum levels of MBL and the development of lupus (22–24). Villareal et al. reported that Spanish patients carrying a genetic variant in codon 54 of MBL, one of the previously reported five variants, had a high risk of developing lupus (25). Furthermore, Seelen et al. (26) supported the data described above by investigating the LP activation and several serum auto-Ab levels in lupus patients with all five MBL variants. They indicated that patients carrying MBL variant alleles exhibited impaired serum activation levels of the LP, increased serum anti-cardiolipin and anti-C1q antibodies, and similar levels of serum MBL when compared to patients without those variant alleles. They concluded that the mutant MBL may have low-binding activity to apoptotic cells, and result in defective clearance of auto-Ags, suggesting a protective role for MBL or the LP in auto-Ab production in lupus. Recently, Sato et al. reported that lupus patients with MBL/L-ficolin and properdin deposition in their glomeruli had significantly higher urinary protein excretion levels compared to patients without glomerular deposition, suggesting pathogenic and exacerbating roles for the LP and AP in the development of lupus glomerulonephritis (27). Collectively, whether the role of the LP in lupus pathogenesis is protective or exacerbating remains unclear. Also, direct evidence on the role of the LP in SLE has yet to be demonstrated. In this context, an association between the LP and SLE that involves complement activation is assumed and should be elucidated.

Since total blockade of all complement pathways is apparently not appropriate as a therapeutic strategy for SLE, an individual role for the three different complement pathways should be clarified. In the present study, we generated lupus-prone MRL/*lpr* mice that were genetically deficient for MASP-1/3 by using a backcrossing strategy and confirmed that their sera lacked both LP and AP complement activity. We then analyzed their lupuslike disease serologically and renal pathologically.

#### MATERIALS AND METHODS

#### Mice

MRL/MpJ-*Faslpr*/J (MRL/*lpr*; stock no. 000485) was purchased from The Jackson Laboratory. *Masp1/3<sup>−</sup>/<sup>−</sup>* C57BL/6 mice (4) were backcrossed for seven generations with MRL/*lpr* mice to generate an eighth backcross generation of *Masp1/3<sup>+</sup>/<sup>−</sup>* MRL/*lpr* mice. For efficient and precise backcrossing, each generation of backcrossed offspring was checked for 12 microsatellite markers corresponding to disease-susceptibility regions (*D4Mit12*, *D4Mit17*, *D5Mit13*, *D5Mit24*, *D5Mit145*, *D7Mit39*, *D7Mit57*, *D7Mit211*, *D10Mit11*, *D10Mit20*, *D17Mit16*, and *TNF*) (28), and the *Fas* genotype by PCR with genomic DNA was used as a template. The eighth generation was bred to yield *Masp1/3<sup>+</sup>/<sup>+</sup>*, *Masp1/3+/−*, and *Masp1/3−/−* MRL/*lpr* mice. All animal experiments that included housing, breeding, and using the mice were reviewed and approved by the Animal Experiments Committee of Fukushima Medical University (approval no. 26001 and 28012), and were performed in accordance with the guidelines for the care and use of laboratory animals established by the Committee.

### Assays for C4 Deposition Onto Mannan-Coated Microtiter Plates

The LP activity in sera was determined by C4 deposition assay onto a mannan-coated microplate according to a method by Takahashi et al. (4). A Nunc 96-well optical bottom plate was coated with 100 µL of 10 µg/mL mannan (Sigma-Aldrich) diluted in 50 mM Na-carbonate/bicarbonate buffer (pH 9.5) by overnight incubation at 4°C. The wells were blocked with 1.0% BSA in Tris-buffered saline (TBS) containing 0.1% Tween-20 and 5 mM CaCl2 (TBST/Ca) for 1 h at room temperature (RT). After washing three times with TBST/Ca, serially diluted serum samples in TBS/Ca buffer were added to each well and incubated for 1 h at RT. After washing three times with TBST/Ca, diluted purified human C4 (5 µg/mL in TBS/Ca) was added to each well, and incubated for 30 min at 4°C. After washing three times with TBST/Ca, horseradish peroxidase (HRP)-conjugated antihuman C4 Ab (MP Biomedicals) was added to each well. After incubation for 1 h at RT, a TMB Microwell Peroxidase Substrate 2-Component System (Kirkegaard & Perry Laboratories) was added to each well and incubated for 10 min at RT in the dark for color development. An equal volume of 1 M phosphoric acid was added to the substrate solution to stop the color development of TMB, and then the absorbance at 450 nm was measured by a spectrophotometer DTX880 (Beckman Coulter). The LP activity in the TBS/Ca buffer was also measured as a blank experiment, and the activity in sera was expressed as the difference between A450 in the serum samples and that in the blank experiment.

#### Assays for C3 Deposition Onto Zymosan-Coated Microtiter Plates

The AP activity in sera was determined by zymosan assay according to a method by Takahashi et al. (8). A Nunc 96-well optical bottom plate (Nunc) was coated with 100 µL of 20 µg/mL zymosan (Sigma-Aldrich) suspended in 50 mM Na-carbonate/bicarbonate buffer (pH 9.5) by overnight incubation at 4°C. The wells were blocked with 1.0% BSA in phosphate-buffered saline containing 0.1% Tween-20 (PBST) for 1 h at RT. After washing three times with PBST, serially diluted serum samples in BBS buffer (0.2 M boric acid, 0.14 M NaCl, pH 8.0) supplemented with Mg2<sup>+</sup>-EGTA were added to each well and incubated for 1 h at RT. The wells were then washed with PBST three times, and HRP-conjugated anti-mouse C3 polyclonal antibodies was added to each well and incubated for 1 h at RT, followed by the color development of TMB. The AP activity in BBS buffer was also measured as a blank experiment, and the activity in sera was expressed as the difference between A450 in the serum samples and that in the blank experiment.

#### Immunoprecipitation and Western Blotting for FD

Mouse sera appropriately diluted in TBS were incubated for 1 h at 4°C with 2.5 µg of affinity-purified rabbit anti-FD IgG and 10 µL of protein A-agarose (GE Healthcare). The beads were washed four times with TBS, and then denatured at 100°C for 5 min in 40 µL of TBS containing 0.1% SDS and 50 mM β-mercaptoethanol. Denatured samples were mixed with 2 µL of 10% Triton X-100 and 0.2 µL of *N*-glycosidase F (Merck Millipore) and incubated for 2 h at 37°C. After centrifugation, supernatants were separated by SDS-PAGE and analyzed by western blotting. FD and pro-FD were detected with peroxidase-conjugated affinity-purified rabbit anti-FD raised in rabbit (8). Images were visualized by chemiluminescence with an ECL Prime Western Blotting Detection Reagent (GE Healthcare) according to the manufacturer's instruction.

### Determination of Serum Anti-Double-Stranded DNA IgG Levels

As previously described by Gilkeson et al. (29), Nunc 96-well optical bottom plate was coated with 100 µL of 1 µg/mL S1 nuclease-digested calf thymus DNA diluted in 1 × SSC (0.3 M sodium citrate, 0.03 M NaCl, pH 7.0) by overnight incubation at 37°C. Wells were blocked with 1% BSA in PBST (BSA-PBST) for 1 h at RT. After washing three times with PBST, 1/100-diluted serum samples were added to each well, and incubated for 1 h at RT. After washing three times with PBST, bound anti-dsDNA IgG was detected with HRP-conjugated goat anti-mouse IgG (γ-chain specific; Sigma) followed by the color development of TMB. Serum anti-dsDNA IgG levels were expressed as the difference between A450 in the serum samples and that in the blank experiment.

#### Determination of Serum Circulating IC Levels

Serum circulating IC levels were determined by ELISA according to a method by Stanilova and Slalov (30) with some modifications. A Nunc 96-well optical bottom plate was coated with 100 µL of 1 µg/mL goat anti-mouse C3 Ab (MP biomedicals) diluted in 50 mM Na-carbonate/bicarbonate buffer (pH 9.5) by overnight incubation at 4°C. The wells were blocked with 1% BSA in 50 mM Tris–HCl (pH 7.2) supplemented with 0.05% Tween-20 (Tris–HCl/tw) for 20 min at RT. After washing with Tris–HCl/ tw, serially diluted serum samples were added to the wells, and incubated for 1 h at RT. After washing three times with Tris–HCl/ tw, bound circulating IgG-IC was detected with HRP-conjugated goat anti-mouse IgG Ab (γ-chain specific; Sigma) followed by the color development of TMB. Serum circulating IgG-IC levels were expressed as the difference between A450 in the serum samples and that in the blank experiment.

# Determination of Serum C3 Levels

A Nunc 96-well optical bottom plate was coated with 100 µL of 1 µg/mL goat anti-mouse C3 polyclonal Ab (MP biomedicals) diluted in 50 mM Na-carbonate/bicarbonate buffer (pH 9.5). The wells were blocked with BSA-PBST for 20 min at RT. After washing with PBST, serially diluted serum samples were added to the wells and incubated for 1 h at RT. After washing three times with PBST, bound C3 was detected with HRP-conjugated antimouse C3 polyclonal Ab (MP biomedicals) followed by the color development of TMB. The mouse complement C3 calibrator (Kamiya biomedical company) was used to determine the serum C3 concentration.

# Evaluation of Albuminuria

Urine samples were collected every 2 weeks beginning at 12 weeks of age by placing the mice in metabolic cages for 24 h. Urine collection was performed with ampicillin, chloramphenicol, and gentamycin in the bottom of a collection tube to prevent bacterial growth. A Nunc 96-well optical bottom plate was coated with 100 µL of 2.5 µg/mL rabbit anti-mouse albumin polyclonal antibody (MP biomedicals) in 50 mM Na-carbonate/bicarbonate buffer (pH 9.5). Wells were blocked with BSA-PBST for 1 h at RT. After washing with PBST, serially diluted urine samples in BSA-PBST were added to the wells, and incubated for 1 h at RT. After washing three times with PBST, bound urinary albumin was detected with HRP-conjugated anti-mouse albumin polyclonal antibody in BSA-PBST followed by the color development of TMB. Mouse albumin was used as a standard. Urinary albumin excretion was expressed as milligrams of albumin per mouse per day.

#### Assessment of Renal Pathology

*Masp1/3<sup>+</sup>/<sup>+</sup>* and *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice were sacrificed at 24 weeks of age for pathological evaluation. At the time of sacrifice, kidneys were recovered and divided into two sections. One section was placed in buffered formalin for subsequent embedding in paraffin, sliced, and stained with H&E. The H&Estained slides were assessed *via* light microscopy for glomerular (glomerular inflammation, proliferation, thickness basement membrane, epithelial reactivity, crescent formation, and necrosis) and interstitial pathologies. Pathological scores from 0 to 4+ (0, none; 1+, mild; 2+, moderate; 3+, moderate to severe; 4+, severe) were assigned for each of these features and then added together to yield a final score, and an overall glomerular score was derived.

#### Immunofluorescence Staining

The other kidney section from *Masp1/3*<sup>+</sup>/<sup>+</sup> and *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice was embedded in an OCT compound and frozen in liquid nitrogen for cryosections. The frozen sections were cut into 5-µm slices, fixed with acetone, and stained with FITCconjugated antibodies: goat anti-mouse IgG, C1q, MBL-A, MBL-C, Ficolin-A, Ficolin-B, and C3 diluted 1:100. The severity of glomerulonephritis and IC deposition was determined in a blind manner. Scores ranged from 0 to 4+, where 0 corresponded to a non-autoimmune healthy mouse and 4+ to the maximal alteration observed in the study.

# Determination of Serum Urea Nitrogen (UN) Levels

Serum UN was determined using a UN Colorimetric Detection Kit (Arbor Assays) according to the manufacturer's instructions. Serum from BALB/c mice was subjected to the assay as an indicator for a normal level of serum UN.

# Statistical Analysis

Statistical analysis was performed using a GraphPad Prism 6 software for Mac OS X (GraphPad Software, San Diego, CA, USA). Single groups were compared using an unpaired two-tailed *t* test.

#### RESULTS

#### Ability of LP and AP Activation in Sera From MRL/*lpr* Mice

Our group had previously demonstrated that mice deficient for MASP-1/3 had little-to-no activation of both the LP and AP with an inactive form of FD in their sera at the C57BL/6 background (8). In the current study, we first tested if lupus-prone MRL/*lpr* mice also had altered activation of the LP and AP in MASP-1/3-deficient serum. To assess the activity of the LP and AP in MRL/*lpr* mice, we performed the C4 deposition assay with mannan-coated plates and C3 deposition assay with zymosancoated plates. As expected, sera from *Masp1/3<sup>+</sup>/<sup>+</sup>* wild-type MRL/*lpr* mice had C4 deposition activity on the mannan-coated plates and C3 deposition activity on the zymosan-coated plates in a dose-dependent manner (**Figure 1**). In contrast, both C4 and C3 deposition from the sera of *Masp1/3−/−* MRL/*lpr* mice were significantly lower than those in the wild-type MRL/*lpr* mice. These results indicate that MASP-1 and/or MASP-3 is involved in the activation of the LP and AP in MRL/*lpr* mice as in C57BL/6 mice.

It has been shown that circulating FD in *Masp1/3<sup>−</sup>/<sup>−</sup>* C57BL/6 mice is a zymogen (inactive form of FD or pro-FD) that can be distinguished from the active form of FD by the difference in molecular weight, as activated FD lacks N-terminal peptide QPRGR, which pro-FD possesses (8). As shown in **Figure 2**, the deglycosylated FD in the *Masp1/3<sup>−</sup>/<sup>−</sup>* mice both at the MRL/*lpr* and C57BL/6 background was slightly larger than that in the wild-type mice. These results demonstrate that lupus-prone MRL/*lpr* mice deficient for MASP-1/3 had little-to-no activation of both the LP and AP with an inactive form of FD in sera.

### Serum Levels of IgG Anti-dsDNA Auto-Abs, Circulating ICs and C3 in MRL/*lpr* Mice

Production of IgG anti-dsDNA auto-Ab is strongly associated with lupus-like renal disease in MRL/*lpr* mice. To assess the effect of MASP-1/3 deficiency in anti-dsDNA Ab production, we measured the serum levels of IgG anti-dsDNA Ab in MRL/*lpr* mice by ELISA starting from 12 to 24 weeks of age. As shown in **Figure 3A**, there was a progressive rise in serum anti-dsDNA Ab levels both in the wild-type and *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice after 12 weeks of age. There was a trend toward lower serum anti-dsDNA Ab levels in the *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice compared to the wild-type MRL/*lpr* mice; however, the difference did not reach statistical significance at any point in time until 24 weeks of age. These results indicate that there was minimal or no impact of MASP-1/3 deficiency on serum auto-Ab levels in MRL/*lpr* mice.

We next assessed serum levels of circulating IgG-ICs in MRL/*lpr* mice by anti-C3 anti-mouse IgG sandwich ELISA starting from 12 weeks of age. As shown in **Figure 3B**, there was no statistically significant difference between the groups at any point in time until 24 weeks of age.

Serum C3 levels in patients with SLE is known to show an inverse correlation with disease activity due to its consumption following activation of the complement cascade. We measured serum C3 levels in the MRL/*lpr* mice starting from 12 weeks of age. Those levels in the wild-type MRL/*lpr* mice decreased as the mice aged and showed an inverse correlation with serum IgG anti-dsDNA Ab levels (**Figure 3C**). In contrast, the serum C3 levels in the *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice were maintained at constant levels, and were significantly higher than those in the wild-type MRL/*lpr* mice at 16 weeks of age, and continued to be higher until the time of sacrifice (Week 24). These results show that there was a significant effect of MASP-1/3 deficiency on the reduction of serum C3 consumption in MRL/*lpr* mice.

#### Albuminuria in MRL/*lpr* Mice

Urinary albumin or protein excretion in SLE reflects glomerular damage to the charge/size barrier between the capillary lumen and urinary space, and, therefore, through the glomerular capillary endothelial cells, glomerular basement membrane, and podocytes in glomeruli. To evaluate the effect of MASP-1/3 deficiency on renal function, we measured 24 h of urinary albumin excretion levels of MRL/*lpr* mice, starting at 12 weeks of age. As shown in **Figure 4**, the wild-type MRL/*lpr* mice developed a high level of albuminuria after 18 weeks of age. In contrast, the MRL/*lpr* mice deficient for MASP-1/3 had significantly less albuminuria remaining at less than 0.1 mg/mouse/day during the tested period compared to wild-type littermates. These results suggest a significant role for MASP-1/3 in the development of glomerular disease in MRL/*lpr* mice.

#### Glomerular Deposition of ICs and Complement in MRL/*lpr* Mice

To determine the mechanistic effect of the absence of MASP-1/3 on the reduction of albuminuria observed in MRL/*lpr* mice, mice were sacrificed at 24 weeks of age, and their kidneys were recovered for pathological analysis. To assess glomerular ICs (IgG), C1q, MBL-A, MBL-C, Ficolin-A, Ficolin-B, and C3 deposition, frozen kidney sections were stained with fluorescein-conjugated Abs against mouse IgG, C1q, MBL-A, MBL-C, Ficolin-A, Ficolin-B, or C3. There was no significant difference in glomerular IgG, C1q, MBL-A, or MBL-C deposition levels between wild-type and *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice, while glomerular deposition levels of Ficolin-A and Ficolin-B were low or undetectable in these mice (**Figure 5**; **Table 1**). It was indicated that wild-type and *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice had similar deposition levels of the CP and LP recognition molecules in their glomeruli. Importantly, glomerular C3 deposition was readily evident in wild-type MRL/*lpr* mice, while *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice had significantly reduced levels of glomerular C3 deposition compared to their wild-type littermates. These results suggest that acceleration of the LP and/or AP activation, which is mediated by MASP-1/3, rather than CP activation, plays an important role in C3 consumption in murine lupus.

#### Renal Pathology in MRL/*lpr* Mice

Kidney sections were stained with H&E and assessed by histological scoring for overall glomerular proliferative changes, crescent formation and necrosis, and interstitial inflammation. As expected, the wild-type MRL/*lpr* mice exhibited diffuse glomerulonephritis, including cellular proliferation, inflammation, glomerular expansion, fibrocellular crescents, and interstitial inflammation (**Figure 6A**). The *Masp1/3−/−* MRL/*lpr* mice, however, had significantly less pathological features of glomerular

Figure 4 | Urinary albumin excretion levels in 12- to 24-week-old MRL*/lpr* mice. Values are means ± SD (*n* = 6–8). Values with asterisks are significantly different at \**p* < 0.05 against values from age-matched wild-type MRL/*lpr* mice.

disease compared to their wild-type littermates, with a reduction in mesangial expansion, glomerular inflammation, focal hypercellularity, and crescent formation (reflected in the renal score, **Figure 6A**, *p* < 0.05). On the other hand, both strains exhibited renal interstitial inflammation with no significant difference in disease score between the two groups (**Figure 6B**). These results indicate that MASP-1/3 is intimately associated with progression in glomerulonephritis, but not with renal interstitial inflammation in MRL/*lpr* mice.

#### Serum UN Levels in *Masp1/3−/−* MRL/*lpr* Mice

To evaluate renal function, serum UN levels of *Masp1/3<sup>+</sup>/<sup>+</sup>* and *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice were measured using serum obtained at 24 weeks of age. As shown in **Figure 7**, serum UN levels of *Masp1/3<sup>+</sup>/<sup>+</sup>* MRL/*lpr* and *Masp1/3<sup>−</sup>/<sup>−</sup>* mice were significantly higher than those of non-autoimmune BALB/c mice, while there was no statistically significant difference of the serum UN levels between *Masp1/3<sup>+</sup>/<sup>+</sup>* and *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice.

#### Mortality of MRL/*lpr* Mice

Despite having significantly less albuminuria and pathological features of glomerular disease in the *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice, the absence of MASP-1/3 had no significant beneficial effect on survival in the MRL/*lpr* mice until the time of sacrifice (week 24) (**Figure 8**).

#### DISCUSSION

To determine the role of MASP-1/3 in lupus nephritis, we backcrossed MASP-1/3-deficient C57BL/6 mice into lupus-prone MRL/*lpr* mice for eight generations, and then intercrossed the *Masp1/3<sup>+</sup>/<sup>−</sup>* mice. The results presented in this report indicate that the absence of MASP-1/3 in MRL/*lpr* mice has a significant effect on the activation of the complement and development of lupus-like glomerulonephritis. Consistent with previous results in *Masp1/3<sup>−</sup>/<sup>−</sup>* C57BL/6 mice (4, 8), the MASP-1/3-deficient MRL/*lpr* mice of the current study had little-to-no activation of the LP and AP, and significantly reduced glomerular C3 deposition, albuminuria and pathologic renal scores compared with their wild-type MASP-1/3-producing littermates.

To date, mice deficient for different complement components have been generated, and the roles for each complement component in the development of lupus or lupus-like disease have been analyzed. In lupus nephritis, activation of the complement system is thought to be triggered via the CP. This is because the presence of autoantibodies is a requirement for the development of lupus nephritis (31), and deposition of complement proteins including the CP components C1q, C4, and C3 in the glomeruli are key features of lupus nephritis. However, as summarized in **Table 2**, mice deficient for C1q or C4 showed high antinuclear antibody titers, anti-DNA autoantibody levels, and glomerulonephritis with impairment in the clearance of apoptotic cells (15–17, 32). Although the severity of serum autoantibody levels or glomerular disease is genetic background-dependent, C1q and C4, the complement components of the CP, provide an important protective role against the development of lupus or lupus-like glomerulonephritis in mice. This observation is consistent with patients with homozygous genetic deficiencies of an early component of the CP (i.e., C1q, C1r, C1s, C4A/C4B, and C2), which are strongly associated with the risk of developing SLE or a lupus-like

here. Original magnification 400×.

Table 1 | Scores for glomerular deposition of IgG and complement in 24-week-old MRL/*lpr* mice.


*All glomerular immunofluorescence images were interpreted by three independent persons in a blinded manner, and graded from 0 to 4*+ *(0, none; 1*+*, mild staining; 2*+*, moderate staining; 3*+*, moderate-high staining; 4*+*, high staining) for fluorescence intensity. All values are means* ± *SD (n* = *5). Values with asterisks were significantly different at \*p* < *0.05 against values from age-matched WT mice. n.d., not detected.*

Figure 6 | Assessment of glomerular (A) and renal interstitial (B) pathologies stained with hematoxylin and eosin. Glomerular pathology was graded from the sum of scores for glomerular inflammation, thickness of basement membrane, epithelial cell reactivity, crescent formation, and necrosis. Interstitial pathology was graded using the sum of scores for perivascular inflammation and inflammatory cell infiltration. Scores were graded as 0 to 4+ (0, none; 1+, mild; 2+, moderate; 3+, moderate-high; 4+, high). Original magnification 200× for glomerular pathology, and 40× for interstitial pathology.

disease (33). As for C3, the converging point for activation of all three complement pathways, *C3<sup>−</sup>/<sup>−</sup>* mice in (129 × C57BL/6)*lpr* background exhibited no difference in serum autoantibody levels, glomerular IgG deposition levels or glomerular pathological

scores compared to *C3<sup>+</sup>/<sup>+</sup>* (129 × C57BL/6)*lpr* mice (16). Similarly, C3-deficient lupus-prone MRL/*lpr* mice exhibited no difference in serum autoantibody levels, glomerular pathological scores, or

mice were sacrificed (24 weeks).


*letters indicate that the corresponding complement component defects are pathogenic or protective against the development of murine lupus-like disease, respectively. CP, classical pathway; AP, alternative pathway; LP, lectin pathway; MASP-1/3, mannose-binding lectin-associated serine protease-1/3; ANA, anti-nuclear antibody.*

survival, but had significantly increased levels of glomerular IgG deposition and albuminuria (34). These results suggest that C3 plays a beneficial role in lupus-like glomerular disease *via* clearance of ICs. Indeed, there is an association of inherited human C3 deficiency with IC-related disorders, including membranoproliferative glomerulonephritis, SLE, and vasculitis (35).

In contrast to the complement component of the CP and C3, no association of the deficiencies of the complement factors of the AP with the risk of developing lupus in humans has been reported, suggesting at least that there are no protective roles for the AP complement factors against the development of lupus. To address the question whether the AP complement factors play protective or exacerbating roles in the development of lupus, MRL/*lpr* mice deficient for the AP complement factors, FD or FB, were generated, and their lupus-like disease was analyzed. As summarized in **Table 2**, both the FB- and FD-deficient MRL/*lpr* mice exhibited reduced glomerular C3 deposition levels, maintained serum C3 levels, and improved glomerular pathological scores compared to the wild-type MRL/*lpr* mice (18, 19). Interestingly, the MRL/*lpr* mice exhibited somewhat different disease phenotypes between *FB<sup>−</sup>/<sup>−</sup>* and *FD<sup>−</sup>/<sup>−</sup>* genetic backgrounds. Similar to the *FB<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice, the *FD<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice exhibited maintained serum C3 levels, significantly reduced glomerular C3 deposition levels and improved pathological renal scores. However, the *FB<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice exhibited some additional beneficial effects on reduced levels of serum anti-DNA antibody and glomerular IgG deposition, proteinuria, as well as reduced IgG3 cryoglobulin production and renal vasculitis (18). We hypothesize that these additional improvements observed in the *FB<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice are unlikely due to the absence of FB itself but likely due to MHC-linked effects. MRL/*lpr* mice deficient for the *FB* gene, which was located on the MHC class III region, were carrying the H2b/b MHC haplotype, whereas wild-type or *FB<sup>+</sup>/<sup>+</sup>* MRL/*lpr* mice were carrying the H2k/k MHC haplotype (18). These differences perhaps affect Ag presentation leading to differences in pathogenic autoantibody production or immunoglobulin isotype switching. In addition, the MHC-linked effects on MRL/*lpr* mice were tested by generating H2b/b *FB<sup>+</sup>/<sup>+</sup>* congenic MRL/*lpr* mice (36, 37). Those studies showed that, similarly to H2b/b*FB<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice, 70% of H2b/b *FB<sup>+</sup>/<sup>+</sup>* MRL/*lpr* mice spontaneously developed serum IgG3 deficiency with reduced anti-dsDNA autoantibody levels and exhibited significantly reduced albuminuria compared to wild-type (H2k/k *FB<sup>+</sup>/<sup>+</sup>*) MRL/*lpr* mice. Production of IgG3 in MRL/*lpr* mice is one of the major factors responsible for the development of glomerulonephritis in such mice (38). Therefore, H2b/b haplotype-linked IgG3 deficiency accounts for at least part of the reduced proteinuria observed in H2b/b *FB<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice.

In the present study, the *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice, lacking the LP and AP, had significantly reduced glomerular C3 deposition, pathological glomerular disease and albuminuria compared to their wild-type littermates. Similarly, the *FD<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice had significantly reduced glomerular C3 deposition and improved renal pathology but had no protective effect on albuminuria. Thus, it is possible that the reduced albuminuria observed in the *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice was in part dependent on the inability of LP activation. Involvement of the LP in the development of glomerulonephritis in human lupus is largely unknown. Previously reported genetic analyses of lupus patients showed that gene polymorphism of MBL, the recognition molecule of the LP, is linked to lupus susceptibility (22–24). A reduced functional activity of the LP, in relation to the expression of MBL variant alleles, is associated with increased levels of serum autoantibodies (26). These observations suggest a beneficial role of MBL in the clearance of apoptotic material that is somewhat similar to the role of the CP recognition molecule C1q. Meanwhile, previously reported pathoclinical analyses of lupus patients demonstrated deposition of the LP recognition molecules in their glomeruli; MBL in 82%, L-Ficolin in 63.6% (39). In that report, patients with glomerular MBL deposition had a higher mean of proteinuria than patients without glomerular MBL deposition, suggesting involvement of the LP in the development of proteinuria. Another study reported that lupus patients with glomerular MBL/L-ficolin and properdin deposition, which is deposits of the LP and AP components, had significantly higher levels of proteinuria than patients without these glomerular depositions (27). Consistent with our results, that report showed the significance of glomerular activation of the LP and AP in lupus nephritis.

In addition to MBL and ficolins, collectin-10 (CL-10 or CL-L1) and collectin-11 (CL-11 or CL-K1) are known to be MASP-1/3-associated collectins (40, 41). Recently, Wu et al. reported a pathogenic role for collectin-11 in the development of tubulointerstitial fibrosis in murine models of renal ischemia-reperfusion injury (42). However, the contribution of collectin-10 and/or collectin-11 to the development of lupus glomerulonephritis is largely unknown. In the present study, we did not evaluate collectin-10 or collectin-11 deposits in glomeruli or tubulointerstitial regions of MRL/*lpr* mice. Further studies are needed to clarify their contribution to the development of renal injury in lupus.

The association of LP serine proteases MASPs (MASP-1, MASP-2, and MASP-3) and MBL-associated proteins MAps (MAp19 and MAp44) with lupus characteristics has also previously been assessed (43). MAp19 and MAp44, the splicing variants of the *Masp2* and *Masp1* genes, respectively, are thought to have a regulatory function in LP activation (44, 45). Interestingly, plasma concentrations in lupus patients were higher than the healthy controls regarding MASP-1, MASP-3, and MAp44, but not MASP-2. These results suggested that the association of MASP-1 and/or MASP-3 with lupus characteristics goes beyond the role of complement activation. Indeed, unlike MASP-2 or MASP-3, MASP-1 has many substrates including non-complement proteins, and the roles of MASP-1 other than in complement activation have been reported. For example, MASP-1 directly digests fibrinogen and factor XIII, playing a role in coagulation (46, 47). Besides the complement and the coagulation system, MASP-1 plays a role in the kinin generation system, where recombinant MASP-1 or natural MBL-MASPs are able to cleave high-molecular-weight kininogen and liberate bradykinin (48). Although the efficiency of MASP-1-mediated cleavage in the kinin generating system is low compared to that of the plasma kallikrein-mediated cleavage, it could be important when the LP activates locally, such as in the glomeruli in lupus patients. Local activation of the LP in glomeruli could lead to local bradykinin production that could contribute to the development of proteinuria. Furthermore, a significant association of MASP-1 in cellular immunity has been reported. MASP-1 can cleave protease activated receptors on the surface of endothelial cells, which results in the pro-inflammatory activation of endothelial cells, followed by local infiltration of neutrophils and endothelial cell damage (49). These tissue damages could also occur on glomerular endothelial cells in lupus patients with glomerular MBL-MASP-1 deposits. Taken together, the absence of MASP-1 in addition to the absence of MASP-3 is highly beneficial to protect lupus patients from the development of glomerulonephritis.

The absence of MASP-1/3 showed significantly reduced glomerulonephritis and albuminuria in the MRL/*lpr* mice of the present study; however, it did not improve their serum BUN levels or survival rate. One possible reason is that, similar to the wild-type MRL/*lpr* mice, the *Masp1/3−/−* MRL/*lpr* mice developed tubulointerstitial nephritis. Moreover, elevated serum BUN levels reflect tubulointerstitial damage and progression of renal failure as well as decreased glomerular filtration rates (50, 51). This observation suggests that the mechanism(s) underlying glomerular disease, which is significantly associated with MASP-1/3, are distinct from that of in tubulointerstitial nephritis in this model, and the absence of MASP-1/3 alone may not be enough to improve overall renal function in MRL/*lpr* mice. Another possible reason is that MASP-1/3 plays an important role in ontogenesis. Deficiency of MASP-1/3 or its complex partner CL-K1 leads to the development of 3MC syndrome (52), which is characterized by unusual facial features and problems affecting other tissues and organs of the body. Indeed, the total body weight of the *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice in the current study was significantly lower than that of the age-matched wild-type MRL/*lpr* mice (Figure S1 in Supplementary Material). Therefore, it is possible that *Masp1/3<sup>−</sup>/<sup>−</sup>* MRL/*lpr* mice, which lack MASP-1/3 congenitally, have lowered resistance against stress-related autoimmune response. The direct effect of MASP-1/3 regarding complement activation on lupus may be elucidated in further studies using anti-MASP-1/3 agents, such as a specific inhibitor and anti-MASP-1/3 Ab, by suppressing MASP-1/3 therapeutically.

In summary, our experiments revealed that MASP-1/3 is required for the development of glomerulonephritis in MRL/*lpr* mice. The major role of MASP-1/3 for glomerular disease in this model is supposed to be in the activation of the AP that is initially activated *via* the CP. In addition, glomerular MBL deposition observed in MRL/*lpr* mice suggests that the activation of the LP, in which MASP-1 plays a significant role, was also involved in their glomerular disease. Moreover, the absence of MASP-1, which also directly plays roles in coagulation, the kinin generation system and endothelial cells, may contribute an additional beneficial effect to protect from glomerular damage. Therefore, the absence of both MASP-1 and MASP-3 is protective against the development of glomerulonephritis in MRL/*lpr* mice. Although further experiments including analysis in different murine lupus strains are required, both proteases are potential therapeutic targets for glomerulonephritis in murine lupus models and possibly in human lupus patients.

#### ETHICS STATEMENT

All animal experiments that included housing, breeding, and using the mice were reviewed and approved by the Animal Experiments Committee of Fukushima Medical University (approval no. 26001 and 28012), and were performed in accordance with the guidelines for the care and use of laboratory animals established by the Committee.

### AUTHOR CONTRIBUTIONS

TM and NS are equally credited as first authors in this work: TM, NS, and HS designed the study and wrote the manuscript: TM and NS performed the experiments, analyzed the results, and

#### REFERENCES


made figures and tables: YI assisted with the experiments: MT, TF, and HS supervised the study.

#### FUNDING

This research was supported by a Grant-in-Aid for Scientific Research (no. 23790542 and 26670478) from the Japan Society for the Promotion of Science.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01191/ full#supplementary-material.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Machida, Sakamoto, Ishida, Takahashi, Fujita and Sekine. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Both Monoclonal and Polyclonal Immunoglobulin Contingents Mediate Complement Activation in Monoclonal Gammopathy Associated-C3 Glomerulopathy

#### Edited by:

Maciej Cedzynski, Institute for Medical Biology (PAN), Poland

#### Reviewed by:

Edimara S. Reis, University of Pennsylvania, United States John D. Imig, Medical College of Wisconsin, United States Giuseppe Remuzzi, Istituto Di Ricerche Farmacologiche Mario Negri, Italy

#### \*Correspondence:

Sophie Chauvet sophiechauvet@ymail.com

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 29 June 2018 Accepted: 11 September 2018 Published: 02 October 2018

#### Citation:

Chauvet S, Roumenina LT, Aucouturier P, Marinozzi M-C, Dragon-Durey M-A, Karras A, Delmas Y, Le Quintrec M, Guerrot D, Jourde-Chiche N, Ribes D, Ronco P, Bridoux F and Fremeaux-Bacchi V (2018) Both Monoclonal and Polyclonal Immunoglobulin Contingents Mediate Complement Activation in Monoclonal Gammopathy Associated-C3 Glomerulopathy. Front. Immunol. 9:2260. doi: 10.3389/fimmu.2018.02260 Sophie Chauvet 1,2,3 \*, Lubka T. Roumenina2,3,4, Pierre Aucouturier 5,6 , Maria-Chiara Marinozzi 2,5, Marie-Agnès Dragon-Durey 2,3,7, Alexandre Karras <sup>1</sup> , Yahsou Delmas <sup>8</sup> , Moglie Le Quintrec<sup>9</sup> , Dominique Guerrot <sup>10</sup>, Noémie Jourde-Chiche<sup>11</sup> , David Ribes <sup>12</sup>, Pierre Ronco4,13,14, Frank Bridoux 15,16 and Véronique Fremeaux-Bacchi 2,3,5

<sup>1</sup> Assistance Publique-Hôpitaux de Paris, Hôpital Européen Georges Pompidou, Department of Nephrology, Paris, France, 2 INSERM UMRS1138, Centre de Recherche des Cordeliers, Team "Complément et Maladies", Paris, France, <sup>3</sup> Université Paris Descartes Sorbonne Paris-Cité, Paris, France, <sup>4</sup> Sorbonne Université, Paris, France, <sup>5</sup> Assistance Publique-Hôpitaux de Paris, Hôpital Saint Antoine, Department of Immunology, Paris, France, <sup>6</sup> INSERM UMRS 938, Sorbonne Universités, UPMC Univ Paris 06, Hôpital Saint-Antoine, Paris, France, <sup>7</sup> Assistance Publique-Hôpitaux de Paris, Hôpital Européen Georges Pompidou, Department of Immunology, Paris, France, <sup>8</sup> Department of Nephrology, Centre Hospitalier Universitaire de Bordeaux, Bordeaux, France, <sup>9</sup> Department of Nephrology, Hôpital de Foch, Suresnes, France, <sup>10</sup> Department of Nephrology, Centre Hospitalier Universitaire de Rouen, Rouen, France, <sup>11</sup> Aix-Marseille Univ, UMRS 1076 Vascular Research Center of Marseille, Department of Nephrology, AP-HM, Marseille, France, <sup>12</sup> Department of Nephrology, Centre Hospitalier Universitaire de Toulouse, Toulouse, France, <sup>13</sup> Assistance Publique-Hôpitaux de Paris, Hôpital Tenon, Department of Nephrology, Paris, France, <sup>14</sup> INSERM UMRS1155, Hôpital Tenon, Paris, France, <sup>15</sup> Department of Nephrology, INSERM CIC 1402, Centre Hospitalier Universitaire de Poitiers, Poitiers, France, <sup>16</sup> Centre National de Référence Maladies Rares: Amylose al et Autres Maladies à Dépôts d'Immunoglobulines Monoclonales, Université de Poitiers, Poitiers, France

C3 glomerulopathy (C3G) results from acquired or genetic abnormalities in the complement alternative pathway (AP). C3G with monoclonal immunoglobulin (MIg-C3G) was recently included in the spectrum of "monoclonal gammopathy of renal significance." However, mechanisms of complement dysregulation in MIg-C3G are not described and the pathogenic effect of the monoclonal immunoglobulin is not understood. The purpose of this study was to investigate the mechanisms of complement dysregulation in a cohort of 41 patients with MIg-C3G. Low C3 level and elevated sC5b-9, both biomarkers of C3 and C5 convertase activation, were present in 44 and 78% of patients, respectively. Rare pathogenic variants were identified in 2/28 (7%) tested patients suggesting that the disease is acquired in a large majority of patients. Anti-complement auto-antibodies were found in 20/41 (49%) patients, including anti-FH (17%), anti-CR1 (27%), anti-FI (5%) auto-antibodies, and C3 Nephritic Factor (7%) and were polyclonal in 77% of patients. Using cofactor assay, the regulation of the AP was altered in presence of purified IgG from 3/9 and 4/7 patients with anti-FH or anti-CR1 antibodies respectively. By using fluid and solid phase AP activation, we showed that total purified IgG of 22/34 (65%) MIg-C3G patients were able to enhance C3 convertase activity. In five documented cases, we showed that the C3 convertase enhancement was mostly due to the monoclonal immunoglobulin, thus paving the way for a new mechanism of complement dysregulation in C3G. All together the results highlight the contribution of both polyclonal and monoclonal Ig in MIg-C3G. They provide direct insights to treatment approaches and opened up a potential way to a personalized therapeutic strategy based on chemotherapy adapted to the B cell clone or immunosuppressive therapy.

Keywords: complement, alternative pathway activation, C3 glomerulopathies, monoclonal gammopathy, autoantibodies

#### INTRODUCTION

C3 glomerulopathy (C3G) is a heterogeneous group of rare glomerular diseases, characterized by predominant C3 deposition in glomeruli (1–3) and resulting from dysregulation of the complement alternative pathway (AP) (4–6). In physiological conditions, the complement AP is continuously activated at a low level and is amplified on activating surfaces, such as bacteria or dying cells (7). To avoid undesirable auto-amplification, the AP is tightly regulated in the fluid-phase and on cell surfaces by the plasma regulatory proteins factor H (FH), factor I (FI), membrane cofactor protein (MCP, CD46), complement receptor 1 (CR1, CD35), and decay accelerating factor (DAF, CD55) (8). Together, these regulators act by preventing the formation of and by dissociating the AP C3 convertase (FH, CR1, and DAF) and by serving as cofactors for FI-mediated inactivation of C3b to iC3b (FH, MCP, and CR1). Properdin is the only positive regulator of the AP, stabilizing the AP C3/C5 convertase (8, 9). Many of these factors are involved in complement dysregulation in C3G. Rare pathogenic variants in AP genes are identified in ∼25% of C3G patients (5, 6, 10). In most cases, complement dysregulation is acquired, induced by the presence of Nephritic Factors (C3NeF and C5NeF), i.e., autoantibodies targeting the AP C3/C5 convertase (6, 10) or anti-FH antibodies (11, 12). Recently, C3G has been proposed to be included in the spectrum of monoclonal gammopathy of renal significance (MGRS) because of the high prevalence of monoclonal immunoglobulins (MIg) in C3G patients aged over 50, without criteria for multiple myeloma, that reached 30–71% in two small series, and 65% in the French C3G cohorts (13– 16). Although this association and the favorable effect of clonetargeted therapy on renal outcomes (16) suggests a role of MIg in the occurrence of the renal disease, the exact pathophysiological link between MIg and AP dysregulation remains to be elucidated.

The aim of the current work was to determine the mechanism of acquired complement AP dysregulation in patients with MIg-C3G in order to clarify the causal relationship between the MIg and the occurrence of C3G.

#### METHODS

#### Study Population

Between 2000 and June 2014, 201 plasma samples from patients aged over 18 were received at the Laboratory of Immunology (European Hospital Georges Pompidou) for complement exploration in the context of C3G. The diagnosis of C3G was assessed by immunofluorescence according to consensus recommendations, with bright diffuse predominant C3 glomerular staining (≥2+), of at least two orders of magnitude greater than any other immune reactant (i.e., Ig). Patients with trace or weak amounts of IgM staining on glomerular sclerotic lesions were included, but those with weak staining for IgG, IgA or Ig light chains were excluded (2). The diagnosis of DDD was confirmed by demonstration of diffuse, highly electron-dense osmiophilic deposits within the lamina densa by EM. By contrast, the diagnosis of C3GN was established in patients showing deposits of lesser density without the characteristic distribution and "sausage shape" appearance of DDD (1). All patients with positive hepatitis B or C serology, antinuclear antigen autoantibodies, anti- double-stranded DNA antibodies, or cryoglobulinemia are excluded from the French C3G registry.

Search for monoclonal gammopathy was performed by immunofixation in all patients aged over 40. Of the 201 adult patients in the French registry of C3G, 60 patients (G1-G60) had a detectable MIg. Among them, 50 were included in a retrospective clinical study regarding the effect of chemotherapy on renal outcomes (16). In the current study, 41/60 MIg-C3G patients with available blood samples were included (**Supplemental Figure 1**). Two of 28 patients screened for genetic abnormalities carried a rare variant of undetermined significance (p.Asp130Asn in CFH and p.Glu548Gln in CFI), as previously described (16). As 96% of MIg-C3G patients displayed a C3GN pattern on kidney biopsy, 107 adult patients with C3GN without MIg extracted from the French cohort of C3G, and 8 patients with MIg without kidney disease, were used as control population. The local ethics committee approved the study and the study was approved by the Commission Nationale de L'informatique et des Libertés (CCP number 192 12 23) and all legal representative of children gave written informed consent for genetic analysis.

#### Assays for Complement Component and for C3 and C5 Nephritic Factors

EDTA plasma samples were obtained from all patients. Plasma protein concentrations of C3, C4 were measured by nephelometry (Dade Behring, Deerfield, IL, USA). Soluble C5b-9 level determination was done using the MicroVue sC5b-9 Plus EIA Assay (Quidel, San Diego, CA), according to manufacturer instructions. Normal values were established from plasma samples from 100 healthy donors. C3NeF and C5NeF activities were determined by assessing the ability of purified plasma IgG to stabilize the membrane-bound C3bBb and C3bBbP convertases (6).

# ELISA Detection for anti-FH, anti-FI, anti-CR1, anti-C3b and anti-FB Antibodies

ELISA plates were coated with 10 to 15µg/ml of FH (11), FI, FB, C3b (17) (all from Complement Technologies, Tylor, Texas), CR1 (RandD System) in PBS for 1 h, followed by blocking of the plates with PBS-0.4% Tween 20. Plasma was diluted 1/200 in PBS-0.1% Tween 20 and applied for 1 h. Bound IgG or IgA was revealed by anti-human IgG antibody conjugated with HRP (Southern Biotech) or anti-human IgA antibody conjugated with HRP (Sigma) diluted in PBS-0.1% Tween 20, followed by TMB substrate system.

# Study of IgG Binding to CR1 by Surface Plasmon Resonance (SPR)

The interaction of patient IgG with CR1 was analyzed in real time using a ProteOn XPR36 SPR equipment (BioRad, Marne-la-coquette, France). CR1 (RandD System) was covalently immobilized to a GLC sensor chip (BioRad) following the manufacturer's procedure. Protein G purified IgG from the patients or healthy donors (at 100µg/ml) were injected for 300 s in PBS 0.005% Tween 20 containing running buffer. The dissociation was followed for 300 s. The signal from the interspots, reflecting the background binding was subtracted, as recommended by the manufacturer.

#### Study of C3b Interaction With CR1 in Presence of Patients IgG by Surface Plasmon Resonance

IgG from patients with anti-CR1 antibodies were tested for their capacity to alter the C3b binding to CR1 using SPR. CR1 was coupled to individual flow channels of GLC biosensor chip using standard amine-coupling, according to the manufacturer's instruction. Total purified IgG were flowed at a concentration 100 mg/ml followed by injection of C3b (Complement Technologies) at concentrations starting from 1µg/ml. Five concentrations and a running buffer were injected at 30 µl/min in HEPES buffer (10 mM Hepes, 25 mM NaCl, Tween 0.005%, pH 7.4) for 300 s across the immobilized ligand. Data were analyzed using ProteOn Manager software and the data from the blank channel were subtracted. Kinetic parameters were calculated by fitting the obtained sensorgrams into a two-state interaction model.

#### Determination of Light Chain and Heavy Chain Isotype Specificity of Anti-complement Protein Antibodies

The light chain (LC) isotype of antibodies was determined by ELISA. After plasma incubation and washing, isotype-specific goat antibodies directed against kappa and lambda LC (Southern Biotech), diluted in PBS-0.1% Tween 20 were incubated 1 h. Bound Ig was revealed by a Rabbit anti goat IgG Ab (Santa Cruz) diluted in PBS-0.1% Tween 20, followed by TMB substrate system. The ratios of the optical densities obtained with the antik and anti-l Abs (k/l) were calculated for all samples. A k/l ratio < 0.1 or >3 indicated the predominance of anti-complement autoantibody of lambda or kappa LC specificity respectively, as previously described. A ratio between 0.1 and 3 indicated both kappa and lambda reactivity (11). The heavy chain (HC) IgG subtypes of anti FH and anti CR1 IgG Ab were determined by an anti-FH or anti-CR1 ELISA. After plasma incubation and washing, isotype-specific mouse antibodies directed against IgG1, IgG2, IgG3, and IgG4 (NL16 for IgG1, GOM2 for IgG2, ZG4 for IgG3, and RJ4 or IgG4) (Unipath, Bedford, UK), diluted PBS-0.1% Tween 20 were incubated 1 h. Bound IgG was revealed by a rabbit anti mouse IgG Ab (Jackson ImmunoResearch) diluted in PBS-0.1% Tween 20, followed by TMB substrate system.

### Determination of LC and HC Isotype Specificity of Monoclonal Immunoglobulin

The analysis of serum MIg of 29/41 patients was performed by a western blotting. Serum dilutions were adjusted to normalized gamma globulin levels. Proteins were separated by highresolution thin layer agarose electrophoresis and transferred on nitrocellulose sheets. After saturation with skimmed milk, the blots were probed with polyclonal antibodies specific for a, g, m, k or l Ig chains or with monoclonal antibodies specific for IgG subclasses with NL16 for IgG1, GOM2 for IgG2, ZG4 for IgG3, and RJ4 or IgG4 (Unipath, Bedford, UK), followed by peroxydase coupled rabbit anti mouse IgG antibodies (Jackson ImmunoResearch). The signal was developed by chemo luminescence using ECL kit (Perkin Elmer) and MyECL Imager (Thermo Scientific).

# IgG Purification

#### Total IgG Purification IgG were purified from plasma of MIg-C3G patients or from plasma of control patients (healthy donors, patients with positive

C3NeF and patients with MIg but without kidney disease) by using Protein G beads (GE Healthcare), as recommended by the manufacturer. The concentration of the IgG was determined by a Nanodrop spectrophotometer.

#### Purification of Monoclonal And Polyclonal Igs by Chromatography

Monoclonal and polyclonal Ig fractions of 5 patients (with monoclonal IgG) were purified using ion exchange column chromatography. Each plasma sample was dialyzed against 10 mM Tris (pH8). Prepaked diethyl-aminoethyl (DEAE) trisacryl column (Life Science) was equilibrated with 10 mM Tris (pH8). The dialyzed samples were loaded onto the column followed by elution with a 0–0.2 M NaCl gradient in 10 mM Tris buffer (pH8). Serial 1 ml fractions were collected and assayed for protein concentration (280 nm OD). The fractions were tested by agarose electrophoresis and immunofixation to determine which fractions contained polyclonal or MIg.

# Cofactor Assays

C3 protein (20µg/ml; Calbiochem) was incubated at 37◦C for 0, 1, 5, or 10 min with FI (10µg/ml; Complement Technologies) and FH (20µg/ml; Complement Technologies, Tylor, Texas), or soluble CR1 (10µg/ml; RandD Systems) in 10 mM Tris, 150 mM NaCl, pH 7.4 in presence of 100µg/ml of total purified IgG. Samples were boiled and the cleavage of the C3 was probed by a Western blot, using SNAP system (Millipore). After blocking with Tris 10 mM, NaCl 150 mM, 0.1% Tween, 1% BSA, the blots were probed with a 1:5,000 dilution of goat anti-human C3 IgG (Calbiochem) followed by HRP-conjugated rabbit anti-goat IgG (Santa Cruz). The signal was developed by chemiluminescence using ECL kit (Perkin Elmer) and MyECL Imager (Thermo Scientific). Cleavage efficiency was evaluated

TABLE 1 | Comparison of immunological findings in 41 MIg-C3G patients and 107 C3GN adults patients without MIg.


\* C4 level was normal in all patients

\*\*99 on 107 C3GN patients without monoclonal gammopathy were screened for genetics abnormalities of complement proteins. Results are described in Servais et al. (4) and Marinozzi et al. (6).

by the appearance of the α43 band and the disappearance of the α-chain at 10 min and quantitated by densitometry of the scanned images. The ratio between α43 and the β bands (representing the % of C3b cleaved) was plotted vs. the time of incubation.

### C3 Convertase Formation in Normal Human Serum in Presence of Patients' IgG

Purified total IgG from patients or healthy donors were incubated for 30 min at 37◦C with normal human serum diluted 1:3 in presence of EGTA-Mg to block the classical pathway (10 mM MgCl2,10 mM EGTA, 40 mM NaCl Hepes buffer). The generation of C3a was quantified by the Micro Vue C3a Kit (Quidel) according to the manufacturer's instructions. IgG from 8 patients with MIg without kidney disease were used as controls.

# Fluid Phase C3 Convertase Activation in Presence of Patients' IgG

Total purified IgG (100µg/ml) were incubated for 45 min at 37◦C with C3 (25µg/ml), FB (0 to 50 ng), FD (0.05µg/ml) (all from Complement Technologies, Tylor, Texas) in Hepes, 40 mM NaCl supplemented with 10 mM MgCl2. The reaction was stopped by adding DTT-containing sample buffer. The cleavage of C3 was probed by a Western blot, using SNAP system (Millipore). After blocking with Tris 10 mM, NaCl 150 mM, 0.1% Tween, 1% BSA, blots were probed with a 1:5,000 dilution of goat anti-human C3 IgG (Calbiochem) followed by HRP-conjugated rabbit anti-goat IgG (Santa Cruz). The signal was developed by chemiluminescence using ECL kit (Perkin Elmer) and MyECL Imager (Thermo Scientific). Percentage of C3 cleavage revealing convertase formation was characterized by the appearance of α' band and quantitated by densitometry of the scanned images. The ratio between α' and the β bands was calculated at 50 ng of FB. The same experiment was reproduced with monoclonal and polyclonal Ig fractions of 5 patients and with IgG of 3

(C3NeF assay). Samples from 41 MIg-C3G patients and 38 healthy individuals were tested. Results of C3NeF and other antibodies are expressed as percentage of residual stabilization, and in arbitrary units (UA), respectively. For anti-FH, anti-C3b and anti-FB antibodies, we used positive controls as previously described (one patient positive for anti-FH auto-antibodies in the setting of atypical HUS and one patient positive for both anti-C3b and anti-FB auto-antibodies) (11, 17). For the other ELISA assays, results were considered as positive when the OD was upper the mean +2SD (of the OD obtained with IgG from healthy donors). The patient's sample with the higher OD value was then used to determine the UA.

MIg-C3GN patients after chemotherapy adapted to the B cell clone. IgG from 8 patients with MIg without kidney disease were used as controls. The same experiment was reproduced with monoclonal and polyclonal fractions of 5 MIg-C3G patients IgG.

#### C3 Convertase Activation on Immobilized Patient IgG

Coating of ELISA plate was performed at 20µg/ml of purified IgG in PBS for 1 h followed by a blocking of the plates by PBS-0.4% Tween 20. After washing, C3 convertase was formed

by adding C3 (25µg/ml), FB (0–50 ng), FD (0.05µg/ml) (all from Complement Technologies, Tylor, Texas) diluted in Hepes, 40 mM NaCl supplemented with 10 mM MgCl2. The cleavage of C3 was probed by a Western blot and quantified as described above.

#### Statistical Analyses

Data are expressed as median (with range) for continuous variables and percentage for categorical variables. Statistical analyses were performed using the Mann-Whitney and Kruskal-Wallis tests, as appropriate, for comparison of continuous variables. Chi-square or Fisher's exact tests were used for comparison of categorical variables. P-values below 0.05 were considered significant. Results were analyzed using the Graph Pad Prism software.

# RESULTS

# MIg-C3G Is Associated With Biomarkers of C3/C5 Convertase Activation

Forty-one patients from the French registry of C3G met inclusion criteria (**Supplemental Figure 1**). Baseline clinical data and complement biomarkers are detailed in **Supplemental Table 1** and **Table 1**. At diagnosis, 18/41 (44%) MIg-C3G patients had a low C3 level and a normal C4 (**Table 1**). Median C3 level of MIg-C3G patients and C3GN patients without MIg were similar (p = 0.86) (**Figure 1A**). Soluble C5b-9 was increased in 29/37 (78%) MIg-C3G patients and in 47/76 (62%) C3GN patients without MIg (p = 0.09) (**Table 1**). Median sC5b-9 level was significantly higher in MIg-C3G patients compared to patients without MIg (p = 0.005) (**Figure 1B**).

#### Detection of Anti-complement Protein Auto-antibodies in MIg-C3G

Samples were screened for C3 NeF/C5 NeF and auto-antibodies targeting 5 proteins of the AP (**Figures 2A-F**). Anti-FH autoantibodies, C3NeF and anti-FI auto-antibodies were detected in 17% (9/41), 7% (3/41), and 5% (2/41) of MIg-C3G patients, respectively. None had anti-C3b, anti-FB antibodies or C5NeF. Eleven patients were positive for anti-CR1 auto-antibodies (11/41, 27%) (**Figure 2E** and **Supplemental Figure 2A**). The characteristics of the binding of anti-CR1 positive IgG to CR1 by Surface Plasmon Resonance (SPR) are provided (**Supplemental Figures 2B-C**). Overall, anti-complement TABLE 2 | Heavy and light chain characterization of anti-complement protein Ab and monoclonal immunoglobulin.


Abbreviations: Ab, antibody, HC: heavy chain, LC: light chain, MIg: monoclonal immunoglobulin

auto-antibodies were detected in 20/41 (49%) MIg-C3G patients, including 4 patients with combined anti-FH and anti-CR1 antibodies and 1 with anti-FI and anti-FH antibodies.

C3 and sC5b9 levels were similar in patients with or without antibodies (**Figures 3A,B**). Compared to C3GN patients without MIg, MIg-C3G patients had significantly lower frequency of C3NeF [3/41(7%) vs. 44/98(45%); p = 0.0001] and C5NeF (p = 0.002), higher frequency of anti-CR1 auto-antibodies [11/41(27%) vs. 3/84(4%); p = 0.0001] and similar frequency of anti-FH auto-antibodies (**Table 1**).

Functional studies were carried out in patients with anti-FH, anti-CR1 and anti-FI antibodies. We studied the impact of anti-FH antibodies on AP regulation by studying the capacity of FI to cleave C3b in iC3b in presence of FH. We performed a fluid phase cofactor assay in presence of total IgG purified from patients with anti-FH antibodies or healthy donors (HD). The C3b cleavage was revealed by Western Blot and ratio α43 chain on β-chain of C3b, determined by densitometry, was used to determine the % of C3b cleavage. C3b cleavage was significantly decreased in 3/9 patients with anti-FH antibodies (**Figures 3C,D**).

We next studied the functional properties of anti-CR1 antibodies. The presence of anti-CR1 antibodies resulted in decreased capacity (from 12 to 25%) of C3b to bind CR1, as demonstrated by SPR-based technology (**Supplemental Figures 2D–F**). Moreover, by Western blot, significant reduced CR1 cofactor activity for FI was obtained in presence of IgG purified from 4/7 anti-CR1 positive patients (**Figures 3E,F**). In 2 patients with anti-FI antibodies, C3b cleavage by FI in presence of FH was not decreased (data not shown).

#### Study of Light and Heavy Chain Isotype Specificity of Anti-complement Protein Antibodies

MIg heavy chain (HC) and light chain (LC) isotype specificities were determined by immunoblot in 29 patients (**Supplemental Table 2**).

Using ELISA, we determined heavy chain (HC) and light chain (LC) isotype specificity of anti-complement antibodies in 13 positive patients. In 3 cases, anti-FI IgA, anti-FH IgG or anti-FH IgA antibodies displayed similar HC and LC restriction as the MIg. In 10/13 (77%) positive patients, the MIg HC (all of IgG isotype) and/or LC did not match those of the respective auto-antibodies (**Table 2**).

### Patients' Ig Induce Fluid-Phase and Solid-Phase AP Convertase Activation

To test the capacity of total purified IgG (containing the MIg) of MIg-C3G patients to activate complement AP, we measured C3a release in normal human serum (NHS) by ELISA after incubation with patients' IgG or IgG from healthy donors (**Supplemental Figure 3**). For 10/32 patients' IgG, C3a level was above the mean+2SD cut-off obtained with IgG from healthy donors (**Supplemental Figure 3A**). C3a release was similar in MIg-C3G patients with or without anti-complement protein auto-antibodies (**Supplemental Figure 3B**).

To demonstrate that IgG of MIg-C3G patients directly enhance the C3 cleavage into C3b without the influence of auto-antibodies, purified C3, FB, FD were incubated with total IgG from controls (Healthy-donors (HD) and patients with MIg without kidney disease) and MIg-C3G patients and tested in solution or on IgG-coated plate in presence of EGTA-Mg2+. The % of C3 cleavage into C3b was determined by Western blot, by measuring the ratio between α ′ chain and βchain of C3b, determined by densitometry. Mean % of C3 cleavage was 38% in the presence of IgG from HDs in solution (mean + 2DS of the ratio = 56%) and 39% on HD-IgGcoated plate (mean + 2DS of the ratio = 59%). Cleavage of C3 was increased (higher than mean+2SD) in presence of 12/34 MIg-C3G patients' IgG in solution (**Figures 4A,B**) and on 13/34 IgG-coated plates (**Figures 4C,D**). Total IgG purified from 3 patients increased C3b formation both in solution and on coated IgG. Patients' IgG that activated the C3 convertase in solution or on coated phase were named "C3-activating IgG." Altogether 22/34 tested patients' IgG displayed capacity to cleave C3. In both experimental conditions, C3b formation was significantly higher compared to that obtained in presence of total IgG from patients with MIg but without kidney disease (**Figures 4B,D**). C3 cleavage was similar in MIg-C3G patients with or without anti-complement protein antibodies (**Supplemental Figures 3C,D**).

C3 levels were significantly lower in patients positive for C3 activating IgG than in those negative (P = 0.03) (**Figure 4E**), whereas there was no difference for sC5b9 in plasma (p = 0.94) (**Figure 4F**). Plasma sC5b9 levels were upper than twice the normal value in 12/22 (57%) patients with C3-activating IgG and in 3/14 (21%) patients without this capacity (p = 0.04).

FIGURE 4 | AP convertase activation in presence of patients' total purified IgG. (A) MIg-C3G patients' IgG were tested for their capacity to enhance fluid phase AP C3 convertase formation. Cleavage of C3 to C3b by fluid phase C3 convertase was measured by the generation of the α' chain. Result of patient G12 is provided compared to Healthy donor (HD) (B) C3 convertase activity was significantly increased in presence of MIg-C3G patients' IgG compared to Ig from HD or Ig from patients with MIg without kidney disease (MIg w/o KD). In presence of Ig from 12/34 patients, % of C3 cleavage was significantly increased [above the cut-off (mean+2SD)]. (C) MIg-C3G patients' IgG coated on well plates were tested for their capacity to enhance AP C3 convertase formation. Cleavage of C3–C3b by fluid phase C3 convertase was measured by the generation of the α' chain. Result of patient G38 is provided compared to HD (D) C3 convertase activity was significantly increased in presence of MIg-C3G patients' IgG compared to Ig from healthy donors or Ig from patients with MIg without kidney disease (MIg w/o KD). IgG from patients able to enhance C3 convertase in fluid phase or on well plate were named "C3-activating" Ig (E) C3 level of patients with "C3-activating" IgG was significantly increased compared to patients without "C3-activating" IgG. (F) sC5b9 level of patients with "C3-activating" IgG was similar to patients without "C3-activating" IgG.

### Monoclonal Ig Are Able to Enhance Fluid Phase C3 Convertase Overactivation

To identify the components involved in the AP activation, the MIg was separated from polyclonal Ig by chromatography in 2 patients with C3-activating IgG (G12, G20) and 3 patients without C3 activating IgG in fluid phase (G38, G40, G24). In samples from patients G12 and G20, C3b formation was increased in presence of the MIg compared to the polyclonal Ig (**Figures 5A,B**).

We also investigated whether the capacity of total IgG to activate the AP disappeared after complete hematological response (as assessed by negative serum immunofixation) following chemotherapy. Blood samples from 3 patients (G5, G37, G53) in whom total IgG were responsible for C3 cleavage in solution were available. In all three cases, C3 cleavage was significantly reduced in presence of total IgG purified from blood after treatment compared to that obtained with IgG from the same patients at diagnosis (**Figures 5C–E**).

#### DISCUSSION

We described for the first time the mechanisms of complement alternative pathway dysregulation in a peculiar group of patients with C3G associated with monoclonal immunoglobulin (MIg-C3G). We found anti complement antibodies in more than 50% of patients but with different target compare to C3G patients without monoclonal gammopathy suggesting that the two diseases are distinct. Moreover, our results

highlight a contribution of both monoclonal and polyclonal Ig in the inappropriate activation of complement AP in MIg-C3G patients, paving the way to new therapeutic strategies.

In the French cohort of adult C3G without detectable MIg, impaired complement control is driven by C3NeF and by genetic variation in complement genes in 45 and 27%, respectively. Genetic abnormalities were identified in only 7% of tested MIg-C3G patients, suggesting that genetic factors do not play a major role in MIg-C3G. This result is in agreement with those of a recent study in which none of 21 tested patients had any genetic abnormalities (18). Exhaustive screening identified auto-antibodies targeting complement proteins in about 50% of MIg-C3G patients. However, the targets of anti-complement auto-antibodies were different between C3G patients with and without MIg. Indeed, C3NeF was found in only 7% of MIg-C3G patients. This is in agreement with previous small cohort studies, which identified C3NeF in 0/6 and 2/9 MIg-C3G patients (14, 15). The presence of C5NeF stabilizing the C5 convertase has been recently described in 56% of patients with C3GN (6). Interestingly, despite elevated sC5b-9 level in 80% of MIg-C3G cases, C5NeF was negative in all tested patients. The frequency of anti-FH auto-antibodies was low and similar to C3G patients without MIg (11, 12). In contrast, we found that 27% of MIg-C3G patients had anti-CR1 auto-antibodies, undetectable in C3G patients without MIg. Interestingly, CR1 which is expressed by podocytes, emerges as a novel disease-relevant target in C3G (19) and auto-antibodies targeting CR1 have been found in patients with multiple myeloma (20). We further explored functional consequences of these antibodies. Cofactor activity of both CR1 and FH was decreased in 4/7 and 3/9 patients positive for anti-FH or anti-CR1 antibodies, respectively, whereas it was normal in two patients with anti-FI antibodies suggesting that these antibodies have limited functional consequences on AP regulation. C3 and sC5B9 levels were similar in patients with or without anti-complement protein antibodies, confirming the weak contribution of these antibodies in AP dysregulation in MIg-C3G.

The initial assumption was that autoantibodies targeting complement proteins were monoclonal. Indeed, in 1999, Jokiranta et al. demonstrated that a dimeric monoclonal lambda LC, identified in a patient with glomerulonephritis and predominant C3 deposits, was able to bind FH as an auto-antibody, resulting in uncontrolled AP activation in vitro (21). In the current study, we showed a concordance in the heavy and light chain isotypes of MIg and anti-complement protein auto-antibodies in only 3/13 patients. Therefore, our results suggest that in most cases anti-complement protein reactivity is not borne by the MIg. This result is in agreement with other kidney diseases mediated by auto-antibodies, where the implication of monoclonal autoantibodies remains exceptional (22, 23).

Further, we tested a new hypothesis according to which the MIg could serve directly as a complement-activating surface. We designed an experiment to study C3 cleavage without interference with the regulatory proteins and thus without the contribution of anti-complement protein antibodies. In 22/34 (65%) of cases, patients' IgG enhanced C3 cleavage, and therefore they could be considered as C3-activating IgG. Interestingly, C3 level was significantly lower in patients with C3-activating IgG than in those without. Moreover, the percentage of patients with sC5b9 levels higher than twice the normal value was significantly increased in patients with C3-activating IgG compared to those patients without C3-activating IgG. Interestingly, the capacity of patients' IgG to enhance C3 cleavage was not increase in patients with MIg but without kidney disease and the link between an ongoing complement activation in MIg-C3G patients and the MIg remains speculative. The direct role of the MIg in AP activation was strongly suggested in 5 patients. Indeed, in 3 of them, we demonstrated the disappearance of the capacity of total IgG to activate the C3 convertase once the MIg had become undetectable after chemotherapy. In 2 patients, we showed the increased capacity of the MIg to enhance fluid phase C3 convertase activity compared to the polyclonal IgG from the same patients. It is well established that MIg have peculiar physicochemical properties due to different profiles of glycosylation or mutations/deletions of the variable or constant domain (24). These peculiarities are likely to account for the variable capacity of these MIg to enhance C3 convertase in vitro. It is tempting to speculate that the nascent C3b, generated by slow fluid phase activation of C3, binds to MIg and forms a starting point for the subsequent assembly of C3 convertase.

In a recent clinical study, we demonstrated that achievement of rapid and deep hematological response with clone-targeted chemotherapy significantly improved renal survival in MIg-C3G patients and that C3 levels in patients with hematological response were significantly higher compared with pretreatment C3 levels (16). The present provides more support for a link between the monoclonal Ig and renal disease. Therefore targeting the responsible clone should be a therapeutic goal to preserve or improve renal function in these patients.

Our study has some limitations. It is a retrospective study with a relatively low number of patients. Most patients had low amounts of MIg making the MIg purification process difficult or even impossible. These limitations did not allow us to investigate the direct contribution of MIg in AP dysregulation in all patients, and further studies are needed to depict the full pathophysiological spectrum of MIg in C3GP.

In conclusion, our study highlight different complement AP activation mechanisms in C3G associated with MIg compared to C3G without MIg. We demonstrated that IgG isolated from MIg-C3G patients directly activate the AP in 65% of cases and our findings provide further evidence that monoclonal gammopathy is a cause of the disease, particularly in patients with very high levels of sC5b9 at diagnosis. Our results highlight the need to consider chemotherapy targeting the B cell clone in the treatment strategy of MIg-C3G patients.

# AUTHOR CONTRIBUTIONS

The study was conceived and designed by SC and VF-B. SC conducted the experiments and analysis; SC and VF-B were involved in the writing of the manuscript. VF-B and LR reviewed the data analysis; SC, VF-B, FB, and all other authors contributed to the conduct of the study, recruited patients, and were involved in the review of results and final approval of the manuscript.

# DISCLOSURE

VF-B received fees for participation in advisory boards, experts meetings and/or teaching courses from Alexion Pharmaceutical. YD received honoraria from Alexion Pharmaceutical for teaching symposia.

# FUNDING

This work was supported by the EU FP7 grant 2012-305608 (EURenOmics) (to VF-B), the KIDNEEDS research grant 2015 (to VF-B), the ANR research grant (ANR-16-CE18-0015-01, CompC3) (to VF-B), the Fondation du rein (FRM, Prix 2012 FDR) (to VF -B), the Association pour l'Information et la Recherche dans les maladies Rénales génétiques (AIRG France), the Fondation Pour La Recherche Medicale (FDM 20130727355) (to SC) and the Fondation Française pour la Recherche contre le Myélome et les Gammapathies monoclonales (SC, VF-B).

# ACKNOWLEDGMENTS

We gratefully acknowledge Morgane Mignotet for technical support and all colleagues who participated in this study: Dr Rémi Boudet (department of Nephrology, Brives), Pr Eric Daugas (department of Nephrology, Bichat, APHP), Pr Fadi Fakhouri (department of Nephrology, Nantes), Dr Florence Gallen Labbe (department of Nephrology, Valence), Dr Pierre Gobert (department of Nephrology, Avignon), Pr Marc Hazzan (department of Nephrology, Lille), Dr Lucile Mercadal (department of Nephrology, Pitié Salpétrière, APHP), Dr Mathilde Nouvier (department of Nephrology, Lyon), Dr Nicolas Martin Silva (department of Medicine, Caen), Dr Merabet (department of Hematology, Versailles), Dr Eric Renaudineau (department of Nephrology, Saint Malo), Dr Jean Baptiste Philit (department of Nephrology, Chambery), Dr Damien Sarret (department of Nephrology, Val de Grâce, Paris), Dr Aude Servais (department of Nephrology, Necker, APHP), Dr Lili Taghipour (department of Nephrology, Armentières), Dr Aurélien Tiple (department of Nephrology, Clermont-Ferrand), Pr Guy Touchard and Jean-Michel Goujon (department of Pathology, Poitiers).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02260/full#supplementary-material

### REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer GR declared a past co-authorship with two of the authors: VF and PR.

Copyright © 2018 Chauvet, Roumenina, Aucouturier, Marinozzi, Dragon-Durey, Karras, Delmas, Le Quintrec, Guerrot, Jourde-Chiche, Ribes, Ronco, Bridoux and Fremeaux-Bacchi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Unraveling the Molecular Mechanisms Underlying Complement Dysregulation by Nephritic Factors in C3G and IC-MPGN

Roberta Donadelli <sup>1</sup> , Patrizia Pulieri <sup>1</sup> , Rossella Piras <sup>1</sup> , Paraskevas Iatropoulos <sup>1</sup> , Elisabetta Valoti <sup>1</sup> , Ariela Benigni <sup>1</sup> , Giuseppe Remuzzi 1,2,3 and Marina Noris <sup>1</sup> \*

<sup>1</sup> Clinical Research Center for Rare Diseases Aldo e Cele Daccò and Centro Anna Maria Astori, Science and Technology Park Kilometro Rosso, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Bergamo, Italy, <sup>2</sup> Unit of Nephrology and Dialysis, Azienda Socio-Sanitaria Territoriale Papa Giovanni XXIII, Bergamo, Italy, <sup>3</sup> Department of Biomedical and Clinical Sciences, University of Milan, Milan, Italy

#### Edited by:

Tom E. Mollnes, University of Oslo, Norway

#### Reviewed by:

Lambertus Petrus Van Den Heuvel, Radboud University Nijmegen Medical Centre, Netherlands Marc Seelen, University Medical Center Groningen, Netherlands

> \*Correspondence: Marina Noris marina.noris@marionegri.it

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 06 April 2018 Accepted: 19 September 2018 Published: 15 October 2018

#### Citation:

Donadelli R, Pulieri P, Piras R, Iatropoulos P, Valoti E, Benigni A, Remuzzi G and Noris M (2018) Unraveling the Molecular Mechanisms Underlying Complement Dysregulation by Nephritic Factors in C3G and IC-MPGN. Front. Immunol. 9:2329. doi: 10.3389/fimmu.2018.02329 Membranoproliferative glomerulonephritis (MPGN) was recently classified as C3 glomerulopathies (C3G), and immune-complex (IC) mediated MPGN. Dysregulation of the complement alternative pathway, driven by acquired and/or genetic defects, plays a pathogenetic role in C3G. However, alternative pathway abnormalities were also found in IC-MPGN. The most common acquired drivers are the C3 nephritic factors (C3NeFs), heterogeneous autoantibodies that stabilize the C3 convertase, C3bBb. C3NeFs are traditionally detected by hemolytic assays based on sheep erythrocyte lysis, which however do not provide a direct molecular estimation of C3bBb formation and decay. We set up a microplate/western blot assay that specifically detects and quantifies C3bBb, and its precursor, the C3 proconvertase C3bB, to investigate the complex mechanistic effects of C3NeFs from patients with primary IC-MPGN (n = 13) and C3G (n = 13). In the absence of properdin, 9/26 patients had C3NeF IgGs stabilizing C3bBb against spontaneous and FH-accelerated decay. In the presence of properdin the IgGs of all but one patient had C3bBb-stabilizing activity. Properdin-independent C3NeFs were identified mostly in DDD patients, while properdin-dependent C3NeFs associated with either C3GN or IC-MPGN and with higher incidence of nephrotic syndrome. When we grouped patients based on our recent cluster analysis, patients in cluster 3, with highly electron-dense intramembranous deposits, low C3, and mostly normal sC5b-9 levels, had a higher prevalence of properdin-independent C3NeFs than patients in clusters 1 and 2. Conversely, about 70% of cluster 1 and 2 patients, with subendothelial, subepithelial, and mesangial deposits, low C3 levels and high sC5b-9 levels, had properdin-dependent C3NeFs. The flexibility of the assay allowed us to get deep insights into C3NeF mechanisms of action, showing that: (1) most C3NeFs bind strongly and irreversibly to C3 convertase; (2) C3NeFs and FH recognize different epitopes in C3 convertase; (3) C3NeFs bind rapidly to C3 convertase and antagonize the decay accelerating activity of FH on newly formed complexes; (4) C3NeFs do not affect formation and stability of the C3 proconvertase. Thus, our study provides a molecular approach to detecting and characterizing C3NeFs. The results highlight different mechanisms of complement dysregulation resulting in different complement profiles and patterns of glomerular injury, and this may have therapeutic implications.

Keywords: C3 glomerulopathy, membranoproliferative glomerulonephritis, complement alternative pathway, C3 nephritic factors, terminal complement complex, C3 convertase, factor H

#### INTRODUCTION

Membranoproliferative glomerulonephritis (MPGN) is a rare chronic kidney disorder associated with thickening of the glomerular capillary wall and mesangial expansion, which are the result of the deposition of immune-complexes and complement factors (1, 2). Evidence of the participation of the third fraction of complement C3 in this type of lesion, dates back to the middle of the last century (3, 4). Traditionally, on the basis of electron microscopy localization of electron-dense deposits relative to the glomerular basement membrane, MPGN was divided into type I, with subendothelial deposits (5), type II or Dense-Deposit Disease (DDD), with intramembranous highly electron-dense deposits (6), and type III, with subendothelial and subepithelial deposits (7).

An advanced approach toward an etiology-based diagnosis of this disease arose from the more recent immunofluorescence (IF)-based classification that distinguishes between immunecomplex-mediated MPGN (IC-MPGN), with glomerular IgG and C3 deposits, and C3 glomerulopathy (C3G) with predominant C3 deposits (8, 9). C3G is further divided into DDD, with distinctive highly electron-dense deposits within the glomerular basement membrane; and C3 glomerulonephritis (C3GN), with mesangial, intramembranous, subendothelial and sometimes subepithelial deposits (1, 6, 10, 11). According to the current classification, IC-MPGN derives from the deposition of immune-complexes that form in the context of infections, autoimmune diseases and malignancies, and trigger the classical complement pathway (8). C3G instead arises from abnormalities in the control of the complement alternative pathway (8). However the distinction of IC-MPGN as an immune-complex mediated disease and C3G as an alternative pathway-mediated disease leaves a number of questions open. In up to 25–30% of cases classified as IC-MPGN based on the IF finding, an underlying condition could not be identified (11, 12). In addition, in about 16% of patients undergoing repeated kidney biopsies the diagnosis shifted from IC-MPGN to C3G and vice versa (11, 13). Finally, genetic and acquired alternative pathway abnormalities have been found as frequently in patients with IC-MPGN as in those with C3G (14–16), suggesting that dysregulation of the complement alternative pathway may play a role in both conditions.

The most common acquired drivers in primary IC-MPGN and C3G are the C3 nephritic factors (C3NeFs) (17–20) a heterogeneous group of autoantibodies that bind to neoepitopes of the C3 convertase complex C3bBb, the key amplifying enzyme of the complement alternative pathway that cleaves C3 to C3a and C3b (21, 22). C3NeFs stabilize both the fluid-phase and the cell-bound C3 convertase (23), markedly increasing its half-life and inhibiting the accelerated decay mediated by complement regulators such as factor H (FH), decay-accelerating factor (DAF) and complement receptor 1 (CR1) (20, 24, 25). C3NeFs are present in 40-50% of patients with IC-MPGN or C3GN and in 70–80% of patients with DDD (14, 16, 26).

A consensus regarding the detection and characterization of C3NeFs is lacking. Few specialized laboratories routinely measure C3NeFs using different methods: they predominantly use hemolytic assays based on the evaluation of the lysis of sheep erythrocytes carrying preformed alternative pathway C3 convertase (27). There is still uncertainty regarding the pathogenetic role of C3NeFs in IC-MPGN and C3G. C3NeF activity, as measured by the hemolytic assays correlates poorly with disease activity and with circulating complement parameters. Some C3NeFs induce C3 consumption in the fluid phase, but do not always enhance the activation of the terminal pathway (28, 29), other C3NeFs are even associated with normal or near normal C3 levels (30). Recent studies have explored the mechanisms of action of C3NeFs in patients with C3G using several functional assays based on ELISA (31), surface plasmon resonance (31) and modified hemolytic assays (26, 28), which confirmed the heterogeneous nature of C3NeFs. Some C3NeFs were found to require the presence of properdin to carry out their stabilizing activity on the C3 convertase (26, 28, 31) and were called C5NeFs since they also activated the terminal complement pathway, while other C3NeFs were properdin-independent and had no effect on C5-cleavage. In one study, some correlation was found with complement profile and histology diagnosis: indeed, properdin-dependent C3NeFs (C5NeFs) were more frequent in patients with C3GN and high plasma levels of the soluble terminal complement complex sC5b-9, while properdin-independent C3NeFs were associated with DDD and complement activation restricted to the C3 level (28). However, electron microscopy examination was lacking in about half of patients, which questions regarding the distinction between C3GN and DDD. In addition, the above published studies did not include patients with IC-MPGN.

In this study we set up a user-friendly method that specifically detects and quantifies the alternative pathway C3 convertase, C3bBb, and its precursor, the C3 proconvertase C3bB, to investigate the complex mechanistic effects of C3NeFs in the pathogenesis of complement dysregulation in patients with IC-MPGN and C3G. The specific aims were: (1) to characterize and quantify C3NeF activity in stabilizing C3 convertase and its FH-mediated accelerated decay; (2) to investigate the effect of properdin on the activity of C3NeFs; (3) to evaluate whether the results of C3NeFs assays correlate with complement profile, clinical features and histological diagnosis.

Through unsupervised hierarchical clustering we recently identified, among IC-MPGN and C3G patients, four clusters, indicating the existence of different pathogenetic patterns underlying complement activation and glomerular injury (15). A large majority of patients carrying C3NeFs, as measured by the hemolytic assay, were included in clusters 1–3, while patients in cluster 4 had a very low prevalence of C3NeFs. In addition, clusters 1–2 included patients with alternative pathway activation at the C3 and C5 levels, as highlighted by low serum C3 and high plasma sC5b9 levels; while patients in cluster 3 had alternative pathway activation mainly at the C3 level (15). Thus, a further aim of this study was to use the new method to compare the activity of C3NeFs between patients in clusters 1–3.

Finally we investigated the mechanisms by which C3NeFs antagonize the FH-mediated regulatory activity on C3 convertase and whether C3NeFs had any effect on the formation and decay of the C3 proconvertase, issues that have not been investigated in depth so far.

# MATERIALS AND METHODS

# Patients and Controls

Twenty-six patients with primary IC-MPGN or C3G were selected from the previously described cohort (15) based on positivity for C3NeF by hemolytic assay (>20%) (27), as well as on the availability of suitable amounts of serum samples collected outside of treatment with steroids, immunosuppressive drugs or eculizumab.

A diagnosis of MPGN was established in patients with the typical light microscopy pattern with mesangial hypercellularity, endocapillary proliferation, and capillary wall remodeling (1). Patients with glomerular immunoglobulin and C3 deposits at IF were considered IC-MPGN (8). C3G was diagnosed based on the presence of MPGN or mesangial proliferative patterns with "dominant C3" glomerular staining (C3c staining intensity at least two orders of magnitude greater than any other immunereactant including IgG, IgM, IgA, and C1q on a 0 to 3 scale) (10). Based on electron microscopy, C3G was further classified as DDD or C3GN. Patients with MPGN secondary to autoimmune diseases, monoclonal gammopathy, infections (HBV, HCV and HIV), neoplasia or atypical hemolytic uremic syndrome were excluded.

Control sera were obtained from 26 donors with no history of renal disease. Twelve patients with primary IC-MPGN or C3G (IC-MPGN, n = 5; C3GN, n = 4; DDD, n = 3) without C3NeF activity by hemolytic assay were also studied as additional controls. The above patients were selected to be homogenously distributed among clusters 1 to 4 (n = 3 for each cluster) (15).

All participants provided informed written consent. The study adheres to the Declaration of Helsinki and was approved by the Ethics Committee of the Azienda Sanitaria Locale of Bergamo (Italy).

# Serum and Plasma Complement Profile and Genetic Analysis

Complement C3 levels in serum were measured by nephelometry. SC5b-9 and C3d levels were evaluated in plasma by MicroVue sC5b-9 Plus EIA (sC5b-9 Plus; Quidel), and ELISA (Human complement C3d ELISA kit Assaypro LLC), respectively. For the latter tests, blood was collected in ice-cold EDTA tubes and immediately centrifuged at 4◦C to avoid ex vivo complement activation. Plasma was quickly separated and frozen at −80◦C until assayed. Normal ranges were defined as follows: C3, 90–180 mg/dl (mean ± 2 SD of the laboratories of the ASST Papa Giovanni XXIII, Bergamo, Italy); sC5b-9, 127–400 ng/ml (mean ± 2 SD of 50 healthy control subjects); C3d/C3 ratio, 0.01-0.09 (mean ± 2SD of 15 healthy control subjects).

Screening for CFH, MCP, CFI, CFB, C3 and THBD coding sequences was performed by amplicon-based next generation sequencing (14). Rare functional variants (missense, nonsense, indel, or splicing variants with minor allele frequency, MAF <0.001 in 1,000 Genomes and ExAC databases) were selected and defined as likely pathogenetic, or pathogenetic when published functional studies were available. Anti-FH autoantibodies were measured in plasma by ELISA (32).

#### Selective C3bB C3 Proconvertase and C3bBb C3 Convertase Formation and Decay Assays

For the generation of alternative pathway C3bB C3 proconvertase and C3bBb C3 convertase we first tried an ELISA assay previously described by Hourcade et al. (33, 34). ELISA plates (Nunc-Immuno Maxisorp) coated with 3µg/ml C3b (Complement Technology Inc., Tyler, TX) were incubated at 37◦C for 2 h with different concentrations of FB (0–500 ng/ml, Complement Technology Inc.) in the absence or in the presence of FD (25 ng/ml, Complement Technology Inc.), respectively, both diluted in assay buffer (8.1 mM Na2HPO4, 1.8 mM NaH2PO4, 4% BSA, 0.1% Tween20, and 75 mM NaCl) containing 2 mM NiCl2. After washes with wash buffer (8.1 mM Na2HPO4, 1.8 mM NaH2PO4, 0.1% Tween20 and 25 mM NaCl) supplemented with 2 mM NiCl2, the C3bB and C3bBb complexes were detected by ELISA using polyclonal goat anti-human FB antibody (Quidel, San Diego, CA) diluted 1:10,000, followed by HRP-conjugated anti-goat antibody (1:40,000; Sigma Aldrich), both diluted in antibody buffer (8.1 mM Na2HPO4, 1.8 mM NaH2PO4, 4% BSA, 0.1% Tween20 and 25 mM NaCl) supplemented with 2 mM NiCl2. Color was developed using 3,3′ , 5, 5′ -teramethylbenzidine (TMB) substrate (Tema Ricerca srl, Bologna, Italy) and stopped with H2SO<sup>4</sup> 2 M, and absorbance was measured at 450 nm. Each reaction was performed in duplicate and the OD values averaged. As shown in **Supplementary Figure 1A**, the ELISA curves from the reaction of coated C3b with either FB alone or FB plus FD showed dose-dependent superimposable profiles, indicating that the ELISA assay cannot selectively discriminate between C3bB and C3bBb formation. To specifically generate either C3bB or C3bBb complexes, we exploited the selective stabilization abilities of Mn2+and Mg2+, respectively (35, 36). For the generation of C3bB(Mn2+) or C3bBb(Mg2+) complexes, C3b-coated microtiter wells were treated as above, except that the incubation was performed in the presence of 2 mM MnCl<sup>2</sup> for 2 h at 37◦C (**Supplementary Figure 1B**) or 10 mM MgCl<sup>2</sup> for 30 min at 25◦C (**Supplementary Figure 1C**), respectively. In either condition the amount of the complexes formed was too small to be detected by ELISA, likely due to dissociation of FB and Bb from C3bB(Mn2+) and the C3bBb(Mg2+) respectively, during post-reaction incubations with primary and secondary antibodies and the intercurrent washing steps.

