# PHAGOCYTOSIS: MOLECULAR MECHANISMS AND PHYSIOLOGICAL IMPLICATIONS

EDITED BY : Esther M. Lafuente, Florence Niedergang and Carlos Rosales 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 978-2-88966-147-3 DOI 10.3389/978-2-88966-147-3

#### 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

# PHAGOCYTOSIS: MOLECULAR MECHANISMS AND PHYSIOLOGICAL IMPLICATIONS

Topic Editors:

Esther M. Lafuente, Complutense University of Madrid, Spain Florence Niedergang, Centre National de la Recherche Scientifique (CNRS), France Carlos Rosales, Universidad Nacional Autónoma de México, Mexico

Citation: Lafuente, E. M., Niedergang, F., Rosales, C., eds. (2020). Phagocytosis: Molecular Mechanisms and Physiological Implications. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88966-147-3

# Table of Contents

*05 Editorial: Phagocytosis: Molecular Mechanisms and Physiological Implications*

Esther M. Lafuente, Florence Niedergang and Carlos Rosales

*08 The Complement System is Essential for the Phagocytosis of Mesenchymal Stromal Cells by Monocytes*

Caroline Gavin, Stephan Meinke, Nina Heldring, Kathleen Anne Heck, Adnane Achour, Ellen Iacobaeus, Petter Höglund, Katarina Le Blanc and Nadir Kadri

*21 Calreticulin and Galectin-3 Opsonise Bacteria for Phagocytosis by Microglia*

Tom O. J. Cockram, Mar Puigdellívol and Guy C. Brown

*33 Thioredoxin-Interacting Protein Promotes Phagosomal Acidification Upon Exposure to* Escherichia coli *Through Inflammasome-Mediated Caspase-1 Activation in Macrophages*

Sung-Jin Yoon, Dong Hyun Jo, Seung-Ho Park, Jun-Young Park, Yoo-Kyung Lee, Moo-Seung Lee, Jeong-Ki Min, Haiyoung Jung, Tae-Don Kim, Suk Ran Yoon, Su Wol Chung, Jeong Hun Kim, Inpyo Choi and Young-Jun Park


Ayesha Murshid, Thiago J. Borges, Cristina Bonorino, Benjamin J. Lang and Stuart K. Calderwood


Patrizia Bonsignore, Johannes W. P. Kuiper, Jonas Adrian, Griseldis Goob and Christof R. Hauck


Eileen Uribe-Querol and Carlos Rosales


Deborah S. Lew, Francesca Mazzoni and Silvia C. Finnemann

# Editorial: Phagocytosis: Molecular Mechanisms and Physiological Implications

Esther M. Lafuente<sup>1</sup> \*, Florence Niedergang<sup>2</sup> \* and Carlos Rosales <sup>3</sup> \*

<sup>1</sup> Department of Immunology, Ophthalmology and Otorhinolaryngology, School of Medicine, Universidad Complutense de Madrid, Madrid, Spain, <sup>2</sup> Université de Paris, Institut National de la Recherche Médicale, U1016, Centre National de la Recherche Scientifique, UMR8104, Paris, France, <sup>3</sup> Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, Mexico City, Mexico

Keywords: phagocytosis, phagosome, CR3, necrotic debris, efferocytosis, macrophages, microglia, opsonins

**Editorial on the Research Topic**

#### **Phagocytosis: Molecular Mechanisms and Physiological Implications**

Phagocytosis is a conserved cellular process for ingesting and eliminating large (≥0.5µm) particles, including microorganisms, foreign substances, and apoptotic cells. Phagocytosis is performed by many cell types and it constitutes an essential process for tissue homeostasis. It was long considered that only specialized cells, and in particular cells of the immune system, termed professional phagocytes (macrophages, neutrophils, monocytes, dendritic cells, microglia, osteoclasts), accomplish phagocytosis with high efficiency. Other cells, such as retinal pigment epithelial cells or hepatocytes, however, are also potent phagocytes. This Research Topic provides a timely overview of the biological importance of phagocytosis for immune-related functions and for tissue homeostasis, with focus on phagocytosis of apoptotic bodies and necrotic debris that could activate undesirable inflammatory responses. This article collection goes through the latest insights on the mechanisms controlling phagocytic receptor activation, particle engulfment, and phagosome maturation. The review by Uribe-Querol and Rosales provides a bird's-eye view of multiple aspects of the process of phagocytosis. It describes the types of phagocytosis receptors known today and the phases involved in phagocytosis, including (i) detection of the particle to be ingested, (ii) activation of the internalization process, (iii) formation of a specialized vacuole called phagosome, and (iv) maturation of the phagosome to transform it into a phagolysosome. Once the process of phagocytosis is activated, the phagocyte requires profound reorganization of its cell morphology around the target in a controlled manner. This process is limited by biophysical constraints involving the receptors, the membrane, and the actin cytoskeleton. The article by Jaumouillé and Waterman discusses the major physical constraints involved in the particle internalization resulting in the formation of a phagosome and provides an extensive review of the underlying molecular mechanisms coping with these constraints. They focus on the two most-studied types phagocytic receptors, the Fcγ receptors and the complement receptor 3 (αMβ<sup>2</sup> integrin). They revise the importance of receptor accessibility and diffusion at the membrane, the protrusive forces generated by actin polymerization and describe a molecular clutch for the phagocytic integrin αMβ<sup>2</sup> involved in mechanosensing. Integrin phagocytic receptors initiate phagocytosis by a process that seems, at least initially, different from antibody-mediated phagocytosis. The review by Torres-Gomez et al. describes the phagocytic integrins and the molecular events controlling integrin activation and the downstream signaling driving particle engulfment. Authors focus on complement receptor (CR) 3/integrin αMβ2, and on CR4/integrin αXβ2. In addition, they briefly mention other

#### Edited and reviewed by:

Francesca Granucci, University of Milano-Bicocca, Italy

#### \*Correspondence:

Esther M. Lafuente melafuente@med.ucm.es Florence Niedergang florence.niedergang@inserm.fr Carlos Rosales carosal@biomedicas.unam.mx

#### Specialty section:

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

Received: 24 July 2020 Accepted: 13 August 2020 Published: 11 September 2020

#### Citation:

Lafuente EM, Niedergang F and Rosales C (2020) Editorial: Phagocytosis: Molecular Mechanisms and Physiological Implications. Front. Immunol. 11:586918. doi: 10.3389/fimmu.2020.586918

**5**

integrins that do not bind complement, but that also function as phagocytic receptors, such as αVβ<sup>5</sup> and αVβ<sup>3</sup> integrins.

Phagocytosis is more efficient when foreign particles are labeled for phagocytosis by opsonins, which are host-derived proteins that bind specific receptors on phagocytic cells. Important opsonins promoting efficient phagocytosis include antibody (IgG) molecules and complement components. Other opsonins are not so well-known. In the article by Cockram et al. two novel opsonins for bacteria calreticulin and galectin-3, are described. Microglia, the brain-resident macrophages, release calreticulin and galectin-3 when activated by bacterial lipopolysaccharide. Both lectins can bind lipopolysaccharide on the surface of bacteria and are also recognized by the phagocytic receptors LRP1 and MerTK, respectively. These findings provide insight on how this innate immune response of microglia may promote clearance of bacteria in the brain.

Foreign particles can also be recognized directly by phagocytes. This is accomplished by non-opsonic receptors that directly bind pathogen-associated molecular patters (PAMPs) and can induce phagocytosis. Members of the family of C-type lectin receptors, including Dectin-1 (dendritic cellassociated C-type lectin-1), Mincle (macrophage-inducible C-type lectin), and DC-SIGN (dendritic cell-specific ICAM-3-grabbing non-integrin), as well as the mannose receptor have been described as important phagocytic receptors. In the review by Bonsignore et al., the carcinoembryonic antigen-related cell adhesion molecule 3 (CEACAM3) is presented as a novel and important non-opsonic phagocytic receptor of highly specialized, host-restricted bacteria. Authors also discuss the importance of CEACAM3 polymorphisms for human innate immunity against bacteria through phagocytosis.

After the particle has been internalized in a phagosome, this novel organelle suffers dramatic changes both in its membrane composition and in its contents by fusing with other intracellular vesicles. This process known as phagosome maturation, can be determined by phosphoinositide (PI) lipids, which are pivotal determinants of organelle identity, membrane dynamics, and vesicle trafficking. Yet, some microbial pathogens can modify the phosphoinositide pattern in phagocytic cells for their own benefit. The article by Leoni Swart and Hilbi describes how Legionella pneumophila, the causative agent of a severe pneumonia called Legionnaires' disease, alters the phosphoinositide pattern of the phagosome to create a new vesicle known as the Legionella-containing vacuole (LCV), where the bacterium can live and replicate inside lung macrophages. The authors summarize strategies, by which L. pneumophila subverts macrophage phosphoinositide lipids, particularly the important conversion from phosphatidylinositol (3) phosphate to phosphatidylinositol (4) phosphate, to promote LCV formation and intracellular replication. As mentioned, another important aspect of phagosome maturation is the change of its contents into a potent microbicidal environment. Acidification the phagosome is essential for achieving activation of many microbicidal enzymes in the phagosome. The article by Yoon et al. presents a novel mechanism by which macrophages regulate phagosome acidification upon exposure to Gram-negative bacteria. Evidence is provided on how the thioredoxin-interacting protein (TXNIP)-associated inflammasome plays a role in pH modulation through the activated caspase-1-mediated inhibition of NADPH oxidase. This acidification pathway is relevant for controlling bacterial clearance by macrophages.

Concomitant with phagocytosis, primed phagocytes perform additional functions that contribute to eliminate the ingested pathogen and to regulate inflammation. The article by Acharya et al. describes that murine macrophages increase their levels of tumor necrosis alpha (TNF-α), interleukin (IL)-1β, IL-6, and matrix metalloproteinase 9 (MMP-9) during complementmediated phagocytosis. Authors also provide evidence that this upregulation of proinflammatory mediators downstream of CRs may be dependent on calpain-mediated inhibition of IκBα and NF-κB activation. Another cell response associated to phagocytosis is the generation of extracellular vesicles with antibacterial properties (aEV). The article by Lörincz et al. describes that complement receptor CR3 is decisive for both processes in human neutrophils. By using bone marrow derived neutrophils from genetically modified mice, they confirm that CR3 is involved in both processes. Although this receptor plays a critical role in both responses, authors also present evidence for phagocytosis and biogenesis of antibacterial extracellular vesicles to be independent processes regulated by different signaling pathways.

Phagocytosis is not only an important process for eliminating microorganisms but also for eliminating damaged and apoptotic cells. In this manner phagocytosis contributes to maintain tissue homeostasis. Detection of apoptotic cells requires particular receptors for molecules that only appear on the membrane of dying cells. These molecules include lysophosphatidylcholine, and phosphatidyl serine (PS) and they deliver to phagocytes an "eat me" signal. Because some normal cells, including activated B and T cells, may express significant levels of PS on their surface, other molecules, for example CD47, function as "don't eat me" signals. Recognition and elimination of apoptotic cells by macrophages, a process known as efferocytosis, is also essential for the resolution of inflammation. The review by Kourtzelis et al., presents our current understanding about the mechanisms regulating macrophage efferocytosis during resolution of inflammation. In special tissues, such as the brain, phagocytosis also plays an essential role in eliminating dead cells and protein aggregates. A particular type of efficient phagocytes, the microglia, is responsible for this function. In the review by Márquez-Ropero et al., the differences in the origin, lineage and population maintenance of microglia and macrophages is presented. They discuss the principles that govern efferocytosis, and describe the epigenetic, transcriptional, and metabolic rewiring by microglia and its implication in trained immunity and brain function.

Most of the studies on the clearance of cellular debris have been strongly biased toward the removal of apoptotic bodies. As a result, the mechanisms underlying the removal of necrotic cells have remained relatively unexplored. In the review by Westman et al., our emergent understanding of the phagocytosis of necrotic debris is presented. They explain the process from recognition of necrotic cells to their internalization and disposal by phagocytes. The review describes the classes of "find me" and "eat me" signals presented by necrotic cells and their cognate receptors in phagocytes, which in most cases differ from the extensively characterized counterparts in apoptotic cell phagocytosis. An example of the importance of phagocytosis of necrotic and cell debris is found in the disease retinitis pigmentosa, characterized by progressive loss of retinal photoreceptors, resulting in blindness. A special form of this disease is caused by mutations in the clearance phagocytic receptor Mer tyrosine kinase (MerTK) and the failure of retinal pigment epithelial cells to eliminate photoreceptor outer segment debris during diurnal phagocytosis. Now, a novel role for microglia in the onset of retinitis pigmentosa is presented in the paper by Lew et al.. They describe how loss of the phagocytic receptor MerTK causes microglia activation and relocalization in the retina, and that microglia activities accelerate loss of photoreceptors. These findings suggest that therapies targeting microglia may slow down the development of this blinding disease. Another example of phagocytosis eliminating necrotic cells and facilitating uptake of protein antigens is presented in the article by Murshid et al.. They describe how cells undergoing necrosis can release heat shock proteins (HSP) a highly abundant class of molecular chaperones that when released can couple to client binding proteins in the extracellular milieu. HSP can be recognized through the scavenger receptors LOX-1 (class E member oxidized low-density lipoprotein receptor-1) and SCARF1 (scavenger receptor class F member 1), expressed by dendritic cells (DC) and macrophages activating inflammatory pathways and innate immunity. Additionally, these receptors facilitate uptake of HSP and coupled antigens, and their delivery to the proteasome, leading to antigen processing, cross presentation, and stimulation of adaptive immunity.

Elimination of altered, but not dying cells, is also an essential aspect of tissue homeostasis. Several cancer cells can be recognized and eliminated through phagocytosis. Yet, some tumor cells can also block this response by expressing the antiphagocytic (don't eat me") CD47 molecule. In the case of the non-small-cell lung cancer (NSCLC) it has been observed that some tumors develop tyrosine kinase inhibitor (TKI) resistance in patients with EGFR-mutant-mediated NSCLC. In the paper by Nigro et al., a transcriptomic analysis of NSCLC cell lines revealed selective overexpression of CD47 in patients carrying EGFR mutations. Also, CD47 expression became up-regulated following in vitro TKI resistance development. By inhibiting the CD47 protein using a specific monoclonal antibody, the clearance of EGFR-TKI resistant cells by phagocytes is increased. Thus, this report supports that some tumors enhance CD47 expression to avoid phagocytosis, and that CD47 neutralization by specific monoclonal antibody may be a promising immunotherapeutic option for resistant EGFR-mutant NSCLC. An interesting example of elimination of foreign cells through phagocytosis is also presented in the paper by Gavin et al.. A promising tool in the treatment of chronic inflammatory diseases is the administration of mesenchymal stromal cell (MSC). The outcomes of this therapy have been ascribed to the capacity of MSC to release a large variety of immune-modulatory factors. However, after administration most of the infused MSC are undetectable in the circulation within hours. In this report, authors found that upon contact with blood, complement proteins were deposited on MSC and complement opsonization of MSC enhanced their phagocytosis by monocytes. These findings imply that at least some of the MSC immune-modulatory effects could be mediated by monocytes that have phagocytosed them.

Phagocytosis is a complex process performed by many cell types. Much of our knowledge comes from studies of model professional phagocytes. However, one must bear in mind that not all phagocytes behave the same. An important example illustrating this notion, is presented in the article by Zajd et al.. They remind us that macrophages are a heterogeneous and plastic population of cells whose phenotype changes in response to their environment, and they explore the functional differences between peritoneal (pMAC) and bone marrow-derived macrophages (BMDM). Contrary to their hypothesis, authors found that BMDM were generally more responsive and poised to respond to their environment than pMAC. These findings are relevant for future studies, since many times these two types of phagocytes were considered to behave similarly. Finally, the review by Davies et al., describes some properties of a non-professional phagocyte, the hepatocyte, which is very important in drug detoxification and immunity. Liver function can have profound effects in metabolism, inflammation and cancer. In addition to phagocytosis/efferocytosis this review presents the latest findings on other types of endocytosis (entosis, emperipolesis, and enclysis) by hepatocytes.

Phagocytosis is a fundamental process for the ingestion and elimination of microbial pathogens and apoptotic cells. Thus, phagocytosis is vital for tissue homeostasis. In this Research Topic, we have collected a series of articles that gives us better understanding of the process of phagocytosis, however many important questions remain unsolved. For example, how different phagocytic receptors on the same cell work together? What is the role of different phagocytes in tissue homeostasis? An improved understanding of phagocytosis is essential for future therapeutics related to infections and inflammation.

# AUTHOR CONTRIBUTIONS

EL, FN, and CR have contributed to the writing of this editorial article. All authors contributed to the article and approved the submitted version.

**Conflict of Interest:** 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 © 2020 Lafuente, Niedergang and Rosales. 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 Is Essential for the Phagocytosis of Mesenchymal Stromal Cells by Monocytes

Caroline Gavin<sup>1</sup> , Stephan Meinke<sup>2</sup> , Nina Heldring<sup>1</sup> , Kathleen Anne Heck <sup>2</sup> , Adnane Achour <sup>3</sup> , Ellen Iacobaeus 1,4, Petter Höglund2,5, Katarina Le Blanc1,6† and Nadir Kadri 1,3 \* †

<sup>1</sup> Department of Laboratory Medicine, Karolinska Institutet, Stockholm, Sweden, <sup>2</sup> Department of Medicine, Huddinge, Karolinska Institutet, Stockholm, Sweden, <sup>3</sup> Science for Life Laboratory, Division of Infectious Diseases, Department of Medicine Solna, Karolinska Institutet, Karolinska University Hospital, Stockholm, Sweden, <sup>4</sup> Division of Neurology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden, <sup>5</sup> Clinical Immunology and Transfusion Medicine, Karolinska University Hospital, Stockholm, Sweden, <sup>6</sup> Center of Hematology, Karolinska University Hospital, Stockholm, Sweden

#### Edited by:

Esther M. Lafuente, Complutense University of Madrid, Spain

#### Reviewed by:

Masashi Mizuno, Nagoya University, Japan M. Kathryn Liszewski, Washington University School of Medicine in St. Louis, United States

> \*Correspondence: Nadir Kadri nadir.kadri@ki.se

†These authors share last authorship

#### Specialty section:

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

Received: 09 July 2019 Accepted: 05 September 2019 Published: 20 September 2019

#### Citation:

Gavin C, Meinke S, Heldring N, Heck KA, Achour A, Iacobaeus E, Höglund P, Le Blanc K and Kadri N (2019) The Complement System Is Essential for the Phagocytosis of Mesenchymal Stromal Cells by Monocytes. Front. Immunol. 10:2249. doi: 10.3389/fimmu.2019.02249 Mesenchymal stromal cell (MSC) therapy is a promising tool in the treatment of chronic inflammatory diseases. This has been ascribed to the capacity of MSC to release a large variety of immune-modulatory factors. However, all aspects of the mode of therapeutic MSC action in different diseases remain unresolved, mainly because most of the infused MSC are undetectable in the circulation within hours after infusion. The aim of this study was to elucidate the fate of MSC after contact with plasma. We found that upon contact with blood, complement proteins including C3b/iC3b are deposited on MSC. Importantly, we also found that complement bound to MSC enhanced their phagocytosis by classical and intermediate monocytes via a mechanism that involves C3 but not C5. Thus, we describe for the first time a mechanism which might explain, at least partly, why MSC are not found in the blood circulation after infusion. Our results indicate that MSC immune-modulatory effects could be mediated by monocytes that have phagocytosed them.

Keywords: MSC, phagocytosis, monocytes, complement, fate, plasma, live

### INTRODUCTION

Mesenchymal stromal cells (MSC) have emerged as a possible new treatment for several chronic inflammatory diseases including diabetes, graft versus host disease, and multiple sclerosis (1). Their immune-modulatory function has mainly been ascribed to paracrine mechanisms associated with secretion of immunoregulatory mediators including cytokines and growth factors which modulate inflammatory response and balance immune profiles (2). The soluble immune secretomes include prostaglandin E2 (PGE-2), indoleamine 2,3-dioxygenase (IDO), or nitric oxide (NO)(3). In numerous clinical trials, MSC have been infused to the circulation (4) but the infused cells have been difficult to detect in the blood already at short time points after infusion (5). Furthermore, tracing studies of injected MSC have revealed that only few MSC were detectable at the site of injury or inflammation despite encouraging clinical outcomes (6–8). Hence, the actual modes of action of intravenous infusion of MSC in several diseases remain unresolved.

Previous studies have shown that MSC have a very short half-life (9) and that their infusion leads to an instant blood mediated inflammatory reaction (10). Indeed, hypotheses that MSC may be trapped in the lungs where they would interact with local macrophages are gaining in popularity (9, 11, 12). Moreover, Galleu et al. have demonstrated that infused MSC are subject to perforin-induced apoptosis through recipient cytotoxic cells, which favor their phagocytosis by monocytes (13). Also, complement activation by MSC plays a role in immunosuppression of peripheral blood cells via a mechanism that involves CD11b<sup>+</sup> cells (14). On the other hand, another study also suggested that MSC may get injured after contact with blood compounds due to the complement system (15). Thus, further studies are needed to understand the interactions of MSC with different components of the immune system, in order to shed light on their fate after infusion and their mechanisms of action.

The complement system, which comprises more than 30 proteins, plays an important role in innate immunity during inflammatory responses against foreign agents (16). It can be activated through three different pathways; the classical, the lectin and the alternative pathway. The classical pathway which uses the circulating C1q molecule is mainly activated by antibodies bound to the surface of a target cell. The lectin pathway uses mannose-binding lectins that bind carbohydrate molecules at the surface of various pathogens. The alternative pathway is constitutively active at a low level in normal serum via spontaneous hydrolysis of C3. Each of these three pathways leads to the generation of labile C3 convertases, which cleave C3 into C3a and C3b that can thereafter participate in forming distinct complexes. Ultimately, the complement cascade results in activation of C5 that initiates the formation of the C5b-7 complex that finally forms the membrane attack complex (MAC), resulting in cell lysis (17). Complement activation is regulated by soluble and cell surface-bound complement inhibitors, which limit uncontrolled complement activation. These complement regulators, including CD46 and CD55, prevent C3b which binds to the host surface, from either forming C3 convertases or from initiating decay of the complexes (18). Other complement regulators such as CD59 prevent MAC assembly and pore formation in the cell membrane (19).

Receptors for complement components have been previously described in various cell types including monocytes. The complement receptor 3 (CR3), comprising CD11b, and CD18, is expressed by all monocytes and critical for facilitating phagocytosis of complement-opsonized cells or pathogens (20). In addition, an increased percentage of suppressive cells including M2 monocytes was found in vivo after MSC infusion (13, 21–23). Thus, we here hypothesized that MSC interact with complement components in plasma, which might facilitate their phagocytosis by monocytes, explaining their disappearance directly after infusion. We here demonstrate that live complement-opsonised MSC are phagocytosed by classical CD14+CD16<sup>−</sup> and intermediate CD14+CD16<sup>−</sup> monocytes via a mechanism that involves C3 but not C5.

#### MATERIALS AND METHODS

#### MSC Donors, Isolation, and Expansion

The study was approved by the Stockholm regional ethics committee. All patients provided written consent (ethical permit number: DNR 2016/338-32-4). Human bone marrow (BM) derived MSC were isolated from 12 healthy volunteer donors as described previously (24). Briefly, under local anesthesia, 30–50 mL aspirate was obtained from posterior iliac crest bone marrow (BM). MSC were isolated from the BM-mononuclear cell (MNC) fraction by Percoll density gradient centrifugation. Cells were washed and expanded in Dulbecco's modified Eagle's medium (DMEM) low-glucose complete medium, supplemented with 10% heat inactivated fetal calf serum and antibiotic-antimycotic (A/A; Gibco, Grand Island, NY), and plated at a density of 1.7 × 10<sup>5</sup> cells per cm<sup>2</sup> . Cells were prepared for harvest, washed with phosphate-buffered saline (PBS) and detached with 0.05% Trypsin-EDTA (Gibco, Grand Island, NY) for maximum 10 min at 37◦C, thereafter replated at a density of 3,400–4,000 cells per cm<sup>2</sup> and detached at a minimum confluence of 70%. Cells were either replated or cryopreserved in 10% DMSO/DMEM complete medium until further use, in liquid nitrogen. The guidelines of the International Society for Cellular Therapy were applied to analyse the MSC prior to use in research. For in vitro assays, MSC from passage 2–4 were thawed in DMEM complete medium on the day of experiments. Cultures were performed under sterile conditions in humidified atmosphere at 37◦C in 5% CO2. Co-culture experiments were carried out in 96 well-plates (Costar Ultra-low Cluster, Corning) in Roswell Park Memorial Institute 1640 (RPMI) GlutaMAX <sup>R</sup> (Gibco, Grand Island, NY) complete medium, supplemented with 10% heatinactivated pooled human blood type AB serum or 10% FCS, penicillin (100 U/mL) and streptomycin (0.1 mg/mL).

#### Plasma Preparation

Thrombin inhibitor Lepirudin (Refludan <sup>R</sup> ) was added immediately to fresh peripheral blood samples obtained from healthy volunteers. The samples were centrifuged at 2,000 × g for 10 min at 4◦C. The plasma was removed and kept on ice until further use. To focus on the complement system and exclude the coagulation cascade, we used a thrombin inhibitor in both the blood and plasma experiments. Heat inactivated (HI) plasma (30 min at 60◦C) or K3EDTA (final concentration of 10 mM, pH 7.3, Alfa Aesar) were used as negative controls. C3 inhibitor (10µM, Compstatin, CP-20 a generous gift from Professor John D. Lambris, Professor of Research Medicine in the Department of Pathology & Laboratory Medicine at the University of Pennsylvania, Philadelphia, PA, USA) or C5 inhibitor (250µg/mL, Eculizimab, Soliris, Alexion Pharmaceuticals) were used in order to inhibit the binding of complement factor C3 or C5 to the cell surface.

# Blood-Chamber and Blood Isolation Procedure

The blood chamber technique has been previously described (25). Briefly, thrombin inhibitor Lepirudin (final concentration 50µg/mL [50 mg in 1 mL NaCl]) (Refludan <sup>R</sup> ) was added immediately to fresh peripheral blood obtained from healthy donors, and collected in pre-heparinized tubes. As a negative control K3EDTA (pH 7.4) was added at a final concentration of 10 mM. Blood was added into pre-heparinized chambers, where MSC were added and incubated on a rotator at 37◦C at different time points. The experiment was stopped by adding K3EDTA (pH 7.4). In selected experiments one fraction of MSC was exposed to 10µg/mL complement inhibitors (all from Biolegend) for 30 min at 4◦C. The effect of fresh blood on MSC was assessed for viability and C3b/iC3b binding [revealed using anti-C3c FITC which binds to C3b and iC3b fragments on MSC (14)] using FlowSight system (Merckmillipore). Lysis buffer (BD Pharm Lyse <sup>R</sup> , BD Biosciences) was used to remove red blood cells before antibody staining.

#### MSC Expansion and Differentiation

MSC were thawed and seeded in DMEM culture for 1 week. MSC were trypsinised, washed and cultured in DMEM (Gibco) containing 50% plasma ± 10 mM K3EDTA, HI plasma or DMEM complete medium alone for 1, 3, or 24 h. Cells were centrifuged, washed and plated in fresh DMEM complete medium, and thereafter examined for adhesion to plastic and expansion for 6 days. Expanded MSC were used for subsequent in vitro experiments. Retained differentiation capacity of MSC was assessed using media and instruction protocols from either adipogenic (Stempro, Invitrogen) or osteogenic (Miltenyi Biotech, GmbH). Adipocyte and osteocyte differentiations were evaluated by Oil Red O (Sigma-Aldrich) and alkaline phosphatase (Sigma Fast, BCIP/NBT), respectively. Presence of lipid vacuoles or calcium deposits was analyzed under a wide field optical microscope.

#### T Cell Stimulation and Suppression Assay

PBMCs were freshly isolated from buffy coats using density gradient centrifugation on Ficoll-Isopaque (Lymphoprep <sup>R</sup> ; Axis-Shield, Norway), according to the manufacturer's protocol. Human CD3<sup>+</sup> T cells were isolated by negative selection (Miltenyi Biotec; Human Pan T Cell Isolation Kit) according to the manufacturer's instructions. CD3<sup>+</sup> T cells (purity >95%) were then stained with carboxyfluorescein succinimidyl ester (CFSE) (Invitrogen) and activated using anti-CD3/CD28 microbeads (Miltenyi Biotec) for 5 days. Using flow cytometry (BD Fortessa LSR-II), proliferation of T cells was assessed in the presence or absence of MSC. All antibodies used for T cell staining are presented in **Table 1**. Depending on conditions, MSC were treated for 1 h with plasma or heat-inactivated plasma prior to co-culture with T cells. Data were analyzed using FlowJo software (Ashland, OH).

#### Cell Surface Staining

MSC treated with plasma as described above were stained with MSC markers described in **Table 1**. Briefly, after 20 min of incubation at 4◦C with specific antibodies, cells were centrifuged for 5 min at 400x g at 4◦C. The supernatant was removed and the cell pellet of each well was taken up in 200 µL PBS. The contents of each well were then acquired using flow cytometry TABLE 1 | Antibodies used in the current study.


(BD LSRFortessa, BD Biosciences) and data were analyzed using the FlowJo software (Ashland, OH).

#### Licensing Assay

MSC production of interleukin (IL)-6 and Indoleamine 2,3 dioxygenase (IDO) in response to licensing by proinflammatory stimuli was assessed after exposure to plasma. MSC were thawed and exposed to 50% plasma ± 10 mM K3EDTA or 50% heat inactivated plasma. As negative control, MSC were cultured in DMEM complete medium only for 1 h as described for previous experiments. Cells were washed, replated and thereafter licensed with 10 ng/mL tumor necrosis factor (TNF)-α, and 100 U/mL interferon (IFN)-γ for 72 h. For detection of intracellular IL-6 or IDO, GolgiPlugTM (BD Biosciences) was added 5 h prior to the end of the experiment (according to manufacturer's protocol). Cells were acquired using flow cytometry (BD LSR Fortessa, BD Biosciences) and data were analyzed using the FlowJo software (FlowJo, Ashland, OH).

#### Complement Lysis Assay

Freshly thawed MSC were loaded with calcein red-orange acetoxymethyl ester (calcein RO AM) (Molecular Probes) at a concentration of 2.5µg/mL in PBS, and incubated for 10 min at 37◦C. Cells were then centrifuged and resuspended in PBS (control), in 50% plasma or in 50% heat-inactivated plasma as negative control. The blocking antibodies CD46, CD55, or CD59 against complement regulators (**Table 1**) were added and the experiment was analyzed after 1 h incubation at 37◦C. Staining for flow cytometry was performed as described above.

#### Phagocytosis Assay

Freshly thawed MSC were stained with 51 nmol/L pHrodo succinimidyl ester (Molecular Probes) in PBS for 10 min at RT, centrifuged at 500 × g for 7 min and resuspended in DMEM. Cells were incubated for 1 h at 37◦C with either of the following conditions: control DMEM complete medium, or medium with 50% plasma, 50% HI plasma or 50% plasma with 10 mM K3EDTA. MSC under each condition were also divided up in the following fractions: MSC alone, MSC with C3 inhibitor (10µM, Compstatin) or MSC with C5 inhibitor (250µg/mL, Soliris) added, respectively. One further fraction was exposed to complete medium with added C3 complement protein (15 mg/mL, Sigma C2910). MSC were washed and resuspended in complete DMEM medium. The cell line THP-1 or freshly isolated PBMCs containing monocytes were used as phagocytes. To increase the phagocytic activity of THP-1 cells, 15 ng/mL phorbol 12-myristate 13-acetate were added for 15 min at 37◦C. Further, 10µg/mL Cytochalasin D (Sigma-Aldrich) was added to negative control cells in order to block phagocytosis for a minimum of 30 min at 37◦C. MSC were co-cultured with phagocytic cells at 1:1 ratio and incubated for 2 h at 37◦C. Thereafter, cells were centrifuged at 400x g for 5 min and stained on ice for analysis by flow cytometry (BD LSRFortessa, BD Biosciences). Monocytes were identified by forward scatter/side scatter (FSC/SSC) and gating on CD14<sup>+</sup> cells. Phagocytosis was detected by pHrodo fluorescence. The positive gate was set based on the negative control with Cytochalasin D. Cells were acquired using flow cytometry (BD LSRFortessa, BD Biosciences) and data were analyzed using the FlowJo software (FlowJo, Ashland, OH).

#### Statistical Analysis

Statistical analysis using paired t-test or one way ANOVA (described in figure legends) were performed using Graph Pad Prism (Graph Pad Prism Software Incl. San Diego, USA). p < 0.05 was considered statistically significant.

# RESULTS

## Survival and Function of MSC Upon Contact With Plasma

Due to the fast clearance of MSC from blood circulation after i.v. infusion, it remains unclear whether MSC die after contact with plasma. To test this, we incubated MSC with freshly isolated plasma in different time periods stretching from one up to 24 h **(Figures 1A–C)** and data not shown. It has been recently demonstrated that the anti-human C3c antibody detects C3b/iC3b deposition on the surface of MSC after incubation with serum (14). To check for potential MSC interaction with the complement system, we stained for the presence of the complement components C3b/iC3b on the surface of MSC by using the same antibody **(Figures 1A–C)**. MSC from different healthy donors displayed heterogeneity in C3b/iC3b deposition, ranging from 10 to 60% after 1 h of incubation with plasma **(Figure 1C)**. Heterogeneity of C3b/iC3b deposition was also observed when we used purified C3 protein together with MSC **(Figure 1C)**. Using flow cytometry, we found that MSC incubated with plasma displayed the same shape and granularity with no significant changes in viability compared to control MSC **(Figure 1D)**. This was also observed when MSC were incubated with all blood compounds using the blood chamber technique (25) **(Figure 1B** and data not shown**).** It should be noted that the intact coagulation system was inhibited in all experiments by the addition of a thrombin inhibitor. Altogether, these results indicate that MSC survive in vitro during the first hours when they interact with plasma.

Moreover, we found that MSC pre-treated with plasma suppressed the proliferation of activated CD3<sup>+</sup> T cells similarly to MSC pre-treated with heat inactivated (HI) plasma or nontreated MSC, suggesting that the immune-suppressive properties of MSC are not altered after interaction with plasma compounds **(Figure 1E).** In response to high concentrations of IFNγ, and TNFα, MSC pre-incubated with plasma still produced substantial levels of IL-6 and IDO **(Figure 1F)**. Using the appropriate media for MSC differentiation, we found that MSC pre-incubated with plasma differentiate to adipocytes and chondrocytes similarly to control MSC suggesting that the differentiation capacity of MSC was maintained **(**Data not shown**)**. Interestingly, expression levels of the complement regulators CD46, CD55, and CD59 were not affected following contact between MSC and plasma **(Figure 1G)**. Altogether, these results indicate that MSC survive and are fully functional, after contact with plasma and that this protection might be due to the expression of complement regulators.

### Effects of Blocking Complement Regulators on MSC Pre-treated With Plasma

We tested whether expression of complement regulators CD46, CD55, and CD59 is important for the survival of MSC following incubation with plasma. MSC were labeled with calcein which is a cytosolic dye that leaks out if the membrane of MSC is damaged by the membrane attack complex (26). MSC were treated with specific blocking antibodies, washed, then incubated later with plasma. Thereafter, MSC were stained for C3b/iC3b deposition on their surface **(Figures 2A–C)**. Addition of an anti-CD59 specific antibody but not anti-CD46 nor anti-CD55 resulted in calcein leakage when MSC were pre-treated with plasma **(Figures 2A–C)**. Importantly, blocking of any complement inhibitor did not induce calcein leakage when MSC were pretreated with HI plasma or untreated MSC (**Figure S1**). The shape and granularity of MSC were dramatically changed with an accumulation of cell debris when MSC were pre-treated with plasma and antibodies blocking the complement inhibitor CD59 (**Figure S2**). Moreover, inhibition of CD59 led to a pronounced increase of C3b/iC3b deposition on the surface of MSC pretreated with plasma compared to controls **(Figures 2A–C)**. Using FlowSight, we found that blocking of CD59, but not other complement regulators resulted in cell death as shown by 7-AAD staining **(Figure 2D)** and data not shown. Similarly, using flow cytometry, we found that almost all MSC stained positive for the dead cell marker after pre-treatment with complement inhibitor CD59 and plasma **(Figure 2E)**. CD59 was expressed at a much higher levels compared to CD46 and CD55 on the surface of

all analyzed MSC. Thus, the different expression levels of CD46, CD55, and CD59 do not exclude the possibility that both CD46 and CD55 may also be important for protection. Furthermore,

cytometry. Mean fluorescent intensities of MSC from five MSC in two different experiments are displayed with mean and SD.

although this result is interesting, it should be noted that blocking of CD59 by this specific IgG2a isotype may also induce killing of MSC through the activation of the classical pathway (27). Thus,

at this stage, our results which suggest that CD59 may protect MSC from complement lysis should be considered as suggestive, but not conclusive.

# Complement Factors Enhance Phagocytosis of MSC by Monocytes

Phagocytosis is considered to be an important pathway for removal of complement-opsonized cells from circulation (28). We hypothesized that MSC are phagocytosed via a mechanism that involves complement proteins. To address this, we labeled MSC with the pH-sensitive dye pHrodo, which has a low fluorescence at neutral pH that increases with decreasing pH, for instance upon entering phagolysosomes where the pH is significantly reduced (29). This approach allows to clearly address active phagocytosis of MSC by monocytes. As a negative control, Cytochalasin D (CytoD) was added to the co-culture in order to inhibit phagocytosis. We initially used the monocytic cell line THP-1, which has been extensively used in phagocytosis studies (30). Although the phagocytic capacity of THP-1 cells was not high, we consistently observed higher phagocytosis if MSC were pre-treated with plasma, compared to heat inactivated plasma or medium alone **(Figures 3A,B)**. Therefore, we set-up experiments using fresh PBMC from healthy donors **(Figures 3C–E)** and gated on CD14<sup>+</sup> cells in order to analyze phagocytosis by monocytes **(Figure 3C)**. After 2 h of co-culture, around 15% of monocytes showed a higher pHrodo fluorescence indicating that MSC are phagocytosed by monocytes in the absence of plasma **(Figures 3D,E)**. Interestingly, almost half of the monocytes phagocytosed MSC if they were pre-incubated with plasma and washed prior to co-culture **(Figures 3D,E)**. When MSC were pre-incubated with HI plasma, we observed a significant decrease in phagocytosis of MSC by monocytes compared with plasma **(Figures 3D,E)**. These results suggest that plasma factors interacting with MSC are responsible for the increased phagocytic capacity of monocytes.

### Phagocytosis of MSC Is Mediated by Classical and Intermediate Monocytes

In humans, monocytes can be divided into three subsets based on the expression of CD14 and the Fcγ receptor III CD16 (31), including classical monocytes (CD14<sup>+</sup> CD16−), intermediate monocytes (CD14<sup>+</sup> CD16+) and non-classical

washed and co-cultured with THP-1 cells (A,B) or PBMC (D,E). Phagocytosis was analyzed after 2 h co-culture. Phagocytosis inhibitor Cytochalasin D (CytoD) was added as a negative control. (A) Representative plots of flow-cytometric analysis of phagocytosis by THP-1 cells. pHrodo bright fluorescence indicates cells that have phagocytosed labeled MSC. (B) Pooled data of THP-1 cell phagocytosis, bars represent means with SD (n = 4) from two independent experiments. (C) Gating strategy of CD14<sup>+</sup> monocytes in freshly isolated PBMC. (D) Representative plots of flow-cytometry analysis of phagocytosis by monocytes. pHrodo bright fluorescence indicates cells that have phagocytosed labeled MSC. (E) Pooled data from MSC phagocytosis by monocytes, bars represent means with SD (n = 4) of two independent experiments. Statistical significance was determined using ANOVA followed by Holm-Sidak's multiple comparisons test \*p = 0.05, \*\*p = 0.001.

monocytes (CD14<sup>−</sup> CD16+). Here, we made use of the CD14 and CD16 surface markers to identify these three subsets in our assays **(Figure 4A)**. We addressed whether MSC phagocytosis, mediated by plasma factors, involves a specific subset of monocytes. Indeed, monocytes with phagocytosed MSC (pHrodo Bright) were mainly classical CD14<sup>+</sup> CD16<sup>−</sup> and intermediate

CD14<sup>+</sup> CD16<sup>+</sup> monocytes while non-classical CD14<sup>−</sup> CD16<sup>+</sup> monocytes did not phagocytose MSC **(Figures 4B,C)** and **Figure S3**. Since intermediate monocytes represent only a small fraction of the total number of monocytes, the majority of pHrodo bright cells were classical monocytes **(Figures 4B,C)**. The distribution of monocyte subsets that phagocytose MSC pretreated with plasma was similar when we compared them to nontreated MSC or MSC treated with HI plasma (data not shown).

### Mechanism of MSC Phagocytosis by Monocytes

To better understand which plasma components are involved in the phagocytosis by monocytes of MSC pre-treated with plasma, we first tested whether complement factors may play a role in this process. The complement factors C3 and C5 were selected due to their important roles in monocyte-mediated phagocytosis (17, 20, 32). As anticipated, inhibition of complement factors C3 or C5 did not change the percentage of phagocytosis of control MSC or MSC pre-incubated with HI plasma **(Figures 5A,B)**. However, we observed a significant decrease in the phagocytosis of MSC pre-incubated with plasma in the presence of a C3 inhibitor. On the other hand, inhibition of C5 had no effect on monocyte phagocytosis **(Figures 5A,B)**. Following subtraction of basal phagocytosis of MSC (phagocytosis which is observed in plasma free medium), ∼80% of plasma-mediated increase of phagocytosis was due to C3 binding **(Figure 5C)**. In addition, a significant positive correlation between C3b/iC3b binding and the percentage of pHrodo bright monocytes was detected **(Figure 5D).** Altogether, these results indicate that C3 is an important mediator for phagocytosis of MSC by monocytes **(Figure 6)**. Further investigations are required to unravel the exact molecular mechanisms underlying the role of C3 in phagocytosis of MSC by monocytes.

## DISCUSSION

The current consensus in the field of MSC therapy has hitherto been that intravenous infusion of MSC leads to quick clearance of MSC from the blood circulation. The remaining debate has been focused on the fate of administrated MSC (5). Here, we report that MSC survive and conserve their phenotypic and functional activities if they are in contact with complement active plasma. Furthermore, our results demonstrate that the complement factor C3 facilitates phagocytosis of live MSC by classical and intermediate monocytes.

One of the first lines of innate defense of the immune system is the complement system, which during infection or entry of foreign cells will bind to their cell surface and amplify a cascade of enzymatic reactions (16). To prevent uncontrolled complement activation, cells express a number of complement regulators at their surface (18). Based on our and other previous reports showing that MSC express complement inhibitors including CD46, CD55, and CD59 (14, 15, 33), we hypothesized that MSC are protected from complement

shown in (B) and the percentage of phagocytosis observed in plasma free medium was subtracted. (D) Correlation between percentage of C3b/iC3b binding to MSC and percentage of pHrodo bright monocytes from HI plasma and plasma in the phagocytosis experiments.

attacks. Our results suggest that CD59 might protect MSC from complement lysis. However, this result must be taken with caution since the specific isotype (IgG2b) of the anti-CD59 antibody used within the present study may lead to the activation of the classical complement pathway as previously reported (27).

FIGURE 6 | Model of MSC fate after interaction with blood. (A) Increase in C3 binding and decrease in the survival of plasma pre-treated MSC in the presence of anti-CD59. (B) C3 binds to the surface of MSC, probably through the alternative complement pathway. CD59 blocks the membrane attack complex from forming. MSC are phagocytosed by classical and intermediate monocytes, mainly mediated by the presence of C3 on the MSC surface. The receptor binding to C3 and inducing the phagocytosis remains unknown.

On the other hand, our results are also in accordance with Feng et al. who demonstrated that CD59 plays an important role in protection against complement-mediated cytotoxicity (15). However, the same report also suggested a role for CD55, which we did not observe in our study. The latter could be due to different expression levels of CD46, CD55, and CD59 on the surface of MSC. Furthermore, it has also been shown that despite CD59 and CD55 expression, MSC were injured after complement binding, as revealed by release of the fluorescent dye bis-carboxyethyl-carboxyfluorescein (BCECF) from MSC (15). In our study, we did not observe calcein leakage when MSC were treated with plasma suggesting that MSC were not injured. Thus, it is plausible that the different dyes used to detect cytotoxicity might be the reason for the different results obtained. Indeed, calcein has been reported to display higher sensitivity and less spontaneous leakage than BCECF (34). Moreover, plasma pre-treated MSC stained negative for all the three commonly used cell death markers in both flow cytometry and FlowSight techniques and displayed full functional capability favoring the hypothesis that MSC survive after contact with plasma. Our data are in accordance with mouse and human in vivo tracing experiments, which identified MSC in different organs hours to days after infusion (7, 8, 21, 35).

We observed an increase in C3b/iC3b deposition on MSC after CD59 inhibition. Here again, we are not excluding the clear possibility that this could be due to the activation of the classical complement pathway (27). However, recent findings suggest a role for CD59 not only as a major controller of the membrane attack complex (19) but also in C3 regulation (36, 37). Furthermore, although CD55 was suggested to regulate C3 (36), we did not observe any significant changes in C3b/iC3b deposition when CD55 was blocked on MSC. More experiments are therefore required for further clarification. It should be noted that interaction with complement is not unique to bone marrow MSC as adipocyte stromal cells have been shown to interact with complement in rat peritonitis model (38).

Complement deposition on the cell surface serves as target for complement receptors present on mononuclear phagocytic cells in particular monocytes and macrophages (28). Our results revealed that live MSC were targeted by monocytes via a mechanism that involves complement. Such a mechanism might explain in part the observed rapid clearance of MSC when infused to the circulation. Indeed, only 2 h of incubation led to phagocytosis of complement-opsonized MSC by more than 45% of monocytes. A recent study by the Hoogduijn research group used umbilical cord MSC (uMSC) labeled with the lipophilic membrane dye PKH26, which were mixed with monocytes in blood. They found PKH26-labeled uMSC fragments on monocytes 3 h after co-culture (22). Also, Braza et al. found in an asthma model in mice that injected PKH26-labeled MSC were engulfed by lung macrophages within 24 h following i.v. injection (21). However, the use of PKH26 may be disadvantageous since it also could incorporate itself into other cells (in this case to monocytes or macrophages) through a cell-to-cell membrane transfer process called trogocytosis, giving rise to false positive signals without phagocytosis (39, 40). In the current study, MSC were labeled with pHrodo, which has the advantage that it is only fluorescent under acidic conditions, consequently it will only be fluorescent in phagolysosomes (pH 4) but neither outside of the cell nor in the cytoplasm where the pH is around 7 (29, 41). Thus, our data demonstrated that live complement-opsonized bone marrow MSC were indeed phagocytosed by monocytes. Intriguingly, we showed that MSC were also phagocytosed by monocytes in the absence of plasma albeit to a lesser extent. These observations are consistent with the paradigm that MSC can be engulfed by cancer cells in the absence of plasma (42). One possible explanation might be that MSC express adhesion molecules that allow a tight contact with monocytes, which could facilitate their phagocytosis (28, 43). A recent study suggested that apoptosis of MSC is induced by cytotoxic T cells which favor their engulfment by phagocytic cells (13). Our data complement this study and reveal that live MSC can also be subjected to phagocytosis by monocytes. This might explain recent pre-clinical studies showing positive effect of living MSC in the treatment of sepsis (44, 45). Nevertheless, a comparative study on the immunomodulatory potential of monocytes which phagocyte apoptotic vs. live MSC is required.

With further evaluation of the mechanism of complementmediated phagocytosis, we found that addition of compstatin, which blocks the activation of complement at the C3 level (32) significantly affected phagocytosis of MSC by monocytes. Adding the C3 inhibitor reduced C3b/iC3b deposition on the surface of MSC suggesting that binding of C3 at the surface of MSC is associated with complement activation confirming previous observations (14). However, the exact mechanisms underlying how C3 deposition triggers phagocytosis remain to be investigated. An interesting question, which needs further investigation, is which receptor on monocytes is involved in MSC phagocytosis. Monocytes express receptors for different C3 derivative fragments including CR1 (CD35), CR2 (CD21), CR3 (CD11b/CD18), and CR4 (CD11c/CD18) (46). Among these receptors, CR1 and CR2 on monocytes are involved in phagocytosis via interactions with the C3 complement during infection (47). Thus, a similar mechanism might occur for MSC. Interestingly, phagocytosis mediated via complement receptors is not always associated with inflammation (47). Therefore, it is reasonable to speculate that monocytes might be the final destination of MSC after their infusion into the circulation without inducing an excessive inflammation. This might partly explain the mild inflammation detected after intravenous infusion of MSC (48).

Another question addressed in our study is which type of monocytes phagocyte MSC. Moll et al. showed that depletion of CD14+CD11bhigh monocytes was associated with strong decrease in the immunosuppressive function of MSC in vitro in both alloantigen- and PHA-stimulated mixed lymphocyte reactions (14). We found that classical CD16<sup>−</sup> CD14<sup>+</sup> and intermediate CD16<sup>+</sup> CD14<sup>+</sup> monocytes were first in engulfing MSC supporting recent findings (22). However, we did not see any significant changes in expression of the other Fc receptors CD32 or CD64 on phagocytic monocytes which have been involved in complement-mediated phagocytosis during certain infection (data not shown). However, kinetic experiments are needed to see whether non-classical CD16<sup>+</sup> CD14<sup>−</sup> monocytes are also involved in the phagocytosis of MSC at a later time point. Indeed, de Witte et al. suggested that 24 h after engulfment of uMSC, monocytes polarize from CD14<sup>+</sup> CD16<sup>−</sup> to CD14<sup>+</sup> CD16<sup>+</sup> expressing cells (22).

In conclusion, we propose that complement opsonization plays a crucial role in the fate of MSC after intravenous infusion **(Figure 6)**. It mediates their rapid phagocytosis by classical and intermediate CD14<sup>+</sup> monocytes. MSC are protected against complement injury through CD59, which is in contrast to the previous dogma that MSC disappear from circulation due to destruction. Our results provide new insights on the fate of MSC after intravenous infusion, which needs to be taken into consideration in order to improve the therapeutic role of MSC in various diseases.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the manuscript/**Supplementary Files**.

#### ETHICS STATEMENT

The studies involving human participants were reviewed and approved by Ethics committee at Karolinska Institutet. The patients/participants provided their written informed consent to participate in this study. The following are the three ethical permits that cover whole paper. DNR: 2013/39-31/4, DNR: 446/00, and DNR: 2016/338-32-4.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

This study was designed by CG, KL, and NK with the input of SM, NH, EI, AA, PH, and KH. CG, EI, KH, SM, and NK performed experiments, analyzed, and interpreted the results. CG, SM, KL, and NK wrote the manuscript with inputs from EI, PH, and AA. All authors read, commented, and approved the final manuscript.

#### ACKNOWLEDGMENTS

The authors acknowledge the MedH Flow Cytometry facility at Karolinska Institutet, supported by grants from Karolinska Institutet and the Stockholm County Council. Authors thank Cecilia ström, Anton Törnqvist André, and Tachazi Plym Forshell for their technical assistance with laboratory work and useful discussions.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02249/full#supplementary-material

Figure S1 | Complement inhibitors CD46, CD55, or CD59 are not toxic on MSC. Representative contour plots of calcein RO stained MSC and exposed to plasma, HI plasma or medium for 1 h in the presence or absence of complement inhibitors CD46, CD55, or CD59. Data is representative of 5 MSC of two independent experiments.

Figure S2 | Representative contour plots showing shape (FSS) and granulosity (SSC) of MSC after incubation for one hour with or without active plasma, in the presence or absence of complement inhibitor anti-CD59. Data are representative of at least three different experiments.

Figure S3 | Phagocytosis of MSC is mediated by classical and intermediate monocytes. Presence of non classical monocytes among other subsets on gated pHrodo Low MSC. Pooled data of two different donors (PBMC = 4).

limited long-term engraftment and no ectopic tissue formation. Stem Cells. (2012) 30:1575–8. doi: 10.1002/stem.1118


**Conflict of Interest:** 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 Gavin, Meinke, Heldring, Heck, Achour, Iacobaeus, Höglund, Le Blanc and Kadri. 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.

# Calreticulin and Galectin-3 Opsonise Bacteria for Phagocytosis by Microglia

Tom O. J. Cockram, Mar Puigdellívol and Guy C. Brown\*

Department of Biochemistry, University of Cambridge, Cambridge, United Kingdom

Opsonins are soluble, extracellular proteins, released by activated immune cells, and when bound to a target cell, can induce phagocytes to phagocytose the target cell. There are three known classes of opsonin: antibodies, complement factors and secreted pattern recognition receptors, but these have limited access to the brain. We identify here two novel opsonins of bacteria, calreticulin, and galectin-3 (both lectins that can bind lipopolysaccharide), which were released by microglia (brain-resident macrophages) when activated by bacterial lipopolysaccharide. Calreticulin and galectin-3 both bound to Escherichia coli, and when bound increased phagocytosis of these bacteria by microglia. Furthermore, lipopolysaccharide-induced microglial phagocytosis of E. coli bacteria was partially inhibited by: sugars, an anti-calreticulin antibody, a blocker of the calreticulin phagocytic receptor LRP1, a blocker of the galectin-3 phagocytic receptor MerTK, or simply removing factors released from the microglia, indicating this phagocytosis is dependent on extracellular calreticulin and galectin-3. Thus, calreticulin and galectin-3 are opsonins, released by activated microglia to promote clearance of bacteria. This innate immune response of microglia may help clear bacterial infections of the brain.

#### Edited by:

Carlos Rosales, National Autonomous University of Mexico, Mexico

#### Reviewed by:

Harald Neumann, University of Bonn, Germany Tommy Regen, Johannes Gutenberg-University Mainz, Germany

\*Correspondence:

Guy C. Brown gcb3@cam.ac.uk

#### Specialty section:

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

Received: 27 August 2019 Accepted: 25 October 2019 Published: 12 November 2019

#### Citation:

Cockram TOJ, Puigdellívol M and Brown GC (2019) Calreticulin and Galectin-3 Opsonise Bacteria for Phagocytosis by Microglia. Front. Immunol. 10:2647. doi: 10.3389/fimmu.2019.02647 Keywords: calreticulin, galectin-3, opsonin, MerTK, LRP1, microglia, bacteria

# INTRODUCTION

Bacterial infections of the brain are serious, often causing death or irreversible brain damage (1). They are relatively rare in developed countries, but still relatively common in some developing countries (2). Infections include bacterial meningitis, bacterial encephalitis, bacterial meningoencephalitis, and brain abscess (1). Bacterial infection of the brain appears to increase in Alzheimer's disease, and might contribute to this disease (3). Brain infections are serious partly because antibodies and leucocytes have limited access to the brain, so that microglia (brain-resident macrophages) act as the primary innate immune response to bacterial infection of the brain. One of the main ways that microglia combat bacterial infection is by phagocytosing the bacteria (4).

Phagocytosis is the cellular engulfment and digestion by cells of large extracellular particles, including bacteria. Phagocytes are cells specialized in phagocytosis, and their phagocytosis of other cells (the target cell) normally requires: an "eat-me" signal on the target cell, a phagocytic receptor on the phagocyte, and an opsonin to link the receptor and eat-me signal (5). An eat-me signal is a molecule normally found inside the target cell, but when exposed on the surface of the target cell promotes a phagocyte to phagocytose that cell. A phagocytic receptor is a receptor on a phagocyte that, when activated by an eat-me signal or opsonin on the target cell, induces engulfment of the target cell by the phagocyte. Opsonins are normally soluble, extracellular proteins that when bound

**21**

to the surface of cells promote phagocytosis of those cells. Classical opsonins include IgG antibodies and complement components, but also pentraxins, collectins, and ficolins. Here, we identify two novel opsonins of bacteria, galectin-3, and calreticulin.

Galectin-3 (gal-3), also known as Mac-2 or LGALS3, is a protein expressed and released by macrophages and microglia (6–9). It is a galactose-binding lectin (galectin) that preferentially binds to N-acetyl-lactosamine (a disaccharide of galactose and N-acetyl-glucosamine) in glycoproteins or gangliosides (7, 10). Galectin-3 is normally monomeric, but when the Cterminal carbohydrate recognition domain (CRD) binds Nacetyl-lactosamine, the N-terminal oligomerizes, so that galectin-3 can cross link glycoproteins or gangliosides (7, 10). By binding to sugars on target cells and then crosslinking to phagocytic receptors on phagocytes, galectin-3 can potentially act as an eatme signal or opsonin (9, 11). However, it is not known whether galectin-3 can opsonise bacteria.

Calreticulin is a protein normally found in the endoplasmic reticulum where it acts as a chaperone, binding to terminal glucose residues on developing glycoprotein oligosaccharides (12). However, in conditions of apoptosis or endoplasmic reticulum stress, calreticulin can translocate to the cell surface, and some cells such as neutrophils constitutively express calreticulin on the cell surface (13, 14). Surface-exposed calreticulin has been demonstrated to act as an eat-me signal to macrophages (13, 15–17). Surface-exposed calreticulin is recognized by the phagocytic receptor LRP1 on phagocytes (13, 18), although calreticulin is also found associated with LRP1 on the phagocyte membrane where it acts as a co-receptor for LRP1 ligands such as C1q and alpha-2-macroglobulin (19, 20). Note, however, that calreticulin is a soluble protein, and therefore has the potential to be released into the extracellular space (21), where in principle it might act as an opsonin.

In this paper, we show that calreticulin and galectin-3 can be released by a phagocyte, microglia, when activated, and can then bind to and opsonise bacteria for phagocytosis by the microglia.

#### MATERIALS AND METHODS

All experiments were performed in accordance with the UK Animals (Scientific Procedures) Act (1986) and approved by the Cambridge University Local Research Ethics Committee.

#### Cell Culture and Treatments

Primary microglia were prepared from mixed glial cultures from the cortices of postnatal day 4–6 mice or rats as described (22, 23). After removal of meninges and mechanical dissociation, cortices were matured in vitro for at least 6 days using Dulbecco's modified Eagle's medium (DMEM) (ThermoFisher, California, USA) supplemented with 10% performance-plus fetal bovine serum (FBS) (ThermoFisher, California, USA), before removing microglia by shakingoff and plating in culture medium (a 1:2 ratio of old "conditioned" medium: fresh medium) on well-plates precoated with poly-L-lysine (Sigma, Missouri, USA). BV2 mouse microglial cells were maintained in DMEM supplemented with 10% FBS. All tissue culture medium was supplemented with 100 U/ml penicillin/streptomycin (ThermoFisher, California, USA). DH5α Escherichia coli for phagocytosis experiments were grown shaking in LB media at 37◦C. The following reagents were used: lipopolysaccharide (LPS) from Salmonella (serotype: typhimurium) or E. coli, 5-(and-6)-carboxytetramethylrhodamine (TAMRA), Cytochalasin D, D-glucose and D-lactose were from Sigma-Aldrich. Recombinant human Calreticulin protein (Abcam, Cambridge, UK), pHRODO Red succinimidyl-ester and sucrose (ThermoFisher, California, USA), recombinant human LRPAP protein (RAP) (R&D systems, Minneapolis, MN, USA), UNC569 (Calbiochem, MA, USA), UNC2881 (Selleckchem, TX, USA), anti-Calreticulin polyclonal antibody (Enzo Life Sciences, NY, USA), rabbit polyclonal IgG antibody (Southern Biotech, AL, USA). Human galectin-3 was provided as a kind gift from Tomas Deierborg (Lund university). Antibodies were Fc-blocked using an Affinipure Fab fragment goat anti-rabbit IgG (Jackson ImmunoResearch, Cambridge, UK).

# RNA Isolation and RT-qPCR

RNA was isolated from microglial BV2 cells using Monarch Total RNA Miniprep Kit (New England Biolabs, Massachusetts, USA), and cDNA was generated from 1 µg RNA and random hexamer primers using the SuperScript II Reverse Transcriptase Kit (ThermoFisher, California, USA). qT-PCR was run with the Platinum SYBR Green qPCR SuperMix (ThermoFisher, California, USA) using a Rotor-Gene Q cycler (Qiagen, Hilden, Germany). Primers against mouse CALR, LGALS3, and IL6 were used, with ACTB (β-actin) as the internal control (Sigma, Missouri, USA). Relative mRNA levels of target genes were analyzed by comparing fold-changes in the delta-delta threshold cycle, after normalizing against the internal control for each condition.

#### Enzyme-Linked Immunosorbent Assay (ELISA)

Primary microglia from mice—chosen for species-specificity of available antibodies—were plated at densities of 10,000 cells/100 µl/well (96 well-plate format) in culture medium. Cells were treated for 24 h, and cell media was extracted and subjected to a calreticulin ELISA (Abbexa, Cambridge, UK) or galectin-3 ELISA (R&D Systems, Minneapolis, USA) as per manufacturer's instructions. Absorbances at 450 nm were measured using a FLUOstar Optima plate reader (BMG Labtech, Ortenberg, Germany) and represented as protein concentration calculated against a standard curve.

**Abbreviations:** LRP1, low-density lipoprotein receptor-related protein 1; sLRP1, soluble low-density lipoprotein receptor-related protein 1; LPS, lipopolysaccharide; Gal-3, galectin-3; MHCII, major histocompatibility complex II; MerTK, Mer tyrosine kinase; TAMRA, 5-(and-6)-carboxytetramethylrhodamine; TREM2, triggering receptor expressed on myeloid cells 2; TLR4, toll-like receptor 4; mLRPAP (or RAP), mouse LRP-associated protein; CSF, cerebrospinal fluid.

# TAMRA-Conjugated Protein Binding Assays

Calreticulin and galectin-3 were incubated with amine-reactive 5-(and-6)-carboxytetra-methylrhodamine (TAMRA, 50µM) for 20 min at 37◦C before diluting in 15 ml PBS. Proteins were spun down using an Amicon Ultracentrifuge filter (Millipore, Merck, New Jersey, USA) with a molecular weight cut-off 10,000 Daltons to remove unbound TAMRA. E. coli were then resuspended in either the protein-positive fraction, or the protein-free eluant as a control.

#### Microglial Phagocytosis Assays

Primary microglia from rat (chosen due to greater cellular yield compared to mice) were plated at densities of 50,000 cells/200 µl/well (96 well-plate format) in culture medium. For LPS experiments, cells were treated with vehicle or 100 ng/ml LPS within 60 min of seeding. LPS from S. enterica was used in all cases except for **Figure 5C**, which was LPS from E. coli. E. coli were grown shaking overnight in LB media (37◦C). Bacteria were heat-inactivated at 65◦C for 15 min before centrifuging at 6,000× g for 5 min and resuspending in PBS. Bacteria were stained with amine-reactive pHrodo Red succinimidyl-ester at 10µM for 20 min (37◦C) before washing several times in PBS via centrifugation and resuspension to remove unconjugated pHrodo. Bacteria were resuspended in PBS, and either incubated for 90 min with opsonins (followed by several wash steps) or added directly to cells, and maintained in an incubator (37◦C, 5% CO2-infused) for the 1-h phagocytosis assay. Primary microglia were detached via trypsinization and resuspended in 60 µl PBS. Samples were maintained in darkness on ice and taken directly for FACS analysis using an Accuri C6 Flow Cytometer (BD services, San Jose, CA, USA). Sucrose (50 mM), lactose (50 mM), UNC569 (5µM), UNC2881 (200 nM), mLRPAP (250 nM or 500 nM) or cytochalasin D (10µM) were added to cells 60 min prior to bacteria; anti-calreticulin or IgG serotype control (2µg/ml) were added to cells 3 h prior to bacteria.

#### Cell Viability Assay

Cell viability was measured using the DNA-staining dye propidium iodide (Sigma, Missouri, USA). Cells were stained with propidium iodide (1.5µM) for 20 min and staining was quantified via flow cytometry using an Accuri C6 Flow Cytometer (BD, New Jersey, USA).

## Statistical Analysis

All statistical analyses were performed using GraphPad Prism version 8. All results contained herein represent mean values averaged from at least three independent experiments. Standard error of the mean (SEM) is represented as error bars in all cases. Figures comparing just two conditions were analyzed using a Student's t-test; all other data was analyzed using one- or two-way ANOVA and post-hoc Tukey's multiple comparison test.

# RESULTS

Calreticulin is known to function as an eat-me signal or as a phagocytic co-receptor, but it is unclear whether it can act as an opsonin, which would require that it was released from cells, normally inflammatory-activated immune cells. Therefore, we tested whether calreticulin is released extracellularly from microglial cells, and whether any such release was enhanced by inflammatory activation of the microglia. Supernatants from primary mouse microglia ± LPS were isolated, and calreticulin was measured by ELISA (**Figure 1A**). We found that media conditioned with vehicle-treated microglia contained small, but statistically insignificant levels of calreticulin (1.29 ± 0.93 ng/ml), when compared to media which had not seen cells (p = 0.695). However, 24 h of LPS treatment triggered a dramatic increase in calreticulin released from these cells (7.28 ± 1.87 ng/ml), which was significantly higher than that from non-activated cells (p = 0.017), or when compared to cell-free media (p = 0.004). To rule out the possibility of necrotic leakage of calreticulin from cells after treatment, cells were stained with propidium iodide and quantified—no significant increases in necrosis after LPS treatment was observed (data not shown).

Galectin-3 is known to be released from blood-marrow macrophages and BV2 cells in response to LPS stimulation (8, 9, 24). So, we tested whether primary mouse microglia also released galectin-3 into the culture medium using an ELISA. We found that unstimulated microglia released galectin-3 into the medium (1.09 ± 0.185 ng/ml, p < 0.01), and this release was increased by LPS (1.73 ± 0.13 ng/ml) (**Figure 1B**; p = 0.02).

To test whether LPS affected calreticulin expression, RNA was isolated from mouse microglial BV2 cells treated ± LPS for 24 h, and calreticulin RNA was measured by qPCR. LPS was found to increase calreticulin mRNA by 35% (± 4) at this time-point (**Figure S1A**). Similarly, LPS increased galectin-3 mRNA by 19% (± 5) (**Figure S1B**), suggesting that LPS triggers upregulation as well as release of calreticulin and galectin-3 in microglia.

To assay microglial phagocytosis of bacteria, we labeled E. coli with amine-reactive pHrodo Red succinimidyl ester (10µM), which covalently binds the bacteria, but only fluoresces when the bacteria are delivered into the acidic lysosome during phagocytosis. Labeled bacteria were incubated with microglial BV2 cells (used initially to optimize the assay) and phagocytosis was quantified by flow cytometry—there was a linear increase in fluorescence over the first 120 min of incubation (**Figure S2**). Having optimized the assay in BV2 cells, we then switched to using primary rat microglia, because they are more physiologically relevant than BV2 cells, and are obtainable in higher yields than primary mouse microglia. Labeled bacteria were incubated with primary rat microglia for 60 min; phagocytosis was confirmed by microscopy (**Figure 2A**) and quantified by flow cytometry (**Figure 2B**). Cytochalasin D, an inhibitor of phagocytosis, completely abolished the fluorescence increase when applied at 10µM (p < 0.0001; p = 0.528 vs. the control without bacteria), confirming that the fluorescence increase was due to phagocytosis.

Calreticulin and galectin-3 are sugar-binding proteins known to have binding affinity for bacterial LPS (25, 26). LPS (lipopolysaccharide) contains many sugar residues and constitutes much of the surface of gram-negative bacteria. So calreticulin and galectin-3 could potentially bind gram-negative bacteria such as E. coli. To test whether calreticulin or galectin-3

can bind E. coli, we labeled recombinant calreticulin and galectin-3 with the amine-reactive fluorophore TAMRA (5-(and 6-)carboxytetramethylrhodamine) (50µM) and incubated this with heat-inactivated E. coli for 1 h to measure protein binding to the bacteria via flow cytometry. Bacterial fluorescence readings were significantly increased in the presence of TAMRAcalreticulin protein compared to the protein-free control (**Figure 3A**). Similarly, TAMRA-labeled galectin-3 bound to E. coli (**Figure 3A**), indicating that both calreticulin and galectin-3 can bind to bacteria.

We next tested whether calreticulin can opsonize bacteria for microglial phagocytosis. pHrodo-conjugated E. coli were incubated with 500 nM calreticulin for 90 min (followed by several washing steps to remove unbound protein), and this increased their phagocytic removal by microglia by 118% (±29) compared to E. coli not incubated with calreticulin (**Figure 3B**). This opsonization was abolished in the presence of sucrose (**Figure 3C**), consistent with a role for the carbohydraterecognition domain of calreticulin in bacterial binding. Sucrose was not able to significantly inhibit microglial phagocytosis of bacteria in the absence of exogenous calreticulin. This is consistent with the lack of a significant release of calreticulin by non-activated microglia demonstrated previously (**Figure 1A**), and suggests that calreticulin can promote, but is not required, for bacterial clearance by unstimulated microglia.

We have previously shown that calreticulin can opsonise PC12 cells via the microglial phagocytic receptor LRP1, which is inhibited by the LRP1-specific ligand RAP (27). So, to investigate whether exogenous calreticulin mediates microglial phagocytosis of bacteria via LRP1, we tested whether RAP affected the calreticulin-induced microglial phagocytosis of bacteria. We found that treating the cells with RAP inhibited this phagocytosis to control levels (**Figure 3D**). Together, these data suggest that calreticulin opsonized bacteria for microglial phagocytosis via (i) it's carbohydrate-recognition domain and (ii) the microglial LRP1 receptor.

We next tested whether galectin-3 can also opsonize bacteria for microglial phagocytosis. E. coli were incubated with 20 nM galectin-3 (followed by several washes) and added to primary rat microglia for 90 min before measuring phagocytosis. Pre-incubation of the bacteria with galectin-3 enhanced their phagocytosis by microglia by 73% (±24) (**Figure 4A**). Thus, added galectin-3 can also opsonise bacteria. Galectin-3 binds lactose, and lactose can inhibit the ability of galectin-3 to opsonise mammalian cells (9). So, we tested the effect of lactose on our cells. Treating with lactose inhibited microglial phagocytosis of bacteria in the presence or absence of exogenous galectin-3 (**Figure 4B**), suggesting that sugars are involved in this phagocytosis.

It is known that galectin-3 can opsonise mammalian cells by bridging between sugars on the target cells and the phagocytic receptor MerTK to promote microglial phagocytosis of certain cell types (9, 11). To test whether the galectin-3

opsonization of bacteria activates phagocytosis via microglial MerTK, cells were briefly treated (or not) with the MerTKspecific inhibitor UNC569 and bacterial phagocytosis was measured as before. In the presence and absence of exogenous galectin-3, UNC569 strongly inhibited microglial phagocytosis of bacteria (**Figure 4C**), without affecting microglial viability (**Figure 4D**). This is consistent with MerTK being the main phagocytic receptor for microglial phagocytosis of the bacteria in the presence or absence of exogenous galectin-3; and suggests that galectin-3 opsonizes bacteria for microglial phagocytosis via MerTK.

LPS, along with other agonists for microglial TLR4, is known to promote increased phagocytosis of bacteria and other pathogenic species by microglia (28–30). Given LPS induces calreticulin and galectin-3 release from microglia, and that these proteins are capable of opsonizing bacteria for phagocytic removal in culture, we hypothesized that these opsonins may mediate the LPS-induction increase in microglial phagocytosis of bacteria. To investigate this, microglia were stimulated with LPS for 24 h before co-incubation with E. coli for an hour, and phagocytosis was measured as before (**Figure 5A**). LPS did indeed increase microglial phagocytosis

of bacteria by 56% on average (±22%). To test whether this LPS-induced phagocytosis depends on extracellular components (including opsonins), a media swap was performed (removing any factors released by the microglia) immediately prior to bacterial addition and the effect on phagocytosis was determined (**Figure 5B**). Phagocytosis by unstimulated microglia was unaffected by the media swap, but phagocytosis by LPSstimulated microglia was significantly inhibited (p = 0.015) to levels not significantly different from the control (p = 0.944). This indicates that factors released by the microglia in response to LPS are responsible for the increased phagocytosis, consistent with the increased phagocytosis being due to a released

relevant columns.

(C) (note that a statistically significant difference between "–gal-3 – lactose" and "–gal-3 + lactose" was observed). (D) UNC569 (5µM for 2 h) does not promote necrotic death of primary rat microglial cells in culture compared to the "vehicle" control, determined by propidium-iodide staining and quantified via flow cytometry. Manganese chloride (MnCl2, 1 mM) was used as a positive control for necrotic death. Values are means ± SEM of at least three independent experiments. Statistical comparisons were made via one- or two-way ANOVA except for (A), which was by pair-wise student's t-test. NS p ≥ 0.05, \*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001 vs. controls, except where indicated by bars over relevant columns.

opsonin, rather than an increased phagocytic capacity of the microglia themselves.

To test more directly whether extracellular calreticulin mediates LPS-induced microglial phagocytosis of bacteria, microglia were treated for 3 h with a function-blocking anticalreticulin antibody prior to co-incubation with bacteria, and the induction of phagocytosis by LPS was measured. Anticalreticulin was able to inhibit the LPS-induced increase in phagocytosis when compared to a serotype control IgG (p = 0.029, **Figure 5C**). Anti-calreticulin had no effect on phagocytosis in the absence of LPS (**Figure 5C**), consistent with calreticulin being absent in this condition (**Figure 1A**). Together, this indicates that the LPS-induced phagocytosis is at least partly due to the LPS-induced release of calreticulin.

We further tested whether lactose or sucrose could inhibit this LPS induction of phagocytosis. Phagocytosis in the presence of

LPS was significantly inhibited by lactose (p = 0.0012) or sucrose (p < 0.0001) to levels not significantly different from the control in the absence of LPS (**Figure 6A**).

Given that the LRP1 inhibitor RAP was able to inhibit calreticulin-induced opsonization of the E. coli, we tested whether RAP could also inhibit the LPS-induced microglial phagocytosis of the bacteria. We found that RAP significantly inhibited the LPS-induced phagocytosis (p = 0.021, **Figure 6B**). The MerTK-specific antagonist UNC2881 also inhibited LPSinduced microglial phagocytosis of bacteria (p = 0.02; p = 0.26 vs.

the "–LPS – UNC2881" control) (**Figure 6C**), indicating a role for galectin-3 and MerTK. Taken together, these data demonstrate that both calreticulin and galectin-3 are required for the LPSinduction of bacterial phagocytosis, and this is mediated via their carbohydrate-binding domains and the microglial phagocytic receptors LRP1 and MerTK.

# DISCUSSION

As calreticulin has been described alternatively as: (i) an eatme signal and (ii) a phagocytic receptor, and in this paper we find it acts as (iii) an opsonin, it is important to distinguish between use of these terms. An eat-me signal is a signal normally inside the cell that when exposed on the surface of a cell induces phagocytes to phagocytose that cell. The classical eat-me signal is phosphatidylserine, normally found on the inner leaflet of the plasma membrane but exposed on the outer leaflet during apoptosis in order to promote phagocytosis of the apoptotic cell. Calreticulin is normally found within the cell but can translocate to the surface of apoptotic cells to promote phagocytosis of these cells, and therefore can act as an eat-me signal (13). Calreticulin can also translocate to the surface of phagocytes, where it can act as a phagocytic receptor together with LRP1 (31). However, in contrast to typical eat-me signals and phagocytic receptors, calreticulin is a soluble protein, which can potentially dissociate from the surface of cells into the extracellular space, where it has the potential to act as an opsonin. Galectin-3 has a similarly ambiguous status: it has been described as an eatme signal (11), but it is also found in the extracellular space and when binding to the surface of target cells can potentially act as an opsonin (11), and in principle when galectin-3 binds to a phagocytic receptor on a phagocyte it could act as co-receptor.

We found that microglia activated with LPS released calreticulin into the extracellular space. These are two important characteristics of an opsonin, that it can exist as a soluble, extracellular protein, and that it can be released according to need, in this case in response to LPS, a marker of the presence of gram-negative bacteria. Additionally, recombinant calreticulin bound to the surface of the gram-negative bacteria E. coli, and when bound promoted phagocytosis of these bacteria by microglia. Furthermore, LPS-induced phagocytosis of bacteria was at least in part mediated by the released calreticulin. Thus, calreticulin fits the core definition of an opsonin: it binds cells and promotes the phagocytosis of those cells when bound.

So, is calreticulin a eat-me signal, an opsonin or a phagocytic receptor? It meets the criteria for all three, although it would be better described as a co-receptor (with LRP1) than a receptor, as it has no transmembrane signaling capacity. When released from a target cell onto the target cell surface, it acts as an eat-me signal. When released from a phagocyte and binding to a target cell, it acts as an opsonin. When released from a phagocyte onto the surface of a phagocyte and binding to LRP, it can act as a co-receptor (31).

What about galectin-3? We found evidence that: galectin-3 was released from LPS-activated microglia, added galectin-3 could bind E. coli bacteria and stimulate their phagocytosis by microglia, and LPS-induced phagocytosis of bacteria was partly mediated by released galectin-3, indicating that galectin-3 is an opsonin. Galectin-3 has been described as an eat-me signal (11), however, although these authors found that galectin-3 could induce and mediate phagocytosis, they did not find that galectin-3 was released from the target cells onto their surface. Thus, they did not provide evidence that galectin-3 can act as an eat-me signal, rather than as an opsonin. We (9) reported that galectin-3 was released from LPS-activated microglia, and the galectin-3 could bind to PC12 cells only when these cells were desialylated, and when bound could opsonise these cells for phagocytosis by microglia. Galectin-3 preferentially binds to N-acetyl-lactosamine residues on glycoproteins and glycolipids, but this binding is normally blocked by terminal sialic acid residues on mammalian cells (32). On bacteria, galectin-3 normally binds the sugar residues of LPS, although for some bacterial species the binding of galectin-3 to LPS is not mediated by sugar residues (33). Galectin-3 opsonized cells may stimulate microglial phagocytosis via binding to microglial MerTK (9). However, we (34) recently reported that galectin-3 can also bind the phagocytic receptor TREM2, and thus galectin-3 could potentially opsonise via TREM2 on microglia, but this requires further investigation. Galectin-3 can also act as an alarmin via activating TLR4 on microglia (8). Thus, like calreticulin, galectin-3 can have multiple adaptive roles in orchestrating the response to LPS or gramnegative bacteria.

What is the relevance of this work to the in vivo situation, and what is the translational potential of this work for treatment of disease? Calreticulin can be found in the serum of healthy humans (at 5 ng/ml), and increases with inflammatory disease (35). Thus, serum calreticulin has the potential to opsonise bacteria in vivo, and the finding of calreticulin bound to LPS in the sera of patients with chronic bacterial infections suggests that is does (26).

Calreticulin is also found in human CSF (36), and thus has the potential to opsonise bacteria in the brain. Macrophages have been shown to release calreticulin that opsonizes mammalian cells for phagocytosis by the macrophages (37), supporting the possibility that this may occur also with bacteria as targets. Phagocytosis of cancer cells induced by calreticulin on their surface has been shown to result in presentation of antigens from the cancer cell on the phagocyte together with MHCII, resulting in an adaptive immune response to the cancer (14, 15). Indeed, calreticulin has been used as an adjuvant to induce an adaptive response to antigen (38). Thus, phagocytosis of bacteria opsonized with calreticulin might in principle result in presentation of antigens from the bacteria by phagocytes to T cells, but this would have to be tested. If this is correct, it may be worth testing whether calreticulin injections during a bacterial infection help clear the infection by (i) opsonizing the bacteria, and (ii) enhancing phagocytosis and presentation of antigens from the bacteria.

Galectin-3 is also found in human plasma and CSF, and levels increases with disease (39, 40). Galectin-3 levels also increase in mouse and humans during fungal infection and play roles in reducing infection (41). The interaction of galectin-3 with LPS from different species of bacteria enhances the binding and interaction of LPS with neutrophils (33). Galectin-3 has been shown to limit infections by the gram-negative bacteria Helicobacter pylori (42), but also limits infection by gram-positive Streptococcus pneumoniae partly by increasing neutrophil phagocytosis of the bacteria (43). Thus, there is at least indirect evidence that galectin-3 may limit bacterial infections in vivo, potentially by acting as an opsonin, but this requires further investigation. If so, it may be worth testing whether galectin-3 injections during a bacterial infection help clear the infection by opsonizing the bacteria. As with calreticulin, there is also the possibility that injection of dead bacteria opsonized with galectin-3 might induce phagocytosis and antigen presentation by antigen-presenting cells, promoting an adaptive immune response.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

#### ETHICS STATEMENT

The animal study was reviewed and approved by Cambridge University Local Research Ethics Committee. All experiments were performed in accordance with the UK Animals (Scientific Procedures) Act (1986).

#### AUTHOR CONTRIBUTIONS

All experimental work was undertaken by TC. Animal maintenance, instruction and assistance for primary cell culturing was provided by MP. Study design and data analysis was done by TC and GB. The manuscript was written by TC and GB.

#### FUNDING

This work was funded by the Biotechnology & Biological Sciences Research Council UK and the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No. 115976 (PHAGO consortium).

### REFERENCES


## ACKNOWLEDGMENTS

The authors would like to thank David Allendorf, Claire Butler, Miguel Burguillos, and Alma Popescu for their scientific ideas, experimental suggestions, and general support.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02647/full#supplementary-material

Figure S1 | Microglial calreticulin and galectin-3 are upregulated following LPS stimulation. Calreticulin (A) and galectin-3 (B) mRNA were quantified in BV2 microglial cells treated with or without LPS (100 ng/ml) for 24 h, and expression normalized to actin mRNA levels. Values are means ± SEM of at least three independent experiments. Statistical comparisons were made via student's t-test. <sup>∗</sup>p < 0.05, ∗∗p < 0.01.

Figure S2 | BV2 microglia rapidly phagocytose pHrodo-conjugated E. coli bacteria in vitro. Significant levels of phagocytosis were detected within 60 min (compared to the "0 min" control) with phagocytic saturation reached between 120 and 180 min. Values are means ± SEM of at least three independent experiments. Statistical comparisons were made via one-way ANOVA. <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.


Acta Neuropathol. (2019) 2:251–73. doi: 10.1007/s00401-019-0 2013-z


**Conflict of Interest:** 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 Cockram, Puigdellívol and Brown. 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.

# Thioredoxin-Interacting Protein Promotes Phagosomal Acidification Upon Exposure to Escherichia coli Through Inflammasome-Mediated Caspase-1 Activation in Macrophages

Sung-Jin Yoon1†, Dong Hyun Jo2†, Seung-Ho Park 1†, Jun-Young Park <sup>1</sup> , Yoo-Kyung Lee<sup>1</sup> , Moo-Seung Lee<sup>1</sup> , Jeong-Ki Min<sup>3</sup> , Haiyoung Jung<sup>4</sup> , Tae-Don Kim<sup>4</sup> , Suk Ran Yoon<sup>4</sup> , Su Wol Chung<sup>5</sup> , Jeong Hun Kim2,6, Inpyo Choi <sup>4</sup> \* and Young-Jun Park <sup>1</sup> \*

#### Edited by:

Florence Niedergang, Centre National de la Recherche Scientifique (CNRS), France

#### Reviewed by:

Roberta Olmo Pinheiro, Oswaldo Cruz Foundation (Fiocruz), Brazil Sophie Dupré-Crochet, Université Paris-Sud, France

#### \*Correspondence:

Inpyo Choi ipchoi@kribb.re.kr Young-Jun Park pyj71@kribb.re.kr

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Microbial Immunology, a section of the journal Frontiers in Immunology

Received: 04 June 2019 Accepted: 24 October 2019 Published: 12 November 2019

#### Citation:

Yoon S-J, Jo DH, Park S-H, Park J-Y, Lee Y-K, Lee M-S, Min J-K, Jung H, Kim T-D, Yoon SR, Chung SW, Kim JH, Choi I and Park Y-J (2019) Thioredoxin-Interacting Protein Promotes Phagosomal Acidification Upon Exposure to Escherichia coli Through Inflammasome-Mediated Caspase-1 Activation in Macrophages. Front. Immunol. 10:2636. doi: 10.3389/fimmu.2019.02636 <sup>1</sup> Environmental Disease Research Center, Daejeon, South Korea, <sup>2</sup> Fight Against Angiogenesis-Related Blindness (FARB) Laboratory, Clinical Research Institute, Seoul National University Hospital, Seoul, South Korea, <sup>3</sup> Biotherapeutics Translational Research Center, Daejeon, South Korea, <sup>4</sup> Immunotherapy Research Center, Korea Research Institute of Bioscience and Biotechnology (KRIBB), Daejeon, South Korea, <sup>5</sup> School of Biological Sciences, College of Natural Sciences, University of Ulsan, Ulsan, South Korea, <sup>6</sup> Department of Biomedical Sciences, Seoul National University College of Medicine, Seoul, South Korea

In host defense, it is crucial to maintain the acidity of the macrophage phagosome for effective bacterial clearance. However, the mechanisms governing phagosomal acidification upon exposure to gram-negative bacteria have not been fully elucidated. In this study, we demonstrate that in macrophages exposed to Escherichia coli, the thioredoxin-interacting protein (TXNIP)-associated inflammasome plays a role in pH modulation through the activated caspase-1-mediated inhibition of NADPH oxidase. While there was no difference in early-phase bacterial engulfment between Txnip knockout (KO) macrophages and wild-type (WT) macrophages, Txnip KO macrophages were less efficient at destroying intracellular bacteria in the late phase, and their phagosomes failed to undergo appropriate acidification. These phenomena were associated with reactive oxygen species production and were reversed by treatment with an NADPH oxidase inhibitor or a caspase inhibitor. In line with these results, Txnip KO mice were more susceptible to both intraperitoneally administered E. coli and sepsis induced by cecum ligation and puncture than WT mice. Taken together, this study suggests that the TXNIP-associated inflammasome-caspase-1 axis regulates NADPH oxidase to modulate the pH of the phagosome, controlling bacterial clearance by macrophages.

Keywords: thioredoxin-interacting protein, Escherichia coli, caspase, phagosome, macrophage

# INTRODUCTION

Phagocytosis by professional phagocytes, including macrophages, neutrophils, and monocytes, is a crucial process for host defense against bacterial pathogens. Phagocytosis occurs in two distinct phases, namely, bacterial internalization followed by phagosomal maturation (1). To allow for sufficient clearance of pathogenic microorganisms, phagocytes should be able to detect, engulf, and

**33**

ultimately kill the pathogens (2). During phagosomal maturation, the phagosomes interact with endosomes and lysosomes, changing the phagosomal protein composition (3, 4). For example, phagosomes fuse with early endosomes and acquire the small GTPase Rab5 (5). The transition from early to late phagosomes is marked by the conversion from Rab5 to Rab7 (6). Following this, phagosomes acquire lysosomal-associated membrane protein (LAMP)-1 and LAMP-2, which are required for phagolysosomal fusion (4). The further fusion of phagosomes with lysosomes produces phagolysosomes enriched in hydrolytic enzymes, reactive oxygen species (ROS), and antimicrobial peptides, which are required for bacterial killing (4, 7).

During phagosomal maturation, the phagosomes undergo progressive acidification of the lumen, which is mainly achieved through proton pumping by the vacuolar-type H+-ATPase (V-ATPase) (1, 8). As phagosomes mature from early to late phagosomes, luminal pH gradually decreases from being mildly acidic (pH 6.1–6.5) to more acidic (pH 5.5–6.0) (1, 9). Furthermore, the phagolysosome is characterized by a highly acidic luminal pH (as low as 4.5) (1). In addition to its role in bactericidal activity through the generation of ROS, the activity of NADPH oxidase counteracts the activity of V-ATPase, tending toward neutralizing phagosomal pH (10–12). To ensure efficient bactericidal activity, it is therefore important to regulate the activity of proteins such as V-ATPase and NADPH oxidase in the phagosome.

One of the mechanisms responsible for regulating the activity of NADPH oxidase is the inflammasome (8). Inflammasomes are multiprotein complexes that activate caspase-1 and consist of nucleotide-binding oligomerization domain-like receptors (NLRs), such as NLR family pyrin domain containing 3 (NLRP3) and NLR family CARD domain containing 4 (NLRC4) (13). Inflammasome-activated caspase-1 causes the proteolytic processing of pro-interleukin (IL)-1β and pro-IL-18, which results in the secretion of functional IL-1β and IL-18 (14). Similarly, caspase-1 activated by the NLRP3 inflammasome accumulates in phagosomes and modulates buffering through the NADPH oxidase NOX2 to control the pH upon exposure to gram-positive bacteria (8). However, the molecular pathway regulating inflammasome-mediated NADPH oxidase has not been clearly identified.

Recent studies have shown that thioredoxin-interacting protein (TXNIP) activates the NLRP3 inflammasome under various conditions (15–18). TXNIP, first described as vitamin D3 upregulated protein 1 in acute promyelocytic leukemia HL-60 cells (19), is known to be a regulator of oxidative stress, acting as an inhibitor of the activity of thioredoxin (20, 21). In addition to these well-known functions, the dissociation of TXNIP from thioredoxin allows TXNIP to bind to NLRP3 and activate the NLRP3 inflammasome in a ROS-sensitive manner (18). Because of this, we speculated that TXNIP could be a modulator of inflammasome-mediated pH control in macrophages upon exposure to bacterial pathogens.

In this study, we investigated the roles of TXNIP in modulating the phagosomal acidity in macrophages that facilitates the destruction of bacteria through activation of the NLRP3 inflammasome-caspase-1 pathway. We first examined the clearance of bacteria by macrophages obtained from Txnip wild-type (WT) and knockout (KO) mice. Then, we examined the recruitment of proteins to the phagosomes, the pH of the phagosomal lumen, and the ROS levels in Txnip WT and KO macrophages upon treatment with bacteria. To decipher the pathways involved, specific inhibitors of the phosphoinositide 3-kinases (PI3K)/Akt pathway, V-ATPase, and caspases were employed. Based on our findings, we propose that the TXNIP-NLRP3 inflammasome-caspase-1 regulates NADPH oxidase to induce the acidification of phagosomes to clear bacteria in macrophages.

### MATERIALS AND METHODS

#### Animals

The animal study was approved by the Institutional Animal Care and Use Committee of the Korea Research Institute of Bioscience and Biotechnology (KRIBB-IACUC, approval number: KRIBB-AEC-11044). All procedures were performed in accordance with guidelines regarding the use of laboratory animals (National Institutes of Health). WT C57BL/6 mice were obtained from the Korea Research Institute of Bioscience and Biotechnology, and Txnip KO mice were prepared as previously described (16). All mice were housed in a pathogen-free animal facility under a standard light-dark cycle with standard rodent chow and water provided ad libitum. The experimental groups were all ageand sex-matched.

#### Preparation of Peritoneal Macrophage

Mouse peritoneal macrophages were harvested and cultured as described previously (16). Cells were harvested 4 days after intraperitoneal injection of 3% thioglycollate (Sigma). Macrophages were washed and plated in a 24-well-plate at 5 × 10<sup>5</sup> cells per well. After incubation with serum-free RPMI medium for 2 h at 37◦C, the wells were washed three times to remove non-adherent cells, and the culture medium was replaced with RPMI supplemented with 10% fetal bovine serum (FBS) and 1% Antibiotic-Antimycotic (Thermo Fisher).

#### Phagocytosis Assay

Mouse peritoneal macrophage cells were plated in 24-well-plates at 5 × 10<sup>5</sup> cells per well and incubated with GFP-expressing E. coli, which were prepared as previously described (22). For the detection of engulfment, cells were incubated with GFPexpressing E. coli at indicated ratios (macrophage:bacteria CFU) at 37◦C for 30 min. After the incubation, cells were washed three times with cold PBS to remove remaining bacteria, and the cells were scraped. For the detection of remaining bacteria in mouse peritoneal macrophages, cells were incubated with GFP-expressing E. coli at 37◦C for 1 h and washed five times with cold PBS. Then, the culture medium was replaced with RPMI supplemented with 10% FBS, 1% Antibiotic-Antimycotic (Thermo Fisher), and 10µg/ml gentamicin to inhibit the growth of extracellular bacteria for the indicated periods. Cells were analyzed immediately using a FACSCanto II flow cytometer (BD), and the data were processed using the FACSDiva software (BD). For the treatment of inhibitors, cells were incubated with 10µM wortmannin (Selleckchem) or 20 nM bafilomycin A (Selleckchem) for 30 min before the addition of bacteria. For the phagosomal maturation assay using pHrodoTM Red E. coli Bioparticles, cells were plated in 48-well-plates at 2 × 10<sup>5</sup> cells per well and incubated with pHrodoTM Red E. coli Bioparticles at 20 µg per well at the indicated periods. After the incubation, cells were washed three times with cold PBS and then immediately analyzed using a FACSCanto II flow cytometer (BD).

#### Immunostaining

Cells were immunostained as previously described (22). Peritoneal macrophages (1 × 10<sup>5</sup> cells per well) were plated on round glass coverslips in 24-well-plates and incubated with bacteria multiplicity of infection (MOI) of 10. For the phagocytosis of yellow-green fluorescent FluoSpheres beads of size 2.0µm (Thermo Fisher), peritoneal macrophages were plated on round glass coverslips in 24-well-plates and incubated with 5 × 10<sup>5</sup> beads/ml per well for 1 h at 37◦C. After incubation, the cells were washed with cold PBS, fixed for 15 min at room temperature (RT) in 4% paraformaldehyde, and then washed again with cold PBS. Before staining with primary antibodies, cells were permeabilized for 10 min at RT in 0.2% Triton X-100 in PBS and incubated overnight at 4◦C with primary antibodies specific for Lamp1 (Abcam) as indicated. Cells were then washed with PBS and incubated for 2 h at RT with Alexa Fluor 555-conjugated donkey-anti-rabbit IgG (Thermo Fisher). Nuclei were stained with 4′ ,6-diamidino-2-phenylindole (Thermo Fisher). The cells were imaged using a ×60 objective and an IX81 inverted microscope (Olympus). Images were obtained using the DP30BW digital camera (Olympus) and X-Cite <sup>R</sup> 120 XL light source. The acquired images were analyzed using Metamorph 7.1 program (Molecular Devices). To count the yellow-green fluorescent FluoSpheres beads, four areas of each image field of bead-containing macrophages were analyzed. For the inhibition of NADPH oxidase or caspase-1, cells were incubated with GFP-expressing bacteria for 1 h at 37◦C and then incubated with 5µM diphenyleneiodonium (DPI; Selleckchem) and 10µM Z-VAD (Enzo) with 10µg/ml gentamicin for 6 h at 37◦C.

#### Gentamicin Protection Assay

The survival of bacteria was determined with the treatment of gentamicin as previously described (23). Briefly, mouse peritoneal macrophages were incubated with E. coli or GFPexpressing E. coli for 1 h, and then the medium was replaced with the one containing 100µg/ml gentamicin to kill extracellular bacteria. For treatment with inhibitors, the cells were incubated with 20 nM bafilomycin A (Selleckchem) for 30 min before the addition of bacteria. After 1 h, the medium was changed with the fresh one containing 10µg/ml gentamicin at the time. The cells were washed with 1X PBS and lysed with 0.5% Triton X-100 in sterile water for 15 min at RT. Finally, the extract was plated directly onto LB agar plates and incubated at 37◦C overnight.

#### Isolation of Phagosomes

Phagosomes from macrophages were isolated as previously described (9). Briefly, after the incubation of peritoneal macrophages with E. coli, the cells were washed in cold PBS, pelleted, resuspended in 1 ml of homogenization buffer (250 mM sucrose, 3 mM imidazole, pH 7.4), and homogenized on ice using a Dounce homogenizer. Phagosomes were then isolated by flotation in a sucrose step gradient during centrifugation for 1 h at 100,000 g at 4◦C. The fraction was then collected from the interface of the 10% and 25% sucrose solutions and resuspended in RIPA buffer.

# Western Blotting Analysis

For Western blotting, the protein extracts were prepared by resuspending cells or phagosomes in the lysis buffer [50 mM Tris, pH 7.5, 1 mM ethylenediaminetetraacetic acid, 150 mM NaCl, 0.1% sodium dodecyl sulfate (SDS), and 1% NP-40] containing a 1X protease inhibitor cocktail solution and 1X phosphatase inhibitor cocktail. Proteins were separated via SDSpolyacrylamide gel electrophoresis (PAGE) and subsequently transferred to Immobilon-P membranes (Millipore). Primary antibodies specific for Rab5 (catalog no. 3547), Rab7 (catalog no. 9367), caspase-1 (catalog no. 24232), and GAPDH (catalog no. 2118) were purchased from Cell Signaling. Primary antibodies specific for Lamp1 (catalog no. sc-19992) and β-actin (catalog no. sc-47778) were purchased from Santa Cruz. Primary antibodies specific for TXNIP (catalog no. K0205-3) were purchased from MBL. Primary antibodies specific for V-ATPase (catalog no. GTX110815) were purchased from GeneTex. The densitometry analysis of Western blot was carried out using Image Lab 6.0.1 (Bio-Rad).