Thus we used a combined microplate and western blot (WB) technique to specifically detect C3bB and C3bBb (37). Briefly, C3bB(Mn2+) complexes were formed by incubating C3bcoated wells at 37◦C for 2 h with FB (1,000 ng/ml), diluted in the assay buffer (8.1 mM Na2HPO4, 1.8 mM NaH2PO4, 0.1% Tween 20, and 75 mM NaCl) supplemented with 4% BSA, and 2 mM MnCl2. C3bBb(Mg2+) complexes were formed by incubating C3b-coated wells at 25◦C for 12 min with FB (1,000 ng/ml) and FD (10 ng/ml,) both diluted in assay buffer supplemented with 0.5% BSA and 10 mM MgCl2. After washes, the protein complexes were detached from microtiter wells with EDTA 10 mM and SDS 1% for 1 h, subjected to 10% SDS-PAGE, and transferred by electroblotting to Hybond-P hydrophobic polyvinylidene difluoride (PVDF) membrane (Amersham Biosciences, GE Healthcare, Euroclone Spa, Milano Italy). Proteins were detected with a polyclonal goat antihuman FB antibody (Quidel, San Diego, CA) followed by HRP-conjugated anti-goat Ab (Sigma Aldrich) and the ECL system (Amersham Biosciences, GE Healthcare, Euroclone Spa, Milano Italy). C3-proconvertase and C3-convertase formation were evaluated by the visualization by WB of the FB (93 kDa) or the Bb band (60 kDa), respectively (**Supplementary Figure 1D**). Images were acquired by the Odyssey Imager instrument (Licor).

For the generation of C3bB(Ni2+) and C3bBb(Ni2+) complexes, C3b-coated wells were incubated at 37◦C with FB (1,000 ng/ml) and FD (10 ng/ml) in assay buffer containing 2 mM NiCl<sup>2</sup> for 30 min. Then the Ni2+-protein complexes formed in the presence or in the absence of FD were evaluated by WB as described above. In the presence of FD, both FB (93 KDa) and Bb (60 KDa) bands were visualized (**Supplementary Figure 1D**) indicating that only a portion of C3bB was converted to C3bBb.

The intensity of the bands detected in WB was estimated by densitometry using NIH Software ImageJ (NIH, USA).

To study spontaneous or FH-mediated decay of C3 convertase and C3 proconvertase over time, C3bBb(Mg2+) and C3bB(Mn2+) were allowed to form for 12 min at 25◦C and 2 h at 37◦C, respectively, as reported above. In some wells the formed complexes were detached from microtiter wells with EDTA 10 mM and SDS 1%, subjected to WB analyses as described above and used as baseline.

Additional wells were then washed and the formed complexes were incubated with selective assay buffers in the presence or in the absence of FH (2,640 ng/ml; physiological molar ratio FB:FH = 1:1.6; Merck) for the following time periods: C3bBb(Mg2+), 2, 8, 16, and 32 min; C3bB(Mn2+), 30, 60, 120, and 240 min.

To study the spontaneous or FH-mediated decay of C3bB and C3bBb simultaneously, C3bB(Ni2+) and C3bBb(Ni2+) were formed together during a 30 min period at 37◦C. In some wells the formed complexes were detached from microtiter wells with EDTA 10 mM and SDS 1%, subjected to WB analyses, and used as baseline. Additional wells were then washed and spontaneous and FH-mediated decay of the formed complexes were monitored by incubating the wells at 37◦C with assay buffer alone or with FH (2,640 ng/ml) for 5, 30, 60, and 120 min. Following washes, the remaining complexes were detached from microtiter wells with EDTA 10 mM and SDS 1% and subjected to WB analyses. The percentage of residual Bb or B band was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of the Bb or B bands after decay and the corresponding baseline Bb or B band density before decay x 100.

#### C3NeF C3 Convertase-Stabilizing Activity Assays

IgGs were isolated from sera by the Melon Gel IgG Purification Kit (Thermo Scientific, VWR International PBI srl, Milano Italy)

To evaluate the ability of C3NeF IgG to stabilize the alternative pathway C3 convertase, C3bBb, against spontaneous and FHmediated decay, we adapted the microplate/WB described above. Three protocols were set up. Each C3NeF IgG sample was analyzed three times.

**Protocol 1**. C3b-coated wells were incubated for 12 min at 25◦C with 1,000 ng/ml FB, 10 ng/ml FD, 100µg/ml IgGs purified from patients or controls, and 10 mM MgCl2, in the absence or in the presence of 2,640 ng/ml FH. In some wells the formed complexes were detached with EDTA 10 mM and SDS 1%, and subjected to WB analyses. The amount of C3bBb formed in the absence of FH was used as the baseline.

In additional wells, the complexes formed in the absence of FH were washed and C3NeF IgG C3 convertase-stabilizing activity (CSA) was evaluated by incubation for 32 min at 25◦C with buffer alone (sCSA) or with buffer containing 2,640 ng/ml FH, (FH-CSA). Following washes, the residual C3bBb C3 convertase complexes were detached from microtiter wells with EDTA 10 mM and SDS 1% and subjected to WB analyses. The intensity of the band detected in WB was estimated by densitometry using NIH Software ImageJ (NIH, USA). The percentage of residual Bb band was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of each Bb band after decay and the corresponding baseline Bb band density before decay x 100.

**Protocol 2**. C3b coated wells were incubated for 12 min at 25◦C with 1,000 ng/ml FB, 10 ng/ml FD, and 10 mM MgCl2. After washing, the complexes formed were incubated for 32 min at 25◦C in the presence of 100µg/ml IgGs purified from patients or healthy controls without (spontaneous decay) or with (FH-mediated decay) 2,640 ng/ml FH. Following washes, the residual C3bBb C3 convertase complexes were detached from microtiter wells. The C3NeF IgGs C3 convertase stabilizing activity was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of Bb bands after decay in the presence of patient IgGs and the Bb bands after decay in the presence of control IgGs.

**Protocol 3**. The C3 convertase-stabilizing activity assay with properdin (PCSA) assay, is similar to the CSA, with the only difference that properdin (P; purified from human serum, Complement Technology Inc.) is included in the protocol during C3 convertase formation. Briefly, C3b-coated wells were incubated for 12 min at 25◦C with 1,000 ng/ml FB, 10 ng/ml FD, 100µg/ml IgGs purified from patients or healthy controls and 10 mM MgCl2, in the presence of 500 ng/ml properdin (Complement Technology Inc.). The C3 convertase formed, C3bBbP, was allowed to decay in the absence (sPCSA) or in the presence of FH (FH-PCSA) for 32 min at 25◦C. After washes, the following steps were identical to those described above. The percentage of residual Bb band was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of each Bb band after decay and the corresponding baseline Bb band density before decay x 100. For FH detection, WB membranes from all three protocols were labeled with a goat anti-FH antibody (Calbiochem, Billerica, MA, dil 1:10,000) followed by HRP-conjugated anti-goat Ab (Sigma Aldrich, dil 1:10,000) and the ECL system (Amersham Biosciences, GE Healthcare).

# Effect of C3NeF on C3bB(Mn2+), C3bB(Ni2+), and C3bbb(Ni2+) Decay

C3b-coated wells were incubated for 2 h at 37◦C with 1,000 ng/ml FB, and 2 mM MnCl<sup>2</sup> in the presence of 100µg/ml IgGs purified from patients P5 and P10 or healthy controls. In some wells the formed complexes were detached from microtiter wells with EDTA 10 mM and SDS 1%, and subjected to WB analyses as described above. In additional wells, the C3bB proconvertase formed, C3bB(Mn2+), was allowed to decay in the absence or in the presence of FH for 30, 60, 120, and 240 min at 37◦C. Following washes, the residual C3bB(Mn2+) complexes were detached and subjected to WB analyses. The intensity of the B band detected in WB was estimated by densitometry using NIH Software ImageJ (NIH, USA). The percentage of residual B band was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of each B bands after decay and the corresponding baseline B band density before decay x 100.

C3bB(Ni2+) and C3bBb(Ni2+) were formed together during a 30 min period at 37◦C, incubating C3b-coated wells with 1,000 ng/ml FB, 10 ng/ml FD, and 2 mM NiCl<sup>2</sup> without (baseline) or with 2,640 ng/ml FH in the absence or in the presence of 100µg/ml IgGs purified from patients P5 and P10 or healthy controls. In some wells the formed complexes were detached from microtiter wells with EDTA 10 mM and SDS 1%, subjected to WB analyses, and used as baseline. Additional wells were washed and spontaneous and FH-mediated decay of the complexes formed without FH were monitored by incubating the wells at 37◦C with assay buffer alone or with FH (2,640 ng/ml) for 30 min. Following washes, the remaining complexes were detached from microtiter wells and subjected to WB analyses. The percentage of residual Bb or B band was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of the Bb or B bands after decay and the corresponding baseline Bb or B band density before decay x 100.

For FH detection, WB membranes were labeled with a goat anti-FH antibody followed by HRP-conjugated anti-goat Ab and the ECL system.

# Statistical Analysis

Continuous variables were analyzed by ANOVA. The Fisher Exact test was used for categorical variables. Linear regression analysis was used for correlations of continuous variables. P values < 0.05 were considered to be statistically significant. Analyses were performed using the R platform v.3.5.0, and the MedCalc v.12.2.1.0 software. The intra and inter-assay coefficients of variation were calculated as SD/mean x 100.

#### RESULTS

#### Description of Patients

The clinical, histologic and biochemical data of the 26 C3NeF<sup>+</sup> patients are shown in **Table 1**. According to the IF-based classification, 13 patients had IC-MPGN and 13 had C3G (C3GN n = 4, DDD, n = 9). Age of onset ranged from 4.9 to 30.6 years and did not differ among histology groups (C3GN: 12.6±4.2; DDD: 12.2 ± 5.4; IC-MPGN: 12.9 ± 6.8 years). Males accounted for 50, 56, and 54% of the C3GN, DDD, and IC-MPGN patients, respectively. At onset all patients had hematuria (micro or gross hematuria) and/or proteinuria. Urinary protein excretion levels did not significantly differ between histology groups (C3GN: 1.8 ± 1.4; DDD: 2.4 ± 2.6; IC-MPGN: 4.7 ± 4.4.g/24 h). Renal function at onset was normal or near normal in all patients (serum creatinine: C3GN: 0.68 ± 0.2; DDD: 0.56 ± 0.2; IC-MPGN: 0.72 ± 0.2 mg/dl) (**Table 1**). C3 levels at onset were low and did not differ between histology groups (C3GN: 40 ± 11; DDD: 19 ± 19; IC-MPGN: 21 ± 15 mg/dl), while plasma sC5b-9 levels were higher in C3GN (1249 ± 736 ng/ml) and IC-MPGN (1895 ± 1291 ng/ml) than in DDD (416 ± 374 ng/ml, P < 0.01) (**Table 1**). No significant correlation was found between serum C3 (r = −0.31, P = 0.15) or plasma sC5b-9 (r = 0.30, P = 0.15) levels and urinary protein excretion at onset. Complement gene likely pathogenetic variants were identified in 4 patients (1 in CFH, 2 in **CFB,** and 1 in THBD, **Table 2**). Anti-FH autoantibodies were identified in two patients (P26 with homozygous CFHR3/CFHR1 deletion and P15 with two copies of CFHR3/CFHR1) (**Table 2**). All patients were positive for C3NeF by the hemolytic assay, as per inclusion criteria (**Table 2**).

### Effect of C3NeFs From Patients With C3G/IC-MPGN on AP C3 Convertase Decay

To evaluate the ability of C3NeF IgGs to stabilize the alternative pathway C3 convertase, C3bBb, and investigate the underlying molecular mechanisms, we developed a microplate/WB assay that specifically detects the Bb component of the C3bBb convertase, exploiting the selective stabilization properties of Mg2<sup>+</sup> on C3bBb.

Since C3NeFs are known to stabilize the C3 convertase against both spontaneous and/or FH-mediated decay (25, 31) we first studied the kinetics of spontaneous or accelerated C3 convertase decay. C3bBb(Mg2+) was allowed to form for 12 min on C3b-coated wells in the presence of FB and FD (baseline). In additional wells the formed complexes were subsequently incubated for different time intervals with buffer alone or with buffer containing FH at a physiological ratio with FB. As shown in **Figure 1A**, in the absence of FH, C3bBb(Mg2+) dissociated


TABLE 1 | Clinical, biochemical

 and histologic parameters

 of patients.

**418**

\*IF IgM staining: 3+;

Normal values : Uprot, <0.2 g/24.h; sCreat: <1.3 mg/dl for adults, <0.5 mg/dl in children <5 years of age, and <0.8 mg/dl in chidren >5 years; C3 levels.: 90–180 mg/dl; plasma sC5b-9 <400 ng/ml.

\*\*Measured by dipstick.


 range.

 were

 were

 are

 mean

 n.a,

in a time-dependent manner. The decay of C3bBb(Mg2+) was strongly accelerated in the presence of FH. Indeed, no Bb band could be visualized after 8 min of decay with FH (**Figure 1A**), confirming that FH is very efficient in Bb displacement from C3b (21). In order to analyze the effect of C3NeFs on the stabilization of the C3bBb(Mg2+) C3 convertase complexes, the above experiments were repeated forming C3bBb(Mg2+) in the presence of 100µg/ml IgGs purified from a C3G patient (P5) strongly positive for C3NeF or from a healthy subject (CTR) (protocol 1, **Figure 2A**). Results showed that IgGs from P5 greatly stabilized C3bBb(Mg2+) complexes by preventing spontaneous decay (**Figure 1B**), as documented by the intensity of the Bb band, which did not substantially change vs. baseline (before the decay). Similarly, P5 IgGs greatly prevented FH-mediated decay (**Figure 1B**). As shown in **Figure 1C**, the capability of P5 IgGs to stabilize C3bBb was dose-dependent. At variance, control IgGs did not affect either spontaneous or FH-accelerated C3bBb decay (**Figures 1C,D**).

Thereafter, 100 µg/ml IgGs from 26 patients who were C3NeF-positive in the hemolytic assay were tested for C3bBb stabilizing activity. IgG from 26 healthy donors were used to set the detection thresholds. Results were considered positive when the percentage of residual C3 convertase after spontaneous or FH-mediated decay was ≥ 37% or ≥12 % of baseline (> mean + 2SD of results from samples with control IgG, **Supplementary Figures 2A,B**), respectively.

Of the 26 tested patients' IgGs, 9 (35%) had C3 convertasestabilizing activity (CSA) against both spontaneous (sCSA+) and FH-accelerated decay (FH-CSA+), one stabilized only spontaneous decay (P8), and one only the FH-mediated decay (P14) (**Table 2**). IgGs from all the C3G/IC-MPGN controls without C3NeF activity were sCSA<sup>−</sup> and FH-CSA<sup>−</sup> (**Supplementary Figures 2A,B**). The mean intra- and inter-assay CVs were 12.5 and 20.6% respectively for sCSA, and 12.4 and 18.4%, respectively for FH-CSA.

Because all components not bound to well surface were removed before the decay, following C3 convertase formation, the positivity in the above test indicates stable interaction between C3NeF IgGs and C3 convertase.

**Figures 2B,C** are representative images of WB analyses showing the effect of IgGs isolated from 3 patients (P19, DDD; P24, C3GN, and P25 IC-MPGN) and one healthy control on C3 convertase stabilization against the spontaneous and accelerated decay. P19 IgG strongly stabilized C3bBb(Mg2+) complexes by preventing both spontaneous and FH-mediated decay. In contrast, P24 and P25 IgGs failed to prevent spontaneous and accelerated decay.

We observed a trend toward a higher prevalence of sCSA<sup>+</sup> and FH-CSA<sup>+</sup> C3NeF IgGs among patients with DDD than in patients in the other histology groups but the difference did not reach statistical significance (**Table 3A**).

A clear-cut difference in the C3 convertase-stabilizing activity of IgGs was found when we grouped the patients according to recently reported cluster analysis (15). Nine patients (the 4 with C3GN, 1 with DDD and 4 with IC-MPGN) fell into cluster 1 and were characterized by low serum C3 and very high plasma sC5b-9 at onset, and subendothelial and frequently subepithelial and mesangial deposits (**Tables 1**, **4**). Eight patients (all with IC-MPGN) were grouped in cluster 2, and showed low serum C3, high plasma sC5b-9, prevalent subendothelial deposits and stronger glomerular staining of IgG and C1q (P < 0.01) than patients in clusters 1 and 3 (**Tables 1**, **4**). Finally, the 9 patients of cluster 3 (8 with DDD and 1 with IC-MPGN) had low serum C3 and normal plasma sC5b-9 levels (P < 0.01 vs. clusters 1 and 2). Cluster 3 patients, including the one with a diagnosis of IC-MPGN, were characterized by intramembranous highly electron-dense deposits (**Tables 1**, **4**).

A significantly higher prevalence of patients who fell within cluster 3 had C3NeF IgGs that stabilized the C3 convertase against spontaneous (sCSA+) and FH-accelerated (FH-CSA+) than did patients from clusters 1 and 2 (**Table 3A**). Levels of sCSA and FH-CSA activities of C3NeF IgGs were significantly higher in cluster 3 than in clusters 1 and 2 (**Figures 3A,B**), while no significant difference in C3NeF IgGs activities was observed between histology groups (**Figures 3C,D**).

C3 convertase-stabilizing activity values of C3NeF IgGs negatively correlated with plasma sC5b-9 levels (r = −0.49, P = 0.01 for both sCSA and FH-CSA activities) measured at the time of IgG isolation, whereas no correlation was found between C3NeF activity measured by hemolytic assay and plasma sC5b-9 (r = 0.15, P = 0.48). Neither C3 convertase-stabilizing activities nor C3NeF activity by hemolytic assay correlated with C3d/C3 ratios (sCSA, r = −032, P = 0.14; FH-CSA, r = −0.27, P = 0.21; hemolytic assay, r = 0.01, P = 0.98). C3 convertase-stabilizing activity values did not correlate with results of the hemolytic assays (sCSA, r = 0.27, P = 0.19; FH-CSA, r = 0.31, P = 0.12).

#### Effect of C3NeF on Preformed C3 Convertase

In order to evaluate the ability of C3NeF IgGs to stabilize preformed C3 convertase, we modified the above assays by adding the IgGs after C3 convertase formation, only during the decay step (protocol 2, **Figure 4A**). Results were considered positive when the ratio of residual convertase in the sample with C3NeF IgGs vs. residual convertase in the sample with control IgGs run in parallel was >1.9 or >2.3 for spontaneous or FH-mediated decay, respectively (ratios between the mean + 2SD and the mean of residual Bb band in the reactions with IgGs from 26 healthy subjects). Of the 26 tested patients' C3NeF IgGs, all those that prevented both spontaneous and FH-mediated decay when added during C3 convertase assembly in protocol 1, also stabilized preformed C3 convertase in protocol 2.

Notably, IgGs from P24 stabilized C3 convertase against both spontaneous and FH-mediated decay only when added during decay (protocol 2, **Figure 4A** and **Table 2**), suggesting that these IgGs interacted weakly with C3 convertase. The IgGs from the other C3NeF<sup>+</sup> patients and from all the C3G/IC-MPGN controls without C3NeF activity by hemolytic assay (**Supplementary Figures 2E,F**) were negative in protocol 2. Representative images of the results of the assay on preformed C3 convertase are shown in **Figures 4B,C**.

FIGURE 1 | Spontaneous and FH-mediated decay of the C3bBb(Mg2+) C3 convertase in the absence or in the presence of IgGs from a DDD patient or a healthy control, by microplate/Western blot (WB) assay. (A) Time course of spontaneous and FH-mediated decay of C3bBb(Mg2+). The complexes were formed by incubating for 12 min at 25◦C C3b-coated wells with 1,000 ng/ml FB, 10 ng/ml FD, and 10 mM MgCl<sup>2</sup> (baseline). In additional wells after washing, the formed complexes were further incubated for 2, 8, 16, and 32 min in the absence (-) or in the presence (+) of 2,640 ng/ml FH. (B) C3bBb(Mg2+) was formed in the presence of IgGs from P5 patient with DDD during 12 min at 25◦C (baseline). The complexes were allowed to decay for 2, 8, 32, and 60 min in the absence (–) or in the presence of 2,640 ng/ml FH (+). (C) C3bBb(Mg2+) was formed in the presence of IgGs (P5-IgGs, 50, and 100µg/ml) from patient 5 or from an healthy subject (CTR-IgGs, 100µg/ml) for 12 min at 25◦C. Spontaneous or FH-mediated decay was monitored by further incubation for 32 min at 25◦C with buffer alone (–) or added with 2,640 ng/ml FH (+), respectively. (D) C3bBb(Mg2+) was formed in the presence of IgGs from a healthy subject (CTR-IgGs) for 12 min at 25◦C. The complexes were allowed to decay for 2, 8, 32, and 60 min in the absence (–) or in the presence of 2,640 ng/ml FH (+). The percentage of residual Bb band (visualized by an anti-FB antibody) was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of each Bb band after decay and the density of the corresponding baseline Bb band before decay x 100 and results are reported in the bottom graphs. Results of a representative microplate/WB experiment of n = 3 for each sample are shown.

FIGURE 2 | Effect of patient and control IgGs on decay of the C3bBb(Mg2+) C3 convertase by microplate/Western blot (WB) assay. (A) Experimental design (protocol 1) (B-C) Representative images of the assay. C3bBb(Mg2+) complexes were formed by incubating at 25◦C for 12 min C3b-coated wells with 1,000 ng/ml FB, 10 ng/ml FD, 100µg/ml IgGs purified from patients or healthy controls and 10 mM MgCl2 in the absence (–, baseline) or in the presence of 2,640 ng/ml FH. Spontaneous or FH-mediated decay of the complexes was monitored by further incubation of C3bBb(Mg2+) formed in the absence of FH for 32 min at 25◦C with buffer alone (decay –) or 2,640 ng/ml FH (decay +), respectively. The percentage of residual Bb band (visualized by an anti-FB antibody) was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of each Bb band after decay and the corresponding baseline Bb band density before decay x 100 and results are reported in the bottom graphs. The membranes were then incubated with an anti-FH antibody and FH band could be visualized at 150 KDa (top). Results of a representative microplate/WB experiment of n = 3 for each sample are shown.


sPCSA<sup>+</sup> 100% 100% 92% 1.000 100% 88% 100% 0.571 FH-PCSA<sup>+</sup> 100% 78% 85% 1.000 89% 75% 89% 0.653

TABLE 3A | Prevalence of patients with C3NeF IgGs with C3 convertase stabilizing activity (CSA) against spontaneous (s) or FH-mediated (FH) decay in the absence (sCSA, FH-CSA) or in the presence of properdin, (sPCSA, FH-PCSA) among histology groups or clusters.

Bold characters indicate statistically significant values.

TABLE 3B | Prevalence of patients with C3NeF IgGs with properdin (P) independent (CSA+/PCSA+) or dependent (CSA–/PCSA+) C3 convertase stabilizing activity against spontaneous (s) or FH-mediated (FH) decay among histology groups or clusters.


Bold characters indicate statistically significant values.

TABLE 4 | Comparison of clinical, and biochemical parameters at onset and histology features between clusters.


EM, electron microscopy; u-protein, proteinuria; s-Creatinine, serum creatinine.

Data are expressed as mean ± SD or as % of positive patients.

\*Significantly different vs. cluster 3.

◦Significantly different vs. clusters 1 and 3.

#Significantly different vs. clusters 1 and 2.

Bold characters indicate statistically significant values.

#### Effect of Properdin on C3NeF Stabilizing Activity on C3 Convertase

Previous studies indicated that some C3NeFs need properdin to exert their stabilizing activity on C3 convertase (26, 28, 31, 38, 39).

In order to investigate the possible synergic effect of properdin (P) with C3NeF IgGs in stabilizing C3 convertase, P (at a physiological molar ratio with FB) was added during C3 convertase formation together with FB, FD and IgGs in C3bcoated wells for 12 min (protocol 3, **Figure 5A**). Thereafter, the C3bBbP complexes were allowed to dissociate for 32 min in the presence or in the absence of FH. Results were considered positive when the percentage of residual convertase after spontaneous or FH-mediated decay was ≥ 42 or ≥10 %, respectively (> mean + 2SD of results from samples with IgGs isolated from 26 healthy donors, **Supplementary Figures 2C,D**). We found that of the 17 C3NeF IgGs that did not stabilize the C3 convertase in the assay without properdin (protocol 1), 13 had stabilizing activity both on spontaneous and FHaccelerated decay of C3bBbP C3 convertase in the presence of properdin (sPCSA<sup>+</sup> and FH-PCSA+), three (P2, P9, and P16) prevented C3bBbP spontaneous decay only (sPCSA<sup>+</sup> and

plots show significantly higher C3 convertase stabilizing activities against spontaneous (sCSA) and FH-mediated (FH-CSA) decay (protocol 1) in patients of cluster 3 than cluster 1 and 2 patients (A,B). No significance difference was observed among histology groups (C,D). The boxes represent the values from the 25th to 75th percentiles. The horizontal bars are the medians. Vertical lines are the 95% confidence intervals. Empty circles are values outside the 95% confidence intervals. The dashed horizontal lines show the limit of positive values (set at >mean+2SD of results with control IgGs from 26 healthy subjects). \*\*P < 0.01.

FH-PCSA−), while one (from P29) had no effect. The latter patient carries a likely pathogenetic variant in CFB and showed low C3NeF activity in the hemolytic assay (**Table 2**). Finally, the C3NeF IgGs that stabilized C3 convertase in the absence of properdin also stabilized the C3 convertase in the assay with properdin (**Table 2** and representative images in **Figures 5B,C**). IgGs from all the C3G/IC-MPGN controls without C3NeFs were negative also in the C3 convertase stabilizing assay with properdin (**Supplementary Figures 2C,D**). The mean intra- and inter-assay CVs were 8.2 and 18.6% respectively for sPCSA, and 6.7 and 23%, respectively for FH-PCSA.

C3NeF IgGs with properdin-dependent C3 convertasestabilizing activity (sCSA−/sPCSA<sup>+</sup> and/or FH-CSA−/FH-PCSA+) were more frequently found among patients with C3GN (3 of 4 for both) or IC-MPGN (sCSA−/sPCSA+: 9 of 13 and FH-CSA−/FH-PCSA+: 8 of 13 patients) than among patients with DDD (sCSA−/sPCSA+: 3 of 9 and FH-CSA−/FH-PCSA+: 1 of 9 patients) (**Table 3B**), however, difference reached statistical significance only for stabilization against FH-mediated decay.

The difference in properdin-dependence of C3NeFs was more evident between clusters: patients who fell within clusters 1 and 2 were more likely to have properdin-dependent C3NeFs (cluster 1: sCSA−/sPCSA+: 7 of 9 and FH-CSA−/FH-PCSA+: 6 of 9 patients; cluster 2: sCSA−/sPCSA+: 6 of 8 and FH-CSA−/FH-PCSA<sup>+</sup> 5 of 8 patients) than patients in cluster 3 (sCSA−/sPCSA+: 2 of 9 and FH-CSA−/FH-PCSA+: 1 of 9 patients) (**Table 3B**).

10 ng/ml FD, and 10 mM MgCl2. The complexes formed were then incubated in the presence of 100µg/ml IgGs from patients or healthy controls without (spontaneous decay, decay –) or with (FH-mediated decay, decay +) 2,640 ng/ml FH. The C3NeF IgG C3 convertase-stabilizing activity was quantified as the ratio of the densities (in Pixel<sup>2</sup> ) of Bb bands (visualized by an anti-FB antibody) after decay in the presence of patient IgGs/ Bb band after decay in the presence of control IgGs (ratio patient Bb band/control Bb band) and results are reported in the bottom graphs. The membranes were then incubated with an anti-FH antibody and FH band could be visualized at 150 KDa (top). Results of a representative microplate/WB experiment of n = 3 for each sample are shown.

Patients carrying C3NeFs with properdin-dependent C3 convertase-stabilizing activity (sCSA−/sPCSA<sup>+</sup> or FH-CSA−/FH-PCSA+) had higher plasma sC5b-9 levels than patients carrying C3NeFs that were also active in the assay without properdin (sCSA+/sPCSA<sup>+</sup> or FH-CSA+/FH-PCSA+, **Figures 6A,B**, **Table 5** and **Supplementary Table 1**). C3d/C3 ratios were significantly higher (P < 0.01) than in healthy subjects (0.05 ± 0.02, n = 15) in the overall group of C3NeF<sup>+</sup> patients (0.24 ± 0.20, n = 23) as well as in the subgroups of patients with properdin-dependent or properdin-independent C3NeFs (**Figures 6E,F** and **Table 2**). C3 and C3d/C3 values did not significantly differ between patients with properdin dependent and properdin independent C3NeFs (**Figures 6C–F**).

In the presence of properdin, levels of spontaneous C3 convertase stabilizing activity (sPCSA) of C3NeF IgGs did not differ between histology groups or clusters (**Figures 7A,C**), while stabilizing activity against FH-mediated decay (FH-PCSA) was significantly higher for C3NeF IgGs from cluster 3 vs. clusters 1 and 2 (**Figures 7B,D**).

#### Clinical and Histology Parameters in C3G/IC-MPGN Patients According to Type of C3 Convertase Stabilizing Activity

We compared clinical parameters at onset and during follow-up and bioptic findings in the groups of patients

percentage of residual Bb band was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of each Bb band (visualized by an anti-FB antibody) after decay and the corresponding baseline Bb band density before decay x 100, and results are reported in the bottom graphs. The membranes were then incubated with an anti-FH antibody and FH band could be visualized at 150 KDa (top). Results of a representative microplate/WB experiment of n = 3 for each sample are shown.

with properdin-dependent vs. patients with properdinindependent C3 convertase-stabilizing activity. Patients with properdin-dependent activity (sCSA−/sPCSA<sup>+</sup> or FH-CSA−/FH-PCSA+) were more likely to have subendothelial deposits in kidney biopsy, and to develop nephrotic syndrome during disease course, while intramembranous highly electron dense deposits were more prevalent in patients with properdin-independent activity (sCSA+/sPCSA<sup>+</sup> or FH-CSA+/FH-PCSA+) (**Figure 8**, **Table 5** and **Supplementary Table 1**).

TABLE 5 | Clinical features, complement assessment, genetic screening and histologic features in patients classified according to the type of C3 convertase stabilizing activity (spontaneous decay).


\*Degree of mesangial proliferation, endocapillary proliferation, interstitial inflammation, interstitial fibrosis, and arteriolar sclerosis, as well as IF findings were graded using a scale of 0 to 3, including 0, trace (0.5+), 1+, 2+, and 3+. Quantitative variables are expressed as mean (±S.D.) unless otherwise specified. Serum C3, reference 90-180 mg/dl; serum C4, reference 10-40 mg/dl; plasma sC5b-9, reference ≤400 ng/ml. LPV, Likely pathogenetic variants. sCSA-/sPCSA-, patients without C3 convertase stabilizing activity; sCSA-/sPCSA+, patients with properdin-dependent C3 convertase stabilizing activity; sPCSA+/sPCSA+, patients with properdin-independent C3 convertase stabilizing activity. Bold characters indicate statistically significant values.

#### C3NeFs Do Not Prevent FH Interaction With C3 Convertase

To investigate the mechanisms by which C3NeFs stabilize the C3 convertase, we then evaluated whether IgGs from patients with C3G/IC-MPGN prevent the interaction of FH with C3 convertase. For this purpose, WB membranes of products recovered after the decay of C3bBb(Mg2+) complexes with C3NeF IgGs or control IgGs added either only during C3bBb(Mg2+) formation in the absence or in the presence of properdin (protocols 1 and 3) or only during decay (protocol 2), were labeled with an anti-FH antibody. As shown in the top of the representative WB images of **Figures 2B,C**, **4B,C**, **5B,C**, FH band could be clearly seen in all samples with no relevant difference between C3NeF IgGs and control IgGs. These results were confirmed with all the C3NeF IgGs of the 26 tested patients.

Altogether, the above results suggest that (1) C3NeF IgGs added during C3 convertase formation did not affect the interaction of FH with C3b and with nascent C3 convertase; (2) C3NeF IgGs added during decay did not compete with FH for the interaction with the C3 convertase.

### Effect of C3NeF on C3 Convertase Formation

To evaluate whether C3NeFs had an effect on C3bBb(Mg2+) formation in the presence or in the absence of FH, we analyzed the density of the Bb bands at the baseline of protocol 1 experiments, in which the IgGs were added during the C3bBb formation. In the reactions with control IgGs the amount of C3bBb(Mg2+) formed in the presence of FH was lower compared to C3bBb(Mg2+) formed in the absence of FH (ratio Bb band + FH /Bb band without FH: 0.32 ± 0.08). FH-mediated inhibitory effect on C3bBb(Mg2+) formation was significantly attenuated in the presence of C3NeF IgGs that in protocol 1 stabilized C3 convertase against FH-accelerated decay (FH-CSA<sup>+</sup> IgGs: ratio Bb band +FH /Bb band without FH: 0.58 ± 0.19, p < 0.05 vs. control IgGs).

At variance, FH-CSA<sup>−</sup> C3NeF IgGs did not affect FHmediated inhibitory effect on C3bBb(Mg2+) formation (ratio Bb band + FH /Bb band without FH: 0.32 ± 0.07).

Representative images are shown in **Figures 2B,C**.

#### Effect of C3NeF on C3bB C3 Proconvertase

To evaluate whether C3NeF IgGs stabilized the alternative pathway C3 proconvertase C3bB, we modified the microplate/WB assay, exploiting the selective stabilization properties of Mn2+. C3bB was formed on C3b-coated wells in the presence of FB and Mn2<sup>+</sup> buffer, and we detected by WB the B component of the C3bB(Mn2+) proconvertase recovered from wells. The kinetics of C3bB(Mn2+) assembly and decay and the effect of FH at physiological molar ratio with FB, are shown in **Figures 9A,B**. In these experimental conditions, no Bb band of the C3 convertase was observed either during assembly or after decay, since C3bBb(Mn2+) is highly unstable and once formed it rapidly dissociates (35).

C3bB(Mn2+) proconvertase did not spontaneously decay, as shown by the intensity of B band that did not change over 240 min incubation. FH band could be detected in the WB of the products detached from the wells after each time interval, indicating that FH binds C3bB. However, FH did not affect either C3bB(Mn2+) formation or decay (**Figures 9A,B**), suggesting that in our condition FH neither competed with FB for the interaction with C3b nor displaced FB from C3b.

We then repeated the above experiments using C3NeF IgGs from two patients (P5 and P10) selected for being strongly positive in stabilizing C3 convertase C3bBb(Mg2+). As shown in **Figures 9C,D**, C3NeF IgGs added during C3bB(Mn2+) assembly did not modify the intensity of the B band at any time point either in the absence or in the presence of FH.

Finally, we performed additional microplate/WB experiments by incubating C3b-coated wells with FB, FD in the presence of Ni2<sup>+</sup> buffer, a condition that stabilizes both C3bB and C3bBb, as demonstrated by the detection of B and Bb bands on WB after 30 min assembly (**Supplementary Figure 1D**).

The complexes were allowed to decay by incubation for different time points with buffer alone or buffer containing FH. We observed a slight spontaneous decay of C3bB(Ni2+) at 30 min, thereafter the intensity of B band remained stable (**Figure 10A**). FH did not affect C3bB(Ni2+) formation or decay (**Figures 10A,B**), confirming the results obtained in the Mn2<sup>+</sup> ion selective conditions. At variance, FH reduced C3bBb(Ni2+) convertase formation and accelerated the decay, as shown by progressive disappearance of the Bb band on WB (**Figure 10A**).

Next, we assessed the effect of C3NeFs on the formation and decay of C3bB(Ni2+) and C3bBb(Ni2+). For this purpose, IgGs from P5 or P10 or from a healthy control were added to C3b-coated wells with FB and FD in the presence or in the absence of FH, followed by 30 min of decay. As observed in Mn2<sup>+</sup> experiments, C3NeFs from P5 and P10 did not affect C3bB(Ni2+) assembly and decay substantially (**Figures 10B,C**). At variance, P5 and P10 IgGs strongly limited FH-mediated C3bBb(Ni2+) decay, and attenuated FH-mediated inhibitory effect on C3bBb(Mg2+) formation, as compared with control IgG (**Figures 10B,C**).

#### DISCUSSION

To characterize the diverse mechanisms through which the heterogeneous autoantibodies C3NeFs influence the function and regulation of the alternative pathway C3 convertase in primary IC-MPGN and C3G, we established a WB-based assay that monitors the formation and the spontaneous and regulated decay of the C3bBb convertase, through specific detection of the Bb band. The assay turned out to be very flexible and allowed us to study the effect of C3NeFs on spontaneous and/or FHaccelerated C3bBb decay, disclosing a subgroup of properdindependent C3NeFs and investigating the mechanisms underlying C3NeF activity. Finding that IgGs from all C3G/IC-MPGN controls without C3NeFs by hemolytic assay did not have C3 convertase stabilizing activity, supports the specificity of the new WB-based assays for analyzing C3NeFs.

Several methods have been reported in the literature for the evaluation of C3NeFs. Classic and modified hemolytic assays

healthy subjects). \*\*P < 0.01.

(26–28, 31, 40), performed on the surface of C3b-coated sheep erythrocytes are laborious and require expertise. In addition the read-out, based on the release of hemoglobin from lysed sheep erythrocytes, is an indirect measure of the terminal complement activation by residual C3 convertase sites on the cell surface and does not make it possible to directly evaluate the molecular mechanisms through which C3NeFs stabilize the C3bBb complex.

Functional molecular assays of C3NeFs have recently been reported. Surface Plasmon resonance with flowing C3NeFs, FB and FD over C3b immobilized on the chip allowed real-time analysis of C3 convertase stabilization (31), but this method is not universally available and required purified patients' IgG at high concentration.

Other, simpler ELISA-based assays have been described, in which C3b-coated microplate wells are incubated with FB plus FD in Mg2<sup>+</sup> buffers, followed by detection with anti-FB antibodies. These assays detect the capacity of C3NeFs to stabilize the C3 convertase in the absence or in the presence of properdin (31). To analyze the effect C3NeFs on C3 convertase

dissociation and on its accelerated decay mediated by FH and other complement regulators (31) the ELISA assays are done in the presence of Ni2+, which prolongs the half-life of C3bBb. However, this experimental condition does not distinguish between the alternative pathway C3 convertase and C3 proconvertase, because through WB analysis of the ELISA products of the reaction with Ni2<sup>+</sup> in the presence of FD, here we visualized both the Bb and the B bands of C3bBb and C3bB complexes. The latter finding is consistent with earlier evidence that Ni2<sup>+</sup> strongly stabilizes the proconvertase C3bB (41).

To overcome these issues, we took a step forward and modified the microplate assay, combining the selectivity of Mg2<sup>+</sup> for C3bBb with the high sensitivity of the WB technique. In the absence of exogenous properdin, the Mg2<sup>+</sup> microplate/WB assay revealed C3NeF IgGs capable of stabilizing C3bBb in 11 of 26 IC-MPGN or C3G patients. The finding that 9 of these C3NeFs had dual action, preventing both spontaneous and FHaccelerated C3 convertase decay, confirms published data that the large majority of C3NeFs also impair the decay-accelerating activity of regulatory proteins (25, 31). Since C3NeF IgGs were added during the first step of C3 convertase formation and unbound IgGs were washed out before the decay, the positivity in the microplate/WB test would suggest that C3 convertasestabilizing IgGs rapidly and irreversibly bound to nascent C3bBb complexes. Essentially the same results were obtained when the assay was repeated adding C3NeF IgGs only during decay, indicating that C3 convertase-stabilizing IgGs also efficiently interacted with preformed C3bBb. One C3NeF IgG that was negative in the first protocol turned positive in the second setting, suggesting that it may act through reversible interaction with C3 convertase.

It is relevant that in the presence of exogenous properdin all but one patients' IgGs had C3 convertase-stabilizing activity, including IgGs negative in the assay without exogenous properdin, which confirms previous evidence that some C3NeFs require additional stability conferred by properdin to block C3 convertase decay (26, 28, 31). Properdin is the only known positive regulator of the alternative pathway of complement, and acts by extending the half-life of the C3 convertase 10 fold (42). It also facilitates the switch of the C3 convertase to the C5 convertase and stabilizes the C5 convertase complex (43, 44). The finding that C3NeFs with properdin-dependent C3 convertase-stabilizing activity associated with higher plasma sC5b-9 levels than C3NeFs with properdin-independent activity may be explained considering previous studies with hemolytic assays modified to interrogate C5 convertase stabilization, which demonstrated that in the presence of properdin a subset of C3NeFs enhances C5 cleavage and terminal pathway activation (28, 29, 45). In a cohort of patients with C3G Marinozzi et al. (28) found that plasma sC5b-9 levels correlated with C5 convertasestabilizing activity of C3NeFs, which is consistent with our data. In addition, Paixao-Cavalcante et al. (31) showed that properdin-dependent C3NeFs enhanced C5 convertase activity while properdin-independent C3NeFs had no effect on C5 convertase activity. The former C3NeFs efficiently cleaved C3, but were "weak" C3 convertase binders and required the addition of properdin for the detection of their C3 convertase stabilizing activity (31).

We speculate that C3NeFs, which in the microplate/WB assay presented here were active in the assay without added properdin, recognize epitope(s) in the C3 convertase that are absent and/or masked in the C5 convertase complex, and cause C3-restricted

FIGURE 9 | Formation and decay of C3bB(Mn2+) C3 proconvertase in the absence or in the presence of C3NeF IgGs, by microplate/WB assay. (A) Time course of C3bB(Mn2+) C3 proconvertase formation. The complexes were obtained incubating C3b-coated wells at 37◦C for 1, 2, 4, and 8 h with 1,000 ng/ml FB and 2 mM MnCl<sup>2</sup> in the absence (-) or in the presence (+) of 2,640 ng/ml FH. The amount of C3bB formed was calculated as the density of B band (93 KDa), and reported in the bottom graph as Pixel<sup>2</sup> \* 10<sup>6</sup> . (B–D) Time course of C3bB(Mn2+) C3 proconvertase spontaneous and FH-mediated decay. C3bB(Mn2+) complexes formed in 2 h at 37◦C in the absence (B) or in the presence of C3NeF IgGs purified from patients P5 (C) and P10 (D), were further incubated with buffer alone (decay –) or with buffer added with 2,640 ng/ml FH (decay +), respectively, for 30, 60, 120, and 240 min at 37◦C The percentage of residual B band was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of each B band after decay and the corresponding baseline B band density before decay x 100 and results are reported in the bottom graphs. The membranes were then incubated with an anti-FH antibody and FH band could be visualized at 150 KDa (top). Results of a representative microplate/WB experiment of n = 3 are shown.

complement activation in vivo, as reflected by low C3 and normal sC5b-9 levels. At variance properdin-dependent C3NeF would bind different epitopes that are present both in the C3 and the C5 convertase, resulting in vivo in complement activation

(Continued)

FIGURE 10 | buffer alone (decay –) or buffer added with 2,640 ng/ml FH (decay <sup>+</sup>). (B-C) C3bB(Ni2+) and C3bBb(Ni2+) were formed for 30 min at 37◦C in NiCl<sup>2</sup> buffer without (baseline –) or with 2,640 ng/ml FH (+) in the presence of IgGs purified from an healthy subjects (CTR-IgGs, B) or from patients P5 and P10 (C, P5-IgGs and P10-IgGs). The complexes formed in the absence of FH were then allowed to decay for 30 min (C) at 37◦C with buffer alone (decay –) or buffer added with 2,640 ng/ml FH (decay +). The percentage of residual B (93 KDa) or Bb (60 KDa) bands was calculated as the ratio of the densities (in Pixel<sup>2</sup> ) of each B or Bb band after decay and the corresponding baseline B or Bb band density before decay x 100 and results are reported in the bottom graphs. The membranes were then incubated with an anti-FH antibody and FH band could be visualized at 150 KDa (top). Results of a representative microplate/WB experiment of n = 3 are shown.

until the terminal pathway with low C3 and high sC5b-9 levels.

Remarkably, we found that 6 of the 9 patients with DDD carried C3NeF IgGs that exerted C3 convertase-stabilizing activity also without properdin addition, and had normal sC5b-9 levels. In line with our data, using a modified hemolytic assay Zhang et al (26) reported that of 32 patients with DDD, 88% of those with C3NeF activity were positive in the assay without properdin. In another report (28), 63% of C3NeF-positive DDD patients had C3NeFs targeting the C3 convertase but not the C5 convertase.

As for properdin-dependent C3NeFs, we identified them mostly in patients with either C3GN [3 of 4] or primary IC-MPGN (9 of 13) and high plasma levels of sC5b-9. The association of properdin-dependent C3NeFs with C3GN and terminal complement pathway activation in vivo is consistent with recent findings that in most (67%) C3GN patients, C3NeFs target both the C3 and C5 convertases (28), and that circulating markers of terminal pathway activity are altered more in C3GN than in DDD (46). Another study documented low properdin levels in C3NeF negative C3G patients and serum properdin consumption was a marker of C5 convertase dysregulation, indeed properdin levels inversely correlated with sC5b-9 in plasma (47). Properdin levels were normal in C3G patients with C3NeF (47).

To the best of our knowledge, no other studies have formally explored the mechanisms of action of C3NeFs in IC-MPGN yet. Old reports proposed an association between properdin-dependent C3NeFs and MPGN type I and III (38, 39), but they antedated the current IF-based classification. Finding comparable properdin-dependency and comparable levels of C3 convertase-stabilizing activity in C3GN and primary IC-MPGN patients here would suggest the existence of a common pathogenetic mechanism that causes complement alternative pathway dysregulation in the two histology groups. We hypothesize that in IC-MPGN patients, disease was initiated by an as yet unidentified trigger (11, 12), causing glomerular immune-complex deposition and possibly the formation of C3NeF IgGs, the latter resulting in a switch from acute classical complement pathway-driven disease to chronic alternative pathway-driven disease. Although the IF-based classification of C3G/IC-MPGN is an advance toward an etiology-based approach of these diseases, it is likely too simplistic to rely only on C3 dominance for distinguishing patients with abnormalities of complement activation and those with immune complexmediated disease, as suggested by a recent review (48). Our results highlight the need of a more advanced classification, based on the underlying pathogenetic patterns of complement activation. Along this line are data that the stratification of patients according to C3NeF activity was more clear-cut when we compared patients grouped on the basis of cluster analysis (15). Thus, patients in cluster 3, characterized by highly electron-dense intramembranous deposits, low C3 levels and mostly normal sC5b-9 levels, had a higher prevalence of C3NeFs stabilizing C3 convertase and higher stabilizing activity in the assay without properdin than patients in clusters 1 and 2. Instead, about 70% of patients in clusters 1 and 2, with a high prevalence of subendothelial, subepithelial and mesangial deposits, had properdin-dependent C3NeFs, which fits with low C3 levels and high sC5b-9 levels measured in vivo [(15) and present data]. Altogether these findings highlight two different mechanisms of complement dysregulation by C3NeFs, with different degrees of unchecked activity of the C3 and C5 convertases resulting in different complement profiles and different patterns of glomerular injury. Thus, prevalent dysregulation of the C3 convertase by properdin-independent C3NeFs may result in the constant, slow accumulation of C3 breakdown products in the GBM, generating denser intramembranous deposits. In patients with properdin-dependent C3NeFs, both C3 and C5 convertase dysregulation is high, and terminal complement activation products are sequestered in the glomerulus and participate in the formation of more amorphous deposits along the glomerular membrane layers. The significantly higher prevalence of nephrotic syndrome during disease course in patients with properdin-dependent C3NeFs supports the role of a pathogenetic mechanism involving the terminal complement pathway in altering glomerular permselectivity properties.

Another important observation from this study is that the stabilizing effect of both properdin-independent and properdindependent C3NeFs against FH-mediated C3 convertase decay is not due to competition between C3NeFs and FH, because in the microplate/WB assays the binding of FH to either nascent C3bBb, during the formation step, or to preformed C3bBb during the decay step, was not reduced in the presence of C3NeFs. This finding indicates that C3NeFs and FH bind different molecular domains in the C3bBb complexes, and may be relevant for further studies aimed at identifying the neopitope(s) recognized by the different C3NeFs, an issue that remains ill-defined (49). Considering that FH binds C3b (50), and is very efficient in Bb displacement from C3b (21), it is possible to speculate that C3NeFs could be directed against epitopes on Bb, as previously proposed (51).

Notably, the C3NeFs that stabilized C3 convertase against FHaccelerated decay also partially but significantly prevented the inhibitory effect of FH on C3 convertase formation. This effect could not be attributed to an antagonistic effect of C3NeFs with FH, since C3NeFs do not interact with the single C3 convertase component proteins, as demonstrated by earlier studies with C3b, FB, and Bb immobilized on microwell plates (31), nor do they prevent FH binding in the assays reported here. We hypothesize that C3NeFs rapidly stabilized C3 convertase to prevent the FH-induced dissociation of newly formed C3 convertase complexes.

Whether C3NeFs can have any effect on the complement alternative pathway C3 proconvertase enzyme is not known. To address this issue, two of the C3NeFs that exhibited the highest stabilizing activity on C3 convertase were tested in the microplate/WB assay with Mn2<sup>+</sup> buffer, designed to selectively interrogate C3 proconvertase stabilization. C3bB complexes did not undergo FH-accelerated decay, which is consistent with other data in the literature (52, 53), and the addition of C3NeFs did not alter the amount of residual C3bB complexes. The results were not restricted to experiments with the Mn2<sup>+</sup> buffer, because when we repeated the test in Ni2<sup>+</sup> buffer, no effects of either FH or C3NeFs were observed on C3bB proconvertase decay, while in the same conditions FH dissociated the C3bBb C3 convertase and C3NeFs antagonized the FH-mediated decay-accelerating effect.

Intriguingly, we observed that independently of the ions (Mn2<sup>+</sup> or Ni2+) used in the reaction buffer, C3 proconvertase complexes formed in the same amounts in the absence and in the presence of FH, suggesting that FH did not compete enough with FB for binding to C3b as to prevent C3bB assembly. A plausible explanation for these results derives from biophysical studies showing that the binding affinity between C3b and FH molecules is lower compared with the affinity between C3b and FB (35, 50). FH binding to negatively charged proteoglycans in cell glycocalyx and on cell surfaces shifts FH from a latent to an active conformation, which significantly increases the affinity for C3b (54). It is possible that in a cell-based context proteoglycans could influence the interactions of C3b with FB or FH in favor of the latter (55).

In summary, our study provides a mechanistic approach to detecting and characterizing C3NeFs in patients with C3G and IC-MPGN and identifies two distinct pathogenetic mechanisms through which C3NeFs may cause complement dysregulation and glomerular disease, confirming at molecular level previous studies with hemolytic and ELISA-based assays (28, 31). C3NeFs that stabilize C3 convertase in the assay without properdin associate with prevalent dysregulation of the C3 convertase and the formation of highly electron-dense deposits in the GBM, while properdin-dependent C3NeFs result in both C3 and C5 convertase dysregulation, high plasma sC5b-9 levels and more amorphous and broadly distributed glomerular deposits.

#### REFERENCES


The identification of different types of C3NeFs with distinct functional specificities may have an impact on patient management. Our results suggest that a therapy inhibiting C5 activation (56–58) or targeting properdin (59) could potentially benefit patients with properdin-dependent C3NeFs. Patients with properdin-independent C3NeFs might benefit from emerging drugs, such as FD or FB inhibitors that target the C3 convertase of the complement alternative pathway (60).

Finally, results of mechanistic studies highlighted new information, which will be relevant to further studies aimed to clarify the way by which C3NeFs induce complement dysregulation: (1) most C3NeFs bind strongly and irreversibly to C3 convertase; (2) C3NeFs and FH recognize different epitopes in C3 convertase; (3) C3NeFs bind rapidly to C3 convertase and antagonize the decay-accelerating activity of FH on newly formed complexes; (4) C3NeFs do not affect the formation and stability of the C3 proconvertase.

#### AUTHOR CONTRIBUTIONS

RD, MN, and GR designed research, interpreted data, and wrote the paper. RD, PP, RP, and EV performed the research and analyzed the data. PI collected clinical data and performed cluster analysis. AB analyzed the data and critically revised the manuscript.

#### ACKNOWLEDGMENTS

The authors thank Drs Caterina Mele and Matteo Breno and Marta Alberti for next generation sequencing, Dr Veronique Fremeaux-Bacchi for C3NeF evaluation by hemolytic assay, Kerstin Mierke for editing the manuscript and Manuela Passera for secretarial assistance. This work was partially supported by Progetto DDD Onlus—Associazione per la lotta alla DDD, Milan, Italy. RP is the recipient of a research contract from Progetto DDD Onlus—Associazione per la lotta alla DDD. EV is the recipient of a fellowship from Fondazione Aiuti per la Ricerca sulle Malattie Rare ARMR ONLUS (Bergamo, Italy). The funding sources had no role in study design, or in the collection, analysis and interpretation of data, nor in the writing of the report or in the decision to submit the paper for publication.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02329/full#supplementary-material


**Conflict of Interest Statement:** MN has received honoraria from Alexion Pharmaceuticals for giving lectures, and for participating in advisory boards and research grants from Omeros, Alnylam, and Chemocentryx. GR has consultancy agreements with AbbVie<sup>∗</sup> , Alexion Pharmaceuticals<sup>∗</sup> , Bayer Healthcare<sup>∗</sup> , Reata Pharmaceuticals<sup>∗</sup> , Novartis Pharma<sup>∗</sup> , AstraZeneca<sup>∗</sup> , Otsuka Pharmaceutical Europe<sup>∗</sup> , Concert Pharmaceuticals<sup>∗</sup> .

<sup>∗</sup>No personal remuneration is accepted, compensation is paid to his institution for research and educational activities. None of these activities have had any influence on the results or interpretations in this article.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Donadelli, Pulieri, Piras, Iatropoulos, Valoti, Benigni, Remuzzi and Noris. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# complement activation During ischemia/reperfusion injury induces Pericyte-to-Myofibroblast Transdifferentiation regulating Peritubular capillary lumen reduction Through perK signaling

*Giuseppe Castellano1 \*† , Rossana Franzin1†, Alessandra Stasi1 , Chiara Divella1 , Fabio Sallustio1,2, Paola Pontrelli1 , Giuseppe Lucarelli3 , Michele Battaglia3 , Francesco Staffieri4 , Antonio Crovace4 , Giovanni Stallone5 , Marc Seelen6 , Mohamed R. Daha6,7, Giuseppe Grandaliano5‡ and Loreto Gesualdo1‡*

*1Nephrology, Dialysis and Transplantation Unit, Department of Emergency and Organ Transplantation, University of Bari Aldo Moro, Bari, Italy, 2 Department of Basic Medical Sciences, Neuroscience and Sense Organs, University of Bari Aldo Moro, Bari, Italy, 3Urology, Andrology and Renal Transplantation Unit, Department of Emergency and Organ Transplantation, University of Bari Aldo Moro, Bari, Italy, 4Veterinary Surgery Unit, Department of Emergency and Organ Transplantation, University of Bari Aldo Moro, Bari, Italy, 5Nephrology, Dialysis and Transplantation Unit, Department of Medical and Surgical Sciences, University of Foggia, Foggia, Italy, 6Division of Nephrology, Department of Internal Medicine, University of Groningen, University Medical Centre Groningen, Groningen, Netherlands, 7 Department of Nephrology, Leiden University Medical Centre, Leiden, Netherlands*

Pericytes are one of the principal sources of scar-forming myofibroblasts in chronic kidneys disease. However, the modulation of pericyte-to-myofibroblast transdifferentiation (PMT) in the early phases of acute kidney injury is poorly understood. Here, we investigated the role of complement in inducing PMT after transplantation. Using a swine model of renal ischemia/reperfusion (I/R) injury, we found the occurrence of PMT after 24 h of I/R injury as demonstrated by reduction of PDGFRβ+/NG2+ cells with increase in myofibroblasts marker αSMA. In addition, PMT was associated with significant reduction in peritubular capillary luminal diameter. Treatment by C1-inhibitor (C1-INH) significantly preserved the phenotype of pericytes maintaining microvascular density and capillary lumen area at tubulointerstitial level. *In vitro*, C5a transdifferentiated human pericytes in myofibroblasts, with increased αSMA expression in stress fibers, collagen I production, and decreased antifibrotic protein Id2. The C5a-induced PMT was driven by extracellular signal-regulated kinases phosphorylation leading to increase in collagen I release that required both non-canonical and canonical TGFβ pathways. These results showed that pericytes are a pivotal target of complement activation leading to a profibrotic maladaptive cellular response. Our studies suggest that C1-INH may be a potential therapeutic strategy to counteract the development of PMT and capillary lumen reduction in I/R injury.

#### Keywords: complement system, pericytes, ischemia–reperfusion, tubulointerstitial fibrosis, capillary rarefaction, C1-inhibitor, C5a

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Trent M. Woodruff, The University of Queensland, Australia Lourdes Isaac, Universidade de São Paulo, Brazil Marina Noris, Istituto Di Ricerche Farmacologiche Mario Negri, Italy*

#### *\*Correspondence:*

*Giuseppe Castellano giuseppe.castellano@uniba.it*

*† These authors have contributed equally to this work.*

*‡ These authors have shared senior authorship.*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 22 December 2017 Accepted: 23 April 2018 Published: 23 May 2018*

#### *Citation:*

*Castellano G, Franzin R, Stasi A, Divella C, Sallustio F, Pontrelli P, Lucarelli G, Battaglia M, Staffieri F, Crovace A, Stallone G, Seelen M, Daha MR, Grandaliano G and Gesualdo L (2018) Complement Activation During Ischemia/ Reperfusion Injury Induces Pericyte-to-Myofibroblast Transdifferentiation Regulating Peritubular Capillary Lumen Reduction Through pERK Signaling. Front. Immunol. 9:1002. doi: 10.3389/fimmu.2018.01002*

**Abbreviations:** I/R, ischemia/reperfusion; PMT, pericyte-to-myofibroblast transdifferentiation; C1-INH, C1-inhibitor; CTRL, control group; TGF-β1, transforming growth factor beta 1; PDGFRβ, beta-type platelet-derived growth factor receptor; NG2, neuronal glial antigen 2; Id2, inhibitor of DNA binding 2; ERK, extracellular signal-regulated kinases; SMAD, small mother against decapentaplegic; MMP9, matrix metallopeptidase 9; ADAMTS1, a disintegrin and metalloproteinase with thrombospondin motifs 1; CTGF, connective tissue growth factor; MAPK, mitogen-activated protein kinase.

# INTRODUCTION

Ischemia/reperfusion (I/R) injury remains an unavoidable consequence after renal transplantation and the principal cause of delay graft function (DGF) (1). After brain death, the decreased blood flow induces a persistent rarefaction in peritubular capillaries (2, 3), whereas the following reperfusion exacerbates the pro-inflammatory response by activation of complement and coagulation (4). Pharmacologic treatments to prevent graft deterioration after I/R are currently lacking. Recently, genetic fate-mapping studies have identified pericytes as the major source of scar-forming myofibroblasts during progressive chronic kidney disease (5–8). Pericytes are mesenchymal-derived cells embedded in the capillary basement membrane and in direct contact with endothelial cells (9). Pericytes contribute to microvessel stability and show regeneration potential; renal pericytes regulate cortical and medullary flow by contracting or dilating in response to various stimuli released by the neighboring endothelial and tubular cells (10). Interestingly, data on cerebral ischemia showed that pericytes died "*in rigor*," causing an irreversible constriction of capillaries that exacerbates the hypoxia (11, 12). Renal pericytes are PDGFRβ+/NG2+ cells that during their abnormal transdifferentiation in myofibroblasts upregulated the pro-fibrotic marker αSMA. The physiological role of the receptor tyrosin kinase PDGFRβ is to bind the platelet-derived growth factor B (PDGF-B) released by endothelial cells. The PDGFRβ signaling promoted the pericytes activation, migration, and the recruitment to the vascular wall of newly formed blood vessels. Nerve/glial antigen 2 (NG2) is a proteoglycan associated with pericytes during vascular morphogenesis (13).

Complement plays a pivotal role in renal I/R injury mediating tissue damage and amplifying innate and adaptive immune response (14–17). C1-esterase Inhibitor (C1-INH) blocks complement activation of the classical, lectin (14, 18, 19), and alternative pathways (20–22). Currently, C1-INH is used as treatment for hereditary angioedema (23), but several studies are evaluating the therapeutic potential in renal transplantation (24–26). In previous studies, we demonstrated that C1-INH is able to prevent the C5b-9 deposition along peritubular capillaries, limiting endothelial dysfunction and renal fibrosis during I/R (18, 27). The aim of our study was to investigate the involvement of complement in pericyte activation in the early phase of I/R injury.

# MATERIALS AND METHODS

#### Animal Models

Animal studies were carried out under protocol approved by Ethical Committee of the Italian Ministry of Health. Briefly, I/R was induced in pig by clamping the renal artery for 30 min followed by reperfusion, as described previously (18). A biopsy was performed before ischemia (T0). Pigs were divided into two groups: control (CTRL, *n* = 5, vehicle infused) and C1 Inhibitor treated group (C1-INH, *n* = 5). Five minutes before the beginning of the reperfusion, rhC1-INH was injected in the ear vein (500 U/kg). Biopsies were performed at 15, 30, and 60 min and 24 h after reperfusion. All animals were sacrificed 24 h after the procedure. Controlateral kidney was not removed for ethical concerns. A mouse model of renal bilateral I/R was performed in C5aR1<sup>−</sup>/<sup>−</sup> mouse with C57BL/6 backgrounds, as previously described (28).

#### Immunohistochemistry

Renal sections underwent deparaffination and heat-mediated antigen retrieval (citrate buffer, pH = 6.00) as previously described (18). For caspase3 and Ki67 detection, sections were permeabilized with Triton 0.25% for 5 min, then blocked by Protein Block Solution (DakoCytomation, USA) for 10 min. Incubation was performed with antibodies against: Caspase-3 (Novus Biologicals, Abingdon Science Park, UK), PDGFRβ (Abcam, Cambridge UK), Ki-67 (Novus Biologicals), and detected by the Peroxidase/DAB Dako Real EnVision Detection System (Dako, Glostrup, Denmark). The peroxidase reaction was shown by a brown precipitate, counterstained with Mayers hematoxylin (blue). Negative controls were prepared by incubation with a control irrelevant antibody. Images were scanned by Aperio ScanScope CS2 device and signals were analyzed with the ImageScope V12.1.0.5029 (Aperio Technologies, Vista, CA, USA).

#### Analysis of Peritubular Capillaries Area

The peritubular capillaries area was calculated by Image J software. The cortical area of the entire biopsy acquired by Aperio ScanScope was analyzed in a stepwise fashion as a series of 10 consecutive fields, avoiding the arterioles, venules, and capillaries, which has a diameter upper than 50 µm. Values from all consecutive images for each biopsy were averaged.

#### Cell Culture

Human placental-derived pericytes (PromoCell, Heidelberg, Germany) were grown in Serum Free Pericyte Growth Medium (PromoCell) at 5% CO2 and 37°C. Once they have reached the 70%, confluence cells were stimulated with human recombinant C5a (Biovision, San Francisco, CA, USA) at 10<sup>−</sup><sup>7</sup> M and human recombinant TGFβ-1 (10 ng/ml, Biovision). All experiments were performed at early P3–P5 passages. For pERK inhibition, cells were pretreated with SC1 (Pluripotin, Abcam) at 1–3–5 µM for 6–24 h, the cells were stimulated by C5a for indicated times. For C5aR inhibition assay, mouse monoclonal anti-C5aR (Abcam) was preincubated (1:10) for 1 h before the C5a exposition.

#### Confocal Laser Scanning Microscopy

Renal sections and cultured pericytes were stained or double stained for αSMA (Santa Cruz Biotechnologies), PDGFRβ, NG2 (Abcam), C3 (HycultBiotech), and C5b-9 (Dako). For C3 and C5b-9 stainings, frozen kidney slides were used. For each experiment, 5 × 104 cells were seeded on a cover slip, grown to 70% confluence, and fixed in 3.7% paraformaldehyde for 5 min. After blocking, slides were incubated with primary antibodies, overnight at 4°C and with secondary antibodies (Alexa Fluor, Molecular Probes, Eugene, OR, USA). TO-PRO-3 was used to counterstain nuclei. Negative controls were prepared by isotype control antibody. Image acquisition was performed with confocal microscope Leica TCS SP2 (Leica, Wetzlar, Germany).

#### FACS Analysis

After incubations, cells were washed, detached by ice cold PBS-EDTA, permeabilized by Intraprep reagents (Instrumentation Lab), then incubated with FCR blocking reagent (Miltenyi Biotec) for 10 min at RT and with APC-conjugated anti-PDGFRβ (LSBio), FITC-conjugated anti-collagen I (Millipore, Millimarck) or mouse monoclonal anti-C5aR unconjugated (Abcam) for 20 min at RT in the dark. After washing, cells were re-suspended in FACS buffer. For apoptosis analysis, 5 × 105 cells for each condition were washed with cold PBS 1× and double-stained with FITC-conjugated Annexin V/Propidium Iodide. Data were obtained using a FC500 flow cytometer (Beckmann Coulter) and analyzed by Kaluza software. Three independent experiments were performed. The area of positivity was determined using an isotype-matched mAb.

#### MTT Assay

Cultured pericytes proliferation was measured by MTT Cell Proliferation Assay Kit, according the manufacturer instructions (Sigma-Aldrich). Briefly, 2 × 104 cells/well were seeded in a 96-well plate, and then cells were treated with C5a, TGFβ1 (as indicated), PDGFBB (10 ng/ml) for 24 h.

#### RNA Extraction and qPCR Analysis

RNA from pericytes was extracted using the miRNeasy Kit (Qiagen), 500 ng of total RNA was retrotranscribed with QuantiTect Kit (Qiagen). qPCR was carried out with SsoAdvanced™ Universal SYBR® Green Supermix (Biorad) and the Light Cycler@96 (Roche). Primer list sequence in **Table 1**.

#### Western Blot

Protein lysates were homogenized by RIPA buffer with phosphatase and protease inhibitors. Proteins (30 µg) were separated in 4–15% polyacrylamide gel and then transferred to PVDF membrane (0.2 mM) by Trans-Blot Turbo (BioRad, Hercules, CA, USA). After blocking in BSA at 5%, the membranes were incubated overnight with the following primary antibodies: pSMAD2/3 (Abcam), SMAD 2/3, pERK, extracellular signalregulated kinases (ERK), Id2, matrix metallopeptidase (MMP9) (Santa Cruz Biotechnology, Inc.) and then with secondary


antibody (hrp-conjugated, Santa Cruz). The same membrane was probed with mouse monoclonal anti-βactin antibody (1:20,000; Sigma). The electrochemiluminescence system was used to detect the antibody binding (Amersham, UK). The chemiluminescent signal was acquired by Chemidoc and quantified using Image J software.

#### Statistical Analysis

Graphs were displayed using GraphPad Prism Software 5. Data were expressed as median ± interquartile range (IQR) and compared with a Mann–Whitney test for tissue immunostainings. For FACS, qPCR, MTT, and WB data were expressed as the mean ± SD. Statistical analysis was assessed using unpaired Student's *t*-test. A *p-*value of <0.05 was considered significant.

# RESULTS

#### C1-INH Treatment Preserved Pericytes Phenotype After Renal I/R Injury

To investigate the possible dysfunction of renal pericytes during I/R injury, biopsies were analyzed for the expression of PDGFRβ, a marker of pericyte (**Figure 1**). Under normal condition (T0, **Figure 1A**), PDGFRβ+ cells were detected in interstitial peritubular capillaries (**Figure 1A**, zoom1), in arterioles (**Figure 1A**, zoom2), mesangial cells, and Bowman's capsule (**Figure 1A**, zoom3). I/R injury caused a significant decrease in PDGFRβ expression of pericytes in peritubular capillaries, a process that in the CTRL group began after 30 min and persisted until 24 h after reperfusion (**Figures 1B,D,F**). Treatment with C1-INH was unable to prevent early PDGFRβ downregulation (**Figures 1C,E**) but gave a significant preservation of peritubular PDGFRβ expression at 24 h after I/R injury (**Figure 1G** compared to **Figure 1F**). Notably, the PDGFRβ expression of mesangial cells was not significantly down regulated.

Next, we used PDGFRβ and NG2 co-expression to specifically label pericytes (**Figure 2**). In swine kidney, pericytes markers were localized in the interstitial peritubular capillaries (**Figure 2A**). We found that all the perivascular NG2<sup>+</sup> cells were PDGFRβ+; on the contrary, we found that mesangial cells were PDGFRβ+/NG2−. After 24 h of I/R, the total number of PDGFRβ/NG2 double positive pericytes significantly decreased; in accordance with **Figure 1F**, PDGFRβ+/NG2<sup>+</sup> cells were barely detectable in the interstitial peritubular capillaries after I/R (**Figures 2B,E**). In contrast, C1-INH treated pigs were protected in pericytes phenotype as shown by a significant recovery in the number of PDGFRβ+/NG2<sup>+</sup> cells in the interstitial peritubular capillaries regions. Interestingly, after 24 h from the C1-INH treatment, other PDGFRβ+/NG2<sup>−</sup> cells (i.e., vascular smooth muscle cells) were protected from the PDGFRβ downregulation.

To further confirm the complement deposition at peritubular capillary level, we investigated the co-localization of PDGFRβ with C3 and C5b-9 after 30 min of reperfusion (Figures S1 and S4 in Supplementary Material) on frozen renal tissue. We found C3 and C5b-9 deposition around peritubular regions after 30 min; interestingly, C1-INH significantly counteracted this

complement activation (Figures S1C–E and S4 in Supplementary Material).

# I/R Injury Did Not Affect Pericytes Viability *In Vivo*

To study the possible occurrence of pericyte apoptosis during I/R injury as observed in cerebral ischemia (12), we stained 30 min, 60 min, and 24 h serial sections for PDGFRβ and for the active form of Caspase 3 (Casp3) (**Figure 3**). After 30 and 60 min from reperfusion, no PDGRFβ/Casp3 double positive cells were detected in peritubular capillaries (**Figures 3A,B** arrowheads). As previously shown (18), apoptosis occurred predominantly in tubular epithelial cells (**Figures 3A,B** *right,* brown nuclei) (Figure S1A in Supplementary Material), and not in pericytes that were PDGFRβ+/Casp3<sup>−</sup>. In addition, we investigated whether renal pericytes proliferation could be detected, labeling serial sections for PDGFRβ and Ki-67, an antigen that marked nuclei in G1, S, and G2 cell cycle phases (29). Remarkably, 24 h after I/R injury, no Ki-67 positive cells could be found in interstitial peritubular capillaries (**Figures 3C,D**). In conclusion, in our model, cellular apoptosis and proliferation occurred at the level of tubular epithelial cells and not within cells of peritubular capillaries (Figure S1B in Supplementary Material).

# Complement Modulation Abrogated I/R Injury-Induced PMT and Attenuated Capillary Lumen Reduction

Next, we investigated the possible occurrence of PMT after I/R. Normal kidney showed αSMA expression predominately in smooth muscle cells (wall of renal arteries, **Figures 4A,D**, dotted arrow). Before ischemia, we found PDGFRβ+/αSMA<sup>−</sup> pericytes in interstitial peritubular capillaries (**Figures 4A,B**). After 24 h from I/R injury, perivascular cells upregulated αSMA together with an intense reduction in PDGFRβ expression, indicating PMT (**Figures 4C,D**). The co-localization of these two markers was more evident in arterioles and peritubular capillaries as shown in **Figure 4D**. Mesangial cells, which originate from the same FOXD1<sup>+</sup> embryonic mesenchymal precursors of pericytes, expressed PDGFRβ (30, 31). However, PDGFRβ expression by mesangial cells remained unaffected and no increase of PDGFRβ+/αSMA<sup>+</sup> was detected in glomerular cells (**Figure 4H**). C1-INH treatment significantly reduced the number of PDGFRβ+/αSMA<sup>+</sup> cells in the peritubular capillaries, preserving the physiological pericytes phenotype (**Figures 4E–G**). These data demonstrate that that inhibition of C1 activity is associated with decreased PMT. Interestingly, we assessed the occurrence of PMT in a mouse model of renal I/R injury. In the Wtype sham, we

contrast, 24 h of C1-INH treated pigs showed PDGFRβ/NG2 marker restoration (C,F). (H) Results are expressed as median ± interquartile range of the numbers of PDGFRβ+/NG2+ cells/high power fields (HPF) of five independent pigs for each group, \**p* < 0.05. Magnification, 630×, scale bar = 50 μm. (G) Isotype control staining for NG2.

detected the PDGFRβ expression at level of peritubular capillaries, where PDGFRβ+/αSMA<sup>−</sup> cells were detected (**Figure 5A**). One day after reperfusion, a significant reduction of PDGFRβ and an increase of aSMA in the peritubular capillaries were observed in the Wtype (**Figure 5B**). On the contrary, C5aR1<sup>−</sup>/<sup>−</sup> mice showed significantly lower number of PDGFRβ+/αSMA<sup>+</sup> cells compared with the Wtype (**Figures 5C,I**).

To investigate the possibility that the occurrence of PMT might influence microvessel luminal diameter, capillaries lumen area was measured in PDGFRβ-stained renal sections (**Figure 5G**). Compared to T0 (**Figure 5D**), we found that PDGFRβ downregulation was evident in the interstitial capillaries characterized by lumen reduction (**Figure 5E**; % area fraction: T24CTRL 3.95 ± 2.36% versus T0 11.30 ± 2.6%). Interestingly, the treatment with C1-INH restored basal capillary area fraction (**Figure 5F**; T24 C1-INH 12.06 ± 3.8% versus T24 CTRL). We found that the restoration of capillary lumen was statistically significant (**Figure 5H**).

# C5a Induced PMT Without Affecting Pericytes Viability *In Vitro*

To validate our findings *in vitro*, we next evaluated the PDGFRβ expression in pericytes culture stimulated with the complement anaphylotoxin C5a and TGFβ, a classic mediator of PMT (7, 32). After C5a stimulation, we found that the number of PDGFRβ+-pericytes significantly decreased indicating the

phenotypic occurrence of PMT (**Figures 6A,B** PDGFRβ+ pericytes: C5a 48.32 ± 7.89 versus basal 79.98 ± 10.45%, *p* < 0.05). This activation persisted even after 48 h from stimulation (data not shown). By using the AnnexinV/Propidium Iodide and MTT test, we also found that pericytes did not undergo to early and late apoptosis upon C5a activation (**Figure 6C**, early apoptosis: C5a 14.57 ± 0.82% versus Bas 11.19 ± 1.52%, ns, Figure S2A in Supplementary Material) nor downregulated their proliferative response (**Figure 6D**). On the contrary, TGFβ stimulation upregulated the percentage of early apoptotic cells during pericytes activation.

We also observed that C5a-stimulated pericytes acquired spindle-like shape morphology similar to fibroblasts. Performing immunofluorescence analysis, we found an increased expression of αSMA in stress fibers (PDGFRβ/αSMA, **Figure 7A**), indicating the acquirement of a contractile phenotype. Moreover, the morphologic changes were accompanied by increased collagen I protein synthesis (**Figure 7B**, C5a 24 h: 63.17 ± 8.22% versus basal: 19.86 ± 15.07%, *p* < 0.05) as well as connective tissue growth factor mRNA expression (**Figure 7C**). Interestingly, C5a increased metalloproteinase MMP9 protein level and ADAMTS1 gene expression, which are usually involved in pericytes

Figure 5 | PMT is modulated by C5aR1 and associated with capillary lumen reduction 1 day after ischemia/reperfusion (I/R) injury. Renal I/R injury was performed for 24 h in C5aR1-deficient mice as showed in (A–C,I) and in a swine model treated with C1-INH as showed in (D–H). (A) Confocal images showing interstitial peritubular capillaries pericytes in a mouse model of I/R injury co-labeled with PDGFRβ (green) and αSMA (red). In the Wtype sham, PDGFRβ+/αSMA+ perivascular cells were barely detectable. After 24 h of I/R injury (B), the number of PDGFRβ+/αSMA+ cells increased predominately at peritubular capillaries level. (C) C5aR1−/<sup>−</sup> mice were protected from PDGFRβ loss and αSMA increase. Results are expressed as median ± interquartile range (IQR) of the numbers of PDGFRβ+/αSMA+ cells/ high power fields (HPF) of five independent mouse for each group (I). Magnification, 50×. Representative IHC images of PDGFRβ-stained renal biopsies used to measure the peritubular capillaries area (D,G). After 24 h of I/R injury (E), microvessels appeared constricted with a significant decrease in luminal diameter. (F) The treatment with C1-INH restored basal capillary area fraction. Scale bar, 100 µM. (G) Schematic panel showing the calculation of capillary lumen area using Image J Software. (H) Graph indicating the mean of capillary lumen area. Results are expressed as median ± IQR of the capillary area fraction (%), *n* = 3 for each group.

detachment during PMT in experimental UUO (Figures S2B,C in Supplementary Material) (33).

Finally, we evaluated the modulation of Id2, a critical DNAbinding protein implicated in the regulation of the profibrotic/ mesenchymal terminal differentiation (34–36). As expected, Id2 was highly expressed in normal cultured pericytes. By 6 h from C5a stimulation, we found a significant downregulation of Id2 in pericytes undergoing PMT; the downregulation was still detectable at 18 h and was comparable to the effects of TGFβ, a negative modulator of Id2 (**Figure 7D**).

#### C5a Signaling Induces PMT by Canonical and Non-Canonical TGF**β** Pathway *via* pERK

To identify the intracellular signaling involved in the C5a-induced PMT, cultured pericytes were incubated with complement anaphylotoxin and TGF-β1 (**Figures 8** and **9**). TGFβ contributes to renal fibrosis by the activation of canonical (SMAD 2/3) pathway and non-canonical [mitogen-activated protein kinase (MAPK)] pathways (37) (Figure S3 in Supplementary Material). Since also C5a, by interaction with the C5aR, induces the activation of ERK/MAPK pathway (38–40), we evaluated pERK protein modulation, as a possible common mediator of TGFβ and C5a signaling. We found by FACS analysis that pericytes expressed the C5aR at cytoplasmic and membrane level (**Figure 8A**). In addition, pericytes significantly upregulated the C5aR1 mRNA, which increased after 18 h of stimulation (**Figure 8B**). As shown in **Figure 8C**, C5a increased ERK phosphorylation after 15 and 30 min.

Next, we investigated whether the C5aR-induced signaling could play a role in promoting the PMT. We used anti-C5aR specific neutralizing antibody to inhibit the C5a binding to C5aR. Anti-C5aR prevented the increase of collagen I induced

stress fibers (B) FACS analysis showed increased collagen I expression in permeabilized cells, after C5a exposition. (C) mRNA expression levels of connective tissue growth factor (CTGF) (CCN2) were determined by qPCR. C5a stimulated pericytes showed a significant increase after 3 and 6 h of incubation. The fold change of CTGF expression was normalized to GAPDH. The histograms represent the mean ± SD, *n* = 3. (D) Western Blot showed a significant reduction of Id2 protein compared to basal condition, β-actin protein expression was used for normalization (\**p* < 0.05, \*\**p* < 0.01).

by C5a exposition (**Figure 8D**, *bottom right*). These results indicate that pericytes expressed the C5aR and that C5aR activation *via* pERK (non-canonical TGFβ pathway) might promote PMT. To further validate that pERK pathway was also pivotal in TGFβ-canonical signaling, we evaluated the effect of C5a stimulation on SMAD2/3 phosphorylation. First, we found the C5a increased the amount of pSMAD2/3 at 15 and 30 min (**Figure 9A**). Further time course at 6 and 18 h revealed that C5a led to a persistent increase of total SMAD2/3 complex (**Figure 9B**). Finally, we tested the effect of SC1 (Pluripotin) on C5a-induced PMT. SC1 is a dual kinase (ERK1, MAPK3) inhibitor that blocks ERK1/2 phosphorylation of at Thr-202/ Tyr-204 (41). Analysis of ERK phosphorylation showed an inhibition at the concentration of 1 µM for 24 h (0.26 ± 0.18 fold compared to untreated cells 2.4 ± 1.18); higher concentrations (3–5 µM) interfered with cellular viability and were not considered. Pretreatment of pericytes with 1 µM SC1 for 24 h reduced ERK phosphorylation (**Figure 8C**), blocked C5a-induced SMAD3 phosphorylation at 15 min (SC1 1 µM SC1 for 24 h + C5a 15 min: 0.69 ± 0.32 compared to C5a 15 min: 1.32 ± 0.25) (**Figure 8D**) and significantly reduced the C5a-induced collagen I production at 12 h (SC1 1 µM SC1 for 24 h + C5a 12 h: 33.10 ± 12.15 compared to C5a 12 h: 8.34 ± 4.39) (**Figure 8D**, left bottom). These data support the role of C5a in promoting PMT by the activation of both TGFβcanonical and non-canonical pathway.

#### DISCUSSION

In the present study, we demonstrated that inhibition of the complement system in I/R injury prevents the occurrence of PMT and the reduction of peritubular capillaries lumen areas. In particular, C5a had a pro-fibrotic activity driving pericytes toward a maladaptive dysfunctional phenotype by modulation of pERK activation.

Pericytes are mesenchymal-derived cells that interact with endothelial cells, releasing trophic factors such as VEGF and PDGF-BB (42). Recently, several studies demonstrated that pericytes are one of the myofibroblast precursors during development of tissue fibrosis (6, 43–45). Our results showed that complement is involved in transdifferentiation of pericytes in the early phases after kidney transplantation. Complement system is a key player in renal I/R injury (17, 46, 47), and C1-INH, a serine protease inhibitor used for the therapy of hereditary angioedema (23) might offer a new strategy for the prevention of I/R injury (25, 48, 49). Previously, it has been shown C1-INH treatment significantly prevents fibrosis, improves early and long-term renal function (26) and protected from TGFβ pathway activation (49). In our swine model of I/R, we found at early time after reperfusion (30 min) the deposition of complement components (i. e C4d, C3c, and C5b-9), along the peritubular capillaries (18), the areas where pericytes niche are localized. We also found the co-localization of PDGFRβ together with C3 (Figure S1 in Supplementary Material) and C5b-9 (Figure S4 in Supplementary Material). In accordance, starting from 30 min, the downregulation of PDGFRβ in our model began exactly along peritubular capillaries, without involvement of PDGFRβ expression in mesangial cells and larger arteries. The treatment with C1-INH, reduced the C4d peritubular deposition (18), and significantly restored pericytes markers expression, thereby accelerating the pericytes recovery. Even if pericytes identification requires transmission electron microscopy and fate-tracing analysis (5) several studies suggested that a PDGFRβ+ perivascular cell population is involved in collagen release, since specific PDGFRβ blocking reduced fibrosis development (32). Additionally, NG2 (50–52) was used to specifically label pericytes. We demonstrated a significant reduction in PDGFRβ/NG2 double positive cells after 24 h of I/R. Interestingly, our data are in line with Hosaka and colleagues (53) showing that in the PMT occurring during the tumor growth and metastasis, the loss of PDGFRβ and NG2 is not due to pericytes death nor proliferation. We found that renal I/R injury did not induce apoptosis of pericytes but their activation characterized by a maladaptive response not leading to cellular death but transdifferentiation toward a myofibroblast phenotype. In accordance, complement system did not affect pericytes viability *in vitro*. Furthermore, our data on renal pericytes are different compared with a model of cerebral ischemia *in vivo*, were pericytes died by apoptosis *in rigor mortis* and induced an irreversible constriction of micro vessel (12). Another difference in our data regarded the pericyte proliferation in the early stage of I/R. Using a rat model of I/R (54) and a transgenic reporter mice to determine the contribution of pericytes to fibrosis, previous reports described the increased proliferation of pericytes, starting from 48 to 72 h after injury (5). This difference might be explained by the fact that our analysis was conducted at 24 h; in addition, we analyzed a swine model.

Several studies revealed the importance of kidney pericytes for peritubular capillary integrity (55) and the microvascular rarefaction following ischemic acute kidney injury (AKI) (56, 57). After I/R, microvessels showed CD31 reduction and αSMA increase indicating the occurrence of endothelial-to-mesenchymal transition, which also contributes to kidney fibrosis (2, 27, 53); in this contest, αSMA, a marker of activated myofibroblasts, amplifies cell contractility with reorganization of stress fibers(9, 58). Here, we demonstrated that αSMA increase is also associated to microvascular pericytes. In accordance with Gomez and colleagues (59), we hypothesized that PMT might lead to direct constriction of vessels with reduction of capillaries density and lumen area. This process has been described in cerebral I/R injury where pericytes led to irreversible constriction of capillaries, exacerbating tissue hypoxia (11, 12). Interestingly, we found that treatment with C1-INH was capable to maintain the capillary lumen area by counteracting PMT. We recognized that C1-INH, beyond targeting classical and lectin pathway can inhibit the contact, coagulation, and fibrinolytic pathway involved in blood flow dysfunction (24). This can result in a reduced thrombi formation and a systemic improvement of vascular stability. However, PMT inhibition by C1-INH could provide a new mechanism to preserve graft from capillary rarefaction and reduction of lumen area.