## ROS Detection

For the detection of the amount of superoxide, mouse peritoneal macrophages were incubated with GFP-expressing E. coli (MOI of 20) at 37◦C for 1 h, washed three times with cold PBS, and incubated with 1µM dihydroethidium (DHE, Thermo Fisher) at 37◦C for 20 min. For the demonstration of the effects of the clearance after the engulfment, cells were incubated with GFP-expressing E. coli (MOI of 20) at 37◦C for 1 h, washed three times with cold PBS, and then incubated with 100µg/ml gentamicin to remove the extracellular bacteria. After 1 h, the culture medium was replaced with RPMI supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, and 10µg/ml gentamicin at the indicated periods. After 6 h of incubation of bacteria with macrophage, cells were incubated with 1µM dihydroethidium (DHE, Thermo Fisher) at 37◦C for 20 min. The GFP-positive population represented the cells engulfing bacteria. The fluorescence intensities of both GFP-positive and -negative populations were analyzed to estimate the amounts of superoxide. To measure the ROS production after 6 h of phagocytosis, mouse peritoneal macrophages were incubated with bacteria, not expressing GFP, at the indicated periods and washed three times with cold PBS. Then, cells were incubated with 2µM H2DCFDA (Thermo Fisher) at 37◦C for 30 min. Data were analyzed using a FACSCanto II flow cytometer (BD) and the FlowJo software (BD). For the kinetic assay of ROS production, mouse peritoneal macrophages were incubated with 1µM dihydroethidium (DHE, Thermo Fisher) and 2µM H2DCFDA (Thermo Fisher) at 37◦C for 20 min, washed three times with PBS, and incubated with E. coli. The fluorescence was detected using SpectraMax <sup>R</sup> iD3 (Molecular Devices) and analyzed using SoftMax Pro 7.0.3 (Molecular Devices).

#### pH Detection

Mouse peritoneal macrophages were plated in 24-well-plates at 5 × 10<sup>5</sup> cells per well, incubated with the pHrodoTM Red E. coli Bioparticles (Thermo Fisher) at 37◦C for 30 min, and then washed three times with the cold Live Cell Imaging Solution (Thermo Fisher). pH values were compared with a standard curve obtained by the incubation with standard pH buffers (ranging from pH 4.5–7.7, Thermo Fisher). The cells were immediately analyzed using a FACSCanto II flow cytometer (BD).

#### ELISA

Mouse peritoneal macrophages were incubated with bacteria at 37◦C for the indicated periods, and then the culture medium was harvested. The amounts of IL-1β were analyzed using the DuoSet ELISA (R&D). After the treatment of the TMB substrate reagent (Thermo Fisher), absorbance was measured at 450 nm using the SpectraMax iD3 Multi-Mode Microplate Reader (Molecular Devices).

### Caspase-1 Activity Assay

The activity of caspase-1 was assessed using the Caspase-1/ICE Colorimetric Protease Assay Kit (cat. no. ALX-850-211, ENZO), as recommended by the manufacturer. Briefly, mouse peritoneal macrophage protein extracts were prepared by the Cell lysis buffer contained in the Assay Kit. The protein extracts (100 µg per each sample) were incubated with 200µM YVAD-pNA substrate at 37◦C for 2 h, and the absorbance at 405 nm was read using the SpectraMax iD3 Multi-Mode Microplate Reader (Molecular Devices).

#### Bacterial Infection

For bacterial infection, mice were injected intraperitoneally with live E. coli (5 × 10<sup>6</sup> CFUs/g) at the age of 8–10 weeks (16). For the cecal ligation and puncture experiments, the cecum was ligated at half the distance between the distal pole and the base of the cecum. Cecal puncture (through-and-through) was performed from mesenteric toward antimesenteric direction after the ligation (24, 25).

#### TUNEL Assay

The TUNEL assay was performed as previously described (22). Briefly, the tissues were fixed in formalin and embedded in paraffin prior to sectioning. The TUNEL assay was performed using the DeadEnd colorimetric TUNEL System (Promega), according to the manufacturer's instructions.

#### Statistical Analysis

The statistical significance of differences in mean values was calculated using unpaired, two-tailed Student's t-test. The statistical significance of the effect for time and treatments was analyzed using two-way ANOVA. The statistical significance was determined using GraphPad software version 8.1.1 (Prism, La Jolla, CA). A p < 0.05 was considered statistically significant.

#### RESULTS

#### TXNIP Regulates the Clearance of Bacteria From Macrophages

Macrophages are professional phagocytes that engulf large particles (≥0.5µm) including microorganisms and are capable of destroying pathogens (1). To investigate the roles of TXNIP in the regulation of phagocytosis, peritoneal macrophages were isolated from WT and Txnip KO mice. There was no difference in the engulfment of yellow-green fluorescent FluoSpheres beads or of GFP-expressing E. coli in both types of macrophages at 1 h (**Figures 1A,B**). To evaluate the effects of TXNIP on phagosome maturation, the fluorescence of the GFP-expressing E. coli in the macrophages was measured by flow cytometry and confocal microscopy at different times after bacterial uptake. To do this, after 1 h of treatment with bacteria, the medium was replaced with fresh medium to remove any extracellular bacteria. Interestingly, Txnip KO macrophages showed higher fluorescence levels of GFP-expressing E. coli than WT macrophages (**Figure 1C**; **Supplementary Figure 1A**). Furthermore, immunofluorescence microscopy also showed that Txnip KO macrophages retained more GFP-expressing E. coli than WT macrophages (**Figure 1D**). Based on these findings, we performed a phagocytosis assay using live imaging. In these data, we can observe that the reduction of the fluorescent bacteria is higher in the WT than in the Txnip KO macrophages (**Supplementary Video 1**). Consistent with these findings, the levels of remaining extracellular bacteria were significantly higher in Txnip KO macrophages than in WT macrophages (**Supplementary Figure 1B**). These results indicated that Txnip KO macrophages are less able to clear E. coli after engulfment compared to WT macrophages. To make sure, we assessed whether TXNIP could regulate the number of engulfed bacteria in macrophages. The gentamycin protection assay revealed that the number of bacterial cell colony-forming units (CFUs) present over time in extracts of bacterially exposed Txnip KO macrophages was higher than in WT macrophages (**Figures 1E,F**). These results indicate that TXNIP has an essential role in regulating the clearance of bacteria by macrophages.

#### TXNIP Controls the Acidification of Phagosomes in Macrophages After the Engulfment of Bacteria

Rab5 and Rab7 are markers for early and late phagosomes, respectively, whereas LAMP-1 and V-ATPase are markers for phagolysosome fusion (4, 26, 27). Although Txnip KO macrophages showed a decreased clearance of bacteria after engulfment, there were no differences in the protein expression levels of Rab7, LAMP-1, and V-ATPase in phagosomes isolated from WT and Txnip KO macrophages after treatment with bacteria. However, Txnip KO macrophages recruited less Rab5 into the phagosomes at 1 and 2 h after the treatment of bacteria (**Figure 2A**; **Supplementary Figure 2**), although there were no differences in total Rab5 expression in whole-cell lysates of Txnip WT and KO macrophages (**Supplementary Figures 3A,B**).

FIGURE 1 | that engulfed GFP-expressing E. coli at 1 h after incubation with bacteria assessed at different macrophage to bacteria [colony-forming units (CFUs)] ratios. (C) The proportion of WT and Txnip KO peritoneal macrophages that retained GFP-expressing E. coli at the indicated time points after treatment with bacteria at a multiplicity of infection (MOI) of 20 (\*\*\*P < 0.001 compared with WT). (D) Representative images of WT and Txnip KO mouse peritoneal macrophages at the indicated time points after treatment with GFP-expressing E. coli at an MOI of 10. Scale bars, 20µm. (E) Representative LB agar plates after overnight incubation with cell extracts derived from WT and Txnip KO peritoneal macrophages incubated with E. coli for the indicated times. (F) CFUs on LB agar plates after overnight incubation with cell extracts derived from WT and Txnip KO mouse peritoneal macrophages incubated with E. coli for the indicated times. Data are expressed as the mean ± SD (n = 3, \*\*P < 0.01, \*\*\*P < 0.001 compared with WT).

These results suggest that TXNIP might be involved in the recruitment of Rab5 to the phagosomes in the early phase (1–2 h after the treatment of bacteria). To investigate whether these effects on differential Rab5 expression were related to the clearance of the bacteria in the late phase (4– 6 h after infection with bacteria as shown in **Figures 1C–F**; **Supplementary Figures 1A,B**), we pretreated macrophages with wortmannin (**Supplementary Figure 3C**), an inhibitor of the PI3K/Akt pathway, which modulates Rab5 recruitment to phagosomes (26). Interestingly, wortmannin decreased the engulfment of bacteria in both Txnip WT and KO macrophages but did not affect the different patterns of bacterial clearance between both cell types (**Figure 2B**).

As stated above, WT and Txnip KO macrophages showed similar patterns of recruitment of Rab7, LAMP-1, and V-ATPase to the phagosomes (**Figure 2A**). On the other hand, we found that there was a striking difference in phagosome maturation between WT and Txnip KO macrophages in terms of acidification (**Figures 2C,D**). To assess phagosomal maturation, macrophages were incubated with pHrodo Red E. coli Bioparticles, which is an indicator for phagolysosomal maturation, for the indicated times. The data showed that WT macrophages had higher levels of phagolysosomal maturation compared with Txnip KO macrophages (**Figure 2C**). The pH, which indicates acidification in the phagosomes, was also significantly higher in phagosomes from Txnip KO macrophages than in phagosomes from WT macrophages (**Figure 2D**). V-ATPases are known to play essential roles in the acidification of phagosomes (27). As shown in **Figure 1C**, we assessed whether V-ATPase activity inhibition could reduce the number of engulfed bacteria in WT macrophages compared to that in Txnip KO macrophages. When WT macrophages were pretreated with bafilomycin A (**Supplementary Figure 3C**), an inhibitor of V-ATPase (28, 29), they showed dysfunction in bacteria clearance with a definite defect in destroying, but not reducing the number of engulfed GFP-expressing E. coli (**Figure 2E**; **Supplementary Figures 4A,B**). Similar to that in WT macrophages, V-ATPase inhibition also totally prevents the bacteria clearance upon phagocytosis in Txnip KO macrophages (**Figure 2E**; **Supplementary Figures 4A,B**). Our observation suggests that V-ATPase inhibition halts the clearance of engulfed bacteria and not only reduced it. So, the data suggest that pathways regulating the acidification of phagosomes other than V-ATPase might be impaired in these cells. Nevertheless, an immunofluorescence analysis for the staining of Lamp1, known as a late phagosomal marker, showed that the bacteria are more fluorescent in the Txnip KO macrophages. In contrast, it is difficult to distinguish their shape in the WT macrophage (**Figure 2F**). These data suggest that TXNIP can regulate E. coli survival in the phagosomes by lowering the phagosomal pH.

### TXNIP Regulates the Level of Superoxide in Macrophages After the Engulfment of Bacteria

ROS produced by NADPH oxidase causes active and stable alkalization of the phagosomal lumen (10, 12), counteracting the activity of V-ATPase (8). Because the total amount of ROS produced during phagocytosis immediately increased upon bacterial infection, we assessed the kinetics of ROS production in WT and Txnip KO macrophages upon phagocytosis of bacteria during 1 h 30 min. After 30 min of treatment with ROSdetection dye, peritoneal macrophages were incubated with E. coli, and then the ROS level was measured at the indicated times. The results of kinetics showed a similar increase in ROS production between WT and Txnip KO macrophages (**Supplementary Figures 5A,B**). Thus, we conclude that TXNIP is not relevant for ROS production in the initial stage of phagocytosis. Next, we conducted a FACS analysis to determine whether the level of superoxides was affected by TXNIP and analyzed the levels of superoxides in WT and Txnip KO macrophages after the internalization of E. coli. After 1 h of treatment with GFP-expressing E. coli, we measured ROS levels in WT and Txnip KO macrophages. As in earlier experiments, there was no difference in the proportion of GFP-positive cells in WT and Txnip KO cells (**Supplementary Figure 5C**). Upon infection with bacteria, ROS levels increased in both WT and Txnip KO macrophages in terms of the mean fluorescence intensities (MFIs) of dihydroethidium (DHE) and total H2DCFDA dyes (**Supplementary Figures 5D–F**). It is noteworthy that GFPnegative cells showed similar intensities in DHE compared to control cells. In contrast, the proportion of bacteria-laden macrophages was higher in Txnip KO macrophages than in WT macrophages 6 h after infection with bacteria (**Figure 3A**). Also, we have measured the ROS production after 6 h of phagocytosis. The production of ROS increased to higher levels in Txnip KO macrophages than in WT macrophages at 6 h (**Figures 3B–D**). Although we did not continuously measure the level of total ROS from initiation of phagocytosis to late time points, our results showed that the level of ROS production increased more in Txnip KO macrophages at the late time. Thus, these results indicated that TXNIP might regulate NADPH oxidase-induced superoxide production following infection of macrophages by E. coli upon phagocytosis at the late time.

To investigate the effects of NADPH oxidase on the removal of bacteria, we treated macrophages with DPI to inhibit NADPH oxidase (**Supplementary Figure 6A**). DPI treatment decreased

treatment with E. coli for the indicated times. (B) The proportion of WT and Txnip KO peritoneal macrophages, treated with or without wortmannin (10µM), which retained (Continued) FIGURE 2 | GFP-expressing E. coli after exposure to GFP-expressing E. coli at a multiplicity of infection (MOI) of 20 for the indicated times. (n = 3) (C) The proportion of WT and Txnip KO mouse peritoneal macrophages showing PE-positive phagosomes after treatment with pHrodoTM Red E. coli Bioparticles for the indicated times. Data are expressed as the mean ± SD (n = 3, \*\*P < 0.01, \*\*\*P < 0.001 compared with WT). (D) Estimated phagosomal pH in WT and Txnip KO mouse peritoneal macrophages after treatment with pHrodoTM Red E. coli Bioparticles for the indicated times. (E) The proportion of WT and Txnip KO mouse peritoneal macrophages, treated with bafilomycin A (20 nM), which retained GFP-expressing E. coli after incubation with GFP-expressing E. coli at an MOI of 20 for the indicated times (n = 3). (F) Representative images of bacteria-laden WT and Txnip KO mouse macrophages 6 h after a 1-h treatment with GFP-expressing E. coli and removal of extracellular bacteria. Scale bar, 10µm.

the proportion of Txnip KO macrophages containing GFPexpressing bacteria (**Figure 3E**). There were no differences in the CFUs present in cell extracts from WT and Txnip KO macrophages after 1 h of bacteria treatment (**Figure 3F**). On the other hand, the CFUs present in cell extracts from Txnip KO cells were higher than those in WT cells at 3 and 6 h after infection with bacteria (**Figure 3G**). Interestingly, the CFUs present in both the WT and Txnip KO cell extracts were decreased by DPI treatment. In addition, there was a change in bacterial shape in the phagosomes of DPI-treated Txnip KO macrophages, indicative of bacterial degradation (**Supplementary Figure 6B**). In line with these results, DPI treatment after 2 h of phagocytosis reduced the level of DHE dye in both WT and Txnip KO cells after bacterial treatment (**Figures 3H,I**). These data indicate that decreasing superoxide levels by inhibiting NADPH oxidase at late stages of phagocytosis restores the clearance of engulfed bacteria in Txnip KO macrophages.

## The TXNIP-NLRP3 Inflammasome-Caspase-1 Axis Regulates the Clearance of Bacteria From Macrophages

Having shown that TXNIP regulates ROS productions by controlling NADPH oxidase, we next tried to determine the mechanism by which TXNIP controls the NLPR3 inflammasome. The interaction between TXNIP and NLRP3 causes caspase-1 activation and IL-1β secretion, and the lack of TXNIP inhibits the formation of the NLRP3-ASC-caspase-1 complex by nitrosylation (15, 16, 18, 30). Therefore, TXNIP is crucial to control the activity of the NLRP3-ASC-caspase-1 complex both directly and indirectly. During bacterial infection, it has been demonstrated that the activation of caspase-1 by NLRP3 inflammasome inhibits the action of NADPH oxidase (8). We assessed whether TXNIP could regulate the activity of NLPR3 inflammasomes following bacterial infection. Treatment with E. coli increased IL-1β secretion in both WT and Txnip KO cells, but the magnitude was lesser in Txnip KO cells than in WT cells (**Figure 4A**). Similarly, Txnip KO macrophages expressed lower levels of the activated form of caspase-1 (Casp1 p10) than WT macrophages (**Figure 4B**; **Supplementary Figure 7**). To assess the contribution of caspase-1 in phagosome maturation, we used the caspase-1 inhibitor, ZVAD. Following bacterial infection, caspase-1 activity increased in WT macrophages but not in Txnip KO macrophages. This increased caspase-1 activity seen in WT macrophages was completely ablated by ZVAD treatment (**Figure 4C**). Under the same conditions, ZVADtreated WT macrophages retained the same level of intact GFP-expressing bacteria as Txnip KO macrophages (**Figure 4D**). Together, these data suggest that TXNIP improves phagosome maturation by the activation of caspase-1 through regulation of the inflammasome.

#### Bacterial Clearance Is Reduced in Txnip KO Mice

To confirm that the delay in bacterial clearance caused by TXNIP loss contributed to the death of mice, we performed an intraperitoneal injection of E. coli (10<sup>8</sup> CFU) in WT and Txnip KO mice. Txnip KO mice were more susceptible to intraperitoneally administered E. coli than WT mice (**Figure 5A**). The CFUs present in the blood, liver, and spleen of WT and Txnip KO mice challenged with bacteria were markedly higher in Txnip KO mice than in the same organs from WT mice (**Figure 5B**). In addition, in the Txnip KO mice, splenomegaly was more evident (**Figure 5C**), and the level of apoptosis in the liver parenchyma was higher (**Figure 5D**). We further assessed the role of TXNIP on mouse survival in other models of bacterial challenge. The cecum ligation and puncture (CLP) model is the most widely utilized model of sepsis (24, 25). Similar to the intraperitoneal injection model, Txnip KO mice were more susceptible to CLP (**Figure 5E**), displaying a similar mortality pattern. In addition, the CFUs present in the blood and extracts from liver and spleen were also higher in Txnip KO mice (**Figure 5F**). These data suggest that TXNIP has an essential role in the defense mechanism used by animals to combat bacterial infection.

Based on these findings, we propose a schematic model to describe the role of TXNIP in the phagosome maturation of macrophages (**Figure 6**). In this model, the TXNIPinflammasome-caspase-1 axis regulates NADPH oxidase to modulate the pH of the phagosome, deciding the clearance of bacteria from macrophages.

#### DISCUSSION

In this study, we demonstrated that the TXNIP-NLRP3 inflammasome-caspase-1 pathway regulates NADPH oxidase to affect the acidification of the phagosomes in macrophages (**Figure 6**). Txnip KO macrophages showed less secretion of IL-1β and activation of caspase-1 upon treatment with bacteria. This might lead to disinhibition of NADPH oxidase by activated caspase-1, resulting in increased ROS levels and pH. Accordingly, Txnip KO macrophages fail to clear bacteria adequately.

Txnip KO macrophages showed a clear defect in the clearance of engulfed E. coli, but not in engulfment itself. Accordingly,

proportion of wild-type (WT) and Txnip knockout (KO) mouse peritoneal macrophages that retained GFP-expressing E. coli 6 h after treatment with GFP-expressing E. coli at a multiplicity of infection (MOI) of 20. (B) The distribution of WT and Txnip KO mouse peritoneal macrophages based on the intensity of the DHE dye after

(Continued)

FIGURE 3 | treatment with GFP-expressing E. coli at an MOI of 20 at 6 h. (C) The mean fluorescence intensity (MFI) of the DHE dye in the total, GFP-positive, and GFP-negative WT and Txnip KO mouse peritoneal macrophages after treatment with GFP-expressing E. coli at an MOI of 20 at 6 h. Data are expressed as the mean ± SD (n = 3, \*\*P < 0.01, \*\*\*P < 0.001 compared with WT). (D) The MFI of the H2DCFDA dye in WT and Txnip KO mouse peritoneal macrophages after treatment with E. coli at an MOI of 20 at 6 h. Data are expressed as the mean ± SD (n = 3, \*\*\*P < 0.001 compared with WT) (E) The MFI of the GFP in WT and Txnip KO mouse macrophages at 1 and 6 h after treatment with GFP-expressing E. coli at an MOI of 20 with or without DPI treatment. Data are expressed as the mean ± SD (n = 3, \*\*P < 0.01, \*\*\*P < 0.001 compared with WT) (F) CFUs on LB agar plates after overnight incubation with cell extracts derived from WT and Txnip KO mouse peritoneal macrophages treated with E. coli for 1 h at an MOI of 20 (n = 3). (G) Colony-forming units (CFUs) on LB agar plates after overnight incubation with cell extracts derived from WT and Txnip KO mouse peritoneal macrophages treated with E. coli at an MOI of 20 for the indicated times, with or without DPI treatment. Data are expressed as the mean ± SD (n = 3, \*\*P < 0.01, \*\*\*P < 0.001 compared with WT). (H) The distribution of WT and Txnip KO mouse peritoneal macrophages based on the intensity of the DHE dye at 6 h after the 1-h treatment with GFP-expressing E. coli at an MOI of 20, with or without DPI treatment, and removal of extracellular bacteria. (I) The MFI of DHE for (H). Data are expressed as mean (n = 3, \*P < 0.05, \*\*\*P < 0.001 compared with WT).

we investigated the effects of TXNIP depletion on several aspects of phagosome maturation, including the recruitment of phagosomal proteins, the phagosomal pH, and bacterial degradation. Even though clearance during the late phase was affected, there were no differences in the recruitment of Rab7, LAMP-1, and V-ATPase to the phagosomes. Instead, Txnip KO macrophages demonstrated clear changes in pH, ROS production, the secretion of IL-1β, and the activation of caspase-1.

It is known that ROS generated by NOX2 is important for removing bacteria in the phagosome, forming LAP (LC3 associated phagosome), a type of phagosome created by LC3, and presenting the antigen (31–34). Although NOX2 has microbicidal activity, the formation of superoxide by NOX2 can inhibit phagosomal acidification by V-ATPase, and the cellular damage caused by sustained ROS can induce apoptosis (8, 35, 36). The removal of bacteria by NOX2 and the formation of LAP are important mechanisms in the early stages of macrophage contact with bacteria, but if the activity of this enzyme is not continuously controlled in the phagosome at later times, it will affect the viability of the bacteria in the phagosome. In commensal bacteria, the low level of ROS generated by NADPH oxidase modulates the redox-sensor regulatory signaling pathway and supports symbiotic effects of the bacteria (37). In this context, the NADPH oxidasedependent generation of ROS has a double-edged sword effect in multicellular organisms and is connected to the adaptation or survival of the microbiome.

Acidification is the key to the many facets of phagosome maturation. It is a tightly regulated process that begins after the phagocytic cup has closed and phagosome luminal pH goes from 7 to 4. These changes precede the fusion with acidic compartments, and this early acidification event requires delivery of the V-ATPase (17). This proton-transporting enzyme is recruited from endosomes and lysosomes and is assembled on the membrane of the nascent vacuole (15). However, how the pH is then regulated remains poorly defined for the phagosomal state. In early phases following phagocytosis of gram-positive bacteria, there is rapid acidification of the phagosomal lumen, and the NLRP3-dependent inflammasome complex generates activated caspase-1, which regulates the superoxide generated by NOX2. In this early process, it has been shown that gram-negative bacteria cannot trigger this mechanism (8). However, our results suggest that TXNIP modulates the levels of superoxide to regulate the activity of caspase-1, generated by the NLRP3 inflammasome complex at later stages, and induces bactericidal activity toward engulfed E. coli by acidifying the phagosomal lumen in order to activate acidic enzymes present in the phagosome. Although the bacteria clearance of Txnip KO macrophages incubated with DPI did not completely restore to the level of that in WT macrophages, the survival of engulfed bacteria in Txnip KO macrophages was reduced compared with that of KO macrophages incubated without DPI at the late time (**Figure 3**). Our result of isolated phagosomes showed that Rab5 is less recruited into the phagosomes of Txnip KO macrophages, and the reduced level of Rab5 did not influence the acidification of phagosomal lumen at the late time. Commonly, Rab5 is known as a regulator of an early endosome or phagosome formation, and this protein is needed in phagosome maturation for acidification of the phagosomal lumen and is essential to reduce bacterial evasion from the host cells (1, 5). In this context, our results indicated the possibility that TXNIP can regulate the formation of Rab5 dependent phagosomes and the acidification against invading bacteria in the early time. As a result, the difference of Rab5 level into the phagosome may be influenced by the bacterial clearance in the Txnip KO macrophage incubated with DPI (**Figure 2A**). To evade these processes, some pathogens have evolved mechanisms such as buffering their local environment in an attempt to maintain a beneficial neutral pH. For this reason, strict regulation of phagosomal maturation is crucial for bactericidal activity in macrophages and to regulate the immune system.

Gram-negative bacteria are associated with pneumonia, bloodstream infections, and urinary tract infections, and these bacteria can easily acquire antibiotic drug resistance genes (38). Macrophages are an important part of the inflammatory response to these bacteria, modulating the activity of the immune system via the production of cytokines and chemokines and providing clearance through their phagocytic machinery (2, 39, 40). In a sophisticated phagocytosis system to protect against invading bacteria, the survival of E. coli in the phagosome machinery is essential for it to adapt to the host or else a mutation that results in the acquisition of antibiotic resistance is required (37, 41–43). In order to survive within macrophages, bacteria have developed strategies to escape the phagocytic machinery using bacterial effector protein, actin modulation, and alkalization of the phagosomal

FIGURE 4 | The thioredoxin-interacting protein (TXNIP)-inflammasome-caspase-1 axis regulates the clearance of bacteria from macrophages. (A) The levels of interleukin (IL)-1β in conditioned media from wild-type (WT) and Txnip knockout (KO) mouse peritoneal macrophages after treatment with E. coli for the indicated times. Data are expressed as the mean ± SD (\*\*\*P < 0.001 compared with WT). (B) The protein expression levels of caspase-1 P45 and caspase-1 P10 in mouse WT and Txnip KO mouse peritoneal macrophages after treatment with E. coli for the indicated times. (C) Caspase-1 activity, with or without ZVAD treatment, in mouse WT and Txnip KO mouse peritoneal macrophages 6 h after the treatment with E. coli. Data are expressed as the mean ± SD (\*\*\*P < 0.001 compared with WT). (D) Representative images of bacteria-laden WT and Txnip KO mouse macrophages 6 h after a 1-h treatment with GFP-expressing E. coli with or without ZVAD treatment and removal of extracellular bacteria. Scale bar, 10µm.

FIGURE 5 | (n = 3, \*P < 0.05, \*\*P < 0.01 compared with WT). (C) Representative images of spleens from WT and Txnip KO mice at the indicated times after the intraperitoneal injection of E. coli (10<sup>8</sup> CFU). (D) (Left) TUNEL-positive cells in liver sections from WT and Txnip KO mice 24 h after the intraperitoneal injection of E. coli (10<sup>8</sup> CFU). (Right) The number of TUNEL-positive cells in five randomly selected fields in liver sections from WT and Txnip KO mice 24 h after the intraperitoneal injection of E. coli (10<sup>8</sup> CFU). Data are expressed as the mean ± SD (\*P < 0.05 compared with WT). (E) Survival of WT and Txnip KO mice after cecum ligation and puncture (n = 7). (F) CFUs and representative images of LB agar plates after overnight incubation with extracts of blood, liver, and spleen from WT and Txnip KO mice obtained 24 h after cecum ligation and puncture. Data are expressed as the mean ± SD (n = 3, \*P < 0.05, \*\*P < 0.01 compared with WT).

lumen (1, 22, 44, 45). In this context, complete phagosomal maturation is crucial to destroy bacteria and to regulate the host immune system. Our results demonstrate that the regulation of NADPH oxidase-derived superoxide via the TXNIP-NLRP3 inflammasome-caspase-1 axis induces a continuous acidification of the phagosomal lumen that inhibits bacterial survival.

phagosomes, and the resultant clearance of bacteria in Txnip WT and KO macrophages.

Recently, numerous studies have shown that the cytosolic levels of TXNIP are important in regulating diversity cellular signaling pathways, such as the redox system, inflammation, glucose uptake, and apoptosis (17, 30, 46–48). In the same manner, differential TXNIP expression levels are thought to be important in regulating cellular or tissue homeostasis during bacterial infection. Our results show that the TXNIP-NLRP3 inflammasome-caspase-1 axis is a modulator of the phagosomal pH in macrophages exposed to bacteria. Without this effective system, macrophages will fail to clear E. coli even in the presence of increased ROS levels. We suggest that TXNIP might play a role in the destruction of pathogenic E. coli through the effects of inflammasome-mediated caspase-1 on NADPH oxidase.

# DATA AVAILABILITY STATEMENT

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

The animal study was reviewed and approved by The animal study was approved by Institutional Animal Care and Use Committee of the Korea Research Institute of Bioscience and Biotechnology (KRIBB-IACUC, approval number: KRIBB-AEC-11044) and all procedures were performed in accordance with guidelines regarding the use of laboratory animals (National Institutes of Health). Written informed consent was obtained from the owners for the participation of their animals in this study.

# AUTHOR CONTRIBUTIONS

S-JY, DJ, and S-HP performed experiments, data analysis, contributed to the study design, and drafting of the manuscript. J-YP and Y-KL assisted with some experiments and data analysis. M-SL, J-KM, SC, and JK were responsible for setting up the experiments, interpretation of data, and critically reviewing the manuscript. HJ, T-DK, and SY provided useful suggestions. IC and Y-JP supervised the study, collected the data, and wrote the manuscript.

#### REFERENCES


#### FUNDING

This work was supported by the Korea Research Institute of Bioscience and Biotechnology (KRIBB) Research Initiative Program and the Bio & Medical Technology Development Program of the National Research Foundation (NRF), funded by the Korean government (MSIP) (NRF-2015M3A9E6028953, 2015M3A9E6028949, and 2019R1A2C3002034), and the Basic Science Research Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Education (2017R1A6A3A04004741). The funders had no role in study design, data collection, and interpretation or the decision to submit the work for publication.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02636/full#supplementary-material

Supplementary Video 1 | (Left) Txnip WT and (Right) Txnip KO mouse peritoneal macrophages were infected with GFP-expressing E. coli. Phagocytosis was initiated approximately 5 min after cells were exposed to bacteria.


**Conflict of Interest:** 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 Yoon, Jo, Park, Park, Lee, Lee, Min, Jung, Kim, Yoon, Chung, Kim, Choi and Park. 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.

# Different Calcium and Src Family Kinase Signaling in Mac-1 Dependent Phagocytosis and Extracellular Vesicle Generation

Ákos M. Lorincz, Viktória Szeifert, Balázs Bartos, Dávid Szombath, Attila Mócsai and ˝ Erzsébet Ligeti\*

Department of Physiology, Semmelweis University, Budapest, Hungary

#### Edited by:

Carlos Rosales, National Autonomous University of Mexico, Mexico

#### Reviewed by:

Yebin Zhou, AbbVie, United States Ian Dransfield, University of Edinburgh, United Kingdom

#### \*Correspondence:

Erzsébet Ligeti ligeti.erzsebet@ med.semmelweis-univ.hu; ligetierzsebet@gmail.com

#### Specialty section:

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

Received: 15 September 2019 Accepted: 29 November 2019 Published: 17 December 2019

#### Citation:

Lorincz ÁM, Szeifert V, Bartos B, ˝ Szombath D, Mócsai A and Ligeti E (2019) Different Calcium and Src Family Kinase Signaling in Mac-1 Dependent Phagocytosis and Extracellular Vesicle Generation. Front. Immunol. 10:2942. doi: 10.3389/fimmu.2019.02942 Encountering opsonized particles by neutrophils results in phagocytosis of the particle and generation of extracellular vesicles with antibacterial property (aEV). The aim of the present study is to compare the involvement of different receptors and receptor-proximal signaling pathways in these two parallel processes. Investigating human neutrophils from peripheral blood, we show that complement receptors are decisive for both processes whereas immunoglobulin binding Fc receptors (FcR) only participate moderately in phagocytosis and pattern recognition receptors induce mild EV production but only minimal phagocytosis. Studying bone marrow derived neutrophils of genetically modified animals we verify that the involved complement receptor is CR3, also known as the β<sup>2</sup> integrin Mac-1. We show that genetic deletion of the adaptor molecules FcRγ chain or DAP12 does not influence either process, suggesting potential redundant function. Combined absence of the Src family kinases Hck, Fgr, and Lyn drastically impairs phagocytosis but does not influence aEV production. In contrast, deletion of PLCγ2 has no influence on phagocytosis, but reduces aEV formation. In accord with the essential role of PLCγ2, aEV biogenesis both from murine and from human neutrophils is dependent on presence of extracellular calcium. Absence of external calcium prevented the generation of antibacterial EVs, whereas the spontaneous EV formation was not influenced. We thus show that phagocytosis and biogenesis of antibacterial EVs are independent processes and proceed on different signaling pathways although the same receptor plays the critical role in both. Our data reveal the possibility in neutrophilic granulocytes to modulate aEV production without disturbing the phagocytic process.

Keywords: extracellular vesicles, phagocytosis, neutrophils, signaling, antibacterial effect, calcium, Src family kinase, complement receptor

# INTRODUCTION

Phagocytosis by neutrophilic granulocytes is significantly promoted by opsonization. The most effective opsonins are immunoglobulins reacting with different Fc receptors (FcR) and complement fragments recognized by complement receptors (CR). The most abundant CR on the surface of neutrophils is CR3, also known as macrophage antigen 1 (Mac-1), a β<sup>2</sup> integrin (1). In addition to phagocytosis, Mac-1 is involved in adhesion and spreading of the cells as well as in

**48**

adhesion-dependent superoxide production and degranulation (2–4). Recently we found that Mac-1 also plays a critical role in production of extracellular vesicles with antibacterial capacity (5).

Generation of extracellular vesicles (EVs) is a common property of different cell types from the simplest prokaryotes up to the most differentiated eukaryotic cells (6, 7). Investigation of the physiological and pathological roles of EVs has skyrocketed in the last decade and by now EVs are considered as important elements of intercellular communication. Some of the established functions of different EV types such as antigen presentation (8), anti-inflammatory (9), or antimicrobial effects (10) represent protective mechanisms for the host organism. Others, such as transfer of oncogenic receptors (11) or dissemination of antibiotic resistance (12) are more frightening. Hence, medical therapy may require - depending on the exact condition enhancement or reduction of EV production. For the time being such interventions are hampered by the lack of knowledge on the molecular mechanisms of biogenesis.

Diversity of EVs has been widely documented (13, 14). There are clear examples indicating that properties of EVs released even from the same cell vary depending on environmental factors (15–17). Investigating EV production from neutrophilic granulocytes, we have previously characterized three different types of medium-sized EVs (also called microvesicles or ectosomes): those produced from freshly isolated cells spontaneously (sEV), or upon activation with opsonized particles (aEV) and EVs produced during apoptosis of the cells (apoEV). We found differences in the number, protein content and protein composition of these EV populations (18). The most striking difference was in their functional property, as aEVs impaired the growth of bacteria in a concentration-dependent way whereas the other two types had absolutely no antibacterial effect (5). However, the cellular processes governing the production of these divergent EV populations are poorly explored.

Searching for the molecular mechanism of generation of antibacterial EVs, we identified Mac-1 as the key cell surface receptor triggering aEV production. The aim of the current study was to investigate the relationship of the two Mac-1 dependent processes and to characterize the initial steps of the signaling pathway leading from receptor activation eventually to the release of aEVs or phagocytosis. Using neutrophils isolated from human peripheral blood or from genetically modified mice, we show that the two processes are independent and proceed on clearly distinguishable signaling pathways.

# MATERIALS AND METHODS

#### Materials

HBSS with or without calcium and magnesium and glucose was from GE Healthcare Life Sciences (South Logan, UT, USA), Zymosan A from Sigma Aldrich (St. Louis, MO, USA), Ficoll-Paque and Percoll from GE Healthcare Bio-Sciences AB (Uppsala, Sweden), HEPES (pH 7.4) from Sigma. All other used reagents were of research grade. GFP-expressing and chloramphenicol resistant S. aureus (USA300) was a kind gift of Professor William Nauseef (University of Iowa).

### Preparation of Human PMN and EV

Venous blood samples were drawn from healthy adult volunteers according to procedures approved by the National Ethical Committee (ETT-TUKEB No. BPR/021/01563-2/2015). Neutrophils were obtained by dextran sedimentation followed by a 62.5% (v/v) Ficoll gradient centrifugation (700g, 40 min, 22◦C) as previously described (18). The preparations contained more than 95% PMN and <0.5% eosinophils. PMNs (typically 10<sup>7</sup> cell in 1 mL HBSS) were incubated with or without activating agent for 30 min at 37◦C in a linear shaker (80 rpm/min). After activation, cells were sedimented (500g, Hermle Z216MK 45◦ fixed angle rotor, 5 min, 4◦C). Upper 500 µL of the supernatant was filtered through a 5µm pore sterile filter (Sterile Millex Filter Unit, Millipore, Billerica, MA, USA). The filtered fraction was sedimented (15700 g, Hermle Z216MK 45◦ fixed angle rotor, 5 min, 4◦C), and the pellet was carefully resuspended in the original incubation volume. Protein concentration of EV was determined by the Bradford protein assay using BSA as standard (5).

### Transgenic Mice

Triple Src-family kinase (SFK) knock-out mice in which all three Src-family kinase isoforms identified in neutrophils (Hck, Fgr, and Lyn) were missing (19) were obtained from Clifford Lowell (University of California, San Francisco, CA). Complete CD18-deficient (Itgb2tm2Bay/tm2Bay, referred to as CD18−/−) mice (20) were obtained from A. Beaudet (Baylor College of Medicine, Houston, TX). CD11b-deficient (Itgamtm1Myd/tm1Myd , referred to as CD11b−/−) mice were purchased from The Jackson Laboratory (21). CD11a-deficient (Itgaltm1Hogg/tm1Hogg , referred to as CD11a−/−) mice were obtained from N. Hogg (Cancer Research UK, London, UK) (19). FcRγ-chain deficient (Fcer1gtm1Rav/tm1Rav, referred to as FcRγ <sup>−</sup>/−) mice were purchased from Taconic (22). Dap12-deficient (referred to as Dap12−/−) mice (22) were obtained from Lewis Lanier (University of California, San Francisco, CA). Heterozygous mice carrying a deleted allele of the PLCγ2-encoding gene (Plcg2tm1Jni , referred to as PLCγ2 <sup>−</sup>/+) were obtained from James N. Ihle (St. Jude Children's Research Hospital, Memphis, TN, USA) (23) and have been backcrossed to the C57BL/6 genetic background for more than 10 generations. Because of the limited fertility of homozygous PLCγ2 <sup>−</sup>/<sup>−</sup> mice, the mutation was maintained in heterozygous form as described (24). Syk+/<sup>−</sup> mice (25) were originally obtained from Victor Tybulewicz (National Institute of Medical Research, London, UK). Syk−/<sup>−</sup> bone marrow chimeras carrying a Syk-deficient hematopoietic system were generated by transplanting Syk−/<sup>−</sup> fetal liver cells into lethally irradiated wild-type recipients (26). All transgenic mice were backcrossed to the C57BL/6 genetic background for at least 6 generations. All transgenic mice were 11–20 weeks old. Age- and sexmatched C57Bl/6 animals were used as controls. Genotyping was performed by allele-specific PCR. WT control C57BL/6 mice were purchased from Charles River or the Hungarian National Institute of Oncology. Mice were kept in sterile, individually ventilated cages (Tecniplast, Buguggiate, Italy) in a conventional facility. All animal experiments were approved by the Animal Care Committee of the National Authority for Animal Health (Budapest, Hungary).

#### Isolation of Murine PMN and EV

Murine neutrophils were isolated from the bone marrow of the femurs, humeri, and tibias of intact mice by hypotonic lysis followed by Percoll gradient centrifugation (62.5% v/v, 700g, 40 min, 22◦C) using sterile and endotoxin-free reagents as previously described (19). Cells were kept at room temperature in Ca <sup>2</sup><sup>+</sup> and Mg2+-free medium until use (usually <30 min) and pre-warmed to 37◦C before activation. Neutrophil assays were performed at 37◦C in HBSS supplemented with 20 mM HEPES, pH 7.4. PMNs (10<sup>7</sup> cell in 1 mL HBSS) were incubated with or without activating agent for 30 min at 37◦C on a linear shaker (80 rpm/min). After incubation, PMNs were sedimented (1,000 g, Hermle Z216MK 45◦ fixed angle rotor, 5 min, 4◦C) and the upper 800 µL of the supernatant was filtered through a 5µm pore sterile filter. The filtered fraction was sedimented again (30,000 g, Beckmann JA-17 fixed 25◦ angle rotor, 30 min, 4◦C). The sediment was resuspended in HBSS at the original volume and used immediately for further analysis according to previous observations (27).

#### Opsonization

Zymosan (5 mg in 1 mL HBSS) was opsonized with 500 µL pooled human or murine serum for 30 min at 37◦C. For complement-free opsonization zymosan (5 mg in 1 mL PBS) was opsonized with 500 µL human serum pretreated with 20 mM EDTA. After opsonization zymosan was centrifuged (5,000g, 5 min, 4◦C, Hermle Z216MK 45◦ fixed angle rotor), and washed once in HBSS.

Bacteria (OD<sup>600</sup> = 1.0 in 800 µL HBSS) were opsonized with 200 µL pooled human or murine serum or with EDTA pretreated human serum for 30 min at 37◦C. After opsonization, bacteria were centrifuged (8,000 g, 5 min, 4◦C, Hermle Z216MK 45◦ fixed angle rotor), and washed once in HBSS.

# EV Analysis and Quantification by Flow Cytometry

Human EVs were labeled with RPE conjugated monoclonal anti-CD11b (1µg/mL, Tonbo Biosciences, USA, clone M1/70) (28), FITC conjugated anti-CD18 monoclonal Ab (1µg/mL, Dako) or FITC conjugated annexinV (BD Biosciences) for 20 min at 37◦C and then washed in HBSS. Murine EVs were labeled with RPE conjugated monoclonal anti-CD11b (1µg/mL, Tonbo Biosciences, USA, clone M1/70) (29) or RPE conjugated monoclonal anti-CD18 (1µg/mL, BD Biosciences, clone C71/16) (30) or PerCP-CY 5.5 conjugated monoclonal anti-Ly6g (1µg/mL, BD Biosciences, clone 1A8) (31) or FITC conjugated AnnexinV (BD Biosciences) for 20 min at 37◦C and then washed in HBSS. Isotype controls were from identical manufacturer, annexinV labeling was controlled in 20 mM EDTA containing medium.

For flow cytometric detection of EVs a Becton Dickinson FACSCalibur flow cytometer was used as described previously in Lorincz et al. (27). Briefly pure HBSS medium was used for setting the threshold to eliminate instrument noise then fluorescent beads (3.8µm SPHERO Rainbow Alignment Particles from Spherotech Inc., USA) were detected to set the upper size limit of EV detection range. After the measurement of an EV preparation the number of isotype control events and the 0.1% TritonX-100 detergent non-sensitive events were subtracted to calculate the true EV number. To avoid swarm detection, the flow rate was held below 1,000 events/s (3,750 events/µL) during measurements. Samples were re-measured after a 2 fold dilution to control linearity of measurements. Linearity was controlled in a broader range previously (32). FC data were analyzed with Flowing 2.5 Software (Turku Center for Biotechnology, Finland).

### Quantification of Phagocytosis

Neutrophils (10<sup>6</sup> in 1 mL HBSS) were incubated with GFP expressing S. aureus (USA300, 10<sup>7</sup> in 1 mL HBSS) for 30 min at 37◦C in a linear shaker (80 rpm/min). For kinetic measurements samples were taken in every 10 min, for endpoint measurements phagocytosis was stopped after 20 min. Samples were diluted 5-fold in ice-cold PBS and analyzed by flow cytometry. To avoid coincidental co-detection of cells and bacteria and to control linearity of the measurement every sample was measured again after a 2-fold dilution. Percentage of phagocytosing PMNs was calculated by detecting GFP positive and negative cells.

To prove that cell associated bacteria are internalized (and not only bound by the surface), we carried out confocal microscopic imaging with X-Y projections (**Figure 1B**). For microscopic control of phagocytosis, cells were 5-fold diluted in ice-cold PBS and sedimented onto a coverslip for 30 min. Later cells were fixed with 4% (w/v) paraformaldehyde and analyzed with a Zeiss LSM710 confocal laser scanning microscope equipped with 40×/1.3 and 63×/1.3 oil immersion objective (Plan-Neofluar, Zeiss). Images were analyzed with LSM Image Browser software (Zeiss).

# Activation of Adherent Neutrophils

Selective activation of Mac-1 complex of adherent human neutrophils was performed in 6 well tissue culture plates (Biofil, Hungary) coated overnight with 0.2 mg/mL BSA or 50µg/mL C3bi (both from Merck, Darmstadt, Germany) as previously described (33). To obtain immobilized immune complex–coated surfaces, human lactoferrin (20µg/mL; Sigma-Aldrich, USA) was covalently linked to poly-l-lysine (Sigma-Aldrich, USA) coated 6 well plates and then treated with polyclonal antilactoferrin (LTF) IgG (1:400 dilution; Sigma-Aldrich, USA) or non-specific IgG (1:400 dilution; Sigma-Aldrich, USA) for 1 h as previously described (34). Unbound immunoglobulin was removed by washing the plate by HBSS three times. The isotype control serves to test the unspecific binding of applied antibodies.

#### Bacterial Survival Assay

Opsonized bacteria (5 × 10<sup>7</sup> /50 µL HBSS) were added to 500 µL EV (derived from 5 × 10<sup>6</sup> PMN) suspended in HBSS. During a 40 min co-incubation step at 37◦C the bacterial count

RM-ANOVA coupled with Tukey's post hoc test. (B) Confocal microscopic images of human neutrophils after 20 min phagocytosis of non-opsonized (UL), partially (UR), and completely (LL) opsonized GFP expressing S. aureus; red color shows CD11b labeling. X-Y projections of engulfed bacteria with the respective side views (LR). Representative images out of 4 independent experiments. (C–E) Quantification of phagocytosis of WT vs. CD18 KO (C), CD11b KO (D), CD11a KO (E) murine PMN. Data were compared using Student's t-test; n = 6, 6, 6 ± SEM. \*\*P < 0.01; \*\*\*P < 0.001; \*\*\*\*P < 0.0001.

decreases or increases depending on the samples' antibacterial effect and the growth of bacteria. At the end of the incubation, 2 mL ice-cold stopping solution (1 mg/mL saponin in HBSS) was added to stop the incubation and lyse EVs. After a freezing step at −80◦C for 20 min, samples were thawed to room temperature and inoculated into LB broth. Bacterial growth was followed as changes in OD using a shaking microplate reader (Labsystems iEMS Reader MF, Thermo Scientific) for 8 h, at 37◦C, at 650 nm. After the end of growth phase the initial bacterial counts were calculated indirectly using an equation similar to PCR calculation, as described previously (35).

#### Statistics

Comparisons between two groups were analyzed by two-tailed Student's t-tests or ANOVA. Exact statistical tests are indicated in the figure legends. All bar graphs show mean and ± SEM. Difference was taken significant if P value was < 0.05. <sup>∗</sup> represents P < 0.05; ∗∗ represents P < 0.01; ∗∗∗ represents P < 0.001. Statistical analysis was performed using GraphPad Prism 6 for Windows (La Jolla, CA, USA).

# RESULTS

#### Comparison of Receptors Involved in Phagocytosis and EV Generation Initiated by Opsonized Particles

We first carried out a detailed analysis on the involvement of different receptors in phagocytosis, using differently opsonized particles (**Figure 1**). Following the process by flow cytometry up to 30 min, we could detect only minimal phagocytosis of non-opsonized bacteria by human neutrophils (**Figure 1A**). If bacteria were treated with complementdepleted serum and so opsonized mainly by antibodies that activate different Ig-binding FcR, we observed phagocytosis in ∼30% of the cells (**Figure 1A**). In contrast, particles opsonized in full serum, allowing thus the activation of both Fc and complement receptors, induced significantly greater phagocytosis, and bacteria were detectable in ∼80% of the investigated neutrophils (**Figure 1A**). In **Figure 1B** we show the results of similar experiments carried out by confocal microscopy, verifying that bacteria were in fact engulfed, not only associated to the surface of the cells. The kinetic experiments presented in **Figure 1A** indicated that under our conditions phagocytosis was completed in 20 min. Therefore, in the following experiments only data obtained by flow cytometry at 20 min are shown.

In order to confirm the complement receptor playing major role in phagocytosis under our conditions, we tested neutrophils from genetically modified mice. Earlier studies (1) suggested CR3-which is identical with the β<sup>2</sup> integrin Mac-1—as the most important complement receptor in neutrophilic granulocytes. In accordance with previous findings (3), deletion of either CD18, the common β chain of all neutrophil β<sup>2</sup> integrins, or CD11b, the specific α chain of Mac-1 resulted in drastic decrease of phagocytosis (**Figures 1C,D**). In contrast, depletion of CD11a, the α chain of the integrin LFA-1, which does not serve as complement receptor, had no influence on phagocytosis (**Figure 1E**).

Next we investigated EV production from human neutrophils under similar conditions. As shown in **Figure 2A**, non-opsonized particles induced moderate EV generation, which was not further increased if particles were opsonized by antibodies. Fully opsonized particles, which were able to stimulate both Fc and CR receptors, initiated maximal EV release. To substantiate the difference in the role of Fc and CR receptors in EV biogenesis, in the following experiments we stimulated the two types of receptors selectively on coated surfaces (**Figures 2B,C**). C3bi, the specific ligand of CR3 receptor induced significant and consistent increase of EV production from adherent neutrophils (**Figure 2B**), whereas no significant change of EV release was observed if the cells were seeded on an immune complex surface (**Figure 2C**). Thus, under our experimental conditions Mac-1/CR3 seems to play a decisive role both in phagocytosis and in EV generation. In the following experiments we focused on the signaling process downstream of Mac-1.

FIGURE 2 | Participation of different receptors in activated EV generation from human neutrophils. (A) Comparison of phagocytosis and EV production after 20 min activation by different opsonized particles. +SEM, n = 21 (EV) or 8 (phagocytosis). Data were compared using RM two-way ANOVA coupled with Dunett's post hoc test. (B) Generation of EV by neutrophils adherent to non-specific BSA or specific C3bi surface. (C) Generation of EV by neutrophils adherent to immune complex surface. Data were compared using paired Student's t-test; n = 4 and 7. \*P < 0.05; \*\*\*P < 0.001; \*\*\*\*P < 0.0001.

FIGURE 3 | Role of adaptor proteins in aEV generation and phagocytosis of murine neutrophils. Comparison of sEV and aEV production of WT vs. FcRγ chain KO (A) and DAP12 KO (C), murine PMN. Comparison of phagocytosis of fully opsonized bacteria by WT vs. FcRγ chain KO (B) and DAP12 KO (D) murine PMN. Data were compared using Student's t-test; n = 24, 8, 7, 6 +SEM. \*P < 0.05; \*\*\*P < 0.001.

#### Role of the Adaptor Proteins FcRγ and DAP12 in Mac-1 Initiated EV Generation and Phagocytosis

Outside-in signaling of β<sup>2</sup> integrins was shown to involve tyrosine kinases, but the short intracellular segments of both chains lack suitable signaling sequences. The adapter molecules FcRγ chain and DAP12 which contain conserved immune receptor tyrosine based activation motives (ITAMs) were shown to play a critical role in downstream signaling of β<sup>2</sup> integrins (36). Therefore, we investigated first the involvement of these two molecules in EV generation and phagocytosis in neutrophils obtained from genetically deficient mice. EV production was detected both under resting conditions (spontaneous, sEV) and upon stimulation with full murine serum opsonized zymosan particles (activated, aEV).

Deletion of FcR γ chain prevents signal transduction via all FcR in murine neutrophils (34) and it was shown to transmit Mac-1 signaling as well (36). In FcRγ KO animals we did not see any change in EV production as compared to the wild type, and observed only a minor, statistically non-significant decrease of phagocytosis (**Figures 3A,B**). The latter finding is in accord with the moderate activity of FcRs shown in **Figure 1**. Similar results were obtained in neutrophils from DAP-12 deficient animals (**Figures 3C,D**) indicating that neither adapter molecule plays an exclusive role in Mac-1 signaling, and suggesting a potential redundant function.

# Src-Family Kinases and Syk Are Dispensable for EV Generation but Required for Phagocytosis

Phosphorylation of the ITAM sequences of the adapter molecules is carried out by the Src family kinases (SFK), of which Hck, Fgr, and Lyn are expressed in murine neutrophils (37). We investigated aEV production and

FIGURE 4 | Role of tyrosine kinases in aEV generation and phagocytosis of murine neutrophils. Comparison of sEV and aEV production of WT vs. SFK KO (A) and Syk KO (C) murine PMN. Comparison of phagocytosis of fully opsonized bacteria by WT vs. SFK KO (B) and Syk KO (D) murine PMN. Data were compared using Student's t-test; n = 17, 8, 7, 5 +SEM. \*P < 0.05; \*\*P < 0.01; \*\*\*P < 0.001.

FIGURE 5 | Role of PLCγ2 and availability of extracellular calcium in aEV generation and phagocytosis of murine neutrophils. Comparison of sEV and aEV production of (A) and phagocytosis of fully opsonized bacteria by (B) WT vs. PLCγ2 KO murine PMN. Quantification of aEV production from (C) and phagocytosis by (D) WT murine PMN in the presence and absence of extracellular Ca2+. Data were compared using Student's t-test; n = 5, 5, 5, 4 +SEM. \*P < 0.05 and \*\*P < 0.01.

phagocytosis in neutrophils from triple KO mice, which lack all three SFKs.

Phagocytosis of fully opsonized bacteria was significantly impaired in the absence of all three SFKs. Surprisingly, EV generation was not affected at all and reached similar values as detected with wild-type cells (**Figures 4A,B**). Thus, at this point we observed serious divergence in the signalization process of phagocytosis and aEV generation.

Phosphorylation of the ITAM sequences of the adaptor proteins by SFKs allows the binding and activation of Syk tyrosine kinase that plays important role in the transmission of several SFK mediated cell functions (4). Syk-deficient animals are not viable but the function of the hematopoietic compartment can be studied following transplantation with wild-type or Syk-deficient bone marrow. The data shown in **Figures 4C,D** were obtained using neutrophils of transplanted animals. Generation of aEVs from Syk-deficient neutrophils was not significantly different from the extent obtained with neutrophils following transplantation of wild-type bone marrow, although the increase compared to sEV generation proved not to be statistically significant either (**Figure 4C**). Phagocytosis by neutrophils from transplanted animals was very low, nevertheless significant difference was obtained between neutrophils produced after transplantation with wildtype or Syk KO bone marrow (**Figure 4D**). Thus, neither SFKs nor Syk seem to be essential for the production of aEVs in neutrophilic granulocytes whereas they are required for maximal phagocytosis.

#### Requirement for Calcium Signaling Is Different in EV Generation and Phagocytosis

The γ2 isoform of phospholipase C was shown to be essential in organization of several neutrophil functions initiated by β<sup>2</sup>

integrin stimulation (23). Hence we investigated aEV production and phagocytosis in neutrophils of PLCγ2 deficient animals. Phagocytosis was not influenced, but aEV production was seriously impaired (**Figures 5A,B**). In fact, there was only very little difference between spontaneous and activated EV generation from PLCγ2 KO neutrophils (**Figure 5A**). Thus, the key enzyme of inositol trisphosphate (IP3) generation and consequent calcium signaling seems to be required for aEV generation but it is dispensable for phagocytosis.

Next, we investigated whether presence of calcium in the extracellular space is essential for aEV generation or phagocytosis. As shown in **Figures 5C,D** for murine neutrophils, aEV production was almost completely inhibited in the absence of calcium, whereas phagocytosis was not affected at all. Interestingly, neither the absence of calcium, nor the lack of PLC γ2 decreased spontaneous EV release.

In the following experiments, we wanted to substantiate our observations also in human neutrophils, where the antibacterial function could be tested as well. EV generation was followed on the basis of two parameters: number of vesicles detectable by flow cytometry and total protein content of the sedimented vesicles. Both parameters indicated defective aEV production in the absence of extracellular calcium whereas sEV release was not affected (**Figure 6A**). In contrast, there was no difference in phagocytosis in the presence or absence of external calcium (**Figure 6B**). Thus, human PMN behaved similar to their murine counterparts.

Antibacterial effect of the separated vesicles was followed under conditions where the effect of EVs of identical protein content was tested upon bacterial survival. Intact human neutrophils were applied as positive control. As indicated in **Figure 6C**, only EVs produced upon stimulation with opsonized particles in the presence of external calcium were able to impair bacterial growth whereas EVs produced upon the same stimulation but in the absence of calcium lacked the antibacterial effect.

#### DISCUSSION

In this study we investigated two cellular processes initiated by opsonized particles and show that both in generation of EVs with antibacterial property and in phagocytosis the multifunctional surface molecule Mac-1 plays the central role. However, our data indicate that aEV production is independent of phagocytosis in spite of being triggered by the same receptor. On one hand aEV generation proceeds also on coated surface where phagocytosis is not possible (**Figure 2B**) and on the other hand we observed major differences in the implicated signaling pathway (**Figures 4**–**6**). Our findings suggest that the cytoskeletal rearrangement related to phagocytosis is neither sufficient nor required for EV generation in neutrophils.

Mac-1 is one of the two major β<sup>2</sup> integrins which plays an important role in slow rolling and adherence of neutrophilic granulocytes (2, 3, 38). β<sup>2</sup> integrin-dependent adherent activation of neutrophils which consists of spreading, superoxide production and degranulation was shown to involve activation of SFKs, Syk tyrosine kinase and PLCγ2 enzymes (23, 25, 39–41). Specific Mac-1 dependent degranulation was shown to depend on SFK and Syk kinase activity (42). Our current data indicate that the two investigated Mac-1 dependent processes are only partially dependent on this previously characterized signaling pathway. Phagocytosis entails the SFK and Syk tyrosine kinases, but it proceeds undisturbed in the absence of PLCγ2 and extracellular calcium (**Figure 7**). Much to our surprise, enhanced EV production triggered by Mac-1 did not depend on activity of SFK and was not affected by their absence. In contrast, activated EV production was seriously compromised by the lack of PLCγ2, the enzyme producing the calcium mobilizing messenger molecule IP3. Similarly, no significant increase of EV production could be achieved by Mac-1 stimulation in the absence of extracellular calcium either in murine or in human neutrophils. Last, but not least EVs produced upon encountering opsonized particles in the absence of extracellular calcium, had no antibacterial effect. The fact that phagocytosis was not affected by the absence of extracellular calcium indicates that receptor binding was functioning properly. Apparently, generation of antibacterial EVs requires concurrent calcium entry and intracellular calcium signaling. In different cell types Ca ionophores were shown to initiate EV generation (43, 44), although the functional properties of those vesicles were not investigated in details. In contrast, here we demonstrate - to our knowledge the first time - the key role of calcium signaling in biogenesis and function of EVs triggered by physiological stimuli via an identified receptor.

Freshly isolated resting neutrophils produce a basal amount of EVs spontaneously (sEV) which served as reference in all our experiments. Interestingly, this constitutive EV production did not depend on any signaling element investigated in our study. This fact clearly indicates the existence of distinguishable signaling pathways in production of EVs of different properties and functions.

Mac-1 is a multifunctional molecule with over 40 identified ligands (45). The ligands investigated in detail were shown to bind to partially overlapping but distinct sites and some were suggested to behave as biased agonist (46). Our studies extend this picture with a Mac-1 dependent pathway that proceeds without participation of SFK (and probably Syk) kinases. Previous studies indicated that both SFK, Syk and PLCγ2 seem to be dispensable for β<sup>2</sup> integrin dependent migration (19, 23, 25, 47), although the involvement of Mac-1 remained questionable. In another study (42) Mac-1 dependent elastase release was only partially inhibited in the absence of all 3 SFK. All these observations suggest the possibility of multiple parallel signaling pathways in organization of different cellular responses triggered by the same cell surface receptor. Thereby our findings raise the potential of selective modulation of aEV production without interference with the phagocytic process.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/supplementary material.

#### ETHICS STATEMENT

The studies involving human participants were reviewed and approved by Budapest Capital Government Office, Public Health Direction. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Animal Care Committee of the National Authority for Animal Health (Budapest, Hungary).

#### AUTHOR CONTRIBUTIONS

ÁL and EL designed the experiments. ÁL, VS, BB, and DS carried out the experiments. AM provided expertise with genetically modified animals. ÁL and EL summarized the results and wrote the manuscript. EL obtained the funding.

#### FUNDING

Experimental work was financed by a grant from the National Research, Development and Innovation Office, Hungary (OTKA K119236).

#### ACKNOWLEDGMENTS

The authors were indebted to Dr. Tamás Németh for help with murine bone marrow transplantation; to Ms. Regina Tóth-Kun for expert and devoted technical assistance; to Ms. Edina Simon for genotyping the animals; and to C. Lowell, A. Beaudet, N. Hogg, L. Lanier, J. Ihle, and V. Tybulewicz for providing the transgenic mice.

#### REFERENCES

1. Erdei A, Lukacsi S, Macsik-Valent B, Nagy-Balo Z, Kurucz I, Bajtay Z. Non-identical twins: different faces of CR3 and CR4 in myeloid and lymphoid cells of mice and men. Semin Cell Dev Biol. (2017) 85:110–21. doi: 10.1016/j.semcdb.2017. 11.025

<sup>2.</sup> Lu H, Smith CW, Perrard J, Bullard D, Tang L, Shappell SB, et al. LFA-1 is sufficient in mediating neutrophil emigration in Mac-1-deficient mice. J Clin Invest. (1997) 99:1340–50. doi: 10.1172/JCI119293

<sup>3.</sup> Li X, Utomo A, Cullere X, Choi MM, Milner DA Jr, Venkatesh D, et al. The beta-glucan receptor Dectin-1 activates the integrin Mac-1 in neutrophils via Vav protein signaling to promote Candida albicans clearance. Cell Host Microbe. (2011) 10:603–15. doi: 10.1016/j.chom.2011.10.009


through the Syk tyrosine kinase. Proc Natl Acad Sci USA. (2004) 101:6158– 63. doi: 10.1073/pnas.0401602101


targets pathologic inflammation while maintaining protective host-defense. Nat Commun. (2018) 9:525. doi: 10.1038/s41467-018-02896-8

47. Mazzi P, Caveggion E, Lapinet-Vera JA, Lowell CA, Berton G. The Src-family kinases Hck and Fgr regulate early lipopolysaccharide-induced myeloid cell recruitment into the lung and their ability to secrete chemokines. J Immunol. (2015) 195:2383–95. doi: 10.4049/jimmunol.1402011

**Conflict of Interest:** 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 Lorincz, Szeifert, Bartos, Szombath, Mócsai and Ligeti. 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.

# Phagocytosis of Necrotic Debris at Sites of Injury and Inflammation

#### Johannes Westman<sup>1</sup> , Sergio Grinstein1,2,3 \* and Pedro Elias Marques <sup>4</sup>

*<sup>1</sup> Program in Cell Biology, Hospital for Sick Children, Toronto, ON, Canada, <sup>2</sup> Department of Biochemistry, University of Toronto, Toronto, ON, Canada, <sup>3</sup> Keenan Research Centre of the Li Ka Shing Knowledge Institute, St. Michael's Hospital, Toronto, ON, Canada, <sup>4</sup> Laboratory of Molecular Immunology, Department of Microbiology, Immunology and Transplantation, Rega Institute for Medical Research, KU Leuven, Leuven, Belgium*

Clearance of cellular debris is required to maintain the homeostasis of multicellular organisms. It is intrinsic to processes such as tissue growth and remodeling, regeneration and resolution of injury and inflammation. Most of the removal of effete and damaged cells is performed by macrophages and neutrophils through phagocytosis, a complex phenomenon involving ingestion and degradation of the disposable particles. The study of the clearance of cellular debris has been strongly biased toward the removal of apoptotic bodies; as a result, the mechanisms underlying the removal of necrotic cells have remained relatively unexplored. Here, we will review the incipient but growing knowledge of the phagocytosis of necrotic debris, from their recognition and engagement to their internalization and disposal. Critical insights into these events were gained recently through the development of new *in vitro* and *in vivo* models, along with advances in live-cell and intravital microscopy. This review addresses the classes of "find-me" and "eat-me" signals presented by necrotic cells and their cognate receptors in phagocytes, which in most cases differ from the extensively characterized counterparts in apoptotic cell engulfment. The roles of damage-associated molecular patterns, chemokines, lipid mediators, and complement components in recruiting and activating phagocytes are reviewed. Lastly, the physiological importance of necrotic cell removal is emphasized, highlighting the key role of impaired debris clearance in autoimmunity.

#### Edited by:

*Florence Niedergang, Centre National de la Recherche Scientifique (CNRS), France*

#### Reviewed by:

*Ian Dransfield, University of Edinburgh, United Kingdom Barbara Bottazzi, Humanitas Clinical and Research Center, Milan University, Italy*

#### \*Correspondence:

*Sergio Grinstein sergio.grinstein@sickkids.ca*

#### Specialty section:

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

Received: *14 October 2019* Accepted: *10 December 2019* Published: *09 January 2020*

#### Citation:

*Westman J, Grinstein S and Marques PE (2020) Phagocytosis of Necrotic Debris at Sites of Injury and Inflammation. Front. Immunol. 10:3030. doi: 10.3389/fimmu.2019.03030* Keywords: cell death, necrosis, apoptosis, phagocytosis, inflammation, cell debris, "find-me," "eat-me"

#### INTRODUCTION

Cell death is inherent to living multicellular organisms. It is a key regulator of homeostasis, being required during development, growth and maintenance of tissues; it is also a turning point in the immune response. Healthy humans lose billions of cells per day constitutively via the process of apoptotic cell death. Apoptosis, the prototypical form of programmed cell death, was described morphologically in the early seventies (1) as involving cell shrinkage and chromatin condensation, followed by fragmentation of the entire cell into smaller, sealed apoptotic bodies. These apoptotic bodies are promptly cleared by neighboring phagocytes and parenchymal cells through phagocytosis, in this case termed efferocytosis (meaning "carrying to the grave"), without initiating an inflammatory response or disturbing tissue homeostasis.

While apoptosis has been studied most extensively, there are many other ways for cells to experience death. The intrinsic activity of organisms often puts them in contact with extreme temperatures, strong mechanical forces and harmful chemical agents. These situations frequently culminate in a catastrophic form of cell death with loss of plasma membrane integrity and pro-inflammatory properties named necrosis (2). Necrotic cell death can either be accidental or programmed (e.g., pyroptosis and necroptosis), leading to the release of intracellular contents into the extracellular environment. Necrosis differs qualitatively from apoptosis, which is clearly demonstrated by the lack of conversion of necrotic cells into apoptotic bodies, a process that requires enzymatic activity and energy. Importantly, these differences also predict that the means of clearance of the cell debris generated by necrosis vs. apoptosis may be drastically different.

Efferocytosis has received a great deal of attention in the past decades, and is by now a well-understood process involving dozens of described receptors and molecular effectors (**Figure 1**). Because of the profusion of studies, a casual reader may be left with the mistaken impression that efferocytosis is the only means of clearance of cell debris in the body. This is certainly not the case, as is most graphically shown by the existence of apoptosis-defective organisms, such as mice deficient in the initiator caspases 2 (3) and 9 (4), and effector caspases 3 (5), 6 (6), and 7 (7), that nevertheless develop and survive rather normally! Clearly, other mechanisms of cell death and debris clearance must exist. The main purpose of this chapter is to review the clearance of cell debris of necrotic origin. Parallels will be drawn between apoptosis and necrosis, stressing how each mode of cell death may produce different "find-me" and "eat-me" signals that will ultimately lead to clearance of debris by different cell types and phagocytic receptors. In addition, the immunological consequences of defective clearance of cell debris will be discussed: this can take the form of delayed tissue regeneration upon injury or even severe autoimmunity in the long-term. In collating the available information on necrotic cell clearance, this review aims to shed new light on diseases in which necrotic debris are central, such as in atherosclerosis, liver injury, arthritis, severe trauma, lupus, and many others.

# APOPTOSIS AND EFFEROCYTOSIS

Approximately 200 billion cells undergo turnover (ostensibly by apoptosis) every day in the human body (8). Yet, few apoptotic cells are found in the steady state in healthy humans, suggesting that these cells are rapidly cleared. In order to orchestrate efferocytosis, three main signaling programs are required. First, chemotactic "find-me" signals are produced to attract professional phagocytes toward the dying cell. Second, "eat-me" signals appear on the surface of the apoptotic cell, which will help phagocytes recognize and engulf it. Lastly, the internalized apoptotic body is degraded in the phagolysosomal compartment by proteases, DNAses and lipases.

During apoptosis, cellular components are modified by the activity of caspases and packaged into sealed vesicles—the socalled apoptotic bodies—that expose phosphatidylserine (PS) (9). The activation of initiator caspases (2, 8–10) leads to the cleavagedependent activation of the effector caspases (3, 6, 7), which, being promiscuous proteases, cause the widespread cleavage of proteins in the cell (10). This, in turn, promotes the degradation of nuclear and cytoskeletal proteins and the activation of accessory enzymes, such as the caspase-activated DNAse (CAD), that degrades chromosomal DNA (11). The concomitant cleavage of nuclear scaffold proteins such as lamins leads to nuclear fragmentation (12), while proteolysis of actin, fodrin, and gelsolin (13) causes cell shrinkage and membrane blebbing. In addition, caspase activation is central to drive PS exposure on the outer leaflet of the plasma membrane, a key event in apoptotic cell recognition and clearance (14). Thus, caspase activity is largely accountable for the morphological and biochemical hallmarks of apoptosis, including the auto-digestion of cellular components and the generation of "find-me" and "eat-me" signals. Caspases can be activated when proteases that are normally secreted are released into the cytosol. For example, neutrophil elastase induces the unfolded protein response in vascular endothelial cells, promoting apoptosis via caspase-3/7 activation (15). The best-characterized apoptotic "find-me" signals are the nucleotides ATP and UTP (16), the chemokine CX3CL1 (17), ICAM3 (18), and the lipids lysophosphatidylcholine (LPC) (19) and sphingosine 1-phosphate (S1P) (20). Interestingly, these signals can be released as soluble mediators or become exposed on the surface of apoptotic microparticles, which detach from the main apoptotic body and are capable of diffusing in the extracellular environment (21).

Apoptotic bodies are easily engulfed by leukocytes (professional phagocytes) and are cleared from the tissue without any inflammatory impact. This process depends largely on the exposure of PS on the outer leaflet of the membrane, an evolutionary conserved "eat-me" signal for apoptotic cells. PS is recognized by a plethora of receptors, including TIM-1, TIM-3, TIM-4, BAI1, MerTK, and the stabilins 1 and 2, which will cause internalization of the apoptotic bodies by phagocytes (2, 22–25). Also, scavenger receptors such as CD36, might be able to interact directly with exofacial PS due to its negative charge (26). In addition to direct receptor-binding to PS, several soluble molecules were described to bridge phagocyte receptors to the phospholipid. They may originate from the phagocyte, the dying cell or the interstitial fluid. Examples of phagocyte-derived bridging molecules include milk fat globule EGF factor 8 (MFG-E8), developmental endothelial locus 1 (Del-1), growth arrest-specific 6 (Gas6), protein S and the complement factor C1q. Bridging proteins interact with PS via different PS-binding domains. For example, MFG-E8 secreted by macrophages and immature dendritic cells binds to PS on apoptotic cells via its Ca2+-independent discoidin-like C2 domain, while interacting with ανβ3/<sup>5</sup> integrins on the phagocyte membrane, resulting in cell engulfment (27, 28). In contrast, Gas6 and protein S bind PS via their γ-carboxyglutamic acid (Gla) domain. Unlike the discoidin-like C2 domain, binding of the Gla domain to PS requires Ca2+, in this way promoting apoptotic cell internalization (29, 30). C1q binds to apoptotic cells via its cationic globular head, and interacts with calreticulin-CD91 on phagocytes to promote efferocytosis (31, 32). Another phagocyte-derived protein, Annexin A1, can be translocated to the plasma membrane to interact with PS exposed on the apoptotic cell target, and this may contribute significantly to the anti-inflammatory effects of apoptotic cell clearance (33).

Complement C1q and additional bridging molecules such as IgM and collectins were proven to bind to "defects" in the plasma membrane of the apoptotic cell, including the presence of phosphatidylcholine, phosphatidylethanolamine, lyso-phospholipids, carbohydrates, and DNA (34). Collectins, such as mannose-binding lectin (MBL) and complement C1q, bind late apoptotic cells and also drive engulfment via interaction with CD91 and calreticulin on the macrophage in vitro. Calreticulin is an endoplasmic reticulum (ER) localized chaperone that normally facilitates folding and quality control of N-glycosylated proteins. As cells undergo apoptosis, calreticulin escapes and translocates to the plasma membrane, where it acts as an "eat-me" signal that is recognized by CD91 on phagocytes (35). In addition, a variety of other receptors and adaptor molecules have been reported to contribute to efferocytosis. These include Fcγ receptors, β2-glycoprotein I receptor, lectins, CD14, ABC transporters, scavenger receptors, and complement components [reviewed in (31, 36–42)]. Together this indicates that there is marked redundancy in receptors and ligands for the engulfment of apoptotic cells.

Interestingly, although PS exposure is a hallmark of apoptosis, forced PS exposure on viable cells does not trigger internalization (43). This is due to the presence of "don't eat-me" signals on viable cells, including CD31, CD46, CD47, and CD61, which disable target cell engulfment. The downregulation of "don't eat-me" signals, such as CD47 and its binding partner SIRPα, contributes to internalization of apoptotic bodies, indicating that a coordinated effort between the dying cell and the phagocyte likely exists (44).

# NON-APOPTOTIC CELL DEATH

Necrosis is generally considered to be a drastic and uncontrolled form of cell death, characterized morphologically by nuclear and organellar swelling (oncosis) and plasma membrane rupture (**Figure 1**) (10). Due to the loss of membrane integrity, the intracellular contents are spilled out by the dying cell. The exposure of necrotic cell content (or debris) is abrupt and lacking in processing, causing it to be released in a disorderly fashion into the tissue, without the specific cues of its apoptotic counterpart. This causes necrotic debris to be potent inducers of inflammation, through activation of pattern-recognition receptors such as toll-like receptors (TLRs), NOD-like receptors (NLRs), and C-type lectins (CLECs), among others. Generally, necrotic cell death is problematic to tissues, prompting the need for immediate removal of debris, delaying the regeneration required after injury and sustaining collateral inflammatory damage. Interestingly, the recent identification of signaling pathways that are activated before and during necrosis have prompted reconsideration of this type of cell death as multiple, distinct types of events, at least some of which are tightly regulated and not always accidental. Recent subclassifications of necrosis include pyroptosis, necroptosis, parthanatos, ferroptosis, oxytosis, ETosis, and secondary necrosis (45).

**Pyroptosis** is a type of necrotic cell death caused by extensive inflammasome activation. It occurs in cell types that express inflammasome components, such as macrophages, upon infection (e.g., with the intracellular pathogen Salmonella typhimurium) or LPS treatment (46). Pyroptosis is caused by the formation of gasdermin D pores, which assemble at the plasma membrane after proteolytic processing of their precursor by inflammasomes containing activated caspase-1 or -11 (47, 48). Gasdermin pores create a path for the release of IL-1β, but secondarily cause cell lysis by excessive permeabilization of the plasma membrane. **Necroptosis** differs from other modalities of necrosis by the involvement of receptor-interacting protein kinase 1 (RIPK1) and RIPK3, which recruit and phosphorylate the mixed lineage kinase domain-like protein MLKL (45). Subsequently, MLKL oligomerizes, translocates to the inner leaflet of the plasma membrane and promotes membrane permeabilization and cell death (49, 50). Curiously, necroptosis requires the inhibition of caspase-8, which otherwise causes the cells to die by apoptosis. This may restrict the relevance of this death pathway in vivo, since caspase-8 inhibition may only occur in some viral infections (51). **Parthanatos** is a necrotic mode of cell death that depends on poly(ADP-ribose) polymerase proteins (PARPs). PARPs are typically activated by DNA breaks from ultraviolet light and by alkylating agents (52). By causing poly (ADP-ribosyl)ation of target proteins, PARPs may deplete cells of NAD<sup>+</sup> and consequently of ATP, causing necrotic cell death.

**Ferroptosis** is the necrotic cell death induced by irondependent oxidative stress (53). It was postulated that iron catalyzes the lipid peroxidation triggered by the ferroptosisinducing molecules erastin and RSL3, or by inhibiting the glutamate/cystine antiporter. It was later found that these pathways converge on the reduction of intracellular glutathione (GSH) levels and impaired GSH peroxidase 4 (GPX4) activity, leading to the accumulation of lipid-based reactive oxygen species and cell death (54). As expected, iron-chelators and lipophilic antioxidants were found to be potent inhibitors of ferroptosis (55). A related oxidative stress-dependent necrotic cell death, **oxytosis**, involves GSH depletion, 12-lipoxygenase activation and opening of cGMP-gated channels on the plasma membrane (56). This leads to calcium influx and activation of the calpain-cathepsin cascade, causing lysosome membrane permeabilization and necrosis (57).

In contrast to the "passive" nature of classical necrosis, one of the necrotic death pathways involves purposeful intracellular content exposure. Since its description in 2004 by Zychlinsky and collaborators (58) (neutrophil) extracellular traps (ETs), which consist primarily of extruded DNA, have been studied extensively. Subsequent work determined that cells may die during ET production, a process dubbed NETosis (59, 60). NETosis was later shown to occur in several cell types other than neutrophils, such as monocytes (61), mast cells (62), and eosinophils (63), making the name ETosis more appropriate. Generally, ETosis requires NADPH oxidasedependent reactive oxygen species production, leading to chromatin decondensation, nuclear disruption and release of chromatin complexed with granular/cytoplasmic proteins (59), although the mechanisms underlying the process may vary between cell types.

Necrosis may also take place even after apoptosis has occurred. If apoptotic cells are not cleared in a timely fashion, the apoptotic bodies may decay and lose plasma membrane integrity, leaking their contents in a similar manner as a primary necrotic cell would (64). This event is named secondary necrosis and it shares common features with both apoptotic and necrotic cell death. The intracellular debris produced by secondary necrosis undergo apoptotic caspase processing, causing it to be qualitatively different from primary necrotic debris (65). For instance, secondary necrotic debris are considerably smaller, contain digested chromatin, prostaglandin E2 and high levels of uric acid, but very low ATP levels (65). These change drastically the manner by which the organism deals with the debris, as exemplified by the higher efficiency of complement C1q and DNAse I in degrading chromatin from secondary necrotic cells (66) and the potent anti-inflammatory polarization of macrophages elicited by C1q-covered late apoptotic debris (67).

#### NECROTIC "FIND-ME" SIGNALS AND THEIR RECEPTORS

In contrast to apoptotic "find-me" signals, necrotic cells may not have enough time or energy to process their own signals. A myriad of molecules has been shown to be released by dying cells, many of which fall into the category of damage-associated molecular patterns (DAMPs): bona fide cellular components that are normally concealed inside the cell, but that become exposed to the extracellular environment upon cell damage or death. Some well-established DAMPs include mitochondria-derived N-formylated peptides, DNA and RNA, the nuclear protein HMGB1, histones, actin, calcium-binding S100 proteins, heatshock proteins (HSPs), ATP and uric acid, among many others (68, 69). Necrotic cells may also release pre-stored inflammatory mediators, such as IL-1α, IL-33, and chemokines, which may directly or indirectly recruit phagocytes to the vicinity. In addition, the occurrence of necrosis and the consequent exposure of "unusual" molecules promptly activates the proteolytic cascades of complement and coagulation. The activation of complement on necrotic debris can itself generate several "findme" signals, including the powerful chemoattractant C5a. Below, we discuss established necrotic "find-me" signals. After reading this section the reader may agree that neutrophils and monocytes respond to a "complex pool of exogenous signals, of which no single cue is absolutely required for migration" (70).

#### Formyl-Peptides

Formyl-peptides (or N-formylated peptides) are classically generated in the course of bacterial protein synthesis, which is initiated by N-formyl-methionine residues. Mitochondria, sharing the bacterial ancestry, initiate protein synthesis similarly, thus creating a formylated protein reservoir inside the eukaryotic cell. Upon necrosis, the release and cleavage of mitochondrial proteins produces formyl-peptides, causing massive leukocyte activation and recruitment in a variety of necrotic states (71, 72).

Formyl-peptides are powerful chemoattractants. The most used analog, fMLP, activates neutrophils in the picomolar range (73, 74). It is considered an end-target chemoattractant, which bypasses the signaling of intermediate chemotactic molecules such as CXCL8 (IL-8) and LTB4 (75). Formyl-peptides bind FPR1, FPR2, and FPR3 receptors, although the classical chemotactic effects are mainly mediated by FPR1 activation. Whereas, FPR1 and FPR2 are expressed in several cell types, especially neutrophils and macrophages, FPR3 is much less understood (76). Upon ligand binding, FPR1, induces multiple intracellular signaling pathways: Gα activation signals via the MAPK pathway and the small GTPases CDC42 and RAC to stimulate migration and phagocytosis; Gβγ transduce signals via PI3Kγ and PLCβ to stimulate superoxide production and transcriptional regulation in phagocytes (77).

The role of formyl-peptides as a necrotic "find-me" signal is firmly established in the literature. In a seminal paper where McDonald et al. (71) used focal thermal injury of the liver, FPR1 activation of neutrophils was the key step required for migration into the necrotic area. The neutrophils initially traveled to the liver stimulated by an intravascular gradient of CXC chemokines. Upon reaching the edge of the necrotic site, the neutrophils switched to a FPR1-dependent migration mode, presumably chasing formyl-peptides emanating from the mitochondria of necrotic cells. Furthermore, the dependence of neutrophils on formyl-peptide gradients for recruitment to necrotic sites was confirmed in clinically-relevant disease models of druginduced liver injury (78) and liver ischemia-reperfusion (79). Interestingly, when the necrotic injury is extensive enough, as in severe trauma (crushes, fractures, burns) or acute liver failure, the formyl peptides can be released in such a significant amount that they cause systemic inflammation, affecting lung function in mice and humans (72, 78).

Though most studies have focused on neutrophil chemotaxis and activation by formyl-peptides, macrophages also express FPR1 and are sensitive to formyl-peptide stimulation. Human PBMCs produce significant amounts of CXCL8 in the presence of mitochondrial extracts containing formyl-peptides (80). Interestingly, the response is stronger when formyl-peptides are applied in conjunction with other stimulatory DAMPs such as HMGB1. This suggests that formyl-peptides may indirectly recruit phagocytes to necrotic sites by inducing the production of additional chemoattractants (CXCL8) by resident macrophages. Formyl-peptides induce monocyte recruitment in vitro (81, 82), however, the relevance of FRP1 signaling in monocyte recruitment to necrotic sites in vivo remains elusive.

Importantly, questions about formyl-peptides as necrotic "find-me" signals remain. For example, the mechanism by which peptides are retained in necrotic areas in order to signal to leukocytes is unclear. Chemokines interact with glycosaminoglycans in order to form a gradient in the vasculature (83), but no mechanism has been proposed for the gradient formation of formyl-peptides. One should also consider that necrotic formyl-peptides may be very heterogeneous, varying in peptide length from a few to several amino acids. This may also impact the localization and agonistic activity of the formylpeptides in vivo.

#### Chemokines

Chemokines are chemotactic cytokines that dictate the localization and mobilization of leukocyte populations in the organism (84). Chemokines are produced in a constitutive fashion and/or in response to stimuli such as those that activate pattern-recognition receptor (85). In the context of necrosis, chemokines play the roles of primary and secondary "find-me" signals, meaning that they can originate from both the dying cells and from healthy bystander cells. However, the multitude and promiscuity of chemokines and chemokine receptors adds a significant layer of complexity to the study of these mediators in vivo (84). CC and CXC chemokines can be released essentially by any cell type, including resident leukocytes (85). For instance, the chemokine CXCL1 can be produced by endothelial cells, pericytes, hepatocytes, macrophages, and fibroblasts (86–89). In addition, leukocytes can produce chemokines in an autocrine fashion, such as when neutrophils secrete CXCL2 during transendothelial migration (87) and Kupffer cells that release CCL2 during necrotic injury (90). Thus, there is an abundance of chemokine sources that can direct the migration of phagocytes to necrotic sites, reflecting the importance of chemokines as necrotic "find-me" signals.

The chemokines CXCL1 and CXCL2 (in mice) or CXCL8 (in humans), among others, have been known as powerful chemoattractants for neutrophils for decades (91). They activate CXCR1 and CXCR2 receptors to induce neutrophil polarization and migration, an effect strongly dependent on PI3Kγ signaling (92). CXC chemokines were shown to be the first signal guiding neutrophils to sites of focal necrosis in the liver (71) and suffice to induce neutrophil accumulation in zebrafish in vivo (93, 94). In contrast to formyl-peptides, the CXC chemokines were actually shown to form an intravascular gradient in the vicinity of necrotic areas, which is required for proper neutrophil recruitment to the injury site. These chemokine gradients are built on heparan sulfate proteoglycans expressed by the endothelium; they are long lasting and extend hundreds of microns from the site of injury (71, 94). This promotes the recruitment of patrolling neutrophils from the vasculature far from the original insult area. Moreover, CXCL1/CXCL2 signaling via CXCR1/CXCR2 receptors act in conjunction with formyl-peptides in the case of widespread hepatic necrosis (e.g., drug-induced liver injury), in which both pathways are required for maximal neutrophil recruitment to the interior of necrotic areas (95). Despite being considered redundant, evidence shows that CXCL1 and CXCL2 act on neutrophils sequentially to promote successful diapedesis and recruitment to inflamed muscle (87). Conversely, there is mounting evidence that neutrophils expressing lower levels of CXCR1 may transmigrate in reverse into the bloodstream (96). This has been corroborated by observations of neutrophil reverse migration (e.g., away from the site of necrosis and back into the bloodstream) in both zebrafish (97) and mice (98). This suggests that the role of CXC chemokines as "find-me" signals may be much more complex than anticipated, regulating initially the recruitment of neutrophils to necrotic areas and subsequently directing their egress back to the vasculature.

CC chemokines such as CCL2, CCL3, and CCL5 are notably active in cells of the monocytic lineage. Even though they are present in a variety of parenchymal and non-parenchymal cells, the CC chemokine receptors, especially CCR2, are highly expressed in monocytes and macrophages (85). Using transgenic mice, Dal-Secco et al. identified two different monocyte subsets that are recruited to necrotic sites: a classical pro-inflammatory CCR2hi-CX3CR1low population and an alternative patrolling CCR2low-CX3CR1hi population (99). They showed that CCR2hi monocytes migrate to the edge of the necrotic area after the initial wave of neutrophil recruitment, and this was dependent on CCR2 expression. The monocytes persisted in the necrotic area for days, where they transitioned in situ into a CX3CR1hi population. The reprogramming of the monocyte population was dependent at least partially on the cytokines IL-4 and IL-10, and was required for the timely resolution of the necrotic injury. Interestingly, CCL2 and CCL3 were found to be significantly increased in the necrotic liver of humans, correlating to a CCR2-dependent recruitment of CD68-positive monocytes to the necrotic areas (100).

Of note, chemokine receptors other than the ones mentioned above may control different leukocyte populations, playing roles that are still unclear in the context of necrotic debris. CXCR3 and its ligands CXCL9 and CXCL10 seem to control the population of NK and NKT cells, such that deficiency of CXCR3 causes a significant reduction of both cell populations in necrotic liver (101). Similarly, the chemokine CXCL12, known to control neutrophil egress from the bone marrow via CXCR4 (102) may control later events in tissue necrosis, such as revascularization (103).

#### Leukotriene B4 (LTB4)

LTB4 is a mediator derived from membrane phospholipids. The activation of cytosolic phospholipase 2 (cPLA2) initiates the cleavage of phospholipids to generate arachidonic acid. This fatty acid is used as substrate by the lipoxygenase 5-LOX to produce LTB4, among other intermediate eicosanoids. LTB4 is a powerful chemoattractant to neutrophils. It activates the BLT1/ LTB4R1 receptor, which, coupled to Gαi, stimulates neutrophil migration via Src-family kinases and Rho GTPases (104). Since LTB4 synthesis demands several enzymatic steps, it is unlikely to be released by necrotic cells as a DAMP. Instead, it can be produced in a matter of minutes by leukocytes such as neutrophils, macrophages, and mast cells upon demand (104). However, it has been recently demonstrated that the enzymatic machinery for LTB4 synthesis can be localized to multivesicular bodies and secreted as exosomes in vitro (105). In this way, LTB4 may also be produced independently of the cell and travel in the aqueous environment concealed in exosomes, increasing its diffusion range and persistence in the tissue.

LTB4 is considered an "intermediate target" chemoattractant. Nevertheless, it is required for the rapid migration and concentration of neutrophils in focal necrotic sites, a phenomenon dubbed "neutrophil swarming" (106). LTB4 is produced by neutrophils recruited to necrotic foci in order to further amplify neutrophil recruitment to the area, forming the typical densely-populated clusters that are associated to neutrophil swarming. Neutrophil-derived LTB4 can act as a signal relay molecule (107) that is necessary for cell-cell communication to produce optimal aggregation of neutrophils at the injury site (106). Moreover, LTB4 was shown to act in conjunction with other necrotic "find-me" signals such as formyl-peptides and chemokines (106, 107), supporting the idea that there is no absolute necrotic "find-me" signal, but instead, a synergistic pool of signals that vary in chemotactic potency and range to mediate an integrated response.

The importance of LTB4 as a necrotic "find-me" signal has been confirmed in several models other than laser-induced focal skin injury. In spinal cord injury, LTB4-BLT1 signaling was required for the recruitment of neutrophils to the injury site (108). Interestingly, BLT1 knockout or pharmacological blockage of the receptor reduced neutrophil recruitment significantly, but did not alter monocyte recruitment to the injured spinal cord area. In the K/BxN model of inflammatory arthritis, inhibition of 5-LO led to a significant reduction of neutrophil migration to arthritic joints and amelioration of the disease (109). There, LTB4 was also produced locally by infiltrating neutrophils. In drug-induced liver injury, deficiency of 5-LO prevented mortality associated with acetaminophen overdose, which was correlated with reduced recruitment of phagocytes to the necrotic liver (110).

It is clear that LTB4 is an essential necrotic "find-me" signal to neutrophils. Yet, many questions pertaining its production and release still remain. For example, the transport of lipid mediators across membranes is still poorly defined, as is its mode of release from the nanoscopic exosomes. It would be interesting to assess whether exosomes are able to bind to the vasculature, perhaps stimulating leukocyte recruitment at long range. In addition, based on the role of LTB4 in the skin, one could wonder if it is especially relevant in tissues with abundant extracellular matrix, where it would be better retained and less prone to degradation.

# Hydrogen Peroxide (H2O2)

H2O<sup>2</sup> is a reactive oxygen species commonly generated in organelles such as mitochondria and phagosomes (111). The signaling capabilities of H2O<sup>2</sup> are not limited to mammalian cells: it also serves as a major chemotactic signal in other species, such as zebrafish (Danio rerio) and Drosophila. Niethammer et al. showed formation of a H2O<sup>2</sup> gradient minutes after wounding zebrafish, which extended up to 200µm from the site of injury (112). The H2O<sup>2</sup> was created by the activity of the enzyme Duox, a NADPH oxidase expressed in epithelial cells, and was necessary for rapid leukocyte recruitment to the injury site. In Drosophila, hemocytes also respond to H2O<sup>2</sup> emanating from the wound (113). In this species, H2O<sup>2</sup> was also derived from Duox and inhibition of the enzyme by siRNA knockdown or using diphenylene iodonium blocked the recruitment of hemocytes to the wound. Interestingly, it was later shown that wounding of tissues in zebrafish and flies causes a calcium wave across the tissue, which precedes and is responsible for the activation of Duox via its EF-hand motif, initiating the production of the H2O<sup>2</sup> gradient (114).

Leukocytes must have a mechanism to sense this transient H2O<sup>2</sup> gradient emanating from the injured cells. The redox sensor is seemingly the Src family kinase Lyn, which is activated by wound-derived H2O<sup>2</sup> and mediates recruitment of neutrophils to injury sites in zebrafish (115). Oxidation of cysteine C466 by H2O<sup>2</sup> activates Lyn, which in turn contributes to neutrophil migration toward the wound. Of interest, H2O<sup>2</sup> is also chemotactic in murine (116) and human neutrophils (115), and Lyn is expressed in all mammalian leukocytes, with the exception of T cells (which nevertheless express related Srcfamily kinases). Thus, the role of H2O<sup>2</sup> as a necrotic "find-me" signal spans several species and leukocyte types. Beyond the direct effects that H2O<sup>2</sup> has on phagocyte recruitment to injury sites, it can also regulate the chemotactic responses to other "findme" signals, such as fMLP, LTB4, and CXCL8 (117). Indeed, generation of reactive oxygen species by the NADPH oxidase at the leading edge of neutrophils is important to oxidize and inhibit the phosphoinositide phosphatase PTEN, maintaining high levels of PI(3,4,5)P<sup>3</sup> at the leading edge and supporting the directional migration of neutrophils (118).

#### Purines

Nucleotides are among the earliest molecules released by damaged and dying cells (119). Nucleotide sensing occurs via P2Y and P2X receptor families, which are G protein-coupled receptors and nucleotide-gated ion channels, respectively. These receptors are numerous and vary in sensitivity to different nucleotides (e.g., ATP, ADP, UTP), but the majority of the studies have focused on the role of ATP and its degradation products. ATP is very abundant in the cytoplasm, ranging from 3 to 10 mM (120). When released actively or passively by cells, ATP is rapidly hydrolyzed by ectonucleotidases into ADP, AMP, and adenosine (119). Yet, despite its very short half-life outside the cell, ATP nevertheless has pivotal effects in leukocyte activation and migration.

Chen and collaborators demonstrated that neutrophils exposed to a gradient of fMLP release ATP at the leading edge of the cell, amplifying the chemotactic response to the formyl-peptide. This effect was mediated by ATP signaling via P2Y2 receptors and subsequently by activation of A3 receptors by adenosine derived from ATP hydrolysis (121). P2Y2 activation by ATP was also required for chemotaxis of human neutrophils toward CXCL8 (122), but in this case adenosine signaling did not play a role. Moreover, it was demonstrated that macrophages utilize the same autocrine ATP amplification loop to migrate toward C5a. Blockage of P2Y2 also impaired macrophage chemotaxis in vitro and in vivo (123). Clearly, purinergic signaling is involved in phagocyte migration to several stimuli, but this is not sufficient to characterize it as a chemotactic agent. Indeed, it was demonstrated that ATP itself is not directly chemotactic to macrophages. Instead, it induces lamellipodial extensions and chemokinesis (increased random displacement) (124), without directing the migration. These studies suggest an indirect effect of ATP, that though not acting as a chemoattractant, acts in an autocrine capacity in phagocytes to maximize the response to other chemoattractants, including fMLP and chemokines.

ATP was initially implicated as a "find-me" signal of apoptotic cells (16). The authors showed that ATP and UTP released during apoptosis were required for monocyte migration toward supernatant of apoptotic cells, in a P2Y2-dependent manner. Also, the migration of monocytes toward apoptotic cells in vivo was impaired in the absence of P2Y2. In necrotic injuries, the role of purinergic signaling is even more interesting. Applying focal necrotic injury to the liver, it was shown that ATP is required for invasion of peritoneal macrophages into the necrotic area (125). Curiously, the peritoneal macrophages, which can be found floating in the peritoneal fluid, took this avascular route to the necrotic site by recognizing ATP released from the dead cells, which prompted the macrophages to arrest at that site. The use of apyrase (to degrade ATP and ADP) and P2X7 blockage reduced significantly the infiltration of macrophages from the peritoneum to the injury site. In the case of focal injury of the brain, the extension of microglial processes to the area of injury was also found to be mediated by ATP (126). The rapid convergence of microglial extensions to the necrotic site took place without displacement of the main cell body and was dependent on ATP and P2Y receptors. Uderhardt and collaborators found a similar response of peritoneal macrophages to a focal necrotic injury. Sessile resident macrophages extended membrane processes toward dead cells in order to cloak the debris from patrolling neutrophils, thereby minimizing inflammation (127). The extension of the macrophage processes could be blocked by apyrase or joint inhibition of P2X and P2Y receptors, indicating an elevated redundancy in purinergic signaling. In zebrafish, wounding causes ATP release and P2Y receptor activation, which in turn activates Duox to produce H2O2, recruiting phagocytes to the injury site (128). In this species, the effects of the nucleotide are not limited to phagocytes, as ATP is also involved in rapid wound closure by stimulating epithelial cell motility (129).

It is important to highlight the differences in the function of P2Y and P2X receptors. As mentioned, P2Y receptors, especially P2Y2, have been implicated in regulating the migration of phagocytes to diverse necrotic "find-me" stimuli. P2Y receptors are metabotropic, transducing signals via RhoA, Rac and PLCβ, leading to cytoskeletal rearrangement and increased intracellular calcium (130). P2X receptors, on the other hand, being ion channels activated by nucleotides, signal in a fundamentally different way. A classic example is the role of P2X7 in inflammasome activation in macrophages. In this instance, P2X7 mediates K<sup>+</sup> efflux from cells stimulated by ATP, a major step in the activation of the inflammasome complex and caspase-1 that eventually culminates in the release of interleukin-1β (IL-1β). IL-1β is able to prime the production of several other chemotactic agents such as chemokines and lipid mediators, but like P2X7, lacks intrinsic chemotactic activity.