Ischemia/reperfusion injury is primarily mediated by complement activation, with C5a playing a pivotal role also in inflammation response and allograft rejection (14, 15, 17, 60, 61). In this paper, we demonstrated for the first time that C5a can induce PMT with important pro-fibrotic effects. We stimulated cells with C5a because represents the most potent pro-inflammatory and chemotactic mediator (62) with specific receptors demonstrated at the level of renal resident cells. Nevertheless, despite the C5a pivotal pathogenic role, in clinical trials, the C5 inhibition by the human monoclonal antibody Eculizumab has been shown not to be sufficient to prevent DGF as well as antibody-mediated rejection (63, 64). Specifically, even after Eculizumab treatment, a residual C5 activity has been demonstrated (65). This could explain why not all patients benefit of an anti-C5 therapy in C3-mediated kidney diseases or during strong complement activation (24). Therefore, strategies that act upstream of C5 activation to prevent opsonization and generation of C3 activated products (i.e., C3a, iC3b, C3b, and later C5a) have been evaluated (24). As endogenous serine protease inhibitor, C1-INH has an excellent safety compared to Eculizumab (66, 67) and indirectly inhibits the release of the reactive late component C5a (68).

The C5a/C5aR pathway has extensively shown to cause recruitment of neutrophils and macrophages and exacerbate tubular injury in acute kidney injury. C5aR deficiency on renal cells or circulating leukocytes can significantly ameliorate renal injury (14, 16, 28, 39, 69).

The C5a-induced PMT was characterized by: the acquirement of αSMA stress fibers, the production of extracellular matrix proteins as collagen I, the downregulation of the antifibrotic BMP7-Id2 signaling (70–72), and finally by the activation of TGFβ canonical and non-canonical pathway (70–77). Our data connecting complement activation with fibrosis are in accordance with other disease model such as lung fibrosis where complement might lead to SMAD2/3 dependent and independent pathway activation, shifting the initial acute inflammatory response in a chronic profibrotic state (77).

We also analyzed ERK activation at early times after C5a stimulation, since ERK is a common downstream mediator of C5aR, TGFβ non-canonical signaling, and a possible inducer of TGFβ canonical pathway (73, 75) (Figure S3 in Supplementary Material). First, we showed that human pericytes expressed the C5aR, both at cytoplasmic level and on membrane surface. Second, by blocking the C5aR, we demonstrated that C5aR signaling is involved in collagen I release. As observed by other group (77), C5a can activate SMADs proteins, independently from TGFβ. In this signaling, ERK could act as bifurcation point to induce both the non-canonical and the canonical TGFβ pathways (78). After ERK phosphorylation blockade by SC1 (pluripotin), a dual kinase (ERK1, MAPK3) inhibitor (41), we found a significant reduction of C5a-induced SMAD3 activation and of C5a-induced collagen production. In accordance with recent evidences (79, 80), these results suggest that ERK might regulate TGF-β/Smad signaling. Therefore, next to TGFβ (81), also innate immune signaling (59) (i.e., anaphylotoxins) might lead to an amplification of interstitial extracellular matrix accumulation by generating myofibroblast *via* PMT after AKI (82). In accordance with our *in vitro* findings, the *in vivo* C5aR1 deficiency protected from PMT, indicating that C5a/C5aR1 is involved in tubulointerstitial fibrosis as shown by Martin et al. (83).

In conclusion, our data suggest that in the early phase of I/R injury, renal pericytes are a major target of complement activation resulting in maladaptive response and PMT. Considering the pivotal role of renal pericytes in preserving vascular homeostasis and maintaining blood perfusion, our data offer new insight into the pathogenic mechanisms regulating vascular capillary reduction and fibrosis development in AKI with potential future therapeutic application.

#### ETHICS STATEMENT

The study was approved by the ethical committee of the Ministry of Health, Italy. This work was supported by University of Bari "Aldo Moro" Ministero della Salute (Ricerca Finalizzata 2009 granted to GC and GG) and an unrestricted research grant from Pharming Group.

# AUTHOR CONTRIBUTIONS

GC, RF, AS, CD, FS, PP, MB, and FS performed the experiments design; RF, AS, CD, FS, and GC performed experiments; MB, AC, GL, and GS provided the pig animal model samples; MS and MD contributed the mouse model. GC supervised all the experiments. GG and LG supervised the project. All authors were involved in data interpretation. GC and RF wrote the paper, all authors had final approval of the submitted and published versions.

# FUNDING

This work was supported by University of Bari "Aldo Moro" and Regione Puglia (Ph.D. in Biotechnology applied to Organ and Tissue Transplantation to RF) and Ministero della Salute (Ricerca Finalizzata 2009, GR-2009-1608662, granted to GC and GG). The authors thank Eustacchio Montemurno for image in Figure S3 in Supplementary Material. The authors thank Beatris Oortwijn and Edwin van Amersfoort from Pharming Group NV, Leiden, the Netherlands for the unrestricted research grant supporting the study.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01002/ full#supplementary-material.

Figure S1 | Ischemia/reperfusion (I/R) injury did not induce apoptosis or proliferation within perivascular compartment. Quantification of Casp3<sup>+</sup> (A) and Ki-67<sup>+</sup> (B) cells by IHC after I/R injury with distinction between tubular and perivascular cells. At different times from reperfusion, the Casp3+ and Ki-67+ were detected predominately at tubular level. Results are expressed as mean ± SD of Casp3+ or Ki-67+ cells/high power fields (HPF). Immunofluorescence images showing interstitial peritubular capillaries pericytes co-labeled with PDGFRβ (green) and C3 (red). In T0, PDGFRβ+/C3<sup>+</sup> perivascular cells were barely detectable (C). After 30 min of I/R injury [(D), rectangle area was zoomed in (G)], the number of PDGFRβ+/C3+ cells increased predominately at peritubular capillaries level [(D), arrowheads]. C1-INH treatment limited the C3 deposition (E). Isotype control staining was used as negative control (F). Results are expressed as median ± interquartile range of the numbers of PDGFRβ+/C3+ cells/HPF of five independent pigs for each group (H). Magnification, 50×.

Figure S2 | C5a did not trigger apoptosis and activated expression of matrix metallopeptidase (MMP9) and ADAMTS1 in pericytes. Human pericytes were treated with C5a and TGFβ for 6, 18, and 24 h. (A) Cell apoptosis or necrosis was analyzed by flow cytometry after Annexin V/propidium iodide (PI) staining. The units of the *Y* and *X* axes are fluorescence intensity. Early apoptotic cells are Annexin-V+/PI−; late apoptotic cells are both Annexin-V + 7/PI+; and necrotic cells were Annexin-V−/PI+. Data are expressed as apoptosis (early and late %) or necrosis (%) (as indicated in Figure 5C). (B) Western blotting demonstrated the increased expression of active form (85 kDa) of MMP9. (C) qPCR demonstrated the upregulation of ADAMTS1 mRNA after C5a and TGFβ exposition. *p* < 0.05 versus basal.

Figure S3 | Schematic pathway showing the possible cross-talk between C5aR and TGFβ canonical and non-canonical pathway. C5aR, a G proteincoupled receptor for C5a anaphylotoxin, promotes the MAPK signaling activation (Ras/Raf/MEK), inducing the extracellular signal-regulated kinases (ERK) phosphorylation and transcription of pro-inflammatory and pro-fibrotic genes. pERK activation is also one of the final effector factors of SMADindependent TGFβ pathway (non-canonical pathway) that include various branches of MAP kinase pathways, Rho-like GTPase signaling pathways and phosphatidylinositol-3-kinase/AKT pathways (*not showed*). Independently from TGFβ presence, C5a could activate SMAD-independent signaling leading to activation of profibrotic pathway. In addition, pERK could be involved in the activation of SMAD-dependent TGFβ pathway (canonical pathway, green arrow and factors) inducing the SMAD2/3 phosphorylation (red dotted arrow). As common downstream mechanisms, the C5a exposition led to transcription of profibrotic gene and proteinase for detachment. Blocking of pERK, by SC1 (Pluripotin) a dual kinase (ERK1, MAPK3) inhibitor of Thr-202/Tyr-204 phosphorylation could interferes with C5a-induced transcription of proinflammatory and with SMAD3 phosphorylation. This could lead to the decrease of extracellular matrix protein accumulation by perivascular pericytes (C5aR, complement component C5a Receptor 1; MAPK, mitogen-activated protein kinase, ERK, extracellular signal-regulated kinases; SMAD, small mother against decapentaplegic).

Figure S4 | C5b-9 deposition occurred at peritubular capillary level and is modulated by C1-INH after 30 min of ischemia/reperfusion (I/R) injury. Immunofluorescence images showing interstitial peritubular capillaries pericytes co-labeled with PDGFRβ (green) and C5b-9 (red). In T0, PDGFRβ+/C5b-9<sup>+</sup> perivascular cells were barely detectable (A). After 30 min of I/R injury, the number of PDGFRβ+/C5b9+ cells increased predominately at peritubular capillaries level and glomerular level (B). C1-INH treatment limited the C5b-9 deposition (C). Results are expressed as media ± SEM of the numbers of PDGFRβ+/C5b-9+ cells/high power fields of five independent pigs for each group (G). Scale bar in (A–C): 100 µm. Boxed area was enlarged in (D–F) (scale bar: 50 µm).

#### 1. Ferenbach DA, Bonventre JV. Mechanisms of maladaptive repair after AKI leading to accelerated kidney ageing and CKD. *Nat Rev Nephrol* (2015) 21. Nielsen EW, Waage C, Fure H, Brekke OL, Sfyroera G, Lambris JD, et al. Effect of supraphysiologic levels of C1-inhibitor on the classical, lectin and alternative pathways of complement. *Mol Immunol* (2007) 44(8):1819–26. doi:10.1016/j.molimm.2006.10.003


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Castellano, Franzin, Stasi, Divella, Sallustio, Pontrelli, Lucarelli, Battaglia, Staffieri, Crovace, Stallone, Seelen, Daha, Grandaliano and Gesualdo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Complement C3 Produced by Macrophages Promotes Renal Fibrosis via IL-17A Secretion

Yanyan Liu1†, Kun Wang1†, Xinjun Liang<sup>2</sup> , Yueqiang Li <sup>1</sup> , Ying Zhang<sup>1</sup> , Chunxiu Zhang<sup>1</sup> , Haotian Wei <sup>1</sup> , Ran Luo<sup>1</sup> , Shuwang Ge<sup>1</sup> and Gang Xu<sup>1</sup> \*

<sup>1</sup> Division of Internal Medicine, Department of Nephrology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, <sup>2</sup> Hubei Cancer Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China

Complement synthesis in cells of origin is strongly linked to the pathogenesis and progression of renal disease. Multiple studies have examined local C3 synthesis in renal disease and elucidated the contribution of local cellular sources, but the contribution of infiltrating inflammatory cells remains unclear. We investigate the relationships among C3, macrophages and Th17 cells, which are involved in interstitial fibrosis. Here, we report that increased local C3 expression, mainly by monocyte/macrophages, was detected in renal biopsy specimens and was correlated with the severity of renal fibrosis (RF) and indexes of renal function. In mouse models of UUO (unilateral ureteral obstruction), we found that local C3 was constitutively expressed throughout the kidney in the interstitium, from which it was released by F4/80+macrophages. After the depletion of macrophages using clodronate, mice lacking macrophages exhibited reductions in C3 expression and renal tubulointerstitial fibrosis. Blocking C3 expression with a C3 and C3aR inhibitor provided similar protection against renal tubulointerstitial fibrosis. These protective effects were associated with reduced pro-inflammatory cytokines, renal recruitment of inflammatory cells, and the Th17 response. in vitro, recombinant C3a significantly enhanced T cell proliferation and IL-17A expression, which was mediated through phosphorylation of ERK, STAT3, and STAT5 and activation of NF-kB in T cells. More importantly, blockade of C3a by a C3aR inhibitor drastically suppressed IL-17A expression in C3a-stimulated T cells. We propose that local C3 secretion by macrophages leads to IL-17A-mediated inflammatory cell infiltration into the kidney, which further drives fibrogenic responses. Our findings suggest that inhibition of the C3a/C3aR pathway is a novel therapeutic approach for obstructive nephropathy.

Keywords: complement component 3, macrophage, renal fibrosis (RF), IL-17A, IgAN

# INTRODUCTION

Renal fibrosis (RF) has become an important worldwide health problem and represents a major economic burden to society. IgA nephropathy (IgAN) is one of the most common causes of chronic kidney disease (CKD), and the prognosis of IgAN is more closely associated with the severity of interstitial injury and fibrosis than that of glomerular lesions (1). As the common consequence of all forms of CKD, RF is characterized by deposition of extracellular matrix and associated with inflammatory cell recruitment, angiogenesis, lymphangiogenesis, an myofibroblast formation (2–4). Although researchers have confirmed that chronic inflammation, oxidative stress,

#### Edited by:

Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France

#### Reviewed by:

Marie-Agnes Dragon-Durey, Université Paris Descartes, France Ranjit Kumar Sahu, University of Virginia, United States

> \*Correspondence: Gang Xu xugang@tjh.tjmu.edu.cn

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 29 May 2018 Accepted: 26 September 2018 Published: 22 October 2018

#### Citation:

Liu Y, Wang K, Liang X, Li Y, Zhang Y, Zhang C, Wei H, Luo R, Ge S and Xu G (2018) Complement C3 Produced by Macrophages Promotes Renal Fibrosis via IL-17A Secretion. Front. Immunol. 9:2385. doi: 10.3389/fimmu.2018.02385 proteinuria, and abnormal activation of complement are involved in the development and progression of CKD, the pathogenesis of RF remains largely unknown (5–8).

The complement system is a crucial part of the immune system and consists of multiple categories of components. The complement component 3 (C3), a 180 kDa glycoprotein, plays a central role in activation of the complement system. Its activation is required for both classical and alternative complement activation pathways. Circulating C3 is produced by the liver, and its extrahepatic production has been observed in other specialized cells, including mast cells, fibroblasts, smooth muscle cells, and macrophages (9, 10). These cells synthesize C3, presumably through their bioactive products, and have an important role in regulating other aspects of autoimmunity, inflammation, and pathogen host defense. Studies have confirmed that locally synthesized C3 appears to have a stronger influence on rejection than circulating C3 (11). Other reports have demonstrated that the epithelial and vascular tissues at local sites of inflammation could secrete complement components (12). Despite Xavier and Cui have demonstrated that complement C3 activation and macrophage infiltration may play important roles in the progression of interstitial fibrosis in UUO mice and human hypertensive nephropathy, the specific mechanism of local synthesis in renal interstitium by immune cells has not been thoroughly investigated (13, 14).

As early as the 1970s, Pepys first discovered the interactions between complement and adaptive immunity by observing complement-depleted mice that were unable to mount potent antibody responses (15). Previous studies have found that complement C3 deficiency is associated with impaired T cell responses in several disease models, including infections, tumors, autoimmune disease, and renal transplantation (16). More recently, researchers observed that the complement components C3, fB, fD, and C5 were upregulated as regulators of T cell immunity, as well as the C3a receptor and C5a receptor. In some kidney transplant studies, the absence of C3 was correlated with defective T cell priming, reduced T cell proliferation, and cytokine production after donor-specific restimulation (17, 18). Moreover, C3 deficiency or blockade was shown to attenuate the expansion of Ag-specific CD4<sup>+</sup> and CD8<sup>+</sup> T cell responses to Listeria monocytogenes in mice, and the regulation of T cell functionality by C3 might not involve the C5aR signaling pathway (19).

In this study, we aimed to identify the roles of locally synthesized C3 in the development of RF in a unilateral ureteral obstruction (UUO) model and determine whether this synthesis contributes to M1 cell responses. In addition, we studied the relationships among C3, macrophages and Th17 cells, which are involved in interstitial fibrosis.

#### MATERIALS AND METHODS

#### Renal and Blood Samples From Patients With IgAN

Between December 2016 and July 2017, patients aged 18–63 years who underwent kidney biopsy at Tongji Hospital were recruited, and renal biopsy specimens were examined retrospectively. Forty-one patients (20 men and 21 women; mean age = 38.10 ± 12.03 years) with a pathologic diagnosis of IgAN were enrolled. Their renal and/or blood samples were obtained at the time of diagnosis. Patients were excluded from this study if they met the following conditions: <18 years of age, an inability to provide informed consent, presence of active infection, and pregnancy. Our protocol was approved by the institutional review board or ethics committee at each center. Written informed consent was obtained from all patients.

#### Animal Model

The animals were purchased from Charles river Laboratories (Beijing, China). Unilateral ureteral obstruction (UUO) is a popular experimental model of renal injury. Mice aged 6– 8 weeks were anesthetized followed by a lateral incision on the back of the mouse. Subsequently, the left ureter was exposed and tied off with two 4.0 silk suture. Shamoperated mice underwent an identical procedure but without ureteric ligation. The therapeutic experiment was performed with the Compstatin analog Cp40 (dTyr-Ile-[Cys-Val-Trp(Me)- Gln-Asp-Trp-Sar-His-Arg-Cys]-mIle-NH2;1.7kDa) which was produced by solid-phase peptide synthesis (GL Biochem Co., Ltd., Shanghai, China), and SB290157, a C3a receptor antagonist, which was purchased from Sigma-Aldrich. UUO and sham-operated mice were treated with Cp40 (1 mg/kg) via subcutaneous injection every 12 h and SB290157 (30 mg/kg) via intraperitoneal injection daily. After 7 or 14 days, the mice were sacrificed by cervical vertebra dislocation, and then, peripheral blood, spleen, and renal tissues were collected. The mouse kidneys were fixed in 4% formalin for 24 h, processed through dehydration in a graded series of alcohol and embedded in paraffin (Wuhan Goodbio Technology Co., Ltd., Wuhan, China). The remaining sample was frozen in liquid nitrogen for later use. All animal studies were performed in accordance with our university's guidelines for animal care.

#### IHC and Immunofluorescence

Paraffin-embedded renal sections (3µm) were subjected to Masson's trichome staining as previously reported (20). Paraffinembedded renal sections (4µm) were deparaffinized in xylene and rehydrated in graded alcohol. The endogenous peroxidase activity was blocked with 3% H2O<sup>2</sup> at room temperature for 15 min, and non-specific proteins were blocked with 10% goat serum for 30 min. Sections were then incubated overnight with antibodies against F4/80, α-SMA (Abcam, Cambridge, MA, USA), C3 (Novus, Littleton, Colorado, USA), iNOS (Santa Cruz, Dallas, Texas, USA), CD68 (Long Island Biotech, Shanghai, China), CD3, CD4, and CD8 (Thermo Scientific, Waltham, MA, USA) at 4◦C, followed by incubation with an HRPconjugated secondary antibody and subsequently visualized with diaminobenzidine substrate and hematoxylin counterstaining. For double-labeling immunofluorescence studies, following incubation with primary antibody, sections were incubated with FITC-conjugated goat anti-rabbit (1:100; Abcam) and Goat antirat Alexa Fluor 594 (1:100; Invitrogen Corporation, Carlsbad, CA) for 45 min at 37◦C and then counterstained with DAPI (Vector Laboratories). C3 expression was quantificated by the percentages of positive area at ×400 magnification on 10 fields exclusive of blood vessels and glomerulus per section from six mice in each group. All scorings were carried out by observers blinded to the experimental groups.

#### Western Blotting

Protein concentrations were quantified by a BCA protein assay kit (Beyotime Institute of Biotechnology, Shanghai, China), and 20 µg of protein was used for gel loading. GAPDH primary antibody (mouse, Santa Cruz, USA) was used at a dilution of 1:3,000, and TGF-β1, collagen I, PDGFR-β, and α-SMA (Abcam, Cambridge, MA, USA) primary antibodies were used at a dilution of 1:2,000. C3 primary antibody (rabbit, Novus, USA), p-ERK, ERK, p-p65, p-STAT3, and p-STAT5 (rabbit, CST, USA) were used at a dilution of 1:1,000. iNOS primary antibody (mouse, Santa, USA) and Arginase1 (rabbit, Santa, USA) were used at a dilution of 1:200. C3aR primary antibody (mouse, Abcam, USA) was used at a dilution of 1:1,000. The secondary antibody was used at a dilution of 1:3,000. Western blotting analysis was performed as previously described (21). The signals were detected using enhanced chemiluminescence (ECL) (Amersham Pharmacia Biotech, Piscataway, NJ).

#### Mouse Cell Preparation

C57BL/6 mice were euthanized by cervical vertebra dislocation, and then, the whole body vasculature was flushed with a 20 ml injection of fresh PBS through a cardiac puncture. Kidneys were harvested and cut into small pieces and placed in RPMI1640 medium containing 2 mg/ml collagenase IV (GIBCO) and 100 mg/ml DNase I (Roche) for 45–60 min at 37◦C with intermittent agitation. After tissue disaggregation, cells were filtered through a 40µm cell strainer (BD Falcon, Franklin Lakes, NJ). Mononuclear cells from kidneys were then washed with cold PBS, counted, and used for flow cytometry.

#### PBMC Isolation

Patients and normal subjects donated 5 ml of blood collected in heparinized tubes. Blood was diluted 1:1 with PBS and overlaid onto lymphocyte separation medium (TBD sciences, Tianjin, China). After centrifugation, 3 ml of the interface containing the PBMCs was collected and diluted to 6 ml with PBS, then washed twice with cold PBS and counted. The PBMCs were collected for flow cytometric analysis.

# CFSE Labeling

A CFSE stock solution (5 mM) was prepared fresh by dissolving lyophilized CFSE (Sigma-Aldrich, USA) in DMSO. Splenocytes were obtained from the spleens of naïve mice, and labeled with CFSE at 5µM in PBS for 15 min at 37◦C. Excess CFSE was quenched by adding three volumes of ice-cold FBS and incubating the cells for 5 min on ice. CFSE labeled cells were then washed three times with PBS and cultured with or without stimulation.

# T Cell Activation

The 96-well assay plate precoated with anti-CD3 and anti-CD28 Ab (BD Pharmingen) was incubated at 37◦C for 4 h. Splenocytes (1 × 10<sup>6</sup> cells/well) were obtained from the spleens of naïve mice and were cultured for 3 days in 96-well plates in medium containing IL-2 (10 ng/mL; R&D Systems) as well as IL-12 (10µg/mL; R&D Systems), or IL-4 (4 ng/mL; BD Pharmingen).

# Flow Cytometric Analysis

A single renal cell suspension was prepared and stimulated with PMA/Ionomycin/Golgi-plug for 4 h. The cells were incubated with different primary antibodies or the appropriate isotype control antibodies at 4◦C for 30 min. The following antibodies were used PerCP/Cy5.5-conjugated anti-human CD14 (Biolegend), PerCP/Cy5.5-conjugated anti-mouse CD4 (Biolegend), APC-conjugated anti-mouse F4/80 (Biolegend), and PerCP/Cy5.5-conjugated anti-mouse CD11b (Biolegend).After cellular surface staining, cells were fixed and permeabilized with Cytofix/Cytoperm Soln Kit for intracellular staining with Alexa Fluor 488-conjugated anti-human C3 (Abcam) and PEconjugated anti-mouse IL-17A (eBioscience). All flow cytometric analyses were performed using an LSR II Flow Cytometer (Beckman-Coulter) and Flowjo software.

### ELISA

To quantify IL-17A levels in the kidney, samples were analyzed using a mouse IL-17A ELISA (R and D Systems) according to the manufacturer's instructions. All measurements were performed in duplicate.

#### Real-Time PCR

Real-time PCR was performed as previously described (22). Realtime PCR was carried out using the LightCycler 480 system (Roche, Pleasanton, CA, USA) with the following primers: mouse C3, forward 5′ -ACTGTGGACAACAACCTACTGC-3′ , reverse 5 ′ -GCATGTTCGTAAAAGGCTCGG-3′ ; mouse IL-6, forward 5 ′ -TAGTCCTTCCTACCCCAATTTCC-3′ , mouse reverse 5′ -T TGGTCCTTAGCCACTCCTTC-3′ ; mouse IL-1β, forward 5′ - GAAATGCCACCTTTTGACAGTG-3′ , reverse5′ -TGGATGCT CTCATCAGGACAG-3′ ; mouse TNF-α, forward 5′ -CCTGTA GCCCACGTCGTA G-3′ , reverse 5′ -GGGAGTAGACAAGGTA CAACCC-3′ ; mouse MCP-1, forward 5′ -TTAAAAACCTGGA TCGGAACCAA-3′ , reverse 5′ - GCATTAGCTTCAGATTTAC GGGT-3′ ; mouse Collagen 1, forward 5′ -GTCCTAGTCGAT GGCTGCTC-3′ , reverse 5′ -CAATGTCCAGAGGTGCAATG-3′ ; mouse α-SMA, forward 5′ -GGAGAAGCCCAGCCAGTCGC-3′ , reverse 5′ -AGCCGGCCTTACAGAGCCCA-3′ ; mouse PDGFRβ, forward 5′ -GGGTCCGTTCCAGAAAATGT-3′ , reverse 5′ -GA CAAGGGACCGGGGTCCAA-3′ ; mouse Arginase, forward 5′ -A GCGCCAAGTCCAGAACCATA-3′ , reverse 5′ -CCATGCAAG TTTCCACTTGT-3′ ; mouse iNOS, forward 5′ -TGGAGCGAGT TGTGGATTGTC-3′ , reverse 5′ -GTGAGGGCTTGGCTGAGTG A-3′ ; mouse GAPDH, forward 5′ -AGGTCGGTGTGAACGGA TTTG-3′ , reverse 5′ -GGGGTCGTTGATGGCAACA-3′ ; mouse C3, forward 5′ -ACGGCATCCTCTGTCATCT-3′ , reverse 5′ -A CGGCATCCTCTGTCATCT-3′ ; TGF-β1, forward 5′ -CGCAAC AACGCCATCTATGA-3′ , reverse 5′ -ACCAAGGTAACGCCAG GAAT-3′ . The relative amounts of mRNA were normalized to GAPDH and were calculated using the 2−11Ct approach as previously reported (23).

#### Statistical Analysis

All statistical analyses were conducted using SPSS 12.0 (SPSS, USA). The values are expressed as the mean ± SEM. Graphpad Prism 5 software (GraphPad Software, La Jolla, CA, USA) was used for the statistical analysis with Student's t-test or oneway ANOVA where appropriate. The threshold for statistical significance was set at P < 0.05.

### RESULTS

#### Renal Complement C3 Expression Is Elevated and Correlated With Infiltrating CD68<sup>+</sup> Monocytes/Macrophages in Human IgAN Biopsies

Complement was previously shown to play a key role in IgAN pathogenesis, which involves the aberrant activation of the classic, alternative, and mannose-binding lectin pathways. We recruited IgAN patients whose renal biopsy specimens were reassessed blindly by a single pathologist using the Oxford classification. Notably, in renal biopsy specimens, C3 expression was observed in both renal tubules and the interstitium, and a positive correlation was found between pathologist-assessed Masson's trichrome staining and C3 expression, although the correlation between C3 expression in the interstitium and serum C3 was not statistically significant (**Figures 1A,B**). However, the intensities of C3 in the interstitium were significantly positively correlated to BUN, serum creatinine (SCr), and urine proteinuria/UCr (ACR) and negatively correlated to the eGFR (**Figure 1B**). No correlation was found between the intensities of C3 in the interstitium and ALB. In addition, along with exacerbated RF and enhanced mononuclear leukocyte infiltration, C3 expression increased significantly (**Figure 1C**).

However, evidence of local secretion of complement components by infiltrating cells during IgAN is still absent. To address this, we assessed C3 expression in renal tissues and performed flow cytometric analysis of C3 secretion in peripheral blood of patients with IgAN. C3 secretion was increased significantly in peripheral blood monocytes from IgAN patients (**Figures 1D,E**). Accordingly, C3 expression and monocyte infiltration were examined in paraffinembedded sections of IgAN tissues by immunofluorescence staining. C3 expression was detected in the interstitium with significant co-staining of macrophages, which was observed as double-positive cells in patients, while healthy individuals showed little or no co-staining (**Figure 1F**). Altogether these results suggest that infiltrating macrophages as well as monocytes are a major source for C3 synthesis in kidney tissue.

# Renal Expression of C3 Is Upregulated in the Mouse UUO Model

C3 deposits within the glomerulus have been well-characterized in previous studies. To assess whether expression of C3 in the renal interstitium and kidney tubules is upregulated following UUO-induced renal injury, we performed immunochemical staining, real-time PCR, and Western blot analyses to measure its expression. Extensive stromal fibrosis was detected by Masson's trichrome staining, and increased C3 synthesis, α-SMA expression, and infiltrating macrophages were detected by immunohistochemistry (IHC) in mouse kidneys after the operation compared with those of the sham-operated mice; these parameters peaked on day 14 (**Figures 2A–C**, **Supplement Figures 1B–E**). Consistent with the histopathological results, C3 mRNA (**Supplement Figure 1A**) and protein (**Figures 2D,E**) levels were significantly increased after 7 and 14 days of UUO. Meanwhile, elevated MCP-1, IL-6, IL-1β, and TNF-α mRNA expression was also observed in UUO mice (**Figure 2F**).

#### Macrophages Are the Major Source of Complement C3 Production in the Kidney Following Obstructive Injury

Our IHC studies showed C3 overexpression in the interstitium of obstructed kidneys. To further define which cells were the major source of C3 production in UUO kidneys, we performed immunofluorescence staining using CD3, ly6G, F4/80, and α-SMA to identify T cells, PMNs, macrophages, and myofibroblasts, respectively. **Figure 2A** shows macrophages detected in the interstitium with significant C3 co-staining, shown as doublepositive cells, in UUO mice, while the sham group showed little or no co-staining. Given that macrophages with different activation phenotypes play distinct roles, we then proceeded to verify which subset of macrophages expressed C3 in vitro. Bone marrow-derived macrophages (BMDMs) were assessed after 7 days of culture with L929 supernatant. As previously reported, classically activated macrophages (M1) expressed high levels of iNOS and little Arginase-I, and alternatively activated macrophages (M2) expressed high levels of Arginase-I and little iNOS (**Figure 3A**). We found that C3 was primarily expressed by M1 cells, as shown by immunofluorescence staining and Western blotting (**Figures 3B,C**). Our results show that both tubules and interstitial cells secrete C3. Macrophages in the interstitium are likely to be affected in the microenvironment. As shown in **Figure 3D**, the addition of C3 to the induced differentiated macrophages in vitro could stimulate increased iNOS, IL-6, and IL-1β expression and decreased Arginase-I, TGF-β, and TNF-α expression, indicating that C3 could induce macrophage differentiation into M1.

#### Macrophage Depletion Reduces Complement Expression and Renal Fibrosis

Our above results showed that macrophages, especially M1 macrophages, contribute significantly to C3 secretion. We depleted F4/80<sup>+</sup> macrophages using clodronate liposomes in UUO mice. As depicted in **Figure 4A**, mice were sacrificed after 7 and 14 days of UUO, and the results showed that macrophages were reduced in UUO mice injected with clodronate compared

histograms show the increased the percentages of C3+CD14<sup>+</sup> in peripheral blood mononuclear in the IgA patients (n = 41) compared with the healthy (n = 37). The

error bars represent the SEM. \*\*P < 0.01. (F) Immunofluorescence staining of CD68 (red) and C3 (green) in the kidney. Scale bar, 50µm.

shown as mean ± SEM. n = 6 per group, Scale bar, 50µm. (D) The expression levels of C3 and fibrotic markers (α-SMA, PDGFR-β, and Collagen I) were detected by Western blot, and (E) the histogram shows the relative intensity for each marker normalized to GAPDH. n = 6 per group. (F) Real-time PCR showing relative renal mRNA levels of MCP-1, TNF-α, IL-6, and IL-1β in sham and UUO mice. The error bars represent the SEM. \*\*P < 0.01; \*\*\*P < 0.001.

to those injected with control PBS liposomes. Similarly, UUO mice treated with clodronate exhibited significantly decreased C3 expression. The reduction in C3 expression was associated with reduced α-SMA expression and decreased tubulointerstitial fibrosis measured by Masson's trichrome staining (**Figures 4B–F**). In addition, the mRNA and protein levels of C3, α-SMA, and PDGFR-β were reduced in UUO mice receiving clodronate liposomes (**Supplement Figure 2**). These results further support the pathogenic function of macrophage infiltration with increased C3 expression, which leads to RF.

# C3 Deficiency Attenuates Fibrosis and Infiltration of Inflammatory Cells in UUO-Induced Renal Fibrosis

To investigate the role of C3 in the pathogenesis of UUO, we used a peptidic C3 inhibitor, Compstatin analog Cp40, to

block C3 activation. Masson's staining and α-SMA expression analysis showed that UUO mice injected with 1 mg/kg Cp40 had much less severe interstitial fibrosis than control peptide-injected mice (**Supplement Figures 3A–D**). Western blot analysis also indicated that the α-SMA and PDGFRβ levels were decreased in the Cp40-injected UUO mice (**Supplement Figures 3E,F**). In addition to the attenuated tubulointerstitial fibrosis, renal infiltration of F4/80<sup>+</sup> macrophages, CD3+T cells, CD4+T cells, and CD8+T cells was significantly reduced in Cp40-injected UUO mice compared with peptide-injected mice (**Figures 5A–E**). Meanwhile, elevated MCP-1, IL-6, IL-1β, and TNF-α mRNA expression in UUO mice was markedly limited by Cp40 (**Figure 5F**). These data indicate that C3 mediates the infiltration of T cells and macrophages into the kidney in response to obstructive injury.

#### C3aR Blockade Substantially Attenuates Renal Fibrosis in UUO Mice

C3a is one of the proteins formed by the cleavage of C3 and plays a large role in the immune response. Given that increased generation of C3 was observed in IgAN patients and UUO mice, we proceeded to investigate whether the renal parenchymal loss in UUO mice could be attenuated by C3aR antagonism, SB290157. SB290157 is a selective antagonist of complement anaphylatoxin C3a receptor, a 74 amino acid proinflammatory mediator and chemotactic peptide. It effectively blocks C3aR in humans, rat, guinea pig, and mouse. As shown in **Figures 6A–C**, the progression of renal interstitial fibrosis was dramatically retarded by daily i.p., injection of SB290157, which was initiated on day 7 after the UUO operation. And injection of SB290157 did not affect C3 expression in renal interstitium of UUO mice. Similarly, along with attenuated RF and TGF-β1 release were significantly reduced by treatment with C3aRA (**Figures 6**D,**E**). In addition, we showed that renal infiltration of F4/80+macrophages, CD3+T cells, CD4+T cells, and CD8+T cells was significantly reduced in SB290157-injected UUO mice compared with that in peptide-injected mice (**Figures 6F–G**). Additionally, elevated MCP-1, IL-6, IL-1β, and TNF-α mRNA expression in UUO mice was markedly limited by SB290157 (**Figure 6H**). Taken together, these data suggest C3-C3aR

signaling promotes RF, and C3aR may be a therapeutic target for renal disease.

#### Blocking C3-C3aR Signaling Attenuates Renal Fibrosis by Inhibiting IL-17A Production in UUO Mice

As shown in **Figure 7A**, renal mRNA levels of IL-17A were substantially increased in UUO mice compared with sham control mice. In addition, IL-17A levels in the serum of UUO mice were significantly increased in the early and late stages compared with those in the serum of sham control mice (**Figure 7B**). Consistent with the ELISA and mRNA data, the FACS results revealed that 8.48 and 10.9% of CD4<sup>+</sup>

renal cells in obstructed kidneys expressed IL-17A following UUO, respectively, whereas ≤5.3% were IL-17A<sup>+</sup> in shamoperated mice, and this effect was strongly inhibited by Cp40 and SB290157 (**Figures 7C–H**). In addition, we performed analysis of CD11b+F4/80+IL-17<sup>+</sup> cells ratio in kidney from C3 blockade UUO mice and UUO mice. The results showed that CD11b+F4/80+IL-17+cells were around 1% in mononuclear cells in two groups, and only slightly changed after blockade C3 with CP40 (**Supplement Figures 4A,B**). Similar results were confirmed in 14 days of UUO mice (**Supplement Figures 4C,D**). Thus, we identified that the main producer of IL-17A in the UUO mice were T cells, which were strikingly increased after unilateral ureteral ligation.

To investigate the mechanism by which C3a promotes IL-17A expression, we isolated spleen cells from naïve wildtype mice on anti-CD3/anti-CD28-coated plates in vitro. We extended our CFSE dilution assay using splenocytes, and T cells exhibited high levels of proliferation in the presence of mC3a. Inhibition by SB290157 abrogated T cell proliferation; therefore, mC3a is required for T cell proliferation (**Figures 8A,B**). More importantly, blockade of C3a by the C3aR inhibitor drastically suppressed IL-17A expression in C3a-stimulated T cells (**Figures 8C,D**). In this study, we found that C3 deficiency significantly reduced the IL-17A production in obstructed kidneys. Given that IL-17A can stimulate chemokine expression in renal and immune cells, our findings establish C3a as a mediator by which IL-17A initiates infiltration of inflammatory cells during obstructive injury. Thus, complement C3 activation represents a key event for triggering the production of IL-17A during obstructive injury, thereby shaping renal microenvironments.

As the ERK signaling pathway was reported to be indispensable for C3a-mediated effector responses during kidney transplant, we defined the role of ERK signaling in IL-17A release upon C3a stimulation. The results showed that exposure to C3a led to phosphorylation of ERK, STAT3, and STAT5 and activation of NF-kB in T cells (**Figures 9A–D**). To further conformed the impact of C3a on activation of ERK signaling pathway, we knocked down endogenous C3aR in T cells by using specific small-interfering RNAs (siRNAs). As shown in **Supplement Figure 5**, two siRNAs targeting C3aR specifically knocked down endogenous C3aR protein in T cells. siRNA # 2 with higher efficiency was chosen for the subsequent studies. The results showed that T cells depleted in C3aR can inhibit phosphorylation of ERK, STAT3, and STAT5 and activation of NF-kB, which active

expression in these groups (n = 6); original magnification, ×400. Scale bar, 50µm. Quantitative analysis of interstitial fibrosis (B), a-SMA (C), and C3 positive cells (C) were shown as mean ± SEM. (D) The expression levels of C3, TGF-β1, and fibrotic markers (α-SMA, PDGFR-β, and Collagen I) were detected by Western blot. (E) The histogram shows the relative intensity for each marker normalized to GAPDH. (F) IHC staining showing F4/80 macrophages, CD3+T cells, CD4+T cells, and CD8+T cells in these groups (n = 6); original magnification, ×400. Scale bar, 50µm. (G) Quantitative analysis of F4/80, CD3, CD4, and CD8 positive cells were shown as mean ± SEM. (H) Real-time PCR showing relative renal levels of MCP-1, IL-6, IL-1β, and TNF-α in these groups (n = 6). The error bars represent the SEM. \*P<0.05; \*\*P<0.01; \*\*\*P<0.001.

the SEM. \*P < 0.05; \*\*P < 0.01; \*\*\*P < 0.001. The data were pooled from three independent experiments.

by C3a (**Figures 9E–H**). Taken together, our data indicated that C3a could induce ERK and STAT3 signaling pathway activation and promote the IL-17A production in T cells in vitro.

Combined with the data obtained from the mouse and cell experiments, the results obtained from IgAN patients further support the notion that local C3 secretion by macrophages leads to renal fibrogenic responses (**Figure 10**).

#### DISCUSSION

In human samples, we showed that circulating and local C3 was highly expressed in peripheral blood and renal tissues from patients with IgAN. Then, we observed that most of the C3 deposited in the interstitium and its expression were associated with the severity of renal interstitial fibrosis. Previous studies have predominantly investigated C3 glomerulopathy, which is defined as a kidney disease caused by complement dysregulation that results in variable glomerular inflammation (24–26). In most of the cases, C3 deposition in the glomeruli, as shown by immunofluorescence, had no immunoglobulins due to alternative pathway activation. However, further elucidation of C3 deposition in the renal interstitium is needed for a better understanding of its initiation and exacerbation. In our study, the contribution of C3 secretion by macrophages that infiltrated in the kidney appeared to be more important than C3 in the intravascular space as shown by IHC and immunofluorescence in renal tissues. (1) Immunofluorescence of human patients showed that C3 expression was more intense in macrophages

\*\*\*P < 0.001. The data were pooled from three independent experiments.

than parenchymal cells, and (2) C3 was barely expressed in renal tubules of UUO mice. We further showed that local C3 levels were highly correlated with RF. Consistent with these results, cultured mouse BMDMs under stimulations mimicking the inflammatory microenvironment, M1 and M2 all showed strongly increased secretion of C3. Furthermore, we found that C3 facilitates IL-17A production in T cells, contributing to the development of renal inflammation and fibrosis in UUO mice. In addition, blockage of C3 by Cp40 attenuated RF of UUO mice. Our proposed mechanistic network for the pathogenic role of C3 in the development of RF is also summarized in **Figure 10**.

Renal fibrosis is considered to be a common end point of various types of CKD, and its biological significance depends on the cell types contributing to collagenous and non-collagenous extracellular matrix production (27–33). The associated processes include vascular leakage, leukocyte recruitment, angiogenesis, and the appearance of myofibroblasts. Currently, most data have focused on the precursor cells of renal myofibroblasts, including circulating bone marrow-derived cells, or the transition from epithelial or endothelial cells,

pericytes, and resident fibroblasts (34–39). However, these types of transdifferentiation were all triggered by innate and adaptive immune responses through production of proinflammatory and profibrotic molecules. As a model of tubulointerstitial fibrosis, UUO model was thought best for this approach, as it is a rapid and reproducible model of RF in mice, and it mimics the main steps of tubulointerstitial fibrosis in humans (40–42). In this study, we explored the interactions between the innate and adaptive immune systems in the process of promoting RF. In the early stage during the pathogenesis of RF, macrophages are present at the affected areas (43). As an essential component of innate immunity, macrophages also regulate adaptive immune responses by recruiting other immune cells, such as neutrophils, mast cells and lymphocytes (44–49). Our results confirmed that depletion of kidney macrophages by clodronate significantly attenuated RF and function in UUO models, which has been demonstrated by previous reports.

Macrophages are highly heterogeneous cells subdivided according to their distinct functions (50, 51). In CKD, M1 macrophages are increased during early injury and inflammation and persistently surround regions of damaged tissue. Subsequently, macrophages switch to an anti-inflammatory (M2) phenotype and contribute to resolution of inflammation (43). Our evidence showed that resident kidney macrophages can secrete C3, and M1 macrophages secrete higher amounts of C3 than M2 macrophages.

Prominent roles of CD4+T cells in chronic diseases have been reported in previous studies (52, 53). Our colleagues showed that massive CD4+T lymphocyte infiltration was observed in the fibrotic kidneys of patients and UUO mice (54). After antibodies were used to deplete CD4<sup>+</sup> T cells, HE and Masson's trichrome staining results showed less inflammatory infiltrates and attenuated interstitial fibrosis in CD4+T lymphocytedepleted mice compared with UUO mice. When we depleted C3 and C3a, the infiltrating inflammatory cells, including

macrophages, CD3+T cells, and CD4+T cells, decreased in UUO mice, as demonstrated by IHC and flow cytometric analyses, indicating that CD4+T cell differentiation occurs after UUO. The data suggest that the C3a/C3aR signaling pathway.

In inflammatory kidney diseases, IL-17A producing by T lymphocytes contributes significantly to the pathogenesis of RF (55). A newly-published study has shown that IL-17 acts an inhibitory factor in TGF-β-induced renal fibroblast activation using the UUO model with IL-17−/−mice (56). The explanation for this discrepancy might be that the IL-17 cytokine family consists of six members (IL-17A–F), and the potential effect of the other IL-17 family members in renal autoimmunity and inflammation is unknown. Another study has shown that macrophages and neutrophils are major source of IL-17 in different diseases (57–60). But Tamassia noted that the human neutrophils are unable to express and produce IL-17A, IL-17B, or IL-17F in vitro (61). Our data show that IL-17A was mainly produced by CD4+Tcell, rather than F4/80+macrophages in UUO mice, which was consistent with previous reports (55, 62).

In addition, IL-17A, which is secreted by CD4+T cells, has been shown to play a major role in post-transplantation allograft rejection and in immune responses in the kidney (63–65). In the presence of TGF-β and IL-6, T cells differentiate into Th17 cells. In this regard, we observed the expression of chemokines (MCP-1, IL-1β, and IL-6) and lymphocyte infiltration in the obstructed kidneys of mice on days 7 and 14 after UUO and IL-17A secretion by CD4+T cells in the UUO mice. The production of IL-17A is negatively regulated by the anaphylatoxins, and C3a signaling elevates Th17 responses, while C5a signaling suppresses Th17 cell differentiation in experimental allergic asthma. As illustrated in **Figure 7**, we found that C3a could promote proliferation of T cells and increase IL-17A secretion by CD4+T cell and activation of the ERK and STAT3 signaling pathways in vitro. These phenomena can be suppressed by a C3aR inhibitor, suggesting that the C3a/C3aR pathway participates in the pathogenesis of UUO injury by modulating IL-17A expression, consequently promoting local inflammation and RF.

Combined with the data obtained from patient samples, relevant animal experiments and cell models, our study defines a novel mechanism by which C3 participates in renal inflammation and fibrosis. In response to injury, local C3 secretion by macrophages leads to IL-17A-mediated inflammatory cell infiltration into the kidney, which further drives fibrogenic responses. Our findings suggest that inhibition of the C3a/C3aR pathway could constitute a novel therapeutic approach for obstructive nephropathy.

#### ETHICS STATEMENT

The study was approved by the Ethical Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology Institutional (Certificate Number: IRB ID:TJ-A20160201).

# AUTHOR CONTRIBUTIONS

The experiments were conceived and designed by YyL, XL, and CZ. Experiments were performed by KW, YL, YZ, and HW and data analyzed by KW, RL, and SG. The paper was written by YyL and GX with input from all authors.

#### FUNDING

This work was partly supported by International (regional) cooperation and exchange projects, (NSFC-DFG, Grant no. 81761138041), the National Natural Science Foundation of China (No.81372244, 81572287, 81772499, and 81470948), Hubei Provincial Health and Family Planning Youth Project of China (no. WJ2015Q007), and the Major Research plan of the National Natural Science Foundation of China (Grant no. 91742204).

#### ACKNOWLEDGMENTS

The authors thank all of our colleagues working in the Department of Nephrology, Tongji Hospital of Tongji Medical College, Huazhong University of Science and Technology.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02385/full#supplementary-material

Supplement Figure S1 | Complement C3 is increased in the obstructed kidney. (A) Real-time PCR showing relative renal mRNA levels of C3, fibrotic markers (α-SMA, PDGFR-β, and Collagen I) in sham control and UUO mice. (B) IHC staining showing F4/80 and C3 protein expression in the sham control and UUO mice; original magnification, ×400. Quantitative analysis of F4/80 (C) and C3 (D) positive cells were shown as mean ± SEM. n = 6 per group, Scale bar, 50µm. (E) High magnification of C3 expression in IgA Patient and UUO mouse. The error bars represent the SEM.∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001.

Supplement Figure S2 | Macrophage depletion protects against renal fibrosis by inhibiting C3 expression. UUO mice were intravenously injected with clodronate

#### REFERENCES


liposomes and control liposomes. On days 7 and 14, mice were sacrificed, and the left kidneys were collected. The expression levels of C3 and fibrotic markers (α-SMA, PDGFR-β, and Collagen I) were detected by real-time PCR (A) and Western blot analyses (B), respectively. (C)The histogram shows the relative intensity for each marker normalized to GAPDH. n = 6 per group. The error bars represent the SEM. <sup>∗</sup>P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001.

Supplement Figure S3 | C3 deficiency reduces C3 expression and fibrosis during UUO. UUO mice were subcutaneously injected with control peptide and Cp40. On days 7 and 14, mice were sacrificed, and the left kidneys were collected. (A) Masson's trichrome staining indicates collagen deposition. IHC staining showing α-SMA and C3 protein expression in these groups (n = 6); original magnification, ×400. Scale bar, 50µm. Quantitative analysis of interstitial fibrosis (B), a-SMA(C) and C3(D) positive cells were shown as mean ± SEM. (E) The expression levels of C3, TGF-β1, and fibrotic markers (α-SMA, PDGFR-β, and Collagen I) were detected by Western blot. (F) The histogram shows the relative intensity for each marker normalized to GAPDH. n = 6 per group. The error bars represent the SEM. <sup>∗</sup>P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001.

Supplement Figure S4 | CD11b+F4/80+macrophages are not the main producer of IL-17A in kidney of UUO mice. Flow cytometric analysis of kidney cell suspensions from the obstructed kidneys injected with or without Cp40 at (A) 7 days and (C) 14 days post UUO; n = 6. Cells were stimulated in vitro with PMA/Ionomycin/Golgi-plug for 4 h. Specific staining of cell markers (anti-F4/80, anti-CD11b) and intracellular staining for IL-17A were performed. The F4/80<sup>+</sup> cells were gated. Among them, the CD11b+IL-17+cells were further gated for the analysis. Plots are gated for live CD11b+F4/80<sup>+</sup> IL-17A<sup>+</sup> macrophages; numbers indicate events in the quadrants as percentages of all gated events. (B,D) Quantifications of CD11b+F4/80<sup>+</sup> IL-17A<sup>+</sup> cells as percentages of all kidney cells isolated from the C3 blockade UUO mice and UUO mice. The error bars represent the SEM. ∗∗∗P < 0.001. The data were pooled from three independent experiments.

Supplement Figure S5 | T cells were knocked down endogenous C3aR by using three small-interfering RNAs (siRNAs). (A) The expression levels of C3aR was detected by Western blot. (B) The histogram shows the relative intensity for each marker normalized to Actin. n = 3 per group. The error bars represent the SEM. ∗∗∗P < 0.001.

of focal retinal degeneration. Invest Ophthalmol Vis Sci. (2017) 58:2977–90. doi: 10.1167/iovs.17-21672


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Liu, Wang, Liang, Li, Zhang, Zhang, Wei, Luo, Ge and Xu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Complement System in Dialysis: A Forgotten Story?

*Felix Poppelaars1 \*, Bernardo Faria1,2,3, Mariana Gaya da Costa1 , Casper F. M. Franssen1 , Willem J. van Son1 , Stefan P. Berger1 , Mohamed R. Daha1,4 and Marc A. Seelen1*

*1Department of Internal Medicine, Division of Nephrology, University Medical Center Groningen, Groningen, Netherlands, 2Nephrology and Infectious Diseases Research and Development Group, University of Porto, Porto, Portugal, 3Department of Nephrology, Hopsital Braga, Braga, Portugal, 4Department of Nephrology, Leiden University Medical Centre, Leiden, Netherlands*

Significant advances have lead to a greater understanding of the role of the complement system within nephrology. The success of the first clinically approved complement inhibitor has created renewed appreciation of complement-targeting therapeutics. Several clinical trials are currently underway to evaluate the therapeutic potential of complement inhibition in renal diseases and kidney transplantation. Although, complement has been known to be activated during dialysis for over four decades, this area of research has been neglected in recent years. Despite significant progress in biocompatibility of hemodialysis (HD) membranes and peritoneal dialysis (PD) fluids, complement activation remains an undesired effect and relevant issue. Short-term effects of complement activation include promoting inflammation and coagulation. In addition, long-term complications of dialysis, such as infection, fibrosis and cardiovascular events, are linked to the complement system. These results suggest that interventions targeting the complement system in dialysis could improve biocompatibility, dialysis efficacy, and long-term outcome. Combined with the clinical availability to safely target complement in patients, the question is not if we should inhibit complement in dialysis, but when and how. The purpose of this review is to summarize previous findings and provide a comprehensive overview of the role of the complement system in both HD and PD.

#### Keywords: complement, kidney, dialysis, hemodialysis, peritoneal dialysis

#### INTRODUCTION

An estimated 2.6 million people are treated for end-stage kidney disease (ESKD) worldwide (1). The majority of ESKD patients are dialysis-dependent. The choice between peritoneal dialysis (PD) and hemodialysis (HD) involves various determinants. Nonetheless, there is no major difference in

*Poppelaars F, Faria B, Gaya da Costa M, Franssen CFM, van Son WJ, Berger SP, Daha MR and Seelen MA (2018) The Complement System in Dialysis: A Forgotten Story? Front. Immunol. 9:71. doi: 10.3389/fimmu.2018.00071*

**Abbreviations:** AP, alternative pathway; C1-INH, C1 esterase inhibitor; C3aR, C3a-receptor; C5aR, C5a-receptor; C5aRA, C5a-receptor antagonist; CARPA, complement activation-related pseudo allergy; MCP, membrane cofactor protein; CD55, decay accelerating factor; CD59, membrane attack complex-inhibitory protein; CP, classical pathway; CR1, complement receptor 1; CR3, complement receptor 3; CRP, C-reactive protein; CV-event, cardiovascular event; DAF, decay accelerating factor; ESKD, end-stage kidney disease; HD, hemodialysis; IgG, immunoglobulin G; IgM, immunoglobulin M; IL, interleukin; LDL, low-density lipoprotein; LP, lectin pathway; MAC, membrane attack complex; MBL, mannose-binding lectin; MCP-1, monocyte chemoattractant protein-1; PD, peritoneal dialysis; sC5b-9, soluble C5b-9; sCR1, soluble complement receptor 1;

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Nicholas Rhys Medjeral-Thomas, Imperial College London, United Kingdom Michael Kirschfink, Universität Heidelberg, Germany*

> *\*Correspondence: Felix Poppelaars f.poppelaars@umcg.nl*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 29 November 2017 Accepted: 11 January 2018 Published: 25 January 2018*

#### *Citation:*

TGF-β, tumor growth factor beta; TNF-α, tumor necrosis factor alfa.

mortality between HD and PD patients (2). Although considerable progress has been made in survival rates of dialysis patients, cardiovascular morbidity and mortality remain extremely high (3). Both traditional risk factors (such as hypertension, dyslipidemia, and diabetes), as well as non-traditional risk factors (such as oxidative stress, endothelial dysfunction and chronic inflammation), contribute to the high cardiovascular risk (4). In order to lower the high morbidity and mortality rates in dialysis patients, the chronic inflammation seen in these patients must be tackled. The systemic inflammation in dialysis patients can be attributed to the (remaining) uremia, the underlying renal disease, comorbidities, and dialysis-related factors (5). The latter represents an issue that has been present in dialysis throughout history, and still remains unresolved, namely bioincompatibility.

#### BIOCOMPATIBILITY

The term "biocompatible" refers to the "capacity of a material/ solutions to exist in contact with the human body without causing a (inappropriate) host response" (6). The biocompatibility of the materials used in dialysis remains an important clinical challenge. In HD, the membrane provokes an inflammatory response, as it is the site where blood has direct contact with a foreign surface (7). Additionally, PD fluids containing high glucose levels, hyperosmolarity and acidic pH are considered biologically "unfriendly" and this lack of compatibility causes peritoneal membrane damage (8). Improving biocompatibility in HD and PD is a critical factor to ensure dialysis adequacy and enable long-term treatment (7–9). The challenge of biocompatibility is not confined to dialysis but equally important for other medical devices in contact with either tissue or blood (10). The incompatibility reaction is complex and poorly understood; however, platelets, leukocytes, the complement, and the coagulation system have been shown to be involved (11, 12). In general, incompatibility will lead to inflammation, thrombosis, and fibrosis (11–13). These events will negatively impact the clinical performance and lead to adverse events. The complement system is an important mediator of incompatibility because it can discriminate between self and non-self (14). In accordance, complement has been shown to be activated during cardiopulmonary bypass (15), low-density lipoprotein (LDL) apheresis (16), plasmapheresis (17), and immunoadsorption (18). Additionally, the complement system is also involved in biomaterial-induced complications of medical devices that are not in direct contact with the circulation, such as surgical meshes and prostheses (19, 20). Yet, it should be emphasized that the trigger by which complement is activated is different and depends on the properties of the biomaterial used (20). Proposed mechanisms of indirect complement activation include: (1) immunoglobulin G binding to the biomaterial initiating the classical pathway (CP); (2) lectin pathway (LP) activation by carbohydrate structures or acetylated compounds; or (3) activation of the alternative pathway (AP) by altered surfaces, e.g., plasma protein-coated biomaterials. In addition, complement initiators can also directly bind to the biomaterial, leading to complement activation (20). Irrespective of the pathway, complement activation always leads to the cleavage of C3, forming C3a and C3b (**Figure 1**). Increased levels of C3b result in the generation of the C5-convertase, cleaving C5 in C5a, a powerful anaphylatoxin and chemoattractant, and C5b. Next, C5b binds to the surface and interacts with C6–C9, forming the membrane attack complex (MAC/C5b-9) (14).

#### HEMODIALYSIS

Hemodialysis is a general term including several techniques such as low or high-flux HD (diffusion-based dialysis) and online haemodiafiltration (combined convective and diffusive therapy). Overall, HD remains the most-used form of renal replacement in adult ESKD patients (1). The dialysis membrane can be divided into two main groups, cellulose-based and synthetic membranes (7, 21). In the past, HD membranes were based on cuprophane (a copper-substituted cellulose) because these were inexpensive and thin-walled. The disadvantage of cellulose-based membranes was the immunoreactivity due to the many free hydroxyl-groups. Subsequently, modified cellulosic membranes were developed to improve biocompatibility by replacing the free hydroxyl-groups with different substitutions (especially acetate). The following step was the development of "synthetic" membranes, such as polyacrylonitrile, acrylonitrile-sodium methallyl sulfonate, polysulfone, polycarbonate, polyamide, and polymethylmethacrylate membranes. Nowadays, synthetic membranes are the most commonly used in clinical practice (21). The benefits of these membranes are the varying pore size and reduced immunoreactivity. The complement system is critical in the bioincompatibility of extracorporeal circulation procedures, because complement is abundantly present in blood. Moreover, innate immune activation during HD is a neglected but potentially vital mechanism that contributes to the high morbidity and mortality in these patients (4).

#### Complement Activation in HD

In the 1970s, HD was already known to affect the complement system (22). Several studies have since then looked at complement activation during HD, the complement pathway responsible and additional mechanisms contributing to complement activation. In the past, an important adverse event in dialysis was the "first-use syndrome," named after the fact that these reactions were most severe with new dialyzers. This incompatibility reaction was the result of complement activation by the membrane and closely resembles the pseudo-anaphylactic clinical picture that is nowadays known as complement activation-related pseudoallergy (CARPA) (23, 24). Furthermore, these early studies provided important information on the kinetics of complement activation. During HD, C3 activation, resulting in the generation of C3a, peaks during the first 10–15 min, whereas terminal pathway activation, resulting in C5a and C5b-9 formation occurs at a later stage of dialysis (25). Over the past decades, membranes have been developed with improved biocompatibility. Nonetheless, even with modern "biocompatible" HD membranes significant complement activation still occurs (23, 26, 27). During a single HD session soluble C5b-9 (sC5b-9) levels and C3d/C3-ratios in the plasma increase up to 70% (23, 26). Yet, this is most likely an underestimation of the amount of complement activation, since these values represent fluid phase activation. Complement activation takes place in the plasma (the fluid phase), but also on

Figure 1 | The complement system. A schematic view of activation of the complement system and its regulation. The classical pathway (CP) is initiated by C1q binding to immune complexes or other molecules (e.g., CRP), thereby activating C1r and C1s resulting in the cleavage of C2 and C4 thereby forming the C3-convertase (C4b2b). The lectin pathway (LP) is initiated by mannose-binding lectin (MBL), ficolins, or collectin-11 binding to carbohydrates or other molecules (e.g., IgA), thereby activating MASP-1 and MASP-2, forming the same C3-convertase as the CP. Subsequently, the C3-convertase cleavages C3 into C3a and C3b. Activation of the alternative pathway (AP) occurs *via* properdin binding to certain cell surfaces (e.g., LPS) or by spontaneous hydrolysis of C3 into C3(H2O). Next, binding of factor B creates the AP C3-convertase (C3bBb). Increased levels of C3b results in the formation of the C5-convertases, which cleaves C5 in C5a, a powerful anaphylatoxin, and C5b. Next, C5b binds to the surface and interactions with C6–C9, generating the membrane attack complexes (MAC/C5b-9). Several complement regulators (either soluble and membrane-bound) prevent or restrain complement activation. C1 esterase inhibitor (C1-INH) inhibits the activation of early pathway activation of all three pathways, while C4b-binding protein (C4BP) controls activation at the C4 level of the CP and LP. Factor I and factor H regulate the C3 and C5-convertase. Furthermore, the membrane-bound inhibitors include complement receptor 1 (CR1), membrane cofactor protein (MCP) that acts as an co-factors for factor I and decay accelerating factor (DAF) which accelerates the decay of C3-convertases. The membrane-bound regulator Clusterin and CD59 prevents the generation of the C5b-9.

surfaces (the solid phase) (14). Fittingly, in addition to fluid phase activation, complement depositions have also been shown on the surface of the HD membranes (28).

Different studies have tried to dissect the pathway responsible for complement activation in HD. Early evidence emerged from a study by Cheung et al., demonstrating AP activation by cellulose membranes (29). Initially, the involvement of the CP or LP was excluded, since it was reported that plasma C4d concentrations remained unaffected during HD (30). However, others were able to show C4 activation by cellulose membranes (31, 32). The increase in C4d levels correlated with the rise in C3d levels, implying that the CP or LP is (at least partly) responsible for the complement activation seen in HD (32). More recently, a role for the LP was demonstrated in complement activation by polysulfone membranes (33, 34). An elegant study by Mares et al., using mass spectrometry, showed a 26-fold change in eluate-toplasma ratio for ficolin-2 (previously called L-ficolin), suggesting preferential adsorption by the membrane (33). A follow-up study using proteomics analysis of dialyzer eluates revealed that C3c, ficolin-2, mannose-binding lectin (MBL) and properdin were most enriched (28). In addition, plasma ficolin-2 levels decreased by 41% during one HD session, corresponding with the excessive adsorption to the membrane. The decrease in plasma ficolin-2 levels was associated with C5a production and leukopenia during HD (28). The adsorption of properdin to the dialyzer, confirms earlier studies regarding AP activation by HD (28, 29). To summarize, the principal mechanism of complement activation in HD is the binding of MBL and ficolin-2 to the membrane, resulting in LP activation; while, simultaneously, properdin and/or C3b bind to the membrane resulting in AP activation (**Figure 2**). The latter is supported by the evidence that in C4-deficient patients, systemic complement activation and C3b deposition on the HD membrane are reduced during dialysis but not abolished (31). These results show the importance of the LP, while demonstrating the crucial contribution of the AP.

A second mechanism that could modulate complement activation during HD is the loss of complement inhibitors *via* absorption to the membrane. In HD, polysulfone membranes

were shown to absorb factor H and clusterin (28, 33). Factor H is an important inhibitor of C3, while clusterin prevent terminal pathway activation thereby stopping the formation of C5a and C5b-9 (**Figure 1**) (14). The loss of these inhibitors would cause dysregulation of the AP, leading to further complement activation in the fluid phase (i.e.*,* in the circulation) in HD patients.

#### Effector Functions and Clinical Implications of Complement Activation

Complement activation will lead to the generation of effector molecules, which can result in a variety of biological responses (14). In HD, the most important effector functions of complement activation are the induction of inflammation, promoting coagulation and impaired host defense due to accelerated consumption of complement proteins (20, 35, 36).

The generation of C3a and C5a during HD promotes recruitment and activation of leukocytes (37, 38). Leukocyte activation results in the oxidative burst and the release of pro-inflammatory cytokines and chemokine's, such as interleukin (IL)-1β, IL-6, IL-8, tumor necrosis factor-α, monocyte chemoattractant protein-1, and interferon-γ. More specifically, the activation of PMNs by C5a leads to the release of granule enzymes such as myeloperoxidase and elastase (39–41). Furthermore, complement activation in HD patients results in the upregulation of adhesion molecules on leukocytes, especially complement receptor 3 (CR3). The C5a-activated leukocytes will then bind C3 fragments (iC3b) deposited on the membrane *via* CR3, leading to leukopenia (20, 28, 39). Likewise, CR3 on PMNs is also important for the formation of platelet–PMN complexes, which can contribute to both inflammatory and thrombotic processes (42). The crosstalk between activation of the complement and coagulation system has correspondingly been described in HD. It has been demonstrated that C5a generation during HD leads to the expression of tissue factor and granulocyte colony-stimulating factor in PMNs, shifting HD patients to a procoagulative state (35). In conformity, plasma C3 levels have been shown to positively correlated with a denser clot structure in HD patients (43). On the other hand, the coagulation system has also been shown to impact complement activation (44).

Inflammation and coagulation are principally involved in the pathogenesis of cardiovascular disease. Accordingly, complement has been associated to the susceptibility to cardiovascular disease in HD patients (26, 27, 45–47). Plasma C3 levels, prior to a HD session, were found to be higher in patients who develop a cardiovascular event (CV-event) than HD patients who remained event-free. Moreover, an association was found between C3 levels and the development of CV-events (27). A similar trend of higher C3 levels in HD patients who develop a CV-event was seen in our study (26). A possible explanation would be that higher C3 levels prior to HD might reflect the potential for HD-evoked complement activation. Additionally, another association was found for baseline sC5b-9 levels with the occurrence of CV-events as well as mortality. This association was complex and showed an U-shaped relationship, indicating that both high and low sC5b-9 levels led to a higher risk, whereas HD patients with mid-range values were protected (27). Furthermore, a common factor H gene polymorphism was found to be an independent predictor of cardiovascular disease in HD patients (47). Homozygous HD patients for the Y402H polymorphism had an odds ratio of 7.28 for the development of CV-events compared to controls. This polymorphism affects the binding sites for heparin and C-reactive protein (CRP) and it has, therefore, been hypothesized that the reduced binding of factor H to the patient's endothelial cells would increase their risk of a CV-event. Alternatively, the link between the factor H polymorphism and the cardiovascular risk in HD patients could be mediated through CRP, since factor H binds CRP and thereby undermines its pro-inflammatory activity (48, 49). The Y402H polymorphism of factor H results in inadequate binding to CRP and thus leaves the pro-inflammatory activity of CRP unchecked. Furthermore, several studies have demonstrated that CRP levels in HD patients are associated to cardiovascular mortality (50–52). *Buraczynska* et al. revealed that in HD patients the complement receptor 1 (CR1) gene polymorphism C5507G is independently associated with the susceptibility for cardiovascular disease (46). Whether this effect is mediated *via* the complement inhibitory capacity of CR1 or *via* the recently discovered function of CR1 in the binding and clearance of native LDL remains to be elucidated (53). Another study showed that low serum C1q-adiponectin/ C1q ratios were linked to cardiovascular disease in HD patients (45). The mechanism behind this connection is not understood but it has been demonstrated that adiponectin protects against activation of C1q-induced inflammation (54). Thus, in HD patients increased complement activation, as well as increased complement activity and the loss of complement inhibitors have all been linked to a higher risk of cardiovascular disease (**Table 1**). Recently, our group showed that low MBL levels are also associated with the occurrence of cardiovascular disease in HD patients (26). The higher risk in these patients was attributed to CV-events linked to atherosclerosis. In support of this, low MBL levels have been linked to enhanced arterial stiffness in HD patients (55). Accordingly, Satomura et al. demonstrated that low MBL levels were an independent predictor of all-cause mortality in HD patients (56). We, therefore, postulate that in HD patients, low MBL levels promote cardiovascular disease by enhancing atherosclerosis due to the inadequate removal of atherogenic particles.

In HD patients, little is known about the changes in complement components overtime. The plasma levels of C3 have been shown to decrease after 12 months compared to baseline (27). In this study, the C3 levels also negatively correlated with the dialysis vintage. In addition, the ability to activate complement has also been shown to be decreased in HD patients compared to healthy controls (23). In theory, these acquired deficiencies of complement proteins could explain the higher infection and sepsis risk seen in HD patients. Conversely, there was no association between low MBL levels and the risk of infection in HD patients (57). However, the authors concluded that this might be due to a compensation mechanism of higher ficolin-2 and MASP-2 levels in MBL-deficient individuals. Furthermore, another study found that long-term HD patients have decreased levels of clusterin, factor B and factor H compared to short-term HD patients (58). Thus far, no study has analyzed the link between HD-acquired complement deficiencies and infection risk. The clinical consequences of the HD-induced ficolin-2 reduction would be the most interesting to examine (28, 33). It is highly likely that this reduction would have a tremendous impact on HD patients' health and outcome. A genetic deficiency in ficolin-2 has not been reported to date, highlighting the essential function of this component within host defense. In conformity, ficolin-2 has been shown to be involved in the elimination of numerous pathogens (59).

#### Therapeutic Options

Several types of interventions have been proposed or tested in HD patients to decrease inflammation or target cardiovascular risk factors with mixed success. Hence, the clinical need for better therapeutic options that limit the inflammation and decrease cardiovascular risk in HD patients is on-going. The complement system is considered to be a promising target during HD to limit the inflammation and decrease cardiovascular risk (60). Therapies modulating HD-induced complement activation have focused on three treatment strategies: (1) reduction in the complement activating-capacity of the HD membrane; (2) the use of non-specific complement inhibitors (e.g., anticoagulants with a complement inhibitory property); and (3) specific complementdirected therapies.

Prevention is better than cure; therefore, creating a truly biocompatible membrane would, therefore, be ideal to prevent complement activation during HD. Much progress has been made with the development of more biologically compatible membranes by surface modifications and reducing protein retention. Today, the most common HD membranes contain sulfonylgroups (7). To further improve biocompatibility, it is vital to understand the structures that initiate complement activation as it has the potential to develop HD membranes with enhanced biocompatibility. In modern HD membranes, ficolin-2 seems to


*a Data are presented as hazard or OR plus 95% confidence interval.*

*OR, odds ratio; HR, hazard ratio; HD, hemodialysis; MBL, mannose-binding lectin; CR1, complement receptor 1; CV-event, cardiovascular event; sC5b-9, soluble C5b-9.*

be an important mediator in HD-induced complement activation (28, 33). Ficolin-2 is unfortunately a highly promiscuous molecule with numerous binding partners, several of which are acetylated compounds (59).

Anticoagulants have been used extensively to render biomaterial-blood incompatibility, through inhibition of the coagulation, contact and complement system. The effect of citrate anticoagulation on complement activation has widely been studied in HD. Citrate has calcium-chelating properties and thereby reduces complement activation (61, 62). During the initial phase of HD with cellulose membranes, citrate anticoagulation reduced C3a levels by almost 50% compared to heparin (63). However, no complement inhibition was seen by citrate anticoagulation during HD in other studies with cellulose or synthetic membranes (64–66). Heparinoids are also known to prevent complement activation, although this inhibition is strictly concentration dependent (67). Although heparin has been tested extensively in HD, sadly none of these studies determined the effect on complement activation.

In the past decade, numerous complement inhibitors have been developed; two are currently used in the clinics and others are now undergoing clinical trials. Purified C1 esterase inhibitor (C1-INH) is a protease that is clinically used to treat hereditary angioedema. Eculizumab, a C5 antibody is used for the treatment of paroxysmal nocturnal hemoglobinuria and atypical hemolytic uremic syndrome (14, 68). In HD, specific complement-directed therapies have predominantly been evaluated in experimental settings, still valuable information has been uncovered and shown that the use of complement inhibitors are a promising tool to reduce the inflammatory response and subsequent consequences in these patients (60). The potential of complement inhibition in HD is further underlined by the successful use of complement inhibitors for biomaterial-induced complement activation in cardiopulmonary bypass systems (19). In patients undergoing cardiopulmonary bypass surgery, treatment with soluble CR1 (sCR1/TP30), an inhibitor of C3, lead to a decrease in mortality and morbidity as well as a reduced need for intra-aortic balloon pump support (69). Consequently, soluble complement inhibitors may be equally effective in HD, since there is the recurrent need of complement inhibition for short periods. Specifically, the short half-life of sCR1 matches the need for restricted complement inhibition in HD, which is only needed during dialysis, after which complement activity should be reestablished between sessions. This approach would also prevent complications of long-term immunosuppression. In a pre-clinical monkey model of HD, another C3-inhibitor (compstatin) was used to attenuate HD-induced complement activation (70). Despite the use of HD membranes with high biocompatibility and standard heparin treatment in their study, severe complement activation still occurred in monkeys. In this study, animals received a bolus injection prior to the HD and a continuous infusion of compstatin during the 4 h HD procedure. Treatment completely blocked complement activation and C3 activation products stayed at basal levels throughout the HD session. Strikingly, a second treatment regimen with only a bolus injection of compstatin at the start of the session was also sufficient to abolished complement activation throughout the procedure. Furthermore, complement inhibition lead to the increase of IL-10, an anti-inflammatory cytokine. Unfortunately, the effect of complement inhibition on other inflammatory markers could not be assessed, since one HD session was insufficient to induce substantial levels of pro-inflammatory cytokines. Next to inhibition of the central component C3, blockage of early complement components may be equally successful. C1-INH forms a therapeutic option, since HD leads to LP activation and C1-INH could attenuate this (67). Additionally, C1-INH also affects the coagulation and contact system, which could add to the success of this therapeutic approach. Given the strong involvement of complement activation effector molecules in HD, more specifically C5a, another attractive option would be the inhibition of C5 or C5a-receptor antagonists (C5aRA) (35). This could be either done by the anti-C5 antibody or by C5aRA. Eculizmab blocks the generation of C5a and C5b-9 and could thus be more effective than C5aRA. However, the long half-life and the high costs form important disadvantages. In contrast, C5aRA tends to be more cost-effective (71). These drugs could significantly reduce activation of leukocytes and thereby inflammation in HD. Currently, the most likely candidate to be used in HD is PMX-53, a C5aRA, since this compound is currently tested in different clinical trials (72). Another promising approach is coating biomaterials with complement inhibitors (20). One of these molecules, the 5C6 peptide is a molecule that has strong binding affinity toward factor H without modifying its inhibitory activity. More importantly, polystyrene surfaces coated with 5C6 were shown to bind factor H and thereby prevent complement activation when exposed to human plasma, thus enhancing biocompatibility (73). However, it is unknown whether the reduction of systemic factor H levels by 5C6 during HD could have undesirable consequences, such as seen in factor H-deficient individuals. Finally, the cost of the different complement inhibitors should be taken into account, considering the high frequency of treatments required in HD patients.

#### PERITONEAL DIALYSIS

Peritoneal dialysis is the most common used dialysis technique at home and is equally effective as HD for the treatment of CKD (74). Nevertheless, the advantages of PD include; better preservation of residual renal function, lower infectious risk and higher satisfaction rates. Despite the good results seen with PD, this dialysis technique remains underused (1). In PD, unlike in HD, no synthetic membrane is used. In contrast, the peritoneum in the abdominal cavity of the patients acts as a semi-permeable membrane allowing diffusion between the dialysis fluid and the circulation. The osmotic gradient during PD is based on high glucose levels in the dialysate. However, glucose acts as a double edge sword, since it serves as an osmotic agent but it is also responsible for the incompatibility reaction. The peritoneal membrane is made up of an inner mesothelial layer and these cells are, therefore, directly in contact with the dialysis fluid. Long-term exposure to dialysate leads to tissue remodeling of this layer resulting in peritoneal fibrosis (75). This progressive fibrosis forms a major limitation for chronic PD treatment. Another common complication in PD is peritonitis (76). Patients who develop peritonitis can have irreversible peritoneum damage, PD failure and significant morbidity or even mortality. For this reason, avoiding PD failure due to peritonitis or fibrosis remains a challenge for nephrologists (77).

#### Complement Activation in PD

The link between the complement system and PD seems less obvious, because there is no direct contact with blood. However, mesothelial cells produce and secrete different complement factors, including C4, C3, and C5 till C9 (78, 79). In accordance, different studies have found the presence of complement in the peritoneal dialysate. Additionally, the amount of C3 in the PD fluid does not depend on the serum concentration, suggesting that the C3 originates from local production (80). The study by Oliveira et al. found strong protein abundance of Factor D in six adult PD patients (81), whereas a similar approach in 76 PD patients by Wen et al. found significant protein expression of C4 and C3 only (82). Altogether, proteomic analyses of the dialysate of healthy PD patients has revealed the presence of C4, C3, Factor B, Factor D, Factor H, Factor I, and C9 (81–85). Proteomic profiling in the peritoneal fluid of children identified a total number of 189 proteins, of which 18 complement components (84). The discrepancies between the various proteomic studies could be explained by differences in the underlying cause of renal failure, since diabetic patients on PD have been shown to have lower levels of C4 in the dialysate compared to controls (83). Obviously, other patient's characteristics such as ethnicity and differences in the accuracy and sensitive of the analysis have to be taken into account as well. Complement production by mesothelial cells has been shown to be increased in uremic patients and it can be further stimulated upon exposure to PD solutions containing glucose (78, 79). Next to complement production; mesothelial cells also express important complement regulators; e.g., MCP, DAF, and CD59 (79, 80).

Systemically, PD patients have lower MBL levels compared to HD patients and healthy controls, even after adjusting for the effect of mutations (86). This could indicate loss of systemic MBL *via* the peritoneal route, independent of the reduced renal function. However, MBL has so far not been assessed in peritoneal dialysates. Furthermore, serum levels of C1q, C4, C3d, factor D, and properdin were shown to be higher in pediatric PD patients compared to healthy controls, however, not in comparison to patients with ESKD (87). Overall, the higher plasma levels of the complement components are likely caused by increased synthesis by the liver due to the pro-inflammatory state in ESKD patients. Moreover, the increased levels of C3d in PD patients are believed to be the consequence of reduced elimination of factor D by the kidney, creating enhanced AP activation. However, while systemic complement activation (the fluid phase) is similar between PD patients and patients with ESKD, higher intravascular complement depositions (solid phase) have been shown in children with PD compared to non-PD children with ESKD. Omental and parietal arterioles from PD patients demonstrated a higher presence of C1q, C3d, and C5b-9 (88).

Evidence has also been provided for complement activation in the peritoneal cavity in PD patients (80, 89). Previously, it was demonstrated that the dialysate/serum ratios of factor D and C3d were elevated in PD, whereas the dialysate/serum ratios of C3, C4, and properdin were decreased (89). The high dialysate levels of C3d demonstrate local complement activation, while the comparatively low dialysate/serum ratios of complement components are likely caused by intraperitoneal complement consumption. In accordance, the presence of sC5b-9 in the peritoneal dialysate has also been shown. In the dialysate of PD patients, sC5b-9 levels up to 200 pg/µg of total protein level have been reported (80). Considering the high molecular weight of sC5b-9 (>1,000 kDa), it is very likely that the sC5b-9 in the dialysate is produced in the peritoneal cavity and does not originate from the circulation.

One of the proposed mechanisms of complement activation in PD patients is that PD therapy modifies the expression of complement regulators on the peritoneal mesothelium, leading to local complement activation (**Figure 3**). In accordance, CD55 expression is lower on mesothelial cells from PD patients than non-CKD patients and the reduced expression of CD55 is accompanied by higher peritoneal levels of sC5b-9 (80). Likewise, complement regulators were also shown to be downregulated in arterioles of PD patients. Furthermore, the C5b-9 deposition seen in the arterioles of PD patients correlated with the level of dialytic glucose exposure (88). However, this is probably not the only mechanism responsible for complement activation in PD patients. Hypothetically, cellular debris as a result of direct peritoneal damage by bioincompatible PD fluids as well as antibodies against microorganisms could contribute to local complement

activation during PD. Unfortunately, most of the reviewed studies are relatively old and there is, therefore, a need for novel studies to assess the effect of newer PD solutions on complement production and activation.

#### Effector Functions and Clinical Implications of Complement Activation

During PD, complement activation occurs locally within the peritoneal cavity and leads to the generation of opsonins, anaphylatoxins, and the MAC. The effects of complement activation during PD include the induction of tissue injury, inflammation, coagulation, and fibrosis. However, complement activation in PD patients has also been linked to long-term effects such as cardiovascular risk (88). In different experimental models, complement

consequently peritoneal fibrosis.

activation during PD leads to direct damage of the peritoneum. The complement-induced peritoneal damage seems to be mediated *via* activation of the terminal pathway, specifically C5a and C5b-9 (90–92). Additionally, complement activation leads to inflammation. In a rat model of peritoneal fluid infusion, the numbers of neutrophils increased significantly overtime, and this process was largely dependent on C5 activation. In conformity, intraperitoneal injections with C3a and C5a in mice lead to the influx of leukocytes, predominantly neutrophils (93). The effect of C5a is mediated *via* C5aR1, while the effect of C3a is presumably mediated *via* the C3a-receptor. The crosstalk between activation of the complement and coagulation system has also been described in PD. Thrombin anti-thrombin complexes increased significantly in experimental models of PD and this process was partly dependent on C5 activation (92). Mizuno et al. showed that intraperitoneal complement activation leads to fibrin exudation on the surface of the injured peritoneum (94). Altogether these findings indicate that activation of the coagulation system by the PD therapy is at least (partly) complement dependent. The fibrin exudate can also be a sign of PD-associated fibrosis.

The link between fibrosis and complement is relatively new; nevertheless, recent evidence suggests that complement activation promotes the progression to fibrosis after tissue injury (95). In PD, high peritoneal transport is associated with progression of peritoneal fibrosis (96). Proteomics analysis of PD fluid showed enhanced expression of C3 in patients with high transporter status, while expression of C4 is lower in low transporters (82, 97). Furthermore, in PD mesothelial cells undergo epithelialto-mesenchymal transition, resulting in the accumulation of myofibroblasts and consequently peritoneal fibrosis (98). In other disease models, complement has been shown to induce epithelial-to-mesenchymal transition (99). This effect is mediated *via* the C5aR1, since in rodent models of infection–induced peritoneal fibrosis C5aR1<sup>−</sup>/<sup>−</sup> mice were protected against fibrosis (100). The C5aR1 is also involved in the production of profibrotic and inflammatory mediators by peritoneal leukocytes (100). In addition, Bartosova et al. reported that in the peritoneal arterioles of PD patient's, high abundance of complement deposition was found to correlate with TGF-b signaling (88). More specifically, C1q and C5b-9 deposition were associated with an increased phosphorylation of SMAD2/3, and enhanced vasculopathy. Interestingly, the TGF-b–SMAD pathway has also been recently linked to cardiovascular disease (101). Encapsulating peritoneal sclerosis is another long-term complication of PD, which is the result of abnormal thickening and fibrosis of the peritoneum, leading to a fibrous cocoon thereby encapsulating the intestines causing obstruction (102). The exact cause of this rare complication is unknown, but it is linked to the bioincompatibility of the glucose-based PD solutions (103). The bioincompatibility of these solutions presumably promotes the expression TGF-b, thereby stimulating the transition of mesothelial cells to myofibroblasts. Recently, a prospective proteomics study identified complement components as a possible biomarker of encapsulating peritoneal sclerosis (85). Factors B and factor I were elevated in the PD fluid of patients up to 5 years prior to developing encapsulating peritoneal sclerosis. In patients with stable membrane function, factor I was present in the PD fluid in lower amounts and decreased overtime, while factor B was barely detectable in the PD fluid of controls. However, whether the elevated levels of these complement factors are merely an acute phase response or involved in the pathogenesis remains to be investigated. Yet, based on the current literature, complement activation is likely to play a role in the mechanisms of peritoneal fibrosis. Nevertheless, additional studies are needed to further elucidate the specific role of the complement system in this process.

Peritonitis is another common complication with significant morbidity and mortality. Complement has been proposed to be involved in the risk of PD patients for peritonitis. First, a variation in the FCN2 gene was shown to be more prevalent in PD patients with a history of peritonitis (104). In addition, local activation will lead to a further decline of already low levels of complement components in PD fluid and may thereby additionally impair host defense. Complement activation products have also been suggested as a biomarker during peritonitis. Mizuno et al. showed that C4, C3, and sC5b-9 levels in the peritoneal fluid are significantly higher in PD patients with poor prognosis after peritonitis (105). Complement markers in peritoneal fluid have, therefore, the potential to serve as a biomarker for the prediction of the prognosis of PD-related peritonitis. Finally, the risk of peritonitis could form a major Achilles heel for complement inhibition in PD.

#### Therapeutic Options

Treatment aimed at attenuating or blocking complement activation in PD has mostly focused on the terminal pathway. The advantage of this approach is the elimination effector functions of C5a and/or C5b-9, while proximal complement functions stay intact. *In vitro*, inhibition of the C5aR1 on peritoneal leukocytes, isolated from PD fluid, reduced bacteria-induced profibrotic (TGF-β) and inflammatory (IL-6 and IL-8) mediator production (100). In addition, the systemic administration of a C5aR1 antagonist in a rat model of PD prevented influx of inflammatory cells and reduced tissue damage of the peritoneal cavity (91). Furthermore, blockage of C5 in PD improved ultrafiltration and additionally reduced activation of the blood clotting system (92). Other studies have confirmed these results; showing that C5 blockade significantly increased the ultrafiltration volume *via* reduced peritoneal glucose transport, most likely by preventing C5a-induced vasodilatation (106). In contrast, C3 inhibition through complement depletion by cobra venom factor, also led to diminished chemoattractant release, neutrophil recruitment and enhanced ultrafiltration (106). Anticoagulants have also been tested for the treatment of the inflammatory reaction to PD fluids (106, 107). The addition of low-molecular-weight heparin to the PD fluid not only prevented thrombin formation but also inhibited the complement activation, neutrophil recruitment, and improved ultrafiltration (107). In brief, results about complement inhibition in PD look promising, but many hurdles remain to be solved.