# Complement

The complement system comprises an evolutionary ancient set of fluid-phase proteins and receptors, present in vertebrates and invertebrates. It is at the core of the immune system and mediates a cross-talk between innate and adaptive responses (131–134). The complement cascade is activated by a myriad of self and non-self molecules, which initiate the proteolytic cleavage of complement proteins into fragments that deposit onto the target or are released into the extracellular fluid to signal to neighboring cells and leukocytes.

Necrotic debris can initiate complement activation through all 3 pathways: classical, alternative and lectin (133). Exposure of intracellular components such as DNA or mitochondria activate complement directly (135, 136), and natural IgM and IgG autoantibodies can bind necrotic debris to initiate complement by the classical pathway (137). In addition, there are numerous adaptors and pattern-recognition receptors that detect necrotic debris and initiate the complement cascade by themselves, including mannose-binding lectin (MBL), pentraxins, ficolins, and histidine-rich glycoprotein (34). In the specific context of this section, the production of complement C3a and C5a fragments (anaphylatoxins) is central, since these stimulate chemotaxis of leukocytes (138). Their respective receptors, C3aR and C5aR, that are expressed primarily in myeloid cells, are G protein-coupled receptors signaling via PI3K activation, MAPK activation and intracellular calcium mobilization (139).

The activation of complement on dying or necrotic cells, measured by the deposition of C3b/iC3b, has been demonstrated in several tissues, including the liver (140–142), muscle (143, 144), brain (145, 146), joint (147, 148), and intestines (149, 150). The presence of complement deposited on damaged tissues was already strong indication that C3a and C5a were being generated, but subsequent studies focused on the exact role of each fragment in disease. In muscle injury induced by cardiotoxin, it was shown that complement is activated via the alternative pathway (spontaneously), and that C3a-C3aR signaling was required for monocyte migration to the tissue (143). Deficiency in C3aR reduced the recruitment of monocytes to the injured muscle significantly, although neutrophil migration was unaffected. Moreover, C5aR was not required for the migration of either monocytes or neutrophils to the muscle, indicating specificity of C3a activity in this setting. In liver injury by ischemia/reperfusion, neutrophil migration requires complement activity. C5a is produced early during injury and formation of the complement membrane attack complex (MAC) plays an additional role in amplifying neutrophil recruitment, likely via release of IL-1β and additional DAMPs (140).

Complement inhibition in intestinal ischemia/reperfusion injury, a severe model of intestinal damage, also minimizes neutrophil recruitment and disease severity (149, 150). Interestingly, whilst complement inhibition presumably inhibited the generation of C3a and C5a in the injured intestine, it also inhibited the production of another chemoattractant, LTB4. This shows again a synergistic relationship between different classes of "find-me" signals, acting simultaneously or sequentially to guide leukocyte recruitment to necrotic debris. In the joints, the synovium is a site of both synthesis and deposition of complement (131). The alternative complement pathway plays a major role in the pathogenesis of arthritis from the initiation phase (when synoviocytes can be damaged directly by complement) to the chronic inflammatory stage (147, 148). Importantly, both C3aR and C5aR are required for the recruitment of neutrophils and macrophages to the damaged joint (148), showing yet again a degree of redundancy in the role of anaphylatoxins in phagocyte recruitment to the joint. Altogether, there is abundant evidence that complement by-products are released during necrosis and that they play a role in attracting phagocytes to injury sites.

### NECROTIC "EAT-ME" SIGNALS AND THEIR RECEPTORS

Apoptotic cells have to be cleared quickly and efficiently to prevent secondary necrosis, which would lead to the release of intracellular components and inflammation. Similarly, cells dying from primary necrosis need to be removed efficiently, as they could be a detrimental source of autoantigens and may trigger excessive inflammation (**Figure 2**). In the case of necrotic debris, the mechanism of recognition by professional phagocytes is not fully understood. As the necrotic cell can be disintegrated into small debris, it has been suggested that engulfment of necrotic cells resembles macropinocytosis, in which macrophages develop membrane ruffles which protrude around the target material (151–153). As reviewed briefly above, multiple receptors implicated in the clearance of apoptotic debris have been described. By contrast, much less is known about the receptors and ligands involved in the uptake of necrotic debris. Remarkably, as the evidence emerges, it is becoming apparent that some necrotic "eat-me" ligands overlap with the equivalent apoptotic signals. For example, necrotic cells also expose PS, although the mechanism underlying such exposure differs drastically. In line with this, many of the molecules that bridge PS for efferocytosis (complement, collectins and pentraxins) have also been shown to bind necrotic cells. Nevertheless, there are "eat-me" signals that apply uniquely to necrotic debris. For instance, complement C1q deposition represents a hallmark of necrotic debris, but it is absent on apoptotic debris (154). Another distinguishing "eat-me" signal is annexin A1, which is translocated to the plasma membrane of necrotic cells to promote phagocytic uptake (155). Below, we will summarize and discuss necrotic "eat-me" signals, comparing and contrasting them to apoptotic "eat-me" signals.

#### Complement

Opsonization of targets by complement components C1q, C3b, and C4 alerts phagocytes bearing complement receptors such as CR1, CR3, and CR4. Moreover, it is clear that a functioning

complement system is required for efficient handling of dying and dead cells. Recent evidence points to a role of complement deposition in the clearance of late apoptotic/necrotic cells, rather than early apoptotic cells (154). In fact, most apoptotic cells are cleared while at an early apoptosis stage, when complement plays a minor role. It is only when apoptotic cells persist into late apoptotic/secondary necrotic stages that complement opsonization enhances recognition by phagocytes (156).

Phagocytosis of late apoptotic/necrotic Jurkat cells is impaired in individuals with deficiencies in C1q, C2, C3, or C4. In contrast, the MBL and alternative pathway did not participate in phagocytosis of debris, suggesting that opsonization by C3 fragments and the involvement of the classical pathway are mostly responsible for the clearance of necrotic cells (157). Similarly, complement components C3 and C4 bind immediately to necrotic peripheral blood lymphocytes. In contrast, irradiated lymphocytes undergoing apoptosis only displayed a weak binding of complement components for up to 2 days. At day 3, when secondary necrosis had ensued, C1q, C3b, and C4 all bound with higher affinity (154). Also, the clearance of necrotic cells is increased in presence of serum, and adding C1q to C1q-depleted serum markedly increased uptake of primary necrotic cells (158). Another study investigated complement deposition on viable, early apoptotic and late apoptotic (secondary necrotic) Jurkat cells (159). In this study, binding of C3 and C4 to early apoptotic cells was similar to that of viable cells, while secondary necrotic cells had a substantial binding of C3, C4, and at some extent C1q. The necrotic cells also bound IgM, and depletion of plasma IgM abolished most of the complement binding, supporting a role for the classical pathway of complement activation on late apoptotic/necrotic cell clearance (159).

Macrophages were shown to engulf apoptotic cells after C1q and MBL opsonization. Calreticulin released from dying cells bound macrophages via CD91/LDL receptor related protein 1, and was shown to recognize the collagen tails of C1q and MBL attached to the surface of the dying cell (31). Besides IgM, C1q binding to necrotic cells can be initiated via molecules of the acute-phase protein pentraxin family. The classical (short) pentraxins—including serum amyloid protein (SAP) and Creactive protein (CRP)—are produced in the liver in response to IL-6 and play a role as opsonins by binding to cellular debris and late apoptotic cells (160, 161). The long pentraxin PTX3 is produced by hematopoietic and stromal cells as a response to a primary pro-inflammatory signal such as LPS, IL-1β, and TNF-α. PTX3 has multiple functions including complement activation on necrotic cells that results in cell clearance and reduced tissue damage (162). However, PTX3 can also limit excessive complement activation by promoting deposition of complement Factor H, a major inhibitor of the alternative pathway of the complement system. Under normal conditions, Factor H binds to self-surfaces, where it inactivates accidental C3b deposition on healthy cells. By directing Factor H to the surface of dying cells, PTX3 limits tissue damage while still increasing phagocytic clearance (163). Moreover, both human and mouse pentraxins recognize FcγRI and FcγRII, and binding of pentraxins to cellular surfaces results in phagocyte activation (164). In line with data suggesting complement deposition on late apoptotic/necrotic cells, the collectin MBL was found to bind to both late apoptotic and necrotic cells, but not early apoptotic cells. MBL binding initiated C4 deposition onto the necrotic cells and addition of C1q inhibited MBL opsonization of cells (165). Taken together, these observations imply that complement deposition and recognition function as a mechanism for the clearance of necrotic cells and as a backup for clearance of late apoptotic material undergoing secondary necrosis.

#### Phosphatidylserine

Although PS was long thought to be an exclusive marker of apoptosis, evidence is gathering that exposure of PS is also a hallmark of necrosis. In a study comparing internalization of apoptotic and necrotic cells, macrophages were shown to selectively engulf apoptotic and necrotic cells, while leaving living cells untouched. The engulfment of both apoptotic and necrotic cells was PS-dependent, suggesting that externalization of PS is a common trigger for the clearance of both types of cell debris (152). Annexin A5 (or Annexin V), commonly used as a marker of apoptosis due to its PS binding capacity, also binds to necrotic cells, supporting the occurrence of PS exposure during necrosis. Moreover, treatment with recombinant Annexin V inhibited phagocytosis of both apoptotic and necrotic cells by mouse macrophages, suggesting that PS exposure is required in both instances (152).

The difference in morphology between apoptotic and necrotic cells suggests that the mechanisms of PS exposure differ. PS exposure was observed in necrotic neurons of the nematode Caenorhabditis elegans, where it was facilitated by the homolog of the calcium-dependent scramblase TMEM16F and by CED-7, a member of the ATP-binding cassette (ABC) transporter family. However, rupture of the necrotic cells into particles can also readily account for exposure of PS without invoking specific externalization mechanisms. Zargarian et al. demonstrated that necroptotic cells also expose PS as an "eatme" signal as phosphorylated MLKL translocates to the plasma membrane. The externalization of PS by necroptotic cells induced recognition and phagocytosis; they stained positive for Annexin A5 and exposed PS prior to overt permeabilization. The dying cells also released PS-exposing extracellular vesicles, thereby alerting neighboring cells of the impending cell death (166).

Although both apoptotic and necrotic cells expose PS, the efficiency of their clearance differs drastically: the engulfment of necrotic cells is considerably less effective, both quantitatively and kinetically. The mechanisms underlying this difference remain obscure, but down-regulation of "don't eat-me" signals in apoptotic, but not necrotic cells is a distinct possibility. Importantly, clearance of necrotic cells is carried out not only by phagocytes like macrophages, but also by non-professional phagocytes. In comparison to macrophages, engulfment by nonprofessional cells is slow and engulfment events were only detectable after 2.5 h. But, by taking up neighboring necrotic cells, non-professional cells remove a portion of the billions of cells that die daily during normal turnover (167).

## Annexin A1

Annexin A1 was first believed to translocate to the surface of apoptotic cells, where it was proposed to function as a bridging protein that facilitates their phagocytic uptake (168, 169). However, this interpretation was recently revised, as it was demonstrated that annexin A1 rarely translocates in apoptotic cells; instead, its translocation to the cell surface is rather a hallmark of secondary necrosis (155). As proposed earlier for apoptosis, in necrotic cells annexin A1 is believed to function by bridging PS to the phagocyte surface to promote uptake. This interaction also dampens the secretion of proinflammatory cytokines by the macrophages that ingested the necrotic cell. This implies that clearance of necrotic debris can have anti-inflammatory effects. After translocation, annexin A1 is proteolytically cleaved at the cell surface by ADAM10, which generates a small peptide with chemotactic activity toward monocytes, thus generating a monocytic "find-me" signal for the necrotic debris (170).

## Histidine-Rich Glycoprotein (HRG)

HRG is an abundant 75 kDa plasma glycoprotein that has a multi-domain structure known to interact with many ligands including Zn2+, heparin, heparan sulfate and plasminogen. HRG has been shown to function as an adaptor molecule that tethers plasminogen to glycosaminoglycan-bearing surfaces to regulate plasminogen activation (171). HRG was also demonstrated to distinguish between apoptotic cells and necrotic cells by binding to cytoplasmic ligands exposed by necrotic cells. This interaction, mediated by the amino-terminal domain of HRG, results in an opsonic function, encouraging the phagocytosis of the necrotic cell. In contrast, HRG does not opsonize apoptotic cells and thus, may play an important physiological role in the selective clearance of necrotic debris (34).

## CD14

Initially, it was thought that the macrophage plasmalemmal glycoprotein CD14 was specific for recognition and clearance of apoptotic cells, as treatment with an anti-CD14 antibody reduced the phagocyte interaction with apoptotic but not necrotic cells (172). CD14 also recognizes LPS and it was initially thought its interaction with apoptotic cells occurs also via its LPS-binding domain, but this view was subsequently revised (173). Indeed, unlike LPS, binding of macrophages to apoptotic cells does not generate pro-inflammatory signaling. Later studies found a significant role for CD14 also in the clearance of necrotic cells. (158). ICAM-3 on the surface of dying cells may serve as the ligand recognized by CD14 (174).

#### Scavenger Receptors

Scavenger receptors were originally discovered by their capacity to recognize and remove modified lipoproteins. They are structurally diverse and recognize a variety of ligands, including DAMPs, oxidized PS and phosphatidylcholine (175). As PS is exposed on necrotic cells, this raises the possibility that PS-binding scavenger receptors may function as receptors for necrotic cells. SR-B1 and CD36 are class B scavenger receptors, and were the first cell surface receptors appreciated to recognize anionic phospholipids such as PS (26). CD36 for instance, which is highly expressed in macrophages, is involved in the phagocytosis of necrotic lymphocytes in vitro (158). Antibody blockage of CD36 caused a significant, yet partial, reduction in the macrophages ability to bind and internalize necrotic cells. Macrophages, dendritic cells and endothelial cells also express the scavenger receptor class F (SCARF1), that can recognize and engulf apoptotic cells via C1q (176). Given the earlier findings that C1q binds to late apoptotic/secondary necrotic cells, SCARF1 can potentially operate as a receptor for necrotic cells. In fact, loss of SCARF1 impaired uptake of dying cells, and SCARF1-deficient mice had accumulation of dying cells in tissues, leading to generation of autoantibodies to DNA-containing antigens and development of lupus-like disease (176).

#### PHAGOCYTE-INDEPENDENT CLEARANCE OF NECROTIC DEBRIS

During pregnancy, a large number of multinucleated fragments of dying syncytiothrophoblasts are shed daily into the maternal circulation. These trophoblasts, shed by the placenta, are rapidly cleared from the circulation by endothelial cells during normal pregnancy in order to prevent clogging of the maternal pulmonary circulation. Indeed, failure to clear such fragments often results in pre-eclampsia. Endothelial cells can internalize dying trophoblasts regardless of whether they are apoptotic or necrotic. However, while engulfment of apoptotic trophoblasts does not induce endothelial cell activation, phagocytosis of necrotic trophoblasts causes endothelial activation and ICAM-I expression (177).

The organism also counts with acellular routes for degradation of necrotic debris. It has been shown that serum components such as the nuclease DNAse I, the complement protein C1q and the protease plasmin act in synergy to degrade chromatin even in the absence of leukocytes. For instance, the binding of C1q to necrotic chromatin strongly enhances the activity of DNAse I, even though C1q lacks nuclease activity (178). It was postulated that C1q was able to enhance the access of DNAse I to necrotic DNA, improving the degradation of debris. Moreover, plasminogen was shown to penetrate necrotic cells, where it was activated into plasmin (179). The proteolytic activity of plasmin caused the cleavage of histone H1, which in turn facilitated the cleavage of DNA by DNAse I. The synergy between these enzymes is required for the fast and effective breakdown of necrotic chromatin.

## "DON'T EAT-ME" SIGNALS

Effective engulfment of dead cells entails not only the exposure of "eat-me" determinants, but requires a reduction of surface "don't eat-me" signals. Eukaryotic cells display CD47, a surface protein that is recognized by SIRPα, expressed by myeloid cells (44). CD47 functions by directly binding SIRPα on macrophages and monocytes, signaling inhibition of phagocytosis that is partly due to impaired myosin assembly at the phagocytic synapse (180).

Aging and the subsequent elimination of erythrocytes by efferocytosis correlates with a decrease in their surface CD47 (181). The importance of CD47 as a "don't eat-me" signal was demonstrated by Kojima and colleagues, who showed that dysregulation of CD47 signaling contributes to the development or atherosclerotic plaques. In this setting, instead of downregulating CD47, dying cells upregulated it, making apoptotic cells resistant to phagocytic clearance and thereby driving plaque formation. Interestingly, administration of a blocking CD47 antibody reversed this effect, stimulating efferocytosis and reducing atherosclerosis, making CD47 a potential drug target for the clinic (182).

CD47 can alter also the phagocytosis of necrotic debris. One explanation why necrotic debris are engulfed at a slower rate than apoptotic cells is that they have, comparatively, an increased surface expression of CD47. Moreover, CD47 was found to be clustered on necrotic cells, and these clusters stimulated RhoApMLC signaling in macrophages that promoted "nibbling" of the necrotic cells, rather than whole-cell internalization (183). This process—commonly known as trogocytosis—is shared by amoeba, lymphocytes, neutrophils and macrophages, and polarization of CD47 might explain the preferential nibbling over whole-cell engulfment.

CD46 is a widely expressed complement regulatory protein. It inhibits complement by binding C3b and C4b and acting as a cofactor for their proteolytic cleavage (184). CD46 is a "don't eat-me" signal that is lost during apoptosis and necrosis. In both types of dying cells, CD46 is clustered and shed in microparticles alongside nucleic acids and PS (185). The loss of CD46 correlates with an increase in deposition of C1q and C3b on the dying cells. However, only necrotic cells proceed to form membrane attack complexes, because they also undergo significant reduction in the expression of the complement regulators CD55 and CD59. This study indicated that the dying cells specifically lose "don't eat-me" signals that block complement activation in healthy cells, allowing them to be opsonized by complement and engulfed (185).

Several "don't eat-me" signals that have been implicated in apoptosis have not yet been investigated in the context of necrosis yet. CD31 (also known as platelet-endothelial cell adhesion molecule 1, PECAM-1) is an important "don't eat-me" signal, acting as a repulsive signal through homotypic CD31- CD31 interactions between cells. The ligation of CD31 on viable leukocytes promotes cell detachment. Apoptotic cells that lack CD31 bind tightly to leukocytes and are subsequently engulfed (186). Plasminogen activator inhibitor (PAI)-1, is a member of the serpin family of serine protease inhibitors. It appears to co-localize with calreticulin on viable neutrophils, where it is thought to impair signaling to macrophages. This impairment is lost during neutrophil apoptosis, suggesting that PAI-1 is a "don't eat-me" signal (187). Also, the urokinase receptor (uPAR), which normally plays a role in fibrinolysis, cell migration and adhesion, was shown to modulate efferocytosis (188). Macrophages from uPAR-deficient mice demonstrated enhanced ability to engulf viable neutrophilsin vitro and in vivo. In line with this, expression of uPAR was reduced in apoptotic neutrophils compared to viable neutrophils, suggesting that uPAR is also a bona fide "don't eat-me" signal that is downregulated in apoptotic cells (189). Proteinase 3 (PR3) is a neutrophil granular protein that is co-externalized with PS during neutrophil apoptosis (190). PR3 impairs phagocytosis of apoptotic neutrophils by macrophages via inhibition of calreticulin function, a powerful "eat-me" signal. Another regulator is CD24, a heavily glycosylated GPI-anchored surface protein. It interacts with Siglec-10 in leukocytes to dampen inflammation in a variety of diseases (191). Recently, CD24 was described as a major "don't eat-me" signal exploited by tumor cells to evade the immune response (192). CD24 is overexpressed by a variety of tumor cells, inhibiting phagocytosis by neighboring tumor-associated macrophages, which express high levels of Siglec-10. Blockage of CD24 by a monoclonal antibody reduced tumor growth in vivo, suggesting that inhibition of this "don't eat-me" signal suffices to enable phagocytosis of live cancer cells.

## IMMUNE CONSEQUENCES OF DEFECTIVE DEBRIS CLEARANCE

The generation of necrotic debris is a severe occurrence; yet, it is immediately met by a barrage of fluid-phase proteins, mediators and cells, which cause it to be essentially uneventful. Tissue inflammation resolves in a timely manner and immune responses against self-develop very rarely. However, if the organism fails to contain or clear the necrotic debris appropriately, tissue inflammation is prolonged and autoimmunity can ensue. Mutations that impair the ability of leukocytes to recognize or eliminate debris have been connected to defects in tissue regeneration and to diseases such as systemic lupus erythematosus (SLE). Below, we highlight a series of studies describing the catastrophic consequences of tampering with the response to necrotic cell death.

Inhibition of phagocyte recruitment or function at necrotic sites results in a clear defect in recovery from injury. Depletion of neutrophils prevents the clearance of debris from necrotic sites, leading to an impairment of regeneration and revascularization of the focal injury (98). Moreover, inhibition of monocyte recruitment to necrotic foci, whether due to CCR2 deficiency or to interference with their transition into CX3CR1<sup>+</sup> cells delays the regeneration of the necrotic injury (99). Similar observations were made in complement-deficient models, reinforcing the notion that the removal of necrotic debris by phagocytes is paramount to tissue repair.

As described above, complement contributes "find-me" and "eat-me" signals to necrotic cells, and several studies have shown its major role in tissue regeneration (132, 140, 193–195). In the long term, defects in the complement cascade have been strongly associated to the development of SLE. Although a multi-factorial disease, SLE and related lupus-like syndromes are clearly connected to mutations or deficiency in C1q, C2, C3, and C4 complement factors (196). In addition, the disease has been associated to decreased expression of complement receptors CR1 and CR2 (133). Defects in complement activation, such as in the classical pathway, also yield organisms susceptible to autoimmunity, as is the case of IgM-deficient mice (197). Interestingly, a mutation of CD11b (ITGAM) has been correlated to the development of SLE as well (198). Phagocytes express high levels of CD11b, which is used as both complement receptor (CR3) and as adhesion molecule (Mac-1). Whether the polymorphism affects the phagocytic or adhesive functions of CD11b is still unclear, but the findings nevertheless provide further indication that impairment of the ability of phagocytes to clear debris causes immediate and long-term disadvantages to the host.

The exposure of extracellular DNA is a key factor in SLE development. The accumulation of DNA in tissues and bloodstream has to be rapidly counteracted by the activity of DNAses to minimize inflammation and autoimmunity. An abundant source of DNAse activity in the organism is serum, which contains two major nucleases, DNAse I and DNAse IL3 (199, 200). The two enzymes have nonredundant roles in DNA/chromatin degradation; DNAse I acts preferentially against internucleosomal "naked" DNA, whereas DNAse IL3 cleaves chromatin (protein-bound DNA) with high activity (201). DNAse I deficiency causes mice to develop anti-nuclear antibodies and SLE (200). DNAse IL3 deficiency is also sufficient to cause autoimmunity and SLE in mice (202, 203). In humans, mutations of DNAse I (204), DNAse IL3 (205) and DNAse III (TREX1) (206) have already been implicated in the incidence of SLE or lupus-like disease. Global deficiency in DNAse II, an isoform found in lysosomes, is embryonically lethal due to the accumulation of undigested DNA from red cell nuclei inside macrophages, which mount a type I interferon response that leads to embryonic demise (207). Interestingly, the lethality can be abrogated by simultaneously knocking out STING, a central adaptor of cytoplasmic DNA immunity (208), suggesting that leakage of undegraded debris from the phagosome to the cytosol fuels the deleterious response. The deficiency in DNAse II also correlates with worsening of heart failure (209) and development of polyarthritis in mice (210). It is clear that removal of necrotic debris is a multi-layered response: the phagocytes must be able not only to reach and ingest the debris, but also to effectively degrade it in order to avoid overt inflammation and autoimmunity.

Like DNA, actin is a very abundant component of cells and a DAMP conserved across species (211, 212). Necrotic cells expose actin after the plasma membrane integrity is breached, and the released actin is recognized by Clec9A (DNGR-1), a C-type lectin receptor expressed primarily in dendritic cells (211, 213). Clec9A specifically recognizes filamentous (F)-actin, which—remarkably—persists in necrotic cells even after their death. F-actin is able to bind Clec9A even when forming

complexes with actin-binding proteins such as spectrin, αactinin, and myosin II (213, 214). Importantly, the recognition of actin-rich necrotic debris by dendritic cell Clec9A prompted the cross presentation of self-antigens to CD8 T cells, a mechanism that explains how autoimmunity is initiated by the exposure of necrotic debris. Exposure of F-actin can be further regulated, as it is depolymerized by circulating DNase I. Full F-actin depolymerization requires ATP, which could be present as necrotic cells release ATP (described in section Purines) (215).

## CONCLUDING REMARKS

Consideration of necrotic cells as an important, ongoing contributor to overall cell death provides a different vantage point of how debris are sensed and cleared and their contribution to inflammation and autoimmunity. Clearly, inhibiting the host's ability to eliminate and process necrotic debris has harmful effects. Strikingly, therapies for acute inflammation are largely confined to the use of anti-inflammatory drugs. In the short term, this approach reduces tissue inflammation and the associated symptoms (swelling, pain), but it comes at the cost of delayed resolution of injury. Prevention of inflammation

#### REFERENCES


retards debris clearance, re-growth of parenchymal cells and tissue regeneration. Thus, an alternative approach would be to stimulate debris clearance in addition to minimizing the uncomfortable symptoms. This can be accomplished by the application of pro-resolving mediators such as resolvins, lipoxins, hydrogen sulfide, IL-10, and annexin A1, which can stimulate clearance mechanisms without the damaging effects of excessive inflammation (216, 217).

#### AUTHOR CONTRIBUTIONS

JW and PM prepared the figures. JW, SG, and PM wrote the manuscript.

#### FUNDING

JW was supported by The Swedish Society of Medicine and The Foundation Blanceflor Boncompagni Ludovisi, née Bildt. SG was supported by Canadian Institutes of Health Research grant FDN-143202. PM was supported by the Rega Foundation (Belgium), by Fonds Wetenschappelijk Onderzoek (FWO—Belgium), and by a Marie Sklodowska-Curie Individual Fellowship.

change in apoptosis. Science. (1997) 278:294–8. doi: 10.1126/science.278. 5336.294


modulate receptor binding and cellular responses. Biochemistry. (1999) 38:12959–68. doi: 10.1021/bi990711d


**Conflict of Interest:** 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 © 2020 Westman, Grinstein and Marques. 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.

# Immunological Outcomes Mediated Upon Binding of Heat Shock Proteins to Scavenger Receptors SCARF1 and LOX-1, and Endocytosis by Mononuclear Phagocytes

Ayesha Murshid<sup>1</sup> , Thiago J. Borges 1,2, Cristina Bonorino3,4, Benjamin J. Lang<sup>1</sup> and Stuart K. Calderwood<sup>1</sup> \*

<sup>1</sup> Department of Radiation Oncology, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States, <sup>2</sup> Renal Division, Schuster Family Transplantation Research Center, Harvard Medical School and Brigham and Women's Hospital, Boston, MA, United States, <sup>3</sup> Laboratório de Immunoterapia, Departmento de Ciências Básicas da Saúde, Universidade Federal de Ciências da Saúde de Porto Alegre, Porto Alegre, Brazil, <sup>4</sup> Department of Surgery, School of Medicine, University of California, San Diego, San Diego, CA, United States

#### Edited by:

Carlos Rosales, National Autonomous University of Mexico, Mexico

#### Reviewed by:

Xiang-Yang Wang, Virginia Commonwealth University, United States Chunqing Guo, Virginia Commonwealth University, United States

> \*Correspondence: Stuart K. Calderwood scalderw@bidmc.harvard.edu

#### Specialty section:

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

Received: 07 November 2019 Accepted: 11 December 2019 Published: 10 January 2020

#### Citation:

Murshid A, Borges TJ, Bonorino C, Lang BJ and Calderwood SK (2020) Immunological Outcomes Mediated Upon Binding of Heat Shock Proteins to Scavenger Receptors SCARF1 and LOX-1, and Endocytosis by Mononuclear Phagocytes. Front. Immunol. 10:3035. doi: 10.3389/fimmu.2019.03035 Heat shock proteins (HSP) are a highly abundant class of molecular chaperones that can be released into the extracellular milieu and influence the immune response. HSP release can occur when cells undergo necrosis and exude their contents. However, HSPs are also secreted from intact cells, either in free form or in lipid vesicles including exosomes to react with receptors on adjacent cells. Target cells are able recognize extracellular HSPs through cell surface receptors. These include scavenger receptors (SR) such as class E member oxidized low-density lipoprotein receptor-1 (LOX-1, aka OLR1, Clec8A, and SR-E1) and scavenger receptor class F member 1 (SCARF1, aka SREC1). Both receptors are expressed by dendritic cells (DC) and macrophages. These receptors can bind HSPs coupled to client binding proteins and deliver the chaperone substrate to the pathways of antigen processing in cells. SR are able to facilitate the delivery of client proteins to the proteasome, leading to antigen processing and presentation, and stimulation of adaptive immunity. HSPs may also may be involved in innate immunity through activation of inflammatory signaling pathways in a mechanism dependent on SR and toll-like receptor 4 (TLR4) on DC and macrophages. We will discuss the pathways by which HSPs can facilitate uptake of protein antigens and the receptors that regulate the ensuing immune response.

Keywords: heat shock proteins, scavenger receptor, immunity, macrophage, dendritic

# INTRODUCTION

Heat shock proteins (HSPs) are the major components of a primordial cellular responses to proteotoxic stress, and the resultant production of many HSP species is collectively described as the heat shock response (HSR) (1) (**Table 1**). Classic activators of the HSR such as heat shock, lead to rapid denaturation of intracellular proteins resulting in dysfunctional intermediate

TABLE 1 | Mammalian heat shock proteins.


conformational states that are prone to aggregation. Cell survival necessitates an almost immediate and abundant induction of the HSPs that halt the aggregation cascade and permit refolding of cellular proteins and restoration of normal protein function (18). Beyond protein refolding, HSPs also function to facilitate trafficking of their client substrates between subcellular compartments and mediate protein-protein interactions. They are hence referred to as molecular chaperones (18, 19). The HSP family contains a number of members that belong to different protein families, but each has in common a role in stepwise protein folding (**Table 1**) (2, 4, 5). However, in addition to their various intracellular functions, HSPs have also been detected in the extracellular spaces and circulation. This phenomenon may be the result of the death of stressed cells and release of the highly amplified HSP cohort (6). Although HSPs lack the leader sequences required for the classical secretion pathways (20) they may be released from viable cells by non-canonical secretion pathways (21). Alternative mechanisms of HSP secretion have been described. These pathways include the secretion of free Hsp70 by a pathway similar to that utilized for non-canonical secretion of interleukin-1 beta, involving secretory lysosomes, or alternatively HSPs may be packaged in lipid vesicles known as exosomes and released to the extracellular milieu of tumors and normal tissues (7, 21–23). The extracellular HSPs may exert a number of effects on adjacent cells after their secretion, such as activation of antigen-presenting cells (APCs), including monocytes, dendritic cells (DC) and macrophages, as well as causing increased mobility and metastasis in target cells as has been observed in wound healing and cancer scenarios (3, 24–26). Hsp70 and Hsp110 have been utilized effectively in anticancer vaccines, in which they function as carriers of antigenic peptides that can be efficiently taken up and processed by APCs and presented to T lymphocytes (8, 9, 16, 17, 27– 29). Understanding how HSPs are bound by acceptor cells and taken up is therefore important in determining the properties and function of extracellular HSPs.

#### AN OVERVIEW OF HSP RECEPTORS

Most of the biological effects of extracellular HSPs identified to date have involved their binding to surface receptors on target cells prior to their internalization (10, 30). However, the entire spectrum of dedicated high affinity receptors for the HSPs have not been identified in studies carried out so far. The first protein to be identified as an HSP receptor was CD91/alpha2 macroglobulin receptor, which is a low density lipoprotein (LDL) binding protein currently known to be a highly versatile receptor for over 30 other ligands (31). This multi-subunit protein appears to be a common receptor for most of the HSPs involved in immune responses. There was some controversy originally regarding the significance of this finding, as CD91 was suggested to be the receptor involved in antigen cross-presentation by DC in response to HSP vaccines, although most types of DC do not appear to express endogenous CD91 (11, 30). However, CD91 has since been shown to be a receptor for Hsp90α in wound healing and cancer metastasis scenarios and signaling pathways downstream from the receptor appear to mediate effects of the chaperone on cell motility, a key property in wound healing and metastasis (32). The class E and F scavenger receptors LOX-1 and SCARF1 are the major receptors for HSP-peptide complexes, mediating antigen uptake and processing (10–12, 33) (**Figure 1**). The scavenger receptors, although not structurally related, share common functions including the binding, endocytosis and thus detoxification of oxidized LDL by vascular endothelial cells (33, 34). They are key players in the removal of oxidized LDL from the circulation and protection from the morbidity associated with atherosclerosis (35). Both LOX-1 and SREC1 are also expressed on DC and macrophages and play key roles in antigen crosspresentation mediated by HSP- peptide complexes (HSP-PC) (11, 36). In this review, we will describe the roles of these SR members in mediating extracellular HSPs-triggered responses, focusing mainly in their interaction with the Hsp90α. A really puzzling feature of this system is that most SR members are not structurally related but bind to a common ligand, while HSPs of different chaperone families often bind to the same scavenger receptor species, although also lacking structural relationship (12, 33) (**Figure 1**, **Table 1**).

It is not clear which property of HSPs prompts their binding to scavenger receptors. However, in addition to binding to oxidized LDL, members of the scavenger receptor family can bind to proteins with other modifications (acetylated LDL) as well as to polyanionic ligands such as poly-IC, findings which may cast some light on interactions with HSPs (11, 33). HSPs have been shown to be phosphorylated and acetylated, modifications that would increase their net negative charge (37, 38). Future studies would be required to clarify this issue. When LDL particles are oxidized, they assume a net negative charge and additionally phospholipid moieties are added to the LDL particle protein apolipoprotein B100 (39–41). These phospholipid residues fit into a hydrophobic tunnel formed by surface LOX-1 dimers (42). HSP binding to scavenger receptors may therefore involve the ability of the chaperones to recognize hydrophobic patches on client proteins as well as the charge interactions mentioned above (5).

#### HSP RECEPTORS AND DETOXIFICATION OF HSP-PEPTIDE COMPLEXES AND DEAD CELLS

The primary role of the scavenger receptor family seems to be removal of oxidized LDL from the circulation (35). It is

also possible that this role may be recapitulated in interactions with HSP-PC, a complex which may flood the local circulation after tissue trauma, a result of the large scale cell death by necrosis that may ensue. Such complexes may be able to prime immune and inflammatory responses in damaged tissues and it may be incumbent on mononuclear phagocytes to rapidly bind and endocytose such structures using the scavenger receptors (43, 44). It is known that LOX-1 and SCARF1 can also bind cell corpses and remove them from the extracellular spaces (45, 46). Uptake of HSP-PC may thus be part of a general detoxification process exerted by scavenger receptors, operating in damaged tissues. The scavenger receptors could also be involved in uptake and removal of HSP-containing exosomes given their abilities to bind lipid structures such as oxidized LDL and cell corpses (46). SCARF1 is a paralog of the cell corpse receptor CED-1 expressed in C. elegans (35, 47). In addition, more closely related paralogs of CED-1 have been unearthed and could be putative HSP receptors. These include Drosophila gene draper and the mammalian MEGF10, MEGF11, and MEGF12 (48– 51). Each of these proteins contains multiple EGF-like motifs in the extracellular domain that may be recognition sequences for apoptotic bodies and play roles in dead cell clearance (**Figure 1**). Another protein with multiple EGF-like motifs in its extracellular domain that can bind to HSPs and apoptotic cell corpses is the Class H scavenger receptor FEEL-1/stabilin-1 (30, 33, 52) (**Figure 1**). Its role in responses to extracellular HSPs is currently unclear.

# PATHWAYS OF SCAVENGER RECEPTOR-MEDIATED ENDOCYTOSIS

The properties of the SR as endocytic receptors with a wide range of selectivity makes them effective intermediaries in sampling the local extracellular milieu of APC for potentially antigenic molecules. Thus, both LOX-1 and SCARF1 are expressed in DC and other mononuclear phagocytes (11, 36). There are a number of pathways by which extracellular molecules can enter cells. These include endocytosis, a process which involves the association of molecules with cell surface invaginations, uptake in an actin-dependent manner, and then fusion of the engulfed vesicles with intracellular endosomes. The major canonical pathway is clathrin-mediated endocytosis, a process that involves pit-like structures inserted into the plasma membrane which are lined with clathrin, a trimeric protein that stabilizes the pits (53). Molecules, sometimes associated with receptors, are then engulfed in clathrin coated vesicles that are found in the majority of cells. There is a second, less prevalent pathway, involving the protein caveolin found in structures known as caveolae, 50 nm invaginations that can also mediate endocytosis of extracellular molecules (54). However, both LOX-1 and SCARF1 have been shown to take up their ligands in a clathrin and dynaminindependent manner, utilizing a more unconventional endocytic pathway (36, 55). The mechanisms involved in endocytosis mediated through LOX-1 seem to be currently unclear although more information has accumulated regarding SCARF1. Upon

ligand binding SCARF1 is internalized by DC via the GPI-AP (glycophosphatidylinositol-anchored proteins) enriched early endosome (GEEC) pathway (**Figure 2**) (56, 57). This pathway is mediated by uncoated tubular vesicular structures called clathrin independent carriers (CLICs) that mature into the early endocytic compartment (GEECs) (58, 59). The pathway is specialized for uptake of GPI-AP such as the folate receptor. Thus, uptake of Hsp90α- peptide complexes was not inhibited by antagonists of clathrin- and caveolin-dependent endocytosis, characteristic of the GEEC pathway (36). Endocytosis of Hsp90αpeptide complexes was however inhibited by blocking the activity of Rho GTPase CDC42, a protein shown to be involved in actin polymerization and uptake of GPI-AP through the GEEC pathway. SCARF1 became co-localized, after binding to Hsp90αpeptide complexes, with CD59, a marker GPI-AP protein that utilizes the GEEC pathway (36, 60). Proteins internalized through the GEEC pathway, such as GPI-AP are frequently associated within plasma membrane microdomains such as lipid rafts (61). These are regions of the membrane enriched in cholesterol and glycosphingolipids that are immiscible with the bulk membrane and appear to diffuse freely through this membrane (62, 63). SCARF1 is not a GPI-AP protein even though it has been shown to enter the GEEC pathway. However, another protein modification that may target transmembrane proteins such as SCARF1 to lipid rafts is S-acylation of cysteine residues close to the transmembrane domain with saturated palmitate residues capable of dissolving in the cholesterol and glycosphingolipid milieu that comprises the partitioned microdomains. SCARF1 contains five cysteine residues (Cys - 440, 441, 443, 444, 445) adjacent to the transmembrane domain (amino acids 422- 442) (35, 62, 63). Thus, cysteine palmitoylation, and perhaps interaction with other proteins in the lipid rafts, may potentially recruit SCARF1 to this region. The nature and extent of partner proteins associated with SCARF1 in the rafts is not clear, although the receptor was shown to interact with the non-receptor tyrosine kinase c-Src (36). Although c-Src is likewise not a member of the GPI-AP family, it also associates with the rafts after S-acylation (63). Inhibition of c-Src activity prevented the cross-presentation of antigens associated with Hsp90α suggesting a key role for signaling through this tyrosine kinase in the antigen presentation pathways (36). Phosphorylation of key tyrosine residues within internalization motifs in the intracellular domain regulates the endocytosis of many receptors, although consensus sequences for internalization such as the NPXY motif found in CED-1 are not observed in the SCARF1 sequence (**Figure 1**) (35). The mechanism of regulation of SCARF1 endocytosis by c-Src thus remains to be defined. LOX-1 function has also been linked to its entry into lipid rafts and cholesterol lowering drugs inhibit its function (64).

#### SCAVENGER RECEPTOR-MEDIATED ACCESS TO ANTIGEN PROCESSING PATHWAYS

Binding of antigenic polypeptides to HSPs in DC allows them to enter the pathway of antigen cross presentation and be processed

in the cytoplasm and presented on major histocompatibility Class I (MHC I) proteins (13). Most of the antigens presented on MHC Class I proteins are derived from proteolytic processing of intracellular proteins via the classical Class I pathway (65). However, DC are specialized to take up extracellular antigens using receptors such as Fc, CLEC9A, DC-SIGN, DEC205 and mannose receptor, and thus funnel them into the Class I pathway, permitting surveillance of the extracellular spaces (66–70). HSPs can also bind external antigens and funnel them into the Class I pathway through LOX-1 and SCARF1 (36). For instance, following transit through the GEEC compartment, Hsp90αpeptide-SCARF1 complexes are translocated to early endosomes (36). In the case of full-length chaperoned proteins such as intact ovalbumin, antigen processing is carried out after its digestion in the proteasome in the cytoplasm and then antigenic peptides are taken up into MHC class I molecules by TAP (transporter associated with antigen processing) in the ER (36). This latter process clearly requires the chaperoned protein to escape the confines of the early endosome and enter the cytoplasm in order to be taken up by the proteasome. Hsp90α is known to facilitate this step as well as to maintain the client protein intact until it reaches the proteasome (71). Hsp90α-peptide complexes internalized in association with SCARF1 can also be processed within the endosome and antigens loaded onto MHC class II molecules prior to recycling to the plasma membrane (14). This is an essential step in efficient activation of T cell immunity and the activation of CD4+ T helper cells. Normally uptake of soluble antigens into the Class I and Class II pathways involves separate

receptors (13). This scavenger receptor can therefore facilitate antigen processing by APC through the two major antigen presentation pathways and appears to play an integral role in the functioning of HSP based immunity (13). LOX-1 also participates in processing of HSP bound tumor antigens and approximately 50% of ovalbumin (sample antigen) cross-presentation appeared to be mediated through this receptor (36). Although HSP-PC binding to SCARF1 can facilitate antigen presentation through the MHC class I and class I pathways it is not clear whether this interaction can induce co-receptor induction in APC- a crucial step in adaptive immunity. For instance, creation of efficient anticancer vaccines employing large HSP family members has employed fusion of the HSPs to prokaryotic danger signals to boost inflammatory signaling and co-receptor induction (9). Interestingly, SCARF1 expression and Hsp70 vaccine anti-tumor activity is dependent on TLR2 and TLR4 function suggesting upregulation in inflammatory conditions (28).

# LIPID RAFTS AND CELL SIGNALING AFTER HSP BINDING TO CELL SURFACE RECEPTORS

In addition to the import of tumor antigens by DC, HSPs may carry out key cell signaling roles within the lipid rafts of mononuclear phagocytes. The association of proteins with lipid rafts may permit them to concentrate at foci within the plasma membrane. This property depends on the ability of the rafts to diffuse within the bulk membrane and thus potentially bring together cooperating signaling molecules (62, 63, 72). As mentioned above, an example of this process is the association of SCARF1 with c-Src after Hsp90α binding, an interaction that may promote endocytosis and phagocytosis through activation of the kinase, recruitment of Cdc42 and association with the actin cytoskeleton (36) (**Figure 3**). SCARF1 entry into c-Src containing lipid rafts was also required for inflammatory cytokine release in mouse macrophages (73). This process involved association of SCARF1 with the pro-inflammatory Toll Like Receptor 4 (TLR4) after exposure to bacterial lipopolysaccharides (15). The association with SCARF1 in lipid rafts led to downstream signaling through TLR4, activation of the c-jun kinase, p38 MAP kinase and NF-kB signaling pathways and upregulation of interleukin 6 synthesis (73) (**Figure 3**). These inflammatory signaling processes required cholesterol, actin polymerization and CDC42 activity. SCARF1 and LOX-1 may also be able to recruit other signaling molecules and exposure to outer membrane protein A (OmpA) from Klebsiella pneumoniae led to recruitment of TLR2 and cytokine synthesis by the scavenger receptors (74). SCARF1 also cooperates with TLR2 in recognition of hepatitis virus non-structural protein by DC (75). In a similar vein, SCARF1 was shown to associate with TLR3 after exposure of macrophages to double stranded RNA and stimulate signaling through the NF-kB, MAP kinase and the IRF3 pathways (76). SCARF1 and LOX-1 may therefore play key roles in associating with cell signaling molecules and creating activating foci through the concentration of lipid rafts after binding eukaryotic or prokaryotic ligands (73, 76–78).

# SCARF1, LOX-1, AND INFLAMMATION

While it is clear that SCARF1 and LOX-1 can mediate immunity by binding HSP-associated antigens and promoting antigen cross-presentation, the effects of HSPs on inflammation are less clear (79). Discrepancies in the field were originally ascribed to the use by some investigators of purified HSPs associated with bacterial PAMPs: indeed, HSP and endotoxins undergo complex interactions that mediate inflammatory responses (80). However, Hsp70—TLR4 association and subsequent inflammatory signaling is regularly observed in vivo and under conditions in which endotoxin contamination of the chaperones seems unlikely [reviewed in (81)]. Nonetheless, in the case of purified Hsp90α, it was shown that while both this chaperone and LPS could bind to SCARF1 and lead the receptor to enter a lipid raft compartment, only exposure to LPS led to significant levels of pro-inflammatory signaling through this mechanism; Hsp90α alone, although entering the lipid raft compartment did not trigger inflammatory signaling (73). In addition, it has been shown that some prokaryotic HSP paralogs tend to be antiinflammatory in contrast to the mammalian isoforms (82). The answer to this conundrum appears to be at least partially that another key class of HSP receptors is expressed on mononuclear phagocytes—the sialic acid-binding immunoglobulin-type lectins (Siglecs) (83, 84). The Siglec family of receptors bind to selfsialic acid residues either in cis or in trans and this interaction leads to suppression of inflammatory responses in mononuclear phagocytes (83). Upon binding to the antigen, the intracellular regions of the Siglecs become activated by phosphorylation of immunosuppressive ITIM (immunoreceptor tyrosine-based inhibition motif) domains that associate with phosphatases Shp-1 and Shp-2, leading to immune suppression (85). Hsp70 has been shown to bind to Siglecs after tissue damage suggesting a mechanisms for the immunosuppressive roles of the chaperone (86). A further complication to this scenario is that human cells can express a receptor pair SIGLEC-5 and SIGLEC-14 that can contain similar ligand binding domains and either ITIM or ITAM (immunoreceptor tyrosine-based activation motif) sequences and thus, dependent on context, are either immunosuppressive (SIGLEC-5) or immunostimulatory (SIGLEC-14) (87). Clearly further studies are essential to clarify the nature of the signaling complexes on mononuclear phagocytes that determine response to HSPs in terms of both endocytosis and inflammatory cell signaling.

# CONCLUSIONS

(a) The scavenger receptors LOX-1 and SCARF1 mediate binding and endocytosis of HSPs such as Hsp70, Hsp90, and Hsp110. The HSPs are taken up by a clathrin-independent mechanism involving the GEEC pathway. At least in the case of SCARF1, endocytosis requires the activity of the c-Src kinase which can bind to the receptor in lipid raft microdomains.


Three HSP binding receptors SCARF1/SREC-I, FEEL-1, and LOX-1 are shown as well as related proteins. Locations of atypical EGF-like domains are indicated in—CED-1, hSCARF1, hFEEL-I/Stabilin-1 and MEGF10. Each share EGF-like consensus repeats in the extracellular domains. Tyrosine-based sorting signals are known to interact with the phospho-tyrosine domain of clathrin adaptors (NPXY for CED-1, FXNPXY and YXXØ for hMEGF10) are shown in the figure. SCARF1 does not contain these motifs and is not internalized through clathrin-mediated endocytosis. FEEL-1 is expressed mainly in intracellular compartments. A dileucine based (DXXLL for hFEEL-1) sorting signal is present in the cytosolic tails of hFEEL-1 and can also be found in mannose 6-phosphate receptors that mediate sorting between trans-Golgi network (TGN) and endosomes. LOX-1, although sharing many properties with SCARF1, including HSP binding and internalization, does not contain EGF-like motifs in its extracellular domain extracellular domain and belongs to the C-type lectin family.

Under resting conditions, SCARF1 is shown in the bulk membrane domain containing a range of surface proteins which are either transmembrane proteins such as SCARF1, GPI-AP proteins or proteins anchored to the inside of the membrane such as c-Src. Upon Hsp90α binding, SCARF1 becomes localized into lipid raft domains and co-localized with c-Src. Within 5 min of ligand binding, Hsp90α–SCARF1 complexes enter the GEEC compartment and are internalized (36). We also show proteins that remain in the bulk membrane and are not internalized through the GEEC pathway.

We show ligand (HSP) binding by SCARF1 leading to its recruitment to lipid raft microdomains in the plasma membrane. SCARF1 then coordinates interaction of c-Src, CDC42 and TLR4 and signaling through the NF-kB and MAP kinase pathways upstream of inflammatory cytokine expression.

# 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 NIH grants RO1CA119045 and RO1CA047407.

# REFERENCES


#### ACKNOWLEDGMENTS

We thank the Department of Radiation Oncology, Beth Israel Deaconess Medical Center for the encouragement and support.


by cellular and humoral innate immunity receptors. Immunity. (2005) 22:551– 60. doi: 10.1016/j.immuni.2005.03.008


**Conflict of Interest:** 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 © 2020 Murshid, Borges, Bonorino, Lang and Calderwood. 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 Receptor-Mediated Phagocytosis Induces Proinflammatory Cytokine Production in Murine Macrophages

Durga Acharya1†‡, Xiao Rui (Lisa) Li 1†‡, Rebecca Emily-Sue Heineman2‡ and Rene E. Harrison<sup>1</sup> \* †

*<sup>1</sup> University of Toronto Scarborough, Toronto, ON, Canada, <sup>2</sup> Department of Biological Sciences, University of Toronto Scarborough, Toronto, ON, Canada*

#### Edited by:

*Carlos Rosales, National Autonomous University of Mexico, Mexico*

#### Reviewed by:

*Zsuzsa Szondy, University of Debrecen, Hungary Luisa Martinez-Pomares, University of Nottingham, United Kingdom*

\*Correspondence:

*Rene E. Harrison harrison@utsc.utoronto.ca*

#### †Present address:

*Durga Acharya, Xiao Rui (Lisa) Li, and Rene E. Harrison, Department of Cell & Systems Biology and the Department of Biological Sciences, University of Toronto Scarborough, Toronto, ON, Canada*

> *‡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 September 2019* Accepted: *12 December 2019* Published: *14 January 2020*

#### Citation:

*Acharya D, Li XR, Heineman RE-S and Harrison RE (2020) Complement Receptor-Mediated Phagocytosis Induces Proinflammatory Cytokine Production in Murine Macrophages. Front. Immunol. 10:3049. doi: 10.3389/fimmu.2019.03049* Macrophages are professional phagocytes that are uniquely situated between the innate and adaptive arms of immunity with a high capacity for phagocytosis and proinflammatory cytokine production as well as antigen presentation. Phagocytosis is a critical process to eliminate microbes, apoptotic cells and other foreign particles and is accelerated by host-generated opsonins, such as antibodies and complement. Early phagocytosis studies established the paradigm that FcγR-mediated phagocytosis was more proinflammatory than Complement Receptor (CR)-mediated uptake in macrophages. Using qPCR, cytokine antibody arrays and ELISA, we revisited this research question in primary macrophages. Using qPCR we determined that CR-mediated phagocytosis increases levels of TNF-α, IL-1β, IL-6, and MMP-9, compared to FcγR-mediated phagocytosis and control unstimulated cells. We confirmed these findings at the protein level using cytokine antibody arrays and ELISAs. We next investigated the mechanism behind upregulated cytokine production during CR-mediated phagocytosis. IκBα protein levels were reduced after phagocytosis of both IgG- and C3bi-sRBCs indicating proteolytic degradation and implicating NF-κB activation. Inhibition of NF-κB activation impacted IL-6 production during phagocytosis in macrophages. Due to the roles of calpain in IκBα and integrin degradation, we hypothesized that CR-mediated phagocytosis may utilize calpain for proinflammatory mediator enhancement. Using qPCR and cytokine antibody array analysis, we saw significant reduction of cytokine expression during CR-mediated phagocytosis following the addition of the calpain inhibitor, PD150606, compared to untreated cells. These results suggest that the upregulation of proinflammatory mediators during CR-mediated phagocytosis is potentially dependent upon calpain-mediated activation of NF-κB.

Keywords: macrophage, phagocytosis, cytokine, inflammation, complement, Fc receptor

# INTRODUCTION

Macrophages, meaning large phagocytes, are a heterogeneous population of cells that are important for maintaining homeostasis, surveillance and killing of pathogens through phagocytosis (1). Phagocytosis is an evolutionarily conserved defense mechanism by which macrophages capture and kill pathogens and remove apoptotic cells into specialized intracellular compartments. Phagocytosis

**86**

is mediated by scavenger receptors, Fcγ Receptors (FcγRs), and Complement Receptors (CRs) (2). FcγRs recognize immunoglobin G (IgG) that flags target pathogens and mediates their recognition by immune cells (3, 4). In mice, the classes of Fcγ receptors includes the activating receptors FcγRI, FcγRIII, and FcγIV and the inhibitory FcγRIIB receptor (5). The intracellular domains of Fcγ receptors contain the immunoreceptor tyrosine-based activation motif (ITAM) that activates downstream signaling cascades (6, 7). The ligation of IgG-opsonized particles to FcγRs leads to receptor crosslinking which activates Src tyrosine kinase resulting in tyrosine phosphorylation of the ITAM motif (8). ITAM phosphorylation is important for the recruitment and activation of Syk (9). Activation of Src and Syk facilitates the activation of downstream Cdc42, RhoA, and Rac1 that control F-actin dynamics and contribute to the formation of pseudopods around the particle (2, 10, 11). Alternatively, the inhibitory FcγRIIB activates signaling cascades by recruiting inositol polyphosphate-5 phosphatase SHIP1 to counteract signaling from activating FcγRs (7).

The other major opsonin is the complement fragment, C3bi, which binds to targets directly or through IgM, and is recognized by CRs on phagocytes (2). The complement system contains 30 soluble and membrane bound proteins and the opsonin, C3bi, is generated from a series of biochemical reactions (12). The complement system is activated by pattern recognition receptors that have evolved to recognize antibodies, mannan-binding lectin, ficolins, C-reactive protein, C1q, and IgM (13). There are three pathways by which target molecules are opsonized by C3bi; the classical pathway, the lectin pathway and the alternative pathway. Each pathway leads to the formation of C3b/C3bi despite the differences in the initiation of the biochemical cascade. The classical pathway is activated upon the binding of C1q directly to pathogen surface or to immune complexes (12). The lectin pathway is activated by the recognition of pathogen molecular patterns by lectin receptors. Lastly, the alternative pathway is activated by the direct reaction of the C3 molecule with carbohydrate, lipid, and/or protein motifs found on the pathogen surface (14). Ultimately, each pathway generates the C3 convertase, C4bC2a for the lectin pathway, or C3bBb, for the alternative pathway. The C3 convertase leads to the production of C3bi, an opsonin (15). Receptors for the C3b protein fragments include CR1 and CRIg. While receptors for C3bi include CR2, CR3 (also known as Mac-1, CD11b/CD18, or αmβ2), and CR4 which belong to the β2 integrin family (14). CR3 requires an inside-out activation signal to effectively bind and internalize C3bi-coated particles (16). CR-mediated phagocytosis does not rely on Syk like FcγR-mediated phagocytosis but does require RhoA (17, 18). Unlike FcγR-mediated phagocytosis that involves pseudopods, CR-mediated phagocytosis leads to the formation of membrane ruffles that capture C3bi-opsonized target particles (19).

At the macroscopic level, inflammation is characterized by the presence of redness, swelling, heat, pain, and loss of tissue function (20). At the cellular level, inflammation is orchestrated with the help of histamines, cytokines, chemokines, and prostaglandins that are secreted by immune cells. A long-standing belief in the scientific community is that phagocytosis mediated by FcγRs is more potent at inducing inflammation than CR-mediated phagocytosis in macrophages. This is based on early studies in primary murine macrophages, a FcγR knockout mouse model and human monocyte-derived macrophages. The original murine macrophage study investigated whether phagocytosis induced production of arachidonic acid, a potent inflammatory mediator of vascular permeability (21). They concluded that phagocytosis via Fc receptors, but not the CR receptors, induces arachidonic acid production in murine resident peritoneal macrophages (21). Examination of human monocyte-derived macrophages revealed significant upregulation of arachidonic acid and H2O<sup>2</sup> after uptake of IgG-opsonized, but not C3biopsonized particles (22). Together these studies established the paradigm that phagocytosis of IgG-opsonized targets was proinflammatory, while C3bi-opsonized targets were non-stimulatory to macrophages.

Our lab was interested in understanding how and why phagocytosis, an effector response, would enhance inflammation. We assessed levels of inflammatory cytokines produced by murine macrophages challenged with either IgG-opsonized or C3bi-opsonized particles. Interestingly, we observed significantly more proinflammatory cytokine production, at the mRNA and protein level, after CR-mediated phagocytosis vs. FcγRmediated phagocytosis. To examine the mechanism behind proinflammatory cytokine production during phagocytosis, we examined signaling within the AP-1 and NF-κB gene expression pathways after particle engulfment. We show that CR-mediated phagocytosis activates NF-κB and provide evidence for a role for calpain in mediating proinflammatory cytokine production in macrophages. Together these studies address the current paradigm in the field and provide mechanistic insight into proinflammatory cytokine production in macrophages during phagocytosis.

# MATERIALS AND METHODS

### Reagents and Antibodies

Dulbecco's modified Eagle's medium (DMEM) and fetal bovine serum (FBS) and fetal calf serum (FCS) were obtained from Wisent Inc. (Saint-Jean-Baptiste, QC). Phosphate buffered saline (PBS) was purchased from ThermoFisher Scientific Inc. (Waltham, MA). Murine macrophage-colony stimulating factor (M-CSF) and granulocyte macrophage-colony simulating factor (GM-CSF) were purchased from PeproTech Inc. (Rocky Hill, NJ). Piceatannol was from Calbiochem Novabiochem Corp. (San Diego, CA).

Sheep red blood cells (sRBCs) (10% suspension) were purchased from MP Biomedicals LLC (Irvine, CA). 3.87µm polystyrene divinylbenzene beads (LBs) were purchased from Bangs Laboratories Inc. (Fishers, IN). Rabbit polyclonal anti-IκBα, rabbit polyclonal anti-NF-κB, rabbit polyclonal antip44/42 MAPK, rabbit polyclonal anti-phospho-p44/p42 MAPK antibodies, and protease/phosphatase cocktail inhibitor were all purchased from Cell Signaling Technology (Danvers, MA). Horseradish peroxidase (HRP)-conjugated secondary mouse and rabbit antibodies, AffiniPure donkey polyclonal anti-rabbit and donkey anti-human cyanine CyTM 5 and AffiniPure donkey polyclonal anti-rabbit cyanine CyTM 3 secondary antibodies were purchased from Jackson Immunoresearch Laboratories Inc. (West Grove, PA). Alexa Fluor <sup>R</sup> rhodamine phalloidin was purchased from Thermo Fisher Scientific. Complement 5 (C5) deficient human serum, Phorbol 12-Myristate 13-Acetate (PMA), Lipopolysaccharide (LPS) from Salmonella enterica serovar typhimurium, interferon-γ (IFN-γ), IgG and IgM opsonins, JSH-23, and Bortezomib were all purchased from Sigma Aldrich Canada Co. (Oakville, ON).

#### Primary Macrophage Isolation and Cell Culture

Primary bone marrow-derived macrophages (BMDMs) were obtained and differentiated from the femur and tibia of 4–8 weeks old C57BL/6 female mice (Charles River Laboratories, Saint Constant, QC). Isolated cells were washed with PBS and resuspended in DMEM containing 10% heat-inactivated FBS, supplemented with 100 IU/mL of penicillin, 100µg/mL streptomycin, and 20 ng/mL of M-CSF. The cells were then transferred to a 37◦C incubator at 5% CO<sup>2</sup> for 24 h. After the incubation, the floating cells were collected by centrifugation and the adherent cells were discarded. The cells were resuspended in DMEM containing 10% heat inactivated FBS, supplemented with 100 IU/mL of penicillin, 100µg/mL streptomycin and 100 ng/mL of M-CSF, plated onto T75 flasks and incubated at 37◦C with 5% CO2. The progenitor cells were then allowed to differentiate into BMDMs for 5 days. Adherent cells were removed from tissue culture flasks through gentle scraping. For experiments, 7.5 × 10<sup>5</sup> cells/well were plated into 6-well dishes or 2.5 × 10<sup>5</sup> cells/well were plated into 24-wells after which the BMDMs were grown to 70–90% confluency for 1 day prior to phagocytosis assays. Peritoneal macrophages were attained by lavage of the peritoneal cavity with 5 ml of ice cold 3% FCS in PBS. For each experiment, peritoneal macrophages were extracted and pooled from 2 mice. Cells were resuspended in DMEM containing 10% heat inactivated FBS and 1.5 × 10<sup>6</sup> cells were plated into a 6 well and allowed to adhere for 2 h. Cells were washed twice with warmed PBS to remove non-adherent cells and serum starved for 1 h prior to PMA activation and phagocytosis. Animal studies were conducted under protocols approved by the University of Toronto Local Animal Care Committee.

#### Opsonization and Phagocytosis Assays

Sheep RBCs or 3.87µm LBs were used as target particles for macrophages in phagocytosis assays. BMDMs were incubated in serum-free DMEM at 37◦C at 5% CO<sup>2</sup> for 1.5 h before the phagocytosis assays. Prior to opsonization, 100 µL of sRBCs (10% suspension) were washed with 500 µL of PBS and then opsonized with 2 mg/mL of anti-sheep rabbit IgG or 0.06 mg/mL of anti-sheep rabbit IgM for 1 h on a rotator at RT. The LBs were opsonized with 2 mg/mL of human IgG or human IgM. After opsonization, IgM-opsonized sRBCs/LBs were washed with 500 µL of PBS supplemented with 0.5 mM CaCl<sup>2</sup> and 0.5 mM MgCl2. C5-deficient human serum (50 µL) was then added to the sRBCs/LBs and incubated in a water bath at 37◦C for 30 min. The C3bi/IgM-opsonized sRBCs/LBs were then washed twice with 500 µL of supplemented PBS before being used in phagocytosis assays. Forty to Sixty microliter of opsonized particles were added to macrophages to induce an average uptake of 1–2 particles per macrophages. CR was activated in macrophages with either 100 nM PMA (19) in serum-free DMEM for 7 min or 300 U/ml GM-CSF for 1 h (23, 24) prior to the addition of C3bi-sRBCs/LBs.

## mRNA Isolation and qPCR

For qPCR assays, macrophages were challenged with either C3bior IgG-opsonized sRBCs or LBs for 3 h. Unbound sRBCs or LBs were removed by washing with cold PBS 3 times. mRNA was extracted using the Invitrogen PureLink RNA Mini Kit (Thermo Fisher Scientific). A total of 1 µg of RNA was used for cDNA synthesis (Invitrogen SuperScript III First-Strand Synthesis SuperMix) after DNase treatment (Invitrogen DNase I amplification Grade kit). The expression levels of TNF-α, IL-6, IL-1β, and MMP-9 were measured using qPCR with the primers listed in **Table 1** below. The PCR reaction was carried out with an initial denaturing step at a temperature of 95◦C for 3 min, followed by 40 cycles of 30 s at 95◦C (denature), 30 s at 55– 62◦C (annealing) and 30 s at 72◦C (extension). A melt curve analysis was also conducted with each qPCR run to detect any non-specific primer binding. The housekeeping genes ActB or 18s (Realtimeprimers.com; Elkins Park, PA) were used to normalize the data. iQTM SYBR Green (Bio-Rad; Mississauga, ON) was used as the detection method and the qPCR reaction was carried out with a DNA Engine Opticon System (Bio-Rad Laboratories Inc., Hercules, CA). The data was analyzed using the method of double delta Ct analysis (2−11Ct).

#### Cytokine Antibody Array and ELISAs

A mouse cytokine antibody array (Mouse Cytokine Array C1) was purchased from Ray Biotech Inc. (Peachtree Corners, GA). BMDMs were plated onto 6-well tissue culture plates and incubated at 37◦C and 5% CO<sup>2</sup> for 24 h. BMDMs were challenged with IgG-sRBCs or C3bi-sRBCs for 16 h. Conditioned medium was collected and the array performed according to manufacturer's instructions. The blots were imaged using a Chemidoc system from Bio-Rad. Densitometric analysis was performed with ImageJ software. The negative control spots


on the array were selected as background and the positive control spot containing only biotinylated antibody was used to normalize the spot intensity (similar to a housekeeping protein). Relative changes in cytokine expression levels were determined by normalizing conditions to the control condition (BMDMs treated with PMA only).

ELISA Duosets (R&D Systems, Inc., Minneapolis, MN) were performed with the conditioned medium obtained from BMDMs 24 h after the addition of opsonized particles. For some experiments, unbound sRBCs were removed after 1 h of incubation with macrophages. For NF-κB inhibition experiments, 100 nM of JSH-23 or 10 ng/ml Bortezomib was added at the time of opsonized-particle addition. To inhibit Syk kinase, 50µM of Piceatannol was added 30 min before phagocytosis. ELISAs were performed according to the Manufacturer's instructions to measure the protein expression of TNF-α, IL-6, and IL-1β, along with the use of the DuoSet ELISA ancillary reagent kit. Color intensity was measured utilizing the Synergy Neo2 Multi-Mode Microplate Reader (BioTek Instruments Inc., Winooski, VT). Data was analyzed through a Dunnett's multiple comparisons test to the control condition without PMA.

#### Immunoblotting and Densitometry Analysis

BMDMs were challenged with either IgG- or C3bi-opsonized sRBCs for 1.5 h. Total cell lysates were extracted by scraping cells in RIPA buffer. The RIPA buffer was made by diluting 5X RIPA buffer [containing Tris-HCl (250 nM pH 7.4), NaCl (750 mM), Triton X-100 (5%), Sodium Deoxycholate (2.5% w/v), SDS (0.5%)] (Bio Basic Canada Inc., Markham, ON) with ddH2O, and 1X protease/phosphatase inhibitors [containing aprotinin (80µM), bestatin (4 mM), E-64 (1.4 mM), leupeptin (2 mM), pepstatin A (1.5 mM) (Sigma)]. Protein concentrations were determined at 750 nm using the DCTM Protein Assay kit (Bio-Rad) according to Manufacturer's instructions. Equal amounts of protein were loaded and separated using 10% SDS-PAGE. After transfer, membranes were blocked for 1 h with either 5% skim milk or 5% BSA in Tris Buffered Saline containing Tween20. Primary antibodies were diluted in blocking solutions at the following dilutions: IκBα (1/1,000), p-ERK (1/1,000), or ERK1/2 (1/1,000) and incubated with membranes overnight at 4◦C. The loading control was β-actin (1/5,000). Membranes were then incubated with either rabbit or mouse HRP-coupled secondary antibodies (1/1,000) for 1 h at RT. Blots were washed 3 times for 5 min and then visualized using SuperSignal West Pico Chemiluminescent Substrate Kit (Thermo Fisher Scientific Inc.). Densitometric analysis was conducted from three independent experiments using ImageJ software. The blots used for analysis were not saturated or over-exposed.

#### Immunostaining and Fluorescent Imaging

After 3 h of phagocytosis, macrophages were fixed with 4% paraformaldehyde in PBS for 20 min. Any external LBs were detected using a donkey anti-human AffiniPure CyTM5 secondary antibody against human IgG. The cells were permeabilized using 0.1% Triton X-100 in PBS supplemented with 100 mM glycine for 20 min. Cells were then blocked using PBS containing 5% FBS for 1 h. Cells were incubated with NF-κB (1:200) diluted in PBS with 1% FBS for 1 h followed by labeling with a donkey anti-rabbit CyTM3 secondary antibody (1:1,000) in PBS with 1% FBS for 1 h. For nuclear staining, cells were washed twice with ddH2O and incubated for 10 min with DAPI (1:10,000). Cells were then mounted using Dako Fluorescent Mounting Medium (Agilent Technologies Inc., Santa Clara, CA). Cells were visualized using a 63x oil-immersion objective using an inverted Zeiss epifluorescent microscope equipped with AxioVision software (Carl Zeiss Canada Ltd., North York, ON). Linear adjustments (contrast and brightness) to images acquired by epifluorescence microscopy were conducted using Axiovision software. Non-linear adjustments were not made. Figures were prepared using Adobe Illustrator CS6 (San Jose, CA).

### Calpain Inhibitor Assay

BMDMs were challenged with opsonized sRBCs as described above for 1 h. Then 100µM of calpain inhibitor diluted in DMSO (PD150606, Sigma Aldrich) was added to each condition for 1 h. mRNA was isolated followed by qPCR for cytokine expression analysis, as described above. For the cytokine antibody array, phagocytosis was performed for 8 h, with 100µM of PD150606 added 1 h after phagocytosis. After 8 h, the conditioned supernatant was collected, and the cytokine array assay performed as described earlier.

# Data Analysis and Statistics

The statistical analysis was conducted on Prism from GraphPad Software Inc. (La Jolla, CA). A two-way ANOVA was used followed by Tukey's or Dunnett's test for multiple comparisons. Differences with a confidence interval of 95% between the specified conditions and the control (generally the PMA-stimulated macrophages unless otherwise indicated) was considered significant during analysis of the data. The data is represented as the mean ± standard error of the mean (S.E.M.) from three biological replicates.

# RESULTS

# CR-Mediated Phagocytosis Is Proinflammatory

It is widely believed that FcγR- but not CR-mediated phagocytosis is proinflammatory in macrophages (21, 22, 25). However, the mechanism by which FcγR signals to transcription factors (e.g., NF-κB) has not been elucidated. We revisited the early biochemical studies of phagocytosis in murine macrophages and used qPCR and sensitive, quantitative protein assays to examine the levels of proinflammatory mediators. To begin, we investigated mRNA levels of TNF-α, IL-1β, IL-6, and MMP-9 in bone marrow-derived murine macrophages (BMDMs) after phagocytosis of IgG- or C3bi-opsonized targets. These inflammatory proteins play an important role in the initiation, progression and resolution of inflammation and thus were chosen as markers of inflammation (26). PMA was used to activate the macrophage complement receptors, αMβ2, to enable uptake of C3bi-opsonized sRBCs (27). However, PMA can signal downstream to activate NF-κB (28) so we used BMDMs only stimulated with PMA as the negative control, to which the other conditions were normalized to. Phagocytosis assays were carried out in serum-free DMEM to avoid non-specific activation of receptors by serum components (29). The qPCR data presented in **Figure 1A** shows the transcript levels of TNF-α, IL-1β, IL-6, and MMP-9 in control cells and BMDMs after ingesting either IgG-sRBCs or C3bi-sRBCs for 3 h. PMA-stimulated BMDMs challenged with C3bi-sRBCs produced significantly higher levels of TNF-α, IL-1β, IL-6, and MMP-9 compared to control BMDMs. Significant upregulation of these proinflammatory genes was not observed in BMDMs that had internalized IgG-sRBCs. Unexpectedly, resting BMDMs challenged with C3bi-sRBCs produced similar levels of IL-6 compared to PMAstimulated BMDMs challenged with C3bi-sRBCs (**Figure 1A**). Immunofluorescence inspection of these cells revealed a basal level of phagocytosis of C3bi-sRBCs even in the absence of PMA (data not shown). To ensure that the proinflammatory effects we observed were not due to components on the sRBC membranes, we repeated these experiments using opsonized latex beads (LBs) (**Figure 1B**). Very similar results were observed with opsonized LBs, compared to opsonized sRBCs (**Figure 1A**), with ingestion of complement opsonized LBs invoking higher expression of proinflammatory mediators vs. IgG-oponized latex beads (**Figure 1B**). Classically activated BMDMs stimulated with LPS/IFN-γ were used as a positive control (30) (**Figure 1C**).

## CR-Mediated Phagocytosis Induces More Cytokine Secretion Than FcγR-Mediated Phagocytosis

We next extended and validated our mRNA data with protein analysis of proinflammatory mediators. We first employed a cytokine antibody array to analyze the levels of secreted cytokines (**Figure 2A**). BMDMs were stimulated with PMA prior to phagocytosis which was allowed to proceed for 16 h to accumulate detectable cytokines within the media. The conditioned media was collected and analyzed using the cytokine array kit (**Figure 2B**). As a positive control, PMA-stimulated BMDMs were also treated with LPS/IFN-γ for 16 h (**Figure 2B**). Densitometric analysis of three replicate experiments showed enhanced proinflammatory mediator secretion following phagocytosis of C3bi-sRBCs with significantly higher IL-6 secretion compared to control cells and BMDMs that had ingested IgG-sRBCs (**Figure 2C**).

We extended the time period of phagocytosis to 24 h and utilized ELISAs to examine select proinflammatory cytokines. Significantly more TNF-α and IL-6 was secreted from BMDMs after ingestion of C3bi-sRBCs, compared to cells undergoing phagocytosis of IgG-sRBCs (**Figure 3A**). This trend was consistent when the cytokine levels detected by ELISA were normalized to the phagocytic index (number of ingested particle/100 macrophages) for each opsonized target (**Figure 3B**). To see if this was limited to bone marrow-derived macrophages, we also investigated proinflammatory cytokine production during phagocytosis in mouse peritoneal primary macrophages. Conditioned media after 24 h of phagocytosis in mouse peritoneal macrophages was subjected to ELISA and both IL-6 and TNF-α levels were significantly increased after ingestion of C3bi-sRBCs, compared to IgG-sRBCs, or conditioned media from control peritoneal macrophages (**Figure 3C**). Typically for our assays, we expose macrophages to an excess of opsonized targets over the experimental time period. To ensure that unbound sRBCs were not dying/ degrading and inducing an inflammatory reaction in macrophages, we washed off unbound sRBCs after 1 h of phagocytosis and compared cytokine levels to macrophages exposed to a constant supply of sRBCs. Levels of IL-6 and TNF-α in conditioned media was not significantly different in the sRBC "washout" experiments compared to cytokine levels in BMDMs continuously exposed to opsonized targets (**Figure 3D**). We were next interested in whether particle internalization itself induced proinflammatory cytokine production or whether particle internalization induced the inflammatory cascade. We pretreated macrophages with 50µM Piceatannol for 30 min to inhibit Syk kinase (31, 32) prior to phagocytosis assays. We monitored phagocytosis and while there were less ingested IgG-sRBCs and C3bi-sRBCs compared to untreated controls, the results were not significant (**Figure 3E**). At 24 h cytokine levels were measured and a significant reduction in IL-6 was observed in Piceatannol-treated macrophages ingesting C3bi-sRBCs, compared to untreated control cells (**Figure 3F**). TNF-α levels were also reduced after CR-mediated phagocytosis compared to untreated BMDMs ingesting C3bisRBCs but these results were not significant (**Figure 3G**). Based on the continued uptake of sRBCs in the presence of Piceatannol (**Figure 3E**) the effects of particle uptake inhibition on proinflammatory cytokine production remain inconclusive. Finally, we wanted to assess whether PMA was amplifying the proinflammatory cytokine expression in macrophages. To address this we activated integrins in BMDMs using 300 U/ml GM-CSF (23, 24). We first confirmed that treatment of macrophages with GM-CSF enhanced phagocytosis (**Figure 3H**). We next performed phagocytosis for 24 h in GM-CSF-stimulated macrophages and collected conditioned media to analyze by ELISA. Both IL-6 (**Figure 3I**) and TNF-α (**Figure 3J**) were significantly upregulated in BMDMs ingesting C3bi-sRBCs after GM-CSF stimulation compared to macrophages not undergoing phagocytosis.

#### p-ERK Levels Remain Constant, but IκBα Protein Levels Are Reduced After Ingestion of C3bi- or IgG-sRBCs in BMDMs

Gene upregulation of many proinflammatory mediators is mediated by NF-κB or AP-1 (33). Previous work on CR3 has shown that co-activation with dectin receptors induces TNFα and IL-6 secretion via the Syk-JNK-AP-1 signaling pathway (34). Since ERK is an upstream element of AP-1 activation (35), we first investigated the phosphorylation status of ERK1/2 using immunoblotting. Since we observed gene expression changes after 3 h of phagocytosis, we investigated earlier time points for ERK activity. Phagocytosis was performed in BMDMs for 1.5 h and lysates collected for immunoblotting (**Figure 4A**). Densitometry analysis of immunoblots from three biological

*(Continued)*

FIGURE 1 | (50 ng/mL) and IFN-γ (100 ng/mL) for 3 h, compared to resting cells. Expression levels were normalized using β-actin and the fold change was calculated relative to the control condition with PMA. A two-way ANOVA followed by Tukey's multiple comparison was performed and significance of each condition was evaluated relative to PMA-stimulated control cells (\**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001). Data are plotted as the mean ± S.E.M. from three independent experiments.

three independent experiments.

replicates indicated that there was no significant difference in the level of p-ERK in either C3bi-sRBC-challenged BMDMs or BMDMS ingesting IgG-sRBCs (**Figure 4B**). Stimulation of BMDMs with LPS/IFN-γ did cause enhanced levels of phosphorylated ERK1/2 (**Figure 4B**), as expected (36).

We next turned our investigation toward the NF-κB signaling pathway and examined the protein levels of IκBα using immunoblotting. Proteosome degradation of IκBα is a critical signaling event to release of cytosol-sequestered NF-κB, allowing it to move into the nucleus for gene activation (37). The phagocytosis experiments were repeated for 1.5 h and lysates collected and probed for IκBα (**Figure 5A**). TLR-activation is known to induce NF-κB nuclear translocation (38) and we saw an expected reduction in IκBα levels after treatment of BMDMs with LPS/IFN-γ (**Figure 5B**). BMDMs ingesting both C3bi- and IgG-sRBCs showed significantly lower levels of IκBα compared to control cells. These results encouraged us to look further into the involvement of the NF-κB signaling pathway in proinflammatory cytokine expression during phagocytosis in macrophages.

FIGURE 3 | Conditions of proinflammatory cytokine production during CR-mediated phagocytosis in macrophages. (A) Quantification of ELISA results for TNF-α, IL-6, and IL-1β from conditioned media of BMDMs after 24 h of phagocytosis. Control cells were cells treated with PMA alone. A two-way ANOVA followed by Dunnett's multiple comparison was performed. The significance of each condition was evaluated relative to PMA-stimulated control cells (\**p* < 0.05, \*\**p* < 0.01). Data are plotted as the mean ± S.E.M. from three independent experiments. (B) Normalization of ELISA results for TNF-α, IL-6, and IL-1β from conditioned media of BMDMs after phagocytosis to the phagocytic index of each opsonized target. A two-way ANOVA followed by Dunnett's multiple comparison was performed. The significance of each condition was evaluated relative to PMA-stimulated control cells (\*\**p* < 0.005, \**p* < 0.05). Data shows mean ± S.E.M. from three independent experiments. (C) Peritoneal macrophages challenged with C3bi-opsonized sRBCs produce more pro-inflammatory cytokines than IgG-opsonized sRBCs. ELISA data representing secreted IL-6 and TNF-α by peritoneal macrophages primed with PMA followed by phagocytosis of C3bi- or IgG-sRBCs. Data shows mean ± S.E.M. from three independent experiments (\*\*\**p* < 0.001, \**p* < 0.05 relative to control cells). A one-way ANOVA followed by Tukey's multiple comparison was performed. (D) Quantification of secreted IL-6 and TNF-α through ELISA with or without washout of opsonized-sRBCs after 1 h. Supernatant was removed after 1 h containing sRBCs and replaced with new serum free DMEM and supernatant was collected after 23 h. sRBC treatments without wash out refers to sRBCs persisting in media for the full 24 h until collection. Data are plotted as the mean ± S.E.M. from 3 independent experiments (*†††p* < 0.001, *††††p* < 0.0001 relative to IgG-sRBC wash out), (‡‡‡‡*p* < 0.0001 relative to IgG-sRBC). A one-way ANOVA followed by Tukey's multiple comparison was performed. (E) The effects of Piceatannol treatment on the *(Continued)*

FIGURE 3 | phagocytic index in BMDMs (number of ingested particles/ 100 macrophages). Immunofluorescence images were taken, and the phagocytic index was calculated as number of sRBCs per 100 macrophages. Data are plotted as the mean ± S.E.M. from three independent experiments. (F) The effects of Piceatannol treatment on secreted IL-6 levels and (G) TNF-α levels after phagocytosis for 24 h, detected using ELISA. Data is plotted as the mean ± S.E.M. from three replicate trials (\*\*\**p* < 0.001 relative to control) (*††††p* < 0.0001 relative to C3bi without Piceatannol). A one-way ANOVA followed by Tukey's multiple comparison was performed. (H) Stimulation of BMDMs with GM-CSF (300 U/ml) for 1 h increases phagocytosis of C3bi-sRBCs. Phagocytosis was allowed to proceed for 1 h and then cells were fixed and imaged. The phagocytic index was calculated as number of sRBCs per 100 macrophages. Data plotted as the mean ± S.E.M. from three independent experiments (I) ELISA data representing secreted cytokine levels of IL-6 and (J) TNF-α from BMDMs challenged with IgG- or C3bi-sRBCs after either: no treatment, 7 min PMA pretreatment, or 1 h pretreatment with GM-CSF. Data plotted as the mean ± S.E.M. from three independent experiments (\*\**p* < 0.01 relative to no sRBC with GM-CSF) (\*\*\*\**p* < 0.0001 relative to no sRBC with or without GM-CSF). A one-way ANOVA followed by Tukey's multiple comparison was performed.

# BMDMs That Ingest C3bi-LBs Have Less Cytosolic NF-κB Than BMDMs That Internalize IgG-LBs

Using immunofluorescence, we next directly examined the subcellular distribution of NF-κB, as a proxy to its gene expression activity (37). Phagocytosis was performed for 3 h with opsonized LBs for compatibility with the NF-κB antibody. Control cells were stimulated with LPS/IFN-γ for an equivalent time-period. Immunostaining revealed 3 distinct phenotypes of NF-κB subcellular distribution (**Figure 6A**). NF-κB immunostaining was either distinctly nuclear (with empty cytosol), or cytosolic (with empty nucleus) or "both," for cells with NF-κB strongly labeling throughout the cell (**Figure 6A**). For each condition, we quantified the subcellular distribution of NF-κB from at least 100 BMDMs (**Figure 6C**). As expected, the positive control cells activated with LPS/IFNγ showed predominantly nuclear NF-κB, which interestingly, was reduced with PMA treatment (**Figures 6B,C**). While

there was no significant difference in the number of cells containing nuclear NF-κB localization after phagocytosis of either opsonized particles, compared to control cells, there was a difference in the number of cells exhibiting purely cytosolic NF-κB. After internalization of C3bi-LBs, significantly fewer BMDMs treated with PMA had cytosolic distributions of NFκB compared to PMA-treated control cells or those ingesting IgG-LBs (**Figure 6C**). To more directly test the involvement of NF-κB in phagocytosis-induced proinflammatory cytokine production, we employed NF-κB activation inhibitors. We utilized a 4-Methyl-N1-(3-phenyl-propyl)-benzene-1,2-diamine (JSH-23) and Bortezomib (Bzb) to inhibit NF-kB activity (39, 40). JSH-23 is a novel aromatic compound that inhibits NFkB activity in RAW 264.7 macrophages by preventing nuclear translocation of NF-kB without affecting IkBa degradation (39). Bzb is a reversible 26S proteasome inhibitor (40). During CRmediated phagocytosis, we inhibited NF-kB using either 100 nM

exposed to LPS/IFN-γ for 3 h. Arrowheads indicate BMDMs with nuclear NF-κB. Scale bars = 20µm. (C) Quantification of NF-κB subcellular distribution in BMDMs

*(Continued)*

FIGURE 6 | after phagocytosis or treatment with LPS/IFN-γ, with and without PMA stimulation. Cells were characterized as either having solely nuclear or cytosolic NF-κB immunostaining, or both in which strong signal for NF-κB was observed in the cytosol and the nucleus. *n* > 100 BMDMs. A two-way ANOVA followed by Tukey's multiple comparison was performed. The cytosolic localization of PMA-stimulated BMDMs challenged with C3bi-LBs was compared with PMA-stimulated BMDMs and BMDMs ingesting IgG-LBs (\**p* < 0.05). Significantly more LPS/IFN-γ-stimulated BMDMs showed nuclear NF-κB and less cytosolic staining, which is not indicated on graph. Data are plotted as the mean ± S.E.M. from three independent experiments. (D) Quantification of ELISA detection of secreted IL-6 and (E) TNF-α after treatment of BMDMS with the NF-kB inhibitors JSH-23 and Bortezomib during CR-mediated phagocytosis. Data is plotted as the mean ± S.E.M. from three independent experiments (\*\**p* < 0.01 relative to control condition) (*††p* < 0.01, *† p* < 0.01 relative to C3bi-sRBC). A one-way ANOVA followed by Tukey's multiple comparison was performed.

of JSH-23 or 10 ng/ml BzB and measured IL-6 and TNF-α levels after 24 h using ELISA. BMDMs challenged with C3bi-sRBCs in the presence of Bzb or JSH-23 produced significantly less IL-6 than untreated BMDMs (**Figure 6D**). The TNF-α levels secreted by BMDMs in the presence of the NF-kB inhibitors also showed lower levels compared to untreated macrophages ingesting C3bisRBCs but these results were not significant (**Figure 6E**).

## Calpain Inhibition Reduces Phagocytosis-Induced Upregulation of Inflammatory Cytokines

To gain insight into the mechanism by which phagocytosis promotes proinflammatory cytokine and chemokine production, we turned our attention to calpain. We became interested in this cysteine protease since CR3 is an integrin and calpain is involved in integrin signaling through degradation of talin (41). Importantly, IκBα is another substrate of calpain (42), implicating a potential involvement in phagocytosis-induced cytokine production. Calpain has been shown to indirectly play a role in FcγR-mediated phagocytosis as the cleavage of ASAP2 by calpain reduced particle uptake during FcγRmediated phagocytosis (43). We employed the calpain inhibitor, the PD150606 peptide, which inhibits the protease core domains of both calpain-1 and calpain-2 (44) and examined gene expression and cytokine and chemokine protein expression during phagocytosis in BMDMs. To avoid particle internalization issues in the presence of the calpain inhibitor, we allowed phagocytosis to occur for 1 h prior to addition of 100µM PD150606. For qPCR studies, we continued phagocytosis for an additional hour in the presence of the peptide prior to cell extraction. We investigated TNF-α and IL-1β as they had the most robust expression after 3 h of phagocytosis of C3bi-sRBCs (**Figure 1A**). In untreated cells, after 2 h of phagocytosis of C3bi-sRBCs, we again saw upregulation of TNF-α and IL-1β as well as upregulation of IL-1β following IgG-sRBC ingestion (**Figure 7A**). Cells treated with 100µM PD150606 had significant less IL-1β levels for both phagocytic targets, compared to untreated cells (**Figure 7A**). The CRmediated upregulation of TNF-α was also significantly blunted by PD150606 treatment (**Figure 7A**).

We next investigated how calpain inhibition would affect the expression of cytokines at a secreted protein level during phagocytosis. Due to the instability of the peptide, the assays were only carried out for 8 h after which conditioned media was exposed to a cytokine antibody array. In control cells, significant levels of IL-6 and CCL5, a chemokine which recruit monocytes, NK and T cells to sites of inflammation (26), were detected following uptake of C3bi-sRBCs which was not observed in control cells or after ingestion of IgG-sRBCs (**Figures 7B,C**). However, the secreted protein levels of IL-6 and CCL5, were significantly lower from BMDMs treated with PD150606 to inhibit calpain, compared to untreated BMDMs ingesting C3bi-sRBCs (**Figures 7B,C**). Together, these results implicate calpain involvement in proinflammatory cytokine and chemokine secretion induced by CR-mediated phagocytosis.

# DISCUSSION

A long-standing belief in the scientific community is that phagocytosis mediated by FcγRs is potent at inducing inflammation whereas CR-mediated phagocytosis in macrophages is non-inflammatory. Recent efforts have concentrated on understanding CR-mediated phagocytosis given the prompt, pivotal role of complement in clearing infections. Low levels of bacteria can be quickly resolved by circulating complement, while rapidly proliferating pathogens necessitate time consuming antibody-mediated solutions of the adaptive immunity (45). Studies so far have revealed striking differences between the modes of phagocytosis induced by the opsonic receptors in macrophages, including a strong requirement for the RhoA GTPase, and not rac1 and cdc42, for CR-mediated phagocytosis (46).

We revisited the inflammatory capacity of opsonic receptor signaling using quantitative mRNA and protein approaches. Surprisingly, we observed enhanced proinflammatory gene and secreted protein levels during CR-mediated phagocytosis. This deviates from previous work in mouse macrophages (21) which may be explained in part from experimental discrepancies including differences in macrophage subtype and proinflammatory readout. The original study by Aderem et al. (21) examined murine resident peritoneal macrophages and detected levels of arachidonic acid, which is a signaling fatty acid that can become inflammatory if generated by phospholipase A2 (47). Arachidonic acid is metabolized to make prostaglandins and leukotriene B4, that cause arterial dilation to enhance immune cell infiltration into the sites of inflammation (47). While arachidonic acid levels were not increased following CR activation in the original study, it is possible the downstream bioactive inflammatory mediators were enriched. We decided to examine proinflammatory cytokines and chemokines during phagocytosis, which serve as important chemical messengers to trigger the inflammation cascade including neutrophil recruitment, edema and hemorrhaging. Both mRNA and protein levels were examined during CR-mediated phagocytosis and compared to FcγR-mediated phagocytosis and control BMDMs. CR-mediated phagocytosis provoked significant increases in

BMDMs after 8 h of phagocytosis, with 100µM of PD150606 included during CR-mediated phagocytosis. (C) Densitometry analysis of IL-6, IL-12, TNF-α, CCL2, MCP-5, and CCL5 from replicate array blots. Expression level was normalized to positive biotinylated antibody signal spots and then to the control cells + PMA condition for each cytokine. A two-way ANOVA followed by Tukey's multiple comparison was performed (\*\*\**p* < 0.001, \*\*\*\**p* < 0.0001 relative to control condition), ( *††p* < 0.01, *††††p* < 0.0001 relative to untreated C3bi-sRBC). Data are plotted as the mean ± S.E.M. from four biological replicates.

cytokine transcript and secretion of the potent pyrogen TNF-α (26) in murine macrophages. Similar results were observed for IL-6, which along with TNF-α, is secreted rapidly upon macrophage stimulation to induce fever and stimulate the acute phase of the immune response (26). Additionally, the cytokine array revealed a CR-specific upregulation of the chemokine MCP-5/CCL12, which recruits eosinophils, monocytes and lymphocytes to sites of inflammation (26).

A priority for our work was resolving the mechanism behind CR-mediated proinflammatory cytokine expression. Our preliminary observations indicated that phagocytic receptors signal downstream to NF-κB, but not to AP-1. Upregulation of cytokines can be induced by stimuli that lead to the phosphorylation and ubiquitination and degradation of IκBα, which results in the nuclear translocation of active NF-κB (38). While we did not observe significantly more NF-κB in the nucleus during phagocytosis, we did see significantly reduced cytosolic sequestration of NF-κB during CR-mediated phagocytosis. Importantly, inhibition of NF-κB activation blunted proinflammatory cytokine production during CRmediated phagocytosis. Our immunoblotting data of IκBα, indicated that both CR- and FcγR-mediated phagocytosis reduce IκBα levels in macrophages, supporting NF-κB activation during phagocytosis. Although the potential activation of NF-κB was not exclusive to CR-mediated phagocytosis, it may explain the slight increase in proinflammatory cytokines observed during FcγR-mediated phagocytosis in BMDMs. Calpain was implicated in cytokine gene induction during CR-mediated phagocytosis as inhibition of calpain significantly reduced cytokine transcripts and secreted protein levels in BMDMs ingesting C3bi-sRBCs. Future studies should directly examine whether IκBα is a substrate for calpain in macrophages and how it coordinates with the 26S proteasome in mediating IκBα degradation. Calpain inhibitors are being developed as effective anti-inflammatory drugs (48) and our results may provide mechanistic insight into their effectiveness.

During an immunological challenge, the goal of immune cells is ultimately to remove the invader with minimal self-tissue damage. The production of cytokines during an immunological challenge is likely temporally-regulated by the potency of the stimulus. LPS is a known potent stimulus of the inflammatory cascade but the potency of opsonins has remained unclear. Enhanced proinflammatory cytokine expression by macrophages ingesting complement-coated targets may help explain the early role for complement in infections to mobilize the adaptive immunity. We only observed slight, but not significant, induction of proinflammatory cytokines during FcγR-mediated phagocytosis in murine macrophages. These results could be due to the absence of FcγRIIa in mice (49). Future studies should directly compare the macrophage species to resolve these discrepancies, but in light of Fc signaling, our data may be relevant to human thrombosis patients who frequently have FcγRIIa mutations (50). The bulk of FcγR signaling and cytokine induction research has been in the context of opsonized pathogens, where co-receptor signaling is involved, and amplified, by FcγR activation (49, 51). Additionally, it is

#### REFERENCES


very difficult to try to fit FcγR and CR signaling into defined inflammatory categories because negative feedback mechanisms induced via inhibitory receptors can dampen the response by inducing anti-inflammatory molecules (5, 52).

While we observed significant upregulation of proinflammatory mediators during CR-mediated phagocytosis, the fold-change increases were often an order of magnitude lower than LPS stimulation. Thus, while detectable, the "danger signal" induced by opsonic receptor ligation is modest compared to signals induced by microbe-associated molecular patterns. This is to be expected as normal clearance of sRBCs is not a state of infection. We employed sRBCs and latex beads as target particles to tease out individual opsonic receptor signaling effects on cytokine induction. Additionally, these "inert" particles are relevant to phagocytosis events in vivo. RBC opsonization and phagocytosis are observed during senescence and in patients with beta thalassemia (53). Latex beads are composed of polystyrene, an emerging FDA-approved biomaterial (54) and serve as a model for foreign bodies known to be opsonized and inflammatory in nature (55). Thus, these studies add insight into opsonic receptor signaling and importantly reveal novel aspects regarding the under-studied CR mode of phagocytosis. Together our work indicates that opsonic receptors signal differently to the nucleus and additional differences will become apparent as other molecular players are investigated in detail.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article.

#### ETHICS STATEMENT

The animal study was reviewed and approved by the University of Toronto Local Animal Care Committee.

#### AUTHOR CONTRIBUTIONS

DA, XL, and RH performed the experiments, data analysis, prepared the figures, and wrote the first draft of the manuscript. REH provided feedback and guidance on all aspects of the project and wrote the final version of the manuscript.

#### FUNDING

This project was funded by a Natural Sciences and Engineering Research Council grant (RGPIN 2017-06087) and a grant from Canadian Institutes of Health Research grant (PJT-166084) to REH.


the risk of thrombosis in patients with essential thrombocythemia. Eur J Haematol. (2014) 93:103–11. doi: 10.1111/ejh.12307


**Conflict of Interest:** 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 © 2020 Acharya, Li, Heineman and Harrison. 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.

# Phosphoinositides and the Fate of *Legionella* in Phagocytes

A. Leoni Swart and Hubert Hilbi\*

Faculty of Medicine, Institute of Medical Microbiology, University of Zürich, Zurich, Switzerland

Legionella pneumophila is the causative agent of a severe pneumonia called Legionnaires' disease. The environmental bacterium replicates in free-living amoebae as well as in lung macrophages in a distinct compartment, the Legionella-containing vacuole (LCV). The LCV communicates with a number of cellular vesicle trafficking pathways and is formed by a plethora of secreted bacterial effector proteins, which target host cell proteins and lipids. Phosphoinositide (PI) lipids are pivotal determinants of organelle identity, membrane dynamics and vesicle trafficking. Accordingly, eukaryotic cells tightly regulate the production, turnover, interconversion, and localization of PI lipids. L. pneumophila modulates the PI pattern in infected cells for its own benefit by (i) recruiting PI-decorated vesicles, (ii) producing effectors acting as PI interactors, phosphatases, kinases or phospholipases, and (iii) subverting host PI metabolizing enzymes. The PI conversion from PtdIns(3)P to PtdIns(4)P represents a decisive step during LCV maturation. In this review, we summarize recent progress on elucidating the strategies, by which L. pneumophila subverts host PI lipids to promote LCV formation and intracellular replication.

#### *Edited by:*

Carlos Rosales, National Autonomous University of Mexico, Mexico

#### *Reviewed by:*

Hayley J. Newton, The University of Melbourne, Australia Yuxin Mao, Cornell University, United States

> *\*Correspondence:* Hubert Hilbi hilbi@imm.uzh.ch

#### *Specialty section:*

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

*Received:* 14 November 2019 *Accepted:* 08 January 2020 *Published:* 30 January 2020

#### *Citation:*

Swart AL and Hilbi H (2020) Phosphoinositides and the Fate of Legionella in Phagocytes. Front. Immunol. 11:25. doi: 10.3389/fimmu.2020.00025 Keywords: *Dictyostelium discoideum*, effector protein, endoplasmic reticulum, host-pathogen interaction, macrophage, pathogen vacuole, type IV secretion, vesicle trafficking

#### *LEGIONELLA PNEUMOPHILA*—AN AMOEBAE-RESISTANT ENVIRONMENTAL BACTERIUM

Legionella spp. are obligate aerobic, Gram-negative bacteria, which are ubiquitously found in technical and natural water systems, where they colonize different niches (1, 2). The facultative intracellular bacteria replicate in planktonic form as well as in biofilms (3–5), and they infect environmental predators such as nematodes (6–9) and protozoa (10–12). Complex, ecologically relevant interactions take place in the aquatic niches inhabited by Legionella spp.; e.g., nematode larvae rupture Legionella-infected amoebae and thus are exposed to a highly virulent form of the bacterial pathogen (9).