#### CONCLUSION

In conclusion, biocompatibility remains an important clinical challenge within dialysis. Due to bioincompatibility, complement is systemically activated during HD, while PD leads to local complement activation. Moreover, important effector functions of complement activation include promoting inflammation and coagulation. In addition, long-term complications of dialysis, such as infection, fibrosis, and cardiovascular events, are linked to the complement system. These results indicate the possibility for complement interventions in dialysis to improve biocompatibility, dialysis efficacy, and long-term outcome.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

FP and MG performed the literature search. MD, SB, and MS helped with the interpretation of the literature. BF and CF provided the review with clinical information and the clinical relevance. FP, BF, and MG wrote the review. WS, CF, SB, MD, and MS critically reviewed the manuscript prior to submission.


collagen-induced arthritis in mice. *Biochem Biophys Res Commun* (2009) 378:186–91. doi:10.1016/j.bbrc.2008.11.005


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Poppelaars, Faria, Gaya da Costa, Franssen, van Son, Berger, Daha and Seelen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Intradialytic Complement Activation Precedes the Development of Cardiovascular Events in Hemodialysis Patients

Felix Poppelaars <sup>1</sup> \* † , Mariana Gaya da Costa1†, Bernardo Faria1,2,3, Stefan P. Berger <sup>1</sup> , Solmaz Assa<sup>4</sup> , Mohamed R. Daha1,5, José Osmar Medina Pestana<sup>6</sup> , Willem J. van Son<sup>1</sup> , Casper F. M. Franssen<sup>1</sup> and Marc A. Seelen<sup>1</sup>

#### Edited by:

Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France

#### Reviewed by:

Conrad Anthony Farrar, King's College London, United Kingdom Christina Karatzaferi, University of St Mark St John, United Kingdom

\*Correspondence: Felix Poppelaars felix.poppelaars@gmail.com

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 04 June 2018 Accepted: 21 August 2018 Published: 13 September 2018

#### Citation:

Poppelaars F, Gaya da Costa M, Faria B, Berger SP, Assa S, Daha MR, Medina Pestana JO, van Son WJ, Franssen CFM and Seelen MA (2018) Intradialytic Complement Activation Precedes the Development of Cardiovascular Events in Hemodialysis Patients. Front. Immunol. 9:2070. doi: 10.3389/fimmu.2018.02070 <sup>1</sup> Division of Nephrology, Department of Internal Medicine, University of Groningen, University Medical Center Groningen, Groningen, Netherlands, <sup>2</sup> Nephrology and Infecciology Group, INEB/I3S, University of Porto, Porto, Portugal, <sup>3</sup> Department of Nephrology, Hospital Braga, Braga, Portugal, <sup>4</sup> Department of Cardiology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands, <sup>5</sup> Department of Nephrology, University of Leiden, Leiden University Medical Center, Leiden, Netherlands, <sup>6</sup> Nephrology Division, Federal University of São Paulo, São Paulo, Brazil

Background: Hemodialysis (HD) is a life-saving treatment for patients with end stage renal disease. However, HD patients have markedly increased rates of cardiovascular morbidity and mortality. Previously, a link between the complement system and cardiovascular events (CV-events) has been reported. In HD, systemic complement activation occurs due to blood-to-membrane interaction. We hypothesize that HD-induced complement activation together with inflammation and thrombosis are involved in the development of CV-events in these patients.

Methods: HD patients were followed for the occurrence of CV-events during a maximum follow-up of 45 months. Plasma samples were collected from 55 patients at different time points during one HD session prior to follow-up. Plasma levels of mannose-binding lectin, properdin and C3d/C3 ratios were assessed by ELISA. In addition, levels of von Willebrand factor, TNF-α and IL-6/IL-10 ratios were determined. An ex-vivo model of HD was used to assess the effect of complement inhibition.

Results: During median follow-up of 32 months, 17 participants developed CV-events. In the CV-event group, the C3d/C3-ratio sharply increased 30 min after the start of the HD session, while in the event-free group the ratio did not increase. In accordance, HD patients that developed a CV-event also had a sustained higher IL-6/IL-10-ratio during the first 60 min of the HD session, followed by a greater rise in TNF-α levels and von Willebrand factor at the end of the session. In the ex-vivo HD model, we found that complement activation contributed to the induction of TNF-α levels, IL-6/IL-10-ratio and levels of von Willebrand factor.

Conclusions: In conclusion, these findings suggest that early intradialytic complement activation predominantly occurred in HD patients who develop a CV-event during follow-up. In addition, in these patients complement activation was accompanied

**484**

by a pro-inflammatory and pro-thrombotic response. Experimental complement inhibition revealed that this reaction is secondary to complement activation. Therefore, our data suggests that HD-induced complement, inflammation and coagulation are involved in the increased CV risk of HD patients.

Keywords: complement, kidney, cardiovascular risk, hemodialysis, biocompatibility, innate immunity, C1-inhibitor

#### INTRODUCTION

Renal replacement therapy (RRT) represents a cornerstone in the treatment of patients with end stage renal disease (ESRD). Hemodialysis (HD) remains the most common form of RRT (1). Despite being lifesaving, HD comes with a risk (2). The life expectancy and quality of life of patients on dialysis is inferior to the general population. Overall, HD has been associated with increased cardiovascular morbidity and mortality (3). Previous studies have suggested that the innate immune system plays a key role in the development of cardiovascular disease in HD patients (4).

The complement system is a major component of innate immunity and activation of this system induces an inflammatory response (5). Complement activation can occur via three pathways: the classical pathway (CP), lectin pathway (LP), and alternative pathway (AP). Regardless of the trigger, all pathways lead to the cleavage of C3 resulting in the formation of C3b, the large fragment and C3a, an anaphylatoxin. Ultimately, C3b is broken down progressively to iC3b and then to the more stable fragment C3d. The functions of the complement system were thought to be limited to opsonization and lysis of pathogens. However, nowadays this system is known to have numerous functions and complement has been shown to be involved in the pathogenesis of various diseases (6).

For decades, HD has been known to activate the complement system (7). In dialysis, complement activation is mainly caused by the interaction of blood with the HD membrane (4). Regardless of the efforts to improve biocompatibility, complement activation still occurs in HD, even with modern membranes (8–10). It has been hypothesized that complement activation leads to HDinduced inflammation and thereby increases the subsequent cardiovascular risk (4). In accordance, several studies have shown an association between complement and cardiovascular events (CV-event) (8, 11–14). However, the link between complement activation products and CV-events remains poorly characterized (15). Only Lines et al. reported an association in HD patients between soluble C5b-9 and cardiovascular risk (15). Furthermore, previous experimental studies proposed a link between HD-induced complement activation, pro-inflammatory cytokines, and the coagulation system (10, 16).

We hypothesize that an unfavorable complement profile is seen in HD patients who will develop a CV-event. To investigate the mechanism of increased cardiovascular risk in HD, we measured complement activation, pro-inflammatory cytokines and pro-thrombotic factors during one HD session in patients that developed a CV-event during follow-up and compared this to patients without a CV-event during follow-up. Furthermore, we used an ex-vivo model of HD to further elucidate the role of complement activation as a trigger for inflammation and coagulation in HD.

#### MATERIALS AND METHODS

#### Study Population and Design

A cohort of 55 hemodialysis patients from Dialysis Center Groningen and the University Medical Center Groningen were followed for a maximum of 45 months. The original cohort was composed out of 109 patients; however, due to a lack of samples only 55 patients could be included for this study. The protocol has been previously described (2). In short, patients were included if the duration of HD therapy was longer than 3 months. Patients with severe heart failure (NYHA class IV) were excluded. Patient characteristics were extracted from patient records.

#### Dialysis Settings

Patients were on maintenance HD treatment for three times a week with a low-flux polysulfone hollow-fiber dialyzer (F8; Fresenius Care, Bad Homburg, Germany). The hemodialysis sessions lasted for 4 h. The blood and dialysate flow rates were 250–350 and 500 mL/min, respectively. A constant ultrafiltration rate was used. Dialysate composition was as follows: acetate, 3.0 mmol/L; bicarbonate, 34 mmol/L; calcium, 1.5 mmol/L; chloride, 108 mmol/L; glucose, 1.0 g/L; magnesium, 0.5 mmol/L; potassium, 1.0 or 2.0 mmol/L; sodium, 139 mmol/L The dialysate temperature was kept on 36.0 or 36.5◦C. Blood samples were taken just before the start of the dialysis session, and after 30, 60, 180, and 240 min.

#### Definition of Endpoint

The end-point of the study was defined as the time to the first CVevent. CV-events included cardiac, cerebrovascular, or peripheral vascular events. Occurrence of a cardiac event was defined as a ischemic heart disease (unstable angina pectoris, myocardial infarction, Coronary Artery Bypass Grafting (CABG) and/or Percutaneous Coronary Intervention (PCI), sudden cardiac death and congestive heart failure. In order to classify as acute myocardial infarction, two out of the following three criteria had to be present: clinical status, elevated heart enzymes, and EKG changes. Cerebrovascular events were defined as stroke,

**Abbreviations:** AP, Alternative pathway; BSA, Body surface area; C5b-9, Membrane attack complex; CV-event, Cardiovascular event; ESRD, End stage renal disease; ELISA, Enzyme-linked immunosorbent assay; HD, Hemodialysis; IL-6, Interleukin 6; IL-10, Interleukin 10; LP, Lectin pathway; MBL, Mannosebinding lectin; RRT, Renal replacement therapy; PCI, Percutaneous coronary intervention; TNF-α, Tumor necrosis factor-α; vWF, Von Willebrand factor

ischemic insult, or newly diagnosed >70% stenosis of the extracranial carotid artery. Strokes and ischemic insults had to be verified by CT or MRI. Peripheral vascular disease was defined as having intermittent claudication with angiographically or sonographically proven stenosis >50% of the major arteries of the lower limbs or ulcers caused by atherosclerotic stenosis or surgery for this disorder. Transplantation was a censoring event and the transplantation date was considered as the final follow-up date (17).

#### Ex-vivo Model of Hemodialysis

An ex-vivo model of HD was used as previously described (18). In brief, a closed circuit was assembled using a pediatric polysulfone hollow-fiber dialyzer (FX paed; Fresenius Care, Germany) and blood lines (SN-Set ONLINEplus BVM 5008-R, Fresenius Care, Germany). The total volume of the circuit was approximately 50 mL. Perfusion was achieved using a Masterflex <sup>R</sup> peristaltic pump (Cole-Parmer, USA) and was flow-controlled (TS410 tubing flow module, Transonic systems Inc, USA) to reach a perfusion flow of approximately 140 to 160 mL/min. The temperature was kept constant at 37◦C and controlled by an external heater. Whole blood was taken from healthy volunteers (n = 3) and anticoagulated with low-molecular weight heparin (1 U/mL). Initially, the circuit was primed with NaCl 0.9% and perfused for approximately 20 min to remove air bubbles. Prior to perfusion, the dialysate compartment was filled with NaCl 0.9% and closed. Next, freshly drawn heparinized blood was added to the circuit, while the same volume of the saline solution was discarded. The system was perfused with recirculating blood for 4 h. Samples were collected at the start of the perfusion and after 30, 60, 120, 180, and 240 min. To investigate the effect of complement inhibition, two consecutive sessions were performed for each healthy volunteer. During one session, 200 units of C1-inhibitor (Cinryze©, Viropharma, USA) were added to the blood prior to perfusion, whereas the other session without C1-inhibitor (C1-INH) served as a control.

#### Inflammatory Markers and Pro-thrombotic Factors

In the HD cohort, TNF-α was measured by Quantikine HS Human Immunoassay (R&D System Inc., USA). Furthemore, IL-6 and IL-10 were determined using a quantitative sandwich enzyme immunoassay technique (R&D System Inc., USA). Lastly, Von Willebrand Factor (vWF) was measured by enzymelinked immunosorbent assay (Dakopatts, UK). In the ex-vivo HD model, TNF-α, IL-6, IL-10, and vWF were measured using a human magnetic luminex assay (R&D Systems Inc, USA) according to the manufacturer's instructions.

#### Quantification of Complement Proteins

C3d was measured by sandwich enzyme immunoassay as previously described (19). Quantitative antigenic assay for C3 was performed by the radial immunodiffusion technique with monospecific anti-sera (19). The C3d/C3 ratio was determined by dividing the C3d values in µg/mL by the C3 concentration in mg/mL. Additionally, Properdin and MBL concentrations were measured as described earlier (19, 20).

#### Statistics

Statistical analysis was performed using IBM SPSS 22.0 (IBM Corporation, Chicago, IL, USA). Normally distributed data are presented as mean ± standard deviation, whereas nonnormally distributed data are shown as median with interquartile range. Nominal data are displayed as total number of patients with percentage [n (%)]. Differences between two groups were assessed with the student t-test, whereas the paired t-test was used to compare values of a single variable during different time points within the HD session. A one-way ANOVA was used when assessing for differences in multiple groups, followed by Bonferroni's post-hoc comparisons tests. The association between different variables and the incidence of CV-event were assessed by Cox proportional hazard regression. The Harrell's C statistic is the equivalent of the area under the ROC curve, if the outcome is binary (21).

#### Ethics

This study was performed in accordance to the Declaration of Helsinki and was approved by the Medical Ethical Committee from the University Medical Center Groningen. All participants signed informed written consent.

# RESULTS

# Patients Characteristics

Blood samples from 55 patients on maintenance HD were available, of which 35 were male and 20 female (**Table 1**). The mean age was 62 ± 15 years and baseline dialysis vintage was 1.2 years [IQR: 0.6–3.9 years]. The median follow-up of the study was 32 months and during this time 17 patients (31%) developed a CV-event, whereas 16 patients died (29%). In our study, the causes of death were cardiovascular (44%), infection (12.5%), discontinuation of the HD treatment (12.5%), or unknown (31%). Among the patients that developed CV-events, 35% had acute coronary syndrome, 17% needed coronary artery bypass surgery, 11% developed congestive heart failure, 17% had a cerebro-vascular accident and 17% developed peripheral vascular disease. Next, we created two different groups; the 17 patients that developed a CV-event during follow-up (CV-event group) and the 38 patients that remained event-free (event-free group).

# Complement Activation in the HD Patients

To assess complement activation, we determined the C3d/C3 ratio in 55 patients during one HD session at the start of the follow-up. The C3d/C3-ratio at baseline was not statistically different between the patients that would develop a CV-event (7.0 ± 6.2) compared to the patients that would not (9.0 ± 7.4). Surprisingly, at the end of the HD session the C3d/C3-ratio was also not statistically different between the two groups (CVevent group: 11.8 ± 8.5, event-free group: 12.9 ± 10.0). However, when the intradialytic C3d/C3-ratios were compared between the two groups, clear differences were seen (**Figure 1** and **Table 1**). At 30 min intradialysis, there was a significant increase in the C3d/C3-ratio in the CV-event group compared to the patients who remained event-free. During these initial 30 min, the C3d/C3-ratio increased by 3.29 fold in the CV-event group and

#### TABLE 1 | Baseline characteristics of our study population of hemodialysis patients with and without a cardiovascular event.


P\* indicates P-value for the difference in baseline characteristics between the patient with and without a cardiovascular-event. Differences were tested by Student's t-Test or Mann– Whitney U test for continuous variables and with χ 2 test for categorical variables. Data are presented as mean ± SD or median [IQR]. P# indicates P-value for univariate Cox-regression for the occurrence of CV-event. Data are presented as beta coefficient with corresponding P-value.

CV, cardiovascular; BSA, body surface area; ADPKPD, autosomal dominant polycystic disease; FSGS, focal segmental glomerulosclerosis; HbA1c, Hemoglobin A1c; hsCRP, high sensitive C-reactive protein; ACE inhibitor, angiotensin-converting-enzyme inhibitor; AT2-receptor antagonists, Angiotensin II receptor antagonists.

by only 1.26 fold in the event-free group (P < 0.01). In addition, Cox regression analysis was performed to assess the association between C3d/C3 ratio at 30 min and occurrence of a CV-event (**Table 2**). In the crude model, C3d/C3 ratios were associated with a hazard ratio of 1.06 (95% CI 1.02-1.09; P < 0.001). After adjustment for age and gender, variables with P < 0.1 CV-event group".


Model 1: crude. Model 2: adjusted for age and gender. Model 3: adjusted for BSA, aspirin and primary chronic pyelonephritis. Model 4: adjusted for DM, cardiovascular history and hypertension. Model 5: adjusted for HD vintage , UF rate and UF volume.

Data are presented as hazard ratio (HR) plus 95% confidence interval (CI). BSA, body surface area; DM, diabetes mellitus; HD, Hemodialysis; UF, ultrafiltration rate.

in univariate analysis (BSA, chronic pyelonephritis as primary renal disease and use of aspirin), cardiovascular risk factors (CV history, DM and hypertension) or characteristics of HD (Ultrafiltration rate, ultrafiltration volume and dialysis vintage), the association between C3d/C3 ratio at 30 min and CV-event remained significant. Subsequently, the Harrell's-C statistics was determined to further confirm the potential relationship between complement activation and CV-events. Plasma C3d/C3 ratio at 30 min had a Harrell's-C statistics of 0.71 (95% CI 0.55–0.88; P = 0.01).

We next set out to assess the contribution of the AP and LP to HD-induced complement activation. Due to a lack of samples, properdin and MBL levels were measured in a subgroup of 30 patients (**Figure 2**). In this subgroup, there were 11 patients in the CV-event group and 19 patients in the event free group. MBL and properdin levels were comparable between the two groups at the start and at the end of the HD session. Conversely, at 30 min intradialysis, MBL levels decreased significantly in the event-free group but not in the CV-event group (P < 0.05). Furthermore, properdin levels were significantly lower at 30 min in the CVevent group, compared to the event-free group. To summarize, MBL consumption was seen in the event-free group implying LP activation, while lower properdin levels were observed in the CV-event group suggesting AP activation.

#### Inflammatory and Pro-thrombotic Factors in the HD Patients

We determined cytokines and Von Willebrand factor (vWF) to investigate if complement activation during HD was accompanied by a pro-inflammatory response and a prothrombotic state. During HD, distinct time-courses for levels of vWF were observed between the two groups (**Figure 3**). In the CV-event group, vWF levels increased steadily during the session

FIGURE 2 | Intradialytic levels of properdin and Mannose-binding lectin. Course of plasma mannose-binding lectin (MBL) and properdin in patients that developed a cardiovascular event (CV-event) during follow-up and in those that remained CV-event free (no CV-event). The data is presented as mean ± SEM. (A) The levels of MBL were determined at the start of hemodialysis session and 30 and 240 min after. (B) The levels of properdin were determined at the start of hemodialysis session and 30 and 240 min after. Differences between the two groups were assessed by the student t-test and a one-way ANOVA followed by Bonferroni's post-hoc comparisons tests was used to compare levels at different time points within one group (\*P < 0.05). The hashtag above the bars denotes a significant difference between the two groups (#P < 0.05), whereas the asterisk above the bars denotes a significant difference compared to baseline within the group. The number of subject is 11 in the "CV-event group" and 19 in the "No CV-event group".

(P < 0.05). Furthermore, compared to the event-free group, the CV-event group had significantly higher levels of vWF at 180 and 240 min intradialysis (P < 0.05). Cytokines such as tumor necrosis factor-α (TNF-α) may initiate inflammation and are therefore believed to play a role in dialysis-related cardiovascular risk. Levels of TNF-α rose significantly during the HD session in both groups (**Figure 4A**). In the CV-event group, levels peaked at 180 min after the start of the HD session (P < 0.01) and were significantly higher than in the event-free group (P < 0.05). Furthermore, in the event-free group, the maximum TNF-α levels were reached at the end of the session (P < 0.001).

To evaluate the relation between anti-inflammatory cytokines and pro-inflammatory cytokines, we determined the IL-6/IL-10 ratio (**Figure 4B**). Interestingly, IL-6/IL-10 ratios were the highest in both groups at the start of the HD session and showed a decreasing trend during the dialysis session, although not significant compared to baseline. Moreover, at 60 min intradialysis an important decrease in the IL-6/IL-10 ratio occurred in the event-free group, indicating a shift toward a less inflammatory profile. However, IL-6/IL-10 ratios remained elevated in the HD patients that developed a CV-event during follow-up, revealing a significant difference between the groups at this time point (P < 0.05). Overall, enhanced levels of proinflammatory and pro-thrombotic mediators seem to prelude the development of CV-events in HD patients.

#### Ex-vivo Model of Hemodialysis

To further evaluate the effect of HD-induced complement activation on inflammation and coagulation, we used an ex-vivo model of HD. During the 4 h of perfusion, the C3d/C3 ratio increased progressively from 4.7 ± 0.6 at baseline to 55.8 ± 12.5 after 240 min (**Figure 5A**). MBL and properdin levels were determined to discriminate between complement activation via the AP and/or the LP. Both MBL and properdin levels decreased

FIGURE 3 | Levels of von Willebrand factor during hemodialysis. Course of von Willebrand factor (vWF) in patients that developed a cardiovascular event (CV-event) during follow-up and in those that remained CV-event free (no CV-event). The data is presented as mean ± SEM. vWF was determined at the start of hemodialysis session and 60, 180 and 240 min after the start of the session. Differences between the two groups were assessed by the student t-test and a one-way ANOVA followed by Bonferroni's post-hoc comparisons tests was used to compare levels at different time points within one group (\*P < 0.05, \*\*P < 0.01). The hashtag above the bars denotes a significant difference between the two groups (#P < 0.05), whereas the asterisk above the bars denotes a significant difference compared to baseline within the group. The number of subject is 17 in the "CV-event group" and 38 in the "No CV-event group".

significantly over time. After 4 h, MBL levels were reduced by 55.2% (P < 0.05) and properdin levels by 34.4%, respectively (**Figure 5**). We next assessed inflammatory and pro-thrombotic factors. Similarly to complement activation, the HD model resulted in a significant increase in TNF-α, IL-6/IL-10 ratio, and vWF levels after 240 min of dialysis (**Figure 6**).

Finally, we evaluated the effect of complement inhibition in our model to test if complement activation acts as a trigger for inflammation and coagulation in HD. C1-INH was added and significantly reduced C3d/C3 ratios compared to controls, namely from 55.8 ± 13 to 33.1 ± 24 (**Figure 5A**; P < 0.05). HDinduced consumption of properdin and MBL was not prevented by the use of C1-INH.Furthermore, TNF-α levels were 1654 ± 631 ng/mL after 240 min in the control session, while in the session with C1-INH levels were reduced to 48.7 ± 74.7 ng/mL(**Figure 6A**). Correspondingly, a similar trend was seen after 240 min in the IL-6/IL-10 ratio (control session 1086 ± 630 pg/mL, C1-INH session 51 ± 78 pg/mL; **Figure 6C**) and for vWF levels (control session: 98.2 ± 22.7 pg/mL, C1-INH session: 1.7 ± 3.3 pg/mL) when the control session was compared to C1-INH session (**Figure 6B**). To summarize, C1-INH addition was able to inhibit HD-induced complement activation and thereby reduce vWF, TNF-α and IL-6/IL-10 by 98, 97, and 95% respectively.

#### DISCUSSION

Hemodialysis treatment comes with the balance between the dangers of advanced uremia and the inherent risks related to this form of RTT (22, 23). The higher cardiovascular risk seen in this population is not only related to ESRD but it is also associated with the HD procedure itself (2). Innate immunity has been proposed to be the missing link in the mechanism of CV-events in HD patients (4). We observed distinct differences in molecular profiles during HD of patients that will later develop CV-events compared to those who remained eventfree during follow-up. At the start of dialysis, a unique peak in complement activation was only seen in patients in the CV-event group. Furthermore, enhanced inflammation and coagulation accompanied the complement activation seen in HD patient that will develop CV-events. Moreover, these processes arose long before the actual development of the CV-event. Altogether these three elements showed different dynamics, with complement activation possibly initiating these processes. In accordance, complement inhibition in our ex-vivo model not only decreased complement activation but also diminished pro-inflammatory and pro-thrombotic mediators.

Despite significant advances in the biocompatibility of HD membranes, complement activation remains an undesired but relevant issue (8, 15). Higher levels of complement components as well as loss of complement inhibitors have been associated with a higher risk for cardiovascular disease in HD patients (8, 11– 14). Recently, complement activation prior to a HD session was associated with the occurrence of CV-events in HD patients (15). Here, we showed that activation of C3 during dialysis is linked to the development of CV-events. Our study is the first, to our knowledge, to assess the relationship between intradialytic complement activation and subsequent outcome. In accordance, previous studies have shown that activation of the complement system peaks during the first 15 to 30 min of the HD session (24). However, the mechanism by which complement activation increases the risk for cardiovascular disease remains largely unknown.

The LP and AP initiate complement activation during HD (25, 26). In our study, we only found MBL consumption in the event-free group, implying that this decrease is actually beneficial. In accordance, MBL has been proposed to be involved in the removal of atherogenic particles, thereby decreasing atherosclerosis. Our previous data showed that higher MBL levels in HD patients were associated with protection against cardiovascular disease (9). We also found a rise in properdin levels in the event-free group. Properdin, unlike other complement factors, is produced by leukocytes, predominately neutrophils (27). Therefore, the increase in properdin is presumably the result of leukocyte activation by the HD membrane leading to degranulation (28). Since, this rise was not seen in the CV-event group, we speculate this was due to properdin consumption by AP activation in these patients.

We found higher TNF-α levels and IL-6/IL-10 ratios in patients that would develop a CV-event. TNF-α and IL-6 are potent cytokines that can initiate a powerful pro-inflammatory reaction (29, 30). If this response is not contained, it can lead to hypotension, organ dysfunction, and eventually result in death. Elevated levels of these cytokines have also been related to an increased risk for CV-events in the general population and in HD patient (31–34). In contrast, IL-10 is a major anti-inflammatory cytokine with the ability to suppress the production and secretion of pro-inflammatory mediators in leukocytes, thereby effectively controlling the inflammation (35). The IL-6/IL-10 ratio has previously been linked to outcome after inflammatory disorders and to the development of HDinduced left ventricular dysfunction (36–38). In ex-vivo models, the induction of IL-6 during the bio-incompatibility reaction was shown to be completely complement-dependent, while the induction of TNF-α was only partially complement-dependent (16). In addition, in a primate model of HD, complement inhibition lead to enhanced levels of IL-10, demonstrating the relationship between the two systems (39).

Thrombosis is a key element in the development of cardiovascular disease. Previously, Péquériaux et al. reported that vWF is a good predictor of CV-events in patients undergoing RRT (40). Von Willebrand factor is a glycoprotein involved in hemostasis but vWF is also a marker of endothelial cell activation (41). We found significantly higher levels of vWF in the group of patients who developed CV-events, which could be evidence of a prothrombotic state. vWF is produced in endothelial cells and megakaryocytes, but also stored in the granules of platelets (42). Considering our ex-vivo lacks endothelial cells, the vWF is most likely derived from platelets. The release of vWF could either be the direct effect of complement activation or via C5a-activated leukocytes (43, 44). The link between the complement system and thrombosis is not new in HD (45). Complement receptors on leukocytes are important for the formation of plateletleukocytes complexes, which contributes to thrombotic processes (46). In addition, complement activation during HD induces the production of pro-coagulation factors (47). Moreover, plasma levels of C3 correlated with a denser clot structure in HD patients (48).

The complement system is a strong mediator of the bioincompatibility reaction. Therefore, we proposed that

FIGURE 4 | Levels of tumor necrose factor alpha and the ratio of interleukin-6 to interleukin-10 during hemodialysis. Course of tumor necrose factor alpha (TNF-α) and ratio of interleukin-6 (IL-6) to interleukin-10 (IL-10) in patients that developed a cardiovascular event (CV-event) during follow-up and in those that remained CV-event free (no CV-event). The data is presented as mean ± SEM. (A) The levels of TNF-α were determined at the start of hemodialysis session and 60, 180 and 240 min after the start of the session. (B) Levels IL-6 and IL-10 were determined at the start of hemodialysis session and 60, 180, and 240 min after. The IL-6/IL- 10 ratio was calculated by dividing the IL-6 (in pg/mL) values by the IL-10 levels (in pg/mL). Differences between the two groups were assessed by the student t-test and a one-way ANOVA followed by Bonferroni's post-hoc comparisons tests was used to compare levels at different time points within one group. (\*P < 0.05, \*\*P < 0.01, \*\*\*P < 0.001). The hashtag above the bars denotes a significant difference between the two groups (#P < 0.05), whereas the asterisk above the bars denotes a significant difference compared to baseline within the group. The number of subject is 17 in the "CV-event group" and 38 in the "No CV-event group".

FIGURE 5 | Complement levels during ex vivo hemodialysis. Two different sessions were performed with whole blood of three healthy donors; one session with C1-inhibitor (C1-INH) and one session without, the control session. The data is presented as mean ± SEM. (A) C3d/C3 ratios were measured to determine complement activation. (B) MBL levels significantly decrease over time during the session. (C) Properdin levels were reduced during the session, although not significantly. Differences between the two groups were assessed by the student t-test and a one-way ANOVA followed by Bonferroni's post-hoc comparisons tests was used to compare levels at different time points within one group (\*P < 0.05, \*\*P < 0.01). The hashtag above the bars denotes a significant difference between the two groups (#P < 0.05), whereas the asterisk above the bars denotes a significant difference compared to baseline within the group.

with whole blood of three healthy donors; one session with C1-inhibitor (C1-INH) and one session without, the control session. The data is presented as mean ± SEM. (A) TNF-α levels significantly increased in the control session, but not in the session with C1-INH. (B) Correspondingly, vWF showed a significant rise in the control session, while C1 INH additionlead to significantly lower levels. (C) The IL-6/IL-10 ratio progressively increased over time, although not statistically significant. Differences between the two groups were assessed by the student t-test and a one-way ANOVA followed by Bonferroni's post-hoc comparisons tests was used to compare levels at different time points within one group (\*P < 0.05). The hashtag above the bars denotes a significant difference between the two groups (#P < 0.05), whereas the asterisk above the bars denotes a significant difference compared to baseline within the group.

complement activation has an essential role in orchestrating the inflammatory response in HD. In accordance with the result observed in the HD patients; our ex-vivo model demonstrated that the dialyzer induces complement activation, inflammation and enhances coagulation. We then wondered if HD-induced inflammation could be attenuated by complement inhibition. Addition of C1-INH to the ex-vivo HD model significantly diminished HD-induced complement activation and also almost completely abolished the induction of TNF-α levels, L-6/IL-10 ratios and vWF levels. Similarly, Kourtzelis et al, also demonstrated in an ex vivo model of HD the induction of coagulation during 2 h of perfusion. In their study, compstatin was used to block complement activation at the level of C3 (10). We postulate that HD-induced complement activation results in the formation of anaphylatoxins, thereby resulting in the activation of peripheral blood mononuclear cell and platelets initiating a pro-inflammatory and pro-thrombotic response. Previously, several reports demonstrated that bioincompatibility-induced inflammation relies mainly on complement, whereas granulocyte enzyme release was predominantly C3-dependent, leukocyte activation and prothrombotic mediators were largely dependent on C5 (16, 46). Altogether, our results support the hypothesis of the complement system as a key component in HD-induced inflammation and coagulation, which subsequently leads to a higher risk for CV-events.

We are aware that our study has strengths and limitations. Although the study has a long follow up, the samples were only collected during a single hemodialysis session. Our study could have benefited from a second assessment of these parameters during another dialysis session in the same patients, to assess reproducibility and increase reliability. Furthermore, while complement activation was seen in the patients during HD as well as in our ex-vivo model, clear differences were present. Moreover, in patients a significant peak of complement activation was seen during the first 30 min of HD. In contrast, in our ex-vivo model a continuous rise of complement activation was seen until the end of the session. Obviously these discrepancies arise due to the differences of in-vivo to ex-vivo. For instance, in the ex-vivo model blood recirculates without re-entering the human body, therefore it lacks the interaction with endothelial cells, liver and other organs. Lastly, the size of our cohort could be considered small and therefore might impact the statistical analysis. However, due to the long follow up, we achieved a relatively high number of CV-events which increases the power of the study in the comparisons between the CV-event group and the event free group.

There is a growing body of data supporting a role for the complement system in the development of cardiovascular disease. Ekdahl et al. proposed that complement activation initiates an inflammatory cascade and amplifies pro-thrombotic processes (4). For the first time, to our knowledge, we demonstrated intradialytic differences in complement activation, inflammation and a pro-thrombotic factor in HD patients that will develop a CV-event compared to HD patients that will not. Furthermore, we showed that complement inhibition during HD resulted in decreased levels of the pro-inflammatory and prothrombotic mediators. Future studies have to determine what the ideal target is to inhibit complement in HD to attenuate these processes and to determine if this decreases the risk of CV-events in HD patients.

#### AUTHOR CONTRIBUTIONS

FP, MG, MD, CF, and MS research idea and study design. FP, MG and SA data acquisition. FP, MG, BF, SB, MD, JM, WvS, CF, and MS data analysis/interpretation. FP and MG statistical analysis and wrote the manuscript. All authors were involved in editing the final manuscript. All authors read and approved the final manuscript.

# REFERENCES


to polysulfone hemodialysis membranes. Artif Organs. (2017) 41:E285–95. doi: 10.1111/aor.12954


J Physiol. (2012) 590:1023–34. doi: 10.1113/jphysiol.2011.2 25417


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Poppelaars, Gaya da Costa, Faria, Berger, Assa, Daha, Medina Pestana, van Son, Franssen and Seelen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Complement in the initiation and evolution of Rheumatoid Arthritis

#### *V. Michael Holers and Nirmal K. Banda\**

*Division of Rheumatology, Department of Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO, United States*

The complement system is a major component of the immune system and plays a central role in many protective immune processes, including circulating immune complex processing and clearance, recognition of foreign antigens, modulation of humoral and cellular immunity, removal of apoptotic and dead cells, and engagement of injury resolving and tissue regeneration processes. In stark contrast to these beneficial roles, however, inadequately controlled complement activation underlies the pathogenesis of human inflammatory and autoimmune diseases, including rheumatoid arthritis (RA) where the cartilage, bone, and synovium are targeted. Recent studies of this disease have demonstrated that the autoimmune response evolves over time in an asymptomatic preclinical phase that is associated with mucosal inflammation. Notably, experimental models of this disease have demonstrated that each of the three major complement activation pathways plays an important role in recognition of injured joint tissue, although the lectin and amplification pathways exhibit particularly impactful roles in the initiation and amplification of damage. Herein, we review the complement system and focus on its multi-factorial role in human patients with RA and experimental murine models. This understanding will be important to the successful integration of the emerging complement therapeutics pipeline into clinical care for patients with RA.

# *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Trent M. Woodruff, The University of Queensland, Australia Daniel Ricklin, Universität Basel, Switzerland*

*\*Correspondence: Nirmal K. Banda nirmal.banda@ucdenver.edu*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 26 March 2018 Accepted: 27 April 2018 Published: 28 May 2018*

#### *Citation:*

*Holers VM and Banda NK (2018) Complement in the Initiation and Evolution of Rheumatoid Arthritis. Front. Immunol. 9:1057. doi: 10.3389/fimmu.2018.01057*

Keywords: complement, arthritis, classical pathway, lectin pathway, alternative pathway, mannose-binding protein-associated serine proteases, inflammation

#### COMPLEMENT SYSTEM AND ITS ACTIVATION

It was Buchner who, at the University of Munich, discovered a blood born substance that was able to destroy bacteria. He named it "alexin." The term "complement" was subsequently introduced by Ehrlich as part of his grand model of the immune system (1–6). Although initially considered primarily in the context of resistance to infection, the complement system, as an important arm of

**Abbreviations:** CP, classical pathway; LP, lectin pathway; MBL, mannose-binding lectin; AP, alternative pathway; CICs, circulating immune complexes; CRDs, carbohydrate recognition domains; FLS, fibroblast-like synoviocytes; CII, bovine type II collagen; PAD2 or PAD4, peptidyl arginine deiminase type 2 or 4; CDA, clinical disease activity; RF, rheumatoid factor; ACPA, anti-citrullinated protein antibodies; anti-CarP, anti-carbamylated protein; CIA, collagen-induced arthritis; PIA, pristane-induced arthritis; ZIA, zymosan-induced arthritis; PGIA, proteoglycan-induced arthritis; SCWA, streptococcal cell wall arthritis; CAIA, collagen antibody-induced arthritis; mBSA, methylated bovine serum-induced arthritis; KBxN STA, KBxN serum transfer model of arthritis; Crry, mouse complement receptor-related gene y; MASP-1, mannose-binding lectin-associated serine protease-1; MASP-2, mannose-binding lectin-associated serine protease-2; MASP-3, mannose-binding lectin-associated serine protease-3; MAp44, mannose-binding lectin-associated protein of 44 kDa a.k.a. MBL/ficolin/CL-11 associated protein-1 (MAP-1); FCNs, ficolins; FCN A, ficolin A; FCN B, ficolin B; CL-11, collectin liver 11; mAb, monoclonal antibody; MAC, membrane attack complex.

the innate immune system, has been long recognized to play an important role in tissue damage in many autoimmune diseases, including rheumatoid arthritis (RA). Thus, the complement system responds not only to microorganisms but also mediates inflammation through the orderly activation of a cascade of multi-protein enzymes and proteases.

Key functions of the complement system include clearance of foreign microorganisms through specific recognition, opsonization, and lysis (7). The system also plays major roles in the clearance of circulating immune complexes (CICs), apoptotic cells, apoptotic bodies, and dead cells (8, 9). Out of three different types of CICs (small, intermediate, and large), intermediate CICs typically cause the most damage as they get trapped in the tissues or in the joints. These protective functions provide potent properties for the benefit of the host, even in the absence of an adaptive immune response.

Although the complement system limits its pro-inflammatory and anti-inflammatory activities, through action of many inhibitors under normal physiological conditions, these natural complement inhibitors are not enough when the complement system gets over-activated during acute inflammatory conditions and thereby causes more damage than good. The functions of the complement system are not only limited to serum or plasma where these are found in abundance but to each and every tissue or organ of the body which are the direct target of various complement components.

Most proteins of the complement system are normally present in the circulation in an inactive (zymogen) form to be activated *via* proteolytic processing upon the recognition of danger. Interestingly, there exists multiple pathways by which the complement system may be activated, each employing different recognition molecules, which underscores its great complexity. The complement system is activated by three different major pathways: the classical pathway (CP), the lectin pathway (LP), and the alternative pathway (AP) and one minor pathway, the C2/C4 bypass (10) (**Figure 1**). All of these pathways are activated by various antibodies, ICs, molecules or microorganisms, or spontaneously as discussed below.

#### Classical Pathway Activation

The CP is activated by binding of C1q to the heavy-chain crystallizable fragment (Fc) domain of immunoglobulin (Ig). In mice, IgM, IgG1, IgG2a, and IgG2b all have complement activation sites, and these can form CICs when combined with an antigen and complement. C1q leads to the activation of C1r, followed by activation of C1s. C1s cleaves and activates C4 into C4a and C4b and also C2 into C2a and C2b, leading to the formation of C4b2a (CP C3 convertase), which itself cleaves C3 into C3a and C3b (11). C3b further binds to C4b2a to generate the C5 convertase of the CP. This initiates the formation of C5b-9, the membrane attack complex (MAC) (12).

Through its recognition mechanisms, C1 can help to distinguish self from non-self, which is important for the maintenance of self-tolerance and homeostasis (13). Conversely, its pathologic activation has been implicated in many inflammatory and autoimmune diseases, and its activation is limited by C1 esterase

Holers and Banda Complement and Arthritis

inhibitor (C1-INH) (14). Recently, it has been shown that C4a is a ligand for protease-activated receptor (PAR) 1 and PAR4, extending the direct link between the complement and coagulation systems (15). In addition, MAC assembly has been shown on the surface of parasites, and to eliminate Gram-negative bacteria and unwanted host cells (16–18). The MAC can rupture cells with varied composition of lipids and once MAC assembly initiates on cell surfaces other factors can still block it (16). Interestingly, sublytic levels of MAC either causes the release of pro-inflammatory mediators or in other circumstances acts to increases the protection of cells to avoid further innocent bystander cell lysis (19, 20).

# Lectin Pathway Activation

The recognition components of the LP, mannose-binding lectins (MBLs), ficolins (FCNs), and collectins (CLs) bind directly to microbial and other surfaces with exposed carbohydrates and *N*-acetyl groups and activate complement *via* MBL-associated serine proteases (MASPs) (21–23).

Ficolins, which contain a carbohydrate recognition domain (CRD), consist of collagen-like and fibrinogen-like domains and preferentially bind to *N*-acetylglucosamine (GlcNAc) (24–26). There are two mouse FCNs: ficolin A (FCN A) (27) and ficolin B (FCN B) (28); by contrast, humans express three FCNs; ficolin M (FCN-M a.k.a. FCN 1), ficolin L (FCN-L a.k.a. FCN 2), and ficolin H (FCN-H a.k.a. FCN 3) (29–33). Mouse FCN A, but not FCN B, exhibits a splice variant known as FCN A variant (34). FCN A is present in the serum and expressed in liver hepatocytes (35). Mouse FCN B was originally found in the lysosomes of macrophages, similar to human FCN-M, which is also found in the secretary granules of monocytes and neutrophils (36, 37). We have reported that FCN B is also present in the circulation of mice suggesting that it is secreted from macrophages (38).

Mannose-binding lectin is a C-type lectin containing a CRD as well as a collagen-like domain (21, 22). MBL binds to mannosecontaining molecules as well as *N*-acetylglucosamine (21, 22). There are two types of mouse MBLs, i.e., MBL-A and MBL-C, whereas there is only one type in human (39). Four different types of MASPs called mannose-binding lectin-associated serine protease-1 (MASP-1), mannose-binding lectin-associated serine protease-2 (MASP-2), mannose-binding lectin-associated serine protease-3 (MASP-3), and sMAP (also called Map19) (small MBL-associated protein) circulate complexed with MBL, FCNs, and CL (40). Although no specific function has been assigned to sMAP by generating *sMAP<sup>−</sup>/<sup>−</sup>* mice, it was shown that the expression of MASP-2 was also decreased in the sera of these mice because of the MASP-2 gene disruption (41). These authors have also shown by using sera from *sMAP<sup>−</sup>/<sup>−</sup>* mice that sMAP plays a regulatory role in the activation of the LP but it is not clear whether sMAP plays a regulatory role before or after the LP activation. sMAP and MASP-2 compete to bind to MBL, and sMAP has the ability to downregulate the LP (41). MAp44 (also called MAP-1), an alternatively spliced product of the *MASP-1/3* gene, is a natural inhibitor of the interactions between MBLs and FCNs and serves as a major regulator of the LP (42, 43). MASP-1, MASP-2, and MASP-3 consist of an A chain (1CUB, EGF, 2CUB, 1CCP, 2CCP, and the linker region) linked by a disulfide bond to a B-chain (serine protease domain).

Both the CP and LP share C2 and C4 complement components. Similar to the CP, the LP forms the C3 and C5 convertases leading to the formation of MAC. A recent additional breakthrough has been the finding that MASP-3, which is an alternative spliced form of *MASP-1/3* gene, is a positive regulator of the AP of the complement system (44) and MASP-3 exclusively enables FD maturation (45). It has been shown that both MASP-1 and MASP-2 can activate MASP-3, but MASP-3 in resting human blood is also present in an active form (46). *In vitro*, not *in vivo* studies, have shown that MASP-1 is essential for bacterial LPS but not Zymosan-induced AP activation (47), indicating that MASP-1 can regulate a specific AP activation mechanism but not the entire AP.

The third class of LP initiators, designated CLs, are similarly C-type lectins containing CRDs (48). Three different human CLs have been identified: Collectin-10 (a.k.a. collectin liver 1, CL-L1, or CL-10), Collectin 11 (a.k.a. collectin kidney 1, CL-K1, or CL-11), and Collectin-12 (collectin placenta 1, CL-P1, or CL-12) (49–52). CL-K1 is present in various human and mouse tissues (49, 53). It has also been shown that CL-K1 acts as a soluble pattern recognition receptor for *Mycobacterium tuberculosis* (54). Additionally, it binds to l-fucose beside other potential ligands (55). In a renal injury model, CL-11 expression was rapidly upregulated and recognized the abnormal presentation of l-fucose leading to complement activation and tissue injury (56). Interestingly, both CL-11 and MASP-2 have been shown to generate C3d on injured cells (56). All of the abovementioned collectins play an important role in an innate immunity because these can bind to the LPS from various species of bacteria (57–60).

#### Alternative Pathway Activation

The original properdin-dependent pathway, now called the AP of the complement system, was discovered in the 1950s (61). The AP consists of four proteins, factor B (FB), factor D (FD), Properdin (Pf or P), and C3. In contrast to the CP and LP, the AP is activated spontaneously through hydrolysis of C3, thereby generating C3(H2O), which can associate with FB, resulting in the cleavage of FB into Ba and Bb by FD. Therefore, the AP does not require a specific recognition molecule in order to be activated. Nevertheless, as C3 hydrolysis (a.k.a. C3 "tick over") is always happening regardless of the presence of FB or FD (62), the system is always poised for activation. In that process, FB binds to C3(H2O), and can be cleaved by FD to generate C3(H2O)Bb (C3 convertase). The cleavage is much slower without properdin, which is a positive regulator of the AP convertase and can also independently promote the activation of the AP on certain surfaces (63). C3b bound to Bb on a surface is a potent C3 convertase and can cleave C3 to generate more C3b and C3a. The C3 convertase can also combine with another C3b molecule forming an AP C5 convertase. The latter can start the formation of the MAC after cleaving complement component C5 into C5b and C5a. In contrast to activation, complement factor H (FH) is the natural regulator of the AP, and in addition to solution phase AP blockade the binding of this molecule to one or more of the host marker recognition sites enables it to control surface activation of the AP (64). There are also several membrane-bound inhibitory proteins described below which determine the location and activity of the complement system (65). Finally, regardless of the activation route, all of these pathways generate two major potent pro-inflammatory molecules; C3a and C5a, *via* C3 and C5 convertases, respectively, which play a vital role in the pathogenesis of arthritis.

#### C4/C2 Bypass Pathway Activation

A fourth pathway has also been found to be important in the generation of complement pro-inflammatory mediators and termed the "C4/C2 bypass pathway" or "C2-independent" pathway (66). Initially, it was shown that C4 and C2 complement components were not necessary to lyse cells by the CP (67–69) despite the fact that C4 and C2 are important constituents of the C3 convertase. More recently, it was shown that the components of the LP such as MBL in the absence of C2, C4, or MASP-2 induce C3 deposition (70). This could be mediated by LP ligands such as MBL or FCNs. These could activate the AP directly *via* MASP-3, which cleaves proFD into FD (10). C3 activation in the absence of C2, C4, or MASP-2 requires FB as well as a high concentration of serum (70). C3 activation on adherent anti-collagen (anti-CII) antibodies, was also reported at a high concentration using sera from mice lacking C4 by an unknown mechanism (71). This C3 activation on adherent anti-CII antibodies was fully inhibited by an anti-FB inhibitory antibody (71), confirming specific AP activation in the absence of C4. Even in human serum lacking C4, MASP-2-dependent C3 activation was reported *via* a C4 bypass route (72). So, this C4/C2 bypass pathway is operative without forming conventional CP or LP C3 convertases. It has been shown that thrombin is capable of generating the complement activation product C5a in the complete absence of C3 (73), which represents another bypass mechanism to generate C5a and C5b. Recently it has been shown that in the absence of C4, the CP cannot be activated however, LP still retains the capacity to cleave C3 into C3a and C3b. This residual C4/C2 bypass is dependent on MASP-2 (74). This study further demonstrated that MASP-2 dependent cleavage of C3 was inhibited by MASP-2-specific inhibitors. All of these studies are consistent with the presence of a backup or bypass complement pathway that works *in vivo* in the absence of C4 or C2. Whether this pathway, like CP, LP, and AP, is controlled by the well-described complement regulators is not known. Importantly, the relative importance of the C4/C2 bypass pathway in relation to arthritis models has recently been shown (10), and other studies have revealed its importance in some ischemia/reperfusion-related models (72).

#### Complement Mediators of Inflammation and Their Receptors

Regardless of the activation pathway, cleavage of C3 is followed by generation of C3a, C3b, iC3b, C3d, C5a, C5b, and MAC (75, 76). Recently, it has been shown that human plasma kallikrein directly cleaves C3 into C3a and C3b and triggers an amplification loop (77). Interestingly, the cleavage site within C3 is identical to that recognized by C3 convertase and is also inhibited by FH. The cleavage of C3 and C5 by kallikrein or thrombin appears to represent a coordination between the complement system and the coagulation pathways (78). C4 cleavage leads to the generation of C4a, another anaphylatoxin but it is not clear whether it is a chemoattractant. C3a, C4a, and C5a are called anaphylatoxins because they are able to carry out pro-inflammatory activities even at a very low concentration. Thus, complement can contribute to the inflammatory injury through many mechanisms.

Previously, it has been shown that C3a is less potent than C5a while C3a desArg (a.k.a. acylation-stimulating protein) has no inflammatory activity. It has also been shown that C3a is not a chemoattractant for neutrophils but can cause migration of eosinophils (79). But this view has shifted based on new findings. C3a binds C3aR expressed on the surface of neutrophils, eosinophils, and basophils, monocytes/macrophages, and mast cells (80, 81). C3a and C5a can bind and equally activate through their receptors C3aR and C5aR, respectively, present on the surface of basophils and mast cells (82).

C4a is very weak anaphylatoxin which is formed by the cleavage of C4 into C4a and C4b. The view whether or not C4a is a classical anaphylatoxin has been recently questioned because evidence has been provided that C4a is a ligand for PAR1 and PAR4 (15). These authors have shown that C4a showed no activity toward known anaphylatoxin receptors but it acted as a non-traditional agonist for both PAR1 and PAR4.

C5a is a cleaved by-product of C5 after complement activation. C5a is rapidly converted by carboxypeptidases to less potent C5a desArg but still has biological activity, and the view regarding C5a desArg potency has also been challenged. Most of the C5a found in the circulation is in the C5a desArg form (83). The binding affinity of C5a to C5aR been reported 100-fold higher than that of C5a desArg for C5aR (84). Although C5a is considered as the triggering molecule but it has been shown that C5a desArg also acts as an important molecule triggering of local inflammation and also maintain blood surveillance and homeostatic status. This study has elegantly shown that C5a desArg induce cell activation in even higher than C5a, which was dependent on the C5aR because it was inhibited by PMX-53, a C5aR antagonist (85). C5a acts as a chemotactic factor of neutrophils and increases neutrophil adhesion to endothelium (86, 87). C5a binds to C5aR (C5aR1 or CD88) and C5L2 (C5aR2 or GPR77) present on many cells leading to chemotaxis of inflammatory cells, vascular permeability, phagocytosis, and release of pro-inflammatory cytokines and chemokines. C5a amplifies tissue injury and inflammation by triggering release of oxygen free radicals and arachidonic acid metabolities (88). C5a is an essential component of the inflammatory response to bacterial infection. *Porphyromonas gingivalis* expresses a peptidyl arginine deiminase (PAD) with a strong preference for the C-terminal arginine of C5a, disabling protein function resulting in decreased chemotaxis of human neutrophils (89). It has been shown that C5a, released at sites of inflammation, upregulates FcγRIIIa and downregulates FcγRIIb simultaneously (90, 91).

The biological and pathological role of the second C5a receptor, C5L2 is controversial. C5L2 does not bind C3a or ASP/ C3adesArg. C5L2 also binds to C5a and C5a desArg. C5a desArg binds 20-fold to 30-fold with higher affinity with C5L2 than C5aR (84, 92). Although C5L2 binds to C5a with the same high affinity as C5aR but function may depend on the cell type, species, and disease context (93). While C5aR is a G protein-coupled receptor, C5L2 is not which led to the hypothesis that C5L2 functions as a decoy receptor. This view is not universally accepted, however. Evidence has been generated linking C5L2 to both anti-inflammatory and pro-inflammatory functions (93). We have found no evidence for C5L2 playing a role in RA; however, so we will leave this fascinating topic for others to review.

Membrane attack complex formation is the final step of the terminal pathway after cleavage of C5 by C5 convertases of the CP/LP or of the AP leading to the formation of the pore consisting of C5b–C9 complement proteins (94). MAC formation may lead either to necrosis or apoptosis, in part depending on the number of pores formed. The choice of necrosis vs. apoptosis is clinically relevant as necrosis is invariably pro-inflammatory, while apoptosis can lead to the resolution of an inflammatory response (95, 96). In general, eukaryotic cells require more pores than prokaryotic cells to induce death. Interestingly, low numbers of MAC, rather than leading to cell death, induce inflammatory signaling events such as the release of TNF-α and IL-1 (97).

Additionally, complement receptor 1 (CR1 or CD35), complement receptor 2 (CR2 or CD21), complement receptor 3 (CR3 or CD11b/CD18), complement receptor 4 (or CD11c/CD18), and complement inhibitors such as FH, decay-accelerating factor (DAF or CD55), membrane cofactor protein (or CD46), and protectin (CD59) have been shown to play an important role in the complement-mediated injury and also, as we will discuss below, in the pathogenesis of arthritis.

#### Measurement of Complement Activation in Inflammation

Classical pathway activity is commonly measured using the CH50 test. Here, serum can be used to lyse sheep erythrocytes coated with anti-sheep antibodies, and degree of hemolysis is measured. By contrast, the AH50 is the best screening test used to measure the proper functioning of the AP. Low levels in either test indicate a deficiency of one or more components of the CP or AP of the complement system, respectively (98). Function of the LP can be measured by enzyme-linked immunosorbent assay pre-coated with mannan particles, and here C4d bound to mannan can be measured. Furthermore, the CP component C4d has been used as a measure to explore the activation of the CP. Bb levels have been used to as a measure to explore the activation of the AP. Tissue bound or soluble MAC levels have been used as measure of the activation of all pathways of the complement system thus as the most important indicator of complement activation within a microenvironment. To measure complement activation and its split products in serum or plasma, there are excellent standard protocols. Conversely, results can be obtained by using fee for services complement focused laboratories and commercially available kits (99–102).

Most of the stable complement components are measured in serum whereas activated split products are measured in plasma (EDTA-anticoagulated blood) due to the interference of the coagulation system enzymes leading to erroneous results. Furthermore, most measurements of complement activity are focused on serum or plasma with no attention paid to specific cells or tissues (e.g., synovial fluid). Since, in clinical disease, complement causes damage locally, more work must be done to assess complement activity on tissue surfaces and within sequestered regions such as synovial fluid. Measurements of CICs along with complement products such as C1q, C3b, C3d, C3dg, and MASPs levels in the serum and synovial fluid of RA patients along with rheumatoid factor (RF), anti-citrullinated protein antibodies (ACPA), and anti-carbamylated protein (anti-CarP) antibodies can provide a better picture of the local production and their role in the RA pathogenesis. Sometimes it is hard to make any conclusion from measuring C3 or C4 levels alone using serum or plasma or synovial fluid because excessive production due to inflammation masks their consumption and results are confounded by the coagulation pathway. Furthermore, endogenous complement inhibitors of the CP, LP, and AP work only under normal physiological conditions but these are ineffective under pathophysiological conditions due to the hyper-activation of complement. Therefore, their measurement provides less useful information.

#### INITIATION OF RHEUMATOID ARTHRITIS

Rheumatoid arthritis is a chronic inflammatory systemic disease that primarily affects peripheral joints, thereby leading to synovial inflammation followed by cartilage and bone destruction. During the development of the disease, the synovium undergoes proliferation, thickens, and incorporates a large number of infiltrating immune cells to become a new tissue called pannus that causes cartilage and bone damage (103–105). Although the exact origin or initiation or development of RA is unknown, studies have shown associations in patients with active RA with infections in the temporomandibular joints (106, 107) or in the gums due to severe periodontitis (108). Dysbiosis of the microbiome in the oral or gut regions has also been strongly associated with onset of RA (109). Furthermore, interstitial lung disease (110), infections by alphaviruses (mosquito-transmitted viruses) such as Ross River virus (111), Chikungunya virus (112), and HBV (113) are also associated with a risk for the development of RA. RA also exhibits a genetic predisposition, with approximately 50% of this genetic risk contributed by certain HLA-DR alleles (114). Many other genes have been shown to contribute to RA pathogenesis (115). Additionally, environmental exposures such as air pollution, occupational exposure to silica, active smoking, wood burning, and mineral oil have been shown to acts a risk factor for initiating and/or developing RA (116, 117). The hypothesis that smoking and pollution lead to an increased risk of RA paved the way to the hypothesis that initial inflammation and production of RA-related autoantibodies (called ACPA anti-citrullinated protein/peptide antibodies) in the lungs may lead to RA (116, 118). So far, there are no studies showing the direct migration of ACPA from lungs to the peripheral joints to precipitate disease. ACPA is the most reliable and specific biological marker to diagnose RA and these antibodies are increased in RA patients sera almost 10 years prior to clinical diagnosis (116, 118–121). Hundreds of citrullinated proteins have been found in the synovial fluid of RA patients which might contribute to the RA pathology (122–125) but why only few autoantigens such as enolase, fibrinogen, and vimentin generate autoantibodies is not clear. How these citrullinated proteins present locally in the synovial fluid interact directly with various complement proteins and activate the complement system is also unknown.

It has been shown that there is a relationship between ACPA, RF, and systemic bone loss in early RA patients (101). The presence of citrullinated antigens on the surface of osteoclastic linage cells makes these cells the main targets of circulating ACPA leading to pro-osteoclastic events (126, 127). There is an argument for complement involvement in this process. Bacterial antigens, perforin, and the MAC cause calcium influx leading to cytolysis. PAD enzymes that convert peptidylarginine into peptidyl citrulline are calcium-dependent (128–131). Interestingly, perforin and the MAC have been shown to reproduce identical patterns of hypercitrullination seen in the neutrophils present in the synovial fluid of RA patients (132). A huge number of neutrophils are present in the synovial fluid of RA patients and are the major source of intracellular citrullination and PADs for extracellular citrullination (132–136). These data suggest that citrullinated proteins along with activation of the complement system might be contributing to the initiation of RA.

#### Complement Activation on Articular Cartilage Surface and in Synovium in Rheumatoid Arthritis

Earlier studies have shown that autoantibodies to type II collagen present in the serum of RA patients bind to the cartilage components or to antigen present on the surface of articular cartilage (137). Articular cartilage is a hyaline cartilage and connective tissue of the joints. The main cellular component of adult articular cartilage is the chondrocyte. These cells, which make up approximately 1% of the tissue, function to organize collagen into ordered structures and secrete extracellular matrix (ECM) components. The ECM is composed of water, collagen type II, proteoglycans, non-collagenous proteins, and glycoproteins (138, 139). Complement activation due to antibody–cartilage surface interaction in RA patients have shown the abundant codeposition of IgGs and activated complement components (140). Interestingly chondrocytes can also synthesize complement components including C1 and C1 inhibitor (141, 142).

An important piece of evidence linking complement activation to pathogenesis in RA was that C1 staining was negative in normal articular cartilage and positive in degenerating cartilage biopsies from all RA patients examined (143). This study strongly implicated the involvement of the CP in the pathogenesis of RA. C3b was also present on the cartilage surface of RA patients; thus, this study clearly showed that C1s can activate the downstream complement cascade thereby causing irreversible damage. It has also been shown that the level of C1q in serum correlates with clinical disease activity (CDA) in RA patients (144–146). In mouse model of RA, C3b gets deposited first on the surface of cartilage vs. synovium and increased rapidly from 4 to 120 h (**Figure 2**). Interestingly, during this time, there is not enough FH availability on the surface of cartilage as well as in the synovium to protect them from complement-mediated damage (**Figure 2**). The presence of C2, C3, C4, and C5 in rheumatoid synovial fluid had been shown previously (147). Levels of properdin and FB of the AP were depressed. An increase in the levels of C3d, C4d, Ba, and MAC has been found in the synovial fluid of RA patients (148, 149). Normally, IgG containing ICs and also C3 split fragments can be found in the joints of more

Figure 2 | A slow snapshot of the early histopathological analysis from the knee joints of wild-type (WT) mice with collagen antibody-induced arthritis (CAIA) for C3 and FH deposition on the surface of cartilage and in the synovium. A mixture of four monoclonal antibody (mAb) to CII (8 mg/mouse) was injected i.p. to induce arthritis, and mice were sacrificed at 0.5, 1, 2, 4, 8, 24, 72, 96, 120, and 144 h later. A low level of FH was present on the surface the cartilage and in the synovium at all time points with slight non-significant increases at 72, 96, 120, and 144 h. By contrast, C3 deposition on the cartilage surface showed a large increase over baseline beginning at 8 h after injection of the anti-CII mAbs and peaking at 120 h. Thus, an imbalance exists between FH deposition and C3 deposition in the early stages of disease leading to failure to protect the knee joints in mice with CAIA. Histopathologic scoring for inflammation (black solid circle) and cartilage damage (white empty circle) from the knee joints (right and left) was performed following tissue processing and Toluidine-blue staining of sections. C3 deposition in knee joints in the synovium (red solid circle) and on the surface of cartilage (red empty circle) is illustrated, as is FH deposition in the synovium (blue solid circle) and on the surface of cartilage (blue empty circle). The data are expressed as mean of disease/baseline ± SEM (*n* = 3 each time point). Baseline = background levels of inflammation, cartilage damage, and C3 and FH deposition in the knee joints of WT mice without treatment with mAb to CII (*n* = 3). Adapted from Ref. (156). Copyright 2013. The American Association of Immunologists, Inc.

than 90% of RA patients (140, 150) and mediate complement activation. IgG with C3d has been present in the synovial fluid of RA patients and also MAC and Bb levels are elevated in the synovial fluid of RA patients (151, 152). No statistical differences in the levels of C3c and C4 in serum and in the synovial fluid of RA patients have been seen in a cross-sectional study although significant differences were seen in the CICs in both biological fluids (153). DAF expression is increased in RA synovium, while the expression of CD59 significantly decreased in the synovial lining (154, 155). The presence of split components of the complement system on the cartilage surface and in the synovium of RA patients indicate that local complement presence/and or synthesis and activation can attract macrophages for phagocytosis of chondrocytes which can further damage the cartilage and synovium.

#### Complement Activation due to Glycosylation, Citrullination, and/or Carbamylation in Rheumatoid Arthritis

It has been documented that IgG is the most abundant immunoglobulin isotype comprising ~75% of the total serum immunoglobulins (157). IgG triggers its effector function (i.e., Holers and Banda Complement and Arthritis

complement activation *via* its Fc). Therefore, any change due to posttranslational modification in the Fc region such as glycosylation will influence the effector function of IgG mediated by the FcγR (158–162). The glycosylation of the Fc is characterized by presence of a single chain N-linked glycan attached to each heavy chain at asparagine 297. It has been shown that the lack of fucose, sialic acid, and galactose residues on the Fc-N-linked glycans increases the inflammatory capacity of IgG in mice (163–166). IgG in RA patients contains less galactose and sialic acid (167). Interestingly, the glycosylation pattern of ACPA changes before the onset of RA skewing toward more inflammation (168, 169). Pregnancy-induced spontaneous improvement of RA as well as flares after delivery has been linked to pregnancy-related changed in the glycosylation of IgGs (170–172). This study was the first natural evidence of that changes in IgG galactosylation can cause disease pathogenicity in humans. Later on, it was shown that agalactosyl IgG is pathogenic in mice and arthritis could be transferred in mice by injecting agalactosyl IgG (173). The LP pathway component, MBL was shown to be associated with the pathogenicity of agalactosyl IgG (159). Not only the Fc but also the Fab domains of IgG has also been reported to contain N-glycosylation consensus sequences (174). It has been reported that more than 90% of ACPA-IgG molecules carry Fab glycans that are highly sialylated (175). More interestingly, ACPA-IgG purified from synovial fluid of RA patients, could even exceed 100% Fab glycosylation implying that multiple glycans can be attached to the variable domain (176). What role Fab-glycan plays in the functionality of ACPA-IgGs is unknown so far, but studies have shown that ACPA-IgG in RA have a pro-inflammatory Fc glycosylation pattern with reduced galactosylation and sialylation levels (177, 178). It has been shown that ACPA have the capacity to activate the CP and AP of the complement system (179). Thus, ACPA reduced galactosylation and sialylation have more capacity to activate the complement system to generate more vigorous effector response.

Citrullination is a normal physiological process which occurs inside apoptotic cells. Normally, apoptotic cells are scavenged by macrophages. If this system is defective, then PAD enzymes and citrullinated proteins can become externalized to influence the immune system (180).

There are five PAD isoenzymes (PAD1–4 and 6) that regulate key cellular processes (181). The PAD enzymes will citrullinate proteins by converting arginine into citrulline. Cl-amidine is the most widely used pan-PAD inhibitor (182, 183). During inflammation many cells die, and it is common to find citrullinated proteins at the inflamed sites such as in the inflamed synovium of RA patients, suggesting that ACPA could be generated as part of an immune response to self-proteins (184). In filaggrin, fibrin, and vimentin, anti-cyclic citrullinated peptide antibody (ACPA a.k.a. anti-CCP) recognizes the arginine residues modified by PAD enzymes to citrulline (185, 186). The presence of citrullinated proteins, as mentioned above, does not always mean generation of ACPA, however (180). The combination of high specificity (90–99%) and high sensitivity (66–88%) of anti-CPP for diagnosing RA, and above all, correlation with radiological damage has led to the conclusion that these antibodies have a pathological connection to the initiation of RA (185, 187–192). The clinical measure of above 20 U/ml suggests the possibility of RA. Additionally, approximately 20% RA patients are anti-CCP negative. Despite the radiological damage association with anti-CCP antibodies, the levels are not being used to determine the progression of disease since even during remission most of the subjects remains anti-CCP positive. This raises question regarding the direct pathogenicity of these antibodies in RA. It is not clear whether anti-CCP antibodies are the cause or the result of inflammation in RA patients. Whether anti-CCP antibodies are the result of defective coagulation system in RA patients is also not clear. Defective coagulation can in principle modulate the generation of anti-CCP autoantibodies. This is due to the fact that thrombin cleaves fibrinogen into fibrin followed by a clot formation but if fibrinogen is citrullinated then thrombin cannot cleave it resulting in anti-CCP immune response. The target protein is not one citrullinated protein but hundreds of citrullinated proteins as mentioned above.

One study has shown that anti-CCP antibodies activate the complement system *in vitro via* the CP and the AP but not by the LP in RA (179). In this study, anti-CCP antibodies from all 60 patients activated the complement system. This important observation leads to the evidence that complement activation can play very important role in the pathogenesis of RA in ACPApositive patients but not all RA patients are ACPA-positive. Later on, it has been shown that citrullination locally in the joints can increase inflammation indicating the direct target of ACPA (193) and it will be consistent with the accepted paradigm that complement activation at the site of antibody recognition of citrullinated antigens can cause damage.

Furthermore, IgM RF and IgA RF amplify complement activation mediated by ACPA-IC (194). These authors concluded that ACPA-IC incorporating IgM or IgA RF participate in the triggering of the inflammation-promoting activation of complement cascades occurring in RA joints. The ACPA test has been used to classify the RA into two disease subsets, i.e., ACPA-positive (which includes the HLA-DR shared epitope subset) and ACPAnegative (no HLA shared epitope association is present) (195). These authors concluded that ACPA-positive RA is genetically different from the ACPA-negative RA. The possibility remains that ACPA-positive and ACPA-negative RA patients have differential level of IgG galactoyslation and carbamylation patterns, thereby activating different pathways of the complement system. This area has not been explored in-depth and it could provide some clues regarding the direct role of complement system in the pathogenesis of ACPA-negative patients.

Recently, anti-CarP antibodies have been described in 16% ACPA-negative RA patients (196) and up to 46% patients with RA in various clinical studies. This led to the hypothesis that anti-CarP are closely related ACPA. ACPA recognize targets that are the result of the enzymatic process whereby arginine is converted into citrulline, while anti-CarP antibodies are the result of a chemical process in which lysine have been converted into homocitrulline (196). A few studies have shown that, similar to ACPA, anti-CarP antibodies are found before the onset of clinical symptoms of arthritis (119, 196). Anti-CarP antibodies have been found even in animal models of arthritis prior to the onset of disease, and its relevance will be discussed later (197). Furthermore, the significant association with radiological progression of anti-CarP IgG in ACPA-negative RA patients strongly suggested that anti-CarP antibody can also be used as a biological marker to diagnose a high risk RA population (196). Anti-CarP antibodies recognize many carbamylated antigens including human serum albumin, fibrinogen, and alpha-1 antitrypsin (196, 198–201). The comparative importance of anti-CarP vs. ACPA in the initiation of RA is unknown.

#### MOUSE MODELS OF HUMAN RHEUMATOID ARTHRITIS

Mouse models are commonly used to study human autoimmune diseases, including RA. Although mice do not develop arthritis naturally, arthritis that shares phenotypic, biochemical, physiological, and immunological properties similar to human RA can be induced in mice. Human RA often develops rapidly as an inflammation in one or more joints which is then followed by the development of the pannus. This acute inflammatory response to some extent has been replicated in some mouse models of inflammatory arthritis such as pristane-induced arthritis (202), zymosan-induced arthritis (203), proteoglycan-induced arthritis (204), streptococcal cell wall arthritis (205), the SKG mouse model of arthritis (206), and methylated bovine serum-induced arthritis (207). Several models are more adaptive immune-mediated or related, and include collagen-induced arthritis (CIA) (208), collagen antibody-induced arthritis (CAIA) (209), and the KBxN serum transfer model of arthritis (KBxN STA) (210, 211). Others are induced entirely by cytokines and include TNF-α transgenic mice (212) and IL-1Ra knockout mice (213). While some of these models are quite similar, in aggregate they possess different clinical, pathological, and mechanistic features that each representing a subset of the different aspects of human RA. With that caveat, here we will discuss the initiation and evolution of disease in three mouse models of human RA which are dependent on the complement system.

#### Collagen-Induced Arthritis and Complement Activation

Approximately 50 years ago, CIA was first reported in rats following an intradermal injection of CII emulsified in Freund adjuvant (214), and later on in many susceptible strains of mice (208) as well as in non-human primates (215, 216). At present, CIA has become one gold standard mouse model of human RA and is used in many laboratories to examine the effect of therapeutics for treatment of RA. Immunization with bovine type II collagen/Freund's complete adjuvant (CFA) or with chicken CII/CFA results in a severe polyarthritis disease after 3 weeks. Often a second injection is given on day 21, in some cases consisting of a second dose of CII/CFA and in some cases simply CII. The re-exposure to CII antigen continues to activate T and B cells and, in mice with the appropriate H2 alleles, creates an autoimmune disease attacking self-CII. In joints, CIA like human RA is characterized by the presence of activated synovial fibroblast like cells, pannus formation (multi-layered synovium), periosteal bone formation, cartilage surface damage, fibrin deposition, infiltration of macrophages and neutrophils, and finally ankylosis of one or more joints (208, 214). Similar to the anti-CII antibodies generated due to the presence of CII autoantigens in mice, similar antibodies to native or citrullinated CII are also present in human RA and appear to have pathophysiological significance (137, 140, 217).

In CIA, recombinant TNF-α induced an increase in anti-CII antibody levels indicating TNF-α contributes to disease development by both initiation of inflammation and production of autoantibodies (218). Anti-CII autoantibodies are generated in CIA mouse model arthritis and accumulated before the initiation of clinical signs of disease after the booster injection (219). This is somewhat similar situation in human RA where ACPA are present in the early evolution of RA before clinical signs of the disease, thereby suggesting that anti-CII or ACPA first accumulates to initiate the arthritis. In this regard, it has been observed that ACPA levels show an increase 3–5 years before the onset of clinical disease and then stabilize at a high levels (220). Perhaps, a certain threshold level of autoantibodies must be reached in a preclinical stage both in CIA and in human RA to develop disease. Additionally, in RA anti-CII autoantibodies were significantly associated with increased radiographic damage at the time of diagnosis (221).

It is interesting that not all strains of mice are equally susceptible to the CIA. Mice with the *H-2q* allele (MHC class II molecule) are highly susceptible to CIA, for example. It is thought that a particular immunodominant CII peptide region binds to this particular MHC allele with high affinity leading to a powerful anti-CII response (222–225). It is possible that a similar immunodominant CII peptide region binds to human RA associated allele HLA-DR (DR1\*0401) (222, 226). The efficacy of abatacept (Orencia®), a fusion recombinant protein consisting of extracellular domain of T lymphocyte-associated antigen 4 linked to modified Fc of human IgG1, in human RA, clearly implicates T cell activity as important for disease progression, which is mirrored by the requirement for T cell help in CIA. Abatacept selectively inhibits T cell activation by two mechanisms, i.e., by blocking the specific interaction of CD80/CD86 receptors to CD28 and also by binding to CD80 and CD86 receptors on the antigen-presenting cells, thereby inhibiting B cell immune response. Immunization of mice with CII results in activation of CII-specific B cells followed by generation of IgG2a as a part of the humoral response (227, 228). The induction phase of CIA through activation of the CP leads to the activation of the adaptive immune response and the generation of anti-CII antibodies. These anti-CII antibodies then bind to cartilage, thereby leading to the effector phase *via* ICs formation and the activation of complement on the cartilage surface. So CIA pathology like human RA is dependent on both humoral and cell-mediated immunity (224, 229).

A vital role for complement in CIA was first suggested by studies in rats, in which injection of cobra venom factor (CVF), inducing marked activation followed by depletion of complement components, led to a delay in the onset of arthritis until serum C3 levels returned to normal (230). Both IgG and C3 are deposited on the cartilage surface in CIA (231), and C3 depletion (a.k.a. de-complementation) of recipient rats with CVF also prevented passive transfer of CIA with anti-collagen Ig (232). Pretreatment of rats with soluble complement receptor type 1 Holers and Banda Complement and Arthritis

(sCR1), an inhibitor of the classical and AP C3 convertases, led to a delay in the development and progression of CIA (233). Whereas CVF could not alter the course of established CIA, sCR1 injections attenuated inflammation during active disease. Soluble CR1 binds to both C3b and C4b, leading to inhibition of C3 and C5 convertases and decreased activation of both C3 and C5. Furthermore, in CII-immunized mice, gene therapy with sCR1 delayed the development of CIA and decreased its severity (234). In addition, those mice expressing sCR1 exhibited decreased levels of anti-CII as well as markedly reduced lymph node and splenocyte proliferative responses to CII *in vitro*. Recent studies have shown that human TT32 (CR2-CR1), a potent CP and AP inhibitor, compared with human sCR1-10 attenuated CDA in mice with CIA (235). Normally, CR1 is expressed on many cells but a soluble form of CR1 is also present in human plasma therefore synthesized lacking the transmembrane and cytoplasmic domains (236). CR1 consists of four long homologous repeats (A–D), each containing seven SCR repeats (237). The first 10 SCR domains (1–10) of CR1 contain all important modalities required for pan-complement inhibition, acting as cofactors for irreversible proteolytic cleavage of C3b or C4b as well as decay accelerators for AP and CP convertases.

At that point, the relative importance of the CP or LP or AP of the complement system was not clear. One study clearly showed that complement activation by both the CP and the AP plays a deleterious role in CIA (224). Here, it was shown that *C3<sup>−</sup>/<sup>−</sup>* and *FB<sup>−</sup>/<sup>−</sup>* mice were highly resistant to CIA and demonstrated decreased CII-specific IgG Ab response. Repeated injection of CII for 3 weeks in *C3<sup>−</sup>/<sup>−</sup>* mice eventually resulted in the development of a low level of arthritis. Thus, C3 and FB deficiency ameliorate CIA, but do not fully protect against the development (224). Mouse complement receptor-related 1 gene/protein y (Crry), a C3 convertase inhibitor, plays a somewhat similar complement regulatory role as CR1. Transgenic mice overexpressing soluble Crry were generated and used for various complement related studies (238). Nonetheless, the AP activity in Crry-Tg mice was not inhibited as originally expected (238).

There was a suppression of CIA in Crry-Tg mice due to enhanced synthesis of Crry locally in the joint with decreased production of pro-inflammatory cytokines (239). The mice transgenic for Crry exhibited more inhibition of CIA than was recently observed in mice treated with a recombinant Crry-Ig fusion protein (219). It was concluded from these studies that the effects of Crry in CIA may be due both to inhibition of B cell function as well as to local blockade of production of pro-inflammatory cytokines.

More effective suppression of the complement system in disease may result from enhanced levels of complement regulatory proteins locally (knee joints) in tissues. Endogenous expression of complement regulatory proteins appears to be important in resistance to inflammatory disease as blockade of both Crry and CD59 led to more severe CIA in rats (240). These studies show that inhibition of an up-stream complement C3 or its C3 convertase can demonstrate profound effects on the initiation of CIA.

To show the role of downstream complement components such as C5 or C5a-C5aR axis in CIA, administration of anti-C5 inhibitory (BB5.1) antibody was used and was found to both prevent the initiation and decrease the severity of arthritis (219, 241). This inhibitory anti-C5 antibody prevented the cleavage of C5 into C5a and C5b, thereby blocking the terminal pathway. Furthermore, mice lacking C5 were partially resistant to CIA (242). However, in other studies, C5-deficient mice were not resistant to the CIA (243–245). So which component of the complement system C3 or C5 is important for the development of arthritis? Of the several components of complement, current evidence still points to the component C5-generated C5a as the strongest inducer of inflammation (246).

The greater inhibitory effects on CIA of an inhibitory anti-C5 antibody in comparison with Crry-Ig may be attributable to decreased levels of IL-1β and TNF α mRNA in the joints (219). To support ongoing clinical development and clinical trials, an inhibitory anti-C5aR monoclonal antibody not only completely inhibited the disease progression including reduced cartilage and bone destruction but also reduced TNF-α, IL-6, and IL-17A (247). Attempts have been made to design a recombinant vaccine to prevent CIA and also other mouse models of RA by inducing C5a-specific neutralizing antibodies without effecting C5/C5b (248). Injection of anti-rat CD59 induced spontaneous complement-dependent arthritis (240) and mice lacking CD59 are susceptible to antigen-induced arthritis (249). Thus, CIA is a valuable model of human RA to examine therapeutic intervention to block the upstream and downstream pathological by-products of the complement activation.

#### Collagen Antibody-Induced Arthritis and Complement Activation

Collagen antibody-induced arthritis can be induced in mice by injecting a mixture of five monoclonal antibodies known as ArthritoMab™ or Arthrogen-CIA® to different epitopes of the CII (250, 251). These antibodies binds to CII epitopes C11b, J1, D3, and U1 and spread across the entire CII region such as CB8, CB10, and CB11 fragments for better immune complex on the surface of cartilage to initiate arthritis.1,2 To induce CAIA in certain strains of mice and to get a 90–100% penetrance rate, the immunization with LPS following a mixture of four or five ArhritoMabs is essential (156). CAIA, like CIA, does not require the involvement of T and B cells for the priming phase and thus represents only the effector phase (252). There is evolving consensus that ACPAs predict the development of human RA (197, 253–255). It has been shown that arthritis can be introduced in mice by injecting a panel of mouse ACPAs (ACC1, ACC3, and ACC4) directed against citrullinated CII epitope (253, 256, 257). In a subset of human patients, ACPAs appears many years before the onset of disease. Furthermore, ACPAs from RA patients have been shown to activate complement *via* both the CP and the AP (179), but serum from RA patients failed to induce arthritis in mice. A somewhat similar experiment also failed to induce arthritis in DBA mice by transferring mouse mAbs against citrullinated fibrinogen (258). Nonetheless, these anti-citrullinated fibrinogen mAbs enhanced the suboptimal disease already established by

<sup>1</sup>www.mdbioproducts.com (Accessed: March 26, 2018).

<sup>2</sup>www.chondrex.com (Accessed: March 26, 2018).

the development of citrullinated antigens in the joint that are induced by the mixture of anti-CII (258). Thus, mouse models of RA clearly shows the importance of B cell generating anti-CII Abs or ACPAs which trigger the effector phase by activating the complement system. There is no mouse model of RA yet developed showing that anti-CarP autoantibodies such as anti-CII Abs can induce arthritis in mice through complement dependence. Nonetheless, similar to ACPA, their presence has been shown in mice and rhesus monkeys with arthritis (197, 200). Although the presence of ACPA in mice with arthritis is controversial, nonetheless the effector functions of anti-CII antibodies in mouse models have provided a clear picture of the pathophysiological processes or events likely to be involved in the initiation of human RA. One study has shown, using a rabbit arthritis model, that first anti-CarP antibodies might be generated from homocitrulline followed by ACPA (259). This observation alone related to the presence of anti-CarP before ACPA can have a huge impact to understanding the initiation of RA in ACPA-positive and ACPAnegative subset of patients.

It has been debated which pathway of the complement system is relevant in RA and how this pathway gets activated in human RA. CAIA using complement component gene-deficient mice has proven very useful to answer many of these questions. Some previous studies have shown that AP of the complement system is the main contributor because there were correlations between Bb and ICs levels (260, 261). The AP can be activated by IgA (261, 262) consistent with the current views regarding the mucosal (gut or lung) origin of the RA. Studies in CAIA mice have shown that the AP of complement is necessary and sufficient for the development of arthritis (263). In this study, C57BL/6 mice genetically deficient in either the AP protein FB (*FB*<sup>−</sup>/<sup>−</sup>) or in the CP component C4 (*C4*<sup>−</sup>/<sup>−</sup>) were used. CDA was markedly decreased in *FB*<sup>−</sup>/<sup>−</sup> compared with wild-type (WT) mice. Conversely, disease activity scores were not different between *C4*<sup>−</sup>/<sup>−</sup> and WT mice. Analyses of joints showed that C3 deposition, inflammation, pannus, cartilage, and bone damage scores were all significantly less in *FB*<sup>−</sup>/<sup>−</sup> as compared with WT mice. There were significant decreases in mRNA levels of C3, C4, CR2, CR3, C3aR, and C5aR in the knees of *FB*<sup>−</sup>/<sup>−</sup> as compared with *C4*−/− and WT mice with arthritis; mRNA levels for complement regulatory proteins did not differ between the three strains. The authors concluded that the AP is absolutely required for the induction of arthritis following injection of anti-CII Abs (263). In a subsequent study, it was shown that arthritis was not altered in *C1q<sup>−</sup>/<sup>−</sup>* or *MBL A/C<sup>−</sup>/<sup>−</sup>* or in *C1q<sup>−</sup>/<sup>−</sup>/MBL A/C<sup>−</sup>/<sup>−</sup>* (no CP no LP) mice. These *in vivo* CAIA results proved the ability of the AP to carry out pathologic complement activation in the combined absence of intact CP and LP (71). In this study, C3 activation results confirmed the ability of the AP to mediate IC-induced C3 activation using sera from *C4<sup>−</sup>/<sup>−</sup>* or *C1q<sup>−</sup>/<sup>−</sup>/MBL A/C<sup>−</sup>/<sup>−</sup>* or both *C1q<sup>−</sup>/<sup>−</sup>/MBL A/C<sup>−</sup>/<sup>−</sup>* mice (71).

From these studies, it was concluded that the AP amplification loop, with its ability to greatly enhance C3 activation, is necessary to mediate inflammatory arthritis induced by adherent ICs. Then, it was questioned whether CP or LP alone mediate CAIA. Later on, it was reported that *FD<sup>−</sup>/<sup>−</sup>* (CP and LP), *C1q<sup>−</sup>/<sup>−</sup>/FD<sup>−</sup>/<sup>−</sup>* (no CP no AP), and *MBL A/C<sup>−</sup>/<sup>−</sup>/FD<sup>−</sup>/<sup>−</sup>* (no LP no AP) mice all these gene-deficient mice failed to develop to CAIA (264). One thing was common among these gene-deficient mice that there was lack of the AP of complement system. But whether AP is sufficient to initiate and sustain RA in humans is unknown. The AP alone on adherent anti-CII antibodies was capable of generating C5a to a level equal to that observed with WT sera. However, the CP alone, in the absence of the AP, generated 71% less C5a than was observed with WT sera; the LP alone generated minimal C5a (264). Huge activation of C5a using sera from *C1q<sup>−</sup>/<sup>−</sup>/MBL A/C<sup>−</sup>/<sup>−</sup>* (only AP) equivalent to sera from WT mice and their huge susceptibility to CAIA further suggested that enzymes/and or proteases independent from the MBL might be activating or regulating the AP (264). To this end, it was also confirmed by using *MBL A/C<sup>−</sup>/<sup>−</sup>/ FCN A<sup>−</sup>/<sup>−</sup>*, and FCN A*<sup>−</sup>/<sup>−</sup>* mice that these ligands plays no role in CAIA (38). Mice lacking FCN B were partially protected while mice lacking Collectin 11 were susceptible to CAIA (10). FCN B is generated by the macrophages and macrophages infiltrate in the joints in mice with CAIA. These data solidify the role of LP as does the FCN B and MASP-1/3 in regulating the AP of the complement system.

While we have demonstrated that the AP plays a critical role in CAIA, we do not yet understand how it is activated in any molecular detail. Given our data, we suspect that factors independently from the LP ligands (MBL, FCN, and Collectins) such as such as MASPs might be activating the AP in CAIA.

Almost a decade ago, a landmark discovery was made regarding a LP enzyme, MASP-1/3 that entirely shifted the paradigm to the in-depth understanding of the interaction between the LP and AP of complement (265–267). It was shown by using *in vitro* studies that MASP-1/3 can cleave proFD (inactive) into FD (active). In this fashion, it seems that a MASP involved with the recognition of pathogens *via* the LP is also a critical activator of FD, a major component of the AP (265). Consistent with this observation, mice lacking the *MASP-1/3* gene have no LP and also have a defective AP (265, 266). Relating this to CAIA, we found that both *FD<sup>−</sup>/<sup>−</sup>* and *MASP1/3<sup>−</sup>/<sup>−</sup>* mice were resistant to CAIA (268) and there was no change in the status of proFD in *MASP-1/3<sup>−</sup>/<sup>−</sup>* mice before or after the induction of CAIA (38), confirming the *in vivo* role of MASP-1/3 in the cleavage of proFD into FD. *In vitro*, adherent anti-CII antibodies failed to fully restore C3 activation using AP-defective sera from *MASP-1/3<sup>−</sup>/<sup>−</sup>* or *FD<sup>−</sup>/<sup>−</sup>* mice (268) consistent with the *in vivo* CAIA resistance. It was further shown, using *ex vivo* cartilage microparticles (CMP), that MASP-1/3 proteases can cleave proFD in the knee joint microenvironment (269). Here, cultured differentiated 3T3 adipocytes were used as a surrogate for synovial adipose tissue. They produce proFD but not mature FD. On the other hand, fibroblast-like synoviocytes (FLS) derived from CIA synovium, were the main source of MASP-1/3 and were expected to process proFD to mature FD. Using CMP coated with anti-CII mAb and serum from *MASP-1/3<sup>−</sup>/<sup>−</sup>* mice as a source of FB, proFD in 3T3 supernatants was cleaved into mature FD by MASP-1/3 in FLS supernatants. The mature FD was eluted from the CMP and was not present in the supernatants from the incubation with CMP, indicating that cleavage of proFD into mature FD by MASP-1/3 occurred on the CMP. These results demonstrated that pathogenic activation of the AP may occur in the joint through IC adherent to cartilage along with the local production of necessary AP proteins by adipocytes and FLS (269). To provide another proof-of-concept experiment, *in vivo* reconstitution of MASP-1 or MASP-3, by liver derived from *FD<sup>−</sup>/<sup>−</sup>* mice, transplanted under the kidney capsule of *MASP-1/3<sup>−</sup>/<sup>−</sup>* mice and restored the cleavage of proFD into FD in the circulation of *MASP-1/3<sup>−</sup>/<sup>−</sup>* mice (270). Consistent with this, we found that sera from *MASP-1/3<sup>−</sup>/<sup>−</sup>* mice, which have defective AP, only after transplantation restored the full AP activity (270). These data confirmed that MASP-1/3 proteases of the LP are essential for the activation of the AP in mice. A new concept evolved from these studies that liver (generating MASPs) and adipose tissue (generating proFD) might acts in concert to activate the AP, thereby playing a vital role in the development of CAIA in mice (38, 43, 264, 270).

There are a number of ligands which can activate the LP. Presumably, the LP contribution to CAIA involves a subset of ligand interactions. Known candidates include MBL A, MBL C, FCN A, FCN B, and collectin 11. To address this, we examined CAIA in *FCN A<sup>−</sup>/<sup>−</sup>*, *FCN B<sup>−</sup>/<sup>−</sup>*, and *CL-11<sup>−</sup>/<sup>−</sup>* mice as mentioned earlier (10). These studies showed the important role of FCN B ligand of the LP in directly activating MASP-1 or MASP-3 to activate the AP. By contrast, we also observed partial protection in *MASP-2/sMAp<sup>−</sup>/<sup>−</sup>* mice. This was likely due to the involvement of the C4/C2 bypass pathway in CAIA (10). Given that MASP-1 and MASP-3 are splice variants derived from a single MASP-1/3 gene, it has been difficult to separate the two functionally. Nonetheless, our most recent data suggest that MASP-3, compared with MASP-1 or MASP-2, is the main driver of the AP and thus CAIA (271). In these studies, MASP-3 siRNA inhibited CAIA compared with MASP-1 or MASP-2 siRNAs (271). All of the above, *in vivo* CAIA studies, show that MASP-3 proteases of the LP regulate the AP, a finding that has also been confirmed by using *in vitro* studies by various research groups (44, 45, 265, 268, 272). There is a possibility that this amplification driven phenomena is surface-specific or disease specific as MASP-1/3 deficiency did affect kidney pathology in *MASP-1/3<sup>−</sup>/<sup>−</sup>/FH<sup>−</sup>/<sup>−</sup>* mice (273) whether this kidney pathology is different from *FD<sup>−</sup>/<sup>−</sup>* mice is unknown. It has been reported that mechanism of AP activation depends on the activator surface (47). Here, it was shown that MASP-1 inhibition prevent AP activation as well as prevent already initiated AP activity on the LPS surface but not for zymosan-induced AP activation (47). Overall, it appears that AP is not one holistic linear pathway as previously thought but it is an interwoven network of multiple pathways regulated by the LP enzymes.

Using CAIA, the effector roles of C3aR and the C3a–C3aR axis, C5aR, and the C5a–C5aR axis, and MAC deposition have also been dissected. Mice lacking C5aR were more resistant to CAIA than C3aR- or MAC-deficient mice, confirming the pivotal role C5–C5aR axis (274). These results are consistent with the concept of the predominant role of C5 over the role of C3 in the pathogenesis of CAIA and that the C5–C5aR axis is essential for CAIA, although C3aR and the MAC also played important roles. Consistent with this conclusion, *C3−/−* mice were partially protected from CAIA (71) while *C5<sup>−</sup>/<sup>−</sup>* failed to develop CAIA (242).

By contrast, why the inhibitory anti-human C5 antibody (Eculizimab, Soliris®) was not effective against RA in clinical trials is not known. There is a possibility that C5 is generated in high quantities in RA, so even high doses of antibody do not prevent C5a generation in the joints. Alternatively, it may be the case that complement plays a more important role early in disease and that eventually RA evolves to a state where complement is only one of several drivers that can each compensate for the other. In this scenario, it might be the case that only certain RA patients would be effectively treated by Eculizimab. While the liver is the major source of C5, neutrophils, macrophages, and T cells are all known to be sources of C5. Blocking of C5aR in human neutrophils using the small molecule inhibitor; PMX-53, resulted in a dose-dependent block of C5a-mediated activation but why it was unsuccessful and very disappointing in RA clinical trial is unknown (275). Possibly, the drug was cleared rapidly and never reached the joints. No doubt that preclinical studies support targeting C5aR in RA because C5a and C5aR are elevated in the joints of RA and psoriatic arthritis patients and their blockade attenuate leukocyte migration to the synovial fluid (276). Almost complete inhibition of CIA was reported using anti-mC5aR inhibitory antibody (247). Based on these preclinical studies, a fully human antibody that blocks the binding of C5a to C5aR was developed and tested in RA patients by Novo Nordisk, a pharmaceutical company. This company has conducted two Phase I clinical trials in Europe with anti-C5aR in patients with RA, where a good drug safety profile was demonstrated.3 Further results regarding the success or failure related to the phase II clinical trials by Novo Nordisk using anti-C5aR therapeutic antibody in RA patients are unknown at this points. Therefore, it is too early to make any conclusions regarding the therapeutic use of anti-C5aR antibody in the clinical settings. Furthermore, mice treated with GalNAc C5siRNAs targeting liver C5 are resistant to arthritis (277). Recently, with a new approach, CAIA mice injected with anti-C5aR antibody conjugated with C5siRNA inhibited arthritis in mice identical to the *C5aR<sup>−</sup>/<sup>−</sup>* mice with an inhibition of more than 80% of the disease (278). These results in CAIA are promising, showing selective and simultaneous inhibition of both C5aR activity and C5 mRNA production within the C5a–C5aR axis can dampen inflammatory response and attenuate arthritis in mice. Intriguingly, it has been shown in an experimental mouse model of autoimmune hemolytic anemia that C5aR activation does not necessarily involve C5 and C5a (279). This striking observation suggests that we must also consider the coordinate modulation of the FcγR system when interpreting the role of C5aR in RA.

Factor H is known to regulate the AP and due to the concurrent absence of C3 through uncontrolled complement activation in the fluid phase, *FH<sup>−</sup>/<sup>−</sup>* mice are resistant to CAIA (156). There is a variant of human FH gene, i.e., factor H-like-1 (FHL-1). There are five different forms of FHR proteins in humans (FHR-1 to FHR-5) (a.k.a. complement factor H-related proteins). Of these, FHL-1 and FHR are believed to counteract the effects of FH. At present, little is known of the effect of FHL-1 or FHR on CAIA due to the lack of experimental models. FH and FHL-1 have been shown to be expressed and secreted by synovial fibroblasts and were present

<sup>3</sup>http://ClinicalTrials.gov Identifier: NCT02151409 (Accessed: March 26, 2018).

in synovial fluid derived from patients suffering from rheumatoid or reactive arthritis (280). Endogenous FH is capable of inhibiting activation of the AP of complement on cartilage and synovium in joints *in vivo* exposed to a submaximal level of anti-CII mAb (156). This conclusion was derived from experiments in CAIA, with mice treated with rFH19-20 to prevent engagement of fulllength endogenous FH. This takes advantage of the observation that domains 19 and 20 of rFH bind to cartilage. By treating with rFH19-20, the interaction of FH with cartilage is inhibited. To evaluate the *in vivo* importance, it was found that competitive blockade by murine rFH19-20 of the binding of endogenous fluid phase FH to either cartilage or an injured FLS surface significantly increased CAIA in WT mice. Further support for the conclusion that FH plays a key role in regulating AP-induced complement deposition on cartilage and cell surfaces in the joint is derived from studies with *FH<sup>±</sup>* heterozygous-deficient mice. These mice exhibit lower circulating levels of FH but are not more susceptible to CAIA unless they are treated with rFH19-20 to disrupt tissue binding of endogenous FH. Recombinant fH19-20 impairs only surface control of the AP by FH and does not influence the systemic activation of the complement system as indicated by unchanged serum levels of C5a. Thus, FH controls AP activation on cartilage and injured FLS *in vivo* in a manner dependent on the FH SCR19-20 domain, indicating that the AP can be regulated on these joint surfaces. Mice lacking FH do not develop arthritis due to the lacking of C3 present in the circulation (156).

In contrast to FH, there are no studies showing the direct role of mouse FHR proteins which shows some sequence homologies to FH in the pathogenesis of inflammatory arthritis in mice. In mice, various transcripts of FHR proteins have been reported such as FHR-A, FHR-B, and FHR-C (278, 281, 282). One study has shown that recombinant mouse FHR-B bound to human C3b and was able to compete with human FH for C3b binding. FHR-B supported the assembly of AP convertase *via* its interaction with C3b. The authors concluded that mouse FHR-B similar to human FHR-1 and FHR-5 promoted complement activation *via* interaction with C3b and *via* competition with mouse FH (283). Similarly, it has been shown that mouse FHR-A and mouse FHR-B proteins antagonize the protective function of FH using sheep erythrocyte hemolytic assays and in two cell lines, kidney proximal tubular cell line and a human retinal pigment epithelial cell line (ARPE-19) (284). Lack of mouse FHR-C has been linked to an autoimmune disease (278). Still none of these above FHR studies have shown the direct role mouse FHR-A, FHR-B and FHR-C in mice with arthritis for *FH<sup>−</sup>/<sup>−</sup>* are resistance to CAIA and depletion of C3 in these mice occurs in *FH<sup>−</sup>/<sup>−</sup>* mice even in the presence of all mouse FHR proteins when there is no absolute competition.

These findings related to the role of AP in CAIA are very likely to be relevant to the initiation and perpetuation of arthritis in humans (103). Recent preclinical studies have shown that human TT32 (CR2-CR1), a potent CP and AP inhibitor, compared with control human sCR1-10 also significantly attenuated CDA in mice with CAIA (235). In man, circulating autoantibodies, including anti-CII Abs, are present for several years prior to the onset of clinically apparent arthritis (221). Substantial evidence suggests that in RA joint-based inflammation is initiated through Ag/Ab complexes that are present on the cartilage surface (285). The observation that only injured FLS, but not normal FLS expressing complement regulatory proteins, could exhibit C3 binding suggests that cartilage damage may precede injury to the synovium. Initial complement activation by solid phase immune complexes in the cartilage may lead to secondary damage to the FLS and thus to subsequent development of synovitis. Therefore, potent CP and AP inhibitors might be helpful clinically to attenuate cartilage damage seen in human RA. This strategy of using complement inhibitors can be very useful during the early development of RA because once ACPA antibodies are present in subjects without clinical signs of joints damage then there are 50% chances of developing RA with 3-year period. Such clinical trials "Strategy for the Prevention of Onset of Clinically-Apparent RA" or a.k.a. StopRA4 are already in progress.

#### K/BxN Serum Transfer Mouse Model of Arthritis and Complement Activation

About 20 years ago, an additional mouse model of RA, i.e., K/ BxN serum transfer arthritis (STA) was discovered (210). It is also being used extensively to examine the role of effector pathways of the autoantibodies. This RA mouse model is different from the CIA and CAIA models as disease is driven by activation of T cells that recognize a self-peptide (i.e., glucose-6-phosphate isomerase, G6PI) (286). These T cells then help B cells to generate IgG antibodies against G6PI which induces arthritis. Furthermore, either purified IgGs or serum alone from K/BxN arthritic mice, when injected into naïve mice, is capable of inducing severe arthritis (287). So G6PI autoantibodies target the G6PI antigen in the joints thereby inducing arthritis by binding to cartilage. In this model, a pooled serum from several arthritic K/BxN mice is transferred into naïve mice to induce arthritis. The isotype of G6PI autoantibodies is IgG1 which does not activate complement as compared to the anti-CII antibodies used for CAIA (288). Whether G6PI antibodies present in RA patients have any practical diagnostic value is unknown. One study has shown the presence of anti-G6PI antibodies in sera but there were no marked differences in the levels of anti-G6PI antibodies among RA, non-RA patients, and healthy controls. Also, there was no significant difference G6PI antibody levels between the active phase and the inactive phase in RA patients (289). It is also controversial as to whether synovial fibroblasts from RA patients can secrete G6PI. One study has been published showing the presence of a distinct population of cells at the surface of the synovial lining of inflamed RA joints that has a high concentration of G6PI (290). This cell population could be T cells present in the RA synovium (291). Interestingly, serum G6PI concentration, C1q/G6PI-CIC, and G6PI mRNA levels within peripheral blood mononuclear cells were significantly higher in active RA than that in non-active RA (292). This is controversial for it has been shown that G6PI is not a specific autoantigen in RA and only few autoimmune sera contains G6PI (293, 294).

Using the K/BxN STA mouse model, the severity of inflammation has been correlated with the expression of PAD2 and PAD4 in the close proximity of citrullinated fibrinogen (295).

<sup>4</sup>http://ClinicalTrials.gov NCT02603146 (Accessed: March 26, 2018).

Two isotypes of PAD2 and PAD4 have been shown to be highly expressed in the synovium of RA patients (295) and infiltrating cells neutrophils, macrophages, and mast cells are the major source these enzymes indicating local citrullination in the joints can take place. Moreover, anti-PAD4 autoantibodies are present in a subset of RA patients (296). Perhaps these autoantibodies are generated to inhibit the excessive conversion of arginine to citrulline as a defensive mechanism. Interestingly, although PAD4 is required for citrullination, PAD4-deficient mice were not protected from arthritis in the K/BxN STA model (295). Once again, human RA clinical studies reflect a different picture of the GPI autoantibodies and also the role PAD4 than the mouse models of K/BxN STA.

Despite the lack of activation of complement by anti-G6PI antibodies, it is fascinating to note that complement activation is still required for K/BxN mice to progress to RA. Studies have examined mice lacking complement components, C3 or FB or C5 in the context of the K/BxN model and have established that these genes are required for disease development (287, 297, 298), thereby showing that the AP of complement is required. Mice lacking C1q, C4, CR1, and CR2 remained susceptible to disease development in the context of K/BxN STA (287, 297, 299). These studies have shown that CP is not required for disease progression in K/BxN STA. Properdin deficiency rescued mice from complement-mediated injury and ameliorated disease in K/BxN STA and Ab neutralization of properdin in WT mice similarly protected mice from arthritis (300). Mice lacking MAC also were not protected using K/BxN STA (287) showing that MAC is not a significant mediator of disease in this model. By contrast, C6 deficiency has been shown to partially protect mice from CAIA (274) indicating MAC, i.e., the terminal pathways of the complement can play important role. Overall, K/BxN STA have provided very important information regarding the role complement in RA and illustrates the complex nature of human RA. Furthermore, most of the data in K/BxN STA are consistent with CAIA regarding the role of AP in the initiation of arthritis.

#### MOUSE MODELS OF HUMAN RHEUMATOID ARTHRITIS AND COMPLEMENT IN THE PRESENT AND FUTURE

Human RA is a complex disease. This becomes readily apparent when one considers that some patients respond well to TNF-α blockade while others do not and instead respond to T or B cell inhibition. In this regard then, it is useful to have multiple mouse models, each of which uses a different driver to ultimately produce synovitis. Given that all three models require a functional complement system, it suggests that the actions of complement are fundamental components of disease progression.

We might divide the process of disease progression in RA into two general subprocesses: initiation and the effector phase. During initiation, autoantibodies find their way to the cartilage and synovial space. These may be generated in response to a pathogen such as *P. gingivalis* or as a response to altered self (i.e., citrullination). ACPA or anti-CarP antibodies in human patients are detected years before disease becomes apparent. In CIA, bovine CII is introduced causing the production initially of anti-bovine CII and then of anti-self-CII. In CAIA, anti-self-CII antibodies are directly introduced. We believe that these autoantibodies serve to initiate the complement cascade either through the CP or the LP. This early activation of complement then initiates an inflammatory response *via* the production of C3a and C5a. As the autoantibodies are located in the joint space, the response is synovitis. The epitope to which the antibody is directed appears to be mutable. Thus, CIA mouse model replicate somewhat an identical chain of early events mostly in RA patients in the initial phases of arthritis. While, ACPA is present in many RA patients, there exist a population of ACPA-negative RA patients which presumably have initiated disease *via* a different mechanism. Although this finding challenges a universal pathogenic model for a key role of autoantibodies in all types of RA, but our preliminary data from CAIA show that even a tiny amount of anti-CII autoantibodies can still bind to the cartilage surface when these autoantibodies are completely absent in the circulation. Therefore, one cannot rule out the presence of a very low levels, i.e., below threshold levels of ACPA or anti-CarP or anti-CII autoantibodies in other body secretions such as nasal secretions or sputum or gingival crevicular fluid or saliva, when autoantibodies are completely absent in the circulation.

Once other immune cells have infiltrated the joint, synoviocytes have proliferated, and pannus has formed, RA has entered into the effector phase. Here, pannus secretes matrix metalloproteinases which act to destroy bone and cartilage while also secreting a complex mixture of cytokines, prostaglandins, and complement components to maintain the inflammatory state. Complement can play a role here as well, although this may be less prominent. Such a later stage role can be seen in the K/BxN STA model of RA. Anti-G6PD antibodies are of the IgG1 type which do not serve to activate complement. Thus here, RA is initiated by a different mechanism. However, as discussed above, components of the AP appear to be necessary for disease in this model. We suspect that in this model complement is required for the formation of pannus and thus acts in the effector phase of RA. Indeed, we find that components of the AP are essential for CAIA to progress as well.

Considering commonalities among the various mouse models of RA, it seems that the AP of complement is universally shared. Here FD cleaves and activates FB, which in turn is necessary for the formation of the C3 convertase on surfaces to amplify the complement response. MASP-3 generated by the liver has recently been identified as the protease critical for the cleavage of proFD and thus for the activation of the AP. In this regard, it we believe that MASP-3 may serve as an important clinical target for the treatment of human RA.

#### AUTHOR CONTRIBUTIONS

Both authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.

#### FUNDING

This study was supported by National Institutes of Health grant R01AR51749 to VMH (PI) and NKB (Co-I).

#### REFERENCES


microbial-binding activity. *J Immunol* (2010) 185:6096–104. doi:10.4049/ jimmunol.1002185


antigen, and beta-actin. *Arthritis Rheum* (2013) 65:69–80. doi:10.1002/ art.37720


complexes is dependent on N-glycans in IgG antibodies. *Arthritis Rheum* (2008) 58:3081–9. doi:10.1002/art.23865


antigenic target of anti-CarP antibodies in patients with rheumatoid arthritis. *J Autoimmun* (2017) 80:77–84. doi:10.1016/j.jaut.2017.02.008


non-overlapping reactivities. *Ann Rheum Dis* (2011) 70:188–93. doi:10.1136/ ard.2010.131102


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Holers and Banda. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# The c3dg Fragment of complement is superior to conventional c3 as a Diagnostic Biomarker in systemic lupus erythematosus

*Anne Troldborg1,2\*, Lisbeth Jensen3 , Bent Deleuran1,3, Kristian Stengaard-Pedersen1,2, Steffen Thiel <sup>3</sup> and Jens Christian Jensenius3*

*1Department of Rheumatology, Aarhus University Hospital, Aarhus, Denmark, 2 Institute of Clinical Medicine, Aarhus University, Aarhus, Denmark, 3Department of Biomedicine, Aarhus University, Aarhus, Denmark*

#### *Edited by:*

*Robert Braidwood Sim, University of Oxford, United Kingdom*

#### *Reviewed by:*

*Brian Reilly, Texas Tech University, United States Lubka T. Roumenina, INSERM UMRS 1138, France*

> *\*Correspondence: Anne Troldborg annetrol@rm.dk*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 14 December 2017 Accepted: 07 March 2018 Published: 26 March 2018*

#### *Citation:*

*Troldborg A, Jensen L, Deleuran B, Stengaard-Pedersen K, Thiel S and Jensenius JC (2018) The C3dg Fragment of Complement Is Superior to Conventional C3 as a Diagnostic Biomarker in Systemic Lupus Erythematosus. Front. Immunol. 9:581. doi: 10.3389/fimmu.2018.00581*

Keywords: complement system, complement activation, systemic lupus erythematosus, biomarker, diagnostics

# INTRODUCTION

Systemic lupus erythematosus (SLE) is an autoimmune disease involving loss of tolerance to self-antigens, which is manifested in the production of autoantibodies and deposition of complement-fixing immune complexes in injured tissue (1). Activation of the complement system has for decades been known as a significant contributor to SLE pathogenesis (2). However, not all mechanisms leading to complement activation are understood.

In 2012, new classification criteria of SLE were published by the SLICC group (Systemic Lupus International Collaborating Clinics) (3), and expert consensus included low complement protein 3 and 4 (C3 and C4) or low CH50 [complement hemolytic activity (4)] in the classification criteria. There is no explanation for this choice, but it reflects what has been done in the clinic for years. No suggestions of type of assays were given for the measurements of C3, C4, and CH50 in relation to the classification criteria.

The complement system, comprising more than 40 soluble and membrane bound proteins, is activated through three pathways: the classical, the alternative, and the lectin pathway (5). The classical and the lectin pathways are activated through pattern recognition. Serine proteases bound to a pattern-recognition molecule are activated upon recognition of a fitting pattern and through several enzymatic reactions this lead to the activation of the classical and lectin pathway C3-convertase (C4b2a) (6). The alternative pathway is in a state of constant activation, but is at the same time inhibited. There is a continuous hydrolyzes of C3 in the circulation, which potentially can lead to complement activation *via* the alternative pathway of complement activation. Under normal circumstances, this process is inhibited both in the circulation and at the cell surface (7, 8).

There are several ways to measure complement activation. One way is quantification of C3 and C4 protein (9) or fragments of the cleaved complement factors, e.g., C5a or C3a (10). Other assays estimate capacity of erythrocyte lysis (11) or estimate the level of soluble membrane attack complex (12, 13).

The central component in complement activation is C3 (14). When activated, C3 is cleaved into two fragments: C3a and C3b (14). C3b is further cleaved by factor I into iC3b and finally to C3dg and C3c (14). The smallest fragments C3a has a short halflive (15), while the larger fragment C3dg (37 kDa) has a longer plasma half-life of 4 h (16). C3c showed a shorter half-life than C3dg. Because of its size, C3dg can relatively easy be separated by size from the larger C3 molecules that also comprise the C3dg part [C3, C3(H2O), C3b, and iC3b]. These molecules will collectively here be termed C3′. A method for measuring the C3 split product C3dg was previously introduced using precipitation with 11% polyethylene glycol (PEG) to separate C3dg from C3′ (17). It has, however, not gained routine use. The same is the case for the modified rocket immuno-electrophoresis (18). The so-called double-decker rocket immuno-electrophoresis is used in a few clinical laboratories despite its technical challenges (19).

Our objective was to optimize an assay for the measurement of C3dg using precipitation with PEG followed by C3dg determination in the supernatant by immune assays. Furthermore, to evaluate if C3dg was superior to conventional C3 in discriminating between SLE patients and healthy controls in a cross-sectional cohort of SLE patients.

#### MATERIALS AND METHODS

#### Patients

A cross-sectional cohort of 169 SLE patients were included consecutively at the out-patient clinic at the Department of Rheumatology, Aarhus University Hospital (November, 2015 to August, 2016). Inclusion criteria were fulfilment of the 1997 American College of Rheumatology (ACR) classification criteria for SLE (20), age 18 or above, understanding and speaking Danish. Exclusion criteria were infection, ongoing cancer treatment, and incapacitation. After written consent, clinical data including disease activity, SLE disease activity index (SLEDAI) (21), accumulated organ damage, SLICC/ACR (22), and treatment information were collected. Serum and plasma samples for research purposes were drawn at the same time as samples for routine biochemical assessments. Control samples (*n* = 170) were collected from blood donors at the blood bank of Aarhus University Hospital, Denmark (December, 2014 to January, 2015) as previously described (23).

Blood collected in EDTA plasma tubes (8 ml) and polystyrene serum tubes (10 ml) (Alere #367525 and #367896) were centrifuged at 2,000 *g* for 10 min. Plasma and serum were collected and frozen immediately at −80°C. Maximum time from blood drawing until freezing was 2 h.

#### Assays for C3dg

#### Time-Resolved Immuno Fluorometric Assay (TRIFMA)

A standard for the assay was generated by activation of serum according to the recommendations of the study group for the manufacture of the International Complement Standard #2 (24). Ten milliliters of serum were incubated for 4 h at 37°C after the admixture of 1 ml heat aggregated human IgG (#007815; CSL Behring GmbH, Germany, 10 mg/ml TBS, aggregated at 63°C for 1 h) and 0.1 g zymosan (Sigma #Z4250). The activation was stopped by adding 550 µl 0.4 M EDTA and 200 µl Futhan (Sigma, 10 mg/ml H2O). Following precipitation with 11% PEG 6000 (w/v), the sample was centrifuged at 10,000 *g* 4°C for 30 min. The supernatant was collected and used as standard for the assay.

The standard curve was made by dilution of the activated serum standard 1/300 in Tris-buffered saline [0.14 M NaCl, 10 mM Tris, 14 mM sodium azide, with 0.05% (v/v) Tween 20 (TBS/Tween)], and further seven threefold dilutions.

Test samples (EDTA plasma) were pre-diluted 1/4 in TBS, and 40 µl was added to 60 µl TBS, 10 mM EDTA. Samples were kept on ice throughout mixing and dilution to inhibit complement activation. One hundred microliters of 32% PEG in H2O (w/v) were admixed and samples incubated for 1 h followed by centrifugation at 4,000 *g* 4°C for 15 min. The supernatants were withdrawn and diluted 1/200 in TBS/Tween, reaching a 2,000 fold dilution of the starting plasma sample. Duplicates of 100 µl were added to Fluoro Nunc MaxiSorb microtiter plates (Nunc, #437958 or #43791) previously coated by incubation overnight at room temperature (RT) with 100 µl rabbit anti-human C3dg (DAKO cat. no. A0063), mistakenly termed "anti-human C3d" (confirmed by correspondence with DAKO), at 5 µg/ml PBS and blocked by incubation with HSA at 1 mg HSA/ml TBS followed by wash with TBS-Tween. After each step, the wells were washed three times with TBS/Tween. Development after incubation overnight at 4°C was with the same anti-C3dg antibody as used for coating only now the antibody was biotinylated as previously described (25). One hundred microliters of biotin-anti-C3dg, 1 µg/ml TBS/Tween, were added to the wells and incubated 2 h at RT. After washing, 25 ng of Eu3<sup>+</sup>-streptavidin (Perkin Elmer #1244-360) in 100 µl TBS/Tween, 25 µM EDTA was added and incubated for 1 h at RT. After washing, 200 µl enhancement buffer (Ampliqon laboratory reagents #Q99800) was added to each well. Plates were read by time-resolved fluorometry using a DELFIAreader Victor5 + (Perkin Elmer®) full TRIFMA protocol can be found Data Sheet 2 in Supplementary Material.

#### Enzyme-Linked Immune Sorbent Assay (ELISA)

The assay was also tested in an ELISA format identical to the above, except after incubation with biotinylated antibody and wash, 100 µl horseradish peroxidase (HRP)–streptavidin (DAKO #PO397) diluted 1/500 in TBS/Tween was added and incubated 1 h at RT. After washing, 100 µl substrate (Sigma #P4922-capsules and #A9941-tablets) was added and incubated for 30 min at 37°C. Plates were read at 405 nm on Victor5+ full ELISA protocol can be found Data Sheet 1 in Supplementary Material.

#### Assay for C3**′**

The assay was performed analogous to the C3dg TRIFMA (excluding PEG precipitation). Wells were coated with rabbit anti-human C3c (DAKO #Q0368, 1 µg/ml TBS), and the development was with the same antibody biotinylated in-house at 1 µg/ml. The standard curve was constructed with dilutions of the serum pool assigned a value of 1 AU C3/ml. The test plasma samples were diluted 750,000-fold in TBS/Tw, 5 mM EDTA.

#### Optimizing the PEG Precipitation

Plasma proteins were precipitated with increasing concentrations of PEG from 10 to 19% (w/v). Precipitations were done as in the assay described above with PEG being added at 20–38% (w/v). C3dg in the supernatants was estimated as described above.

#### Gel Permeation Chromatography (GPC)

Samples of serum, supernatant of PEG precipitated activated serum and EDTA-plasma supernatant after precipitation with either 11 or 16% PEG (w/v), were subjected to GPC on a Superose 6 10/300 GL column (GE Healthcare). The running buffer was TBS/Tween, 5 mM EDTA. Samples were diluted 1:1 in buffer and 200 µl was loaded on the column. Fractions of 0.25 ml were collected in pure polystyrene microtiter plates (Nunc #249570), which were pre-blocked by incubation with TBS-Tween. C3dg in the fractions was quantified as described above.

#### Western Blots

Samples of supernatants and precipitates after admixture of PEG were added to 1/4 volume sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE) sample buffer [30 mM Tris–HCl, 10% (v/v) glycerol, 8 M urea, 3% (w/v) SDS, 0.1% (w/v) bromophenol blue, pH 8.9]. TBS/Tween was added to reach the desired sample volume (30–45 µl). Onetenth volume of dithiothreitol (DDT), 0.6 M, was added to the samples to be reduced, and iodoacetic acid, 1.4 M, 1/10 vol, was added to all samples applied to gels containing both reduced and non-reduced samples. Samples were denatured at 100°C for 3 min. Proteins were separated on 4–15% gradient gels (Bio-Rad, Criterion TGX gels #567-1083). Following electrophoresis, the proteins were blotted onto nitrocellulose membranes (Bio-Rad #170-4159). The membranes were then blocked by incubation for 30 min at RT in TBS with 0.1% Tween (v/v), washed, and developed with polyclonal rabbit anti-human-C3dg (DAKO #0063) at 1 µg/ml or polyclonal rabbit anti-human-C3c (DAKO #0368) in primary buffer [Tris-buffered saline, 1 mM EDTA, pH 7.4, with 1 mg human serum albumin (CSL Behring #109697) and 100 µg human IgG (CSL Behring #007815) per ml]. Membranes were subsequently washed and incubated with HRP-conjugated goat anti-rabbit IgG antibody (DAKO #P0448) diluted 1/3,000 in secondary buffer (TBS/Tween, 100 µg human IgG/ml, 1 mM EDTA, pH 7.4). After washing, the blots were developed with SuperSignal West Dura extended-duration substrate (Pierce), and emission recorded by a charge-coupled device camera.

#### Storage, Freeze/Thaw, and Diurnal Variation

To test the stability of the assay, regarding handling of samples, we initially tested both serum and EDTA plasma. Samples were tested after 0, 1 h, and 5 h at RT, and at 1 and 5 h at 37°C. Blood collected in EDTA was left for 5 h at RT or 5 h at 37°C before centrifugation and collection of plasma.

To test the stability of the samples to freezing/thawing cycles, a pool of EDTA plasma was aliquoted in 500 µl samples, frozen at −80°C, and thawed 1–9 times. Each thaw cycle was 1 h at RT.

Influence of diurnal variation on complement activation was investigated on six healthy individuals with samples taken at 4 h intervals through 24 h.

#### Comparison With Commercial Assays

Immunoassays were bought from Hycult Biotech, estimating a C3c neo determinant (# HK368-02), Nordic Biosite, estimating a C3d neo determinant (#KSP-305) and Quidel, estimating C4d (#A008). Seventeen SLE samples and seventeen controls were randomly picked and run on all assays according to the manufacture's description.

Twelve SLE patients were randomly picked to have C3d measured at the hospital of Vejle, Denmark, using a doubledecker rocket immuno-electrophoresis method (19).

#### Statistics

Checking the data for normality by Q–Q plots and histograms revealed that Gaussian distribution could not be assumed. Log-transformation did not improve normality significantly. Therefore, non-parametric tests were used for the statistical analysis. The Mann–Whitney *U*-test was used for comparison of plasma levels of the proteins in patients and controls and correlation analysis was performed calculating Spearman's rank correlation coefficient. For comparison of repeated measurements, the Kruskal–Wallis test was used. *p*-Values <0.05 were considered statistically significant. Stata version 12 and GraphPad Prism software package (version 6.0) were used for data management and statistical calculations.

#### Ethics

Clinical investigations were conducted according to the Declaration of Helsinki. The Danish Data Protection Agency and The Regional Committee on Health Research Ethics approved the study (case #1-10-72-214-13).

# RESULTS

#### Patients and Controls

The cohort of SLE patients (**Table 1**) was comparable to other Caucasian cohorts. Controls had a mean age of 45 (SD 15) at inclusion and 90% were female making them comparable to our SLE cohort with respect to age and gender.

#### Assay

To test whether the activation of our standard was successful, we ran GPC of non-activated serum and supernatant of PEGprecipitated serum (**Figure 1A**). Before activation (red) most C3dg is found in the large fragments (C3′), whereas after activation and precipitation of the activated serum, the supernatant (blue) contained only the smaller C3dg fragment. We further PEG-precipitated EDTA plasma to find out which PEG% would



be the best to separate C3dg from C3′ (**Figure 1B**) and found a plateau at 16% PEG, suggesting this would be the optimal concentration.

To compare the previously suggested PEG concentration of 11% (w/v) to our choice of 16% (w/v), we ran GPCs of samples after precipitation with both PEG concentrations (**Figure 1C**) (17). Using 16% PEG-precipitation reduced the C3′ peak (blue **Figure 1C**) and showed the supernatant predominantly contained C3dg.

We then investigated the efficiency of precipitation with 11 and 16% PEG by western blotting. The result shown in **Figure 1D** illustrate that 16% PEG is superior to 11% in precipitating all fragments larger than C3dg. We observed that both free C3dg and larger C3dg-containing components, C3′, were present in the supernatant after PEG-precipitation with 11% (lane 2), whereas precipitation with 16% PEG (lane 1) yielded a much cleaner separation of free C3dg from the other components, supporting the results illustrated in **Figures 1B,C**. A more detailed analysis by western blot is presented in Figure S1 in Supplementary Material in which fragments of C3 is visualized in different samples developed with either anti-C3c or anti-C3d antibodies.

We observed that the concentration of C3dg in serum increased with both the time and the temperature to which the samples were exposed (**Figure 2A**). By contrast, no significant increase in C3dg was seen in EDTA plasma (**Figure 2B**). Freeze–thaw cycles of EDTA-plasma showed an increase in C3dg concentration in samples after four freeze–thaw cycles (*p* < 0.05) (**Figure 2C**). The concentration of C3dg in plasma did not display significant diurnal variation (*p* = 0.19) (**Figure 2D**).

The performance of our assay was compared to three commercially available assays for complement activation products (**Figures 2E–G**). Significant correlations (*p*< 0.05) were observed with all assays. Furthermore, our C3dg results were compared with the results obtained by double-decker rocket immunoelectrophoresis on plasma from 12 randomly picked patients (**Figure 2H**) (*r* = 0.74, *p* < 0.05).

As ELISA is used more frequently than TRIFMA, it was considered expedient to subject the analysis of the PEG supernatants to assay by ELISA. The results were similar when comparing the two assays (*r* = 0.91, *p* < 0.0001) (**Figure 3**).

#### Complement Activation in SLE Patients and Controls

No significant difference was observed between SLE patients and controls for C3′, the value routinely determined as the C3 concentration (*p* = 0.211, **Figure 4A**). SLE patients showed higher C3dg concentrations in plasma compared with controls (*p* < 0.0001, **Figure 4B**). The C3dg/C3′ ratio was calculated for SLE patients and controls and showed a clear difference (*p* < 0.0001, **Figure 4C**).

C3dg and SLEDAI did not show any correlation (*r* = 0.03, *p* = 0.71). C3dg/C3′ correlated to SLEDAI (*r* = 0.28, *p* < 0.001) as did C3 (*r* = −0.49, *p* < 0.001), which was no surprise, since C3 is part of the SLEDAI score.

Panel (D) shows the results of western blotting of supernatants of EDTA-plasma precipitated with 16% (lane 1) or 11% (lane 2) PEG. The samples (corresponding to 0.1 µl plasma) were run non-reduced on the SDS-PAGE. The blot was developed with anti-C3d antibody. It can be seen that the 11% PEG supernatant still contains appreciable amounts of larger C3dg-encompassing molecules. This was repeated three times with similar results.

ROC-curves were made based on the measurements of C3, C3dg, and C3dg/C3 (**Figure 5**). C3dg was superior in separating patients from controls with an area under the curve of 0.96 (CI 0.94–0.98) (**Figures 5A,B**). When estimating sensitivity and specificity based on different cutoffs, C3dg yielded the best combination of both high sensitivity and high specificity (**Figure 5C**), revealing a much larger complement turnover in SLE patients than reflected by the C3 measurements.

#### DISCUSSION

Low complement C3, C4, and CH50 were introduced into the classification criteria of SLE in 2012 (3), reflecting an international acceptance of the importance of complement in the diagnosis. However, in many cases of SLE, low C3 or C4 poorly reflect disease activity, as patients often present with low levels irrespectively of disease activity (26, 27). Our newly developed C3dg assay was clearly superior to the conventional C3 measurement in discriminating SLE patients from controls both with regards to specificity and sensitivity. Thus, C3dg may be a valuable diagnostic biomarker in SLE.

Instead of simply estimating C3 or C4 and interpreting low levels as evidence for complement activation or consumption, it seems more relevant to directly evaluate ongoing complement activation. Several assays have been published for estimating complement activation products without being widely implemented. With a half-life of 4 h, C3dg is an ideal candidate for evaluating ongoing complement activation. As mentioned in the Section "Introduction," some procedures for estimating C3dg use precipitation of large proteins with PEG, followed by various assays for measurement of C3dg in the supernatant, have been published. The PEG concentration suggested in the first paper describing this idea (17), i.e., 11%, has been used in all the subsequent reports. We here re-addressed this issue and as illustrated with both GPC and western blot (**Figures 1B–D**), it appears expedient to increase the PEG concentrations to separate the large C3 components, here termed C3′, from C3dg (**Figure 1**). The PEG-C3dg assay has also been applied to samples from SLE patients. Significantly higher levels of C3dg have consistently

Figure 2 | C3dg measured on samples handled in different ways from the time of blood withdrawal until measurement. (A) shows the result using serum and (B) shows the results using EDTA plasma. Each sample was diluted 100 (tall columns) and 1,000-fold (short columns). (A) Samples frozen *t* = 0, (B) samples frozen after 60 min at room temperature (RT), (C) samples frozen after 60 min at 37°C, (D) samples frozen *t* = 5 h at RT, (E) samples frozen *t* = 5 h 37°C, (F) EDTA plasma left 1 h at RT before centrifugation, (G) EDTA plasma left 5 h at RT before centrifugation. (C) The samples were tested for sensitivity to freeze/thaw cycles using six EDTA plasma samples. Samples were frozen up to 10 times each followed by thawing at RT for 1 h. C3dg was measured after 2nd, 4th, 6th, 8th, and 10th thawing. (D) C3dg diurnal variation. Six controls had blood drawn at six time points during 24 h, and C3dg was estimated. For comparison of the repeated measurements, the Kruskal–Wallis test was used. (E) Comparison between our assay and enzyme-linked immune sorbent assay (ELISA) kit from Hycult Biotech (C3c-neo determinants). (F) Comparison with ELISA kit from Nordic Biocite (C3d). (G) Comparison to ELISA kit from Quidel (C4d). 34 samples [17 systemic lupus erythematosus (SLE) and 17 controls] were analyzed in all kits. (H) Twelve randomly chosen SLE samples had C3d measured using the double rocket immunoelectrophoresis (19). The same samples were measured in our C3dg assay. To assess the correlation between the assays, Spearman correlation was used.

plasma samples using our assay as TRIFMA and as ELISA.

Figure 4 | Comparison of concentrations in plasma of (A) C3, (B) C3dg, and (C) C3dg/C3 ratio in systemic lupus erythematosus (SLE) patients and healthy controls (measured by time-resolved immuno fluorometric assay). The Mann–Whitney *U* test was used for the comparison.

Figure 5 | Separation of systemic lupus erythematosus patients from controls using C3, C3dg, and C3dg/C3. (A) demonstrates ROC-curves for C3 (light gray), C3dg (black), and C3dg/C3 (dark gray). (B) represents calculations of area under the curve with 95% confidence intervals for each of the curves in Figure 1B. (C) shows calculations of sensitivity, specificity, and likelihood ratios for C3, C3dg, and C3dg/C3 using two different cutoffs for each measurement.

been reported in patients compared with controls (11, 17, 19, 26, 28, 29). Also, the improvement gained using the C3dg/C3 ratio have been reported (11). Considering the sizable literature on this subject, it seems surprising that C3dg assays have received little clinical attention.

The C3dg estimation proved stable using EDTA-plasma. It is clear, however, that even with the addition of EDTA, complement activation is not completely inhibited unless samples are kept cold and not thawed multiple times. This underlines the necessity of handling samples similarly when comparing cohorts (30). Control samples were stored for approximately a year longer than patient samples, and at least for EDTA plasma, storage at minus 80°C seemed to stop complement activation, as controls displayed considerable lower C3dg than patients. Another concern, when using blood for measurements is whether or not the component of interest displays diurnal variation. This is the case for several proteins and hormones (31, 32). We, however, observed no significant diurnal variation in complement activation reflected by C3dg.

The plasma concentration of C3 and C4 show inter-individual variation (33). Concentrations depend to a large extend on synthesis; however, SLE patients often show low levels of C4 due to a partial genetic defect (34, 35). Low C4, therefore, does not necessarily reflect consumption. Thus, it is dubious to use a single complement measurement for diagnostic/clinical purposes. Intuitively, therefore, it makes sense to use the C3dg/C3′ ratio, as previously suggested by Röther et al. (29). If initial concentration of intact complement components influences the concentration upon activation, it is reasonable to correct for this by calculating the C3dg/C3 ratio. Whether this measurement can be used for clinical purposes, e.g., to rule out a flare, remains to be tested on a prospective cohort. We demonstrated that the use of C3, as suggested in the SLE classification criteria, was not optimal for the SLE diagnosis. We found that C3dg was significantly higher in SLE patients than in healthy controls, whereas this was not the case for C3. C3dg showed superiority with regard to both sensitivity and specificity compared with C3. The performance of C3dg as a criterion for classification of SLE should be further evaluated by comparing SLE with other diseases.

The study raises the question of why most SLE patients have higher C3dg concentrations in plasma than controls (**Figure 4**). It likely indicates that in most patients there is complement turnover, even under conditions where patients are regarded as having inactive disease (36). The alternative pathway is an essential amplification loop in the activation of the complement system (37, 38). We know that lack of control of the alternative pathway potentially leads to devastating disease (39, 40). A possible explanation for higher C3dg in patients could be a relative lack of inhibition of the alternative pathway, making it easier to tip the balance to non-inhibition in cases of infection, cancer, sun exposure, and trauma, all known causes of SLE flares. Studies on lupus-prone mice have demonstrated exacerbation of disease with inadequate cell level complement inhibition (41). In line with this, low factor H has been associated with lupus nephritis (42). Another possible explanation could be anti-C3-antibodies. As reported by Vasilev et al., anti-C3-antibodies bind to the C3c part of C3 and thus bind C3, iC3b, C3b, and C3c (43). The antibodies likely compete for the same binding site as factor H, thereby hindering inhibition at the cell surface (43). Thus, one would expect patients with anti-C3-antibodies to have higher levels of C3dg because of more complement activation. Future studies are warranted to elucidate this in SLE patients.

We present a modified assay for measuring C3dg. The assay is simple, inexpensive, and stable. The estimation of C3dg directly reflects complement turnover independently of activation pathway. The assay differentiated excellently between SLE patients and healthy controls. We propose that C3dg measurements should be evaluated as a standard complement activation measurement in patients with SLE particularly with respect to classification.

#### ETHICS STATEMENT

Clinical investigations were conducted according to the Declaration of Helsinki. The Danish Data Protection Agency and The Regional Committee on Health Research Ethics approved the study (case #1-10-72-214-13).

# AUTHOR CONTRIBUTIONS

AT and LJ performed the laboratory experiments. AT was in charge of collecting blood samples and handling the blood samples after they were drawn. AT, BD, and KS-P handled patient inclusion and clinical assessments. ST and JJ developed the assays used in the project and supervised laboratory procedures. AT, ST, JJ, and KS-P wrote the manuscript. All authors participated in the editing of the article.

#### ACKNOWLEDGMENTS

The authors would like to acknowledge all patients participating in the project. We thank Kristina Nilsson-Ekdahl for generous advice and help comparing C3dg assays.

# FUNDING

The Danish Rheumatism Association (Grant nr: R122-A3031) and Aase and Ejnar Danielsens Fond supported the project.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00581/ full#supplementary-material.

# REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Troldborg, Jensen, Deleuran, Stengaard-Pedersen, Thiel and Jensenius. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Complement Activation in inflammatory Skin Diseases

*Jenny Giang1 , Marc A. J. Seelen2 , Martijn B. A. van Doorn3 , Robert Rissmann4 , Errol P. Prens <sup>3</sup> and Jeffrey Damman1 \**

*1Department of Pathology, Erasmus Medical Center Rotterdam, Rotterdam, Netherlands, 2Department of Nephrology, University Medical Center Groningen, Groningen, Netherlands, 3Department of Dermatology, Erasmus Medical Center Rotterdam, Rotterdam, Netherlands, 4Center for Human Drug Research, Leiden, Netherlands*

The complement system is a fundamental part of the innate immune system, playing a crucial role in host defense against various pathogens, such as bacteria, viruses, and fungi. Activation of complement results in production of several molecules mediating chemotaxis, opsonization, and mast cell degranulation, which can contribute to the elimination of pathogenic organisms and inflammation. Furthermore, the complement system also has regulating properties in inflammatory and immune responses. Complement activity in diseases is rather complex and may involve both aberrant expression of complement and genetic deficiencies of complement components or regulators. The skin represents an active immune organ with complex interactions between cellular components and various mediators. Complement involvement has been associated with several skin diseases, such as psoriasis, lupus erythematosus, cutaneous vasculitis, urticaria, and bullous dermatoses. Several triggers including auto-antibodies and micro-organisms can activate complement, while on the other hand complement deficiencies can contribute to impaired immune complex clearance, leading to disease. This review provides an overview of the role of complement in inflammatory skin diseases and discusses complement factors as potential new targets for therapeutic intervention.

Keywords: complement, dermatology, skin diseases, innate immunity, psoriasis, hidradenitis suppurativa, lupus erythematosus, bullous pemphigoid

# INTRODUCTION

#### The Complement System

The complement system consists of a network of more than 50 different plasma and membraneassociated proteins. It is a part of the innate immune system and plays a key role in host defense against pathogens as well as in tissue homeostasis. The complement system can be activated through three distinct pathways: the classical, lectin, and alternative pathway (**Figure 1**). Classical pathway activation occurs after binding of the first component C1q to antibody-antigen complexes, cell particles, or certain acute phase proteins such as C-reactive protein or serum amyloid P. Once the complement activation cascade is initiated, the attached serine proteases C1r and C1s become activated. This is followed by cleavage of, respectively, C4 and C2, leading to the formation of a C3 convertase (C4b2a) that can activate C3 (1). The lectin pathway is activated when mannan binding lectin (MBL), ficolins, and/or collectins interact with carbohydrate structures predominantly present on invading pathogens. Consequently, this activates MBL-associated serine proteases 1 and 2 (MASP1 and MASP2), subsequently cleaving C4 and C2. The classical and lectin pathways generate

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*Zoltan Prohaszka, Semmelweis University, Hungary Arvind Sahu, National Centre for Cell Science, India*

#### *\*Correspondence:*

*Jeffrey Damman j.damman@erasmusmc.nl*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 21 December 2017 Accepted: 14 March 2018 Published: 16 April 2018*

#### *Citation:*

*Giang J, Seelen MAJ, van Doorn MBA, Rissmann R, Prens EP and Damman J (2018) Complement Activation in Inflammatory Skin Diseases. Front. Immunol. 9:639. doi: 10.3389/fimmu.2018.00639*

**526**

the shared C3 convertase C4b2a, which in turn generates the opsonin C3b and anaphylatoxin C3a through cleavage of native C3. In contrast to the classical and lectin pathways, the alternative pathway is activated by low grade spontaneous hydrolysis of systemic native C3 ("C3 tickover") or *via* properdin binding to certain cell surfaces (e.g., bacteria). Hydrolyzed C3 [C3(H2O)] can associate with Factor B (FB), which is subsequently activated by Factor D. This results in the formation of a C3 convertase (C3H20Bb) that can finally cleave C3. This fluid phase C3 convertase cleaves C3 to generate C3a and C3b. C3b can covalently bind to nearby structures and provides the basis for generation of the surface bound C3 convertase C3bBb. Alternative pathway activation can also be initiated as an amplification loop when C3b, generated by either one of the three pathways, is deposited on the triggered surface and binds to FB, which also results in the formation of the surface bound C3 convertase C3bBb. C3 convertases can be stabilized by factor P (properdin). C3b, generated by the classical, lectin, or alternative pathway, combines

with the C3 convertases to form the C5 convertases (C4b2aC3b and C3bBbC3b), initiating the terminal complement cascade. C5 convertases cleave C5 into C5b and anaphylatoxin C5a, eventually resulting in the assembly of C5b-9 by combining C5b with C6-C9. C5b-9 can form channels into the cell membrane causing cell lysis. In addition to the capacity to generate C5b-9, complement split products iC3b and C3dg are known to have immunomodulating functions (1–3).

Several soluble and cell-bound regulators are capable of mediating complement regulation on different levels of the complement cascade to limit damage to self-cells (**Figure 1**). C1-esterase inhibitor (C1-INH), C4b binding protein (C4BP), FH, FI, vitronectin, and clusterin are the soluble regulators. However, if complement components still manage to deposit on the cell membrane, inactivation of complement can be achieved by cellbound regulators. These include decay-accelerating factor (DAF/ CD55), membrane cofactor protein (MCP/CD46), complement receptor-1 (CR1), and CD59 (protectin) (4).

#### Complement System and the Skin

Plasma complement proteins are predominantly synthesized in the liver by hepatocytes, although extrahepatic complement can also be produced by other cell types such as endothelial cells, epithelial cells, and immune cells. Furthermore, complement receptors and membrane regulators can also be synthesized by other types of cells, such as leukocytes, fibroblasts, adipocytes, and endothelial cells (5). An important role of extrahepatic synthesized complement components is the protection against micro-organisms and inflammation at tissue or organ level.

The skin is the body's largest organ that serves as an immune and physical barrier against pathogenic organisms, irritant, xenobiotics, allergens, UV-irradiation, and mechanical injury. The epidermis is an active immune organ equipped with immunecompetent cells, including Langerhans cells, keratinocytes, dendritic epidermal T-lymphocytes, and melanocytes, of which the keratinocyte is the predominant cell type. The dermis harbors many immune cells such as dermal dendritic cells (DC) and mast cells. Multiple reports have demonstrated that human keratinocytes are able to produce several complement proteins including complement components C3, C4, and FB of which synthesis can be regulated and enhanced by interleukin- 1α (IL-1α), interferon-γ (IFN-γ), and tumor necrosis factor-α (TNF-α). Furthermore, keratinocytes can also synthesize soluble complement regulators such FH and FI, complement receptors CR1, cC1qR, C5aR1, and CR2 and cell-bound complement regulator proteins MCP, DAF, and CD59. Importantly, IFN-γ can enhance the production of FH and FI locally, thereby preventing epidermal damage which could be caused by locally produced C3, C4, and FB (6, 7). Besides keratinocytes, also melanocytes express cell-bound complement regulator proteins DAF, MCP, and CD59, making these cells less vulnerable to autologous complement attack. For example in vitiligo, melanocytes and keratinocytes seem to express less cellbound regulators, which adds to the cells susceptibility to lysis (8). Langerhans cells and DC are the antigen-presenting cells, which have the capacity to initiate primary immune responses. Human DCs are able to produce most of the complement components and are also qualified to identify soluble and cell-bound complement effector molecules. Interestingly, the functional development of DCs depends on complement production and activation. Hence, T-cell differentiation can be indirectly determined by complement, in particular the anaphylatoxins C3a and C5a (5).

The skin is colonized by a diversity of microbes including commensals and potential pathogens. The skin plays an important role in restraining the invasion of opportunistic or pathogenic organisms. The complement system may partly impact this microbial ecosystem by regulating complement C5aR1 signaling. Inhibition of C5aR1 signaling results in lower microbial diversity, making the skin less resistant to pathogenic organisms (2). Moreover, inhibiting C5aR1 signaling also decreases expression of various chemokines, cytokines, anti-microbial peptides, and pattern recognition receptors (2). In particular, downregulation of IL-12 by complement might impact the differentiation and development of T cells (9).

Overall, the complement system forms an important bridge between the innate and adaptive immune system. Complement activation products are known to mediate chemotaxis, opsonization, and mast cell degranulation, which can contribute to the elimination of pathogenic organisms. Aberrant expression and genetic deficiencies of complement components or regulators are associated with certain skin diseases. Complement activation in different skin diseases has experienced a first period of high-level attention already in the 1970s and 1980s. However, in the following decades the interest in complement has declined, paralleling intensified research into lymphocyte function, adaptive immunity, and immunogenetic associations in skin diseases. The introduction of immunosuppressive agents targeting the adaptive immune system, for example, cyclosporine in the treatment of psoriasis, shifted the interest from innate to adaptive. A second "wave of attention" in complement occurred after the development and introduction of several new complement inhibiting agents. These complement inhibitors are now effectively being used in clinical practise in several diseases, for example, in the treatment of renal disease. The main purpose of this review is to provide an overview of the role of complement in inflammatory skin diseases and to discuss the rationale and potential targets for pharmacological intervention using complement inhibiting therapeutics.

#### PSORIASIS

Psoriasis is a chronic skin disease, affecting approximately 2–4% of the Western population. Although the exact etiology of psoriasis remains unclear, it is considered a multifactorial disease with genetic, environmental, and behavioral factors playing a role in the pathogenesis and course of the disease. Psoriasis vulgaris (plaque psoriasis) is the most common form of the disease (affecting 70% of the patients) and usually presents as symmetrical erythematous papules or plaques covered with thick silvery scales located on the extensor side of the elbows and knees, scalp, and lumbosacral area. The scales are a consequence of a hyperproliferative epidermis, which is reflected on histology by parakeratosis, alternating regular hyperplasia with elongation of the rete ridges, loss or absence of the granular layer, and dilated capillaries of the papillary dermis (**Figure 2**). In addition, the epidermis may show transepidermal migration of neutrophils, and to a lesser extent lymphocytes, to the corneal layer that results in the formation of Munro microabscesses (10, 11). A Munro microabscess is a collection of neutrophils in the stratum corneum and can be considered as one of the characteristic histological findings of psoriasis. Inflammation in the dermis is characterized by a superficial perivascular infiltrate of lymphocytes.

Already in the 1970s and 1980s, several research groups have examined the scale extracts from psoriatic lesions, unraveling the underlying mechanisms of transepidermal leukocyte migration, and Munro microabscess formation. It was found that psoriatic scales exhibit chemotactic and chemokinetic properties compared with scales from non-psoriatic patients (12, 13). Subsequently, the presence of C3a in the extracts of psoriatic scales was demonstrated (14). Deposition of C3b in the presence of immunoglobulins in stratum corneum of skin biopsies from psoriasis patients suggested activation of the classical pathway of complement (15). Furthermore, also C5a was found in psoriatic leukotactic factor (PLF) and appeared to be the most potent

chemotactic factor in psoriatic scales. Several subsequent studies were able to confirm these findings by monitoring elevated levels of anaphylatoxins (C3a, C4a, C5a, C5a des-Arg) in scales of psoriatic lesions (16–19). Deposition of complement and the presence of complement activation products in psoriatic scales might be explained by locally produced and activated complement. In addition, systemic derived complement components and/or systemic complement activation could also be an important source of complement deposition in psoriasis.

Regarding systemic complement activation, sera, and plasma from psoriasis patients showed elevated levels of complement components, complement split products, and regulator proteins including FB, Bb, C3, C3a, C3b, C4, C4a, C4d, C5b-9, C1-INH, C4BP, FH, and FI. These results suggest involvement of both classical and alternative pathways in systemic complement activation (20–24). Marley et al. demonstrated low-circulating properdin levels in psoriatic patients, also indicating alternative pathway activation (25). Ohkohchi et al. found that, although circulating levels of classical complement pathway regulators C1-INH and C4BP were slightly increased, marked increase of alternative pathway regulator proteins were found and associated with disease activity and skin involvement (23). Altogether, these findings support a prominent role for systemic alternative pathway activation in psoriasis. However, the most convincing and controlled study was performed by Fleming et al., taking into account and excluding confounders that could potentially bias results from the previously described studies. These included discrepancy in the involved pathways, infection, and different types of psoriasis. It was found that circulating C5b-9 levels did not correlate with the extent and disease activity. More importantly, the absence of specific associated components, such as the immune complexes C1r-C1s-C1-INH and C3bBbP, in the plasma of psoriasis patients with elevated C5b-9 levels is a direct evidence that systemic complement is not activated in psoriasis (20).

Thus, previously cited literature shows that local complement is activated in psoriatic scales but systemic complement is unlikely to be activated in psoriasis. Therefore, it is reasonable to assume that complement activation products in the circulation originate from spillover of local products into the circulation (26). Important questions in this respect are: what are the triggers for local complement activation and what is the overall contribution of complement activation in the pathogenesis of psoriasis.

There are several hypotheses that might explain the activation of complement and the inherent generation of C5a in psoriatic scales. Firstly, natural auto-antibodies are directed against carbohydrate antigens in the stratum corneum and this results in IgG, C3b, and/or C4b deposition, reflecting classical pathway activation (27, 28). However, one study provided evidence that IgG and C3b binding to stratum corneum also occurred in other parakeratotic lesions such as verruca vulgaris and lichen simplex chronicus. Yet these parakeratotic lesions do not evolve into psoriatic lesions, suggesting that immunoglobulin binding and complement activation within the lesional stratum corneum might not be a consequential pathogenic event in psoriasis (15). Secondly, complement is activated independent of immunoglobulins *via* direct cleavage of C5 by serine proteinases present in the stratum corneum or *via* alternative pathway activation. The alternative pathway can be activated when serum comes in direct contact with the stratum corneum after traumatic injury (29, 30). Thirdly, complement is activated by microbial products, which is supported by the fact that infections can trigger the exacerbation of psoriasis in a much higher number of patients than was previously believed (26, 31). Lastly, complement is activated in the dermis, pervades into the epidermis, causing neutrophils to migrate, and damage epidermal cells (32). Although previously cited studies have shown local complement activation in psoriatic scales, one should argue what is the source of complement in the upper epidermis. In the literature, there have been several reports of keratinocytes being a rich source of producing complement components including C3, FB, FH, and FI. Regarding the terminal complement components (C5–C9), a study revealed keratinocytes to express and release C7 and C9 (33). Moreover, both TNF-α and IFN-γ have been shown to augment the production of C3 by keratinocytes (34). Although systemic complement *activation* is unlikely to play a role in psoriasis as previously described, circulating complement components could also be a rich source for local activation.

Whatever the cause of complement activation in psoriasis might be, it is interesting to speculate about the concept that complement might be one of the initiating factors in the pathogenesis of psoriasis. Psoriasis is regarded as an auto-immune disease and therefore therapy in psoriasis has mainly focused on targeting the adaptive immune system. Clinical and mechanistic studies are generally conducted in patients with stable plaque type psoriasis (psoriasis vulgaris), characterized by an adaptive T-cell predominant phenotype. However, to study the pathogenesis of psoriasis, one has to take into account the stage, activity, and type of psoriasis. Early psoriatic lesions are characterized clinically by pinpoint papules, papulopustular lesions or as pustular psoriasis. Histologically, early psoriatic lesion show influx of neutrophils in the dermal papillae, transepidermal migration of neutrophils (squirting papillae), dermal capillary activation, tortuous dilatation, and angiogenesis (35–37). Importantly, this stage is characterized by predominance of neutrophils, in particular in pustular psoriasis. The pathogenesis is auto-inflammatory and driven by innate immune activation, in contrast to lymphocyte predominant auto-immune activation in stable plaque type psoriasis. In the complex immunopathology of psoriasis, it is thought that during innate immune activation, plasmacytoid dendritic cells (pDCs) are activated together with IL-1β and TNF-α producing keratinocytes in response to dendritic Toll-like receptor activation by DNA (LL-37). In this initiating phase, neutrophil recruitment has mainly been attributed to the release of IL-1β and TNF-α. However, anaphylatoxin C5a is the most potent chemo-attractant for various inflammatory cells including neutrophils, monocytes, and macrophages. It was hypothesized that in the earliest psoriatic neutrophil predominant lesions, complement is activated in the stratum corneum, C5a is released which can activate immune cells and keratinocytes. C5a can migrate to the dermal microvasculature, activating endothelial cells, and mast cell degranulation, which results in capillary tortuous dilatation, so characteristic for (early) psoriasis. To investigate this hypothesis, Tagami et al. analyzed the effect of intradermal PLF (mainly consisting of C5a) injections in guinea pigs. Light microscopy of skin biopsies revealed features reminiscent of actual psoriasis lesions and comparable with injection of C5a *in vivo* in healthy human volunteers (38). These histological features included dense infiltration of neutrophils in the dermis followed by increased epidermopoiesis, epidermal proliferation, and degeneration. In contrast, PLF/C5a did not induce proliferation or influenced the viability of cultured human epidermal cells (32). These results suggest that PLF/C5a requires specific circumstances necessary to exert its effect on skin, most probably the presence of neutrophils from the circulation. Altogether, besides IL-1β and TNF-α, C5a is likely to play a role in the early neutrophil predominant and auto-inflammatory phase of psoriasis. Following auto-inflammatory initiation, the response is shifted toward auto-immune activation including activation of the IL-23/IL-17 axis characteristic for stable plaquetype psoriasis. However, more and more evidence points toward bimodal immune activation in psoriasis. Alternating waves of auto-inflammatory bursts of neutrophils coexist with T-cell driven auto-immune activation (37). Complement might be one of these factors linking innate auto-inflammatory and adaptive auto-immune responses in psoriasis. Two recent studies have provided evidence supporting this theory. Treatment of mice with siRNA targeting C3 reduced skin disease in a mouse model of psoriasis (39). Furthermore, using the imiquimod-induced model of psoriasis, it was found that C3 deficient mice showed significantly reduced levels of IL-1β, TNF-α, IL-17a, and IL-23 in the skin and draining lymph nodes and decreased infiltration of neutrophils (40). These studies clearly show a link between complement activation and activation of IL-23/IL-17 axis of adaptive immunity. Clinical studies targeting complement activation might investigate whether flares of psoriasis can be prevented by targeting innate immunity instead of treating adaptive immune activation in well-established psoriatic plaques.

#### ACNE VULGARIS AND HIDRADENITIS SUPPURATIVA (HS)

#### Acne Vulgaris

Acne vulgaris is a common cutaneous disorder and its pathogenesis is multifactorial including genetic, infectious, and hormonal factors. The distribution of the skin lesions in acne vulgaris reflects that of sebaceous glands. Patients present with comedones predominantly on the face, nose, forehead, and chest (sebaceous areas). Light microscopy reveals open and closed comedones, which are a result of excessive sebum secretion, hyperkeratosis of the sebaceous duct, and follicular infundibulum, subsequently followed by hair follicle dilatation. Excessive dilatation of the follicular infundibulum eventually results in rupture of the epithelial layer. Besides distension, epithelial damage is also attributed to overgrowth of *Propionibacterium acnes* (*P. acnes*) in the hair follicle lumen. *P. acnes* is an anaerobic bacterium and a habitual follicular resident. Although data exist implicating *P. acnes* in the initiation of follicular distension and obstruction, the bacterium is primarily involved in the subsequent inflammatory response of the hair follicle. Once ruptured, secondary inflammatory changes occur, such as granulomatous inflammation in response to epithelial fragments, and chronic-active inflammation (41, 42).

Today, it is still not known what is the sequence of events that is responsible for evolution of a non-inflamed into an inflamed acne lesion. Evidence exists that complement activation could be one of the primary triggers in this inflammatory response in acne. In 1966, Puhvel et al. revealed that high levels of complement fixing antibodies are present in patients with severe acne compared with patients with mild or no acne (43). Scott et al. analyzed noninflamed (white- and black heads) and inflamed (papules and nodules) acne lesions. The most frequent finding was perivascular granular and/or linear pilosebaceous basement membrane zone (of affected units) deposition of C3b in both non- and inflamed acne lesions (44). Similar findings were reported by Dahl et al. in early acne lesions (12). In non-inflamed acne lesions, C3b deposition preceded influx of mononuclear cells while inflamed lesions showed both C3b deposition as well as mononuclear inflammation. Kligman et al. observed that microscopic hair follicle rupture precedes clinical inflammation. In uninvolved skin of patients with acne sometimes microcomedones are seen with early inflammatory changes, with migration of neutrophils into intact follicular epithelium. In a later stage, the follicular epithelium becomes spongiotic and small foci of neutrophils can be found in the hair follicle lumen (45). These findings suggest that in the early stages of acne chemotactic factors are released from the intact hair follicle to recruit neutrophils. Complement activation products, in particular C5a, is the most likely factor involved in the chemokinesis of neutrophils that initiates the conversion of a non-inflamed into an inflamed lesion. Interestingly, C3b deposition was almost exclusively found without concurrent immunoglobulin or C1q deposition, suggesting that the alternative pathway might be primarily involved in the development of acne vulgaris. A candidate triggering factor for complement activation might be *P. acnes*, since bacteria are well known to activate complement. It has been shown that both the alternative and classical complement pathways can be activated in normal human serum upon interaction with comedonal contents as well as *P. acnes* (46–49). Another possibility is the direct activation of the alternative pathway by the stratum corneum (30). Both Dahl et al. and Scott et al. only found isolated granular C3b deposition or granular C3b with immunoglobulin deposition along the dermo-epidermal junction in a few biopsies. This could hypothetically represent immune complex formation with *P. acnes* and possibly subsequent classical pathway activation (12, 44). Whatever the trigger or route of complement activation might be, C5a is released acting as a potent chemokinetic factor in the recruitment of neutrophils. Therefore, it is plausible to hypothesize that the increased influx of neutrophils in acne vulgaris is, at least in part, dependent on complement activation and may therefore be amenable to complement inhibiting therapies.

#### Hidradenitis Suppurativa

Hidradenitis suppurativa (synonym: acne inversa) is a severe inflammatory follicular skin disease causing severe patient discomfort and psychosocial burden. HS is a common, yet unrecognized and underdiagnosed disease with a prevalence of 1–4% of the European population. Patients present in the acute stage with painful inflamed nodules (boils) in the inverse apocrine bearing regions of the body such as the groin and axilla. In a late stage, sinus tract formation occurs with formation of abscesses and scarring. Importantly, systemic therapy with immunosuppressive agents (systemic corticosteroids, dapsone, and cyclosporin) has been investigated in the past decade and has shown limited efficacy.

Although the exact pathogenesis of HS still has yet to be unraveled, different theories have been proposed during the last decade. In general, it is believed that HS is a multifactorial disease in which genetic, environmental factors (such as smoking, microbial colonization, and obesity) and the immune system interact. Follicular occlusion is thought to be the primary event in HS and is caused by follicular infundibular epithelial hyperplasia and hyperkeratosis. It is hypothesized that this could be a result of microbial overgrowth due to a deficient follicular skin immune system or, in contrast, an overactive immune system that reacts to normal skin flora. Eventually, follicular plugging and occlusion will result in hair follicle rupture. Upon rupture, classical histological features are observed as seen in a ruptured epidermal cyst or comedone such as granulomatous inflammation on keratin fibers and chronic inflammation (**Figure 3**). Although such material is rapidly cleared in ruptured epidermal/acne cysts, the inflammatory response in HS persists, leading to chronicity with formation of sinus tracts and scarring (50–52).

Although "HS" is a misnomer since involvement of sweat glands is not a central event in the pathogenesis of HS, HS is indeed characterized by suppuration with deep abscess formation. Although not classified as a neutrophilic dermatosis, influx of numerous neutrophils is a characteristic histological feature of HS. However, it is uncertain whether this is an early or late event in the pathogenesis of this disease. The inflammatory response in HS is characterized by increased levels of IL-1β and TNF-α, both in (peri)lesional skin biopsies as well as circulating levels in HS patients (53). The important role of these cytokines is illustrated with the clinical improvement observed after treatment with TNF-α antagonists infliximab and adalimumab (54). Recruitment of neutrophils has been mainly attributed to TNF-α and IL-1β, however, the role of complement activation has only recently been studied. Blok et al. found induction of complement pathway genes in lesional HS skin biopsies, indicating involvement of the complement system in this disease (55). An open label phase 2 study is currently underway with the C5a inhibitor IFX-1 in patients with moderate to severe HS (NCT 03001622). Recent results from this study demonstrated systemic complement activation in HS patients reflected by increased circulating levels of C5a and C5b-9 (56). Furthermore, increased C5a serum levels in HS patients were able to activate neutrophils, thereby contributing to HS symptoms. Up to 83%, HS clinical response rate was achieved at the endpoint of anti-C5a treatment in this phase 2 clinical study. Thus, targeting the C5a–C5aR axis may represent a promising therapeutic strategy for HS patients, most probably *via* inhibition of systemic neutrophil activation. Besides systemic activation, also local complement could potentially be activated in HS since commensal follicular skin bacteria, similar to acne vulgaris, might function as pathogen-associated molecular patterns that can activate classical and alternative complement pathways. Furthermore and in a later stage after hair follicle rupture, cellular fragments might function as danger-associated molecular patterns, which can also activate complement. Whether complement activation, and aberrant immune activation in general in HS, is a primary or secondary event to an initial hair follicle process remains to be investigated.

#### SYSTEMIC AND CUTANEOUS LUPUS ERYTHEMATOSUS (LE)

Lupus erythematosus is a heterogeneous auto-immune disease, which is characterized by the presence of elevated immune complexes, high titers of auto-antibodies against nuclear and cytoplasmic components and consumption of complement components. Deposition of immune complexes within various organs causes tissue damage, producing a broad spectrum of clinical manifestations ranging from systemic to solely cutaneous lesions. Epidemiological studies have estimated the prevalence of systemic LE (SLE) to be around 0.2–0.7% occurring predominantly in patients older than 40 years of age, with a female/male ratio of 9:1 (57). The cause of SLE is multifactorial with genetic, environmental, and hormonal contributions, also integrated in the American College of Rheumatology (ACR) guideline for SLE. This guideline requires at least four of the 11 clinical and laboratory components to be present before diagnosing SLE. Whereas the revised classification Systemic Lupus International Collaborating Clinics requires four components with one clinical and one immunologic item (58). Studies have shown that skin involvement will eventually develop in about 70% of patients with SLE, although primary cutaneous LE (CLE) mainly develops in the absence of systemic features. The probability of progression from CLE to SLE is approximately 20% (59). There are a few subsets in which CLE can be classified: acute LE, subacute (SCLE), and chronic cutaneous LE (CCLE), of which CCLE can be further subdivided into discoid LE, LE profundus, LE tumidus, chilblain lupus, and bullous LE (60).

Literature over the years suggests that the pathogenesis of SLE and CLE might share common features, along with the developmental consequences of genetic, environmental factors, and immune dysregulation. Several haplotypes and certain alleles of the major histocompatibility complex (MHC) have shown association with different subtypes of CLE. From a mechanistic point of view, MHC polymorphisms may lead to an increase of auto-reactive T-lymphocytes mediated by a selection error in the thymus (61). Furthermore, a panel of single nucleotide polymorphisms (SNPs) has been shown to be associated with CLE, including TYK2, IRF5, and CTLA4. These polymorphisms may have an influence on the IFN cytokine signaling (62).

A well-known trigger of CLE is ultraviolet light (UVL), also one of the ACR criteria of SLE, included as photosensitivity. Firstly, UVL can penetrate through different layers of the skin, causing acute inflammation as well as damage to the DNA, thereby inducing keratinocyte apoptosis (63). Accumulation of apoptotic material due to decreased clearance can result in the formation of immune complexes, which in turn can enhance the IFN-α production by pDCs. IFN-α is an important cytokine in the recruitment of Th1 and cytotoxic T cells (62). Secondly, UVL directly increases the production of TNF-α, interferons, and several interleukins of keratinocytes and dermal fibroblasts. These cytokines are responsible for the infiltration of leukocytes in skin biopsies of CLE (61).

Numerous auto-antibodies, resulting from impaired clearance of immune complexes, can be found in patients with CLE. The role of auto-antibodies in CLE is not completely understood, although some auto-antibodies have been found to be useful indicators in the prognosis of the disease. Patients with anti-RNP, anti-Sm, and anti-aPL antibodies are closely associated with high prevalence of malar rash, while patients with only anti-Ro/SSA antibodies demonstrated to have an increased risk for nephritis (62).

Besides genetic variation in certain MHC class I and II alleles, congenital deficiencies of classical complement pathway components C1q, C1r, C1s, C4, and C2 are strongly associated with the development of SLE (62). In addition, a homozygous SNP of the C1QA gene appears to be highly associated with SCLE (64). Boeckler et al. were the first to demonstrate a high prevalence of partial deficiency of C2, C4 and combined C2/C4 in patients with CLE (65). Furthermore, C4 copy number variation of *C4A* and *C4B genes* is also associated with the risk of SLE (extensively reviewed elsewhere) (66). Two concepts could potentially explain the development of SLE in patients deficient in classical pathway components. Firstly, early classical pathway components are involved in the induction of immune tolerance in germinal centers of lymph nodes. It is well known that CD35/CR1 and CD21/CR2 on follicular DC bind and present complement factors to virgin B-cells in order to differentiate self from non-self B-cells. More specifically, it was demonstrated that CD21, CD35, or C4 deficient mice resulted in high levels of anti-nuclear antibodies and severe lupus-like disease (67). Therefore, complement deficiency results in the breach of self-tolerance and subsequently the development of SLE. Secondly, complement deficiency results to the inability of efficiently clearing apoptotic cells/debris and could render them to become auto-antigens, induce auto-antibody formation and thereby SLE. Most research has been performed on the latter concept, in particular the role of C1q deficiency in LE. There are three sets of findings that link C1q to the development of SLE. Firstly, congenital C1q deficiency is the strongest genetic risk factor known for the development of SLE (68). Secondly and paradoxically, circulating levels of C1q are strongly decreased in SLE as a result of classical complement activation. Thirdly, anti-C1q antibodies can be found in approximately 33% of the patients with SLE and usually coincides with classical pathway activation (69, 70).

One of the main functions of C1q is to act as an opsonin to stimulate removal of apoptotic cell fragments. C1q can directly bind to the surface of apoptotic keratinocytes through its globular heads, which results in the formation of C3 convertases. This is followed by cleavage of C3 and the release anaphylatoxin C3a and the opsonin C3b which results in the clearance of apoptotic cells by phagocytosis (71). Furthermore, the collagenous region of C1q can bind to the calreticulin/CD91 complex, which is expressed on the surface of phagocytes. This binding enhances phagocytosis, leading to ingestion of C1q coated apoptotic cells (72). C1q deficiency therefore directly interferes with the clearance of apoptotic cell debris. Although this is interesting from a mechanistic point of view, most SLE patients are C1q sufficient and therefore other mechanisms underlie the impaired clearance of apoptotic debris in SLE. Impaired clearance of apoptotic cells could lead to secondary necrosis, causing disintegration, and high expression of potential auto-antigens on the surface of apoptotic bodies and blebs (73). This induces the production of auto-antibodies against these "lupus auto-antigens" by B-cells. In turn, auto-antibody-antigen complexes bind C1q thereby activating the classical pathway of complement (74). Together with classical pathway activation as a result of C1q bound to apoptotic blebs, this explains the paradox that SLE patient show strong ongoing complement activation with low and sometimes undetectable serum complement levels (secondary hypocomplementemia). Furthermore, it is thought that prolonged exposure of C1q bound to the surface of apoptotic blebs could become antigenic due to impaired clearance that consequently induces production of auto-antibodies to C1q (75). Recent studies have demonstrated that anti-C1q specifically interacts with C1q bound to early apoptotic cells and not to C1q bound to immunoglobulins or immune complexes. Antigen-C1q-anti-C1q complexes induce classical pathway activation also causing secondary hypocomplementemia. Besides, this results in interference of the uptake of apoptotic cells and impaired C1q-dependent phagocytosis (76, 77).

The deposition of auto-antibody-antigen complexes can be visualized in skin biopsies by immunofluorescence in the so-called "lupus band test," demonstrable as granular or linear deposits of IgG, C3b and occasionally IgA and IgM along the dermo-epidermal junction. The lupus band test can be applied to differentiate SLE from CLE and also as a prognostic parameter for patients with LE. A positive reaction can be observed in both lesional as well as non-lesional skin in SLE. However, in CLE only lesional skin shows a positive lupus band test. More importantly, sun-exposed skin biopsies of healthy individuals may also exhibit a positive lupus band test; therefore, it is essential to perform direct immunofluoresence on sun-protected skin (78). Interestingly, also MBL deposition was recently found in lesional skin of patients with SLE, indicating involvement of the lectin pathway (79).

Lastly, there seems to be no clear difference in the role of complement between SLE and CLE. The complement system, especially C1q, appears to play a crucial role in the pathogenesis of SLE and CLE. The absence of functional C1q can lead to impaired clearance of apoptotic cells, resulting in expression of auto-antigens, and induction of auto-antibody generation. Furthermore, C1q can eventually bind to these surface blebs stimulating the production of auto-antibodies to C1q itself. Despite the increased knowledge concerning the pathogenesis of SLE, it remains a complex disease in which multiple mechanisms are of significance.

#### CUTANEOUS SMALL VESSEL VASCULITIS (CSVV)

Cutaneous small vessel vasculitis is defined as inflammation of the postcapillary venules in the skin. CSVV without presence of systemic vasculitis is currently named as single organ CSVV, according to the updated Chapel Hill Nomenclature for Vasculitis (80). There are also other subtypes of CSVV with systemic involvement such as anti-neutrophil cytoplasmic antibodies (ANCA) associated vasculitis and cryoglobulinemic vasculitis. CSVV is the most common type of vasculitis with an incidence of 15 per million. The majority of the patients are adults with a slight preference for females (81). Over the years, numerous studies have revealed multiple factors taking part in the development of CSVV. The disease seems to be associated with drugs, infections, and systemic disorders such as SLE, RA, or malignancy. However, in many patients there is no identifiable cause and in these cases CSVV is then considered as a primary idiopathic entity (82, 83).

The classical presentation of CSVV usually occurs 7–14 days after exposure to a triggering agent. Palpable purpura are the hallmark of CSVV, and appear as red-purple discolorations on the skin varying from 2 to 5 mm in size. However, CSVV presents as many clinical variants, depending on the severity and duration of the disease including urticaria, pustules, erosions, and ulcers (81).

The histopathologic pattern observed in CSVV is leukocytoclastic vasculitis (LCV), which is characterized by a superficial infiltrate surrounding the postcapillary venules predominantly composed of neutrophils, leukocytoclasis, endothelial swelling, extravasation of erythrocytes, and eventually fibrinoid necrosis (**Figure 4**) (82). Moreover, the presence of eosinophils can be an indicator of drug-induced CSVV.

fibrinoid necrosis (asterisks) and erytrocyt extravasation (solid arrows).

Accumulating evidence has indicated that the complement system plays a significant role in the pathogenesis of CSVV. Immunofluorescence studies have supported this hypothesis by demonstrating deposits of immunoglobulins of IgM- and IgG classes, C3c and fibrinogen in, and around dermal vessel walls in LCV. Grunwald et al. demonstrate these immunoreactants in skin biopsies of early, mature, and healing stages of vasculitis (84). Detection of vascular IgA deposition is defined for Henoch–Schonlein purpura (85). Vascular immunoglobulin and complement were also detected in non-lesional skin, suggesting that deposition of these immunoreactants is a primary event, and not secondary to endothelial damage (81). In addition, Dauchel and colleagues reported that C3d and C5b-9 were significantly increased in plasma of patients with LCV compared with controls, whereas C3, C1q, C2, C4, and FB were within normal ranges. These findings indicate local as well as systemic complement activation in patients with LCV. Moreover, the increased level of C3d and C5b-9 has been shown to correlate with vasculitis activity. Interestingly, circulating C3d and C5b-9 levels did correlate with disease activity but not with the intensity of perivascular cutaneous depositions, suggesting that C3d and C5b-9 plasma levels are caused by systemic complement activation instead of passive diffusion from cutaneous lesions into the circulation (86).

The most widely accepted concept of the pathogenesis is deposition of immune complexes in postcapillary venules, with inflammation occurring after complement activation. As mentioned previously, multiple etiologies can trigger the formation of circulating immune complexes. These consist of foreign antigens and antibodies in slight excess of antigen, and therefore easily get trapped and subsequently deposited in small vessels. These immune complexes lead to activation of the complement cascade *via* the classical pathway. Among the activated complement products, C5a and C5b-9 seem to be mainly responsible for the endothelial damage observed in CSVV (82, 83). C5a induces activation of neutrophils and upregulation of several adhesion molecules (E-selectin, ICAM-1, and V-CAM) on the surface of endothelial cells, allowing neutrophils, and other inflammatory cells to migrate from the circulation into the site of inflammation (87). Activated neutrophils can secrete reactive oxygen species and also dysregulate the expression of serine proteinases elastase and cathepsin G, which eventually leads to endothelium damage (88). Additionally, decreased tissue plasminogen activator (t-PA) in response to released cytokines due to vascular damage leads to abnormal fibrinolysis and may eventually result in fibrin deposition as reflected on histology (89). Under physiologic conditions, various regulators secreted or/and expressed by endothelial cells maintain control of complement activation. These include C1-INH, FH, FI, vitronectin, clusterin, and CD55/DAF (90). Interestingly, Boom et al. observed that DAF expression on endothelial cells of cutaneous vasculitis was almost completely absent, while DAF expression on intraluminal erythrocytes was unaffected. Expression of DAF on the surface can restrain assembly of the terminal cascade by binding to C3b, and thereby interfering with the C5 convertases. The authors suggest that the absence of DAF might be the consequence of local downregulation of DAF synthesis and might play an important role in the pathogenesis of CSVV (91). However, there is only sparse literature available about potent downregulators of DAF.

In aggregate, the complement system appears to be associated with the development of CSVV. The interaction between complement components and activation products, and the endothelial cells seem to play an essential role in the pro-inflammatory response seen in CSVV. Activation of classical pathway of complement has been recognized in CSVV, regarding to the immune complex-mediated process. In particular, the effects of C5a and MAC on endothelial cells and neutrophils may eventually lead to structural and functional damage of the endothelium resulting in CSVV.

#### URTICARIA AND URTICARIAL VASCULITIS (UV)

#### Urticaria

Urticaria is widely held to be one of the most common skin diseases, affecting up to approximately 20% of the population during lifetime. Urticarial lesions are characteristically pruritic, edematous, erythematous papules, or wheals often with pale centers, which can merge into larger plaques and usually resolve within 24 h. Urticaria consists of a wide spectrum ranging from localized wheals to widespread recurrent wheals with angioedema. Coexisting episodes of angioedema appear to be present in approximately 40% of patients with urticaria (92). Angioedema can be characterized as edema in the deep dermal layer and subcutaneous or submucosal tissues. Acute urticaria is defined by disease duration of less than 6 weeks, whereas recurrent urticarial lesions persisting for a period beyond 6 weeks is defined as chronic urticaria (CU). CU can be further divided into chronic inducible urticaria and chronic spontaneous urticaria (CSU). Inducible urticaria is caused by a response to external triggers, including cold, heat, allergens, sunlight, sweat, and pressure (92). CSU is mainly idiopathic, although researchers have identified a subpopulation with an auto-immune etiology. The histopathology of an urticarial lesion mostly involves changes in the upper dermis, consisting of mild dermal edema, and sparse perivascular and interstitial mixed inflammatory infiltrate composed of a variable number of lymphocytes, monocytes, mast cells, eosinophils, and neutrophils (**Figure 5**) (93). The pathophysiological events in the formation of wheals involve activation of dermal mast cells and basophils by various triggers, including anaphylatoxins C3a and C5a, and also physical stimuli. It is well established that binding of antigen to antigen-specific IgE on mast cells and basophils activates and degranulates these inflammatory cells (type 1 hypersensitivity reaction). The most important active substance in urticaria is histamine. Histamine induces vasodilation, increases vascular permeability, and stimulation of sensory nerve endings leading to pain or itching. Other important mediators such as TNF-α, leukotrienes, platelet-activator, and prostaglandin D2 may also promote an inflammatory response (94). The exact trigger or cause of histamine release has not been identified in most patients with CSU. Evidence of an auto-immune etiology in approximately 45% of patients has been presented. Patients with CSU have shown to possess circulating IgG auto-antibodies directed against the α subunit of IgE (anti-FcεR1α) and IgE receptor in their sera, corresponding to a type 2 hypersensitivity reaction (95). Additionally, increased frequency of thyroid dysfunction and thyroid auto-antibodies (anti-microsomal peroxidase and anti-thyroidglobulin) are found in patients with CSU (96).