Upon inhalation of contaminated water droplets, Legionella bacteria reach the lung, where they replicate in and destroy alveolar macrophages, thus causing a potentially fatal pneumonia termed Legionnaires' disease (2). The clinically most relevant and best studied species is Legionella pneumophila; yet, Legionella longbeachae is prevalent in some parts of the world, too (13). The spread of Legionella spp. predominantly occurs through environmental sources; however, a probable person-to-person transmission of L. pneumophila, resulting in the death of the two people involved, was recently reported (14).

Legionella pneumophila replicates intracellularly in amoebae and macrophages by exploiting evolutionarily conserved pathways (15, 16). The pathogen forms a unique, degradationresistant compartment, the Legionella-containing vacuole (LCV), wherein which bacterial replication takes place. The LCV does neither acidify nor fuse with lysosomes, but communicates with several vesicle trafficking pathways including the endosomal, secretory, and retrograde routes (17–21). At later steps of pathogen vacuole maturation, the LCV tightly and continuously associates with the endoplasmic reticulum (ER). Small GTPases of the Arf (22, 23), Rab (24, 25), Ran (26), and Rap (27) families regulate LCV formation and intracellular replication of L. pneumophila. Moreover, large GTPases implicated in eukaryotic membrane fusion and fission play a role in L. pneumophila infection. Atlastin3 (Atl3/Sey1), an ER tubule-resident large GTPase that catalyzes homotypic ER fusions, promotes ER remodeling around LCVs, pathogen vacuole expansion and intracellular bacterial replication (28). Dynamin1-like GTPase (Dnm1l), a mitochondrial large GTPase, mediates L. pneumophila-induced mitochondrial fragmentation and inhibition of host cell respiration (29).

LCV formation requires the Icm/Dot (intracellular multiplication/defective organelle trafficking) type IVB secretion system (T4SS), which is conserved among Legionella spp., and in the case of L. pneumophila translocates more than 300 different "effector" proteins into host cells (30, 31). In eukaryotic cells, the effector proteins subvert essential process such as signal transduction, cytoskeleton dynamics and membrane trafficking (17, 32–37). Distinct effector proteins have been shown to target the small GTPases Arf1 (22), Rab1 (38–41) or Ran (26, 42), the retromer coat complex (43–46), the vacuolar H+-ATPase (47), the autophagy machinery (48–50), or phosphoinositide (PI) lipids (35, 51, 52). Here, we focus on how L. pneumophila subverts host PI lipids to promote LCV formation and intracellular replication.

#### PHOSPHOINOSITIDE LIPIDS—REGULATORS OF ORGANELLE IDENTITY AND MEMBRANE DYNAMICS

Phosphoinositides are minor constituents of eukaryotic membranes (<10% of all phospholipids), but this low abundance class of lipids exert pivotal functions for cellular organelle identity, membrane dynamics and vesicle trafficking (53–56). Accordingly, the production, turnover, interconversion, and subcellular localization of PI lipids are tightly regulated by eukaryotic cells. The core compound of PI lipids is phosphatidylinositol (PtdIns), comprising a diacylglycerol (DAG) moiety and a D-myo-inositol 1-phosphate head group facing the cytoplasmic side of membranes (**Figure 1**). PtdIns

can be reversibly phosphorylated at the positions 3, 4, and/or 5 of the inositol ring, giving rise to seven different mono- or poly-phosphorylated derivatives (53–56). These reactions are catalyzed by organelle-specific PI metabolizing enzymes (PI kinases and PI phosphatases), the activity of which controls compartmentalization and vesicle trafficking within the cell (57, 58).

PI lipids, jointly with small GTPases in their active GTPbound form, recruit peripheral membrane proteins harboring distinct PI-binding motifs, such as the PH, PX, FYVE, ENTH/ANTH, or FERM domains (59). Hence, lipid-protein co-incidence detection, along with specific adaptor proteins, determines organelle identity and vesicle trafficking routes in eukaryotic cells (54, 60). PI-metabolizing enzymes are usually recruited to the cytoplasmic side of cellular membranes by small GTPases; e.g., the endosomal small GTPase Rab5 recruits and activates the class III phosphatidylinositol 3-kinase (PI3K) to produce PtdIns(3)P from PtdIns (61). The small GTPases themselves are localized and activated by specific guanine nucleotide exchange factors (GEFs), which concomitantly displace the guanine nucleotide dissociation inhibitor (GDI) protein from the small GTPase, thus allowing the membrane association of the GTPase. To switch off the signal, the inactivation of small GTPases is catalyzed by specific GTPase activating proteins (GAPs) (61).

The different PIs preferentially localize to distinct subcellular compartments and pathways [(53, 54, 62); **Figure 2**]. Accordingly, PtdIns(4)P and in particular PtdIns(4,5)P<sup>2</sup> are enriched at the plasma membrane, where PtdIns(3,4,5)P<sup>3</sup> and PtdIns(3,4)P<sup>2</sup> transiently accumulate upon signal transduction events and during phagocytosis. PtdIns(3)P is the "signpost" PI lipid of the endocytic pathway, and is enriched on phagosomes and early endosomes, as well as on autophagosomes and

**Abbreviations:** AMPylase, adenylyltransferase; DAG, diacylglycerol; Icm/Dot, intracellular multiplication/defective organelle trafficking; GAP, GTPase activating protein; GDI, guanine nucleotide dissociation inhibitor; GEF, guanine nucleotide exchange factor; LCV, Legionella-containing vacuole; OCRL, oculocerebrorenal syndrome of Lowe; PI, phosphoinositide; PI3/4/5K, PI 3-/4-/5-kinase; PtdIns, phosphatidylinositol; T4SS, type IV secretion system.

multivesicular bodies, which like late endosomes and lysosomes are also decorated with PtdIns(3,5)P2. PtdIns(4)P is the hallmark PI lipid of the secretory pathway and predominantly localizes to the Golgi apparatus and secretory vesicles (53, 54, 56, 62). This PI lipid is formed from PtdIns on the ER and together with PtdIns(3)P also regulates phagosome-lysosome fusion (63).

On certain compartments and along some vesicle trafficking pathways, distinct PIs are functionally coupled, i.e., the product of a given PI-metabolizing enzyme is the substrate of a subsequent modification. This occurs, e.g., in the endocytic pathway, where PtdIns(3)P is phosphorylated to yield PtdIns(3,5)P2, as well as in the secretory pathway, where PtdIns(4)P serves as the precursor of PtdIns(4,5)P<sup>2</sup> at the plasma membrane. In turn, PtdIns(4,5)P<sup>2</sup> is phosphorylated by class I PI3K to transiently yield PtdIns(3,4,5)P<sup>3</sup> during phagocytosis.

#### EUKARYOTIC PI KINASES IMPLICATED IN UPTAKE AND ENDOCYTOSIS OF *L. PNEUMOPHILA*

PtdIns(3,4,5)P<sup>3</sup> and PtdIns(3)P are produced by class I or class III PI3Ks and are major regulators of phagocytosis or the endocytic pathway, respectively. Using the haploid social soil amoeba Dictyostelium discoideum, genetic and pharmacological disruption of class I PI3Ks indicated that these kinases are largely dispensable for uptake of wild-type L. pneumophila, but required for uptake of an icm/dot mutant strain (51, 64). Moreover, using D. discoideum producing a fluorescent probe for PtdIns(3,4,5)P3, live-cell microscopy revealed that this PI lipid accumulated at bacterial entry sites and was cleared within approximately 40 s after uptake, regardless of whether the amoebae were infected with wild-type or icm/dot mutant L. pneumophila. In parallel, plasma membrane PtdIns(4,5)P<sup>2</sup> disappeared from the uptake sites (65).

Similar to amoebae, the uptake of L. pneumophila wild-type, but not the icm/dot mutant strain by replication-permissive human U937 macrophage-like cells was not affected by the class I PI3K inhibitor wortmannin (66, 67). In contrast, wortmannin or LY294002 inhibited the uptake of wild-type as well as icm/dot mutant L. pneumophila by non-permissive murine J774A.1 macrophages (64, 66, 67). The Icm/Dot T4SS controls the uptake of L. pneumophila by phagocytes (68, 69); however, no effectors implicated in the process have been identified. These results suggest that during uptake of L. pneumophila class I PI3Ks are activated and the pathogen evades/inhibits downstream processes in an Icm/Dot-dependent manner to form the replication-permissive compartment.

Dictyostelium discoideum mutant strains were also used to examine the role of endosomal PI kinases, PI phosphatases and phospholipases for intracellular growth of L. pneumophila. Wildtype L. pneumophila replicated more efficiently in D. discoideum lacking two or five class I PI3Ks (51, 64) or in amoebae lacking PIKfyve (70), a PI 5-kinase, which is recruited through its FYVE domain to early endosomes, where it phosphorylates PtdIns(3)P to yield PtdIns(3,5)P2. While it is not clear how lower levels of PtdIns(3,4,5)P<sup>3</sup> promote the intracellular replication of L. pneumophila, the reduction of PtdIns(3,5)P<sup>2</sup> impairs the bactericidal endocytic pathway, which restricts bacterial killing and thus benefits the pathogen (70). The disruption of D. discoideum PTEN (phosphatase and tensin homolog), a PI phosphatase antagonizing PI3Ks, reduces the uptake of L. pneumophila but does not affect intracellular growth (64). Finally, the inhibition of D. discoideum PLC (Phospholipase C), a hydrolase cleaving PI(4,5)P<sup>2</sup> to yield DAG and inositol 1,4,5 phosphate (IP3), also abolishes the uptake of L. pneumophila, but again has no effect on bacterial replication (64).

#### PHOSPHOINOSITIDE CONVERSION ON THE *LEGIONELLA*-CONTAINING VACUOLE

PtdIns(3)P accumulates on LCVs within 1 min after uptake, regardless of whether the vacuole contains wild-type or icm/dot mutant L. pneumophila (71). However, while phagosomes containing icm/dot mutant bacteria remain decorated with PtdIns(3)P, more than 80% of wild-type LCVs gradually lose this PI within 2 h. Concomitantly, major membrane rearrangements take place with PtdIns(3)P-positive membranes being segregated from the LCV and compacted at the cell center. PtdIns(4)P, on the other hand, transiently localizes to early phagosomes harboring wild-type or icm/dot mutant L. pneumophila, but is cleared within minutes after uptake. During the following 2 h, PtdIns(4)P steadily accumulates only on wild-type LCVs, which for at least 8 h maintain a discrete PtdIns(4)P identity spatially separated from the calnexin-positive ER. PtdIns(4)P decorates the LCV for a prolonged time (18 h p. i. and beyond) up to when the bacteria exit from the pathogen vacuole and the infected cell (71). Taken together, within 2 h post-infection, the LCV undergoes a PI conversion, replacing the endosomal PtdIns(3)P with the secretory PtdIns(4)P (**Figure 3**). Importantly, the LCV PI conversion occurs prior to and independently from ER recruitment, and the two compartments appear to remain separate throughout the intracellular life of L. pneumophila.

Mechanistically, the PI conversion on the LCV possibly proceeds along several, mutually non-exclusive pathways: (i) the LCV might communicate and selectively retain PI-decorated vesicles, (ii) L. pneumophila might produce (Icm/Dot-secreted) effectors acting directly as PI interactors, phosphatases or kinases, and/or (iii) the pathogen might subvert host PI metabolizing enzymes (**Figure 3**). Indeed, using D. discoideum producing fluorescent PtdIns(3)P and PtdIns(4)P probes in tandem, we recently showed by high-resolution real-time confocal laser scanning microscopy that nascent LCVs continuously capture and accumulate PtdIns(4)P-positive vesicles derived from the trans-Golgi network (72). The sustained association of the PtdIns(4)P-positive vesicles, but not the LCV-vesicle interactions per se, require a functional T4SS. Thus, L. pneumophila exploits the cellular dynamics of vesicle-bound PtdIns(4)P for LCV formation. At different stages of infection L. pneumophila effectors might modulate the host PI pattern in different ways (73).

As outlined below in detail, L. pneumophila Icm/Dottranslocated effector proteins subvert PI lipids (i) by directly binding PIs (SidC, SidM, RidL, LtpM), (ii) by acting as bacterial PI phosphatases (SidF, SidP), PI kinases (LepB, LegA5), or phospholipases (VipD, PlcC, LpdA), or (iii) by recruiting eukaryotic PI phosphatases or kinases (RalF, SidM). Currently, no effector has been described, which directly modulates the activity of host PI-metabolizing enzyme. In general, L. pneumophila effectors determining the LCV PI pattern might act either in cis (on the LCV membrane) or in trans (in a distance from the LCV). In fact, a number of these effectors have been shown to act in cis, in agreement with their exceptional affinity for specific PI receptors (40, 74–76).

## PHOSPHOINOSITIDE ANCHORS FOR *L. PNEUMOPHILA* EFFECTORS

Legionella pneumophila Icm/Dot substrates translocated to the cytoplasmic face of the LCV can bind to the pathogen vacuole as peripheral membrane protein [e.g., RalF; (77, 78)], as intrinsic membrane protein [e.g., MavN; (79, 80)], through host cell prenylation of a C-terminal CAAX motif [e.g., LegG1, AnkB, LpdA; (81–83)], or through PI lipids [e.g., SidC, SidM, RidL, LtpM; (44, 84, 85); **Figure 4**]. PI lipids bind a plethora of eukaryotic proteins through distinct domains (59), none of which was identified in L. pneumophila effector proteins. However, L. pneumophila produces a battery of effector proteins, which bind through novel domains to PtdIns(4)P (SidC, SdcA, SidM, Lpg1101, Lpg2603, AnkX, LidA) and/or PtdIns(3)P (LepB, RidL, SetA, LtpD, LtpM, RavD, RavZ, AnkX, LidA) (**Table 1**).

The L. pneumophila Icm/Dot substrate SidC and its paralogue SdcA localize to the LCV membrane (115) and almost exclusively bind to PtdIns(4)P [(51); **Figure 4** and **Table 1**]. The 105 kDa effector proteins harbor a unique 20 kDa C-terminal domain termed P4C [PtdIns(4)P-binding domain of SidC], which does not show similarity to any eukaryotic PI-binding motif and was used as a PtdIns(4)P probe in eukaryotic cells (116, 136). SidC and the P4C domain are conserved in Legionella longbeachae, where the 111 kDa effector represents the major PtdIns(4)P binding protein (75). The SidC orthologs of L. pneumophila and L. longbeachae bind PtdIns(4)P with a low dissociation constant (Kd) of ca. 240 or 70 nM, respectively. The crystal structure of SidC revealed a unique PtdIns(4)P-binding domain essential for targeting the effector to the pathogen vacuole (137).

LCVs harboring an L. pneumophila 1sidC-sdcA mutant strain recruit the ER slower and to a smaller extent; yet, the formation of the spatially separated PtdIns(4)P-positive limiting LCV membrane is not affected (28, 51, 65, 116). The interaction with the ER is catalyzed by a 70 kDa N-terminal fragment of SidC (116). The crystal structure of the N-terminal fragment revealed a novel fold (117, 121), comprising a catalytic Cys-His-Asp triad, which is essential for SidC to promote the polyubiquitination of protein substrates on the LCV (118). Indeed, SidC and SdcA act as E3 ubiquitin ligases, which show a broad and nonoverlapping specificity for ubiquitin-conjugating E2 enzymes (118, 119). Hence, the L. pneumophila effector SidC links and subverts two different eukaryotic pathways, phosphoinositide and ubiquitination signaling.

In L. pneumophila-infected phagocytes, SidC decorates the LCV selectively, uniformly and in copious amounts (51, 116). We exploited this feature to isolate LCVs from homogenates of infected host cells by establishing a two-step procedure comprising immuno-affinity enrichment with an anti-SidC antibody, followed by Histodenz density gradient centrifugation (138, 139). Using this protocol, intact LCVs were isolated from D. discoideum amoeba (28, 140), murine RAW 264.7 macrophage-like cells (24, 27) and bone marrow-derived primary macrophages (141). The isolated LCVs were utilized for biochemical fusion experiments (28) and proteomics analysis (24, 27, 140, 141), which identified small GTPases and their effectors (Rab family, Rap1, Ran, RanBP1), large GTPases, components of the endosomal and late secretory trafficking pathways, as well as protein or lipid kinases and phosphatases. LCV localization of some of these proteins was confirmed by fluorescence microscopy using D. discoideum strains producing the corresponding GFP-fusion proteins (24, 26–28, 140, 142).

The Icm/Dot substrate SidM (alias DrrA) localizes to the LCV membrane early during L. pneumophila infection (92) and is the major PtdIns(4)P-binding protein, as it was exclusively identified as such in a non-biased pulldown approach [(84); **Figure 4** and **Table 1**]. In lysates of L. pneumophila 1sidM, no other PI-binding protein (not even SidC) was identified. The 73 kDa effector protein harbors the 12 kDa C-terminal domain P4M [PtdIns(4)P-binding domain of SidM], which does not show similarity to any eukaryotic PI-binding motif or the P4C domain of SidC, but is shared with two other effectors, Lpg1101 (alias Lem4) and Lpg2603 (alias Lem28) [(102); **Table 1**]. The P4M domain has been ectopically produced and used as a PtdIns(4)P probe in eukaryotic cells (143) and Drosophila photoreceptor cells (144). The crystal structure of SidM and biochemical analysis revealed a unique PtdIns(4)P-binding domain and a very high binding affinity (K<sup>d</sup> = 4–18 nM) (40, 74).

SidM, i.e., its central domain, exerts GEF activity toward Rab1-GDI complexes, thus leading to GTP loading and Rab1 activation on LCV membranes (38, 39, 92, 124–127). Moreover, the N-terminal domain of SidM catalyzes the covalent attachment of AMP to Rab1, a reaction termed AMPylation (128), which renders Rab1(GTP) inaccessible to GAPs and causes the constitutive activation of the small GTPase on LCVs (93). The AMPylation reaction is reversible, and the L. pneumophila effector protein SidD can remove the AMP residue from Rab1 by a deAMPylation reaction (145–147). The removal of the covalent modification allows the GAP LepB to inactivate Rab1 (92, 94). Through activation of Rab1, SidM catalyzes the non-canonical pairing of plasma membrane t-SNARE syntaxin proteins (present on the LCV membrane) with the ER-localized v-SNARE protein Sec22b (148, 149). Thus, the SidM-catalyzed activation of Rab1 seems to promote the tethering and fusion of the LCV with ER-derived vesicles, which has been described many years ago (150, 151). In summary, the L. pneumophila effector SidM links and subverts two different eukaryotic pathways, phosphoinositide and small GTPase signaling.

The Icm/Dot substrate LidA supports SidM-dependent recruitment of Rab1 to LCVs (39) and preferentially binds to PtdIns(3)P or with lower affinity to PtdIns(4)P [(84, 103);

4-kinase IIIβ (PI4KIIIβ). OCRL and PI4KIIIβ produce PtdIns(4)P from PtdIns(4,5)P<sup>2</sup> or PtdIns, respectively.

**Figure 4** and **Table 1**]. The 83 kDa effector targets Rab1 and several other host Rab GTPases (152, 153) and binds with high affinity to the GDP- and GTP-bound as well as the AMPylated form of Rab1, thus stabilizing the active conformation of the GTPase and preventing inactivation by GAPs (39, 104, 105).

The Icm/Dot substrate AnkX localizes to LCVs and binds with apparently similar affinity to PtdIns(3)P and PtdIns(4)P [(154); **Figure 4** and **Table 1**]. AnkX covalently attaches a phosphocholine moiety to GDP-bound Rab1 and Rab35 in a process termed phosphocholination, which stabilizes inactive Rab1 at the LCV membrane (86, 87, 155). The CDP-choline-dependent activity of AnkX is reversed by the Icm/Dot-secreted effector Lem3, which dephosphocholinates Rab1 (88, 155).

The Icm/Dot substrate RidL specifically binds PtdIns(3)P and localizes to the LCV, juxtaposed to where the polar Icm/Dot T4SS connects to the pathogen vacuole membrane [(44); **Figure 4** and **Table 1**]. RidL binds the Vps29 subunit of the retromer coat complex, inhibits retrograde trafficking and thereby promotes intracellular bacterial replication (19, 20). Structural studies revealed that a hydrophobic β-hairpin in the N-terminal domain of RidL interacts with Vps29, thus displacing the Rab7 GAP TBC1D5 [a regulator of retrograde trafficking; (43, 45, 46)].

The Icm/Dot substrate RavZ targets autophagosomes and binds PtdIns(3)P on high-curvature membranes trough a C-terminal domain [(49); **Figure 4** and **Table 1**]. RavZ inhibits autophagy by deconjugating Atg8/LC3 from phosphatidylethanolamine (PE) (48). In contrast to the eukaryotic deconjugating factor Atg4, the cysteine protease RavZ irreversible decouples Atg8 from PE by hydrolyzing the amide bond between the C-terminal glycine and an adjacent aromatic amino acid in Atg8.

The Icm/Dot substrates SetA (110, 120) and LtpM (85) localize to LCVs and endosomes through C-terminal PtdIns(3)P-binding domains (**Figure 4** and **Table 1**). The N-terminal domains of these effectors show similarities with glycosyl transferases, and indeed, the purified enzymes were found to exhibit glycohydrolase and glycosyltransferase activity in vitro, using UDP-glucose as a sugar donor. Intriguingly, PtdIns(3)P activates the glycosyltransferase activity of LtpM (85).

The Icm/Dot substrates LtpD (109) and RavD (114) also localize to the LCV through C-terminal PtdIns(3)Pbinding domains (**Table 1**). LtpD might bind to the inositol monophosphatase IMPA1, which has indeed been detected on isolated LCVs (140). LpnE is a 41 kDa L. pneumophila virulence factor that binds to PtdIns(3)P and the eukaryotic PI



AMP, adenosine monophosphate; DAG, diacylglycerol; ER, endoplasmic reticulum; GAP, GTPase-activating protein; GEF, guanine nucleotide exchange factor; LCV, Legionella-containing vacuole; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphoinositide.

5-phosphatase OCRL (see below) [(156); **Figure 4** and **Table 1**]. The Sel1 repeat-containing LpnE is secreted independently of the Icm/Dot T4SS or the Lsp T2SS and promotes uptake of L. pneumophila by phagocytes and intracellular replication (157, 158). Finally, a recent bioinformatics-based screen identified three novel PtdIns(3)P-binding domains, which are present in at least 14 known Icm/Dot substrates, including LepB and RavZ (95).

#### *L. PNEUMOPHILA* PHOSPHOINOSITIDE PHOSPHATASES, KINASES, AND PHOSPHOLIPASES

Legionella pneumophila produces Icm/Dot-translocated effector proteins, which directly modify PI lipids by acting as PI phosphatases, PI kinases or phospholipases (**Figure 4**). The Icm/Dot substrate SidF localizes to the LCV at early time points of infection (2 h) [(122, 123); **Figure 4** and **Table 1**]. The crystal structure of the N-terminal catalytic domain in complex with its substrate PtdIns(3,4)P<sup>2</sup> revealed a positively charged groove in the catalytic center, similar to other PI phosphatases harboring the "CX5R" motif (123). The 102 kDa effector SidF harbors two predicted C-terminal transmembrane motifs, which anchor the protein to the LCV membrane. SidF specifically hydrolyses in vitro PtdIns(3,4)P<sup>2</sup> and PtdIns(3,4,5)P<sup>3</sup> typically occurring on early phagosomes, and it likely contributes to the production of PtdIns(4)P on LCVs, since vacuoles harboring L. pneumophila 1sidF accumulate lower amounts of the PtdIns(4)P-binding effector SidC. Yet, the 1sidF mutant strain is not impaired for intracellular growth.

The Icm/Dot substrate SidP acts as a PI 3-phosphatase in vitro and converts PtdIns(3,5)P<sup>2</sup> to PtdIns(5)P as well as PtdIns(3)P to PtdIns (**Figure 4** and **Table 1**). However, its PI-phosphatase activity was not assessed in L. pneumophilainfected cells, and a 1sidP mutant strain is not impaired for intracellular growth (129). The crystal structure of SidP from L. longbeachae revealed three distinct domains: a large N-terminal catalytic domain, an appendage domain inserted into the catalytic domain, and a C-terminal α-helical domain. Based largely on biochemical studies, SidF and SidP were postulated to produce PtdIns(4)P and hydrolyze PtdIns(3)P on LCVs, thus contributing to the PI conversion on the pathogen vacuole.

The Icm/Dot substrate LepB is a Rab1 GAP (see above), but also shows PI 4-kinase activity specific for PtdIns(3)P [(96); **Figure 4** and **Table 1**]. The effector might contribute to the production of PtdIns(4)P on LCVs, since pathogen vacuoles harboring L. pneumophila 1lepB accumulate lower amounts of the PtdIns(4)Pbinding effector SidC. LepB was proposed to convert PtdIns(3)P on LCVs into PtdIns(3,4)P, which could be hydrolyzed by SidF to yield PtdIns(4)P (96). Interestingly, the Icm/Dot substrate LegA5 (159), a membrane-associated effector toxic for yeast (110, 160), was recently found to be a wortmannin-insensitive, class III-like PI 3-kinase [(101); **Table 1**]. In fact, LegA5 might be a PI 3-kinase producing PtdIns(3)P on LCVs as a substrate for the PI 4-kinase LepB.

The Icm/Dot substrate LppA is another example of a CX5R motif PI phosphatase hydrolyzing in vitro PtdIns(3,4)P2, PtdIns(4,5)P2, and PtdIns(3,4,5)P<sup>3</sup> to yield PtdIns(4)P [(108); **Table 1**]. While LppA appeared like an ideal candidate to produce PtdIns(4)P on LCVs, live-cell microscopy using GFP-P4C as a PtdIns(4)P probe indicated that LppA does not affect the LCV PI pattern. Instead, LppA is a T4SStranslocated hexakisphosphate inositol phosphatase (phytase), which degrades the micronutrient chelator phytate (indeed produced by amoebae), and thereby promotes the intracellular growth of L. pneumophila. Given that the L. pneumophila genome encodes more than 400 proteins with the CX5R (PI) phosphatase signature (123), other (PI) phosphatases are likely produced by the pathogen.

The Icm/Dot substrates VipD, PlcC, and LpdA are lipases, which possess broad range activity against phospholipids including mono-phosphorylated PIs (**Figure 4** and **Table 1**). VipD was identified as an Icm/Dot substrate that impairs membrane trafficking in yeast (130, 131). The effector hydrolyzes PE as well as phosphatidylcholine (PC) (132) and, intriguingly, binds Rab5 as well as Rab22 and acts as a Rab5-activated phospholipase A<sup>1</sup> (133–135). Accordingly, VipD removes PtdIns(3)P from endosomal membranes and thus might promote the evasion of the endocytic pathway by LCVs (133, 134). Analogously, the Icm/Dot substrate PlcC (alias CegC1) is a metallophospholipase C, which hydrolyzes a broad spectrum of lipids including PC, phosphatidylglycerol (PG), and PtdIns (111). The effector can degrade PtdIns(3)P and likely destabilizes target membranes. Finally, the Icm/Dot substrate LpdA is a phospholipase D that binds to membranes through C-terminal prenylation and hydrolyzes PG, PtdIns and PtdIns(3)P as well as PtdIns(4)P yielding phosphatidic acid (PA) (83). While LpdA does not seem to affect the cellular PI pattern, the phospholipase triggers Golgi fragmentation.

#### SUBVERSION OF HOST PHOSPHOINOSITIDE KINASES AND PHOSPHATASES BY *L. PNEUMOPHILA*

In addition to directly modulating PI lipids, L. pneumophila effectors also subvert the host cell PI pattern indirectly by targeting eukaryotic PI phosphatases and kinases (**Figure 4**). The PtdIns(3)P-binding virulence factor LpnE binds mammalian OCRL (Oculocerebrorenal syndrome of Lowe) and its Dictyostelium homolog Dd5P4 (D. discoideum 5-phosphatase 4) via their N-terminal domains (156). The interaction of LpnE with OCRL was recently confirmed by size exclusion chromatography and supported by the crystal structure of the bacterial protein (161). OCRL and Dd5P4 are PI 5-phosphatases, which hydrolyse PtdIns(4,5)P<sup>2</sup> and PtdIns(3,4,5)P<sup>3</sup> to yield PtdIns(4)P and PtdIns(3,4)P2, respectively (162, 163). Dd5P4 is likely catalytically active on LCVs and increases the PtdIns(4)P available for binding by effectors such as SidC or SidM (156). Consequently, LpnE might increase the concentration of PtdIns(4)P on LCVs by recruiting OCRL/Dd5P4, and thereby promote PI conversion. L. pneumophila grows more efficiently in D. discoideum lacking Dd5P4, and thus, the pleiotropic PI 5-phosphatase restricts intracellular bacterial growth. Mechanistic details of this process are not known, but Dd5P4 modulates the recruitment of calnexin, Rab1 and retromer components to LCVs, which might account for growth restriction (156, 164).

The Icm/Dot substrates RalF and SidM possibly contribute indirectly to the modulation of the LCV PI pattern through the recruitment and activation of small host GTPases. RalF is an Arf1 GEF and activates the small GTPase on the LCV [(22, 112); **Figure 4** and **Table 1**]. RalF harbors a C-terminal globular "capping" domain, which regulates GEF activity by auto-inhibition (77). Activated Arf1 recruits PI 4-kinase IIIβ (PI4KIIIβ) to the trans Golgi network (165), and hence, RalF might indirectly increase the PtdIns(4)P concentration on LCVs. Indeed, the depletion by RNA interference of PI4KIIIβ, but not PI4KIIIα or PI4KIIα decreases the amount of the PtdIns(4)P-binding effector SidC on LCVs, suggesting that in absence of PI4KIIIβ the level of PtdIns(4)P is reduced (84). Analogously, SidM recruits and activates Rab1 on LCVs (see above). Activated Rab1 (166) as well as Arf1 (167) recruit OCRL to endosomal membranes. Accordingly, SidM might not only bind to PtdIns(4)P, but also indirectly contribute to an increase of this PI on LCV membranes.

The Icm/Dot substrates LpdA and LecE localize to LCVs and might also indirectly modulate the LCV PI pattern by promoting DAG biosynthesis [(99); **Table 1**]. LpdA is a phospholipase D, which hydrolyzes PC to yield PA (see above). LecE enhances the activity of the eukaryotic PA phosphatase Pah1, which dephosphorylates PA yielding DAG. The second messenger DAG recruits protein kinase D (PKD) and its activator protein kinase C (PKC) to membranes. Activated PKD then interacts with PI4KIIIβ, thereby possibly also contributing to an increase in PtdIns(4)P on LCVs (99).

# CONCLUSIONS AND OUTLOOK

Legionella pneumophila replicates intracellularly in phagocytes within an LCV, a complex compartment tightly associated with the ER. The nascent LCV undergoes a PI conversion from PtdIns(3)P to PtdIns(4)P, and thereby is rerouted from the bactericidal endocytic to the replication-permissive secretory pathway. To modulate the PI pattern in infected cells, L. pneumophila (i) recruits PI-decorated vesicles, (ii) produces effectors acting as PI interactors, phosphatases, kinases or phospholipases, or (iii) subverts host PI-metabolizing enzymes. To this end, at least 21 T4SS-translocated effector proteins have been shown to target the host PI metabolism (**Table 1**). Intriguingly, a number of these effectors harbor 2–3 different functional domains and link PI signaling to other pivotal cellular pathways, e.g., SidC (PI interactor, ubiquitin ligase), SidM (PI interactor, Rab1 GEF, Rab1 AMPylase), LepB (PI interactor, PI 4-kinase, Rab1 GAP), SetA and LtpM (PI interactor, glycosyltransferase), and VipD (Rab5 interactor, phospholipase). LCV formation and the contribution of PI lipids to this process are incompletely understood. Among the more than 300 T4SS-translocated effector proteins of L. pneumophila only about 50 have been thoroughly investigated. Future studies will focus on the structural, molecular and cellular characterization of novel effectors implicated in host cell PI pattern subversion, as well as on the spatiotemporal regulation of effector translocation and function.

# AUTHOR CONTRIBUTIONS

ALS and HH wrote the manuscript.

#### REFERENCES


#### FUNDING

Research in the laboratory of HH was supported by the Swiss National Science Foundation (SNF; 31003A\_153200, 31003A\_175557), the University of Zürich, the Novartis Foundation for Medical-Biological Research, and the OPO foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


membranes using artificial membrane tethering. Biochem J. (2017) 474:1259– 72. doi: 10.1042/BCJ20170015


factor from Legionella pneumophila, reveals the presence of a cap over the active site. J Biol Chem. (2005) 280:1392–400. doi: 10.1074/jbc.M410820200


**Conflict of Interest:** 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 © 2020 Swart and Hilbi. 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.

,

# Enhanced Expression of CD47 Is Associated With Off-Target Resistance to Tyrosine Kinase Inhibitor Gefitinib in NSCLC

#### Edited by:

Esther M. Lafuente, Complutense University of Madrid, Spain

#### Reviewed by:

Soldano Ferrone, Massachusetts General Hospital and Harvard Medical School, United States Patricia A. Possik, Instituto Nacional de Câncer (INCA), Brazil

> \*Correspondence: Jessica Dal Col

> > jdalcol@unisa.it

†These authors share last authorship

#### Specialty section:

This article was submitted to Cancer Immunity and Immunotherapy, a section of the journal Frontiers in Immunology

> Received: 01 October 2019 Accepted: 23 December 2019 Published: 31 January 2020

#### Citation:

Nigro A, Ricciardi L, Salvato I, Sabbatino F, Vitale M, Crescenzi MA, Montico B, Triggiani M, Pepe S, Stellato C, Casolaro V and Dal Col J (2020) Enhanced Expression of CD47 Is Associated With Off-Target Resistance to Tyrosine Kinase Inhibitor Gefitinib in NSCLC. Front. Immunol. 10:3135. doi: 10.3389/fimmu.2019.03135 Annunziata Nigro<sup>1</sup> , Luca Ricciardi <sup>1</sup> , Ilaria Salvato<sup>1</sup> , Francesco Sabbatino<sup>1</sup> , Monica Vitale<sup>1</sup> Maria Assunta Crescenzi <sup>1</sup> , Barbara Montico<sup>2</sup> , Massimo Triggiani <sup>1</sup> , Stefano Pepe<sup>1</sup> , Cristiana Stellato<sup>1</sup> , Vincenzo Casolaro1† and Jessica Dal Col <sup>1</sup> \* †

<sup>1</sup> Department of Medicine, Surgery and Dentistry "Scuola Medica Salernitana," University of Salerno, Baronissi, Italy, 2 Immunopathology and Cancer Biomarkers, Translational Unit, Centro di Riferimento Oncologico di Aviano (CRO), IRCCS, Aviano, Italy

Mutual interactions between cancer cells and the tumor microenvironment importantly contribute to the development of tyrosine kinase inhibitor (TKI) resistance in patients affected by EGFR-mutant NSCLC. In particular, immune recognition-associated proteins with impact on tumor cell clearance through phagocytosis, such as CD47 and calreticulin, could contribute to adaptive resistance and immune escape. Preclinical studies targeting the anti-phagocytic CD47 molecule showed promising results in different cancer types including lung cancer, but no data are available on its role in patients acquiring resistance to EGFR TKI treatment. We analyzed the functional contribution of CD47 and calreticulin to immune surveillance and evasion in a panel of NSCLC cell lines carrying sensitizing or resistant mutations in the EGFR gene, following treatment with the TKI gefitinib and after in vitro development of gefitinib resistance. While CD47 and calreticulin protein levels were markedly variable in both EGFR-mutant and wild-type cell lines, analysis of NSCLC transcriptomic dataset revealed selective overexpression of CD47 in patients carrying EGFR mutations. EGFR inhibition significantly reduced CD47 expression on the surface of pre-apoptotic cells, favoring more efficient engulfment of cancer cells by monocyte-derived dendritic cells. This was not necessarily associated with augmented surface exposure of calreticulin or other molecular markers of immunogenic cell death. Moreover, CD47 expression became up-regulated following in vitro drug resistance development, and blocking of this protein by a specific monoclonal antibody increased the clearance of EGFR-TKI resistant cells by phagocytes. Our study supports CD47 neutralization by specific monoclonal antibody as a promising immunotherapeutic option for naïve and resistant EGFR-mutant NSCLCs.

Keywords: CD47, non-small cell lung cancer, tyrosine kinase inhibitor resistance, phagocytosis checkpoint, innate immunity, cancer immune surveillance

# INTRODUCTION

Non-small-cell lung cancer (NSCLC) accounts for 85% of all diagnosed cases of lung cancers, with the diagnosis taking place often when the disease is at locally advanced or metastatic stage (1, 2). The discovery of oncogenic driver mutations in almost two-thirds of NSCLCs and the introduction of targeted therapies undeniably improved the outcome in patients with oncogene-addicted adenocarcinoma. However, clinical response to treatment is generally temporary and incomplete (3). Recent studies on the molecular determinants of resistance to tyrosine kinase inhibitors (TKIs) identified two major mechanisms either affecting targeted oncogenes ("on-target" resistance) or molecules residing downstream or belonging to parallel and other pathways ("off-target" resistance) (4). In addition, dynamic interactions between tumor cells and the surrounding microenvironment critically contribute to modifying the response to TKI therapy and influence the development of resistant phenotypes (5–7). It has been demonstrated that in epidermal growth factor receptor (EGFR) driven lung tumors, anti-tumor immunity is inhibited by activation of the Programmed death 1 (PD-1)/Programmed death ligand 1 (PD-L1) pathway, leading to suppression of effector T cell function and increased levels of pro-inflammatory cytokines (8). As recently shown, PD-L1 expression in tumor cells adversely affects EGFR-TKI efficacy, especially in NSCLC patients with de novo resistance (9). Furthermore, the secretion from stromal cells of paracrine factors such as interleukin-6 (IL-6), transforming growth factor-β (TGF-β), and hepatocyte growth factor (HGF) promotes MAP-kinase activation and further supports EGFR TKI resistance development by eluding EGFR pathway inhibition (10).

Immune checkpoint inhibitors (ICIs) targeting the PD-L1/PD-1 axis have been recently approved for the treatment of EGFR- and Anaplastic lymphoma kinase (ALK)-positive NSCL tumors after failure of appropriate targeted therapy (11, 12). While the association of EGFR mutations with high PD-L1 expression suggests the potential efficacy for PD-L1 inhibitors in this setting, treatment with ICIs showed limited efficacy in different cohorts of patients previously treated with an EGFR TKI (13–16) and the outcome did not show correlation with the EGFR mutation subtype. The poor response to ICIs in EGFRmutated, TKI-resistant patients suggests that other immuneescape mechanisms are at stake in this clinical phenotype.

No studies to date have examined the effects of EGFR TKIs on immune recognition-associated molecules, such as CD47 and calreticulin (CRT), recently found to affect innate immune surveillance. CD47, originally identified as integrin-associated protein (IAP), is a cell-surface immunoglobulin-like molecule that serves as a "don't eat me" signal via its interaction with signal regulatory protein alpha (SIRPα) on phagocytes (17, 18). Loss of CD47 is permissive to homeostatic phagocytosis of aged or damaged cells (19, 20). While CD47 is ubiquitously expressed at low levels on normal cells, multiple hematologic and solid tumors have been found to express higher levels of CD47 compared to their non-transformed counterparts (21–24). Enhanced expression of CD47 has also been reported in primary NSCLC tumors and cell lines (25). Up-regulation of CD47 expression in human cancers negatively regulates anti-tumor immunity through suppression of phagocytosis, and it has been associated with tumor growth and dissemination (18, 25–28). Conversely, CRT is a highly conserved endoplasmic reticulum chaperone protein, which, upon translocation from the endoplasmic reticulum to the cell surface, provides an "eat-me" signal and facilitates capture by macrophages and dendritic cell precursors of cancer cells undergoing immunogenic cell death (ICD) or other stress conditions (29, 30). Fucikova et al. demonstrated that the expression of CRT in NSCLC correlates with increased accumulation of antitumor immune cells and favorable prognosis (31). Given the emerging critical roles of CD47 and CRT in NSCLC adenocarcinomas, in the present study, we assessed whether the EGFR TKI gefitinib modulates their expression in different EGFR-mutated NSCLCs. Furthermore, we analyzed in these cells the functional contribution of these proteins to immune surveillance, while their potential role in surveillance evasion was tested in subsets of NSCLC cell lines rendered TKI resistant in vitro.

# MATERIALS AND METHODS

# Cell Lines

Human NSCLC cell lines NCI-H1975, NCI-H1299, NCI-H1437, and NCI-H1573 were purchased from ATCC (American Type Culture Collection, Manassas, Virginia). PC9 and HCC827 cell lines were obtained from Cell Biology and Biotherapy Unit, Istituto Nazionale Tumori di Napoli, IRCCS "G. Pascale." The EGFR TKI-resistant cell lines, PC9GR and HCC827GR, were generated by culturing the respective parental cell lines in the presence of increasing concentrations of TKI Gefitinib (from 0.05 to 0.5µM and from 0.1 to 1µM, respectively) for 8 weeks, to reach a concentration 10 times higher than the initial IC50. Cell lines were cultured in RPMI-1640 medium (Gibco) supplemented with 10% fetal bovine serum (Gibco), 1% L-glutamine (2 mM, Lonza), 1% streptomycin-penicillin (EuroClone) at 37◦C in a 5% CO<sup>2</sup> humidified atmosphere. Short tandem repeat (STR) loci for cell lines authentication were evaluated for all cell lines, by using the GenePrint10- System (Promega).

# Antibodies and Reagents

Anti-CD47-APC, anti-CD47-FITC, anti-CD11c-APC-Vio770, anti-CD11-PE, anti-CD14-FITC, anti-CD80-PE, anti CD83-PE, anti-CD86-PE, anti-HLA-DR-FITC, anti-HLA-ABC-FITC, anti-Hsp70-FITC, and 7-Amino-Actinomycin D (7-AAD) fluorescent DNA dye were purchased from Miltenyi. All monoclonal antibodies (mAbs) used in flow cytometry experiments were used at 1:200 titer unless otherwise specified. Anti-calreticulin-FITC mAb (EPR3924; 1:50) and anti-GAPDH were from Abcam. Anti-mouse/human/rat CD47 mAb or mouse IgG isotype control were purchased from Bio X Cell. Anti-phospho-EGFR (Y1608), anti-EGFR, and anti-phospho-Akt (S473) were from Cell Signaling Technology. Secondary antibodies anti-rabbit IgG-HRP or anti-mouse IgG+IgM+IgA-HRP were from Bethyl Laboratories. Gefitinib was purchased from Selleckchem, and its IC<sup>50</sup> was determined for each cell line using Cell-Counting Kt-8 (Dojindo Laboratories) according to the manufacturer's instructions. Gefitinib IC<sup>50</sup> for each cell line, reported in **Table 1**, was used for all the experiments. Recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF) was purchased from Miltenyi, and interferon (IFN)-α (IntronA) from SP Europe.

#### Protein Extraction and Western Blot

For total protein extraction, cells were directly lysed in a buffer containing 50 mM Tris/HCl at pH 7.5, 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 25 mM NaF, 25 mM β-glycerolphosphate, 0.1 mM Na3VO4, 0.1 mM PMSF, 0.2% Triton X-100, 0.3% Nonidet P40, and a cocktail of protease inhibitors (100 X, EuroClone). After incubation on ice for 30 min, the lysate was centrifuged at 13,000 rpm for 15 min at 4◦C. A total of 15 µg of protein per well was separated by 4–15% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. Membranes were then blocked with 5% milk-TBST buffer (TBS plus 0.1% Tween-20) for 1 h at room temperature and incubated with primary antibodies overnight at 4◦C, washed with TBST buffer three times, and incubated with corresponding secondary antibodies at room temperature for 45 min. Signals were detected by the "Pierce ECL Western Blotting Substrate" method (Thermo Fisher Scientific) and analyzed by the ImageLab software, using the Chemidoc image acquisition and analysis tool (BioRad).

#### Flow Cytometry Analysis

To measure the cell surface expression of CD47 and CRT, 2.5 × 10<sup>5</sup> cells were seeded in 6-well plates and treated with gefitinib at their specific IC<sup>50</sup> (see **Table 1**) or corresponding proportions of DMSO solvent for 48 h. Cells were stained with primary antibodies for 20 min at room temperature (RT). Seven-AAD was added to exclude non-viable cells from the analyses. For intracellular labeling with anti-CD47-FITC mAb, cells were fixed with 4% paraformaldehyde (PFA) and permeabilized with 100% cold methanol. Cells were acquired using a BD FACSVerseTM flow cytometer (BD Biosciences) and the data were analyzed using the BD FACSuiteTM software.

#### Differentiation of Dendritic Cells From CD14<sup>+</sup> Monocytes Isolated From Peripheral Blood

Human peripheral blood mononuclear cells (PBMCs) were isolated from whole blood donated by healthy volunteers using Ficoll HyPaque (GE Healthcare). Monocytes were isolated from PBMCs by positive selection onto MACS LS columns using CD14 MicroBeads (Miltenyi) following the manufacturer's protocol. The resulting cell suspensions, containing at least 85% monocytes, were seeded in 6-well plates at a concentration of 2.0 × 10<sup>6</sup> cells/ml and cultured for 3 days at 37◦C in a 5% CO<sup>2</sup> humidified atmosphere in CellGenixTM GMP DC medium (CellGenix GmbH) supplemented with 1% streptomycinpenicillin (EuroClone), and in the presence of 500 ng/ml GM-CSF and 10,000 U/ml IFN-α (32). Differentiation of dendritic cells was confirmed by immune phenotype analysis for the established dendritic cell markers: CD80, CD83, CD86, CD11c, and HLA-DR.

### In vitro Phagocytosis Assay

Dendritic cells were plated in 24-well plates (10<sup>5</sup> cells/well). After 48 h, lung tumor cells treated with gefitinib at their specific IC<sup>50</sup> (see **Table 1**) or DMSO carrier were labeled with DiO cell-labeling solution (Vybrant Cell-Labeling Solution, Molecular Probes) and added to dendritic cells at a 1:1 ratio. Where indicated, tumor cells were incubated with antimouse/human/rat CD47 mAb (10µg/ml, Bio X Cell) or mouse IgG isotype control (10µg/ml, Bio X Cell) prior to culture with dendritic cells. Following 2.5 h co-culture at 37◦C, cells were washed twice with PBS and then labeled with anti-CD11c

TABLE 1 | Histological and mutational characteristics of the six NSCLC cell lines included in the study.


The indicated gefitinib IC<sup>50</sup> values were measured experimentally.

mAbs (1:200, Miltenyi). Phagocytosis was determined by flow cytometry detection of dendritic cells double positive for CD11c and DiO cell-labeling solution.

## Annexin V/PI Apoptosis Assay

Tumor cells were seeded into 6-well plates (2.5 × 10<sup>5</sup> cells per well) and cultured (72 h) in the absence or presence of gefitinib at their specific IC<sup>50</sup> (see **Table 1**). Apoptosis was assessed by cytofluorimetric analysis using FITC-Annexin V Apoptosis Detection Kit (Dojindo Laboratories) according to the manufacturer's instructions.

#### Analysis of CD47 and CRT mRNA in Transcriptomic Databases of NSCL Adenocarcinomas

CD47 and CRT mRNA expression was determined in a publicly available microarray database collecting frozen primary NSCLC specimens with different oncogenic driver mutations (GEO accession number GSE31210) (33, 34). Only the Jetset best probe sets were considered (35), 226016\_at for CD47 and 214315\_x\_at for CRT.

#### Statistical Analysis

Protein expression levels of CD47 and CRT in the different cell lines were compared using ANOVA test with Fisher's post hoc multiple comparison analysis, whereas mRNA expression levels in different groups of patients were compared using the nonparametric Kruskal-Wallis test. The effect of gefitinib on CD47 and CRT surface expression and on phagocytosis was determined by two-tail paired Student's t-test, based on a minimum of three experiments/different donors. The effect of a CD47-neutralizing antibody on phagocytosis was determined using ANOVA test with Fisher's post hoc multiple comparison analysis. Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software, Inc. La Jolla, CA, USA).