IgG auto-antibodies bound to IgE or IgE receptor on mast cells results in the activation of complement *via* the classical pathway followed by generation of C5a and C5b-9. Subsequently, C5a can mediate mast cell activation *via* C5aR1 ligation with subsequent degranulation and release of mediators, resulting in urticarial wheals (97, 98). Accumulated evidence has shown that complement is involved in degranulation of mast cells and basophils in some patients with CSU (97). Firstly, one study demonstrated that heat inactivation of serum complement of patients with CU, decreased the capacity of serum to release histamine from basophils (99). Ferrer and colleagues found that unless serum containing C2 and C5 was added to IgG derived from CU sera, histamine is not released from mast cells, implicating that activation of the classical complement pathway is required for mast cell degranulation (100). In parallel with early studies, the presence of complement, especially C5a, had a role in augmenting IgG-dependent histamine release from basophils (101). Although the immune reaction in urticaria is regarded as a type 1 and type 2 hypersensitivity reaction, also local complement deposition can be found in about one-third of chronic urticarial lesions without signs of systemic activation, suggesting a type 3 hypersensitivity reaction as well (102–105).

#### Urticarial Vasculitis

Urticarial vasculitis is a relatively uncommon disease with an estimated prevalence of 5%, occurring more often in women. Clinical cutaneous manifestations of UV consist of erythematous urticarial papules and plaques that last 24–72 h with a tendency to heal with purpura or hyperpigmentation. Patients with UV can be classified as normocomplementemic UV (NUV) or hypocomplementemic UV (HUV) depending on serum complement levels. Although NUV is mostly idiopathic, HUV is associated with a more severe form of disease and can indicate an underlying disease like SLE or hypocomplementemic urticarial vasculitis syndrome (HUVS) (106). HUVS is characterized by urticaria with hypocomplementemia, arthralgia/arthritis, glomerulonephritis, recurrent abdominal pain, and obstructive lung disease (107). According to retrospective observations, there seems to be no transition among UV subtypes (108). Histopathologically, lesions of UV reveal combined features of urticaria with superimposed vascular damage. In contrast to LCV, vascular damage in UV is more subtle with endothelial activation, sparse karyorrhexis, and focal fibrinoid necrosis present in only the minority of cases (109).

Various reports support the hypothesis that UV is an immune complex-mediated disease. Deposits of immunoglobulins, complement, and/or fibrinogen in the vessel walls are often observed by immunofluorescence in patients with UV. Additionally, immune complexes are also regularly detected in the blood circulation of patients with UV. There are numerous etiologies that can trigger the formation of immune complexes in UV, including auto-antibodies, infections, and medications. Formation of immune complexes leads to activation of the complement cascade *via* the classical pathway and subsequently generation of C3a, C5a, and C5b-9.

Figure 5 | Urticaria. Dermal edema [solid arrows in (A,B)] and a sparse superficial predominantly perivascular and interstitial infiltrate of lymphocytes and eosinophils without signs of vasculitis (dashed arrow).

Release of C5a can lead to a chain of events, including activation of neutrophils, mast cell degranulation, and eosinophil degranulation, which in turn can result in endothelial damage (106).

Wisnieski et al. were the first to suggest that the pathogenesis of HUVS might involve abnormal genetic background (110). Subsequently, one study found that genetic mutations in the DNASE1L3 gene are associated with a familial form of HUVS and SLE. The protein encoded by this gene functions as endonuclease, which is possibly responsible for the removal of DNA during apoptosis (111). This finding supports the link between SLE and HUVS. Another similar feature in both diseases is the presence of auto-antibodies against C1q that can be detected in 100% of the patients with HUVS and in approximately one-third of the patients with SLE (69, 108). Anti-C1q antibodies bind to the collagenous region of C1q and usually coincide with decreased levels of classical pathway complement components as a result of complement activation. Additionally, decreased levels of CH50, C2, and C4 were also found in patients with HUVS (106). These findings implicate that complement might contribute to the pathogenesis of HUVS.

To summarize, UV consist of a wide spectrum of cutaneous, systemic features ranging from urticaria with mild vasculitis to systemic vasculitis combined with hypocomplementemia. Classical activation of complement is involved in the pathogenesis of UV and HUVS.

#### BULLOUS PEMPHIGOID (BP)

In this section, we will review the role of complement in the most common auto-immune bullous dermatosis, BP. Although complement has also been shown to be involved in the pathogenesis of other auto-immune bullous dermatoses, such as mucous membrane pemphigoid, epidermolysis bullosa acquisita, dermatitis herpetiformis, and bullous systemic LE, the reader is referred elsewhere to more detailed publications on this topic (112–114).

Bullous pemphigoid is the most frequently encountered acquired auto-immune blistering disease most commonly affecting the elderly. Clinically, patients present with tense pruritic dome-shaped fluid filled blisters measuring up to several centimeters in diameter. The most commonly recognized BP antigens are BP180 (BPAG2) and BP230 (BPAG1) and less frequently, antibodies can be found against plectins and LAD1. Most antibodies in BP are directed against the NC16A domain of BP180, which is a major non-collagenous extracellular antigenic site. Separation in BP occurs at the level of the lamina lucida. A skin biopsy from an established blister reveals a subepidermal blister, often accompanied by a superficial dermal infiltrate of lymphocytes, eosinophils, neutrophils, mast cells, and monocytes/macrophages (**Figure 6**). These inflammatory cells are also found in the blister fluid in the cell-rich variant of BP. Immunofluorescence demonstrates specific linear IgG and/or C3b deposition along the basement membrane zone. Although deposition of complement fragments such as C3c and C1q is routinely used for diagnostic reasons, activation of the complement system has also been shown to play a crucial role in the development of disease.

Complement components and activation fragments including C1q, C3, C3c, C3d, C4, C4d, C5, C5b-9, FB, FH, and properdin have been found at the basement membrane zone and blister fluid in BP. These findings indicate involvement of both classical and alternative pathways in the pathogenesis of BP (115–117).

Recently, complement split products C3d and C4d have been shown to be deposited along the basement membrane zone in formalin fixed paraffin embedded tissue and might be a potential substitute for direct immunofluorescence in the near future (118–122).

Liu et al. developed an experimental mouse model of BP using a rabbit antibody against murine BP180NC14A, a homolog of human BP180NC16A, which is, however, very poorly conserved in murine protein. This antibody was passively transferred in neonatal mice and developed a subepidermal blistering disease closely mimicking clinical BP. This experimental model of BP has demonstrated a critical role for complement activation in BP. Mice deficient of C5 or depleted of complement by cobra venom factor showed no clinical or histological evidence of BP in contrast to wild-type (WT) or complement sufficient mice. Also, injection of F(ab')2 fragments prepared from the pathogenic rabbit anti-murine BP180 did not show induction of BP, indicating a crucial role for the IgG Fc portion in BP. It is the Fc portion of antibody that expresses antigen-binding sites for C1q, thereby initiating classical complement activation (123). Using the same BP model, C4 deficient and WT mice pretreated with anti-mC1q were resistant to develop BP, also strongly indicating a crucial role for classical complement activation in BP. However, FB deficient mice developed delayed and less intense blisters, indicating that also the alternative pathway is involved in BP, probably acting in concert with classical pathway activation *via* the amplification loop (124, 125). The complement-mediated development of BP appeared to be dependent on mast cells and neutrophils. Mice depleted of circulating neutrophils did not develop BP after injection of anti-mBP180. Reconstitution with C5a or injection with neutrophil chemokine IL-8 in C5 deficient mice regained susceptibility to BP (126). These findings indicate that neutrophil recruitment in BP is C5a dependent. However, mice deficient in mast cells or C5aR1 deficient mast cells failed to develop BP, despite the activation of complement and in the presence of neutrophils. The passive transfer model exhibits extensive mast cell degranulation, preceding neutrophil infiltration, and subepidermal blistering. It was found that complement-dependent chemokinesis of neutrophils is, at least in part, dependent on mast cell degranulation *via* C5aR1 (124, 127).

Although interesting, it is questionable whether this mice model can be easily translated to the human situation since BP180NC14a is poorly preserved in mice. Therefore, a humanized mice model was developed in which mouse BP180NC14A was replaced with the homologous human BP180NC16A epitope cluster region. Mice injected with anti-human BP180NC16A subsequently development BP (128). A similar model was developed by Nishie et al. introducing the human COL17 cDNA transgene into Col17-null mice (129). Importantly, the previous findings of complement-dependent development of BP were confirmed (128).

Notably, although complement plays an important role in the pathogenesis of BP, more and more evidence emerges that also complement-independent mechanisms exist. Several groups have shown direct pathogenic effects of auto-antibodies leading to depletion of COL17 in keratinocytes and contributing to skin fragility in a complement-independent manner (130–132).

Most of the previously cited studies were performed in animal or *in vitro* models, making interpretation of the data limited. However, in a large cohort (*n* = 300) patient study, Romeijn et al. have shown that in the majority of BP skin biopsies, complement deposition could be demonstrated. More interestingly, deposition of complement was related to clinical and serological disease activity strengthening the crucial role of complement in this disease (133).

### COMPLEMENT INHIBITING STRATEGIES IN SKIN DISEASES

For years, many attempts in producing complement-specific drugs have been made with limited success. The main challenge in creating complement-specific drugs has been to discover the balance between sufficient blocking of complement activity in order to prevent local tissue damage and preserving the protection of this system. As outlined in the current review, clinical and experimental studies have demonstrated variable roles of complement in the pathogenesis of skin diseases. The findings and proposed triggers for complement activation in different skin diseases are summarized in **Table 1**. With the impending recognition of neutrophil driven auto-inflammatory diseases, IL-1β and TNF-α inhibitors have now been proven beneficial for the treatment of specific auto-inflammatory diseases such as the use of adalimumab in the treatment of HS (54). The successful use of these agents is primarily based on the inhibition of neutrophil chemotaxis. As described previously, C5a acts as a potent recruiter of neutrophils by binding to C5aR1 expressed primarily by neutrophils. Moreover, C5a can trigger increased vascular permeability on endothelial cells and activate the adaptive immune system. Therapeutic intervention in C5a–C5aR axis could therefore be a promising target for the treatment of neutrophil driven skin diseases such as psoriasis and HS as well as other neutrophilic dermatoses. Selective neutralization of the C5a–C5aR interaction offers opportunities to inhibit complement activation without interfering with membrane attack complex formation and thereby the defensive potential of complement. A decade ago, C5aR antagonist PMX53 has already been evaluated and trialed in psoriasis with initial encouraging results. Orally administered cyclic peptidomimetic PMX-53 was tested in a phase 1b/2a trial in 10 patients with psoriasis and positive results in many disease measures were obtained. However, unfortunately, this drug has a very short half-life of only 70 min which severely limits its use in clinical practice (134). Besides psoriasis, intervention in C5a–C5aR axis has very recently been shown to beneficial in the treatment of HS. Blockade of C5a by an anti-C5a antibody demonstrated up to 83% HS clinical response rate at the endpoint of anti-C5a treatment in a phase 2 clinical study, indicating the importance of complement in this disease (NCT03001622). Furthermore, a promising new therapeutic agent is a small molecule C5aR inhibitor CCX168 (Avacopan). The drug can be orally administered, shows a relatively long halflife, has shown limited side effects and promising results in other neutrophil driven diseases, for example, in ANCA-associated vasculitis (135). These therapeutic agents are preferred over the use of Eculizumab (Soliris) which is a recombinant humanized


Table 1 | Overview of complement in skin diseases and the proposed triggers of complement activation.

*SC, stratum corneum; IMQ, imiquimod; HS, hidradenitis suppurativa; PAMPs, pathogen-associated molecular patterns; DAMPs, danger-associated molecular patterns; LE, lupus erythematosus; CSVV, cutaneous small vessel vasculitis; CU, chronic urticaria; UV, urticarial vasculitis; BP, bullous pemphigoid.*

monoclonal antibody directed against C5. Eculizumab intercepts the complement cascade at the terminal effector pathway by binding tightly to C5 and preventing cleavage to C5b. Consequently, this suppresses the formation of the C5b-9, increasing susceptibility to infection (136).

Besides therapeutic intervention in C5a–C5aR axis in autoinflammatory diseases, intervention in complement activation is likely to be beneficial in BP. This review discussed several experimental mouse studies of BP that demonstrated a crucial role of complement activation in this auto-immune bullous disease. Noteworthy, currently two trials with complement inhibitors are ongoing in BP: TNT-009 is a monoclonal antibody against C1s (NCT02502903) while coversin targets C5 and leukotriene B4 (http://adisinsight.springer.com/trials/700284185).

Altogether, intervention in C5a–C5aR axis and the inherent inhibition of neutrophil recruitment might be the most interesting treatment for complement-mediated inflammatory skin diseases.

#### CONCLUSION

The complement system represents far more than the originally thought function as a non-specific defense mechanism against micro-organisms. By virtue of extensive murine model studies and genome-wide association studies, our knowledge regarding the functions and hidden connections of complement has greatly expanded. Besides the protective role of complement, improper regulated activation of complement can lead to extensive tissue damage. Local and systemic complement activation has been demonstrated in several skin diseases; however, whether

#### REFERENCES


complement activation has pathogenic significance in different skin diseases remains to be investigated. Although not all triggers of complement activation are fully elucidated in different skin diseases, antibody, and immune complex deposition play an important role in (classical) complement activation. However, earlier and more recent experimental studies using mice deficient in complement components have shown involvement of complement in the pathogenesis of BP and experimental psoriasis. Furthermore, preliminary results of a phase IIa clinical study in HS shows promising results using a monoclonal anti-C5a antibody. These result indicate that complement mediates, at least in part, inflammatory skin diseases, in particular neutrophil driven diseases. Therefore, inhibition of neutrophil recruitment through intervention in the C5a–C5aR axis might be the most interesting target of intervention in inflammatory skin diseases.

#### AUTHOR CONTRIBUTIONS

JG, MS, RR and JD wrote the article. MV, EP, and JD edited and approved the final manuscript.


of psoriasis to immunofluorescence findings. *Arch Dermatol Res* (1978) 261(2):135–46. doi:10.1007/BF00447158


and treatment. *Clin Cosmet Investig Dermatol* (2017) 10:105–15. doi:10.2147/ CCID.S111019


vasculitic skin lesions. *J Dermatol Sci* (1991) 2(4):308–15. doi:10.1016/ 0923-1811(91)90055-3


dermatitis herpetiformis. *Lancet* (1973) 2(7822):175–7. doi:10.1016/S0140- 6736(73)93006-7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Giang, Seelen, van Doorn, Rissmann, Prens and Damman. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Collectin-11 (CL-11) Is a Major Sentinel at Epithelial Surfaces and Key Pattern Recognition Molecule in Complement-Mediated Ischaemic Injury

#### Christopher L. Nauser\*, Mark C. Howard, Giorgia Fanelli, Conrad A. Farrar and Steven Sacks

*MRC Centre for Transplantation, School of Immunology and Microbial Sciences, King's College London, Guy's and St. Thomas' NHS Foundation Trust, London, United Kingdom*

#### Edited by:

*Maciej Cedzynski, Institute for Medical Biology (PAN), Poland*

#### Reviewed by:

*Peter Garred, University of Copenhagen, Denmark Cordula M. Stover, University of Leicester, United Kingdom*

\*Correspondence:

*Christopher L. Nauser christopher.nauser@kcl.ac.uk*

#### Specialty section:

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

Received: *29 May 2018* Accepted: *16 August 2018* Published: *06 September 2018*

#### Citation:

*Nauser CL, Howard MC, Fanelli G, Farrar CA and Sacks S (2018) Collectin-11 (CL-11) Is a Major Sentinel at Epithelial Surfaces and Key Pattern Recognition Molecule in Complement-Mediated Ischaemic Injury. Front. Immunol. 9:2023. doi: 10.3389/fimmu.2018.02023* The complement system is a dynamic subset of the innate immune system, playing roles in host defense, clearance of immune complexes and cell debris, and priming the adaptive immune response. Over the last 40 years our understanding of the complement system has evolved from identifying its presence and recognizing its role in the blood to now focusing on understanding the role of local complement synthesis in health and disease. In particular, the local synthesis of complement was found to have an involvement in mediating ischaemic injury, including following transplantation. Recent work on elucidating the triggers of local complement synthesis and activation in renal tissue have led to the finding that Collectin-11 (CL-11) engages with L-fucose at the site of ischaemic stress, namely at the surface of the proximal tubular epithelial cells. What remains unknown is the precise structure of the damage-associated ligand that participates in CL-11 binding and subsequent complement activation. In this article, we will discuss our hypothesis regarding the role of CL-11 as an integral tissue-based pattern recognition molecule which we postulate has a significant contributory role in complement-mediated ischaemic injury.

Keywords: collectin-11, lectin pathway, complement system, innate immunity, renal ischaemia, renal transplantation

# INTRODUCTION

We have seen over many years that innate immune defense systems mounted at epithelial surfaces perform multiple and often non-immune roles. The toll-like receptor system is a classic example that has transformed our perception about the origins of innate immunity and its roles in insect development (1, 2), antimicrobial defense (3) and in the pathogenesis of some inflammatory conditions (4–6). The complement system **Figure 1** has been known about for longer (7), but the diversity of function and, in particular, its role at the interface between innate and adaptive immunity can now be re-visited as a typical model for innate immune function as well as a therapeutic target in a growing number of medical disorders.

and acts as an amplification process for the central complement component, C3, upon which both the lectin and classical pathways converge upon. Recently, MASP-1/3 was also shown to trigger the alternative pathway as well. Through a number of complement convertases the effectors of the complement system are generated. These are the anaphylatoxins, C3a and C5a, the membrane attack complex (MAC, C5b-9) as well as C3b and its metabolite C3d which mediate antigen opsonisation and cell-cell adhesion *(NB This is a generalized overview of the complement system as it specifically relates to the focus of this manuscript and is not meant to be a comprehensive depiction of all parts of the complement system)*.

#### EARLY BEGINNINGS—LOCAL COMPLEMENT SYNTHESIS

For us, the story begins with early reports detailing the significance of local complement synthesis. Colten et al. authored a number of publications that highlighted the capacity of macrophages for producing a wide range of complement components (8, 9). Well cited papers on skin cells, neural cells, gut cells, cardiac cells, and others came into vision, showing that the principle of local synthesis was almost universal in resident parenchymal cells and migratory leukocytes (10–17). Our interest, as a nephrology group starting up in the late 1980s, was caught by kidney expression of a number of complement genes in the context of inflammatory renal disease (18, 19).

It became apparent that among the variety of intrarenal cells studied, the renal tubule cells were a prolific source of complement components. The renal tubule supports vital functions in maintaining health and blocking invasive pathogens such as uropathogenic Escherichia coli. The capacity of renal tubule cells for complement synthesis was first characterized through the work of Daha and others (20–22) in Leiden and further confirmed through histological examination of both healthy and diseased tissue by several research groups (23–25). The link between inflammatory conditions primarily affecting the renal tubules and a high degree of intrarenal complement expression was striking. Notably, these diseases included ischaemic injury of the newly transplanted kidney and transplant rejection (26–28).

At the time, the techniques of tissue- or cell-specific gene deletion were not well enough established to allow interrogation of local complement gene function, in the way that would be pursued now. However, we found that we could harness skills in mouse kidney transplantation to swap kidneys between wildtype and gene-deleted mice, in both directions, to illuminate the relative importance of local complement synthesis (29–31). Our initial focus was on C3, the central and most abundant component of the complement cascade in blood. We had already shown, in human kidney transplantation, that C3 has a significant contribution to the systemic complement pool through intrarenal synthesis (32).

#### COMPLEMENT AND THE TRANSPLANTED KIDNEY

The summation of the research including knockout mouse transplant studies and clinical observational studies has demonstrated a number of important principles. Namely, it showed that local synthesis of C3 had a disproportionate influence on ischaemia-reperfusion injury relative to the systemic pool (30), and an impact was also observed in the process of cell mediated rejection of MHC mismatched kidney transplants (31). It was evident from the analyses that the renal tubule cell was the primary target for complement deposition in these conditions, and the tubule cell was also the primary site of C3 expression (30).

The evidence additionally highlighted the role of donor antigen-presenting cells (APCs) as a source of complement (33). These cells, also known as donor passenger cells, reside in the interstitial space of the donor kidney, specifically around the renal tubules. Within the first 24 h after transplantation, these cells migrate into the recipient lymphoid system where they immunize the recipient against donor MHC antigen (34). Local production of complement was shown to modify the capacity of the APC to prime the antigen-specific T cells that mediate rejection (31, 35–37).

Considerable work has gone into identifying the complement effectors generated downstream of C3 that mobilize the inflammatory and adaptive immune functions against the transplant [reviewed in (38)]. These investigations have resulted in a deeper understanding both of the roles of the anaphylatoxins, C3a and C5a, (39) and the membrane attack complex (C5b-9) on immune cells and parenchymal tissue (40).

In addition to expressing core complement components and activating enzyme-precursors, tissue-resident and migratory cells also display receptors that detect a range of biologically active complement products formed downstream of C3 cleavage (41– 43). This emphasizes the ability for cross talk between tissueresident and migratory cells within the transplant setting. It is helpful to think of the different cells that produce and detect complement as nodes in a local network, whose functions bring together and amplify the innate immune response and regulate adaptive immunity.

# LOCAL TISSUE DEFENSE

Presumably, the local synthesis of complement components serves to enhance the defense against invasive organisms. For example, the synthesis of C3 by renal tract epithelium potentially increases the efficiency with which locally invasive organisms are opsonised and subsequently eliminated at the point of entry. There is strong evidence for such a role, as the renal tubular epithelium constitutively expresses complement, and the production is rapidly upregulated in the presence of infection (44, 45). Further testament to this mechanism is that many common urinary pathogens have developed resistance to complement, including clinically relevant strains of gram negative pathogens (46–48). Not only have these strains been found to resist complement mediated lysis but they can also utilize complement to invade complement receptor expressing tubular epithelial cells (44, 49, 50). The C3b receptor, CD46, is one such receptor used by uropathogenic E. coli to evade extracellular defenses (45, 51) and is an illustration of the means by which complement resistant strains can gain an advantage against the host. Thus, the local pool of complement can be both a protector against infection and a source of tissue injury.

#### EMERGENCE OF COLLECTIN-11

Whereas the last 20-years of research has taught us much about the effector functions of locally derived complement, our knowledge of the trigger mechanisms that localize tissue injury to a particular tissue compartment has lagged behind. This may be because the changes that induce complement activation are different for each organ and as such the studies in different organs have produced mixed and sometimes contradictory results [for more information refer to (52)]. Alternatively, it could be that the focus on circulating complement has not led us to the local mechanisms that drive complement-mediated disease. It is a common observation that measurement of circulating complement does not closely correlate with biopsy evidence of complement activation within an affected organ, and this may have delayed our understanding of local disease mechanisms. If only we understood more about the structures that triggered complement activation and how they are recognized in an organ such as the kidney, we would know more about how to detect and regulate harmful signals for health benefit.

We recently reviewed the evidence for the different pattern recognition molecules that could trigger complement activation in renal ischaemia-reperfusion injury and transplantation (52, 53). There, we considered whether the classical or lectin pathways could mediate the onset of ischaemia-reperfusion injury and found no conclusive evidence of a role for the classical pathway in the genesis of the renal injury within a murine model (54, 55). The lectin pathway also, at first, seemed not to have a key role in the induction of ischaemic renal injury, since the injury—at least in mice–was independent of C4, which is a component shared by both the classical and lectin pathways (56). However, we now believe that more recent findings on CL-11 and the coupled enzyme MBL associated serine protease-2 (MASP-2) reconcile these observations, both in the context of renal ischaemic epithelial cell damage and very possibly in retinal epithelial ischaemic damage (57–59).

CL-11 is a recently described member of the lectin family of pattern recognition molecules, with known antimicrobial functions and ability to trigger complement activation via the lectin pathway (60). Reported in 2006, CL-11 was at first named kidney collectin, or CL-K1, for its abundant expression in normal renal tissue (61, 62). The renaming of CL-11 is appropriate, since it is now known that the molecule is widely expressed (60). The most obvious expression site in the kidney is the renal tubule, for this structure encompasses the largest volume in the kidney, though CL-11 is also present in the glomerular mesangium and epithelium. Despite its strong presence in the kidney, the mean concentration of CL-11 in serum is just 284 ng/mL, by ELISA measurement. Furthermore, CL-11 is known to form heteromeric complexes with Collectin Liver 1 (CL-10) in the serum. Interestingly, this heterocomplex, CL-LK, has been shown to activate complement in vitro (63). However, we hypothesize that it is the local production of complement that accounts for tissue injury. Indeed, CL-11 has been shown to be produced by renal epithelial cells, whereas CL-10 has not been definitively shown to be expressed in the kidney (64) thereby making it less likely that the CL-LK heterocomplex is participating in renal injury.

CL-11 monomers have a similar structure to other C-type lectins such as MBL and consist of a globular head followed by a neck and a collagenous tail. The head contains a carbohydrate recognition domain (CRD), and the tail contains binding sites for MASPs which are required for complement activation. The CL-11 monomers form a triplet structure that self-combines to form oligomers with higher avidity of binding to ligand (65). Since hypoxia- or hypothermia-treated epithelial cells appear to bind CL-11 avidly (57–59), it is proposed that a change in presentation of the stress-induced cellular ligand for CL-11 underpins strong attachment of the oligomeric CL-11 complex. Whether this stress-induced pattern could involve a change in orientation or distribution of ligand, or increased expression or alteration of biochemical structure, is currently under investigation. However, clues to potential binding motifs or patterns can be gleaned from understanding the binding properties of other lectins. Mannose-Binding Lectin (MBL) is a well characterized C-type lectin similar to CL-11 in that it shares a similar structure with a CRD and collagenous tail, and binds oligosaccharide ligands in a calciumdependent manner. In particular, MBL has a higher affinity for ligands that contain mannose or N-Acetyl-D-glucosamine residues (66). More recently, information has been gathered which further characterizes the MBL-oligosaccharide interaction, specifically in relation to MBL binding of lipopolysaccharide (67). Thus, our understanding of potential CL-11 ligands could be narrowed by considering and applying our knowledge of the glycan motifs that MBL recognizes.

Many molecules are normally glycosylated to some extent. In particular, fucosylated molecules are widely synthesized in normal tissues (68), and what is interesting to us is that Lfucose, the preferred monosaccharide recognized by CL-11, is also abundant in the proximal renal tubule (57). These are the very cells that express complement components in abundance. Therefore, the core components of the complement system including C3, C5 (23) and a lectin pathway trigger (CL-11) are expressed within the same hypoxia-sensitive segment of the renal tubule, where a potential binding ligand for CL-11 is also present. In vivo and ex vivo studies of epithelial cell injury suggest that hypoxia- or hypothermia-induced binding of CL-11 is followed by complement activation on the injured cell surface at sites that are specifically marked by CL-11. In a murine model of hypoxia-induced renal tubule cell stress, complement deposition was prevented by CL-11 deletion or by L-fucose blockade of the carbohydrate recognition domain of CL-11 (57). The protective effect of L-fucose blockade could also be demonstrated in wild type mice undergoing renal ischaemic insult (unpublished data). A similar injury mechanism also appears to occur in retinal pigmented epithelial cells, where hypoxia-induced membrane attack complex formation and CL-11 deposition correlate with sites of L-fucose expression (59). Thus, these findings may have potentially broad implications for diseases where complement mediated injury is thought to play a significant role.

#### CLOSING THE GAP

David and colleagues originally described the ability of renal tubule cells to spontaneously activate complement in the presence of normal human serum, and they attributed this to activation of the alternative complement pathway (69, 70). However, subsequent understanding of the lectin pathway and, in particular, the role of CL-11 now suggests another explanation (57). It points to a role for pattern recognition by CL-11 in contact with a damage-associated ligand on hypoxia-activated cells, and in turn the subsequent activation of the complement cascade by lectin-pathway associated serine proteases, i.e. MASPs. Although the alternative pathway may still play a role in hypoxia-mediated renal injury, the emerging data suggests that CL-11 is indispensable for triggering complement activation. Secondary activation of the alternative pathway could then occur either after the formation of C3b (which is an acceptor for factor B of the alternative pathway), or by MASP-1/3-mediated cleavage of complement factor D, which in turn cleaves factor B (60, 71). In the current model, CL-11 is expressed at physiological levels in the kidney and only substantially increases in binding to the tubule following cell stress. Complement activation then occurs at the sites where CL-11-MASP complexes form.

MASP-2 and its relationship with CL-11 are thought to play an essential role in this model. MASP-2 is one of the three MASPs that are physically linked with complement-activating lectins, including CL-11. MASP-2 differs from MASP-1 and MASP-3, in

that only MASP-2 can directly cleave C3 in human and murine sera (64, 72). It is only recently that studies in gene-deleted mice have not only confirmed the importance of MASP-2 in mediating renal, cardiac and intestinal ischaemic injuries, but have shown that the injury in each case was C4-independent (30, 58, 73). The evidence supports a pathway of injury in which stress-induced ligand presentation leads to CL-11 binding and subsequent MASP-2-mediated triggering of complement activation using a route that involves the direct cleavage of C3 by MASP-2 (60, 71, 74). As indicated above, MASP-1/3-mediated activation of the alternative complement pathway could be a supplementary mechanism of injury triggered by CL-11.

In total, the evidence suggests that the renal epithelial cell behaves as a unit of host defense and inflammation, in which CL-11 appears to integrate the detection of cell stress signals with activation of the complement system. We envisage that CL-11 is primarily a tissue-based pattern recognition molecule that binds damage associated molecular patterns (DAMPs) **Figure 2**. In the presence of tissue stress from non-infectious causes, the effector actions of CL-11 appear to be misdirected by glycan ligands that are inappropriately exposed on the renal tubule cell surface. Similar cellular responses to stress such as hypoxia probably exist at other surfaces, such as the retinal epithelium, where hypoxia-induced complement activation in the presence of CL-11 is also described (59). What remains unclear is the biochemical structure of the damage-related glycan ligands (and carrier molecules) recognized by CL-11. Future studies will need to elucidate the nature of the stress signal on cells and how it is induced by hypoxia to fully validate the hypothesis, and show whether the same or similar mechanism operates at other epithelial surfaces. As the identity of these DAMPs begin to emerge, it will be important to determine whether and how those exposed on renal tissue differ from the CL-11 ligands on other tissues and microbial structures, or indeed on developing tissues targeted by CL-11. There is little doubt that investigating CL-11 will provide new tools to prove the wider functions of immune surveillance at different sites, with the potential to develop new clinical agents for detecting or blocking specific patterns.

#### AUTHOR CONTRIBUTIONS

CN and MH participated in manuscript writing, editing, and figure development. GF contributed to figure development. CF and SS contributed to manuscript editing, and writing, respectively.

#### FUNDING

Some of the work described was supported by Medical Research Council grants MR/J006742/1, MR/L020254/1, G1001141, MR/J004553/1, MR/M007871/1, and MR/L012758/1, European Research Council (ERC-2012-ADG\_20120314), UK

#### REFERENCES


Regenerative Medicine Platform, and by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy's and St Thomas's NHS Foundation Trust and King's College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.

blood monocytes. J Exp Med. (1980) 151:501–16. doi: 10.1084/jem.15 1.3.501


initiation complex. J Innate Immun. (2013) 5:242–50. doi: 10.1159/000 345356


of C4 and/or C2. FASEB J. (2017) 31:2210–9. doi: 10.1096/fj.2016 01306R

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Nauser, Howard, Fanelli, Farrar and Sacks. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Heme Drives Susceptibility of Glomerular Endothelium to Complement Overactivation Due to Inefficient Upregulation of Heme Oxygenase-1

#### Edited by:

Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France

#### Reviewed by:

Viviana P. Ferreira, University of Toledo, United States Mariusz Z. Ratajczak, University of Louisville Physicians, United States

#### \*Correspondence:

Marie Frimat marie.frimat@chru-lille.fr Lubka T. Roumenina lubka.roumenina@crc.jussieu.fr

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 29 June 2018 Accepted: 05 December 2018 Published: 20 December 2018

#### Citation:

May O, Merle NS, Grunenwald A, Gnemmi V, Leon J, Payet C, Robe-Rybkine T, Paule R, Delguste F, Satchell SC, Mathieson PW, Hazzan M, Boulanger E, Dimitrov JD, Fremeaux-Bacchi V, Frimat M and Roumenina LT (2018) Heme Drives Susceptibility of Glomerular Endothelium to Complement Overactivation Due to Inefficient Upregulation of Heme Oxygenase-1. Front. Immunol. 9:3008. doi: 10.3389/fimmu.2018.03008 Olivia May 1,2,3†, Nicolas S. Merle1,4,5†, Anne Grunenwald1,3,6, Viviane Gnemmi <sup>6</sup> , Juliette Leon1,5, Cloé Payet 1,4, Tania Robe-Rybkine1,4,5, Romain Paule<sup>1</sup> , Florian Delguste<sup>2</sup> , Simon C. Satchell <sup>7</sup> , Peter W. Mathieson<sup>8</sup> , Marc Hazzan2,3, Eric Boulanger <sup>2</sup> , Jordan D. Dimitrov 1,4,5, Veronique Fremeaux-Bacchi 1,9, Marie Frimat 2,3 \* † and Lubka T. Roumenina1,4,5 \* †

1 INSERM, UMR\_S 1138, Centre de Recherche des Cordeliers, Paris, France, <sup>2</sup> INSERM, UMR 995, Lille, France, <sup>3</sup> University of Lille, CHU Lille, Nephrology Department, Lille, France, <sup>4</sup> Sorbonne Universités, UPMC Univ Paris 06, Paris, France, <sup>5</sup> Université Paris Descartes, Sorbonne Paris Cité, Paris, France, <sup>6</sup> University of Lille, INSERM, CHU Lille, Department of Pathology, UMR-S 1172 - Jean-Pierre Aubert Research Center, Lille, France, <sup>7</sup> Bristol Renal, University of Bristol, Bristol, United Kingdom, <sup>8</sup> University Lodge, University of Hong Kong, Hong Kong, Hong Kong, <sup>9</sup> Assistance Publique – Hôpitaux de Paris, Service d'Immunologie Biologique, Hôpital Européen Georges Pompidou, Paris, France

Atypical hemolytic uremic syndrome (aHUS) is a severe disease characterized by microvascular endothelial cell (EC) lesions leading to thrombi formation, mechanical hemolysis and organ failure, predominantly renal. Complement system overactivation is a hallmark of aHUS. To investigate this selective susceptibility of the microvascular renal endothelium to complement attack and thrombotic microangiopathic lesions, we compared complement and cyto-protection markers on EC, from different vascular beds, in in vitro and in vivo models as well as in patients. No difference was observed for complement deposits or expression of complement and coagulation regulators between macrovascular and microvascular EC, either at resting state or after inflammatory challenge. After prolonged exposure to hemolysis-derived heme, higher C3 deposits were found on glomerular EC, in vitro and in vivo, compared with other EC in culture and in mice organs (liver, skin, brain, lungs and heart). This could be explained by a reduced complement regulation capacity due to weaker binding of Factor H and inefficient upregulation of thrombomodulin (TM). Microvascular EC also failed to upregulate the cytoprotective heme-degrading enzyme heme-oxygenase 1 (HO-1), normally induced by hemolysis products. Only HUVEC (Human Umbilical Vein EC) developed adaptation to heme, which was lost after inhibition of HO-1 activity. Interestingly, the expression of KLF2 and KLF4—known transcription factors of TM, also described as possible transcription modulators of HO-1- was weaker in micro than macrovascular EC under hemolytic conditions. Our results show that the microvascular EC, and especially glomerular EC, fail to adapt to the stress imposed by hemolysis and acquire a pro-coagulant and complement-activating phenotype. Together, these findings indicate that the vulnerability of glomerular EC to hemolysis is a key factor in aHUS, amplifying complement overactivation and thrombotic microangiopathic lesions.

Keywords: atypical hemolytic uremic syndrome, complement system, endothelial cells, heme, heme oxygenase-1, thrombomodulin

### INTRODUCTION

The atypical hemolytic uremic syndrome (aHUS) is a rare, kidney-predominant, thrombotic microangiopathy (TMA) associated with the formation of fibrin-platelet clots in microvessels which trigger mechanical hemolysis (1). A dysregulated complement alternative pathway (AP) plays a key role in aHUS, as suggested by genetic abnormalities in AP proteins found in up to 60% of patients (2, 3) and the efficacy of anti-C5 therapy (4). Although the complement overactivation is systemic, a particular renal tropism of the TMA lesions exists, and injury of others organs is unusual in aHUS (5–10% of patients) (2, 5).

This susceptibility of renal endothelial cells (EC) to complement overactivation remains poorly understood. Despite the morphologic and functional differences between glomerular EC and other microvascular EC, no major differences were observed in the basal expression of complement regulators or complement component C3 deposits in resting state (6–10). The high expression of complement proteins in kidneys could explain the greater susceptibility of glomerular EC to complement attack. The EC-derived C3 and Factor B are, however, present at lower concentrations in kidney compared to the liver-derived C3 and FB in blood, to which endothelium is exposed in vivo. Additional factors are necessary to fully explain the vulnerability of the glomerular endothelium to complement-mediated injury and TMA in aHUS.

Cell-free hemoglobin and heme, released during hemolysis, are potent promoters of pro-inflammatory and pro-oxidant effects (11–13) and can induce EC phenotypical changes, affording them complement activating properties (8, 14, 15). Under hemolytic conditions EC also up-regulate the cytoprotective heme-degrading enzyme heme oxygenase 1 (HO-1) (15), and this overexpression was associated with increased resistance to complement activation in HUVEC (16). Nevertheless, the implication of HO-1 in the protection of glomerular endothelium has not, to our knowledge, been studied.

We here show that endothelial heterogeneity is apparent under hemolytic conditions: the microvascular EC, and particularly the glomerular EC, become vulnerable to injury through differences in their C3 regulation and heme degradation compared with macrovascular EC. These results might explain, at least in part, the renal tropism of complement-mediated TMA lesions in aHUS.

#### MATERIALS AND METHODS

#### Reagents

The oxidized form of heme (hemin, ferriprotoporphyrin IX), designated as heme, (Sigma or Frontier Scientific) and Sn(IV) mesoporphyrin IX (SnMPIX) (Frontier Scientific) were suspended at 10 mM in 50 mM NaOH, 145 mM NaCl, and therafter diluted in the appropriate vehicle (medium for culture cells, PBS for animal experimentation).

#### Animal Experimentation

All experiments were conducted in accordance with the recommendations for care and use of laboratory animals of the French Ministry of Agriculture and with the approval of the Charles Darwin Ethics Committee for animal experimentation (Paris, France) number APAFIS#3764-201601121739330v3. Six to eight-week-old female C57Bl/6 mice from Charles River Laboratories (L'Arbresle, France) were treated with intraperitoneal injection of 200 µL phosphate buffer saline (PBS, Dulbeco) or freshly prepared heme (40µmol/kg corresponding to 27µg/g body weight) at day 0 and 1. At day 2 mice were culled and the organs were then recovered: kidneys, lungs, heart and skin (1 cm²) were preserved directly in plastic molds containing optimal cutting temperature (OCT) compound, placed on dry ice and frozen. Brains were otherwise frozen in isopentane at −70◦C to preserve the tissue architecture before being placed in OCT and frozen.

#### Immunohistochemistry (IHC) and Immunofluorescence (IF)

Frozen organs sections (6µm) were fixed in acetone. The primary antibodies were: Heme Oxygenase-1 (HO-1; rabbit anti-mouse, Abcam Ab13243, 5µg/ml), C3b/iC3b (rat anti-mouse, Hycult biotech, HM1065, 1µg/ml), Thrombomodulin (TM; rabbit anti-mouse, R&D systems, MAB3894, 20µg/ml), CD31 (rat anti-mouse, Abcam Ab7388, 2µg/mL), von Willebrand Factor (vWF; sheep anti-mouse, Abcam, Ab11713, 1/100). Staining was revealed by goat anti-rabbit AF647 (Thermoscientific, A21244, 5µg/mL) or goat anti-rat AF555 (Thermoscientific, A21434, 5µg/mL). Slides were scanned by Nanozoomer (Hamamatsu) and AxioScan (Zeiss) for IHC and IF, respectively.

# Endothelial Cells Assays

#### Cells Culture

Human umbilical vein endothelial cells (HUVEC—a model of macrovascular EC), a human dermal EC line (HMEC—a microvascular EC model), conditionally immortalized (GENC) and human primary (HRGEC) glomerular EC were compared. HUVEC (n = 7 donors) were obtained from CHU Lille (France), HMEC were from ATCC (17) (US), GENC were from Dr. Satchell (Bristol, UK) (18) and HRGEC (n = 3) were from iCelltis (Toulouse, France). Cells were cultured as previously described (9, 10, 18): briefly, the growth medium was M199 20% fetal calf serum and 20% EGM2-MV (Lonza) for HUVEC, and EGM2- MV for the other cell types. HUVEC and HRGEC were used for experiments until passage 4, HMEC between passages 2–7 and GENC between passages 23 and 30. Cells were exposed to heme at indicated doses overnight or for 30 min in serum-free medium OPTI (Thermo Fisher), or overnight to the pro-inflammatory cytokines TNFα and IFNγ (PeproTech) at 10 ng/ml and 10<sup>3</sup> U/ml in complete medium, respectively. Alternatively, HUVEC were exposed to heme or SnMPIX overnight before being rechallenged, or not, with 50µM of heme. Dead cells were removed by washing. Normal human serum (NHS) was used as a source of complement.

#### Flow Cytometry

Cells were washed, detached, labeled and analyzed by flow cytometry (BD LSR II), and the data assessed using FCS Express software (De Novo software). Antibodies were diluted in PBA (PBS, BSA 0.5%, Azide 0.1%): anti-C3c (Quidel, A205, 10µg/ml), anti-FH (antiFH#1, Quidel, A229, 55µg/ml), anti-MCP-PE (Bio Rad, MCA2113, 10µg/ml), anti-DAF (Bio Rad, MCA1614, 10µg/ml). Staining was revealed by goat anti-mouse IgG PE (Beckman Coulter, IM0551). Cell viability was followed by annexin V-APC/DAPI staining (BD Bioscience).

#### RTqPCR

RNA extraction from cells was performed with a Qiagen kit. Quality and quantity of RNA were measured by an Agilent 2100 Bioanalyzer (Agilent Techonologies). RNA Integrity Number was considered acceptable if >9. After standard RT-PCR, amplification of cDNA was proceeded with following probes: actin-4332645, TM-hs00264920-s1, HO-1 hs01110250\_m1, KLF2-Hs00360439\_g1, KLF4-Hs00358836\_m1, NRF2-Hs00975961\_g1, BACH1-Hs00230917\_m1. Data were analyzed by SDS2.3 and RQ manager software. The mean cycle threshold (CT) values for both the target and internal control (β-actin) were determined for each sample. The fold change in the target gene, normalized to β-actin and relative to the expression of untreated HUVEC, was calculated as 2–11CT (19).

#### Western Blot (WB)

Cells were lysed in RadioImmunoPrecipitation Assay (RIPA) buffer and deposited at 10µg/ml on 10% pre-casted gels (Life Technology). After transfer, the HO-1 and actin were probed by a rat anti-human HO-1 IgG2b (R&D systems, MAB3776, 2µg/ml) and a rabbit anti-actin (Sigma Aldrich, A2066, 1/10000). Secondary antibodies were goat anti-rat IgG-HRP (R&D systems, HAF005, 1/5000) and goat anti-rabbit IgG-HRP (Santa Cruz Technology, sc-2004, 1/10000). Blots were revealed by chemiluminescence (Super Signal West, Extended Duration Substrate, Thermoscientific myECL).

# Study of C3b Cleavage

To study the contribution of TM to the cleavage of C3b by Factor I (FI) in the presence of Factor H (FH) as a cofactor, purified C3b (Calbiochem), diluted in Tris buffer (10 mM, NaCl 40 mM, pH7.4) at 20µg/ml (300 µl), was incubated with human albumin (CSL Behring, 10µg/ml), TM (R&D Systems, 10µg/ml, 10 µl), or without protein for 10 min on ice. FH (20µg/ml, 10 µl) was added on ice for 10 min. Each sample was separated into 4 tubes, and 4 µl of FI (10µg/ml) was added per tube. DTT-Blue Reducing sample buffer (13 µL) was added at t = 0, 30 s, 2 min or 10 min to stop cleavage of C3b. Samples were denatured, resolved on a 10% gel and transferred onto a nitrocellulose membrane. The antibodies used for visualization by WB were an anti-C3 polyclonal goat antiserum (Merck Millipore, 204869, 1/5000) and a goat anti-rabbit antibody IgG-HRP (Santa Cruz, sc-2004, 1/5000).

# Patients' Kidney Biopsies

Biopsies from two aHUS patients - P1 carrying C3 p.R161W [C3 gain of function (9)] and P2, having homozygous FH mutation p.C564F (FH deficiency)—were retrieved from the archive of the Pathology Institute of CHU Lille, France. P1 had a hemolytic anemia and histological analysis described typical lesions of glomerular and arteriolar TMA. Hemolysis was corrected at the moment of P2 biopsy, which reported predominant chronic TMA lesions, characterized by vessel wall thickening and thrombotic occlusion of many arterioles. Perls' Prussian blue staining identified hemosiderin deposits. Immunohistochemical analysis for HO-1 (rabbit anti-HO-1, Abcam) was performed on deparaffinized slides using Ventana XT autostainer (Ventana Medical systems). Immunofluorescence with rabbit anti-C3c (Dako) was performed on frozen slides. Secondary antibodies were coupled to Fluorescein IsoThioCyanate—FITC (Sigma Aldrich). A normal protocol, kidney allograft biopsy performed 3 months after transplant was used as a negative control and biopsies of two patients with chronic hemolysis (hemolysis associated with prosthetic heart valve) were used as positive controls for hemosiderin detection and tubular staining of HO-1 (20). Whole slides were scanned or analyzed by an Olympus microscope (Life Sciences Solutions) or Nanozoomer (Hamamatsu).

# Statistics

Statistical analysis was performed with R software (21). The ggplot2 package (22) was used for graphical representations. Data are presented as medians and interquartile ranges. After a Kruskal-Wallis test, a Dunn's test was used for multiple pairwise comparisons. A p-value <0.05 was considered significant.

# RESULTS

# Heme-Induced C3 Fragment Deposition With a Particular Renal Tropism in vivo

We performed a comparative analysis of the deposition of C3 activation fragments in mouse organs under control conditions (injection of PBS) and after administration of heme using IF (**Figure 1**) and IHC (**Supplementary Figure 1**). No structural differences were detected in the kidneys or livers of hemeinjected mice (Periodic Acid Schiff (PAS) and hematoxylin, eosin, saffron (HES) staining), in agreement with previous observations (13).

IHC staining for C3 activation fragments revealed minimal/absent staining in the three tested organs: kidney, heart and liver of PBS-injected mice. In heme-injected mice,

significant deposits were observed in kidney glomeruli, while only minimal effects were observed in heart or liver using this technique (**Supplementary Figures 1A–C**). We also studied the C3 fragment deposition using the more sensitive IF method. C3 staining was detected in kidneys of PBS-injected mice, located along the tubular basement membrane and in the Bowman's capsule (**Figure 1A**); it was minimal/absent in heart, liver, brain, skin and lungs of these mice (**Figures 1B–F**). In the presence of heme, strong C3b/iC3b staining was detected in kidneys (**Figure 1A**, **Supplementary Figure 1D**) contrary to heart and brain (**Figures 1B,D**). To a lesser extent, deposits were also revealed in liver, skin and lungs (**Figures 1C,E,F**). In the heme-exposed kidneys themselves, C3 fragment deposits were observed in vessels, intra-glomeruli, along the tubular basement membrane and in the Bowman's capsule. C3 deposits were detected on the glomerular EC, as seen by double staining for C3b/iC3b and vWF (**Supplementary Figure 1D**).

# Long-Term Exposure to Heme Rendered Glomerular Endothelial Cells Susceptible to Complement Activation in vitro

We then tested whether the particular susceptibility of the glomerular endothelium to C3 activation fragment deposition could be reproduced in vitro, using cultured human EC from different vascular beds (**Figure 2A**). We compared human primary (HRGEC) and conditionally immortalized (GENC) glomerular cells with HUVEC (as a macrovascular EC model) and human microvascular EC line (HMEC), for their capacity to activate complement after exposure to heme. In accordance with the in vivo data, C3 deposits were significantly higher on glomerular EC (GENC and HRGEC) than on HUVEC and HMEC (p < 0.05). Overnight exposure to 50µM heme increased C3 deposition on average ∼2.6 fold on HUVEC, ∼4 fold on HMEC, ∼7.5 fold on GENC and ∼6 fold on HRGEC compared with the baseline. No significant difference was observed in terms of C3 fragment deposition at resting state on the tested EC (**Figures 2B,C**).

Others TMA-related stimuli were tested, such as exposure to pro-inflammatory cytokines TNFα and IFNγ (deemed pertinent as the primary trigger for aHUS could be an infection) or brief exposure (30 min) to heme [since hemolysis has been proposed as a secondary hit for aHUS (8)]. In both cases HUVEC and GENC showed a similar response profile, with increased C3 fragment deposition from NHS, but without significant difference between the cell types in either the resting state or in the 30 min hemeexposed cells, while GENC had lower levels of C3 deposition compared to HUVEC after activation by the pro-inflammatory cytokines (p < 0.0001) (**Figure 2D**).

# The Increase of C3 Deposition on Glomerular EC Correlates With a Weaker FH Binding

Most cells adapt over the long term to resist to heme overload. Therefore, we compared the binding/expression of complement regulators after overnight exposure to heme (**Figure 3**). We detected a heme-concentration-dependent increase of FH binding from NHS, significantly higher on HUVEC, compared with the glomerular EC (p < 0.05) (**Figures 3A,B**). The expression of MCP and DAF did not differ among the four tested EC types at resting state (**Supplementary Figures 2A,B**), while under hemolytic conditions MCP decreased to a similar level for the four cell types and DAF displayed a heterogeneous behavior (**Figures 3C**,**D**).

### TM Is Inefficiently Upregulated on Heme-Exposed Microvascular EC in vivo and in vitro

The susceptibility of microvasculature to TMA could be related not only to lower resistance to complement but also to inefficient thromboresistance or regulation of the coagulation cascade. TM is a key factor in reducing blood coagulation, but may serve also as complement regulator (23). Here, TM accelerated the cleavage of C3b to iC3b by FI, leading to the appearance of the α43 fragment at 30 s with a time-dependent increase; this band was invisible in the presence of albumin (irrelevant protein) or buffer at 30 s, appearing later (2 min) with a time-dependent increase (**Figure 4A**).

In vitro, the expression of TM did not differ among the four tested EC types at resting state (**Supplementary Figure 3**). After prolonged exposure to heme, TM gene expression (**Figure 4B**) increased in HUVEC (gene expression ∼6.1 fold at 50µM heme, compared to baseline), while it was significantly lower in HMEC, GENC, and HRGEC (p < 0.05) compared with HUVEC. The TM level was also evaluated in different organs of heme-injected mice (**Figure 4C**). No difference was detected in glomeruli (strong basal staining) nor in heart or brain of heme- vs. PBS-injected mice. In contrast, a tendency toward an increase of TM was detected in skin, and a significant 18-fold increase of TM staining was found in the large vessels of the livers from heme-injected mice compared with PBS controls (p < 0.05). Surprisingly, an over 30-fold decrease of TM was seen in the lungs' microvasculature of heme-injected mice compared with controls (p < 0.0001) (**Figure 4C**).

#### Heme Oxygenase-1 (HO-1) Is Inefficiently Upregulated on Microvascular and Particularly on Glomerular Endothelium by Heme in vitro and in vivo

The susceptibility of the kidney to TMA lesions could be also related to a reduced resistance to oxidative stress and inefficient heme-degradation. HO-1 is the major heme-degrading enzyme and its expression is inducible after cell stress (15).

In mice, at basal condition (PBS), HO-1 staining was negative in the kidney (**Figure 5A**) as well as in others studied organs (**Figures 5B–F**). After heme injection, HO-1 staining was intense in the proximal tubules, but remained negative in glomeruli (**Figure 5A**, **Supplementary Figures 4A,B**). The same results were observed by IHC. Only minimal HO-1 staining was detected in some cases in heme-exposed glomeruli, corresponding to the glomerular epithelial cells and not endothelial cells in agreement with previous reports for human glomeruli (24); HO-1 expression was not detected in the brains of the heme-treated mice either. HO-1 labeling was positive in heart, liver, skin and lungs of heme-treated animals (**Figures 5B–F**), and visual colocalization with blood vessels was observed. In the skin, the HO-1 expression was particularly strong in the vessels of the hair follicles.

In vitro, heme-treatment induced ∼92.3 fold increased HO-1 gene expression on HUVEC, compared to baseline (**Figure 6C**).This increase was weaker in the other cell types: ∼5.9 fold on HMEC (p < 0.05), ∼26.6 fold on GENC (p < 0.05) and ∼17 fold on HRGEC (p < 0.05). HO-1 protein expression level in HUVEC, and to a lesser extent HMEC, increased with increasing heme concentration (**Figures 6A,B**—note that HMEC are derived from cells isolated from human foreskins and lack hair follicles). Meanwhile, HO-1 expression on both glomerular EC types was only weakly detected.

independent experiments. (C) Quantification of C3b/iC3b fragments deposition on the four EC types, studied as a function of the dose of heme overnight (n = 3–5). (D) Quantification of C3b/iC3b fragments deposition on HUVEC and GENC pretreated with medium only, TNFα/IFNγ overnight, or 50µM of heme (30 min) (n = 3–5).

#### HO-1 Is Inefficiently Upregulated in Patients' Glomerular Endothelium

In kidney biopsies of patients with hemolysis—aHUS or with prosthetic cardiac valve—IHC with anti-HO-1 revealed a tubular staining proportional to the degree of hemolysis observed by hemosiderin deposition by Perls' Prussian blue staining (**Figure 6D**). Glomerular endothelium HO-1 staining was negative (**Figure 6E**). No hemosiderin deposition, HO-1 mesangial or tubular stainings were detected in negative controls. It should be noted that all biopsies, even negative controls, displayed non-specific staining due to capture of the antibody by erythrocytes. Positive staining was nevertheless detected on podocytes, in agreement with previous observations (25). Similar to the mouse results, weak positive C3 staining was detected in some glomeruli of cardiac valve patients, while intense deposits were observed in the patients with complement abnormalities (**Figure 6F**).

FIGURE 4 | Influence of heme on the TM expression in vitro on cultured human EC and in mice. (A) Functional activity of TM as complement regulator, studied using purified protein by western blot. C3b is composed of 2 chains, α and β, visible by WB as 2 bands. The α chain is cleaved by FI to give two new bands: α67 and α43. The intensity of these bands reflects the level of C3b conversion to iC3b. Incubation of C3b with FH and FI led to cleavage of C3b to iC3b in a time-dependent manner. Addition of TM accelerates the cleavage of C3b compared to albumin and buffer controls. (B,C) HUVEC, HMEC, GENC, and HRGEC were incubated with increasing concentrations of heme overnight and mRNA was extracted. Gene expression of (B) TM was measured by RTqPCR, (n = 3). (C) TM staining (in red) was studied by IF in frozen kidney (x15), heart (x15), liver (x8), brain (x8), skin (x15), and lung (x5) sections of mice treated with PBS (negative control) or heme. Representative images of 5 mice per group. Lower panels: quantification of TM staining in organs sections of PBS (white) and heme-injected mice (gray).

#### Glomerular EC Fail to Accommodate a Second Challenge With Heme and Show Enhanced C3 Deposition

The HO-1 up-regulation is accompanied by accommodation and a gain in complement resistance in HUVEC (16). We here compared the C3 deposition on the four EC types after incubation overnight with different doses of heme (to induce HO-1 upregulation), followed by a second challenge with 50µM of heme for 30 min and exposure to serum (**Figure 7A**). The 30 min heme-exposure caused about a 2-fold increase of the C3 deposition on EC not receiving heme overnight (not shown). Exposure to heme overnight followed by 50µM heme for 30 min produced a dose-dependent decrease in C3 deposition on HUVEC, but an increase on glomerular EC (**Figure 7B**). C3 deposits decreased by ∼50% on HUVEC, while they significantly increased by ∼8 fold on GENC and ∼3 fold on HRGEC (p < 0.05) compared with HUVEC. HMEC showed an intermediate phenotype with C3 deposition comparable to baseline.

blot (n = 2–5). (B) Quantification of the western blots (n = 2–5), ratio of the band of HO-1, relative to the band of actin. (C) HO-1 gene expression measured by RTqPCR (n = 3). (D) Hemolysis level on human kidney biopsies was evaluated by the hemosiderin deposits, revealed by Perls' coloration. (E) HO-1 (brown) and (F) C3 (green) staining were performed by IHC and IF, respectively. A normal protocol, kidney allograft biopsy performed 3 months after transplant was used as a negative control. Biopsies of a patient with chronic hemolysis (hemolysis associated with prosthetic heart valve) were used as a positive control. Two patients with aHUS carrying complement mutations were tested.

To find out to what extent this protective effect in HUVEC was HO-1 dependent, the HO-1 expression was induced by pre-incubation with another porphyrin of similar structure to heme (SnMPIX) but which blocks its activity (26–28). Overnight exposure of the cells to SnMPIX caused HO-1 expression, albeit 2–3 fold weaker than that induced by exposure to heme, and no increase of C3 deposition (data not shown). After a second challenge with heme, no protective effect against complement of the pre-incubation with SnMPIX was detected at any of the tested doses (up to 50µM), contrary to heme which produced a protective effect after cells were exposed at 12.5µM heme (p < 0.05) (**Figure 7C**).

#### TM and HO-1 Expression May Be Parallel Phenomena, Dependent on the Transcription Activity of KLF2 and KLF4

HO-1 expression is mainly dependent of the transcription factors NRF2 and BACH1, but no major differences were found for their expression between the four EC types (**Figures 7D,E**), except for a slightly lower expression of NRF2 at resting state for the glomerular HRGEC (p < 0.05).

We then studied transcription factors KLF2 and KLF4 because they control the expression of TM (29), and KLF4 could also modulate HO-1 expression (30). For KLF2, a strong dosedependent effect was noted in HUVEC with ∼9.8-fold increase compared with untreated cells after exposure to 50µM of heme (**Figure 7F**). The increase in the other cell types was lower compared to HUVEC, with only ∼2, ∼3.3, and ∼3.8-fold after exposure to 50µM heme for HMEC, GENC, and HRGEC, respectively. Strong KLF4 gene expression was also observed with increases in heme concentration (**Figure 7G**). Exposure to 50µM heme induced ∼25-fold increase of KLF4 level in HUVEC, but only ∼1.26, ∼1.1, and ∼1.48-fold in the microvascular HMEC, GENC, and HRGEC, respectively. These results may be explained, at least in part, by the shared profile of TM and HO-1 expressions in microvascular EC.

# DISCUSSION

In this study, we discovered a dichotomy in the phenotypic adaptation to hemolysis of macro- vs. microvascular endothelium. Macrovascular EC adapted to heme-overload by up-regulating the heme-degrading HO-1 and the coagulation and complement regulator TM, as well as exhibiting enhanced binding of the complement regulator FH. On the contrary, the microvascular EC, and in particular glomerular EC, were less efficient at each of these processes (**Figure 8A**), which resulted in a susceptibility of glomerular endothelium to hemedriven complement overactivation. These observations may help to explain the microvascular and renal tropism of the complement-mediated TMA lesions of aHUS (**Figure 8B**).

The tropism of TMA lesions for small vessels is a defining feature of this group of syndromes, but the mechanisms remain poorly described. Our results and data from the literature suggest that resting microvascular EC express similar or even higher levels of complement regulators and TM (6, 31, 32), failing thus to explain their susceptibility to TMA lesions. It is well-established that the presence of shiga toxin in typical HUS, complement abnormalities in aHUS, or ADAMTS13 deficiency in thrombotic thrombocytopenic purpura (TTP) are necessary but not sufficient in themselves to trigger TMA (1, 33). Inflammatory insult is a frequently described TMA-trigger. Nevertheless, our results showed that inflammatory mediators increased complement activation and decreased TM expression in both micro- and macrovasular EC, in accordance with previous studies (6, 33– 35). Secondary hits are needed to overcome the tolerable EC stress and to promote overt endothelial injury and disease manifestation. Since hemolysis is a hallmark of TMA, it is tempting to speculate that oxidative stress and cell activation will be particularly noxious for the microvessels, in the kidney as well as other organs, amplifying cell damage, complement activation and thrombosis, contrary to macrovessels.

Hemolysis alone also does not induce renal TMA lesions (13) but may present a secondary challenge in the presence of deregulated complement, as in the case of aHUS (7, 8); indeed, heme activates complement AP directly in serum and this contributes to the C3 deposits found on EC (8, 11, 36–39). In support of this hypothesis, injection of heme resulted in a marked renal C3 deposition in mice, localized predominantly in the glomeruli, and similar to what has been observed for intravascular hemolysis and injection of cell-free hemoglobin (39). In contrast to kidneys, our examination of the heart, liver, brain, skin and lungs showed them to be largely unaffected by injection of heme, or to have exhibited only small increases in C3 deposits after injection (summary in the **Supplementary Table 1**). In agreement with these in vivo data, overnight exposure to heme induced stronger C3 deposition on glomerular EC compared with the macrovascular (HUVEC) and microvascular (HMEC) cell models. Together, these in vivo and in vitro observations point toward a vulnerability of the glomerular endothelium to complement deposition in the presence of heme. However, the inefficient adaptation to hemolysis in terms of HO-1 expression and TM up-regulation was also observed in other microvascular EC, both in vitro and in vivo, while only glomerular EC were subject to deposition of very high levels of C3 activation fragments upon exposure to heme.

These particularly elevated complement deposits in kidneys and glomerular cells may be explained by reduced binding of FH, the main regulator of the AP of complement (3), which is the most frequently affected among aHUS patients [genetic or acquired abnormalities found in >40% of patients (2, 40, 41)]. Moreover, we confirmed that FH is assisted by TM in the inactivation of C3b (23, 42, 43) and that the lower level of TM can further aggravate the inefficient complement regulation. The underlying mechanism(s) behind this reduced FH binding require further investigation, but the combination of decreased FH binding to glomerular EC under hemolytic conditions with FH mutations or autoantibodies could contribute to the susceptibility of glomerular EC to TMA lesions. Interestingly, a recent study also showed that renin, an enzyme produced specifically in the kidney, could cleave C3 and exacerbate the complement activation (44). The hemolysismediated complement activation could be an additional key

Frontiers in Immunology | www.frontiersin.org

(Continued)

FIGURE 8 | explain the microvascular/glomerular cell susceptibility of TMA lesions. Our results suggest that hemolysis is a key factor in the vulnerability of the glomerular EC complement overactivation and TMA. The glomerular EC have a weaker capacity to adapt to hemolysis and to up-regulate cytoprotective and stress-response genes, such as HO-1 and TM, compared to EC from other organs. This could be related to the lower levels of the transcription factors KLF2 and KLF4. The inefficient degradation of heme due to the lack of HO-1 will result in stronger heme-mediated complement activation on the endothelium. The weaker expression of TM will on the one hand contribute to the pro-coagulant phenotype of the EC, and on the other hand contribute to inefficient degradation of the anaphylatoxins C3a and C5a and assist FH and FI in degradation of C3b. Finally, FH will bind less to heme-exposed glomerular EC compared to EC of other organs. This will be further exacerbated in case of complement overactivation, as in aHUS, leading to EC suffering and TMA.

element in this vicious circle, linking renin secretion, vascular damage and renal failure. Thus, the C3 deposits described in aHUS kidney biopsies, which are generally considered nonspecific (45, 46), could reflect the complement activation due to hemolysis and deregulated complement, this last condition explaining why the C3 deposits are greater in aHUS patients than in cases of isolated hemolysis (cardiac valve).

The cells' capacity to express HO-1, to degrade heme and then detoxify it to cytoprotective metabolites biliverdin and CO, plays a key role in vascular protection (15, 47). In vitro studies have shown heme-mediated HO-1 expression in HUVEC (48, 49), and there is clear evidence for endothelial HO-1 expression in large vessels of patients with sickle cell disease (50), as well as in animal models of hemolysis or other stress inducers (51–55). In contrast, we found only HO-1 upregulation to be minimal or absent in glomerular EC (in vitro or in vivo), in aHUS and hemolytic patients, as well as in heme-injected mice. This observation is in agreement with published images of aHUS and under other hemolytic conditions (20, 24, 25, 56, 57) and indicates that glomerular endothelium does not have the capacity to efficiently up-regulate HO-1 in presence of heme. HO-1 production is, however, indispensable to the glomerular protection, as suggested by reported early damages of glomeruli in case of HO-1deficiency in murine model or in human (47, 58, 59). Mesangial proliferation and thickening of the capillary wall in accordance with endothelial swelling and detachment were indeed described in a HO-1 deficient young boy. This early aspect of mesangioproliferative glomerulonephritis was confirmed from sequential kidney samples (at 2, 5 and 6 year-old), while tubulointerstitial damages advanced more progressively. This could to be related to the fact that the production of HO-1 within kidneys would be mainly tubular, while in the glomeruli the major HO-1 source would be the infiltrating macrophages, not the intrinsic glomerular cells (60). Further studies are needed to extend our knowledge on the different ways, that the EC from different vascular beds manage heme homeostasis, based on their specific ability to accumulate it, to transport it within the cell by FLVCR1, to express ferritin, store iron and to produce reactive oxygen species. Some of these parameters have been studied for HUVEC and HMEC, especially in the context of deficiency of the heme transporter FLVCR1a, but data on glomerular EC are lacking (61). Moreover, a recent study in the context of leukemia demonstrated that C3a and C5a trigger phosphorylation of MAPK, followed by downregulation of HO-1 expression in malignant cells (62). If such phenomenon operates in HO-1 expressing endothelium, complement activation by intravascular hemolysis and heme release may weaken the endothelial resistance. In such context, inhibitors of MAPK (such as SB203580) will result in upregulation of HO-1 and enhance resistance to hemolysis-derived products as well as to complement.

A protective role of HO-1 against immune complexesmediated complement overactivation has also been described on HUVEC (16). Here, our results demonstrated a complementprotective effect of pre-incubation with heme in HUVEC. Ideally, this phenomenon should be confirmed after HO-1 silencing. In contrast, the glomerular EC showed a marked increase in C3 deposition. These results, together with the loss of this protective effect upon HO-1 inhibition in HUVEC, suggest that HO-1 activity for heme degradation contributes to protection against complement activation in this cell type—a phenomenon which is not operational in the glomerular endothelium. It is important to note that glomerular EC were able to upregulate HO-1 after incubation with a very potent inducer, such as CoPPIX (data not shown), suggesting that the machinery needed for its synthesis is present and can be triggered after potent stimulation. Nevertheless, induction of massive hemolysis in vivo was not strong enough to mediate HO-1 staining in glomerular EC, contrary to tubuli and podocytes [(13) and data not shown]. Our in vivo experiments also showed that the brain endothelium failed to upregulate HO-1 after heme injection, but this was not associated with concomitant deposition of C3 fragments. Sartain et al. recently demonstrated that brain microvascular EC express higher levels of complement regulators, compared to glomerular EC, and had a better capacity for suppressing the alternative pathway in vitro (32). These findings may help to explain the occurrence of lower levels of cerebral vessels' complement deposits and the fact that in aHUS, as well as in shiga toxin HUS, brain manifestations were found in only a small fraction of patients (63). On the other hand, the inefficient upregulation of HO-1 in the brain vasculature may contribute to the cerebral manifestations of TTP (also a TMA disease). The possible role of HO-1 expression in TTP pathophysiology requires further studies.

Interestingly, KLF2 (64, 65), and KLF4 (66), which are known transcription factors of TM, have also been shown to serve as modulators of HO-1 expression (30). Therefore, the weaker upregulation of TM and HO-1 we observed in microvascular EC could at least in part be related to differences in transcription regulation. Consistent with these data, other workers have reported that KLF2 and KLF4 were not significantly upregulated in glomeruli of aHUS patients compared with controls (67).

The observed differences in gene expression and phenotype among the tested EC here confirm the utility of HUVEC and HMEC as model EC types for macro- and microvascular endothelial, but highlight unique features and responses to activation in glomerular EC, which could perhaps be better modeled by glomerular (primary or cell line) EC in culture. A limitation of our work is that, although we observed correlations, we do not provide direct evidence linking HO-1 and complement deposits in vitro, for which the knock out/knock in strategy would have been useful. Also, providing in vivo evidence that up-regulation of HO-1 could prevent complement activation and TMA lesions in a mouse model of aHUS was outside the scope of this study.

In conclusion, we have shown that when compared with macrovascular EC, microvascular EC, and glomerular EC in particular, are vulnerable to complement-mediated TMA at least in part because of their failure to adapt to hemolysis, to up-regulate cytoprotective and stress-response genes such as HO-1 and TM, and to bind FH (**Figure 8B**). We hypothesize that local, subclinical microthrombosis due to a primary triggering event (infection, pregnancy) will cause mechanical hemolysis in the kidney. By amplifying complement activation and pro-thrombogenic traits, heme will exacerbate endothelial activation. Once the threshold of tolerance is reached, this heme-induced endothelial and complement overactivation could sustain and perpetuate the TMA lesions. Together, our results indicate that the vulnerability of the glomerular EC to hemolysis is a key factor, predisposing them to complement overactivation and TMA, as seen in aHUS. The heme scavenger protein hemopexin has been efficient in preventing hemolysis-mediated C3 deposition in kidneys in a mouse model of intravascular hemolysis and on endothelial cells in vitro (39), as well as in models of stasis in sickle cell disease (68) or other hemolytic conditions (69). Therefore, heme-blocking agents may be explored as novel therapeutic strategies to prevent microvascular injury in TMA diseases.

# REFERENCES


# AUTHOR CONTRIBUTIONS

LR, MF, and OM designed the study. OM, NM, AG, CP, TR-R, RP, VG, and FD performed research. SS and PM provided the GENC cell line. MH performed statistical analyses. LR, MF, OM, VF-B, JD, NM, AG, VG, and EB discussed the data. All authors wrote the manuscript and approved the submission.

#### ACKNOWLEDGMENTS

This work was supported by grants from Agence Nationale de la Recherche ANR JCJC—INFLACOMP 2015-2018 ANR-15-CE15-0001 to LR, ANR JCJC—COBIG ANR-13-JSV1- 0006 to JD, Fondation du Rein under the aegis of the French Medical Research Foundation AMGEN 2014 FdR-SdN /FRM\_FRIMAT and INSERM. The cytometric and microscopy analysis were performed at the Centre d'Histologie, d'Imagerie et de Cytométrie, (CHIC), Centre de Recherche des Cordeliers UMRS1138, (Paris, France) and we are grateful to the CHIC team for the excellent technical assistance. CHIC is a member of the UPMC Flow Cytometry network (RECYF). We are grateful for excellent technical assistance of the CEF team of the Centre de Recherche des Cordeliers for their support with the animal experimentation. The authors thank Dr. Michael HOWSAM (English proof-reader) for editorial assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.03008/full#supplementary-material

hit for atypical hemolytic uremic syndrome. Blood (2013) 122:282–92. doi: 10.1182/blood-2013-03-489245


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 May, Merle, Grunenwald, Gnemmi, Leon, Payet, Robe-Rybkine, Paule, Delguste, Satchell, Mathieson, Hazzan, Boulanger, Dimitrov, Fremeaux-Bacchi, Frimat and Roumenina. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Lectin Pathway of Complement in Myocardial ischemia/Reperfusion injury—Review of its Significance and the Potential impact of Therapeutic interference by C1 esterase inhibitor

#### *Anneza Panagiotou1 , Marten Trendelenburg1,2 and Michael Osthoff 1,2\**

*1Division of Internal Medicine, University Hospital Basel, Basel, Switzerland, 2Department of Biomedicine, University Hospital Basel, University of Basel, Basel, Switzerland*

#### *Edited by:*

*Maciej Cedzynski, Institute for Medical Biology (PAN), Poland*

#### *Reviewed by:*

*Peter Garred, University of Copenhagen, Denmark Robert Rieben, Universität Bern, Switzerland*

> *\*Correspondence: Michael Osthoff michael.osthoff@usb.ch*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 19 March 2018 Accepted: 08 May 2018 Published: 25 May 2018*

#### *Citation:*

*Panagiotou A, Trendelenburg M and Osthoff M (2018) The Lectin Pathway of Complement in Myocardial Ischemia/Reperfusion Injury—Review of Its Significance and the Potential Impact of Therapeutic Interference by C1 Esterase Inhibitor. Front. Immunol. 9:1151. doi: 10.3389/fimmu.2018.01151*

Acute myocardial infarction (AMI) remains a leading cause of morbidity and mortality in modern medicine. Early reperfusion accomplished by primary percutaneous coronary intervention is pivotal for reducing myocardial damage in ST elevation AMI. However, restoration of coronary blood flow may paradoxically trigger cardiomyocyte death secondary to a reperfusion-induced inflammatory process, which may account for a significant proportion of the final infarct size. Unfortunately, recent human trials targeting myocardial ischemia/reperfusion (I/R) injury have yielded disappointing results. In experimental models of myocardial I/R injury, the complement system, and in particular the lectin pathway, have been identified as major contributors. In line with this, C1 esterase inhibitor (C1INH), the natural inhibitor of the lectin pathway, was shown to significantly ameliorate myocardial I/R injury. However, the hypothesis of a considerable augmentation of myocardial I/R injury by activation of the lectin pathway has not yet been confirmed in humans, which questions the efficacy of a therapeutic strategy solely aimed at the inhibition of the lectin pathway after human AMI. Thus, as C1INH is a multiple-action inhibitor targeting several pathways and mediators simultaneously in addition to the lectin pathway, such as the contact and coagulation system and tissue leukocyte infiltration, this may be considered as being advantageous over exclusive inhibition of the lectin pathway. In this review, we summarize current concepts and evidence addressing the role of the lectin pathway as a potent mediator/modulator of myocardial I/R injury in animal models and in patients. In addition, we focus on the evidence and the potential advantages of using the natural inhibitor of the lectin pathway, C1INH, as a future therapeutic approach in AMI given its ability to interfere with several plasmatic cascades. Ameliorating myocardial I/R injury by targeting the complement system and other plasmatic cascades remains a valid option for future therapeutic interventions.

Keywords: C1 esterase inhibitor, complement system, complement inhibition, ischemia/reperfusion injury, mannose-binding lectin, myocardial infarction, inflammation

# INTRODUCTION

Ischemic heart disease is still a major cause of morbidity and mortality worldwide. In the United States, more than 700,000 episodes of acute myocardial infarctions (AMIs) are diagnosed annually (1). AMI is the consequence of rupture or erosion of a vulnerable atherosclerotic plaque in the coronary arteries subsequently leading to total or partial occlusion and tissue ischemia. In patients with total occlusion, emergency reperfusion of ischemic myocardial tissue is the cornerstone of therapy to salvage ischemic tissue from permanent damage. However, abrupt restoration of coronary blood flow with reperfusion of ischemic myocardium may itself trigger additional injury, which is known as ischemia/reperfusion (I/R) injury and may lead to the death of previously viable cardiac tissue (2). Estimates from previous experimental studies suggest that I/R injury may account for a significant (up to 50%) proportion of the final infarct size (3). Several mechanisms and mediators of I/R injury have been previously identified including oxidative stress, inflammation, and endogenous salvage kinase pathways (4). Despite promising results in experimental and early phase human studies (5, 6), interventional strategies targeting cardiac I/R injury have not been successful including the phase 3 trials of cyclosporine or pexelizumab before percutaneous coronary intervention in patients with AMI (7, 8).

Regarding inflammation as one mediator of I/R injury, experimental and clinical studies have shown that reperfusion after transient ischemia results in activation of endothelial cells, the contact and the complement system and attraction of neutrophils to the site of infarction (9, 10). The complement system is a major component of innate immunity, which is involved in both recognition and response to pathogens (11). It is further implicated in an increasing number of homeostatic and disease processes such as the immune complex catabolism, the clearance of dead and dying cells and the modulation of adaptive immune responses (12). Three pathways can activate the complement cascade: the classical, the alternative, and the lectin pathway. After initiation, these three pathways converge at the level of cleavage and activation of complement component C3, which subsequently leads to the generation of the anaphylatoxins (C3a, C5a) and the membrane attack complex (MAC; C5b-9). This review focuses on the role of the lectin pathway in myocardial I/R injury and the potential benefit of therapeutic application of its natural inhibitor, C1 esterase inhibitor (C1INH). In particular, we will focus on the potential advantages of C1INH as a promising candidate for future trials of salvage strategies of hypoxic myocardial tissue after AMI.

# THE LECTIN PATHWAY OF COMPLEMENT

The lectin pathway can be activated by the pattern-recognition receptors (PRR), mannose-binding lectin (MBL), ficolin-1, ficolin-2, ficolin-3, collectin 10 (CL-10), and collectin 11 (CL-11 or CL-K1). These glycoproteins bind to carbohydrate patterns, acetyl groups, or immunoglobulin M exposed on microorganisms or dying host cells (13, 14). In plasma, lectin pathway PRR complex together with MBL-associated serine proteases (MASPs)-1, -2, -3, and two non-protease peptides, sMAP (Map19) and MBL/ficolinassociated protein-1 (MAP-1 or Map44). Whereas MAP-1 was shown to be a natural inhibitor of the lectin pathway (15), the exact function of sMAP is still unknown. After binding of the PRR to their target structures, MASP-1 is activated which in turn activates MASP-2, which is required for generating the C3 convertase (C4b2a) (16). The C3 convertase cleaves C3 into the opsonin C3b and the anaphylatoxin C3a, initiating the formation of an additional anaphylatoxin (C5a) and the MAC (C5b-9). Whereas the function of MASP-2 is strictly limited to the activation of the lectin pathway, several proteolytic functions have been observed for MASP-1, such as cleavage of fibrinogen, factor XIII, prothrombin and thrombin-activated fibrinolysis inhibitor (coagulation cascade), cleavage of kininogen (kallikrein-kinin cascade), and activation of protease-activated receptors on endothelial cells (neutrophil attraction) (17). Although the proteolytic activity of MASP-1 toward these proteins is much lower compared with the primary cleaving or activating enzymes of these proteins or compared with the primary lectin and classical pathway target proteins of MASP-1, it may still be of relevance *in vivo*. For example, a prolonged bleeding time and decreased arterial thrombogenesis was observed in a MASP-1 knock-out rodent model (18, 19). In the context of AMI, activation of the lectin pathway of complement after AMI may impact on I/R injury not only through activation of the complement cascade but also *via* promotion of clot formation (coagulation and fibrinolytic system) and of inflammation (kallikrein-kinin cascade, activation of endothelial cells, and attraction of neutrophils). Hence, inhibition of the lectin pathway, particularly at the level of MASP-1/-2 seems to be advantageous over downstream inhibition of the complement.

# C1 ESTERASE INHIBITOR

The most important natural inhibitor of the lectin pathway of complement is C1INH. C1INH, a member of the serpin superfamily of serine protease inhibitors, is an acute-phase protein that has manifold targets and biological functions. Although the primary function of serpins involves the inhibition of proteases, they are also implicated in additional biological interactions, such as the inhibition of leukocyte rolling and interactions with endothelial cells and microorganisms (20). For example, treatment with C1INH was shown to limit the activation of endothelial cells and their subsequent transition into a procoagulatory and antifibrinolytic state after I/R injury (21). Proteases that are inactivated by C1INH include C1r, C1s (classical pathway of complement), MASP-1 and MASP-2 (lectin pathway), factor XII and plasma kallikrein (contact system), factor XI and thrombin (coagulation system), and plasmin and tissue plasminogen activator (fibrinolytic system) (22, 23) (**Figure 1**). Binding of C1INH to any of its target proteases leads to tight complexes which are subsequently cleared from the circulation and can be summarized as suicide inhibition (9). Decreased plasmatic antigenic levels of C1INH result in uncontrolled production of vasoactive peptides, which leads to the characteristic episodes of local soft tissue swelling observed in hereditary angioedema (HAE) (24).

Regarding the complement system, MASP-1 and -2 seem to be the major target of C1INH with less effective inhibition of the classical pathway (25). Interestingly, the lectin pathway and in particular MASP-1 have been recently implicated in the

pathophysiology of HAE, which underscores the central role of C1INH in controlling the activation of the lectin pathway. C1INH deficiency seems to cause uncontrolled activation of MASP-1, which may aggravate HAE (26).

Currently, three C1INH preparations are approved for the treatment and/or prevention of HAE, two plasma-derived, pasteurized, and nanofiltered (pdC1-INH, Berinert® and Cinryze®), and one recombinant product [rhC1-INH, Conestat alfa (Ruconest®), derived from the breast milk of transgenic rabbits]. Conestat alfa shares an identical protein structure with plasma-derived C1INH, but has a different glycosylation pattern of the amino-terminal domain of the protein (containing abundant oligomannose residues), which is responsible for a markedly shorter half-life compared with plasma-derived C1INH (3 vs. 30 h) (27). In fact, the unique glycosylation pattern introduced by the production of Conestat alfa in the mammary gland of transgenic rabbits may have additional, yet undiscovered consequences. For example, artificial variation of the glycosylation pattern of pdC1INH was previously shown to selectively impact on its target proteases with little impact on C1s inhibition in contrast to its interaction with kallikrein (28). Moreover, an important regulatory function of the amino-terminal domain of C1INH has been identified preventing inhibition of MASP-1-mediated alternative pathway activation (29), which may be influenced by the glycosylation pattern of C1INH. Although comparable inhibition for most target proteases was demonstrated (including C1s, factor XIa, XIIa, and Kallikrein) (30), Conestat alfa seems to target MBL and activation of the lectin pathway more effectively compared with plasma-derived preparations (31). This may be related to the fact that oligomannose-type glycans on average account for 15% of the total amount of glycans of Conestat alfa (compared with less than 1% in pdC1INH) (32), which may expedite the targeting and subsequent inhibition of MBL-MASP-1/MASP-2 complex by Conestat alfa.

Despite the rather broad interference with several cascades and targets, major adverse events or unique toxicities have not been demonstrated in previous studies, with the exception of a potential risk of allergic reactions in patients with rabbit dander allergy receiving Conestat alfa. Previous concerns of an increased thrombotic risk of pdC1INH (33) have not been confirmed in recent trials and registry data of both, pdC1INH and rhC1INH (34, 35). Common side effects described in trials include headache, nausea, and diarrhea. Currently, C1INH is evaluated in interventional clinical trials in the context of transplantation, acute antibody-mediated rejection after transplantation, and contrast-induced nephropathy.

#### THE LECTIN PATHWAY IN EXPERIMENTAL MYOCARDIAL I/R INJURY

Many animals and a limited number of human studies support the concept, that activation of the complement system and in particular the lectin pathway contributes to tissue injury observed after reperfusion of ischemic myocardial tissue (**Table 1**). Collard et al. were the first to report lectin pathway-dependent local activation of the complement system after myocardial I/R injury (36). Co-localized deposition of MBL and C3 in rat hearts was detected following myocardial I/R but not after sham surgery or myocardial ischemia only. Shortly thereafter, the same group demonstrated in a rat model of myocardial I/R injury that selective inhibition of MBL-A decreased infarct size and limited tissue injury and C3 deposition (37). In addition, infiltration of reperfused myocardial tissue by neutrophils was attenuated. Walsh et al. were the first to demonstrate the crucial involvement


Table 1 | Experimental studies investigating the role of the lectin pathway in murine myocardial I/R.

*fB, factor B; fD, factor D; I/R, ischemia and reperfusion; LAD, left anterior descending artery; mAbs, monoclonal antibodies; MBL, mannose-binding lectin; rhMBL, recombinant human MBL; PMN, polymorphonuclear leukocytes; sIgM, secreted immunoglobin.*

of the lectin pathway when compared with the classical and alternative pathway in myocardial I/R injury (38), an important finding, which contradicted earlier studies that had implicated the classical pathway as the most important mediator of tissue injury after I/R (39). According to Walsh et al., myocardial I/R injury was strongly attenuated in MBL knock-out mice, whereas mice with an intact lectin pathway but lacking C1q, the PRR of the classical pathway, or factor D of the alternative pathway were not protected. Reconstitution with recombinant MBL abrogated the protective effect. These results were replicated in a mouse model of diabetes with two weeks of hyperglycemia followed by myocardial I/R injury (40).

As a previous study had suggested a sequence of binding of IgM to stressed endothelial cells followed by complement activation of the lectin pathway, the same group studied myocardial I/R injury in reconstitution experiments using mice lacking natural IgM and MBL (41). Interestingly, both, MBL and IgM were found to be required for increased myocardial C3 deposition, neutrophil infiltration, and loss of left ventricular function after reperfusion of ischemic myocardial tissue. Hence, binding of natural IgM to neoepitopes in ischemic tissue was suggested as a prerequisite for subsequent MBL-mediated complement activation and tissue injury.

Schwaeble et al. investigated myocardial tissue injury in MASP-2 and C4-deficient mouse models, i.e., in the absence of any residual lectin (MASP-2) and presumably of any classical/lectin pathway activity (C4), and further downstream after binding of MBL or other lectin pathway PRR to stressed and injured cardiomyocytes (42). Lack of MASP-2 but not of C4 led to significantly less I/R-induced myocardial damage, which suggests that MASP-2-mediated activation of the lectin pathway is a major requirement for the inflammatory myocardial tissue damage after I/R. Activation of the two other complement pathways does not seem to be sufficient in the absence of any lectin pathway activity. Importantly, the authors identified a MASP-2-dependent, C4-independent route of complement activation, which likely also involves MASP-1 (29). This finding highlights the importance of inhibiting the activation of the complement system as far upstream as possible. In line, treatment of mice with an anti-MASP-2 monoclonal antibody (mAb) administered 12–18 h before coronary artery occlusion and reperfusion led to significantly smaller infarct sizes (15 vs. 26% in isotype control treated animals) (43). Similarly, treatment with MAP-1, an endogenous lectin pathway inhibitor, which displaces MASP-2 from the MASP-MBL/ficolin-3 complex and inhibits MBL- and ficolin-3-dependent complement activation *in vitro*, was associated with a decreased infarct size and C3 deposition in both, wildtype (WT) and MBL-supplemented MBL knock-out mice (15).

Finally, Pavlov et al. confirmed the concept of targeting MBL to attenuate myocardial I/R injury in a humanized mouse model (44). These humanized mice produce functional human MBL while lacking murine MBL-A and -B. Inhibition of human MBL by a monoclonal anti-human MBL antibody preserved cardiac function after myocardial I/R, attenuated myocardial fibrin deposition, and prevented ferric chloride-induced arterial thrombogenesis.

Summarizing the evidence from experimental models, the lectin pathway was shown to be crucially involved in mediating myocardial I/R injury. Conceptually, myocardial I/R injury seems to be initiated by binding of natural IgM to neoepitopes on stressed or dying cardiomyocytes with subsequent binding of MBL to IgM and activation of MASP-1/-2 during reperfusion. The latter is a major requirement for subsequent complement-mediated inflammation including formation of the membrane-attack complex, induction of apoptosis, and attraction of neutrophils. The significance of other lectin pathway PRR in acute myocardial I/R injury remains to be determined.

#### THE LECTIN PATHWAY IN HUMAN MYOCARDIAL I/R INJURY

Inter-individual serum concentrations of lectin pathway PRR and proteases show a considerable degree of variability in humans (45). In particular, the distribution of MBL plasma levels is huge and ranges from undetectable to approximately 10 µg/ml secondary to well-characterized exon and promoter polymorphisms in the MBL2 gene (46). Very low (<0.1 μg/ml, resembling a knock-out setting in murine models) and low (<0.5 µg/ml) MBL levels are found in approximately 10 and 30% of the population worldwide, respectively. Several studies have investigated associations of lectin pathway protein levels, in particular MBL, with outcome after AMI to confirm the evidence from murine studies (**Table 2**).

In a pilot study of 74 patients with ST elevation myocardial infarction (STEMI) and successful reperfusion, sufficient plasma MBL levels (defined as >0.8 µg/ml) were independently associated with a significant cardiac dysfunction [defined as left ventricular ejection fraction (LVEF) <35%] (47). In line with the concept of downstream complement activation following binding of MBL to stressed endothelial cells and/or cardiomyocytes and subsequent consumption of complement components, plasmatic sC5b-9 levels were significantly lower in patients with cardiac dysfunction. In the largest analysis to date, MBL levels were investigated in 890 patients with STEMI receiving placebo in the setting of a randomized controlled trial of the C5 inhibitor pexelizumab (48). In this study, MBL deficiency was defined as a level ≤0.1 µg/ml. Interestingly, 30-day mortality was markedly lower in MBL deficient patients (0.8 vs. 5.5%), whereas creatine kinase levels, a marker of myocardial injury and infarction size, and complement activation product levels were similar in the two groups. The authors speculate that the observed reduction in mortality was mainly driven by protection from fatal arrhythmias in MBLdeficient patients, a well-appreciated consequence of reperfusion


*CAD, coronary artery disease; LVEF, left ventricular ejection fraction; MBL, mannose-binding lectin; NSTEMI, non ST-elevation myocardial infarction; RCT, randomized controlled trial; STEMI, ST elevation myocardial infarction; AMI, acute myocardial infarction.*

injury (49). However, these results have not yet been confirmed in a cohort of STEMI patients undergoing percutaneous coronary interventions with contemporary stents and antithrombotic therapy.

In a small cohort of 55 STEMI patients, Schoos et al. investigated the association of ficolin-2/-3, MBL, and MAP-1 levels with left ventricular remodeling and infarct size as assessed by cardiac magnetic resonance imaging after percutaneous coronary intervention and at 6-month follow-up (50). Ficolin-2 levels significantly increased from admission to day 4 in contrast to other lectin pathway proteins, which may indicate consumption of ficolin-2 during STEMI. Higher baseline ficolin-2 levels were associated with left ventricular dilatation after STEMI. A similar finding was observed for the combination of higher ficolin-2 and MBL levels or higher ficolin-2 and lower MAP-1 levels indicating that the overall activation of the lectin pathway (rather than higher or lower levels of a single lectin pathway protein or protease) may be the key parameter influencing left ventricular dilatation after STEMI. However, lectin pathway proteins were not associated with infarct size, left ventricular function, or remodeling after 6 months, similar to the lack of association between MBL and infarct size observed in the larger study of Trendelenburg et al. (48). Zhang et al. investigated the role of MASP-2 in myocardial ischemia in the setting of AMI and separately in open heart surgery (51). For the purposes of this article, we only report the results of 29 AMI patients being described in this study. MASP-2 levels determined within two days after admission were almost 50% lower compared with healthy individuals or patients with stable coronary artery disease (CAD). This may be consistent with activation of the lectin pathway during AMI with subsequent consumption of MASP-2, similar to the previously cited observation of reduced ficolin-2 levels on admission. Although follow-up samples were lacking, it seems unlikely that genetically determined lower MASP-2 levels were already present before the AMI or that a temporary change in MASP-2 levels triggered the AMI in these patients.

In another study, lower MASP-2 and higher MASP-1 levels were observed in patients with myocardial infarction compared with patients with stable CAD and healthy controls (only MASP-1) (52), whereas MAP-1 levels were similar in these groups. However, the sample size was limited (*n* = 49 AMI patients) and protein levels were not associated with the severity of cardiovascular disease. Finally, levels of MAP-1, MASP-1, and -3 were analyzed in 192 AMI patients and 140 healthy controls (53). Whereas protease levels were significantly higher in AMI patients compared with healthy controls, they did not correlate with final infarct size or LVEF at 30 days. Importantly, results were in agreement with the previous study by Holt et al. only with regards to elevated MASP-1 levels but not with regards to observed MAP-1 levels. Again, follow-up samples were not available, and hence it remains to be determined if the observed changes in lectin pathway proteins are the consequence or the cause of the initial AMI event.

In summary, results from the cited studies indicate a potential involvement of the lectin pathway of complement in human myocardial I/R injury. In particular, levels of the PRR ficolin-2 and the activating protease MASP-2 were significantly lower in AMI patients on admission, whereas the levels of the endogenous inhibitor MAP-1 were higher (in only one of two studies), which may indicate activation of the lectin pathway during AMI with consumption of MASP-2 and ficolin-2. With regards to the observed increased MASP-1 concentrations in AMI patients, it remains to be determined if this is a result of an acute-phase reaction or if higher levels of MASP-1 may itself trigger an AMI event under certain circumstances given its clot promoting activity (54).

Importantly, associations of lectin pathway proteins with myocardial dysfunction, infarct size, and outcome were mostly lacking with the exception of an association of higher MBL levels and a higher net activation of the lectin pathway with mortality and ventricular dilatation after STEMI, respectively. However, involvement of MBL as a major initiator of the lectin pathway has not been demonstrated at the tissue level after human myocardial I/R injury to date. Future studies are required to investigate in more detail if PRR or protease levels are indeed associated with outcome after AMI and are causally involved in I/R injury in humans.

In general, evidence for the involvement of the lectin pathway in humans is much weaker than in animal studies, which is related to the limited samples size in most studies, a lack of analysis of a contemporary cohort of patients, and that most individual lectin pathway proteins have only been analyzed in a single study (with the exception of MBL, MASP-1, and MAP-1) and results have not yet been confirmed in subsequent studies. Although the significant variation of MBL levels in humans permits an analysis of complement-deficient patients in analogy to the analysis of knock-out compared with WT animals, this requires large cohorts. Evolutionary, other PRR may have evolved as a consequence of MBL deficiency and may compensate and in the case of I/R injury augment tissue injury independent of MBL. Unfortunately, experimental data regarding the significance of ficolins or collectins in myocardial I/R injury are lacking.

#### C1INH IN EXPERIMENTAL MYOCARDIAL I/R INJURY

The complement system was thought to play a major role in initiating an inflammatory response secondary to ischemia and reperfusion following studies 30 years ago (55). In particular, the classical pathway had initially been implicated in reperfusion damage of the heart (56, 57). To inhibit the classical pathway and subsequent complement activation at an early step, several studies examined the effects of treatment with C1INH on myocardial tissue damage, long before the central role of C1INH in the inhibition of the lectin pathway on the one hand and the pivotal role of the lectin pathway in myocardial I/R injury on the other hand was discovered (**Table 3**).

In a feline model of myocardial infarction, administration of pdC1INH before reperfusion led to a 65% reduction in cardiac tissue injury, and a markedly attenuated increase of creatine kinase compared with treatment with buffered saline solution (39). In addition, neutrophil activity/accumulation in the reperfused myocardial tissue was significantly reduced, which may be related to the interaction of C1INH with activated endothelial cells or a reduction in locally generated leukocyte chemo-attractants such as C3a and C5a. The latter was subsequently confirmed in a pig

#### Table 3 | Effect of C1INH in murine models of myocardial I/R injury.


*pdC1INH, plasma-derived C1 esterase inhibitor; iC1INH, inactive C1INH; C1s-INH-248, selective C1s inhibitor; PMN, polymorphonuclear leukocytes; cardiac CK, cardiac creatine kinase; LAD, left anterior descending artery; LAD left anterior descending artery; sCR1, soluble complement receptor 1; WT, wild-type.*

model of myocardial I/R injury (58) administering pdC1INH as an intracoronary bolus before reperfusion. Local C3a production, which markedly increased after reperfusion, was significantly attenuated in the C1INH group compared with the placebo group indicating that C1INH may indeed suppress local complement activation. In a rat model of myocardial infarction, Murohara et al. sought to investigate the differential role of the classical and alternative complement pathway in myocardial I/R (59). Rats were treated with either pdC1INH or an alternative pathway inhibitor (soluble complement receptor 1) immediately before reperfusion. Interestingly, C1INH treatment was associated with a significantly attenuated creatine kinase release and neutrophil accumulation in ischemic myocardial tissue, whereas targeted inhibition of the alternative pathway was clearly inferior in this model. Finally, Buerke et al. confirmed the cardioprotective effect of different doses of plasma-derived C1INH in a rat model (60). Release of creatine kinase and local accumulation of neutrophils was attenuated in a step-wise fashion depending on the dose of C1INH with the greatest effect observed for the highest dose (100 U/kg). In addition, expression of adhesion molecules in the affected vascular endothelium was markedly reduced after administration of C1INH, which may explain the decreased local accumulation of neutrophils (61).

As a consequence of serious thromboembolic events in 13 newborns and babies who had received 500 IU/kg of pdC1INH to prevent capillary leakage after cardiopulmonary bypass operation, Horstick et al. investigated the effect of systemic administration of different doses of plasma-derived C1INH in the same pig model as had been previously used (33). In contrast to their previous model, pdC1INH was administered intravenously 10 min before reperfusion (40, 100, or 200 IU/kg) and without concurrent heparin. Experiments with 200 IU/kg were terminated after the first three pigs had developed severe coagulation disorders. Interestingly, while 40 IU/kg of C1INH reduced infarction size to a similar degree as in the previous model (>50%), there was a lack of beneficial effect with higher doses. The severe adverse events associated with the higher dose may be related to the significant inhibition of bradykinin release and of activators of the fibrinolytic system. The latter may have been prevented by co-administration of heparin, a principal anticoagulation agent in human STEMI patients (62).

In a rat model of AMI, Fu et al. observed a reduced C3 expression and apoptosis in the affected myocardial area associated with the administration of pdC1INH (63, 64). Interestingly, the amino-terminal domain of C1INH was implicated in the antiapoptotic effect and not the protease activity.

As C1INH is a multi-action multiple-target inhibitor, it is difficult to identify the most important pathway or target protease inhibited by C1INH and responsible for the observed beneficial effect in myocardial I/R injury. While results from previous studies consistently demonstrated a clear benefit of C1INH in myocardial I/R injury, which was at least partly attributable to its inhibitory effect on complement activation, three studies have left more questions than answers with regards to the mechanism of action of C1INH and its effectiveness in cardioprotection after myocardial I/R injury. In particular, the study by Lu et al. questioned the importance of the complement system in mediating myocardial I/R injury in comparison to other mechanisms such as neutrophil influx (65). In their mouse model, a large dose of pdC1INH (400 IU/kg) or inactive C1INH (iC1INH) was administered intravenously 5 min before coronary reperfusion. iC1INH was generated by trypsin incubation, which results in the loss of its reactive center (66) and hence its ability to inhibit target proteases such as C1s. Interestingly, both C1INH and iC1INH similarly reduced myocardial damage and myocardial neutrophil influx in WT mice. In addition, C1INH treatment was effective even in C3 knock-out mice. These data are consistent with a mainly complement-independent cardioprotective mechanism of action of C1INH with the exception of its potential influence on protease-independent actions of the PRR of the lectin pathway [such as direct induction of apoptosis (67)] or its influence on lectin proteases *via* its glycosylated amino-terminal domain. The authors speculate that C1INH mainly acts *via* the inhibition of neutrophil influx across endothelial cells interacting with selectin ligands, which may be mediated by its sugar moieties rather than its proteolytic function. Two aspects warrant further comments. First, the dose of both preparations of C1INH used in the study was very high compared with previous animal models, which may be associated with additional modes of action, particularly in the case of iC1INH. Second, while the authors' dismissed a reaction of iC1INH with C1s of the classical pathway, their work does not fully exclude a residual protease activity of MASP-1/-2 after trypsin digestion. This may be of importance, as activation of MASP-1 in particular may mediate inflammatory actions independent of downstream complement activation (68). The beneficial effect of C1INH in the C3 knock-out model may be mediated by a similar mechanism, i.e., inhibition of MASP-1 and its pro-inflammatory, complement cascade-independent function. Third, previous work has identified C1INH functions that are independent of its proteolytic activity and mediated by its glycosylated, amino-terminal nonserpin domain, such as its interaction with the endotoxin lipopolysaccharide (69) or with selectins on endothelial cells (20). Similar nonserpin actions of C1INH on lectin pathway PRR and proteases have not yet been examined or described, but may contribute to the effect observed in the study by Lu et al. (65).

Similarly, Buerke et al. investigated the effect of a C1s-specific C1INH preparation (C1s-INH-248) in a rabbit model of myocardial I/R injury (70). This novel molecule is a specific inhibitor of the classical pathway of the complement system but does not inhibit MASP-1, kallikrein, and activated factors XII and XI. C1s-INH-248 was associated with attenuated myocardial injury as demonstrated by diminished plasma creatine kinase activity and local neutrophil accumulation. In addition, C1s-INH-248 was more effective than pdC1INH administered at doses up of 200 IU/kg. The authors speculated that cardioprotective effects of C1s-INH-248 may be related to the inhibition of the classical pathway and of the interaction of neutrophils with endothelial cells, and that the lectin pathway does not play a dominant role in their model. Again, the authors do not provide evidence of a total lack of interference of this C1s-specific C1INH with functions of the lectin pathway (in particular with MASP-2).

Finally, Schreiber et al. observed a lack of effect of intracoronary application of C1INH in a pig model of acute myocardial ischemia followed by urgent coronary bypass grafting (71). In contrast to all previous studies, C1INH did not reduce infarct size or release of creatine kinase, which may be explained by several ways. First, the dose of C1INH administered was rather low compared with previous studies (approximately 11–15 IU/kg). Second, C1INH was only administered locally and 60 min after induction of CPB. CPB is well known to systemically trigger activation of the complement system (72, 73), and hence late and local administration of C1INH was probably not able to prevent systemic (and even local) complement activation. In addition, evidence of an epicardial shunt was found which again may have impacted on the limited effect of locally applied C1INH. Ideally, C1INH should be administered before institution of CPB in models of coronary bypass grafting, as this would be practicable in the human setting, too.

In summary, significant cardioprotection by C1INH was evident in all but one experimental myocardial I/R injury studies. Limitations include the use of different doses, routes of administration, and AMI models with different durations of ischemia and reperfusion. However, due to its manifold actions, the exact mechanism of C1INH being responsible for the observed attenuated injury remains to be elucidated. Apart from the inhibition of complement activation, attenuated infiltration, and accumulation of neutrophils in ischemic tissue and a reduction of endothelial cell activation may be the dominant modes of action of C1INH. Unfortunately, experimental studies examining the impact of C1INH administration on the lectin pathway of complement and its consequences in myocardial I/R models are lacking. This is of importance, as a link between the lectin pathway of complement and activation of endothelial cells and subsequent recruitment of leukocytes has been previously demonstrated (16, 17, 68), and protease-independent actions of the PRR of the lectin pathway [such as direct induction of apoptosis (67) or amplification of inflammation *via* the alternative pathway] may play a role in myocardial I/R injury. A limitation of modified C1INH preparation as used in previous studies is that the difference in its function compared with original C1INH has not been comprehensively elucidated, in particular toward its effect on the lectin pathway and regarding a potential unwanted modification in the aminoterminal glycosylation.

#### C1INH IN HUMAN MYOCARDIAL I/R INJURY

Although multiple experimental studies have demonstrated the beneficial effect of complement inhibition in general and of C1INH in particular, no complement inhibitor is currently approved for the treatment of AMI or other I/R injuries. Interventions specifically and exclusively targeting the lectin pathway of complement are lacking in humans, whereas the effect of C1INH has been already investigated in three clinical trials and one case series (**Table 4**). Similar to the above-mentioned experimental studies, only pdC1INH was used. Bauernschmitt et al. were the first to report on the experience of pdC1INH treatment in three patients undergoing emergency coronary artery bypass grafting (CABG) after failed percutaneous coronary intervention (74). All three patients developed severe postoperative myocardial dysfunction which impaired weaning from CPB. Termination of CPB in all three patients was associated with the administration of a single dose of pdC1INH, which led to improved left ventricular function and hemodynamic stabilization.

In an open-label, dose-escalation study, de Zwaan et al. treated 22 patients with acute STEMI with pdC1INH during 48 h (75). The majority of patients had received antifibrinolytic therapy at least 1–2.5 h before C1INH administration, and only three patients underwent acute percutaneous coronary intervention. Drug-related adverse events were not observed, but creatine kinase and troponin T levels were reduced in comparison to a historical control population.

In a randomized, open-label study 28 patients undergoing emergency CABG after STEMI were treated with pdC1INH at aortic unclamping followed by another bolus dose 6 h later, and were compared with 29 similar patients receiving placebo and 10 patients undergoing elective CABG without evidence of recent STEMI (76). Again, drug-related adverse events were not observed. Administration of C1INH prevented the significant decline in C1INH activity (indicating consumption of C1INH) observed in placebo and control patients, whereas there was no difference in complement fragments. Interestingly, postoperative troponin T increase was only attenuated in patients receiving C1INH and surgical revascularization less than 6 h after symptom onset but not in patients with a longer ischemic interval.

#### Table 4 | Effect of C1INH in clinical trials of AMI.

In the largest study to date, Fattouch et al. randomized 80 patients with STEMI undergoing emergency CAGB to treatment with C1INH or placebo in a double-blind manner (77). C1INH was administered as an intravenous bolus before aortic unclamping followed by an intravenous infusion of 500 IU over 3 h after surgery. C1INH treatment resulted in significantly lower troponin T levels, an attenuated increase in complement activation fragments (C3a and C4a) and a shorter intensive care unit stay. Again, a significant decline of C1INH activity was only prevented in the active comparator group, and adverse events related to C1INH were not observed.

In summary, clinical studies of pdC1INH in human myocardial I/R suggest, that treatment of pdC1INH is safe and potentially effective in this setting. Of note, there is a lack of studies investigating C1INH in patients with AMI undergoing contemporary management with drug-eluting stents and stateof-the-art antithrombotic therapy. Mortality of AMI has significantly declined over the last decades as a result of modern drug and interventional treatment (78) and thus a potential benefit of pdC1INH as demonstrated in studies more than 10 years ago may not imply a similar positive effect in contemporary AMI patients. In addition, adverse drug reactions of C1INH administration must be meticulously evaluated in future studies given the potential of an increased bleeding but also thrombotic risk in the era of potent antithrombotic therapy and drug-eluting stents.

#### DISCUSSION

Acute myocardial infarction remains a leading cause of morbidity and mortality worldwide despite early and successful reperfusion strategies, which have been shown to significantly limit the size of the myocardial infarct and improve clinical outcomes. Although its existence remains controversial, reperfusion injury after restoration of blood flow has been regarded as a critical contributor to myocardial damage paradoxically limiting the beneficial effects


*CABG, coronary artery bypass grafting; CPB, cardiopulmonary bypass; ICU, intensive care unit; pdC1INH, plasma-derived C1 esterase inhibitor; STEMI, ST elevation myocardial infarction; AMI, acute myocardial infarction.*

of myocardial reperfusion. With respect to the modulation of the complement response, several complement inhibitors targeting different proteins of the complement cascade have been successfully investigated in experimental models of myocardial I/R injury (79). These models have underscored the potential of attenuating myocardial tissue damage and ventricular remodeling by complement inhibition. However, with the exception of a single randomized placebo-controlled phase 3 trial, there is a lack of high-quality clinical studies of complement inhibition in human AMI (8). In addition, single-target interventions such as inhibition of C5 are probably inadequate to address the manifold inflammatory reactions *via* several cascades and pathways after AMI. For example, pexelizumab only attenuated the increase in C5 and interleukin-6 levels but had no impact on the increase and decrease of several other pro- and anti-inflammatory proteins, respectively (80).

#### FUTURE PERSPECTIVES

Based on the data as outlined above, we would like to suggest two potential strategies of complement inhibition for future clinical trials that are not mutually exclusive.

The first strategy involves targeted inhibition of the lectin pathway of complement. As the activity of the lectin pathway essentially depends on MASP-2 as the central enzyme (81), selective inhibition of MASP-2 shortly after myocardial ischemia and before reperfusion seems like an obvious next step. However, several caveats have to be considered apart from the availability of a suitable anti-MASP-2 antibody for future clinical trials.

MASP-2 inhibition, although effective in the above mentioned animal model, may be partially bypassed by the function of at least two lectin pathway proteins, such as MBL and MASP-1. For example, targeted MAPS-2 inhibition will not impact on several pro-inflammatory functions of MASP-1 such as endothelial cell activation (82) and the promotion of clot formation (54) *via* cleavage of thrombin substrates and the activation of platelets (68). Most importantly, evidence regarding the significance of the lectin pathway in human AMI is still premature and limited. Hence, a strategy that again only targets a single protein of the complement system (although further upstream in the complement cascade as in the pexelizumab trials) may be of limited or uncertain benefit in clinical trials of AMI. The complement system has also been implicated as a mediator of regenerative processes after myocardial tissue injury (83), and hence it is imperative to study the effect of short-term complement inhibition on outcomes after at least 30 days.

Treatment with C1INH may be regarded as a potential solution to this challenge. There are several advantages of using C1INH compared with isolated MASP-2 inhibition but also some caveats. In contrast to the previously mentioned strategies of complement inhibition, C1INH is a multiple-target, multiple-action inhibitor. Myocardial I/R injury is not mediated or caused by a single protein or even pathway, rather the opposite is true, i.e., several pathways are activated simultaneously and act in parallel. In addition, the relative contribution of each involved protein and pathway to the net damage is unknown and may vary in the line with the significant heterogeneity of AMI itself and of the diverse patient populations that suffer from AMI.

Evidence from multiple experimental studies mentioned in the present review point to a relevant and protective effect of C1INH in myocardial I/R injury. Moreover, results from small clinical trials, though conducted more than 10 years ago, seem to confirm findings from animal models. Similarly important as effectiveness is the fact that adverse drug reactions of C1INH administration were not reported in the setting of human AMI (75–77) and of acute rejection following renal transplantation (84–86).

Potential disadvantages of C1INH include unwanted effects when interfering with the coagulation and the fibrinolysis system at the same time. As a matter of fact, thromboembolic complications were noted in neonates receiving high-dose C1INH during cardiopulmonary bypass surgery and in an animal model of myocardial I/R injury (33). However, there was no safety signal in the above-mentioned clinical trials of C1INH.

When designing clinical trials of C1INH in AMI several aspects and pitfalls have to be addressed such as the required duration of treatment. In the previous pexelizumab trial in STEMI patients, complement inhibition was sustained for at least 24 h with its activity having returned to baseline after 48 h. However, elevated serum complement levels have been demonstrated during the first 10 days after AMI (87).

The second question is which type of C1INH should be investigated in future clinical trials of AMI. PdC1INH seems to be the obvious choice since it has been utilized in every experimental and human myocardial I/R injury study to date. Another advantage is its significantly longer half-life compared with Conestat alfa (30 vs. 3 h) (88). However, Conestat alfa significantly decreased ischemic damage when administered up to 18 h after induction of cerebral ischemia and reperfusion, whereas pd1INH was only effective when given at the time of reperfusion (31, 89). Similarly, the formation of plasmatic functional MBL/MASP-2 complexes after transient cerebral ischemia was attenuated only in mice receiving Conestat alfa but not pdC1INH.

Another important aspect involves the requirement of inhibiting target proteases and non-protease targets at the site of acute inflammation. To maximize effectiveness and minimize adverse reactions, an ideal C1INH preparation should exert its inhibitory function preferably or exclusively in ischemic and/or reperfused myocardial tissue similar to the concept of targeting cancer cells by mAbs. In contrast to cancer, the speed of inactivation is also crucial in I/R injury, as for example MBL-MASP-1/-2 complexes clustered together on ischemic endothelial or myocardial cells may escape inactivation by C1INH long enough to activate downstream effector molecules.

Whereas the protein structure is identical, the glycosylation pattern of Conestat alfa at the amino-terminal domain is markedly different from pdC1INH preparations (32). In particular, exposed mannose residues are significantly more prevalent in Conestat alfa compared with pdC1INH (15 vs. <1%), which may influence the binding preference toward lectins and in particular to MBL. Indeed, Gesuete et al. demonstrated high-affinity binding of Conestat alfa to MBL but not of pdC1INH (31). Although this may not directly influence the degree of inhibition by its serpin domain, it may potentially impact on the speed and location of inhibition in the human body. By binding to MBL, Conestat alfa may be hijacked to the primary site of inflammation, where it may immediately inhibit its target proteases before they are able to activate downstream effector molecules. By contrast, pdC1INH is certainly able to limit random complement activation in plasma but may allow considerable escape of inactivation of complement complexes at the tissue level. Interestingly, Conestat alfa remained confined on the ischemic endothelial wall in co-localization to MBL, whereas pdC1INH was also found in the area around the ischemic vessels (31).

However, due to the lack of comparative studies of pdC1INH vs. Conestat alfa in myocardial I/R injury it is premature to draw any definitive conclusions about any difference in efficacy of the two preparations in the setting of AMI.

#### SUMMARY AND CONCLUSION

In this review, we have summarized current concepts and evidence addressing the role of the lectin pathway as a potent regulator of myocardial I/R injury in murine models and the human setting. As it still remains to be determined if administration of a complement inhibitor after myocardial ischemia and before reperfusion is effective, we may have to leave the beaten path and "should strike out on new paths" (John D. Rockefeller,

#### REFERENCES


1839–1937) avoiding single-target interventions and investigate pleiotropic compounds such as the natural inhibitor of the lectin pathway, C1INH, that also interferes with other important proinflammatory pathways. Given the evidence from several animal models and previous small clinical trials and the lack of major concerns regarding adverse events, there is ample reason to embark on larger clinical trials with C1INH in STEMI patients. In our opinion, ameliorating myocardial I/R injury by targeting the lectin pathway of complement remains a valid option for future therapeutic interventions.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct, and intellectual contribution to the work including drafting and critical revising the article and approved it for publication.

#### FUNDING

This work was financially supported by a research grant from the Fondation Machaon, Switzerland (a not for-profit private foundation) to MO.


ischemia and reperfusion injury. *Methods Find Exp Clin Pharmacol* (1995) 17(8):499–507.


double-blind study. *Eur J Cardiothorac Surg* (2007) 32(2):326–32. doi:10.1016/ j.ejcts.2007.04.038


**Conflict of Interest Statement:** MO is the principal investigators and MT is a co-investigator of an investigator initiated clinical trial (NCT02869347; https:// clinicaltrials.gov/ct2/show/NCT02869347) of Conestat alfa in the prevention of contrast-induced renal damage and have received an investigator initiated research grant from Pharming Technologies B.V., the manufacturer of Conestat alfa, for this trial. Pharming Technologies B.V. had no influence on the design and content of this review. The views expressed here are the responsibility of the authors only. AP declares no conflict of interest.

*Copyright © 2018 Panagiotou, Trendelenburg and Osthoff. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Complement After Trauma: Suturing Innate and Adaptive Immunity

#### Shinjini Chakraborty, Ebru Karasu and Markus Huber-Lang\*

Institute of Clinical and Experimental Trauma-Immunology, University Hospital of Ulm, Ulm, Germany

The overpowering effect of trauma on the immune system is undisputed. Severe trauma is characterized by systemic cytokine generation, activation and dysregulation of systemic inflammatory response complementopathy and coagulopathy, has been immensely instrumental in understanding the underlying mechanisms of the innate immune system during systemic inflammation. The compartmentalized functions of the innate and adaptive immune systems are being gradually recognized as an overlapping, interactive and dynamic system of responsive elements. Nonetheless the current knowledge of the complement cascade and its interaction with adaptive immune response mediators and cells, including T- and B-cells, is limited. In this review, we discuss what is known about the bridging effects of the complement system on the adaptive immune system and which unexplored areas could be crucial in understanding how the complement and adaptive immune systems interact following trauma.

#### Edited by:

Nicole Thielens, UMR5075 Institut de Biologie Structurale (IBS), France

#### Reviewed by:

Cordula M. Stover, University of Leicester, United Kingdom Christine Skerka, Hans Knöll Institut, Germany

\*Correspondence: Markus Huber-Lang markus.huber-lang@uniklinik-ulm.de

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 30 May 2018 Accepted: 20 August 2018 Published: 24 September 2018

#### Citation:

Chakraborty S, Karasu E and Huber-Lang M (2018) Complement After Trauma: Suturing Innate and Adaptive Immunity. Front. Immunol. 9:2050. doi: 10.3389/fimmu.2018.02050 Keywords: complement, trauma, innate immunity, adaptive immunity, T-cells, B-cells

# COMPLEMENT ACTIVATION DRIVES INNATE IMMUNITY EARLY AFTER POLYTRAUMA

Well-known ancient physician and philosopher Hippocrates from Kos (ca. 460–370 B.C.) had defined some interventions in trauma care, which had been radical at that time, eventually, some of them proving to be clinically relevant (1). Hippocratic Medicine followed the humoral pathology and defined the "four humors" or fluids of the body (among them blood), which had to be rebalanced to achieve healing. Almost as a foresight of the complement response after trauma, Hippocrates clearly pointed out the importance of humoral components in the pathophysiological course of diseases, thus being the first to indicate the involvement of the fluid phase and its potential relevance. Drawing on this age-old concept, complement activation was formerly defined as "disseminated intravascular multiple systems activation" following trauma as in a major burn injury (2). The cellular immune system with "pre-programmed responses" had been proposed to be of less importance after severe tissue injury compared to the mediator systems of complement components and other opsonins (3). Supporting this observation, activation of the humoral systems, including the coagulation and complement systems, was described to occur early after multiple injury because complement activation products, the anaphylatoxins C4a and C3a, were elevated and associated with injury severity and infectious complications (4, 5). These preliminary observations stimulated a growing interest in trauma-associated complement activation, which extended from temporal to spatial activation tendencies. Two studies are of relevance in this regard. In one clinical study, early after trauma (< 48 h) including multiple fractures, blunt abdominal trauma and blunt chest trauma with rib fractures, high C3a/C3 ratio was useful to segregate adult respiratory distress syndrome (ARDS) from non-ARDS patients (6). In a later study, plasma samples from patients were obtained as early as 30 min post-trauma comprising of penetrating injury or severe head injury and with a mean injury severity score (ISS) of 17, without any resuscitative fluid replacement therapy. Here it was found that the soluble terminal complement complex (TCC) sC5b-9 and Bb (activated factor B) were associated with high ISS (7). On comparing the two studies, several similarities and discrepancies can be summarized. Both studies identified an early activation of the alternative pathway, however, opposing in whether classical complement activation preceded or followed alternative pathway activation. Interesting to note in here is the difference of time-points in the two studies (30 min vs. <48 h post-trauma) and the fluid resuscitation that was received by patients in the former study as opposed to none received when blood samples were collected for the latter. Apart from this temporal association, the site and pattern of injury could also modulate activated complement factors and complexes differentially, as was described in a small cohort study proposing that the appearance of the TCC and C3dg as early as 24 h after trauma was mainly seen in patients who had suffered a thorax trauma (8).

When complement activation products were individually implicated in trauma-induced local and systemic trauma complications, several groups found that early C3a generation and depletion of C3 with a resultant increase in the C3a/C3 ratio were associated with the development of acute lung injury, ARDS, sepsis, multiple organ dysfunction and consequently a poor prognosis (6, 9–12). Likewise, anaphylatoxin C5a appears in the circulation of humans within 20 min post polytrauma and its levels have been related to the mortality rate (10). Exposure to C5a may result in significantly delayed neutrophil apoptosis early after trauma, thereby causing enhanced host damage through the recruitment and accumulation of neutrophils at the injury site (13). The underlying causes of rapid local and systemic complement activation via all pathways may be manifold and some remain speculative, for example, tissue hypoperfusion with acidic (micro)environments, natural antibodies, reactive oxygen species, coagulation-complement crosstalk, release of non-specific proteases with subsequent cleavage of complement components, exposure to pathogen-associated molecular patterns (PAMPs) in a synchronic shock situation, artificial surfaces like catheter materials and transfusion of blood products, including complement components in fresh frozen plasma among others (14–18). Uncontrolled complement activity can be further augmented by dysregulated levels of complement regulators, including C4b-binding protein and factor I (10). Additionally, an early depletion of central complement components causing trauma-induced complementopathy with declining complement hemolytic activity and manifestation of signs of immunodeficiency have been reported (10, 19).

While the complement system itself maybe dysregulated in the post-traumatic pathophysiology, activation of various other serine proteases, including proteases of apoptotic systems, like granzyme B and cathepsin D, may intensify the extent of complement activity. Damaged cells can also activate factor VII-activating proteases which, together with other proteases, rapidly cleave the complement factors C3 and C5 and generate the respective anaphylatoxins C3a and C5a (**Figure 1**) (16, 20). In addition to the autonomous action of the complement system after trauma, this ancient fluidphase innate immune system is intimately associated with the cellular immune responses (13, 21–23). C5a functions not only as an effective chemoattractant for neutrophils, but also upregulates various adhesion molecules on the endothelium, evoking the classical signs of inflammation. C5a exposure leads to various acute defensive functions in neutrophils, including enhancement of phagocytic activity, mounting of an oxidative burst, further inflammatory mediator release and generation of neutrophil extracellular traps. On the contrary, after severe trauma or during prolonged systemic inflammatory conditions that promote complement activation, C5a may become "too much of a good thing" (24, 25). C5a-activated neutrophils have been shown to induce endothelial and mesothelial autologous tissue destruction, for example, of the human omentum, which performs several basic innate and adaptive immune functions (26). Recent studies have indicated that C5a also results in a profound morphological polarization of neutrophils and immunometabolic changes by enhancement of the glycolytic flux (27, 28). C5a was able to generate an alkalotic intracellular milieu in neutrophils and a local lactic acidosis microenvironment despite the absence of an ischemic condition or oxygen debt (27). Yet, it remains unclear to what extent these effects may contribute to known metabolic changes and acidosis after severe trauma. By contrast, deficiency in complement-related factors have also been reported in trauma. Blood neutrophils from trauma patients, which appeared unresponsive to C5a, displayed an impaired oxidative burst activity, less C5aR expression, increased C3b binding activity, and clinically more (infectious) complications (**Figure 1**) (29). Chemotactic desensitization of neutrophils to C5a was observed

**Abbreviations:** Apaf-1, Apoptotic protease activating factor 1; APC, antigen presenting cell; ARDS, Adult respiratory distress syndrome; ATP, adenosine triphosphate; BCR, B-cell receptor; BLK, B-cell lymphocyte kinase; BLNK, B cell linker protein; Btk, Bruton's tyrosine kinase; cAMP, cyclic adenosine monophosphate; CD, cluster of differentiation; Cig, cytoplasmatic Ig; CLP, cecal ligation and puncture; CR, complement receptor; CR2-fH, Complement Receptor 2- factor H fusion protein; DAF, decay-accelerating factor; DAMP, damageassociated molecular pattern; DC, dendritic cell; DNA, deoxyribonucleic acidEP-54, response-selective molecular agonist peptide; ERK(1/2), extracellular signalingregulated kinase; Fas, first apoptosis signal; FcRII, Fc gamma receptor; FDC, follicular dendritic cell; fMLP, N-Formyl-L-methionyl-L-leucyl-L-phenylalanine; GLUT1, Glucose transporter 1; HLA-DR, human leukocyte antigen – antigen D Related; Ig, immunoglobulin; IL, interleukin; IP3, inositol trisphosphate; IR, ischemia-reperfusion; ITAM, immunoreceptor tyrosine-based activation motif; ITIM, immunoreceptor tyrosine-based inhibitory motif; LAMTOR, lysosomal adaptor and mitogen-activated protein kinase and mammalian target of rapamycin [mTOR] activator/regulator; LAT1, L-type amino acid transporter 1; Lck, Tyrosine-protein kinase Lck; MAC, membrane attack complex; MAPK, mitogenactivated protein kinase; MHC, major histocompatibility complex; MZ, marginal zone; NAD, nicotinamide adenine dinucleotide; NF-κB, nuclear factor κ-lightchain-enhancer of activated B cells; NLRP3, NACHT, LRR and PYD domainscontaining protein 3; PAMP, pathogen-associated molecular pattern; PARP, poly (ADP-ribose) polymerase; PI3K, phosphoinositide 3-kinase; PLC, phospholipase C; PTK, protein-tyrosin-kinase; Rag1, recombination activating gene 1; RNA, ribonucleic acid; sCR1, soluble complement receptor 1; Src, Proto-oncogene tyrosine-protein kinase Src; Syk, spleen tyrosine kinase; TBI, traumatic brain injury; TCC, terminal complement complex; TGF-β, transforming growth factor beta; Th, T helper lymphocytes; Treg, T regulatory lymphocytes; TUNEL, Terminal deoxynucleotidyl transferase dUTP nick end labeling; WT, wild-type; ZAP-70, Zeta-chain-associated protein kinase 70.

in patients with multiple trauma, burn injury or sepsis, indicating dysfunctional danger associated molecular pattern (DAMP) sensing by neutrophils upon excessive complement exposure (29, 30). In accordance with this, loss of C5aR1, C5aR2, and C3aR on neutrophils after multiple injury was found clinically and was correlated to infectious complications and multiple organ dysfunction (31–33). An integrated clinico-transcriptomic approach investigating RNA from leukocytes revealed that lower C5-expression on day 1 after polytrauma was associated with the development of nosocomial infections (34). Monocytes and macrophages are also major players in the orchestrated post-traumatic immune response and are highly responsive to complement factors. C5a is capable of priming peripheral blood mononuclear cells via p38 mitogen-activated protein kinase (MAPK) pathway activation (**Figure 1**), which in turn results in a significantly enhanced inflammatory mediator release upon secondary exposure to PAMPs (35). In previous studies, the migratory responsiveness of monocytes to C5a but not to fMLP had been found to be suppressed, with a maximum depression from 5 to 7 days after major injury (36, 37). The authors proposed these findings as an explanation for the predisposition of trauma patients to infection and poor wound healing.

When the "first line of defense" represented by physiological barriers, including the skin, is injured, the "second line of defense," represented by innate immune cells and the complement system is challenged. Proteomic analyses of wound fluids after (surgical) trauma revealed the presence of key complement components including C3 and factor B, in addition to a discrete proteomic leukocyte signature, indicating that as part of the second line of defense, both leukocytes and complement implement the clearance of wound debris and regenerative processes (38). Currently, the role of complement in regard to the "third line of defense," as formed by the adaptive immune system, is less well known, more so in the context of trauma.

# TRAUMA: WHERE DO COMPLEMENT AND ADAPTIVE IMMUNITY MEET?

The complement system and the adaptive immune responses after trauma have been described separately in various literature. Nonetheless, post-traumatic complement activation and its effects on lymphocyte phenotype and function remain elusive. At present, few studies mainly involving trauma injury like traumatic brain injury (TBI), burn injury, and splenectomy after blunt abdominal trauma, have brought the complement system and the adaptive immunity on to a common platform. Injury to the central and peripheral nervous system like TBI, peripheral nerve injury and spinal cord injury (SCI) have offered suitable opportunities to partially understand the basis of complement—adaptive immunity cross-talk. In a murine SCI model, IgM was proven to drive not only the recognition of neoepitopes on damaged cells, but also facilitated on-site complement activation and C3d deposition correlating with injury severity (39). This relation of complement activation to the injury was further bolstered through multiple studies, like, controlled brain contusion performed on strain-specific rats was found to have significant local C3 expression and high T-cell infiltration 3 days thereafter (40, 41). Implicating that mature T- and B-cells could be a pre-requisite for the generation of activated complement factors, Rag1(-/-) mice used to perform a closed head injury study had significantly decreased C3a levels compared to wild-type (WT) mice with no intergroup difference in injury severity (42). Thus, due to this experiential dependence of injury severity on complement activation, healing could also be achieved with complement inhibition. In post reperfused stroke it was seen that treatment with C3aR antagonist diminished infiltration of C3aR-expressing T-cells at the ischemic site, helping in neurogenesis (43). Pain responses due to peripheral nerve injury induced by partial ligation of the rat sciatic nerve could also be attenuated by intraperitoneal injection of soluble sCR1, with diminished macrophage and T-cell recruitment at the injury site (**Figure 1**) (44). However, the scope of these studies was limited as they did not investigate the cause/s of the observed effects.

Blunt abdominal injury is another example of trauma injury where patients frequently undergo splenectomy. The function of the complement system in attuning adaptive immune responses in the spleen has been underscored by two studies where the role of dendritic cells (DCs) and B-cell responses due to complement factors or receptors in the splenic marginal zone (MZ) has been described. The uptake and transfer of self-antigens by DCs to secondary lymphoid organs serves the crucial purpose of tolerance induction. However, proven through in vitro studies, CD8α <sup>+</sup> DCs resident in splenic MZ can also actively uptake and internalize circulating iC3b opsonized apoptotic leukocytes and these cells require complement receptor (CR) 3 and not CR4 for this purpose (45). Coming to B-cell mediated responses, C4 deficient mice had higher immune complex localization in the splenic MZ and impaired antibody response and class-switching, which was restored when antigen was directed to the splenic MZ (46). Hence, complement seemed to play a role in modulating self-antigen localization such that peripheral B-cell tolerance is maintained. Splenectomy post-trauma affected immune function in terms of reduced T-cell response to phytohemagglutinin, decreased number of lymphocytes, decreased IgM levels, and no changes in C3, C4, and C5 levels (47). Contradicting this former study, other studies concluded that either serum IgM levels did not vary (48) or there was an increase in B-cell population (49). The commonality among them being the unchanged levels of complement factors, a stark demerit of these conclusions was that activated complement fragments were not measured, rendering one inconclusive as to what exact role the complement system might have had in insinuating the adaptive immune responses. In a later study including polytrauma patients, investigation of complement regulatory surface proteins on lymphocytes from the patients showed significantly high CD59 expression 120 and 240 h post trauma and significantly decreased CD46 expression up to 48 h after trauma, with or without splenectomy (**Figure 1**) (31). Whether an impairment or augmentation of lymphocyte activity is a cause of deregulated complement activation post-trauma was confirmed from burn injury studies, like, generation of C1q degradation peptides in burn patients having an immunosuppressive effect on lymphocytes (50). Serum obtained from major burn injury patients, when subjected to complement inactivating temperatures in vitro could affect mitogen-associated lymphocyte blastogenesis, establishing the fact that complement is putatively necessary for lymphocyte development in a trauma event (51). Extending on this concept, in a pig burn wound model (described as 8 burn wounds inflicted for 20 s with a 170◦C heated copper rod over a 4 × 4 cm area on two flanks), systemic C3 increased significantly from day 9 and up to 60 days post injury, C4 increment was delayed after burn and a concomitant increase in Tcell infiltration at the wound site was seen on day 3 which declined 21 days post burn injury (52). Additionally, a local increase in C3 and C4 was observed 9 and 4 days post burn respectively, though both decreased after 21 and 9 days, respectively.

As evident from the paucity of relevant studies is that the causality of the complement—adaptive immunity interaction after trauma is still missing. In the following discussion, we focus on aspects of adaptive immunity, from antigen presentation to T- and B-cell functions, which have been proven to be under complement-mediated regulation and vice versa, and how such mechanisms are of importance in the "traumatic" context.

#### COMPLEMENT SYSTEM AND ANTIGEN PRESENTATION

Antigen presentation is the first and foremost step in priming lymphocytes for their effector functions. This includes processing of exogenous foreign particles, which are in turn presented by major histocompatibility complex (MHC) class II to CD4+ T-cells and endogenous foreign particles are presented by MHC class I to CD8+ T-cells. MHC class II molecules are principally expressed on professional antigen presenting cells (APCs) e.g., macrophages, DCs and B-cells, while all nucleated cells express MHC class I on their surface. However, in addition to conventional antigen-presentation modes, MHC class I can also cross-present, i.e., exogenous antigens can be presented on MHC class I of professional APCs (53). Reduced antigen-presentation functions were initially reported in postinjury macrophages, further having been frequently described in trauma studies (54, 55). For example, reduced antigen presentation and interleukin (IL)-12 and interferon (IFN)-G production after surgical trauma, a diminished population of HLA-DR+ monocytes early in trauma patients and attenuated IL-15 production by DCs following trauma hemorrhage have been reported (56–58). TBI and its impact could be lessened by targeted inhibition of class II-associated invariant peptide, an essential component in MHC antigen presentation (59). Therefore, antigen processing, presentation and co-stimulation of T-cells could well be under the regulation of activated complement factors, particularly in the event of excessive complement activation which has been observed after trauma. Few studies from the last decade could help one get an idea of the role played by complement factors in the uptake of antigens, self or non-self, through DCs and their effect in generating T-cell responses. The **e**xtensively convoluted pathophysiology of trauma can contribute to cellular apoptosis as seen in the spleen, thymus, and the gut and also in neurons after TBI and SCI (60–63) and requires resolution of unwarranted proinflammatory responses**.** Conversely, the said physiological mechanism can also yield in an imbalance of the inflammatory responses so generated. Suppression of proinflammatory response was identified to be mediated by DCs when a study found C1q and mannose binding lectin as receptors aiding apoptotic cell uptake by immature DCs (iDCs) (64). In connection to this, apoptotic cell removal by microglial cells, which are central nervous system tissue resident DCs, brings about inhibition of IL-6, IL-1α, TNFα, and IL-1β as opsonization with C1q targets apoptotic cells to DC (65). Another study described that C1q opsonized apoptotic cell uptake by iDCs reduced their surface expression of CD86 costimulatory molecule along with diminished subsequent Th1 and Th17 cell proliferation, generating a contracted inflammatory response (66). This may explain the observed delayed development of immune paralysis in trauma patients, giving rise to poor prognosis as well as the susceptibility to sepsis (67). In a human study, severely injured patients have demonstrated monocyte deactivation, decreased HLA-DR expression on B-cells and dramatically influenced T-cell activation (68). Furthermore, a study showed increase in splenic T-cell and decrease in splenic B-cell 26S proteasome activity concomitant with changes in NF-κB expression after trauma-hemorrhage, suggesting the essential involvement of proteasomes in the regulation of signal transduction in splenic T- and B- cells (69). However, this study did not further describe if the antigen presentation machinery in B-cells is affected anyhow, as proteasome degradation is a key step in cytosolic antigen processing pathway of antigen presenting cells (53).

A complement-enriched environment is necessary for antigen presentation and T-cell differentiation, which is strongly dependent on the type of complement activation products generated (70, 71). Self-antigen presentation after degradation of apoptotic cells is guided by activated C3, contributing to MHC class II antigen processing and presentation (72). In several studies, the role of CR1 and 2 in prolonged antigen presentation have been described for antigen-specific B-cells and macrophage-mediated antigen presentation to T-cells (73–77). Studies have also attempted to describe the mechanistic pathway of antigen-loading efficiency of tetanus toxin-opsonized by C3b and C4b particles (78–80), though the type of antigen spoken hereof is of pathogenic source and may not necessarily explain mechanisms applicable in the context of trauma. Yet, the dynamics of C3 fragment mediated opsonization in purportedly regulating antigen processing is an intriguing premise. C3 activation fragments bound to antigen can delay its degradation in the cytosol by evading its fusion with lysosomes and maintain the antigen-loaded internalized cargo within DCs, a process which modulates T-cell responses to self-antigens (72). Additionally, antigens complexed with C3b were shown to form more stable MHC class II complex, modulating the antigen processing pathway in the late endocytic stage, the loading of the antigen onto MHC and the presentation of the antigen to T-cells resulting in T-cell proliferation (72, 78, 79, 81, 82). Furthermore, high intracellular cAMP in C3aR- or C3-deficient APCs impaired antigen presentation (83). While antigen presentation is facilitated by complement activation products, DC maturation could also be inhibited by iC3b, which exerts its effect by increased extracellular signal-regulated kinase 1/2 (ERK 1/2) phosphorylation and reduced p38 MAPK phosphorylation, resulting in decreased IL-12p70 and increased IL-10 (84). Even anaphylatoxin C5a is a potent effector in priming lymphocytes, because modified C5a peptide EP54 (responseselective agonist peptide) was identified as an adjuvant in DC-mediated T-cell priming (85). In addition to activated C3 and C5 products, elucidated effects of another complement factor C1q have been shown to drive DC-mediated functional activation of T-cells and cross-presentation of exogenous antigen to CD8+ T-cells resulting in greater CD8+ T-cell proliferation (86, 87). The gradual momentum gained by these evidences puts forward a need to confirm the same in trauma. While different experimental confirmations have brought forth the importance of complement activation and degradation products on antigen processing and presentation pathways, in vivo and in vitro analyses could help in further elaborating how high levels of C3 and C5 activation and/or degradation products generated during trauma could affect these downstream mechanisms.

#### COMPLEMENT AND ADAPTIVE IMMUNITY

Accumulating evidences supporting a possible interaction between the complement system and T-cells have been discussed in some reviews extensively (88–91). In this section we will discuss the relevant studies and recent developments with respect to complement—adaptive immunity crosstalk from the perspective of trauma-immunology.

#### Anaphylatoxin-Regulated Th17 Response

Trauma severity and its outcome appear to be associated with elevated levels of the anaphylatoxins C3a, C4a, and C5a as well as other activated factors, including factor B and TCC (7, 92). Therefore, these intermediates could affect the resolution of the trauma response when adaptive immunity is involved. Invariably, there are obvious differences between sterile inflammation (like trauma) and infectious inflammation (like sepsis), which although can show similar effects, may or may not employ identical pathways to the resultant observed effect. Therefore, the origin of the antigens involved to enunciate the immune response is important. These antigens can be foreign (e.g., PAMPs), self yet foreign or DAMPs, or self-antigens. In PAMP-driven sepsis mice, treatment with anti-C5a antibodies displayed significantly reduced IL-17 levels compared to controltreated sepsis mice (93). C5a can also increase IL12+ DC migration from the peritoneal cavity to the periphery and could prime both Th1 and Th17 T-cells (94). Delving deeper into the exact effect of Th17 T-cells it was seen that, elevated level of IL-17 is generally observed in sepsis and this can be abrogated by a Rho-kinase inhibitor designating it as a participating downstream signaling mediator (95). Th17 cells mediating proinflammatory responses in trauma have been already reviewed extensively, and a few recent studies have investigated into their modulation by the complement system like elevated recruitment of the proinflammatory Th17-cell cohort in a trauma setting, whose inhibition proved to be protective and supported healing (66, 96–98). Interestingly, the DAMPdriven trauma and its associated Th17 responses is caught up in various contradictory results, albeit without the clarification of an intervening involvement of the complement system. When Inatsu et al., obtained peripheral blood mononuclear cells (PBMCs) from third degree thermal injury patients to culture and stimulate these PBMCs with Candida albicans hyphae in vitro, they found that PBMCs from burn patients had impaired Th17 differentiation (99). Notwithstanding this reported impairment in Th17 phenotype acquisition, subsequent murine studies showed in third degree burn model that Th17 cells were highly recruited at the site of wound as early as 3 h (100) and even 7 days post-injury (101). Few clinical and murine studies could support this conclusion as well for example increased circulating population of CD4+ Th17 T-cells 5 days after trauma in patients with an ISS > 20 (102) and greater IL-17 production by splenocytes in mice which were given an acute scald burn injury of 5 s at 95◦C (103). Apart from systemic effects, the local response to trauma hemorrhage was found to effectively increase the Treg:Th17 ratio in the mesenteric lymph nodes, highly likely due to the concomitant decrement observed in circulatory CD103+ DCs (104). A recent clinical study where 114 trauma hemorrhagic shock (THS) and 50 control patients were recruited also showed that, THS patients who developed sepsis later on had lower Th17:Treg cell ratio (105). As sepsis is a common development in later time-points post-trauma, the relevance of IL-17 as a prognostic marker (67) and the contextual relevance of complementopathy in regulating this response could be further validated. IL-17 has also been identified to predict organ damage through computational network analysis of systemic inflammatory markers in blunt injury patients and neutralizing IL-17A attenuated organ damage in a trauma/hemorrhage mouse model (106). Therefore, it is a prerequisite to disentangle the gradually building yet convoluted concept of how exactly Th17 responses are under intricate modulation/s following trauma. This could be dependent on the time and type of injury, local to systemic differences and the inherent inconsistencies found in human and murine systems, apart from the proven role of complement in sepsis as opposed to the unexplored one in trauma. Complement factors and its regulatory components could also be one of those modulators in trauma injury and its yet untapped functions is an opportunity for further investigations.

# Other Complement-Mediated T-Cell Responses

Gδ T-cells have been recently associated in a few trauma-related burn injury and fracture studies for example in regulation of fracture healing, high recruitment of Gδ T-cells at the site of the wound, wound healing through modulation of myeloid cells and higher neutrophil recruitment in the lungs after trauma-hemorrhage (107–109). A recent study demonstrated how mitochondrial DAMPs can directly interact with Gδ T-cells and induce synthesis of proinflammatory cytokines, including IL-1β and IL-6 (110). Though unavailable in the context of trauma, complement system's role has been elaborated in PAMP-driven systemic inflammatory response like sepsis. Gδ T-cells obtained from cecal-ligation puncture (CLP) mice and transferred to recipient CLP mice followed by in vivo treatment with anti-C5a improved survival rates of the recipients (93). When these T-cells were obtained from anti-C5a pre-treated CLP mice and transferred to recipient CLP littermates, the recipients displayed reduced survival, indicating that the pathogenic role of Gδ T-cells was modulated by C5a (93). Extending upon the effect of C5a, increased C5a generation was shown to stimulate neutrophil release of histones, which in turn can induce lymphocyte apoptosis (111). This study used WT and C5aR1 knockout (C5aR1-/-) mice to induce CLP-sepsis and stained apoptotic cells in the spleen by TUNEL, whereby WT CLP-sepsis mice had up to 5 times higher apoptotic cells per view field. The complement—lymphocyte connection could be confirmed for other complement factors as well, like, functionally defective Tregs, along with defective DCs and impaired B-cell switching were found in a 2 year old male patient with C3 deficiency (112). Hence, the concomitant overactivation of complement factors and following consumptive effect that trauma has on them could very well impede the functioning of the adaptive immune system. A seldom explored complement regulatory protein in traumatic conditions is the decay acceleration factor (DAF), which exerted a protective effect in a pig hemorrhage model and was seen to be upregulated in human neutrophils early after polytrauma (31, 113). Imitating central nervous system inflammation by injecting myelin oligodendrocyte glycoprotein (35–55) peptide showed that Daf1 transgenic mice with a higher cell-surface DAF expression displayed an attenuated disease phenotype and reduced antigenspecific Th1 and Th17 responses, further emphasizing on the role of DAF in T-cell-mediated proinflammatory functions (114). Hence, the robust involvement of complement factors in generation of T-cell responses could affect the mechanisms that are modulated downstream and the evidences explaining these potential mechanisms have been a recent addition to our understanding.

#### Potential Mechanisms in T-Cells

Severe tissue trauma and hemorrhagic shock (HS) is often associated with a significant oxygen deficit as reflected by lactate acidosis and a drop in the base excess (13). The decrease in pH can be manifested systemically and be even more pronounced locally. Low pH environments are known to activate key complement factors and thereby generate C3 and C5 activation products (115) which in principle can contribute to the complement activation and modulation of lymphocyte intracellular mechanisms. Metabolic effects of activated complement at lower concentrations functionally modulated CD4+ T-cells in terms of glutamine utilization and enhanced oxidative capacity of T-cells; whereas a higher concentration of complement resulted in cell death by ATP depletion (116). A combination of TCC and immune complexes trigger CD4+ T-cell activation by inducing phosphorylation of ζ-chain, ZAP-70, Syk, Src, and Lck along with the arrangement of the actin cytoskeleton and effector T-cell functions (117, 118). Upon stimulation, T-cells in C3 knock-out mice have also been shown to display attenuated T-bet transcription factor expression, a factor which determines lineage specificity in CD4+ T-cells (71). Nevertheless, the complement—T-cell association is not just restricted to this. In the last few years, the intimate association of complement mediated signals in T-cell phenotype acquisition has been elaborated through a series of important studies. It was seen that, absence of C5aR and C3aR signaling caused CD4+ T-cells to acquire a Treg cell phenotype in the absence of DCs, through induction of TGF-β and downregulation of complement factor expression by T-cells (119). Thus, autocrine complement activity of Thelper cells gained relevance with a following study showing the generation of intracellular C3a to be driven primarily by cathepsin-L protease synthesized by T-cells (120). A subsequent more recent study in this context showed how T-cell effector function results from intracellular metabolic alterations, modulated by intracellular synthesis of complement products and surface-expressed CD46 (121). Intracellular generation of C3 cleavage products and binding of C3b to CD46 drives the upregulation of glucose and amino-acid transporters and ragulator complex protein LAMTOR which affects cell growth (121). Even intracellular activation of C5 and its effect on NLRP3 inflammasome signaling are necessary for Th1 cell IFN-G synthesis (122). Considering these evidences, it would be interesting to see how and where these mechanisms are affected, when the aforementioned downregulation of CD46 occurs on Tcells following trauma (31) and the role of intracellular C5aR1 in CD4+ T-cells thereafter.

#### Complement-Mediated B-Cell Responses

Raad et al. has extensively reviewed how autoantibody generation is a common observation after central nervous system trauma (103). The review discusses in depth the origin of such autoantigen synthesis, which may result due to generation of autoreactive T-cells or even B-cell hyperactivation. Here, the role of the complement system in expediting lymphocyte sensitization and their responses to self-antigens generated following profuse tissue damage after trauma is left to be explored in detail. Ischemia-reperfusion (IR) injury is a common consequence of trauma where the role of complement has been confirmed together with humoral immune responses (51, 52). Clonally specific B-cells and the activated classical complement pathway were shown to be involved in injury pathogenesis, with IgM implicated as the key mediator (53). Additionally, the relevance of CR2 has been demonstrated in inducing IgG and IgM antibody production and their concomitant effects in IR injury pathogenesis (54). Inhibition of spleen tyrosine kinase (Syk), a membrane signaling protein found in B- and T-cells, elicited protective effects by reducing C3 and IgM deposition in tissues in a model of mesenteric IR injury with remote lung injury (55). Opposing existing concepts, C3 was subsequently shown not to be involved in IR injury, whereby a study revealed that surface-expressed CR2 on marginal-zone B-cells did not require C3 activation products to generate IR injury-induced antibodies (56). While nonspecific and specific humoral immune responses have been evaluated in burn injury and SCI, additional observations have been made in major trauma models (123–125). A study by Faist et al. evaluated Bcell function in 30 patients following major trauma over a 21 day period (126). Although the number of circulating B-cells in the trauma patients was not decreased following injury, the number of terminally maturing CIg+ B-cells were significantly decreased compared to controls up to 21 days post-trauma. Bcells obtained from the recruited patients in this study were analyzed in vitro and synthesis of IgA, IgM, and IgG by these B-cells were significantly reduced on day 1 after trauma. Later, IgG synthesis decreased on day 3 though total IgG levels were supra-normal. IgM synthesis was diminished throughout the study as compared to the controls (126). This post-traumatic IgM deficiency has also been described in other studies, a fact which is supported by the lack of the lymphokine IL-2 found in patients with multiple trauma (127, 128). Additionally, B-cell derived autoantibodies after SCI are co-localized with C1q on injured neurons, though it remains unclear whether complement activation modulates the production of these autoantibodies (50).

The complement system is furthermore involved in the maintenance of homeostasis by clearing debris, apoptotic bodies, and immune complexes (129). It regulates effector functions of natural antibodies, and C3 cleavage products participate in the opsonization and transport of antigen to the B-cell compartments of the secondary lymphoid tissue (130–132). A detailed role of complement regulators has been realized of late, like CR2/CD21 and CR1/CD35 expression by follicular DCs (FDCs), which function to retain the antigen in the lymphoid follicles required for the activation, proliferation, and antibody generation of B-cells (133, 134). CR1/2 and complement C3 support the localization of IgM-immune complexes in the splenic MZ (135), where B-cells concentrate the IgM-immune complexes onto FDCs. Supporting this, decreased MZ B-cells dramatically diminish the amount of IgM-immune complex on FDCs. The localization of antigen on FDCs has been implicated in optimizing the formation of germinal centre, memory B-cell formation, somatic hypermutation, and IgG class switching (135). Furthermore, in the FDCs of the peripheral lymphoid tissues, including the spleen and lymph nodes, CR1 and CR2 assist in autoreactive B-cell clone anergy by binding and presenting the self-antigen to these B-cells (136). This clearly points toward the differential expression of these two antagonistic complement regulators during the development of human B-cells, a phenomenon which may influence the maintenance of peripheral B-cell tolerance. Deficiencies in upstream complement molecules, including C1q, C2, and C4, may result in inefficient clearance, in the absence of which these products become a source of autoantigens for B-cells and a probable source of an autoimmune response (137). In addition to complement receptors, attachment of C4b to selfantigen and localization of these complexes to CD35 on stromal cells within the bone marrow also regulate the selection of potentially autoreactive B-cells (138). B-cell maturation and selftolerance being the major foreground of the few players from the complement arsenal, B-cell class switching was shown to be regulated by C3, as C3-deficient patients had abnormal variation in IgG isotypes, including high IgG3, low IgG2, and no IgG4 response (139). These isotypes can in turn modulate the type of adaptive immune responses generated, having an overall effect in subsequent inflammatory responses generated in the said individuals. This indicates but does not yet prove how complement imbalance post-trauma could contribute to altered B-cell responses though further studies could help explain the same.

#### Complement-Mediated B-cell Signaling

The antigen receptors on B-cells, similar to T-cells, are present as multiprotein complexes. Binding to their cognate ligand initiates an intracellular signaling cascade involving translocation to the nucleus and changes in gene expression that dictate the response of the lymphocyte. C3d has been extensively described in mediating signaling pathways, thus linking the innate to the adaptive immunity (140). In this regard, CR1 and CR2 are the best characterized complement regulators that mediate complement-dependent B-cell signaling. C3d interacts with both CR2 and the B-cell-receptor (BCR), causing activation and proliferation of B-cells and the production of specific antibodies. Studies have demonstrated that C3d can bind to CR2, which mobilizes the B-cell signaling complex consisting of CR2, CD19, CD81, and leu-13. Studies in mice lacking either CD19 or CR2 revealed that the CD19-CR2 complex is essential for proper immune function. This B-cell signaling complex confers important co-receptor activities to the BCR. The BCR consists of a non-signaling membrane immunoglobulin and a transmembrane immunoglobulin heterodimer Igα and Igβ bearing immunoreceptor tyrosine activation motifs (ITAMs). BCR stimulation activates two classes of tyrosine kinases: Srcfamily kinase (Src-protein-tyrosine kinase [PTK], including Lyn, Fyn, Blk, or Lck) and Syk kinase (141). Active Src-PTKs facilitate the phosphorylation and consequent activation of Syk. Activated Syk phosphorylates numerous intracellular substrates mediating B-cell activation, including the ITAMs of Igα and Igβ, and various adaptor proteins, including B-cell linker protein [BLNK or SH2-domain containing leukocyte protein of 65 kDa (SLP-65)]; which in turn leads to the recruitment of Bruton's tyrosine kinase (Btk) and phospholipase C-G(PLC-G) (142– 146). Subsequent phosphorylation of PLC-G generates IP3. This molecule orchestrates a cascade of events resulting in Ca2<sup>+</sup> release from intracellular stores and activation and nuclear localization of the nuclear factor of activated T-cells (NFAT) and nuclear factor "kappa-light-chain-enhancer" of activated B-cells (NF-κB) pathways (147–152).

CD19 is the only molecule within the B-cell signaling complex capable of intracellular signaling. It acts as a membrane adaptor molecule by recruiting signaling intermediaries, including Vav and phosphotidylinositol-3 kinase (PI3K), which enhance Ca2<sup>+</sup> flux, activation of ERK 1/2, and ultimately B-cell proliferation (153). C3d-opsonized Ag can co-ligate BCR with CR2, consequently improving antigen uptake (74, 154). This complex recruits several copies of CR2 and the B-cell signaling complex, upregulating intracellular Ca2<sup>+</sup> mobilization and in turn decreasing the threshold of B-cell activation (4). In conclusion, C3d-CR2 binding resulting in CR2-BCR co-ligation reduces the threshold for cell activation by 10- to 100-fold (155) BCR-CR2 co-ligation is also able to inhibit first apoptosis signal (Fas)-mediated apoptosis. In this context, co-ligation results in increased expression levels of anti-apoptotic proteins, including Bcl-2 and the cellular FADD-like interleukin-1-ß-converting enzyme inhibitory protein short form. Invariably, the increase in Bcl-2 appears to inhibit Apaf-1-mediated apoptosis (156).

Whereas the role of human CR2 in B-cell activation is relatively well-established, much less is known about the exact function of CR1 (CD35). In contrast to CR2, CR1 has been described to mediate inhibitory signals (157–159). CR1 ligands include aggregated C3 and aggregated C3(H2O), which in contrast do not bind to CR2. Furthermore, CR1 binds to the complement components C3b and C4b and has co-factor activities in their cleavage by factor I to iC3b, a substrate for CD21, and iC4b, respectively (160). It was demonstrated that CR1 strongly reduced the intracellular Ca2<sup>+</sup> level and phosphorylation of cytoplasmic proteins. Importantly, this inhibitory activity also occurs in the presence of the costimulatory cytokines IL-2 and IL-15 (157).

In trauma, impaired cellular function, including B-cell activation and antibody production, has been described. A recent study showed that the absence of CR1 and CR2 was protective toward secondary damages following closed head injury, including reduced IgM deposition at the injury site (161), though it was not sufficient to gather how are the downstream signaling mechanisms regulated to confer that protection. Further studies could attempt at deconstructing this yet unexplored premise.

#### FUTURE PERSPECTIVES

Considerable knowledge regarding interaction of the complement system with the adaptive immune system has accumulated over time. Nevertheless, there is a scope for experimental and clinical studies to provide an insight into the biology of complement and B- and T-cell interactions particularly after tissue trauma. It is well established that complement receptors are involved in B- and T-cell signaling and function. In the context of trauma, the broadly described role of CR2 in B-cell-signaling cascades could be investigated upon. While CR2 mediates activation of B-cell signaling, CR1 has B-cell inhibitory functions. In addition to CR1 and CR2, studies have also revealed a minor expression of CR3 and CR4 on B- and T-lymphocytes (162). Their expression on leukocytes, especially macrophages has important functions in inflammatory conditions, including tissue trauma and infection, where CR3 and CR4 enhance antimicrobial functions by potentiating leukocyte cell adhesion (163), though their function on Band T-lymphocytes remains undetermined. Because CR3 is expressed by cytotoxic T-cells, it can be speculated that this receptor may play an important role in lymphocyte functions (162). Like complement regulatory proteins, the receptors for the anaphylatoxins (C3aR, C5aR) appear to be crucial for proper cellular functions of monocytes and neutrophils, though their respective roles in adaptive immune cells have not been fully understood. An early study found C3aR on neutrophils, monocytes, and eosinophils, but not on B- lymphocytes, whereas another study found expression of both anaphylatoxins on tonsillar B-cells (164, 165). While a recent study found C3aR expression both at the mRNA and protein levels within the B-cell (166), to date it remains unclear whether they express C5aR (167).

Several studies have gone on to describe the direct and indirect involvement of the complement system in modulating B-cell responses, albeit not as a consequence of trauma. For example, C3 is required for memory B-cell differentiation and locally generated C5a by peritoneal macrophages is required for chemokine ligand 13 synthesis which is further needed for B1 cell homeostasis (168, 169). Activation of the BCR increased the amount of intracellular C3aR though it could not be detected on the cell surface even after activation. Additionally, B-cellderived C3 and C3aR appear to have a positive impact on allogeneic stimulation of T-cells (166). Studies of T-lymphocytes have confirmed surface expression of C3aR and C5aR, and revealed that C3aR and C5aR signaling have an important impact on functions of T-cell subpopulations, including Th1 and Th17 responses (170). The effects of trauma on lymphocytes has been barely investigated. Only a limited number of publications indicate that B- and T-cell functions are impaired after trauma, but currently no study exists to clarify the role of the complement system in it. This could indicate that different forms of trauma may affect different pathways of complement and lymphocyte activation, but we do not have any evidence to explain which mechanisms are involved in this putative functional cross-talk. In this context, DAMPs appear to be undermines T-cell survival. Severe injury with consequent tissue damage leads to excessive release of self-DAMPs, including histones (21). Histones released by the effect of C5a on neutrophils have also been described to induce T-cell apoptosis (**Figure 2**). Therefore, it may be rational to propose that post-traumatic enhanced histone levels is an indirect effect of C5a anaphylatoxin and contribute to T-cell death by apoptosis. Additional investigation is desirable to understand whether post-traumatic C3aR and C5aR signaling is affected. Regarding cells of the innate immune system, C5a can induce C5aR1-mediated morphological changes of neutrophils together with altered chemotactic activity and neutrophils from septic shock patients even displayed decreased C5aR1 expression (28, 171). Activation of C5aR has been described to also enhance proinflammatory Th1 responses and to reduce antiinflammatory Th2 responses, which should be also addressed in the context of trauma. Because trauma induces a shift toward the Th2 phenotype, it may be speculated that C3aR and C5aR signaling is impaired or their expression is decreased (**Figure 2**).

With regard to lymphocyte signaling, T-cells have been described to generate intracellular C3 and C5, which in turn can be cleaved and activated and are able to bind to receptors, including C3aR, C5aR, and CD46, thus altering signaling, cell growth, and T-cell responses. What remains unclear is whether autocrine- and/or systemically-generated C3 and C5 cleavage products (e.g., by trauma) may play predominant roles in signaling. It is well established that trauma generates excessive complement activation products. Because complement interferes with receptors on different T-cell subpopulations, including Th1, Th2 and Tregs, it is important to investigate how trauma induces changes to T-cell signaling, function and activity. In this regard, CD46 appears to represent an important receptor in the regulation of T-cell activation and function. For Th1 responses, C3b binding to CD46 and subsequent signaling has also been described to play a critical role in cytokine generation including IFN-G, IL-10, and granzyme-B production in Th1 cells and in their metabolic processes (121, 172). Kolev et al. further established that changes in metabolism influence the resulting CD4+ T-cell functions (i.e., IFN-G and IL-10 production), which may also be influenced in trauma (121). Excessive C3b generation after trauma may alter CD46 function or even expression on T-cell and consequently the functioning of Tcells. Following trauma, the presence of CD46 was found to be decreased on T-lymphocytes (31). Trauma may additionally change the concentrations of intracellular C3 and C5 and their activation, so that an autocrine loop alters CD46 signaling and thus T-cell activation. Further investigations could analyze whether excessive complement activation products after trauma influence T-cell survival, because complement products have been described to induce ATP depletion with subsequent T-cell apoptosis (**Figure 2**).

Nonetheless, no study to date has provided a mechanistic explanation of a trauma-induced impact on B-cell signaling. One interesting research field may be inhibitory receptors bearing immunoreceptor tyrosine-based inhibitory motifs (ITIMs), which are required for the regulation of B-cell immune responses. To date, numerous inhibitory receptors including CD22 and FcRII have been characterized as possessing one or more ITIMs within their cytoplasmic domain, which generate and transduce inhibitory signals (173). It may be speculated that after trauma, B-cell activation is impaired through a compensatory mechanism to protect the host from further immune reactions mediated by the adaptive system. It could also be possible that in addition to complement interaction with CR1 and CR2, the complement components may interact with these inhibitory receptors on Bcells, reinforcing inhibitory B-cell signaling and thus impairing B-cell activation. Further investigations in relation to the interaction of complement components with other receptors, particularly those with ITIMs could be considered.

Regarding complement therapeutic strategies, several specific and effective approaches have been recently evaluated for their clinical use in inflammatory disorders, including trauma, HS, and sepsis (21). The majority of these investigations focused on the influence of trauma on innate immune responses, including

on neutrophil and macrophage function, but less, if at all, on the modulation of adaptive immunity. Therefore, it remains to be investigated to what extent neutrophils, macrophages and also adaptive immune cells contribute to metabolic changes after trauma and whether C5a as a "metabolic switch" may be influenced by specific inhibition (27). C5 cleavage could be inhibited ex vivo by small peptide inhibitors, which require future in vivo testing for their clinical potential (174). Whereas systemic inflammatory processes after trauma or during infection appear to significantly benefit from central complement blockades, there is no overall benefit for all injured tissues (22, 175, 176). For example, C5aR1 blockade failed to improve uneventful fracture healing, where femoral osteotomy was performed in mice following which immediate and post-12 h, a C5aR antagonist treatment was given (175). Early blockade of C5aR ameliorated both, the IL-6 levels and neutrophil recruitment at the site of injury, but late blockade did not effect in the same. Therefore, the compartmentalized immune response after trauma, involving most likely both innate and adaptive immune reactions, appears to require a specific complement intervention in the future with both adequate spatial and temporal resolution (177). Elaborating on the temporal aspect, the treatment with anti-C5a early post-blunt chest trauma in rats ameliorated subsequent leukocytosis, reduced white blood cell recruitment in the lung, and also decreased trauma induced TNF-α levels (178). As for local effects, 12 h after blunt chest trauma coupled with sepsis induction, C5 deficient mice exhibited decrease in cytokines IL-6 and monocyte chemotactic protein-1 levels in the lung tissue (179). Local healing processes are actively under the regulation of C5a, as its cognate receptor C5aR1 deficiency (C5aR1-/-) affected fracture healing in a mouse model of femoral fracture (180). A remarkable finding in this same study was that, 14 days after fracture in C5aR2 deficient mice, the recruitment of CD8+ T-cells was significantly higher compared to C5aR1-/ mice with fracture, indicating the differences in the elicited function of C5aR1 and C5aR2. Hence, targeted inhibition of complement at the damaged-tissue site is already partly realized. For example, the chimeric CR2-fH construct (mTT30), which can inhibit neuroinflammatory response after traumatic brain injury (181). Similarly, other complement blockers like TCC inhibitors C6 antisense oligonucleotide (OMCI) (182) and CD59- 2a-CRIg (183) could turn out to be promising inhibitors and as a result, modulators of the adaptive immune responses. However, in the future, such damage-targeted principles need clinical translation and must not only consider the innate immune response but also the adaptive immune response after trauma, thereby aiming to improve any observed B- and Tcell dysfunction. Another emerging field represents vaccines that influence B- and T-cells in the context of prevention or treatment of infectious morbidity and mortality after traumatic injury, including soft-tissue wounds, human or animal bites, or after splenectomy (184). New effective therapeutics on the C3 level, for example, by compstatin as a small peptide inhibitor and on the C5 level, for example, by NDT9513727 as a nonpeptide inverse agonist of C5aR1, need to be tested in the trauma context not only on the innate immune responses but also on the adaptive immune changes (185–187). In the clinic, dysfunctional adaptive immunity after severe tissue trauma requires improved and timely diagnosis. A first approach is

#### REFERENCES


represented by clear definition of the persistent inflammation– immunosuppressive catabolism syndrome with its innate and adaptive immune-suppressive characteristics (188). Up until now, the underlying mechanisms remain unclear and need to be elucidated to improve the long-term quality of life of a patient after severe trauma. Overall, complement may represent an innovative therapeutic target in trauma and critically ill patients, because it principally bridges innate and adaptive immunity.

#### AUTHOR CONTRIBUTIONS

SC is the first author, EK is the second and MH-L is the Senior author of the paper.

#### FUNDING

This work is supported by grants from the German Research Foundation (DFG) to MH-L. (INST 40/479-1).

#### ACKNOWLEDGMENTS

We are grateful to Dr. Stephanie Denk for preparing the figures for this review.


dodecyl sulfate-stable HLA-DR dimer production. Eur J Immunol. (1997) 27:2673–9. doi: 10.1002/eji.1830271029


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Chakraborty, Karasu and Huber-Lang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Sickening or Healing the Heart? The Association of Ficolin-1 and Rheumatic Fever

Sandra Jeremias Catarino<sup>1</sup> , Fabiana Antunes Andrade<sup>1</sup> , Angelica Beate Winter Boldt 1,2 , Luiza Guilherme<sup>3</sup> and Iara Jose Messias-Reason<sup>1</sup> \*

<sup>1</sup> Molecular Immunopathology Laboratory, Department of Medical Pathology, Clinical Hospital, Federal University of Paraná, Curitiba, Brazil, <sup>2</sup> Human Molecular Genetics Laboratory, Department of Genetics, Federal University of Paraná, Curitiba, Brazil, <sup>3</sup> Heart Institute (InCor), School of Medicine, University of São Paulo, São Paulo, Brazil

Rheumatic fever (RF) and its subsequent progression to rheumatic heart disease (RHD) are chronic inflammatory disorders prevalent in children and adolescents in underdeveloped countries, and a contributing factor for high morbidity and mortality rates worldwide. Their primary cause is oropharynx infection by Streptococcus pyogenes, whose acetylated residues are recognized by ficolin-1. This is the only membrane-bound, as well as soluble activator molecule of the complement lectin pathway (LP). Although LP genetic polymorphisms are associated with RF, FCN1 gene's role remains unknown. To understand this role, we haplotyped five FCN1 promoter polymorphisms by sequence-specific amplification in 193 patients (138 with RHD and 55, RF only) and 193 controls, measuring ficolin-1 serum concentrations in 78 patients and 86 controls, using enzyme-linked immunosorbent assay (ELISA). Patients presented lower ficolin-1 serum levels (p < 0.0001), but did not differ according to cardiac commitment. Control's genotype distribution was in the Hardy-Weinberg equilibrium. Four alleles (rs2989727: c.−1981A, rs10120023: c.−542A, rs10117466: c.−144A, and rs10858293: c.33T), all associated with increased FCN1 gene expression in whole blood or adipose subcutaneous tissue (p = 0.000001), were also associated with increased protection against the disease. They occur within the <sup>∗</sup>3C2 haplotype, associated with an increased protection against RF (OR = 0.41, p < 0.0001) and with higher ficolin-1 levels in patient serum (p = 0.03). In addition, major alleles of these same polymorphisms comprehend the most primitive <sup>∗</sup>1 haplotype, associated with increased susceptibility to RF (OR = 1.76, p < 0.0001). Nevertheless, instead of having a clear-cut protective role, the minor c.−1981A and c.−144A alleles were also associated with additive susceptibility to valvar stenosis and mitral insufficiency (OR = 3.75, p = 0.009 and OR = 3.37, p = 0.027, respectively). All associations were independent of age, sex or ethnicity. Thus, minor FCN1 promoter variants may play a protective role against RF, by encouraging bacteria elimination as well as increasing gene expression and protein levels. On the other hand, they may also predispose the patients to RHD symptoms, by probably contributing to chronic inflammation and tissue injury, thus emphasizing the dual importance of ficolin-1 in both conditions.

#### Keywords: *FCN1*, ficolin-1, polymorphism, haplotype, rheumatic fever, rheumatic heart disease

#### *Edited by:*

Joao P. B. Viola, Instituto Nacional de Câncer (INCA), Brazil

#### *Reviewed by:*

Thomas Vorup-Jensen, Aarhus University, Denmark Bryce Binstadt, University of Minnesota Twin Cities, United States

*\*Correspondence:*

Iara Jose Messias-Reason iarareason@hc.ufpr.br

#### *Specialty section:*

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

*Received:* 15 June 2018 *Accepted:* 05 December 2018 *Published:* 18 December 2018

#### *Citation:*

Catarino SJ, Andrade FA, Boldt ABW, Guilherme L and Messias-Reason IJ (2018) Sickening or Healing the Heart? The Association of Ficolin-1 and Rheumatic Fever. Front. Immunol. 9:3009. doi: 10.3389/fimmu.2018.03009

# INTRODUCTION

Rheumatic fever (RF) is an immune-mediated disease occurring in genetically susceptible individuals, as a sequelae of group A (GAS) Streptococcus pyogenes pharyngitis. It affects predominantly children and adolescents in low-income and developing countries, where it remains a considerable public health problem (1, 2).

Five major manifestations reflect target tissue involvement in RF, including synovium (inflammatory arthritis), heart valves (endocarditis), brain (Sydenham's chorea), skin (erythema marginatum), and subcutaneous tissue (nodules). Repeated or severe RF episodes can result in permanent damage to the heart valves, leading to rheumatic heart disease (RHD), the most common acquired cardiovascular disease in young adults (3, 4). Rheumatic heart disease (RHD) is associated with high morbidity and mortality, causing 9 million disability-adjusted life years lost, 33 million cases (5) and 275,000 deaths each year (6). This multifactorial disorder involves multiple genetic and environmental factors, not yet fully elucidated. Well-designed case-control studies strive to unravel the genetic susceptibility to this disease, given that the only genome-wide association study done with RHD has rendered no relevant results (7).

Autoimmunity is known to play a role in the pathogenesis of RF and RHD, with tissue damage being mediated by autoantibodies resulting from molecular mimicry between GAS and heart tissue proteins. It has been shown that GAS molecules such as N-acetyl-β-D-glucosamine (GlcNAc) and M protein display cross reactivity with valve and myocellular contractile proteins of the host. GlcNAc is the immunodominant cell wall antigen of GAS, recognized by ficolins, molecules that comprise important pattern-recognition receptors (PRRs) of the complement (2, 5, 8, 9).

The complement system plays an important role, both in the defense against GAS infection, as well as in the development of autoimmunity in RHD. This system promptly responds to any pathogen, bridging innate and adaptive immune responses (10–14). Ficolins initiate one of the three complement activation pathways, known as the lectin pathway, along with collectins, as mannose-binding lectin (MBL). They occur in oligomeric structures of a basic homotrimer, where each chain is formed by a collagenous strand and a C-terminal fibrinogenlike recognition domain. The oligomers are complexed with homodimers of serine proteases (MASP-1 or MASP-2). Ficolin oligomers bind specific patterns of acetylated residues on the surface of pathogens or altered cells (8, 15). MASP-1 then autoactivates, transactivating MASP-2 leading to subsequent cleavage of downstream complement components [reviewed by Boldt et al. (16)]. The lectin pathway, along with the classical and alternative pathways, converge at the cleavage of C3 and C5, with subsequent C3b opsonization and pathogen phagocytosis or its destruction by membrane attack complex (MAC) pores on the cell membrane (17).

Three human ficolins have been described: ficolin-1 (Mficolin), ficolin-2 (L-ficolin), and ficolin-3 (H-ficolin, Hakata antigen) (15). Unlike any other PRRs, ficolin-1 is expressed by myeloid cells, being found in monocytes, neutrophils and macrophages of the lung and spleen (8). It is also the only one that occurs in both soluble form (0.05–1.0µg/mL in serum) and on the cell membrane (18–20). Apart from other acetylated residues, ficolin-1 has the unique ability to bind sialic acid to capsular polysaccharides of pathogens including Streptococcus agalactiae, as well as to the surface of immune cells (21, 22).

Ficolin-1 is encoded by the FCN1 gene on chromosome 9q34 and contains nine exons. Among the several SNPs described for the FCN1 gene, at least eight are associated with Ficolin-1 levels, four of them are located in the promoter and one in the first exon (23). Polymorphisms of collectin and ficolin genes have been repeatedly associated with infectious and autoimmune diseases (23–28). FCN1 polymorphisms were associated with increased fatal outcome in patients with systemic inflammation (29), susceptibility to rheumatoid arthritis (28) and leprosy (26). Gene polymorphisms of the lectin pathway have already been associated with RF in case-control association studies with MBL (30–32), Ficolin-2 (33), and MASP-2 (34). Within this context, we are the first to investigate a membrane-bound molecule of the lectin pathway, able to activate the complement. More specifically, we evaluated the association of FCN1 polymorphisms and haplotypes, as well as Ficolin-1 serum levels, with the susceptibility to RF and RHD.

#### MATERIALS AND METHODS

#### Subjects and Samples

This study was approved by the local medical ethics committee (CEP/HC 2658.265/2011-11). All patients and control subjects provided written informed consent in accordance with the Declaration of Helsinki. We investigated a total of 193 patients with a history of RF, all of them with ASO (anti streptolysin O) titers higher than 250 units, characterizing a precedent streptococcal infection and diagnosed according to Jones' modified criteria; 55 (28.5%) males and 138 (71.5%) females; with a mean age of 37 years (range = 7–76 years). Among them, 138 had RHD, confirmed by the transthoracic echocardiogram showing rheumatic involvement of the mitral or aortic valves (**Table 1**), and 55 did not present RHD, but had RF history and were designated as "rheumatic fever only" (RFo) patients. None of the patients presented other inflammatory disease, neoplasia, infective endocarditis, or other infections at the time their blood was collected. Values of high-sensitivity C-reactive protein (hs-CRP) levels, C3 levels and C4 levels, previously published by our group, are shown in **Table 1** (30). The control group included 193 blood donors from The Clinical Hospital of the University Federal of Paraná, with a mean age of 37 years (range = 18–64 years), 68 (35%) males, and 125 (65%) females.

**Abbreviations:** RF, Rheumatic Fever; RHD, Rheumatic Heart Disease; RFo, Rheumatic Fever only; FCN1, ficolin-1 gene; LP, Lectin Pathway; CRP, C-Reactive Protein; MAC, Membrane Attack Complex; GAS, Group A Streptococcus; LD, Linkage Disequilibrium; GlcNAc, N-acetyl-β-D-glucosamine; MBL, Mannosebinding lectin.

#### *FCN1* Genotyping

DNA extraction from peripheral blood was performed using The QIAamp DNA Blood Mini Kit, QIAGEN (Hilden, Germany), following the manufacturer's instructions. The genotype method was adapted from a previously described multiplex PCR-SSP (sequence-specific amplification) method (30). Five FCN1 single nucleotide polymorphisms (SNPs) were genotyped: rs2989727 SNP (c.–1981G > A) in the distal FCN1 promoter, rs10120023 (c.–542G > A), rs17039495 (c.–399G > A), rs10117466 (c.–144C > A), and rs10858293 (c.+33T > G) SNPs in the proximal FCN1 promoter, with primers listed in **Table S1**.

TABLE 1 | Clinical characteristics of patients.


hs-CRP, high-sensitivity C-reactive protein. Normal reference values for hs-CRP: <0.1 mg/dL, for C3: 82-160 mg/dL, for C4: 12-36 mg/dL. \*mean; #Data already published by Schafranski et al. (30).

FCN1\_Prom-1981Af or FCN1\_Prom-1981Gf were conjugated with the FCN1\_Prom\_r reverse primer to generate a fragment of 729 bp. PCR conditions were as follows: 0.2µM of SSP primers, 1X Coral Load PCR buffer (Qiagen, Hilden, Germany), 1.6 mM MgCl<sup>2</sup> (Qiagen, Hilden, Germany), 0.5% glycerol, 0.2 mM deoxyribonucleoside triphosphate (dNTP) (Invitrogen, São Paulo, Brazil), 0.03 U/uL of Taq polymerase (Invitrogen, São Paulo, Brazil), 0.1µg/mL DNA, ultrapure water for 15 µL. The amplification protocol starts with a 5 min denaturation step at 94◦C, followed by 10 cycles of 20 s at 94◦C, 30 s at 60◦C, and 30 s at 72◦C; 10 cycles of 20 s at 94◦C, 30 s at 56◦C, and 30 s at 72◦C; 10 cycles of 20 s at 94◦C, 30 s at 52◦C, and 30 s at 72◦C, concluding with 5 min at 72◦C in the final DNA extension step.

FCN1\_Prom-542Af or FCN1\_Prom-542Gf were conjugated with the FCN1\_Prom-144Ar or FCN1\_Prom-144Cr to generate a fragment of 434 bp. PCR conditions differed from those previously mentioned as follows: 0.6µM of SSP primers, 1.7 mM MgCl2, and 1.5% glycerol. The amplification protocol differed from the previous one only by the annealing primer temperatures, which were 57◦C in the first 10, 55◦C in the next 10, and 53◦C in the last 10 cycles.

FCN1\_Prom-399Af or FCN1\_Prom-399Gf were conjugated with the FCN1\_Prom+33Gr or FCN1\_Prom+33Tr to generate a fragment of 470 bp. PCR conditions differed from those previously noticed, as follows: 0.4µM of SSP primers, 1.75 mM MgCl2, 0.5% glycerol. The amplification protocol differed from the first one only by the annealing primer temperatures, which were 58◦C in the first 10, 56◦C in the next 10 and 54◦C in the final 15 cycles.

Interpretation was based on the electrophoretic pattern of the amplified fragments, on agarose gels 1.5% stained with Sybrsafe (Invitrogen, São Paulo, Brazil). This bispecific PCR-SSP approach allows the identification of 8 haplotypes: <sup>∗</sup> 1 (GGGCG), ∗ 3A (AGGCG), <sup>∗</sup> 3A.3C2.A (AGGCT), <sup>∗</sup> 3A.3C2.B (AGACT),



∗ 3B2 (AAGCG), <sup>∗</sup> 3C1 (AAGCT), <sup>∗</sup> 3C2 (AAGAT), and <sup>∗</sup> 3C2.3A (AGGCG), as previously described (26).

#### Ficolin-1 Measurement

Ficolin-1 serum concentrations were measured in 78 patients and 86 controls using enzyme-linked immunosorbent assay (ELISA) SEA786Hu Cloud-Clone Corp. (Texas, USA).

#### Statistics

Allele, haplotype and genotype frequencies were obtained by direct counting (the phase between distantly situated SNPs could be deduced due to the strong LD between the variants, and were verified with the Expectation-Maximization algorithm implemented in the PLINK software. Exact tests of Guo and Thompson for testing the hypothesis of Hardy-Weinberg equilibrium were accomplished using ARLEQUIN v.3.5.2.2 (http://cmpg.unibe.ch/software/arlequin35/). The investigated polymorphisms were evaluated for regulatory effects on gene expression, using information from The Genotype-Tissue Expression (GTEx) Project. Associations with alleles were tested by the Exact Fisher test using the SISA quantitative skill tables and including genotypes and haplotypes, by multivariate binary logistic regression with the software package STATA v. 9.2. Correction for associated demographic factors (sex, age) and clinical factors were applied in the reduced logistic regression model where sample size was adequate, and p-values ≤ 5% were considered significant. Distributions of ficolin-1 levels were tested for normality with the Shapiro-Wilk test. Since ficolin-1 levels presented a non-normal distribution, the two-tailed Mann-Whitney test was used to compare ficolin-1 levels between groups (GraphPad Prism v.7.03).

# RESULTS

### *FCN1* Alleles and Haplotypes Association With RF

Genotype distribution was in equilibrium with the Hardy-Weinberg model in both controls and patients, except for SNPs rs10120023 (c.−542G > A) and rs10117466 (c.−144C > A) in patients (p = 0.017 and p = 0.024, respectively). A total of eight haplotypes were found in patients and seven, in controls. The phase between the variants in the proximal promoter was determined with bispecific PCR-SSP and the phase of these variants with the SNP c.−1981G > A could be deduced due to a strong linkage disequilibrium (LD). Evidence for recombination was found between the SNP rs10858293 (c.33G > T) in exon 1 and promoter variants in patients (**Figure 1**).

The c.−1981A, c.−542A, c.−144A, and c.33T alleles were associated with an increased level of protection against RF, presenting an additive effect with homozygotes protecting more than heterozygotes. All four are also known to be associated with higher FCN1 expression in adipose subcutaneous tissue. Three of them present the same effect in peripheral blood cells (GTex portal) (35). All the protective alleles occur within the <sup>∗</sup> 3C2 haplotype (AAGAT). As expected, this haplotype was also associated with increased protection against RF under the additive model (OR = 0.41, p < 0.0001). On the other hand,

The values shown in bold correspond to significant values.

the common alleles of these SNPs compose the phylogenetically ancestral <sup>∗</sup> 1 (GGGCG) haplotype, which was associated with increased susceptibility to RF (OR = 1.76, p < 0.0001, **Table 2**).

In contrast, the c.−1981A allele was associated with increased susceptibility to valvar stenosis (5/19 or 26.3% vs. 6/89 or 6.7% A/A homozygotes and 13/19 or 68.4% vs. 39/89 or 43.8% A/G heterozygotes in patients with moderate to severe, vs. light or no valvar stenosis, respectively: OR = 3.75 [95%CI = 1.39–10.15], p = 0.009). Similarly, the c.−144A allele presented an increased susceptibility effect for mitral insufficiency (2/12 or 16.7% vs. 1/75 or 1.3% A/A homozygotes and 1/12 or 8.3% vs. 4/75 or 5.3% A/C heterozygotes in patients with moderate to severe vs. light or no mitral insufficiency, respectively: OR = 3.37 [95%CI = 1.15– 9.92], p = 0.027). Although no other allele or haplotype were associated either with RHD or other clinical manifestation, this may be due to the relatively small sample size of RHD patients in our setting.

#### Ficolin-1 Levels

Ficolin-1 serum levels were lower in patients (median: 800.5 ng/mL [324.6–1,715 ng/mL]), compared to controls (1,208 ng/mL [488–2,852 ng/mL, p < 0.0001), but did not differ between RHD and RFo patient groups (**Figure 2**).

Patients with the <sup>∗</sup> 1 (GGGCG) "susceptibility" haplotype presented lower ficolin-1 levels, than those with the <sup>∗</sup> 3C2 (AAGAT) "protective" haplotype (p = 0.03, medians 770.8 and 975.9 ng/mL, respectively) (**Figure 3**).

### DISCUSSION

The complement system plays an important role both in the defense against GAS infection, as well as in the development of RHD (28, 36, 37). Among studies focusing on complement genetic polymorphisms (30–34), this is the first considering the role of a complement membrane-bound molecule of the lectin pathway in the development of RF and RHD. Our results

indicate that FCN1 polymorphisms may play a dual role in the physiopathology of RF. On one hand, they increase resistance to GAS infection and on the other hand, predispose the patient to RHD symptoms, once the infection is established.

A protective role for FCN1 promoter variants against RF has been observed (**Figure 4**). Among the investigated FCN1 alleles, those four occurring within the <sup>∗</sup> 3C2 haplotype that were associated with RF protection and ficolin-1 levels have also been associated with higher FCN1 gene expression (GTEx Portal) and ficolin-1 serum levels in other studies (23, 26, 29). Indeed, it has been suggested that the minor alleles −542A and −144A may facilitate the binding of transcription factors, causing amplified gene expression (29). Thus, it is conceivable that higher FCN1 gene and protein expression could increase resistance against GAS infection due to ficolin-1 anti-bacterial properties. In fact ficolin-1 is able to bind sialic acid on Group B Streptococcus bacteria (20, 38) as well as GAS's carbohydrate A (GlcNAc) which is a preferable ficolin ligand. GAS recognition by ficolins may culminate in complement activation, despite the described complement evasion mechanisms of the bacteria, e.g., through C5a, as well as C2–C9 cleavage (39). Moreover, it is known that GAS infection occurs through fibronectin binding in the extracellular matrix (39). Fibronectin is also a ligand for ficolin-1 (40), thus competition for fibronectin binding sites may occur. Additionally, ficolin-1 anchors to GPCR43, a G-protein coupled receptor on monocytes. Ficolin-1/GPCR43 activation results in signal transduction through NFkB and interleukin-8 (IL8) gene expression. IL8 is a chemokine that attracts phagocytes to the infection site, enhancing bacterial elimination (20). Taken together, these events corroborate the eventual protective effect of ficolin-1 in the development of RF.

On the other hand, these FCN1 variants associated with high ficolin-1 expression may contribute to excessive complement activation, leading to chronic inflammation and tissue injury, thereby predisposing the patient to RHD symptoms such as valvar stenosis and mitral insufficiency in the advanced phase of the disease. In addition, exposure to neoepitopes on bacteriadamaged valvar tissue would not only induce autoantibody production, but also promote complement activation with MAC deposition, thereby increasing tissue injury. C9, the last MAC complement, was indeed exclusively found in RHD patients with mitral stenosis, compared to control subjects (13). Thus, in the advanced phase of the disease, high ficolin-1 levels would probably not be beneficial. Even though ficolin-1 levels did not differ between RHD and RFo patients, it is possible that the quantification of ficolin-1 in serum does not reflect the levels of membrane-bound ficolin-1, and that the valvar damage could be instead related to this last ficolin-1 form. Keeping this balance—pathogen elimination vs. host preservation, has proven to be very difficult in circumstances where ineffective pathogen elimination induces chronic persistence of the infection with autoimmune features, as is the case of RF/RHD. This would explain other apparently opposite associations of ficolin-1, reported formerly in leprosy for −542A, −144C, and +33T (26), −1981A in rheumatoid arthritis (28) and earlier chronic Pseudomonas aeruginosa colonization in cystic fibrosis patients

complement activation, despite GAS's abilities to cleave C5a, as well as C2-C9; ficolin-1/GPCR43 activation on monocytes, transducing signals through NFkB and activating interleukin-8 (IL8) gene expression, leading to chemoattraction of more phagocytes to the infection site. (C) After established infection, structures on damaged/altered cells recognized by membrane-bound or soluble ficolin-1 mediate complement activation and increases tissue injury, which may become irreversible in the cardiac valves.

binding in the extracellular matrix; recognition of GlcNAc in GAS carbohydrate A by phagocyte membrane-bound ficolin-1 and consequent bacteria internalization;

(27), as well as ficolin-1 deficiency in a mouse model of collagen Ab-induced arthritis (41).

Lower ficolin-1 levels were found in RF patients, an effect that might purely indicate considerable ficolin-1 consumption in RF. In contrast to our study, ficolin-1 serum levels were higher in patients with vasculitis syndrome or rheumatoid arthritis, than in those with myositis, whereas no difference was observed among patients with systemic lupus erythematosus and Behcet's disease (42). Ficolin-1 levels further correlated with several inflammatory markers, including C-reactive protein (CRP), serum amyloid protein (SAP) and complement factor C3 (42) and strongly associated with the severity of ischemic stroke, in another group (43). Interestingly, part of the benefit of intravenous immunoglobulin (IVIG) therapy relies on ficolin-1 pulldown (reported for Kawasaki disease, the most common form of acquired heart disease in childhood) (44). IVIG therapy has also been proven beneficial for Sydenham's Chorea associated with rheumatic fever (45). Moreover, anti-ficolin-1 mAb ameliorated symptoms of collagen antibody-induced arthritis (CAIA) in animal model (42). Additionally, acute injury leads to higher FCN1 gene expression, due to specific regulatory proteins such as hypoxia factor HIF-1a (46). This collection of evidence is in line with our results associating high-FCN1 producing promoter variants with heart damage, in later stages of the disease.

This study has some limitations, especially regarding sample size of individuals with measured ficolin-1 levels which probably affected some of the results. Increased ficolin-1 levels were observed in patients carrying the <sup>∗</sup> 3C2 haplotype, however this effect was not evident among controls possibly due to low sample size. In addition we cannot dismiss the possibility that there may be other causal variants responsible for modulating FCN1 expression, not investigated in this study, such as rs12377780 (in intron 1), rs7857015 (5′ upstream) and rs7858307 (3′UTR) (http://raggr.usc.edu/). Well-designed case control studies including higher number of individuals and different FCN1 gene polymorphisms are necessary to better define the

#### REFERENCES


action of ficolin-1in RF. Concluding, we suggest a role for ficolin-1 in fighting GAS infection, with a possible damaging effect when infection succeeds, due to excessive complement activation. Inhibiting the final steps of complement activation may be a therapeutic clue for preventing valvar damage in patients with RHD.

#### AUTHOR CONTRIBUTIONS

FA, AB, LG, and IM-R designed the study and analyzed the data. SC performed all the assays, analyzed the data, and performed statistical tests. All of the authors contributed toward manuscript preparation and revision, and provided final approval of the version to be published.

#### FUNDING

This work was supported by CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico), Fundação Araucária and BNDES (The National Bank for Economic and Social Development).

#### ACKNOWLEDGMENTS

We gratefully acknowledge all the patients for their participation in this study, and the staff of the Molecular Immunology Laboratory of HC/UFPR for their assistance. We are thankful to CAPES and CNPq for research grants to IM-R, and to BNDES (The National Bank for Economic and Social Development) for its research grants to LG.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.03009/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Catarino, Andrade, Boldt, Guilherme and Messias-Reason. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# IL-6 Receptor Inhibition by Tocilizumab Attenuated Expression of C5a Receptor 1 and 2 in Non-ST-Elevation Myocardial Infarction

Hilde L. Orrem1,2,3, Per H. Nilsson1,2,4,5 , Søren E. Pischke1,2,3, Ola Kleveland6,7 , Arne Yndestad4,8,9,10,11, Karin Ekholt 1,2, Jan K. Damås <sup>12</sup>, Terje Espevik <sup>12</sup>, Bjørn Bendz <sup>13</sup> , Bente Halvorsen4,8,9,10,11, Ida Gregersen8,9, Rune Wiseth6,7, Geir Ø. Andersen11,14,15 , Thor Ueland4,8,9,10,11, Lars Gullestad10,11,13, Pål Aukrust 4,8,9,16, Andreas Barratt-Due1,2,3 and Tom E. Mollnes 1,2,4,12,17,18 \*

#### Edited by:

Uday Kishore, Brunel University London, United Kingdom

#### Reviewed by:

Peter Monk, University of Sheffield, United Kingdom Luis Enrique Munoz, Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany Robert Rieben, Universität Bern, Switzerland

> \*Correspondence: Tom E. Mollnes t.e.mollnes@gmail.com

#### Specialty section:

This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology

Received: 05 May 2018 Accepted: 17 August 2018 Published: 12 September 2018

#### Citation:

Orrem HL, Nilsson PH, Pischke SE, Kleveland O, Yndestad A, Ekholt K, Damås JK, Espevik T, Bendz B, Halvorsen B, Gregersen I, Wiseth R, Andersen GØ, Ueland T, Gullestad L, Aukrust P, Barratt-Due A and Mollnes TE (2018) IL-6 Receptor Inhibition by Tocilizumab Attenuated Expression of C5a Receptor 1 and 2 in Non-ST-Elevation Myocardial Infarction. Front. Immunol. 9:2035. doi: 10.3389/fimmu.2018.02035 <sup>1</sup> Department of Immunology, Oslo University Hospital, Rikshospitalet, Oslo, Norway, <sup>2</sup> University of Oslo, Oslo, Norway, <sup>3</sup> Division of Emergencies and Critical Care, Department of Anesthesiology, Oslo University Hospital, Rikshospitalet, Oslo, Norway, <sup>4</sup> KG Jebsen Inflammation Research Centre, University of Oslo, Oslo, Norway, <sup>5</sup> Linnaeus Centre for Biomaterials Chemistry, Linnaeus University, Kalmar, Sweden, <sup>6</sup> Clinic of Cardiology, St. Olavs Hospital, Trondheim, Norway, <sup>7</sup> Department of Circulation and Medical Imaging, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology, Trondheim, Norway, <sup>8</sup> Research Institute of Internal Medicine, Oslo University Hospital, Rikshospitalet, Oslo, Norway, <sup>9</sup> Faculty of Medicine, University of Oslo, Oslo, Norway, <sup>10</sup> KG Jebsen Center for Cardiac Research, University of Oslo, Oslo, Norway, <sup>11</sup> Center for Heart Failure Research, Oslo University Hospital, Oslo, Norway, <sup>12</sup> Centre of Molecular Inflammation Research, Department of Clinical and Molecular Medicine, Norwegian University of Science and Technology, Trondheim, Norway, <sup>13</sup> Department of Cardiology, Oslo University Hospital, Rikshospitalet, Oslo, Norway, <sup>14</sup> Center for Clinical Heart Research, Oslo University Hospital, Ullevål, Oslo, Norway, <sup>15</sup> Department of Cardiology, Oslo University Hospital, Ullevål, Oslo, Norway, <sup>16</sup> Section of Clinical Immunology and Infectious Diseases, Oslo University Hospital, Oslo, Norway, <sup>17</sup> Research Laboratory, Nordland Hospital, Bodø, Norway, <sup>18</sup> K.G. Jebsen TREC, University of Tromsø, Tromsø, Norway

Background: Elevated interleukin-6 (IL-6) and complement activation are associated with detrimental effects of inflammation in coronary artery disease (CAD). The complement anaphylatoxins C5a and C3a interact with their receptors; the highly inflammatory C5aR1, and the C5aR2 and C3aR. We evaluated the effect of the IL-6 receptor (IL-6R)-antagonist tocilizumab on the expression of the anaphylatoxin receptors in whole blood from non-ST-elevation myocardial infarction (NSTEMI) patients. Separately, anaphylatoxin receptor expression in peripheral blood mononuclear cells (PBMC) from patients with different entities of CAD was investigated.

Materials and Methods: NSTEMI patients were randomized to one dose of tocilizumab (n = 28) or placebo (n = 32) and observed for 6 months. Whole blood samples drawn at inclusion, at day 2, 3 and after 6 months were used for mRNA isolation. Plasma was prepared for analysis of complement activation measured as sC5b-9 by ELISA. Furthermore, patients with different CAD entities comprising stable angina pectoris (SAP, n = 22), non-ST-elevation acute coronary syndrome (NSTE-ACS, n = 21) and ST-elevation myocardial infarction (STEMI, n = 20) were included. PBMC was isolated from blood samples obtained at admission to hospital and mRNA isolated. Anaphylatoxin-receptor-expression was analyzed with qPCR using mRNA from whole blood and PBMC, respectively.

**606**

Results: Our main findings were (i) Tocilizumab decreased C5aR1 and C5aR2 mRNA expression significantly (p < 0.001) and substantially (>50%) at day 2 and 3, whereas C3aR expression was unaffected. (ii) Tocilizumab did not affect complement activation. (iii) In analyzes of different CAD entities, C5aR1 expression was significantly increased in all CAD subgroups compared to controls with the highest level in the STEMI patients (p < 0.001). For C5aR2 and C3aR the expression compared to controls were more moderate with increased expression of C5aR2 in the STEMI group (p < 0.05) and C3aR in the NSTE-ACS group (p < 0.05).

Conclusion: Expression of C5aR1 and C5aR2 in whole blood was significantly attenuated by IL-6R-inhibition in NSTEMI patients. These receptors were significantly upregulated in PBMC CAD patients with particularly high levels of C5aR1 in STEMI patients.

Keywords: complement, C5a receptors, C3a receptor, IL-6, myocardial infarction, inflammation

### INTRODUCTION

Inflammation plays a pivotal role in the pathophysiology of coronary artery disease (CAD) from the establishment of the atherosclerotic plaque through rupture or erosion of the plaque leading to partial or total occlusion of the coronary vessel. This might lead to myocardial necrosis and thereby a myocardial infarction (MI). A total occlusion typically leads to ST-elevation in the electrocardiogram whereas a partial occlusion or an occlusion with collateral circulation doses not show these changes and are classified as unstable coronary syndromes. Unstable coronary syndromes with elevated levels of Troponin T, a marker of myocardial necrosis, are classified as non-ST-elevation MI (non-STEMI) whereas without rice in TnT are classified as non-ST-elevation acute coronary syndromes (1). Rapid restoration of coronary blood flow by re-opening of the occluded coronary vessel with percutaneous coronary intervention (PCI), has considerably improved outcome following MI. However, CAD is still associated with considerable morbidity and mortality (2).

Both the myocardial necrosis and the reperfusion of the infarcted myocardium activate inflammatory mechanisms. Innate and adaptive immune mechanisms are involved in this process and act together to orchestrate a response to damage (3). A balanced inflammatory response is required for proper healing following myocardial infarction (MI), whereas excessive inflammation could give rise to collateral tissue damage with detrimental effects on the myocardium (4). The complement system is an important sensor and effector system of innate immunity and plays a role in all phases of CAD (5). The complement system exerts its main inflammatory functions through proteolytic activation of C3 and C5, which upon cleavage liberate the complement anaphylatoxins C5a and C3a. The anaphylatoxins bind to their respective receptors: the C5a receptor 1 and 2 (C5aR1, C5aR2) and the C3a receptor (C3aR) (6), and the C5a-C5aR1-axis seems to be involved in atherogenesis and CAD (7–9). C5aR inherits an inflammatory role following tissue injury stimulating the release of cytokines like tumor necrosis factor (TNF), interleukin (IL)-1β, IL-6, and chemokines, e.g., IL-8 (10), and induce thrombogenicity by upregulation of tissue factor (11). The effect of activating C5aR2 and C3aR are more diverse and the effect of activating these receptors in the context of acute coronary syndromes (ACS) is at present less clear.

IL-6 and complement may both contribute to the progression of cardiovascular diseases (5, 12, 13) but there are limited data on the interaction between these inflammatory proteins. In a mouse sepsis model, IL-6 inhibition reduced the expression of tissue C5aR (14), but to the best of our knowledge the effects of IL-6 inhibition on the anaphylatoxin receptor expression in human CAD have not been investigated. In a recent study, the IL-6 receptor (IL-6R) antagonist tocilizumab reduced C-reactive protein (CRP) and percutaneous coronary intervention (PCI)-related troponin T (TnT) release in patients with non-ST-elevation myocardial infarction (NSTEMI) (15). In the present study, we aimed to investigate the expression of the anaphylatoxin receptors in a sub-group of this patient cohort (15). Additionally, anaphylatoxin receptor expression was investigated in samples from patients with different entities of CAD before any intervention was initiated.

#### MATERIALS AND METHODS

In this study we included two different patient cohorts: one cohort consisting of NSTEMI-patients randomized to antiinflammatory treatment with an IL-6R antagonist or placebo where blood was sampled from inclusion, before treatment and with repeated measurements, and another cohort consisting of patients with different entities of CAD where blood samples were drawn at hospital admission, before treatment was given.

#### NSTEMI Patients Treated With Tocilizumab

The present work is a sub-study of a previously published double-blind, placebo-controlled two-center study on patients (n = 117) admitted with NSTEMI randomized to treatment with the IL-6R inhibitory monoclonal antibody tocilizumab (n = 58) or placebo (n = 59) (ClinicalTrails.gov, NCT01491074) (15). Tocilizumab was administrated as a single dose of 280 mg immediately prior to coronary angiography. This dose provides a complete IL-6 blockade for approximately 2 weeks (15). Briefly, patients between 18 and 80 years of age with NSTEMI scheduled for coronary angiography were included. Exclusion criteria were clinically significant cardiac disease other than CAD, disease or medication affecting inflammation, contraindications to the treatment drug and clinically unstable patients. Patients were included at a median of 2 days after symptom onset. There were no significant between-group differences in baseline characteristics (15). Fifteen age and sex-matched healthy controls were included. A flow chart describing the whole patient population randomized to tocilizumab or placebo and the number of patients with or without PCI, and with early (≤2 days) vs. late (>2 days) inclusion after symptoms onset is shown in **Figure 1**.

We evaluated the expression of anaphylatoxin receptors (C5aR1, C5aR2, and C3aR) in 60 of the patients treated with tocilizumab (n = 28) or placebo (n = 32). These patients represent all patients included at one of the two study centers (St. Olavs hospital). Due to lack of resources, we only investigated patients from half of the original study population. In this subgroup of patients, there was a significant difference in gender, but no other differences in baseline characteristics were found (**Table 1**). The whole study population (n = 117) was included for plasma complement activation analysis.

### Patients With Various CAD Entities

Three patient groups with different entities of CAD, described in detail elsewhere (16), were examined with respect to anaphylatoxins receptor expression in blood samples obtained at admission to hospital. CAD was defined as coronary artery stenosis >50% verified by coronary angiography. The three patient entities were defined as: (i) stable angina pectoris (SAP) (n = 22), defined as episodes with reversible ischemic chest pain, referred to elective coronary angiography. (ii) Non-ST-elevation acute coronary syndromes (NSTE-ACS) that included unstable angina and NSTEMI patients (n = 21), defined as angina at rest or crescendo angina, referred to urgent coronary angiography within 48 h. (iii) STEMI (n = 20) defined as elevated plasma levels of Troponin T (TnT; at least one value above the 99th percentile) together with ischemic symptoms and ST-segment elevation or new left bundle branch block in the electrocardiogram referred to immediate coronary angiography and PCI if indicated (16). Patients that had malignant or chronic inflammatory diseases, intercurrent infections, or were treated with glucocorticosteroids were not included. Age and sex-matched healthy controls (n = 29) were also included.

#### Blood Sampling Protocol NSTEMI Tocilizumab Study

Blood samples drawn at the time of inclusion, i.e., before study medicine was given and angiography performed, at day 2 and 3 following inclusion and after 6 months were included in this sub-study. Blood was collected in EDTA vacutainer tubes (BD Biosciences, Plymouth, UK), kept on crushed ice and centrifuged within 30 min at 2,500 g for 20 min at 4◦C. Plasma was stored at −80◦C until analyzed, and samples were thawed only once. Whole blood (3 mL) was collected in Tempus Blood RNA tubes (ThermoFischer, Paisley, UK) from patients and healthy controls ensuring immediate lysis of all blood cells and stabilization of RNA. Tempus Blood RNA tubes were stored at −80◦C until RNA preparation.




Data are given as mean with (standard deviation, SD) or median with (25 and 75th percentile, IQR) and number with (%). BL, baseline values; PCI, percutaneous coronary intervention; LMWH, low molecular heparin administered before baseline; CRP, C-reactive protein; R, receptor. Bold value indicate statistical significance.

#### Patients With Different CAD Entities and Healthy Controls

Venous blood was drawn from healthy controls and patients with SAP and NSTE-ACS before angiography. Arterial blood was drawn from the arterial cannula immediately before coronary angiography in patients with STEMI. Peripheral blood mononuclear cells (PBMCs) were isolated from heparinized blood in all three patient groups and the healthy controls by Isopaque-Ficoll (Lymphoprep, FreseniusKabi Norge AS, Oslo, Norway) gradient centrifugation within 1 h after sampling, stored at −80◦C as cell pellets until RNA isolation was performed.

#### RNA Isolation and Quantitative PCR (qPCR) NSTEMI Tocilizumab Study

Whole blood RNA purification was performed by Aaros Applied Biotechnology, Aarhus, Denmark. mRNA from the healthy controls was isolated using Tempus Spin RNA isolation Kit (ThermoFischer, Paisley, UK). cDNA was produced using the high capacity cDNA reverse transcriptase kit (Applied Biosystem, Foster City, CA). TaqMan qPCR primers (FAM-MGB dyelabeled) were purchased from Applied Biosystems for the following genes: C5aR1 (HS00704891), C5aR2 (Hs01933768) and C3aR (Hs0026963). Beta-2-microglobulin (HS 00187842) was stably expressed and used as endogenous control. Each sample was analyzed in triplicate and the reaction was run in 96 well-MicroAmp optical reaction plate on a StepOnePlus system (Applied Biosystems).

#### Patients With Different CAD Entities

RNA from PBMC was isolated using RNeasy Mini Kit (Qiagen, Hilden, Germany). cDNA was synthesized using qScript cDNA SuperMix (Qantabio, Beverly MA). SybrGreen primers were used for qPCR (primer sequences can be given upon request) with GAPDH as endogenous control. Each sample was analyzed in duplicate in 384 well-optical reaction plate on a 7900 HT Fast Real-time PCR system.

#### Complement Activation

Plasma complement activation was evaluated by quantification of the terminal complement complex (TCC) in its soluble form (sC5b-9) using an enzyme-linked immunosorbent assay (ELISA) previously described in detail (17). Briefly, the mAb, aE11, which binds to a neoepitope exposed in C9 when incorporated into the C5b-9 complex, was used as capturing antibody and a biotinylated monoclonal anti-C6 (clone 9C4) was used for detection. The level was related to the International Complement Standard #2, defined to contain 1,000 complement arbitrary units (CAU) per mL (17).

#### Data Presentation and Statistical Analysis

Statistical analysis was performed with IBM SPSS Statistics 24 (Armonk, NY) or Graph Pad Prism, version 7 (San Diego, CA). Differences between two groups were tested with t-test or Mann-Whitney U test when the data were not normally distributed. Differences between more than two groups were tested with ordinary one-way ANOVA or with Kruskal-Wallis test dependent on distribution. Change from baseline was calculated for each time point (e.g., time point-baseline). Longitudinal data were analyzed with Friedman test followed by Wilcoxon signed-rank test to compare the specific time point with baseline levels within each treatment group. To compare differences in categorical data between groups the Chi-square test was used. Correlation analysis was measured by the Spearman correlation test. Bonferroni correction was used to correct for multiple testing. Results are given as median with interquartile range or mean with 95% confidence interval (CI). All tests were two-sided and a p-level of <0.05 was regarded as statistically significant.

#### Ethics

Both studies were approved by the Regional Committee for Medical and Health Research Ethics of South-Eastern Norway and the tocilizumab study also by The Norwegian Medicine Agency and both studies were conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent.

# RESULTS

#### The Effect of IL-6R Inhibition on Anaphylatoxin Receptor Expression in Whole Blood From NSTEMI Patients C5aR1

Expression of C5aR1 was significantly lower in the tocilizumab group compared to the placebo group at day 2 and 3 (**Figure 2A**). Compared to baseline and the healthy controls, the expression of C5aR1 at day 2 and 3 was significantly lower in the tocilizumab group, whereas no difference was observed for the placebo group. After 6 months the expression of C5aR1 in the tocilizumab group was still significantly lower compared to baseline, which was not the case for the placebo group. Compared to healthy controls there was no difference at baseline or after 6 months in any of the two patient groups (**Figure 2A**).

#### C5aR2

Expression of C5aR2 was significantly lower in the tocilizumab group compared to the placebo group at day 2 and 3 (**Figure 2B**). Compared to baseline levels, the expression of C5aR2 was significantly lower in the tocilizumab group at day 2 and 3, whereas no such difference was observed in the placebo group. There were no differences between the two patients groups at baseline or after 6 months. Compared to healthy controls, C5aR2 expression was significantly decreased in the tocilizumab group and the placebo group during the whole study period (**Figure 2B**).

#### C3aR

C3aR expression behaved strictly different from the C5a receptors. There were no differences in receptor expression between the tocilizumab group and the placebo group at any

of the time points (**Figure 2C**). In the tocilizumab group there was a significantly higher expression of C3aR at day 2 and 3 when compared to baseline, whereas both patients groups had significantly higher levels at day 2 and 3 compared to healthy controls (**Figure 2C**). At baseline and after 6 months, there were no differences in C3aR expression between the patient groups and health controls (**Figure 2C**).

### Effects of Coronary Intervention and Time From Symptom Onset to Inclusion on the Expression of Anaphylatoxin Receptors

The effect of tocilizumab could potentially depend on whether the patients were treated with PCI or not, or whether they were included early (≤2 days) or late (>2 days) from the onset of symptoms. However, the pattern of the C5aR1, C5aR2 and C3aR expression was virtually identical in patients with or without PCI (**Figures 3A–C**) and in patients included early or late (**Figures 3D–F**). A flow chart of the patients is shown in **Figure 1**.

### Systemic Complement Activation in NSTEMI Patients

To see if inhibition of IL-6R affected complement activation, sC5b-9 was evaluated in all patients in the tocilizumab study (n = 117) from baseline to day 3, and at 6 months follow up. Plasma concentration of sC5b-9 did not change over time in the NSTEMI patients (**Figure 4A**). Tocilizumab had no effect on the degree of systemic complement activation. The same pattern was seen regardless of PCI treatment or not (**Figure 4B**) and independent of early (≤2 days) or late (>2 days) inclusion from the onset of symptoms (**Figure 4C**).

#### Association Between the Expression of Anaphylatoxin Receptors and Key Biomarkers in the NSTEMI Patients During Hospitalization

The original tocilizumab study found a fall in leukocytes in the tocilizumab-group, primarily caused by a decrease in

FIGURE 3 | Effect of coronary intervention and time of inclusion on the expression of C5aR1, C5aR2, and C3aR in NSTEMI patients. Expression level of complement anaphylatoxin receptors C5aR1 (A,D), C5aR2 (B,E), and C3aR (C,F) in patients with non-ST-elevation myocardial infarction (NSTEMI) receiving placebo (n = 32) or tocilizumab (n = 28) divided into two groups according to percutaneous coronary intervention (PCI) (23 placebo and 21 tocilizumab, gray bars) or not (7 placebo and 9 tocilizumab, white bars) (A–C), and divided into two groups according to inclusion ≤2 days (22 placebo and 15 tocilizumab, gray bars) or >2 days (10 placebo and 13 tocilizumab, white bars) from symptom onset (D–F). Baseline levels show the receptor expression at inclusion, i.e., after hospital admission, before treatment was given. Follow-up time points were day 2 and 3, and 6 months. A group of healthy individuals (n = 15) were included as controls. The qPCR results were quantified using the 2−11CT method, normalized to reference genes and presented as fold change with the healthy controls as calibrator. Data are given as median and 95% CI. \*P < 0.05, \*\*P < 0.01, \*\*\*P < 0.001 vs. healthy controls. †P < 0.05, †††P < 0.001 differences in change from baseline between tocilizumab and placebo. §p < 0.05, §§p < 0.01, §§§p < 0.001 vs. baseline.

neutrophils from baseline to day 3 which led to a significant between-group difference in change from baseline (15). The same statistical differences were also found in the sub-group of patients studied here (**Table 2**). We found no correlation between change in neutrophils and change in expression level of any of the three anaphylatoxin receptors in the treatment group (**Supplementary Table 1**). However, in the placebo group there was a significant correlation between change in neutrophils and change in the expression level of C5aR1 and C5aR2 (**Supplementary Table 1**).

In the original tocilizumab study, IL-6 and sIL-6R increased significantly from baseline to day 3 in the tocilizumabtreated patients (15). A similar pattern was found in the sub-group of patients investigated in this study (**Table 2**). In the tocilizumab group no correlation was found between change in expression for any of the anaphylatoxin receptors and IL-6 in the tocilizumab. sIL-6R correlated with the expression of C3aR in the tocilizumab-treated patients (**Supplementary Table 1**). In the placebo group we found a significant correlation between all three anaphylatoxin receptors and IL6 whereas no correlation was found for sIL-6R (**Supplementary Table 1**).

#### Associations Between the Anaphylatoxin Receptors and CRP and TnT in the NSTEMI Patients During Hospitalization

We evaluated whether AUC for the three anaphylatoxin receptors showed any correlation with AUC for CRP and TnT representing the primary and most important secondary endpoint, respectively, in the original tocilizumab study (15) (**Table 3**). There was a significant correlation between C5aR1 and C3aR, but not C5aR2, and CRP in the placebo group, whereas only C3aR was correlated with CRP in the tocilizumab group (**Table 3**; **Supplementary Figure 1**). TnT correlated significantly with all three receptors in the placebo group, whereas only C5aR1 correlated with TnT in the tocilizumab group (**Table 3**; **Supplementary Figure 1**). All correlations between receptor expression and CRP and TnT were positive. The regression plots, however, show negative regression lines since the statistics were calculated on delta-CT values and a decrease in the delta-CT value represents an increase in receptor expression.

#### Expression of Anaphylatoxin Receptors in PBMC From Patients With Different CAD Entities

In order to explore whether the expression of anaphylatoxin receptors is dependent on the severity of CAD, independent



Data are given as mean with (standard deviation, SD) or median with (25 and 75th percentile, IQR). IL, interleukin; R, receptor; s, soluble.

\*\*\*p < 0.001 comparing differences within group from baseline.

†p < 0.05, †††p < 0.001 comparing between-group differences in change from baseline.

TABLE 3 | Spearman Rho correlation between AUC during hospitalization for CRP and TnT and the three anaphylatoxin receptors in the NSTEMI patients.


AUC, area under the curve; TnT, troponin T; CRP, C-reactive protein; R, receptor.

Data: Spearman Rho correlation coefficient with \*p < 0.05, \*\*p < 0.01. Bold values indicate statistical significance.

on any intervention, we investigated the expression of these receptors in PBMCs in samples obtained from patients admitted to hospital comprising three different entities of CAD: SAP (n = 22), NSTE-ACS (n = 21), and STEMI (n = 20).

Whereas C5aR1 expression was significantly increased in all CAD subgroups compared to healthy controls with the highest levels in the STEMI patients (**Figure 5A**), the increase in C5aR2 and C3aR were more moderate, showing significantly increased levels as compared with controls in the STEMI group (C5aR2) and NSTE-ACS group (C3aR) only (**Figures 5B,C**).

#### DISCUSSION

This study demonstrates for the first time that inhibiting IL-6R profoundly attenuated the expression of C5aR1 and C5aR2 in peripheral whole blood in NSTEMI patients. Treatment with PCI is known to cause a reperfusion injury, which in itself can enhance inflammation. However, the effect on the anaphylatoxin receptor expression seen in this study was independent of treatment with PCI or time between debut of symptoms and inclusion. In contrast, C3aR expression was not affected by the IL-6-inhibitory treatment. Moreover, changes in C5aR1 was significantly correlated with changes in TnT during tocilizumab treatment suggesting the beneficial effect of IL-6R inhibition at least partly could involve downregulation of the inflammatory C5aR1.

Inflammation plays a pivotal role in the wake of a MI being essential for cardiac repair (18). However, sustained and excessive inflammation may contribute to increased tissue damage and is associated with worse prognosis in ACS (19). Elevated levels of inflammatory markers like CRP, IL-6 and C5a are related to the detrimental effects of inflammation in CAD (5, 13, 19–22) and anti-inflammatory treatment is suggested to improve outcome after MI (23). Genetic studies suggest that inhibiting either IL-6 or complement could prove beneficial in patients with CAD (24–26), and it has recently been shown that a single dose of tocilizumab attenuates the increase in CRP and PCI-related TnT release in NSTEMI patients (15).

The activation of C5aR1 induces pro-inflammatory effects like recruitment and activation of inflammatory cells and enhanced cytokine and chemokine production. Experimental studies have shown reduction in infarct size and inflammation when the C5a/C5aR1-axis has been attenuated (27–30). Furthermore, lack of C5aR1 on circulating leukocytes led to reduced infarct size and improved clinical outcome in an in vivo mouse model of MI (31). IL-6 inhibition is previously shown to attenuate expression of anaphylatoxin receptors in an experimental model of sepsis (14). Herein, we show a similar pattern in NSTEMI patients with a significant downregulation of C5Ra1 by tocilizumab in the first days following NSTEMI. Notably, this downregulation was significantly correlated with TnT release in the tocilizumab group suggesting that downregulation of C5aR1 might contribute to the attenuated TnT release by tocilizumab seen in these patients (15). The gradual increase in C5aR1 expression in the different CAD subgroups from SAP through NSTE-ACS with the highest level in STEMI patients may further

support a role for this receptor in plaque progression and destabilization.

C5aR2, previously considered a non-signaling receptor, has been shown to have both pro- and anti-inflammatory effects and its function seems to be dependent on cell type, disease context and species (32). In experimental studies of CAD, there is some evidence that antagonizing C5aR2 might have beneficial effects (9). In the present study we showed a downregulation of C5aR2 by tocilizumab in NSTEMI patients. However, changes in C5aR2 were not correlated with changes in TnT during tocilizumab treatment, and in contrast to C5aR1 expression, the changes in C5aR2 expression in PBMC in the different CAD subgroups were rather modest.

In the NSTEMI patient group, a reduction in C5aR2 expression was observed both in the tocilizumab group and the placebo group, throughout the whole study period. The expression of C5aR2 is known to be attenuated in the context of inflammation (33) and the reduced level of C5aR2 in both the placebo group and the tocilizumab group even observed at inclusion might be due to the inflammatory response caused by the MI itself. We did not find the same reduction in C5aR2 expression compared to controls in the PBMC CAD group. The reason for this is unknown, but might be due to differences in time of sampling in relation to the myocardial injury, different expression level in different cell types or different methods. The effect on C5aR2 seen after 6 months might be related to the enhanced inflammation caused by the reperfusion injury caused by treatment with PCI. Also attenuating IL-6, which is a pleiotrop cytokine, might indirectly change the expression of C5aR2. Thus, the effects of C5aR2 in the setting of CAD and myocardial damage are still unclear and needs further investigations.

C3aR was previously regarded as a pro-inflammatory receptor but recent studies support a more complex effector function for this receptor with anti-inflammatory effects in the acute phase of inflammation by preventing neutrophil mobilization from the bone marrow (34). In an experimental study of intestinal ischemia and reperfusion injury, C3aR was shown to ameliorate ischemia-reperfusion injury in mice (35). Herein, we found a marked increase in C3aR expression in NSTEMI patients that was not modulated by tocilizumab. Moreover, C3aR, but not the two C5a receptors, correlated positively with changes in CRP during IL-6 receptor inhibition. Whatever the effect of C3aR, these findings suggest that IL-6 differently affect the expression of the C5a receptors and C3aR.

There was a reduction in the number of leukocytes and particularly neutrophils in the tocilizumab-treated NSTEMI patients as demented in the original study (15). This could, however, not explain the decreased C5aR1 and C5aR2 expression. First, there was no correlation between the change in receptor expression and change in neutrophil levels in the tocilizumab group. Second, the amount of mRNA in all samples was identical coming mainly from granulocytes, lymphocytes and monocytes, which constitute the main amount of nucleated cells in peripheral blood. Also lymphocytes and monocytes express anaphylatoxin receptors. Lymphocytes have previously been found to express C5aR1 (36–39) and the two C5a-anaphylatoxin receptors are typically co-expressed (33). Monocytes also express all three anaphylatoxin receptors shown for the CAD-population in this study. Third, the decrease was explicitly seen for the C5a receptors and not for the C3aR, indicating that the decrease was selective. Taken together this supports a real reduction in expression of C5aR1 and C5aR2.

Orrem et al. IL-6R Inhibition and Anaphylatoxin Receptors

In the present study, we used whole blood and PBMC, precluding us for detecting individual cell populations as would have been possible using cell sorting. There is, however, an advantage of using whole blood for this purpose, since the cells are less manipulated and in vitro changes in cell activity is reduced and the changes are to a greater extent reflecting the in vivo situation.

No correlation between IL-6 or sIL-6R and the three different anaphylatoxin receptors in the tocilizumab-treated patients, were observed. Tocilizumab was administrated in doses high enough to give a total IL-6 blockade for about 2 weeks (15) thus the level of IL-6 or sIL-6R is rather irrelevant since the effect of the cytokine is totally blocked in all patients during the hospital stay. We did find a correlation between the anaphylatoxin receptors and IL-6 and sIL-6R in the in the placebo group consistent with rather little change in both IL-6 and the anaphylatoxin receptors during the time course in this group.

sC5b-9 did not increase in the present study which most likely was due to the relatively small MIs in the NSTEMI patients. Complement is however constantly activated at a low level and acts in the circulation as a humoral alarm system ready to respond to any danger threatening the host (40). Importantly, the absence of significant systemic complement activation does not preclude the presence of local activation with the ability to act at the site of damage. Thus, downregulation of the receptors for C5a might have beneficial effects both locally and systemically.

The present study has some limitations. The number of patients was rather low. Also, the lack of protein data on the anaphylatoxin receptor expression may weaken our conclusions. Finally, it should be emphasized that correlations do not necessarily mean any causal relationship and more mechanistic studies are needed to further explore the role of anaphylatoxin receptors in CAD.

#### REFERENCES


In conclusion, a substantial and statistically highly significant reduction of C5a receptors was observed in NSTEMI patients treated with tocilizumab, and as for C5aR1, the downregulation correlated with attenuated TnT release. C5aR1 expression in PBMC did also reflect disease severity in another separate CAD population. The cross-talk between complement C5aR1 and IL-6 might contribute to the attenuated TnT release during tocilizumab treatment in these NSTEMI patients.

#### AUTHOR CONTRIBUTIONS

HO, TM, PN, AB-D, PA, BH, OK, JD, BB, RW, LG, AY, TE, and SP contributed to conception and design; OK, LG, GA, BH, HO, IG, and KE contributed with acquisition of data; HO, TM, AB-D, PN, OK, GA, PA, BH, IG, KE, SP, AY, and TU contributed with analysis and interpretation of data; HO, TM, PN, AB-D, OK, and PA drafted the article; All authors critically revised the article and approved the final version.

#### FUNDING

This study was financially supported by The Norwegian Council on Cardiovascular Disease, The Odd Fellow Foundation, The Simon Fougner Hartmann Family Fund and the European Community's Seventh Framework Programme under grant agreement n◦ 602699 (DIREKT).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02035/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer RR declared a past co-authorship with one of the authors TM to the handling editor.

Copyright © 2018 Orrem, Nilsson, Pischke, Kleveland, Yndestad, Ekholt, Damås, Espevik, Bendz, Halvorsen, Gregersen, Wiseth, Andersen, Ueland, Gullestad, Aukrust, Barratt-Due and Mollnes. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Analysis of C3 Gene Variants in Patients With Idiopathic Recurrent Spontaneous Pregnancy Loss

*Frida C. Mohlin1 , Piet Gros2 , Eric Mercier <sup>3</sup> , Jean-Christophe Raymond Gris3 and Anna M. Blom1 \**

*1Department of Translational Medicine, Lund University, Malmö, Sweden, 2Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Department of Chemistry, Utrecht University, Utrecht, Netherlands, 3 Laboratory of Hematology, University Hospital, Nimes, France*

Miscarriage is the most common complication of pregnancy. Approximately 1% of couples trying to conceive will experience recurrent miscarriages, defined as three or more consecutive pregnancy losses and many of these cases remain idiopathic. Complement is implicated both in the physiology and pathology of pregnancy. Therefore, we hypothesized that alterations in the C3 gene could potentially predispose to this disorder. We performed full Sanger sequencing of all exons of C3, in 192 childless women, with at least two miscarriages and without any known risk factors. All exons carrying nonsynonymous alterations found in the patients were then sequenced in a control group of 192 women. None of the identified alterations were significantly associated with the disorder. Thirteen identified non-synonymous alterations (R102G, K155Q, L302P, P314L, Y325H, V326A, S327P, V330I, K633R, R735W, R1591G, G1606D, and S1619R) were expressed recombinantly, upon which C3 expression and secretion were determined. The L302P and S327P were not secreted from the cells, likely due to misfolding and intracellular degradation. Y325H, V326A, V3301I, R1591G, and G1606D yielded approximately half C3 concentration in the cell media compared with wild type (WT). We analyzed the hemolytic activity of the secreted C3 variants by reconstituting C3-depleted serum. In this assay, R1591G had impaired hemolytic activity while majority of remaining mutants instead had increased activity. R1591G also yielded more factor B activation in solution compared with WT. R1591G and G1606D showed impaired degradation by factor I, irrespectively if factor H, CD46, or C4b-binding protein were used as cofactors. These two C3 mutants showed impaired binding of the cofactors and/or factor I. Taken together, several alterations in C3 were identified and some of these affected the secretion and/or the function of the protein, which might contribute to the disorder but the degree of association must be evaluated in larger cohorts.

Keywords: reproductive immunology, miscarriage, complement system, C3, mutation

# INTRODUCTION

Spontaneous pregnancy loss, or miscarriage, is the most common complication of pregnancy and includes all pregnancy losses from conception until 24 weeks of gestation. It is believed that as many as 50% of all conceptions and 15% of clinically recognized pregnancies end up in miscarriage (1). The definition of recurrent spontaneous pregnancy loss (RSPL) differs among international societies.

#### *Edited by:*

*Tom E. Mollnes, University of Oslo, Norway*

#### *Reviewed by:*

*A. Inkeri Lokki, University of Helsinki, Finland Francesco Tedesco, Istituto Auxologico Italiano (IRCCS), Italy*

> *\*Correspondence: Anna M. Blom anna.blom@med.lu.se*

#### *Specialty section:*

*This article was submitted to Molecular Innate Immunity, a section of the journal Frontiers in Immunology*

*Received: 25 April 2018 Accepted: 23 July 2018 Published: 07 August 2018*

#### *Citation:*

*Mohlin FC, Gros P, Mercier E, Gris J-CR and Blom AM (2018) Analysis of C3 Gene Variants in Patients With Idiopathic Recurrent Spontaneous Pregnancy Loss. Front. Immunol. 9:1813. doi: 10.3389/fimmu.2018.01813*

**Abbreviations:** AMD, age-related macular degeneration; C4BP, C4b-binding protein; HEK, human embryonic kidney; RSPL, recurrent spontaneous pregnancy loss; WT, wild type.

The American Society for Reproductive Medicine defines RSPL as two or more pregnancy losses (2), affecting around 5% of couples trying to conceive, while The European Society for Human Reproduction and Embryology defines it as three or more consecutive losses (3), affecting around 1% of couples trying to conceive. Over the years, various etiologies have been identified, such as genetic, structural, autoimmune, endocrine, thrombophilic abnormalities, together with infections. However, approximately 40–50% of the cases still remain idiopathic (1).

Pregnancy is a complex process, where on the one hand significant adaptations of the immune system is needed to tolerate the semiallogenic fetus but on the other hand, an active immune system is needed to defend the mother and fetus from infections. The complement system, which is a pivotal part of our innate immunity, has been shown to play an important role for a successful pregnancy. Studies have shown that too little or too much complement at the wrong time during pregnancy can have devastating consequences for the pregnancy outcome. Too much complement activation in the placenta may lead to placental damage, with a potential risk of fetal loss and therefore complement inhibition in the placenta is of great importance. Complement inhibitors, such as CD46, CD55, and CD59, are expressed in the placenta and on the surface of the trophoblast (4–6), and the critical role of these inhibitors is demonstrated by the embryonic lethality in mice deficient in Crry (7), an important mouse complement inhibitor that resembles human CD46. Previously, we hypothesized that alterations in the complement inhibitors C4b-binding protein (C4BP), CD46, and CD55 might be associated with an increased complement activation, at the trophoblastic maternal interface, leading to pregnancy loss. Several alterations in C4BP and CD46 were identified and some of these caused altered expression and/or function of the proteins (8). However, complement is not only needed for fighting infections during pregnancy but also for proper placental and fetal development. Despite the complement inhibitors presence in the placenta, some degree of complement activation is seen during normal pregnancy. One example of this importance is that mice deficient in C1q, had abnormal development of the placenta, reduced litter size and fetal weight (9). These mice demonstrated a preeclamptic-like phenotype with proteinuria and increased blood pressure, which indicates that C1q plays an important role for a successful pregnancy. Furthermore, C1q in human placenta has also been implicated in preeclampsia (6).

Complement component C3 is the key protein of the complement system and composed of two polypeptide chains, α- and β-chain. The importance of C3 is demonstrated by the high plasma concentration of 1–1.5 mg/ml (10). The cleavage of C3 by C3-convertases is the central reaction of all three complement pathways and results in the biologically active components C3a, an anaphylatoxin and C3b, a potent opsonin. C3b has an exposed thioester bond, allowing C3b to rapidly bind covalently to cell surfaces and promote further complement activation. C3b will participate to form the C5-convertase and in the final step of the complement cascade, the membrane-attack complex is formed, which has the ability to lyse some target cells, e.g., Gram-negative bacteria (11). Activation of complement can be limited by degradation and inactivation of C3b by factor I, in the presence of different cofactors, such as factor H, CD46, and C4BP (12). C3, along with C3b and most importantly iC3b (the inactivated form of C3b), has been shown to be embryotrophic factors. These factors are mainly derived from the uterus and promote embryonic growth prior to the development of the placenta. In humans, the most abundant embryotrophic factor is embryotrophic factor 3 and this was later found to contain mainly C3, C3b, and iC3b (13). It has also been demonstrated that rat C3 on the visceral yolk sac is important for early embryonic development and in explant rat embryo culture, adding intact C3 significantly favors development, without the requirement of C3 activation (14). C3 knockout mice also demonstrate smaller blastocysts, higher resorption rates, and smaller sizes of the placenta (15). In humans, C3 levels was shown to be higher in patients suffering three consecutive miscarriages, compared with women who went on having a live birth after two miscarriages (16). *In vitro* data also suggest that mutations in *FOXD1*, a transcription factor that directly seems to target C3, leading to both decreased and increased C3 expression, were deleterious (17). Genetic variants in the *C3* gene have also been reported to be associated with severe preeclampsia (18). Thus, C3 appears to have an important physiological role in the early phase of pregnancy and in placental development and a subtle fine-tuning of the C3 level is necessary for optimal function.

Due to the role of C3 in the physiology and pathology of pregnancy, we hypothesized that maternal mutations and polymorphisms in C3 might be associated with pregnancy loss. Herein, we performed full Sanger sequencing of all coding exons of *C3* in women experiencing idiopathic RSPL and we found several heterozygous non-synonymous alterations. The C3 mutants were expressed recombinantly and some of the alterations affected the secretion and/or the function of the protein, which might contribute to the disorder.

#### MATERIALS AND METHODS

#### Patients and Controls

Our cohort was previously described in detail (8). Briefly, patients with recurrent pregnancy losses were referred to the Department of Gynecology and Obstetrics or the Hematology Laboratory, University Hospital of Nîmes, France. In total, 1,359 women were pre-selected but the focus was on the 962 most severe cases, defined by at least three consecutive embryonic losses before the 10th gestational week, or two consecutive fetal losses at and beyond the 10th gestational week, all occurring in childless women. These women were screened for classical risk factors for pregnancy loss, such as abnormal parental karyotypes, infectious diseases during pregnancy, uterine anatomical abnormalities, diabetes mellitus, thyroid dysfunction, hyperprolactinaemia prior to luteal phase defects, erythroblastosis fetalis, immune thrombocytopenic purpura, feto-maternal alloimmune thrombocytopenia, and antiphospholipid antibodies. Women were excluded from the study if they had any of the above classical risk factors, previous occurrence of superficial or deep vein thrombosis, were positive for constitutional thrombophilia, preeclampsia, or were of non-Caucasian origin, since this might have introduced consistent confounding heterogeneities in the local frequencies of the polymorphisms and mutations under focus. In the end, 453 women fulfilled all of the criteria and after informed consent was obtained, 429 patients were finally recruited. DNA samples from controls were obtained from women referred to the Department of Gynecology and Obstetrics for a systematic medical exam such as implementation of a new contraception or evaluation of the pelvic floor after pregnancy. Women with no previous pregnancy loss but with at least two uneventful pregnancies were selected (*n* = 261) and similar to cases, these controls were also screened for classical risk factors for pregnancy loss. In total, 224 controls were finally recruited, after informed consent was obtained.

#### DNA Sequencing

All exons of C3 were sequenced in 192 patients by Polymorphic DNA Technologies (Alameda, CA, USA). A nested PCR amplification of the specific exons preceded the Sanger dideoxy sequencing. The primers used are listed in **Table 1**. The identified non-synonymous alterations found in the patients were then analyzed in a control group of 192 women. All patients and controls were randomly chosen from the cohort defined above. There were no statistical differences in baseline characteristics (age, BMI, ethnicity, and risk factors for vascular diseases) between the patients and controls (**Table 2**). The NHLBI Exome Sequencing Project database1 was used for investigating the frequencies of the identified C3 non-synonymous alterations in the general population. The PolyPhen software2 was used for determining the potential effect on the protein function. The sequencing data have been deposited at DDBJ/EMBL/GenBank Targeted Locus Study project under the accession KCDD00000000. The version described in this paper is the first version, KCDD0100000.

#### Proteins

Factor B, factor D, factor H, C3, and properdin were purchased from Complement Technology. Factor I (19), C4BP (20), CD55 (21), and CD46 (21) were expressed recombinantly and purified, as described previously.

#### Expression of C3 Variants

The non-synonymous C3 variants of interest (**Table 3**) were expressed recombinantly. cDNA coding for human C3 was used as template and the mutations were introduced by site directed mutagenesis, according to the manufacturer's instruction (Quik-Change Lightning, Agilent Technologies). The primers used for mutagenesis are listed in **Table 4**. Automated DNA sequencing confirmed the correct mutations and the constructs were cloned into the mammalian expression vector pCEP4 (Thermo Fisher Scientific). The constructs were transiently transfected into either human embryonic kidney-293 adherent cells (ATCC) using Lipofectamine 2000 (Thermo Fisher Scientific) or Freestyle 293-F cells (Thermo Fisher Scientific) using FreeStyle MAX reagent (Thermo Fisher Scientific). To evaluate the expression and secretion levels of the mutants, the C3 concentration in the supernatants and cell lysates after transient transfection in adherent cells, were determined by ELISA (described below). The same volume (5 µl) of these cell supernatants was also analyzed by Western Blotting, both under non-reducing and reducing (25 mM DTT) conditions. A polyclonal goat anti-human C3 antibody (Quidel), followed by a HRP-conjugated rabbit anti-goat antibody (Dako), were used for detection. The blots were developed using enhanced chemiluminescence (Merck-Millipore). The supernatants from the Freestyle 293-F cells were collected, concentrated, C3 concentration determined by the same ELISA and used in the functional assays. Supernatant from cells transfected with empty pCEP vector were treated in the same manner and used as negative control in all assays. For some functional assays, C3b-like molecules were needed, and this was achieved by treatment of the C3 cell media with 0.1 M methylamine hydrochloride pH 8.0, for 1 h at 37°C, followed by a buffer exchange to TBS (Zeba Spin desalting columns, Thermo Fisher Scientific). Methylamine will create C3met, a C3b-like molecule, by breaking the labile thioester bond in C3 and inducing a conformational change.

#### C3 Concentration Determination by ELISA

To determine the C3 concentration after protein expression, an ELISA was performed. Two polyclonal antibodies were used, to eliminate the risk that a monoclonal antibody will recognize the mutants differently. Microtiter plates were coated with rabbit antihuman C3c antibody (Dako), diluted in 75 mM sodium carbonate, pH 9.6. The wells were washed with washing buffer (50 mM Tris–HCl, pH 8.0, 0.15 M NaCl, 0.1% Tween) and blocked with 3% fish gelatin (Norland Products) in washing buffer. The samples were diluted in blocking buffer, supplemented with 10 mM EDTA, and incubated on the plate for 1 h at 37°C. Purified C3 from Complement Technology was used as a standard. C3 was then detected using a goat anti-human C3 antibody (Quidel), followed by a HRP-conjugated rabbit anti-goat antibody (Dako). After 1 h incubation at RT for each of the antibodies, the plates were developed using OPD-tablets (Kem-En-Tec). The absorbance at 490 nm was measured spectrophotometrically.

#### Hemolytic Assay

To assess the hemolytic activity of the C3 variants, a hemolytic assay using C3-depleted serum was performed, essentially as described previously (22). Briefly, antibody-sensitized sheep erythrocytes (Håtunalab AB) was incubated in a V-bottom microtiter plate with 1% C3-depleted serum (Quidel) and four different concentrations (0–2 µg/ml) of either wild type (WT) or mutant C3, in DGVB++ buffer (2.5 mM veronal buffer pH 7.3, containing 70 mM NaCl, 140 mM dextrose, 0.1% porcine gelatin, 1 mM MgCl2, and 0.25 mM CaCl2). The plate was incubated for 1 h at 37°C, during shaking, followed by centrifugation to pellet unlysed erythrocytes. The hemolytic activity was then evaluated by measuring the absorbance, i.e., amount of released hemoglobin, in the supernatant at 405 nm.

#### Degradation of C3 by Factor I

C3b can be degraded by factor I, in the presence of cofactors such as factor H, CD46, and C4BP. C3 cell media treated with methylamine (C3met) were used. In a total volume of 15 µl, 20 ng of C3met, 25 ng factor I, and either 200 ng factor H or 100 ng CD46 were added in TBS. For C4BP, 20 ng C3met, 50 ng

<sup>1</sup>http://evs.gs.washington.edu/EVS (Accessed: 20 February, 2018).

<sup>2</sup>http://genetics.bwh.harvard.edu/pph2/ (Accessed: 20 February, 2018).

#### TABLE 1 | Primers used for sequencing of all C3 exons.a


*a The specific exons were first amplified from genomic DNA using nested PCR. The nested PCR product was subsequently sequenced, using Sanger dideoxy method and the inner primer pairs.*

TABLE 2 | Baseline characteristics of the sequenced patients and controls.a


*a No statistical differences were detected between the two groups, for either of the parameters. The significance was assessed with Mann-Whitney test for the quantitative* 

*data and Chi-Square test for the qualitative data.*

*bQuantitative data: median, interquartile range, and range values.*

*c Qualitative data: numbers and percentages.* factor I, and 4 µg C4BP were mixed. The samples were then incubated for 60 and 90 min, and the reaction was terminated by addition of SDS-PAGE sample buffer, containing 25 mM DTT. The samples were heated at 95°C for 5 min, applied on a 10% SDS-PAGE, and transferred to a PVDF-membrane for Western Blotting. The C3 Western Blot procedure is described above. The intensity of the intact α′-chain of C3met and the 46 kDa degradation product were analyzed, and the results are shown as a ratio of the two.

#### Direct Binding Assays

Binding of C3 to the different ligands factor H, CD46, C4BP, factor I, CD55, factor B, and properdin was assessed in plate assays. The ligands were immobilized on microtiter plates at 5 µg/ml (except factor B which was coated at 10 µg/ml) in PBS and the plates were washed and blocked as described above for the C3 ELISA. The different C3 cell media (mock cell medium was used as negative control), treated with methylamine, were then diluted to four different concentrations in a low salt buffer (10 mM Tris, pH 7.2, 25 mM NaCl, 0.05% Tween 20, 4% BSA) and incubated on the plate for 2 h at 37°C. For binding to factor H and CD46,


*a All mutations were found in heterozygous form. None of the found alterations were significantly associated with the disorder, as determined by Fisher's exact test.*

*bNumbering including signal peptide, Met* = *1.*

*c Polyphen predication of the protein function: http://genetics.bwh.harvard.edu/pph2/.*

*dMAF, minor allelle frequency in NHLBI Exome Sequencing Project database.*

TABLE 4 | Primer sequences (5′–3′) used to introduce site directed mutations in *C3*.


*a Nucleotides corresponding to the changed amino acid residue are underlined.*

C3 variants were tested at 250, 125, 62, and 31 ng/ml; for C4BP, properdin and CD55 at 500, 250, 125, and 62 ng/ml; for factor I at 1,000, 500, 250, and 125 ng/ml; and for factor B at 2,000, 1,000, 500, and 250 ng/ml. The amount of bound C3met was then detected using a rabbit anti-human C3c antibody (Dako), followed by a HRP-conjugated goat anti-rabbit antibody (Dako). Both antibodies were incubated for 1 h at 37°C and developed as described above.

#### Factor B Cleavage

The cleavage of factor B by factor D was determined in solution, in the presence of methylamine-treated C3 WT, R1591G, or mock cell medium as negative control. A mixture of 2 µg/ml factor B, 2 µg/ml C3, and 0.1 µg/ml factor D was prepared in Mg2<sup>+</sup> EGTA buffer (2.5 mM veronal buffer, pH 7.3, 70 mM NaCl, 140 mM glucose, 0.1% gelatin, 7 mM MgCl2, and 10 mM EGTA) and incubated for 0.5, 1, 2, and 4 h at 37°C. At each time point, 10 µl was removed from the mixture, and the reaction stopped by addition of SDS-PAGE sample buffer. The samples were heated at 95°C for 5 min, subjected to SDS-PAGE (10% gel, non-reduced), and transferred to a PVDF-membrane for Western Blotting. Intact factor B, together with the cleavage products Bb (60 kDa) and Ba (33 kDa), were then detected using a polyclonal goat antihuman factor B antibody (Complement Technologies), followed by a HRP-conjugated rabbit anti-goat antibody (Dako). Both antibodies were diluted 1:10,000 and developed using enhanced chemiluminescence (Merck-Millipore). The intensity of the intact factor B band and Bb degradation product were analyzed, and the results are shown as a ratio of the two.

#### Structural Analysis

Structure files were downloaded from the Protein Data Bank (23) and visualized using the molecular graphics program PyMOL.3

#### Statistical Analysis

Fisher's exact test was used to assess statistical association between the found C3 alterations and the disorder. The statistical differences in baseline characteristics between the patient and the control group were assessed with Mann–Whitney test for the quantitative data and Chi-square test for the qualitative data. For all experimental data in the figures, statistical significance was determined using either one-way or two-way ANOVA with Dunnett's multiple comparison test. The results are shown as the mean + SD of at least three independent experiments and values of *p* < 0.05 were considered significant (\**p* < 0.05, \*\**p* < 0.01, and \*\*\**p* < 0.001).

# RESULTS

#### Alterations Identified in *C3*

All exons of *C3* were sequenced in 192 patients with RSPL. The exons carrying identified non-synonymous alterations found in

<sup>3</sup>https://pymol.org/2/ (Accessed: 2 March, 2018).

the patient group were then analyzed in a control group of 192 women. Altogether, 13 heterozygous, non-synonymous alterations were found both in patient and control groups (**Table 3**). Seven novel alterations in C3 were found in total, four in the patient group (L302P, Y325H, R1591G, and G1606D) and three in the control group (V326A, S327P, and V330I). The remaining alterations have been identified previously in other patient cohorts or are common polymorphisms (R102G and P314L). We also identified a novel heterozygous deletion (p.S297SfsX5) in two patients, which results in a frame shift and an introduction of a premature stop codon. This mutant was not expressed recombinantly as it could only yield a 30-kDa truncated fragment of C3 without function. None of the found alterations in the patient group were statistically significantly associated with RPSL, as determined by Fisher's exact test. With the available sample size we had a power of at least 0.80 to detect an association for alleles with risk allele frequencies >0.07, assuming a population prevalence of 0.05, an alpha of 0.05, and a genotype relative risk of 2. Power was calculated using the Genetic Power Calculator (24).

# Expression, Secretion, and Characterization of C3 Variants

To be able to elucidate whether the found C3 variants altered the expression, secretion, or function of the protein, all 13 identified alterations were expressed recombinantly. Transient transfections revealed that two of the mutants (L302P, identified in one patient and S327P, identified in one control) were expressed but not secreted from the cells into the supernatant (**Figure 1A**). Five other alterations (Y325H, V326A, V330I, R1591G, and G1606D) yielded approximately half C3 concentration in the supernatant compared with WT. The supernatants from the transient transfections were also subjected to Western Blotting under both non-reducing (**Figure 1B**) and reducing (**Figure 1C**) conditions. All mutants migrated with the same apparent velocity to 185 kDa for non-reduced C3 and to 110 and 75 kDa for the α- and β-chain, respectively, under reducing conditions.

#### Hemolytic Activity

To assess the hemolytic activity of the C3 variants, C3-depleted serum was reconstituted with C3 WT or variants (or

corresponding volumes of mock medium as negative control), added to antibody-sensitized sheep erythrocytes and the amount of hemoglobin released from lysed erythrocytes was measured spectrophotometrically. Four concentrations of C3 were tested (2, 1, 0.5, and 0.25 µg/ml) in the hemolytic assay but for clarity only the result for 2 µg/ml is shown in **Figure 2**. Several variants showed increased hemolytic activity compared with WT while K633R and R1591G showed instead impaired activity. However, for the K633R mutant, the impaired activity was only clearly observed for the one shown concentration. The other three tested concentrations showed the same hemolytic activity as for WT. The rest of the mutants displayed the same activity for all tested concentrations (data not shown).

either C3 wild type (WT), the different C3 variants or the negative control (corresponding volumes of cell media from mock-transfected cells). This was added to antibody-sensitized sheep erythrocytes and the hemolytic activity, i.e., lysis of the erythrocytes, was determined by measuring the absorbance of the supernatant at 405 nm (free hemoglobin). C3 was added in different concentrations but for clarity only one concentration (2 µg/ml) is shown in the figure. The experiment was performed three independent times in duplicates, and the results are shown as mean + SD. Statistical significance of the differences between WT and mutants was determined using one-way ANOVA with Dunnett's multiple comparisons test, \*\*\**p* < 0.001. The C3 alterations identified only in patients are marked in gray.

#### Factor B Cleavage in Solution for R1591G

Since R1591G showed decreased hemolytic activity, we hypothesized that this mutant might cause more cleavage of factor B by factor D in solution and thereby less C3 will be available for the cell surface. Indeed, in the presence of R1591G, an increased cleavage of factor B in solution was observed, compared with C3 WT (**Figure 3**). This was statistically significant after 4 h cleavage but the trend was visible at all tested time points.

#### Degradation of C3 by Factor I

Degradation assays in fluid phase were performed to study if the different C3 variants could be degraded by factor I, in the presence of the cofactors factor H, CD46, and C4BP. C3 cell media were treated with methylamine to create C3met, a C3b-like molecule, and this was incubated for either 30 min (data not shown) or 90 min (**Figure 4**) with factor I and the different cofactors. Two of the mutants (R1591G and G1606D) showed statistically significantly impaired degradation compared with WT, regardless of which cofactor used. Degradation assays using factor H (**Figures 4A,B**), CD46 (**Figures 4C,D**), and C4BP (**Figures 4E,F**) all displayed the same result. The same trend was also seen at 30 min cleavage (data not shown).

#### Binding Assays

Direct binding assays were performed to elucidate if the C3 variants showed altered binding to factor H, CD46, C4BP, factor I, CD55, factor, and properdin. Corresponding volumes of mock medium was used as negative control. Four different concentrations of C3 were tested in each binding assay but for clarity only one concentration is shown in **Figure 5**. Small differences in binding could be observed. R1591G, that showed both impaired hemolytic activity and impaired degradation by factor I, also showed a somewhat decreased binding to factor H, CD46, and properdin. The other mutant, G1606D, that also showed impaired degradation by factor I, also displayed somewhat impaired binding to factor H, CD46, C4BP, factor I, CD55, and properdin.

factor I, together with the cofactor for 90 min. The degradation of C3met was evaluated by Western Blotting [representative blots can be seen in panels (A,C,E)] and the bands corresponding to intact α′-band of C3met as well as the 46 kDa degradation product were quantified by densitometry. Data in panels (B,D,F) show the ratio between the intensity of these two bands. The negative control shown in the blots is C3 WT at 0 min. Each sample was analyzed in a single replicate, and the data are shown as mean + SD from at least three independent experiments. Statistical significance of the differences between WT and mutants was determined using one-way ANOVA with Dunnett's multiple comparisons test, \**p* < 0.05 and \*\*\**p* < 0.001. The C3 alterations identified only in patients are marked in gray.

#### Structural Analysis

Next, we tried to rationalize the functional consequences of the identified and studied alterations in the context of the known C3 (pdb-id: 2A73) (25) and C3b (pdb-id: 5FO7) (26) structures.

All C3 alterations analyzed in this study were mapped onto both the C3 (**Figure 6A**) and the C3b structure (**Figure 6B**). R102G is the common polymorphism also known as C3 S/F (slow/fast) and located in the macroglobulin (MG) 1 domain. Mutants K155Q, located in the MG2 domain; K633R in the LNK domain; R735W in ANA/C3a domain and S1619R in the C345C (CTC) domain, all showed normal expression and normal or only slightly altered function.

Mutants L302P, P314L, Y325H, V326A, S327P, and V330I are all located in the MG3 domain of C3. For two of the mutants, L302P and S327P, C3 was not secreted from the cells after transient transfections. In these mutants, leucine 302 and serine 327 are changed to proline residues, which most probably will cause misfolding and degradation of the protein intracellularly. The others are relatively structurally subtle mutations, changing a proline into leucine, a tyrosine into histidine, a valine into an alanine, and a valine into a leucine in MG3. Structurally, MG3 plays a critical role in the conversion from C3 to C3b and its contacts with MG7 and MG8 are changed drastically when converting C3 into C3b by removal of the anaphylatoxin (ANA/C3a) domain. Apparently, the minor alterations in the MG3 domain affected the secretion

FIGURE 5 | Binding of factor H, CD46, C4b-binding protein (C4BP), factor I, CD55, factor B, and properdin to C3met. Microtiter plates were coated with the different ligands for C3b: (A) factor H, (B) CD46, (C) C4BP, (D) factor I, (E) CD55, (F) factor B, and (G) properdin. C3met, either WT, variants or corresponding volumes of cell media from mock-transfected cells as a negative control, was added to the coated plates and the ability of the variants to bind the different ligands was detected using a polyclonal anti-human C3 antibody. Four different concentrations of C3met were tested but for clarity only one concentration for each ligand is shown in the figure (125 ng/ml for factor H and CD46; 250 ng/ml for C4BP, properdin, and CD55; 1 µg/ml for factor B; and 500 ng/ml for factor I). Each concentration was analyzed in duplicates, and the data are shown as mean + SD from three independent experiments. One-way ANOVA with Dunnett's multiple comparisons test was used for determining the statistical significance of the differences between WT and mutants, \**p* < 0.05, \*\**p* < 0.01, and \*\*\**p* < 0.001. The C3 alterations identified only in patients are marked in gray.

(C) C3 (pdb-id: 2A73) and (D) C3b (pdb-id: 5FO7). Both C3 and C3b are colored in cyan with indicated in purple the neck region (res. 1494–1518), connecting the CTC domains to the C3 and C3b body. R1591 is indicated by spheres and disulfide bridge Cys1518–Cys1590 shown in sticks. (E) For G1606D; the structure of C3b—mini-factor H—factor I (pdb-id: 5O32) is shown. The glycine C[alpha] atom is indicated by a red sphere. Mutation G1606D likely induces local rearrangement affecting its subsequent loop (res. 1617–1618, highlighted in purple) that is directly involved in contacting factor I.

of the protein (all except P314L showed decreased secretion) and also resulted in an increased hemolytic activity.

Mutant R1591G showed impaired hemolytic activity but instead yielded increased cleavage of factor B in the solution. In C3, R1591 is involved in interactions with residues of the "neck" region (res. 1494–1517) in native C3. Mutation R1591G possibly destabilizes these interactions and thereby alters or destabilizes the MG7-MG8 domain arrangement (**Figures 6C,D**). Destabilization of the MG7-MG8 arrangement in C3 is likely to induce generation of C3(H2O) more readily, leaving less native C3 available for hemolytic activity. Moreover, enhanced generation of C3(H2O) also explains the observed increased cleavage of factor B, due to soluble C3 convertase, i.e., C3(H2O)Bb, generation.

Both R1591G and G1606D showed impaired degradation by factor I, independently of which cofactor that was used. Both mutations R1591 and G1606 are located in the C-terminal CTC domain. Interactions between the C3b-CTC domain and the heavy chain of factor I (pdb-id: 5O32) are needed to bind factor I and activate its catalytic serine–protease domain for cleaving C3b (26). Mutation G1606D possibly induces local rearrangements that disturb neighboring contact sites required for factor I binding (**Figure 6E**). R1591G, however, is positioned opposite of the factor I binding interface on the CTC domain. Putatively, a distortion of the "neck" conformation mis-orients the CTC domain for proper factor I binding needed for inducing catalytic activity.

#### DISCUSSION

Miscarriage is unfortunately a common complication of pregnancy and RSPL affects around 1–5% of couples trying to conceive. Evidence-based treatments, e.g., aspirin and heparin treatment for antiphospholipid syndrome and surgical corrections for uterine abnormalities, have improved the outcome for couples over the years but 40–50% of the cases still remain idiopathic. The complement system has been shown to be involved both in the physiology and the pathology of pregnancy. Furthermore, C3—the central protein of the complement cascade—appears to have an important role in the early phase of pregnancy and in the development of the placenta. In this study, we therefore set out to determine whether maternal variations in C3 are associated with idiopathic RSPL.

We performed full Sanger sequencing of all exons of the *C3* gene, in 192 patients with RSPL. The exons carrying identified alterations were then analyzed in a control group of 192 women and a total of 13 heterozygous non-synonymous alterations were found in both groups. Four of these (L302P, Y325H, R1591G, and G1606D) were novel mutations only found in the patient group and three (V326A, S327P, and V330I) were novel ones only found in the control group. We found the two common C3 polymorphisms in many of both the patients and controls; R102G (27) and P314L (28), which both have been associated with age-related macular degeneration (AMD) in many studies (29–31). R102G is also known as the C3S/F (slow/ fast) allele and has been associated not only with AMD but also other immune-mediated disorders, such as IgA nephropathy, systemic vasculitis, partial lipodystrophy, and membranoproliferative glomerulonephritis type II (32). K155Q is a rare mutation, found previously in AMD patients (33) and we found this mutation in one RSPL patient as well. We also identified K633R in two RSPL patients, which has previously been identified to be a rare variant in atypical hemolytic uremic syndrome patients (34). Both R735W and S1619R were each found in one patient and one control and the former has previously been reported before in atypical hemolytic uremic syndrome (35) and the latter in AMD (36). None of the C3 alterations identified in the patients were associated with RSPL. However, we used a relatively small cohort and several of the found mutations are rare variants, which makes it difficult to appropriately statistically analyze the association between rare variants and disease. To be able to conclude a significant association, replication in larger cohorts is needed.

We decided to recombinantly express and analyze all our found alterations, also those only found in the control group. The women in the control group were selected on the basis of no previous pregnancy loss, at least two uneventful pregnancies and no known classical risk factors for pregnancy loss. However, they could potentially have other, unknown medical conditions, not related to pregnancy loss. Transient transfections of all C3 variants revealed that two mutants, L302P (identified in one patient) and S327P (identified in one control), were not secreted from the cells. These proline substitutions will most probably misfold the protein, leading to intracellular degradation. However, it should be noted that all identified C3 alterations in this study were in heterozygous form, which means that the individuals with the L302P and S327P alterations could theoretically have C3 concentration in blood decreased by 50%. We could unfortunately not confirm this, due to lack of blood samples from these individuals. Functional testing of the mutants that were secreted revealed that several of the alterations had increased hemolytic activity compared with C3 WT, e.g., R102G and P314L, which is in accordance with previous reports (37, 38). However, in that previous report, the authors also observed a 1.3-fold lower binding affinity for R102G to factor H, something we did not observe. The reason for this could be the use of different methods; we detected factor H binding using a microtiter plate assay and the former report used surface plasmon resonance. The K155Q mutant was previously reported in AMD (39) with impaired degradation of this mutant by factor I, in the presence of factor H as a cofactor. When using CD46 as a cofactor, the activity was normal. However, in our degradation assay, this mutant showed normal activity, for both the cofactors factor H and CD46. Furthermore, we did not see any altered binding for K155Q to factor H in our microtiter plate assay, which is in accordance with Seddon et al. However, Seddon et al. did see a decreased binding of K155Q to factor H using surface plasmon resonance. The only difference we observed for this mutant was an increased hemolytic activity. R735W was previously tested functionally, for its binding to CD46, factor H, factor B, and soluble CR1, together with its degradation by factor I, using CD46 as the cofactor (35). In that report, no functional changes were observed. In our report, we could see a slightly impaired degradation by factor I, when using C4BP as a cofactor but in accordance with the previous report, the activity was normal when using CD46 as the cofactor. We could also see a somewhat decreased binding to factor H and CD46, together with an increased binding to factor B. However, these changes were only minor.

Two novel mutations, importantly only found in the patient group, displayed the most functional impairment and also decreased secretion of the protein into the cell media after transient transfection: R1591G and G1606D. Both of these mutations showed impaired degradation by factor I, independently of which cofactor that was used (factor H, CD46, or C4BP). Binding of R1591G to factor H and CD46 was slightly impaired but we could not observe any impaired binding to factor I in a microtiter plate assay. However, to become active, factor I needs to bind to C3b in a way that it stabilizes the conformation of its catalytic domain. R1591G is positioned opposite of the factor I binding interface on the CTC domain. A distortion of the "neck" conformation could mis-orient the CTC domain for proper factor I binding needed for inducing catalytic activity. R1591G also displayed a decreased hemolytic activity but instead an increased cleavage of factor B in the solution. This might lead to consumption of C3 in the solution, resulting in less activity on the cell surface. Structural analysis of this mutant indicated that R1591G is involved in interactions with the neck region of native C3 and could possibly destabilize that region. This could also lead to more C3(H2O) generation and subsequently less native C3 available for hemolytic activity. The other mutant, G1606D, also showed impaired degradation by factor I. Structural analysis of this mutant revealed that this amino acid change could induce a significant local rearrangement, which might affect the residues that are involved in binding to factor I. We did indeed observe a slightly impaired binding of this mutant to factor I, together with impaired binding to factor H, CD46 and C4BP, which could all together result in the impaired degradation we see by factor I.

Taken together, several C3 alterations were identified both in patients and controls. Since we used a relatively small cohort, none of the found alterations in the patient group were associated with RSPL in a statistically significant manner. To be able to determine an association with the disease, larger cohorts need to be analyzed. Two mutants (L302P, found in one patient and S327P, found in one control) were not secreted from the cells, probably due to misfolding and two others (R1591G and G1606D), found in patients, showed both decreased secretion and functional impairment. Some of the identified C3 alterations might result in *in vivo* consequences and contribute to RPSL.

#### REFERENCES


#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations and approval by the University Hospital of Nîmes Institutional Review Board and ethics committee (ref# 2001-12-07). All subjects gave written informed consent in accordance with the Declaration of Helsinki.

#### AUTHOR CONTRIBUTIONS

AB and J-CG designed the study. J-CG and EM provided the patient and control samples, together with the clinical data. FM performed the functional experiments. PG performed the structural analysis. FM, AB, PG, and J-CG wrote the manuscript, and all authors approved it.

#### ACKNOWLEDGMENTS

We would like to thank all the patients and controls for their contribution to this study. We would also like to thank Emma Ahlqvist at Lund University for her kind help with power calculations. This work was supported by the Swedish Research Council (2016-01142) and a grant for clinical research (ALF).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Mohlin, Gros, Mercier, Gris and Blom. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*