# RESULTS

# Treatment With Gefitinib Modulates CD47 and CRT Surface Expression in NSCLC Cell Lines

The basal expression levels of CD47 and surface-exposed CRT (ecto-CRT) were investigated in a panel of six NSCLC cell lines harboring different mutations on EGFR as well as on other genes connected or not to the EGFR pathway (**Table 1**). Flow cytometry analysis showed that expression levels of surface CD47 and ecto-CRT are markedly variable among the different cell lines and independent of the type of EGFR mutations (**Figures 1A,B**). Retrospective analysis of a published dataset of microarray expression data from frozen primary lung adenocarcinoma specimens allowed us to investigate CD47 and CRT mRNA levels in patients (N = 226) with different oncogenic driver mutations (GEO accession number GSE31210) (33, 34). As shown in **Figure 1C**, patients with EGFR mutations expressed higher levels of CD47 mRNA compared to patients with KRAS mutations, ALK-fusion, or no mutations in these three driver oncogenes (**Figure 1C**, p ≤ 0.03 Kruskal-Wallis test). In contrast, in the same dataset, CRT mRNA expression showed no significant differences among the four groups of patients (**Figure 1D**).

Next, we assessed the ability of gefitinib to modulate CD47 and CRT surface expression in the cell lines tested. Cells were cultured for 48 h with gefitinib concentrations corresponding to IC<sup>50</sup> identified for each cell line (listed in **Table 1**) or an equivalent DMSO amount. Immunoblotting analysis of phospho-EGFR confirmed the inhibition of EGFR activating phosphorylation by gefitinib (**Figure S1**). Flow cytometry analysis of viable cells showed that gefitinib significantly down-regulated CD47 surface expression in all cell lines, except for HCC827, one of the two cell lines harboring the EGFR exon 19 deletion (del19 mutation) (**Figures 2A,C**). Concomitantly, gefitinib treatment promoted a significant increase of ecto-CRT expression in all cell lines carrying wild-type EGFR and in H1975, whereas no significant difference was induced in HCC827 and a slow, but consistent ecto-CRT reduction was detected in PC9 cells (**Figures 2B,D**).

#### Gefitinib-Induced Modulation of CD47 and CRT Is Independent on Immunogenic Cell Death Promotion

Previous work has shown that the anti-EGFR mAb cetuximab induces immunogenic cell death (ICD) in colon cancer cells by triggering endoplasmic reticulum stress response and CRT translocation, depending on the mutational status of the EGFR signaling pathway (36). Therefore, we next investigated if the effects of gefitinib on CD47 and ecto-CRT expression were associated to the induction of ICD in NSCLC cells. The markers of ICD evaluated together with the ecto-CRT expression were the induction of apoptosis, the surface exposure of CRT heatshock protein 70 (HSP70), and the release of HSP70 and of high mobility group box-1 (HMGB1) (37). Consistently with the spatiotemporal sequence of ICD markers, surface exposure of proteins was measured in pre-apoptotic cells after 48 h of treatment, whereas the release of HSP70 and HMGB1 was measured 72 h after treatment. As shown in **Figures 3A,B**, treatment with gefitinib significantly increased the frequency of cells undergoing apoptosis relative to DMSO-treated cells. Moreover, gefitinib increased the surface expression of HSP70 by ≥50% in all cell lines except in cell lines carrying EGFR exon 19 deletion, PC9, and HCC827 (**Figures 3C,D**), in agreement with findings on ecto-CRT expression in these two cell lines (**Figures 2B,D**). In contrast, as shown in **Figure 3E**, gefitinib increased HSP70 release over control-treated cells only in the PC9 cell line. Lastly, gefitinib-induced apoptosis was not associated with HMGB1 release, which was not detected in supernatants (data not shown).

Taken together, these results, summarized in **Figure 3F**, indicate that while gefitinib-promoted apoptosis is associated with several ICD traits, these are not all detectable in each cell line. Therefore, according to recently published findings (38), the results indicate that EGFR inhibition by gefitinib does not induce ICD in the NSCLC cell lines tested. Moreover, even the concomitant activation of the type I interferon pathway, a well-known important mediator of ICD (39), was insufficient

to induce ICD in these experiments. In fact, the addition of IFN-α in these cultures, while further increasing the frequencies of apoptotic cells, did not augment the effects of gefitinib on CD47 and ecto-CRT expression (**Figures S2A**–**C**) and was also insufficient to induce HMGB1 release (data not shown).

#### Gefitinib-Induced CD47 Down-Regulation Promotes Phagocytosis of Tumor Cells by IFN-Conditioned Dendritic Cells

Since the balance between CD47 and ecto-CRT expression determines the susceptibility of cancer cells to engulfment by phagocytes, based on our results (**Figures 2C,D**), we hypothesized that gefitinib treatment could promote phagocytosis of NSCLC cells by CD47 down-regulation. To test our hypothesis, we performed phagocytosis assay using monocyte-derived dendritic cells from healthy donors and, as target cells, we selected PC9, HCC827, and H1975 cell lines cultured in the presence of gefitinib or DMSO for 48 h. These cell lines were included because of their differential mutational status (EGFR del19 mutation and gefitinib resistant T790M mutation, respectively) and of their differential CD47/ecto-CRT balance following treatment with gefitinib. Cell incubation time was kept at 48 h, since at this time point CD47 down-regulation and ecto-CRT increase were detectable without high levels of apoptosis. Gefitinib-treated and untreated cells were then labeled with the fluorescent tracer DiO (see Methods) and then cocultured for 2 h with monocyte-derived dendritic cells (29, 40). As shown in **Figures 4A,B**, PC9 cells treated with gefitinib were

engulfed more efficiently than untreated controls, in line with the significant down-regulation of CD47 induced by gefitinib in these cells (**Figures 2A,C**). Conversely, gefitinib treatment did not significantly affect HCC827 and H1975 cell uptake by dendritic cells (**Figures 4C–F**). These findings indicate that the marked ecto-CRT up-regulation observed in gefitinib-treated H1975 cells (**Figures 2B,D**) is not sufficient to increase cell susceptibility to phagocytosis in the absence of a substantial decrease of CD47 (**Figures 2A,C**).

Indeed, the addition of a CD47-specific blocking mAb to gefitinib-treated or untreated HCC827 and H1975 cells significantly increased the percentage of DiO-positive dendritic cells, while addition of an isotype control did not affect tumor cell engulfment (**Figures 5A,B**). This indicates that down-regulation of CD47 in gefitinib-treated NSCLC cells is per se sufficient to enhance antitumor immunity by improving cell recognition and engulfment by immune phagocytes.

### CD47 Expression Increases in NSCLC Cells Acquiring Gefitinib Resistance in vitro

The development of TKI resistance is the main limiting factor of this otherwise effective target therapy in patients affected by EGFR-mutant lung cancer. Given the increasing evidence of the involvement of tumor microenvironment in remodeling cancer cell responsiveness to TKI therapy, we investigated the expression of CD47 and ecto-CRT in PC9 and HCC827 cell lines after the acquisition of gefitinib resistance in vitro. Gefitinib-resistant PC9GR and HCC827GR cells were generated upon exposure to gradually increasing concentrations of the drug. Flow cytometry analyses showed a significant increase of surface CD47 expression in both PC9GR and

HCC827GR relative to the parental cell lines (**Figure 6A**). Conversely, the development of gefitinib resistance was not associated with higher expression of CRT on the plasma membrane (**Figure 6B**). As expected, treatment of PC9GR and HCC827GR cells with gefitinib for 48 h did not affect the expression levels of CD47 (**Figure 6C**) and failed to promote tumor cell phagocytosis by dendritic cells (**Figure 6D**). In contrast, blockade of the CD47/SIRPα axis by the addition of a CD47-specific mAb significantly increased PC9GR phagocytosis by dendritic cells in the presence or absence of gefitinib (**Figure 6E**).

#### DISCUSSION

Our study presents compelling evidence supporting the modulation of CD47 as a novel and important determinant of antitumor immunity in the response to EGFR TKI target therapy. In particular, our findings of overexpressed CD47 in transcriptomic analysis of patients with EGFR-mutated NSCLC, as well as in tumor cell lines acquiring EGFR TKI resistance, identify CD47 as target to be further explored for the immunotherapy treatment of naïve and resistant EGFR-mutant NSCLCs.

The development of resistance occurs consistently in patients affected by EGFR-mutant NSCLCs following first-line treatment with gefitinib and other TKIs. Secondary mutations in the EGFR gene, including T790M, account for more than 50% of resistant cases, followed by the MET gene amplification and the activation of other parallel pathways (41–43). In addition, histological transformation to SCLC may occur in 3–10% of cases (11). Recently, the European Medicines Agency extended the approval of immunotherapy with ICIs to EGFR- and ALKpositive tumors after failure of appropriate targeted therapy.

However, the response of these patients to ICI therapy remains poor (13–15), indicating that blocking the PD1/PD-L1 axis is not sufficient to obtain an adequate antitumor immune response.

DiO tracer relative to total dendritic cells (\*p < 0.05, n.s., not significant, paired two-tailed Student's t-test).

Several studies documented that first-generation EGFR TKIs improve the interaction between natural killer (NK) and tumor cells favoring immune-mediated cytotoxicity, indicating

that EGFR inhibitors can enhance innate tumor immune surveillance (44–46). Despite that, the potential impact of this therapy on immune recognition and elimination of cancer cells by phagocytes remains underexplored. Disruption of the CD47/SIRPα axis with specific mAbs may promote cancer cell elimination by macrophages, and it is a potential immunotherapeutic strategy recently described for different cancers, including SCLC and NSCLC (24, 26, 47–49). Our analysis of large transcriptomic dataset identified higher expression levels of CD47 mRNA, but not of CRT mRNA, in primary lung adenocarcinomas with EGFR mutations as compared to those with different oncogenic mutations. Functional relevance of CD47 overexpression is indicated by our data; although gefitinib down-regulates CD47 and increases ecto-CRT in almost all cell lines tested, it is the decrement of CD47 that results in enhanced phagocytosis of cancer cells by dendritic cells. These results suggest that gefitinib could enhance antitumor immunity by improving lung cancer cell recognition and engulfment by immune phagocytes chiefly through CD47 down-regulation, thereby inhibiting tumor cell viability not only through TK-dependent mechanisms but also by enhancing innate anticancer immune responses.

At variance with previous works using anti-EGFR mAbs 7A7 F(ab′ )<sup>2</sup> in mice (50) and cetuximab in humans (36), in our experimental system, gefitinib does not induce ICD, as it fails to induce established ICD markers (e.g. HMGB1 secretion) in the cells where it promotes widespread apoptosis. Moreover, our results show that the activation of type I IFN pathway is not sufficient to improve the immunogenic features of gefitinib-induced apoptosis. On the other hand, the induction of HSP70 surface exposure and its release by gefitinib-treated tumor cells point to additional immune-modulatory effects of the EGFR TKI.

Our findings that gefitinib-induced down-regulation of CD47 promotes cancer cell phagocytosis in responsive cells and that establishment of gefitinib resistance reverts this response indicate a novel immune mechanism for EGFR TKI therapy, warranting further validation in preclinical and clinical studies. The relevance of CD47 as target in this setting is also underscored by the data obtained in PC9GR cells and in H1975 cells, which harbor a T790M mutation, showing that blocking the CD47/SIRP1α axis promotes cancer cell elimination by dendritic cells also in NSCLC cell lines in which gefitinib has minimal or no effect on CD47 expression. On these bases, evaluation of CD47 expression in patients with EGFR-mutant NSCLCs becoming resistant to target therapy may justify the subsequent adoption of CD47-targeting immunotherapeutic options. In support of this strategy, the administration of CD47-specific mAb inhibited in vivo the growth of xenografted tumors developed from patient-derived lung cancer cells or cancer stem cells by recruiting macrophages into the tumor microenvironment (49). Moreover, the administration of CD47-blocking antibodies or targeted inactivation of the Cd47 gene in humanized mouse models markedly inhibited SCLC tumor growth (24). These evidences could support immunotherapy with CD47-blocking agents as a viable option also in patients undergoing histopathological transformation after EGFR-TKI therapy.

#### CONCLUSIONS

Several studies conducted over the last few years support an important role of the tumor microenvironment in mutant NSCL adenocarcinoma, especially in TKI resistance development and its dynamic remodeling by target therapy. Increasing awareness of an essential contribution by innate immunity in tumor immune surveillance and in metastasis control is advocated whereby new immunotherapeutic options are becoming available for management of relapsed EGFR mutant patients. In particular, promotion of tumor

flow cytometric histograms (left) and mean (± SD, N = 3–5) fold changes of treatment-resistant over sensitive cells (right). (C) Representative flow cytometric histogram plots (left) and mean (± SD) fold changes (right) of surface CD47 levels in resistant cell lines treated with DMSO (CTRL) or gefitinib (GEF) as indicated.

(Continued)

FIGURE 6 | Acquisition of resistance to gefitinib abolished drug-induced CD47 down-regulation in PC9GR (\*p < 0.05, \*\*p < 0.01, n.s., not significant, paired two-tailed Student's t-test). (D) Mean ± SD (N = 3 independent healthy donors) of phagocytic activity of monocyte-derived dendritic cells against PC9GR cells in the absence or presence of gefitinib treatment, performed at 4◦C as control and at 37◦C. Histograms represent the percentages of positive cells for both CD11c and DiO tracer relative to total dendritic cells (paired two-tailed Student's t-test. n.s., not significant). (E) Dendritic cells were co-cultured with gefitinib-treated, DiO tracer-labeled PC9GR cells in the presence of IgG isotype control or anti-CD47 mAb. Shown is the mean ± SD (N = 3 independent healthy donors) percent change of CD11c+/DiO<sup>+</sup> tracer double positive dendritic cells, relative to dendritic cells co-cultured with DMSO-treated tumor cells (##p < 0.01, ANOVA with Fisher's post hoc analysis).

cells elimination by phagocytosis could be successfully achieved through the administration of anti-CD47 mAbs. The effectiveness of CD47 blockade also following gefitinib treatment, as shown in our experiments in vitro, supports the development of therapeutic strategies in which anti-CD47 immunotherapy and target therapy may be combined to minimize the development of resistant clones responsible for tumor relapse.

#### DATA AVAILABILITY STATEMENT

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.

# AUTHOR CONTRIBUTIONS

AN performed most of the experiments, analyzed the data, and drafted the figures and parts of the manuscript. LR provided support with the statistical analysis and performed in silico analyses. IS, MV, and MC performed some of the experiments and analyzed the data. BM helped set up phagocytosis assays. FS, SP, and MT critically revised the manuscript. CS made significant contributions to data interpretation, manuscript writing, and revision. VC conceived and supervised the study and critically revised the manuscript. JDC conceived the study, developed the experimental design, analyzed and interpreted the data, and wrote the manuscript. All authors contributed

#### REFERENCES


toward drafting and revising the paper and approved the final manuscript.

#### FUNDING

The work was supported by Fondazione con il Sud (Grant: Brains2South 2015 PDR-0224 to JDC), POR Campania FESR 2007–2013–RETE DELLE BIOTECNOLOGIE IN CAMPANIA—Project Terapie Innovative di Malattie Infiammatorie croniche, metaboliche, Neoplastiche e Geriatriche—TIMING (to CS), and FARB 2017/2018 (to CS and VC), Regione Campania, POR FESR 2014/20 RarePlatNet Project (Az. 1.2, CUP B63D18000380007) (to VC).

# ACKNOWLEDGMENTS

The authors thank Prof. Normanno's lab (Cell Biology and Biotherapy Unit, Istituto Nazionale Tumori di Napoli, IRCCS G. Pascale) for the kind gift of PC9 and HCC827 cell lines and Dr. G. Giurato (Department of Medicine, Surgery and Dentistry Scuola Medica Salernitana, University of Salerno) for kind assistance in analysis of the microarray dataset.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.03135/full#supplementary-material


receptor activation with a specific antibody. J Immunol. (2011) 187:4954–66. doi: 10.4049/jimmunol.1003477

**Conflict of Interest:** 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 © 2020 Nigro, Ricciardi, Salvato, Sabbatino, Vitale, Crescenzi, Montico, Triggiani, Pepe, Stellato, Casolaro and Dal Col. 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.

# CEACAM3—A Prim(at)e Invention for Opsonin-Independent Phagocytosis of Bacteria

Patrizia Bonsignore<sup>1</sup> , Johannes W. P. Kuiper <sup>1</sup> , Jonas Adrian<sup>1</sup> , Griseldis Goob<sup>1</sup> and Christof R. Hauck 1,2 \*

<sup>1</sup> Lehrstuhl Zellbiologie, Fachbereich Biologie, Universität Konstanz, Konstanz, Germany, <sup>2</sup> Konstanz Research School Chemical Biology, Universität Konstanz, Konstanz, Germany

Phagocytosis is one of the key innate defense mechanisms executed by specialized cells in multicellular animals. Recent evidence suggests that a particular phagocytic receptor expressed by human polymorphonuclear granulocytes, the carcinoembryonic antigen-related cell adhesion molecule 3 (CEACAM3), is one of the fastest-evolving human proteins. In this focused review, we will try to resolve the conundrum why a conserved process such as phagocytosis is conducted by a rapidly changing receptor. Therefore, we will first summarize the biochemical and structural details of this immunoglobulin-related glycoprotein in the context of the human CEACAM family. The function of CEACAM3 for the efficient, opsonin-independent detection and phagocytosis of highly specialized, host-restricted bacteria will be further elaborated. Taking into account the decisive role of CEACAM3 in the interaction with pathogenic bacteria, we will discuss the evolutionary trajectory of the CEACAM3 gene within the primate lineage and highlight the consequences of CEACAM3 polymorphisms in human populations. From a synopsis of these studies, CEACAM3 emerges as an important component of human innate immunity and a prominent example of a dedicated receptor for professional phagocytosis.

Keywords: phagocytosis, CEACAM3, pathogenic bacteria, granulocyte, human innate immunity, ITAM, immunoreceptor tyrosine-based activation motif, signal transduction

#### INTRODUCTION

The ability to detect and phagocytose microbes is vital to protect multicellular organisms against dangerous infections. In mammals, this important function is accomplished by dedicated immune cells, the so-called professional phagocytes, encompassing macrophages, dendritic cells, and polymorphonuclear granulocytes (PMNs). They carry out phagocytosis via two distinct mechanisms: On the one hand, they perform opsonin-independent phagocytosis by utilizing receptors such as mannose receptor, scavenger receptor, Siglecs, DC-SIGN, or Dectin-1, which directly recognize and bind microbial surfaces that expose characteristic molecular patterns, such as glycan structures with terminal mannose or sialic acid residues, or fungal β-glucans (1–3). As such types of glycans are found on various microorganisms, including bacteria, fungi, as well as protozoa, and can also occur on endogenous structures, opsonin-independent receptors often detect a broad and diverse range of particles. On the other hand, professional phagocytes are capable of performing opsonin-dependent phagocytosis. Prominent opsonin-dependent receptors

#### Edited by:

Carlos Rosales, National Autonomous University of Mexico, Mexico

#### Reviewed by:

Scott D. Gray-Owen, University of Toronto, Canada Wolfgang A. Zimmermann, Ludwig-Maximilians University München, Germany James L. Stafford, University of Alberta, Canada

#### \*Correspondence:

Christof R. Hauck christof.hauck@uni-konstanz.de

#### Specialty section:

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

Received: 27 September 2019 Accepted: 31 December 2019 Published: 11 February 2020

#### Citation:

Bonsignore P, Kuiper JWP, Adrian J, Goob G and Hauck CR (2020) CEACAM3—A Prim(at)e Invention for Opsonin-Independent Phagocytosis of Bacteria. Front. Immunol. 10:3160. doi: 10.3389/fimmu.2019.03160

**127**

are complement or Fc receptors, which require prior coating of particles with host-derived complement components or specific antibodies before they are able to initiate phagocytosis (4– 6). Therefore, opsonin-dependent receptors can be targeted toward specific microbes, but they cannot support phagocytosis in situations where opsonins are either not present or where they fail to mark the microbial surface, for example, in the case of antigenically variable or encapsulated microorganisms.

Recent work has indicated that, at least in primates, a third group of specialized phagocytic receptors operates, which combines pathogen-specific detection with the immediate action of opsonin-independent receptors. The paradigm for this type of phagocytic receptors is the carcinoembryonic antigen-related cell adhesion molecule 3 (CEACAM3). CEACAM3 is a receptor of the immunoglobulin (Ig) superfamily and a member of the CEA subfamily of Ig domain containing cell adhesion molecules (IgCAMs) (**Figure 1**). In humans, CEACAM3 is selectively expressed by PMNs and plays a prominent role in the opsoninindependent detection and elimination of a small set of humanrestricted bacteria. In this review, we will place CEACAM3 in the context of a growing list of bacterial pathogens expressing CEACAM-binding adhesins and discuss the biochemical and functional evidence that this receptor is an effective phagocytosisinitiating protein and granulocyte activator. Further, we will elaborate the evolutionary trajectory of the CEACAM3 gene within the primate lineage and discuss the significance of human CEACAM3 polymorphisms, which appear to accommodate the recognition of variable bacterial surface antigens.

#### CEACAM FAMILY MEMBERS AND THEIR ROLE AS MICROBIAL TARGETS

Upon the identification of carcinoembryonic antigen (CEA) as a prominent surface protein expressed by human colon carcinomas, it was soon realized that antibodies directed against CEA react with numerous other proteins, especially on granulocytes (8, 9). According to their apparent molecular weights, these proteins were initially termed non-specific crossreacting antigen (NCA) -26, -50, -90, -95, and -160 (10). Screening of a human leukocyte cDNA library with a probe derived from NCA-50 uncovered several transcripts including clone W264 containing a 1259-base pair insert (11). The insert encoded a 252-amino acid protein, which was designated Carcinoembryonic Gene family Member 1a (CGM1a) and later grouped, due to its reactivity with monoclonal antibodies, into the CD66 cluster of differentiation. Besides CGM1a (CD66d), the CD66 antigens comprise biliary glycoprotein (BGP; CD66a), CGM6 (CD66b), NCA-50 (CD66c), and CEA (CD66e), which share 69–92% amino acid sequence identity in their aminoterminal immunoglobulin-variable (IgV)-like domains with CGM1a (11–14). Superposition of known crystal structures of CEA, CD66a, and CD66c reveals that these sequence similarities

CEACAM1, CEACAM3, and CEACAM4 connect the extracellular Ig-domains with functional ITIM (CEACAM1), ITAM-like (CEACAM3), or consensus ITAM sequences (CEACAM4). GPI-anchors of CEACAM5/CEA and CEACAM6 are depicted in gray.

also translate into high structural conservation between these proteins (7, 15).

The growing awareness of the complexity of the CEA family necessitated a major revision of the nomenclature, which led to CGM1a (CD66d) being renamed CEACAM3 (16) (for current and former nomenclature of CEACAM family members discussed in this review, please consult **Figure 1**). While CEACAM3 transcripts and protein have only been detected in human granulocytes and myeloid leukemia cells, the other closely related CD66 antigens are either widely expressed on epithelial and hematopoietic cells (CD66a/BGP/CEACAM1 as well as CD66c/NCA-50/CEACAM6) or exclusively expressed by mucosal epithelial cells (CD66e/CEA/CEACAM5) (17–19). Furthermore, CEACAM3 is distinct from other CD66 antigens in that its extracellular part does only comprise a single IgVlike domain and lacks additional Ig constant (IgC)-like domains, which are present in varying numbers (2-6 IgC-like domains) in CEACAM1, CEACAM5, and CEACAM6 (**Figure 1**) (20). This short stature of CEACAM3 might also be the reason why this receptor does not participate in binding interactions with other CEACAM family members, as the IgC-like domains can stabilize

**Abbreviations:** CEA, carcinoembryonic antigen; CEACAM, CEA-related cell adhesion molecule; Ig, immunoglobulin; ITAM, immunoreceptor tyrosine-based activation motif.

cis- and trans-interactions between CEACAM extracellular domains and thereby contribute to cell–cell adhesion (21, 22). Indeed, CEACAM1, CEA, CEACAM6, and CEACAM8 engage in CEACAM–CEACAM interactions with each other to support cell–cell binding (12, 21–24).

As CEACAM3 does not participate in these binding interactions, what could then be the function of this particular CEA-related protein on professional phagocytes? Clearly, the physiological role of CEACAM3 can only be reconciled in light of the fact that several pathogenic bacteria and fungi take advantage of epithelial CEACAMs as preferred docking sites on the mucosa [for review, see (25)]. Indeed, a growing list of pathogens has been found to express dedicated adhesins to specifically connect to human CEACAM family members, such as CEACAM1, CEA, and CEACAM6, which are exposed on the apical surface of human epithelial cells. CEACAM-binding microorganisms comprise Neisseria gonorrhoeae (causative agent of the venereal disease gonorrhea), Neisseria meningitidis (bacterial meningitis), Haemophilus influenzae (pneumonia, bacterial meningitis), Haemophilus aegyptius (purulent conjunctivitis), Helicobacter pylori (chronic gastritis, stomach cancer), Moraxella catarrhalis (otitis media, sinusitis), Fusobacterium nucleatum (periodontal disease), pathogenic Escherichia coli strains (Adherent-invasive E. coli, Diffusely adherent E. coli; involved in Crohn's disease), and the yeast Candida albicans (candidiasis, systemic infections) (26–35). It is important to mention that almost each of these pathogens employs a structurally distinct adhesive protein to bind human CEACAMs, implying that these adhesins have evolved independently multiple times in a striking form of convergent evolution (7, 15, 32, 36–39). Evidently, there must be strong, but not necessarily a uniform selection pressure on these microorganisms to develop CEACAM-binding adhesins. Several non-mutually exclusive explanations have been put forward to explain this exceptional preference of distinct pathogenic microbes to engage human CEACAMs. One finding relates to the fact that CEACAM1, the target of a large fraction of these adhesins, is also expressed by T cells and that major CEACAM1 isoforms have a negative regulatory role in T cell stimulation and proliferation [reviewed in (40)]. A second hypothesis is based on the fact that a unifying theme for all CEACAMbinding microbes is their outstanding ability to colonize, often throughout the lifespan of an individual, the mucosal surface of either the naso-pharynx, the gastrointestinal, or the urogenital tract. The role of CEACAM engagement in mucosal colonization has been best worked out in the case of N. gonorrhoeae and N. meningitidis and demonstrated that both microbes greatly profit from tight association with CEACAMs, which facilitates successful host colonization (41–43). Aside from their role as a handle by which to anchor to the mucosal epithelia, CEACAM engagement allows bacteria to suppress the exfoliation and delamination of superficial epithelial cells, thereby creating a stable foothold on the mucosa (41, 44, 45). It becomes obvious that pathogens can immensely profit, potentially in multiple ways, from engaging CEACAMs on epithelial cells and this nicely explains the prevalence and independent evolution of CEACAMbinding adhesins among human pathogens. However, why is it then that humans rarely succumb to gonococcal infection or develop severe forms of disease after being colonized by N. meningitidis or H. pylori, which are present in a large fraction of the healthy population? Indeed, CEACAM-binding pathogens such as F. nucleatum, M. catarrhalis, N. gonorrhoeae, or H. pylori, despite being able to efficiently colonize the human mucosa, rarely or only in a minority of the cases lead to a fatal outcome. It is exactly in the context of this scenario that we can now appreciate the role of CEACAM3, a CEACAM family member that does not participate in cell–cell interactions, but is present on the surface of professional phagocytes. In particular, the capacity of CEACAM3 to trigger rapid phagocytosis of attached particles and to activate bactericidal mechanisms of granulocytes will be discussed in the next sections, as these features provide major clues to understand the specialized function of this protein.

#### CEACAM3-INITIATED SIGNAL TRANSDUCTION LEADING TO PHAGOCYTOSIS

CEACAM3's notable status within the CEACAM family is not only due to its small extracellular domain and its cell-typespecific expression pattern, but is also based on a particular sequence motif within its cytoplasmic domain. Similar to the prototypic opsonin-dependent phagocytic receptors of the Fc receptor family, the carboxy-terminus of CEACAM3 encompasses an immunoreceptor tyrosine-based activation motif (ITAM) [for review, see (46)]. To be more precise, the motif found in CEACAM3 does not conform perfectly to the consensus ITAM (D/Ex(7)D/ExxYxxI/Lx(6−8)YxxI/L), but resembles an ITAM-like motif, where the carboxy-terminal leucine/isoleucine residue is substituted by methionine (47–49). The presence of this motif and the expression in professional phagocytes already indicate that CEACAM3 might be involved in phagocytosis of CEACAM-binding bacteria. For several CEACAM-binding pathogens, granulocytes play a major role during symptomatic disease. For example, the purulent exudate containing numerous granulocytes with intracellular, gram-negative diplococci is a diagnostic hallmark of gonorrhea (**Figures 2A,B**). It has long been known that gonococci, which express a CEACAM3 binding adhesin, are recognized and phagocytosed by human granulocytes in an opsonin-independent manner (**Figure 2C**), while isogenic strains lacking a CEACAM-binding adhesin are hardly recognized under these conditions (26, 27, 50–52). Despite the presence of other CEACAM family members such as CEACAM1 and CEACAM6 on the granulocyte surface and despite the fact that CEACAM3 is expressed at lower levels compared to CEACAM1 and CEACAM6, CEACAM3 is the main driving force behind this rapid and efficient opsoninindependent phagocytosis (53). Evidence for the prominent role of CEACAM3 comes from pharmacological, biochemical, genetic, and microbiological approaches: inhibitors that affect CEACAM1 and CEACAM6-mediated uptake in transfected cell lines (such as cholesterol-depleting agents) do not interfere with the opsonin-independent phagocytosis of CEACAM-binding bacteria by granulocytes (54), while inhibitors or blocking antibodies selectively affecting CEACAM3-mediated uptake in

stain of cervical smears from N. gonorrhoeae-infected women. Even at low magnification (A) the presence of granulocytes (arrowheads) and detached epithelial cells (asterisk) can be readily detected. (B) At higher magnification, granulocytes with phagocytosed diplococci (arrow) as well as extracellular bacteria are visible. (C) Scanning electron microscopy highlights the massive lamellipodia (arrowheads) induced on isolated primary human granulocytes upon exposure to CEACAM3-binding N. gonorrhoeae (asterisk).

transfected cell lines also disrupt this process in granulocytes (50). Interference with CEACAM3-specific binding partners or signaling processes by transduction of primary human granulocytes with dominant-negative variants also severely compromises the opsonin-independent uptake of CEACAMbinding bacteria. Selective expression of CEACAM3, but not CEACAM1 or CEACAM6, in murine promyelocytic cells can recapitulate major features of neutrophil activation in response to CEACAM-binding bacteria such as oxidative burst and degranulation (55). Furthermore, microbes expressing particular adhesins, which bind CEACAM1, but not CEACAM3, are hardly phagocytosed by primary human granulocytes in the absence of opsonins (56). Therefore, the immediate and dramatic phagocytic response of human granulocytes exposed to CEACAM-binding bacteria can be mainly attributed to CEACAM3.

A number of studies have addressed the molecular basis of CEACAM3's capability to vigorously trigger opsoninindependent phagocytosis. Most of these investigations, conducted with either transfected human cell lines or primary human granulocytes, have pointed toward a major role of the ITAM-like motif for CEACAM3 functionality in phagocytosis. For example, phosphorylation of the tyrosine residues within this motif (Y230/Y241) is critical for CEACAM3-initiated phagocytosis, as mutation of either tyrosine to a phenylalanine significantly decreases internalization and mutation of both residues results in an additive effect (47, 48, 53, 57). Interestingly, a single tyrosine-to-phenylalanine mutation completely blocked phosphorylation of CEACAM3 (48). Whether this points to a cooperative phosphorylation mechanism requiring both tyrosine residues or is due to inadequate sensitivity of the assay is unclear. Besides the ITAM-like motif, additional structural elements within the cytoplasmic domain possibly contribute to phagocytic signaling as the CEACAM3 Y230F/Y241F double mutant exhibits residual phagocytic activity compared to variants, which lack the complete cytosolic domain (48, 57). In contrast to CEACAM1 and CEACAM6, cholesterol-rich membrane domains (lipid rafts) do not seem to contribute to CEACAM3-mediated phagocytosis, as the CEACAM3 dependent internalization of bacteria is insensitive to severe cholesterol depletion, e.g., by methyl-β-cyclodextrin (54, 58, 59). It has been proposed that a Y-to-F mutation in the ITAM motif generates a binding site for AP-2, which could support an endocytic mode of internalization (48). However, regular endocytosis via AP-2 initiated, clathrin-coated vesicles has an upper size limit of 200 nm (60), implying alternative endocytic processes upon deletion or mutation of the CEACAM3 ITAMlike sequence. Though it is currently unknown which specific cellular processes guide the residual, ITAM-independent internalization of bacteria, the ITAM-dependent events upon CEACAM3 stimulation have been extensively analyzed.

Genetic, biochemical, and pharmacological evidence supports a major role for kinases of the Src family in CEACAM3 phosphorylation (**Figure 3A**). Indeed, the local clustering of CEACAM3 by the multivalent bacteria triggers recruitment and activation of several members of the Src family tyrosine kinases, including Hck and Fgr in granulocytes, while in transfected cell lines, Src, Yes, and Fyn might take over the respective role (48, 52, 61, 62). Due to acyl modification, Src family kinases are constitutively associated with the cytoplasmic leaflet of membranes and are therefore in a prime position to initially phosphorylate the ITAM tyrosine residues. Although the tyrosine kinase Syk is also recruited to nascent gonococci-containing phagosomes in an ITAMdependent fashion, pharmacological inhibition of Syk did not reduce bacterial internalization (63). Only when polystyrol beads larger than 5µm (a typical gonococcal diplococcus is about 1–2µm in size) were coated with anti-CEACAM IgG and used as bacterial surrogate did Syk augment internalization. The fact that Syk facilitates phagocytosis depending on particle size is not unique to CEACAM3 as the same phenomenon has been observed for FcγRmediated phagocytosis (64). Therefore, Syk is dispensable for internalization of gonococci, but it clearly does promote downstream bactericidal activity by enhancing the oxidative burst, degranulation, and the NF-κB-mediated inflammatory response (63).

Upon phosphorylation, the ITAM-like motif in CEACAM3 creates a platform for effectors that drive cytoskeletal remodeling required for phagocytic cup formation (**Figure 3B**). The adaptor molecule Nck binds CEACAM3 in a phosphorylation-dependent manner via its SH2 domain and recruits the WAVE2 complex to sites of bacterial attachment (65). WAVE2 is part of a multiprotein complex that activates Arp2/3-mediated actin nucleation, which drives lamellipodia extension [reviewed in (66)]. Indeed, ablation of Nck or inhibition of WAVE2 impedes lamellipodia formation and bacterial uptake (65). The WAVE2 complex is a coincidence detector, which is activated by the local and temporal co-occurrence of protein

FIGURE 3 | CEACAM3 signaling connections. (A) Phosphorylation of CEACAM3 upon bacterial engagement. Localized clustering of CEACAM3 receptors by multivalent, CEACAM-binding bacteria trigger recruitment and activation of several Src family tyrosine kinases, which in turn phosphorylate the ITAM-like sequence in the cytoplasmic domain of the receptor at two tyrosine residues (Y230 and Y241). This phosphorylation event is critical for downstream signaling processes. (B) CEACAM3 signaling to the cytoskeleton. The adaptor protein Nck binds via its SH2 domain to the membrane proximal CEACAM3 phosphotyrosine residue pY230 and recruits the WAVE2 complex to sites of bacterial attachment. Activation of this complex requires a localized increase in GTP-loaded Rac, which is orchestrated by recruitment of the Rac guanine nucleotide exchange factor Vav. Similar to Nck, Vav binds via its SH2 domain to CEACAM3 pY230. Activation of the WAVE2 complex leads to localized activation of Arp2/3 at the site of infection resulting in particle engulfment by actin cytoskeleton-based lamellipodial protrusions. (C) CEACAM3 signaling to the NADPH oxidase. Residue pY230 serves as an interaction platform for the regulatory subunit of phosphatidylinositol 3′ kinase (PI3K). PI3Ks are responsible for the generation of phosphoinositides (PIPs), which regulate the oxidative burst by recruitment of NADPH oxidase subunits p40phox and p47phox. In addition, PIP hydrolysis by CEACAM3-recruited phospholipase C γ (PLCγ) releases the second messengers inositol-(1,4,5)-trisphosphate (IP3) and diacyl glycerol (DAG). Together, they activate the protein kinase C (PKC), which further stimulates the NADPH oxidase complex. (D) CEACAM3 stimulation of neutrophil transcriptional response. CEACAM3 can trigger the NF-κB-mediated transcriptional regulation of IL-8 and other neutrophil chemotactic factors. CEACAM3-initiated NF-κB activation occurs via two distinct routes: the PKCδ/TAK1 pathway and the CARD9/BCL10 pathway depend on Syk localization to phosphorylated CEACAM3.

tyrosine phosphorylation, acidic phospholipids, and activation of the small GTPase Rac (67). Importantly, all of these events are concomitantly initiated by CEACAM3 engagement. Acidic phospholipids such as phosphatidylinositol-(3,4,5) trisphosphate (PIP3) and phosphatidylinositol-(3)-phosphate (PI3P) sequentially accumulate at the phagocytic cup during engulfment (59, 68, 69). WAVE2 activation strictly depends on the recruitment of the small Rho GTPase, Rac, which is crucial for remodeling of the cytoskeleton (70). Rac1 is also essential for CEACAM3-mediated phagocytosis and is locally and transiently activated at the phagocytic cup (52, 53, 57). Activation of Rac is catalyzed by specific guanine nucleotide exchange factors (GEFs) that exchange Rac1-bound GDP for GTP and CEACAM3 aggregation triggers Rac GTP loading by the GEF Vav (61). Interestingly, Vav itself is activated by Src family kinases (71, 72) and the Vav SH2 domain can interact directly with the phosphorylated Y230 within the ITAM-like motif of CEACAM3 (61). As most other GEFs bind via their pleckstrin homology domain (PH domain) to phosphoinositides, their recruitment is intrinsically dependent on PI3K activity. In contrast, Vav's direct interaction with phosphorylated CEACAM3 could render this process PI3K-independent. Indeed, phosphoinositide 3′ kinase (PI3K) activity is dispensable for CEACAM3-mediated internalization of gonococci (68, 73). The direct interaction between phosphorylated CEACAM3 and a Rac GEF as well as the direct linkage to the WAVE complex via Nck could explain why CEACAM3-mediated phagocytosis is particularly efficient and rapid (69). In vitro, ∼90% of primary human granulocytes have internalized multiple CEACAM-binding gonococci within 20 min of infection, while CEACAM-non-binding gonococci remain almost untouched (48, 69). These studies of CEACAM3-initiated signal transduction have also indicated that the ability of this receptor to trigger phagocytosis relies on cellular constituents, such as Src family kinases, Vav, Rac, and Nck-WAVE, which can be found in almost any mammalian cell type. Maybe this universality of CEACAM3 downstream connections explains why transfection of CEACAM3 cDNA in diverse cell types, from cervical epithelial cells to mouse fibroblasts, is sufficient to convert such non-professional phagocytes into cells avidly and efficiently internalizing CEACAM-binding bacteria. The generic nature of these processes also indicates that studies of CEACAM3-initiated phagocytosis are helpful to delineate the core elements necessary for productive internalization of particles.

#### CEACAM3-INITIATED SIGNAL TRANSDUCTION BEYOND PHAGOCYTOSIS

In professional phagocytes, phagocytosis is tightly coupled to downstream bactericidal processes. One central regulatory switch in this regard is the GTPase Rac, which not only regulates cytoskeletal remodeling, but is also an essential subunit of the NADPH oxidase, the enzyme complex generating the microbiocidal oxidative burst (**Figure 3C**). In contrast to the PI3K-independent Rac activation during engulfment described above, the CEACAM3-induced oxidative burst strictly depends on PI3K activity (73). Since there is only a partial temporal overlap between particle engulfment and the oxidative burst, it is possible that Vav-mediated GTP-loading of Rac during engulfment is disconnected from a putative phosphoinositidedependent GEF, which might activate Rac later during the induction of an oxidative burst. However, PI3K-generated phosphoinositides are required at additional regulatory steps during NADPH oxidase assembly, such as the recruitment of NADPH oxidase subunits p40phox and p47phox through their PI3P-binding PX domains (74). Phosphoinositides also serve as substrate for various lipid phosphatases and phospholipases. Interestingly, the SH2 domain of phospholipase C γ (PLCγ) can bind CEACAM3 in vitro and the isolated SH2 domain is enriched around gonococci-containing phagosomes (48, 75). PLCγ-mediated hydrolysis of phosphatidylinositol-(4,5) bisphosphate (PIP2) produces the second messengers diacyl glycerol (DAG) and inositol-(1,4,5)-trisphosphate (IP3), which in turn trigger increases in cytosolic Ca2<sup>+</sup> and PKC activation. Indeed, intracellular Ca2<sup>+</sup> levels in PMNs rapidly rise upon CEACAM3 engagement and this process does not occur in cells lacking PLCγ (47). Accordingly, upstream PLCγ activity might be required to allow PKC-mediated phosphorylation of multiple NADPH oxidase subunits, which represents an important regulatory step in activation of the oxidative burst [reviewed in (76)].

Though both tyrosine residues within the ITAM-like sequence of CEACAM3 seem to be functionally relevant, the biochemical assays conducted so far have pointed toward the Y230 residue as the central hub for interactions with SH2 domain-containing proteins. This conclusion is based on binding studies with synthetic phospho-peptides and pull-down experiments with recombinant proteins, which demonstrate that the SH2 domains of Src family kinases, PI3K, Nck, or Vav selectively bind to phospho-Y230. Therefore, the CEACAM3 ITAM-like sequence has been likened to a so-called HemITAM sequence found, for example, in the macrophage receptor Dectin-1, where also a single tyrosine residue conveys the phagocytic function (77, 78). Interestingly, the only known negative regulator of CEACAM3-mediated signaling, the adaptor protein Grb14, also targets phospho-Y230 (79). One can easily envision that Grb14 restricts access for other SH2 domain-containing effector proteins and thereby interferes with CEACAM3 mediated phagocytosis. However, the idea that multiple proteins compete for phospho-Y230 of CEACAM3 immediately begs the question how these binding events can be coordinated to allow a productive and orchestrated cellular response. In the future, time-resolved analysis of the various SH2 domainmediated binding events upon bacterial CEACAM3 engagement might help to answer this question. Nevertheless, the emerging overall picture of CEACAM3 phosphorylation-initiated events depicts Y230 within the ITAM-like motif as the minimal structural feature, which directly links CEACAM3 engagement with cytoskeletal remodeling as well as with the initiation of bactericidal responses.

It is interesting to mention that there is a further member of the CEACAM family, CEACAM4, which has a domain architecture similar to CEACAM3, which is selectively expressed in granulocytes, and which harbors a consensus ITAM (D/ExxYxxLx(6−8)YxxI) (**Figure 1**). CEACAM4 is an orphan receptor, as neither an endogenous nor a microbial ligand for this membrane protein has been detected. However, the cytoplasmic domain of CEACAM4 can trigger particle uptake, with both ITAM-embedded tyrosine residues engaging in SH2 domain interactions (80). Therefore, the human CEACAM family appears to harbor additional phagocytic receptors that might function to eliminate microorganisms in an opsoninindependent manner.

Although the main task of PMNs is to clear pathogens through their bactericidal capabilities, there is mounting evidence that they can also shape the inflammatory response [reviewed in (81, 82)]. Indeed, CEACAM3 activation can trigger an inflammatory program in PMNs (**Figure 3D**). CEACAM3 binding by Moxarella catarrhalis activates the CARD9/BCL10 pathway resulting in NF-κB-mediated expression of the potent neutrophil chemotactic factor, interleukin-8 (IL-8) (83). As CEACAM3 expression is restricted to human neutrophils, its contribution to inflammatory responses in an organismal context are difficult to study. Sintsova et al. made use of a transgenic mouse model that harbors a 187-kb human bacterial artificial chromosome encoding CEACAM3, CEACAM5, CEACAM6, and CEACAM7 (CEABAC mice) (84) to study the CEACAM-mediated response to infection with N. gonorrhoeae (85). Global expression profiles were generated from both WT and transgenic neutrophils infected with CEACAM-binding gonococci to discern CEACAM-dependent signatures. A pronounced upregulation of pro-inflammatory cytokines was observed in neutrophils from CEABAC mice, which depends on p38 MAPK activity and the PKCδ/TAK1/NF-κB axis (85). Though the used transgenic neutrophils and primary human granulocytes express other CEACAMs, which could be engaged by CEACAM-binding bacteria (such as CEACAM6 in the case of transgenic murine neutrophils or CEACAM1 and CEACAM6 in the case of primary human neutrophils), inhibitor studies point to CEACAM3-ITAM signaling as the main contributor to neutrophil-initiated inflammatory responses. In vivo, infection of CEACAM-transgenic, but not wild-type mice with N. gonorrhoeae led to increased neutrophil infiltration and increased levels of neutrophil-derived IL-1β and MIP-1α. Accordingly, CEACAM3 signaling on the one hand helps to limit gonococcal survival, but also initiates a vicious cycle, where the bacteria-triggered release of chemotactic cytokines leads to increased neutrophil influx and potentiates the risk of severe damage to the infected tissue (86). In this light, it will be important to understand how CEACAM3 signaling is kept in check to prevent unrestrained inflammatory signaling. In contrast to the initiation of CEACAM3 signaling, surprisingly little is known about its termination. The negative regulatory role of the adaptor protein Grb14 (79) has already been discussed above. Furthermore, phosphorylation of CEACAM3 appears to be counteracted by the cytoplasmic protein tyrosine phosphatase SHP-1, which most likely constrains CEACAM3 effector functions by compromising ITAM functionality (87). Interestingly, neutrophils also express CEACAM1, which contains an immunoreceptor tyrosine-based inhibitory motif (ITIM). However, co-recruitment of CEACAM1 does not seem to have an inhibitory effect on CEACAM3 phagocytic activity (55). Further research is required to address mechanisms that could regulate CEACAM3 activity, including intracellular trafficking of CEACAM3 and cooperation with other phagocytic receptors (e.g., Fc-γ and complement receptors). Considering the potential detrimental effects of excessive CEACAM3 activation, it is highly likely that in human PMNs, additional negative modulators of CEACAM3 signaling operate.

#### CEACAM3 EVOLUTION—A RED QUEEN SCENARIO AT WORK

Based on the functional studies summarized in the previous sections, it is safe to conclude that CEACAM3 represents an effective detector and eliminator of CEACAM-binding bacteria. In its capacity as a phagocytosis-promoting receptor, CEACAM3 can take care of pathogens expressing CEACAMbinding adhesins designed to exploit the human receptors for mucosal colonization. Thereby, CEACAM3 might help to establish a truce between host and pathogen, which could be one of the reasons why CEACAM-binding pathogens are contained in most instances. Looking at this reality from the viewpoint of a microbe, wouldn't it then be smart to avoid expressing a CEACAM3-binding adhesin in the first place? Obviously, bacteria unanimously answered this question with "Yes," as they appear to optimize their adhesins to allow binding to epithelial CEACAMs such as CEACAM1 and CEA, while evading CEACAM3 recognition. Indeed, most CEACAMbinding bacterial adhesins analyzed in this regard show selective binding to either CEACAM1 or CEA. This is true for the CEACAM-binding Opa protein adhesins of N. gonorrhoeae and N. meningitidis as well as for their functional homolog, the OMP P1 adhesin of H. influenzae, which have been studied most thoroughly in this context. For example, a single gonococcal strain encodes 11 distinct Opa adhesins and the complete compendium of Opa proteins has been functionally tested for binding to CEACAM family members in the case of N. gonorrhoeae strains MS11 and VP1. In MS11, 10 out of 11 Opa adhesins bind CEA, while only three of these are also able to be recognized by CEACAM3 (50, 51, 88). For strain VP1, 8 out of 10 tested Opa adhesins bound to CEA or CEACAM1, while only a single Opa protein associated with CEACAM3 (56). In a complementary approach, Sintsova et al. screened a large collection of primary isolates from gonorrhea patients for their CEACAM-binding capacity. Instead of trying to analyze the complete Opa repertoire of each of these strains, the authors used low passage isolates, which are thought to continue to express in vitro the Opa variant previously selected for and expressed in vivo (89). Also in this study, the overwhelming majority with 74% or 80% of the strains bound CEA or CEACAM1, respectively, and only 27% were found to also recognize CEACAM3 (89). While most strains bound CEA and/or CEACAM1, but not CEACAM3, not a single strain could be observed, which showed a reverse binding pattern: recognizing CEACAM3, but not recognizing an epithelial CEA family member (89). An even more skewed situation is found in N. meningitidis, where 11 of 13 Opa proteins of serogroup A, B, and C strains, which have been tested in binding assays with different CEACAMs, associated with either CEA or CEACAM1, but none of those Opa proteins associated with CEACAM3 (90, 91). Similarly, all of the OMP P1 adhesins derived from 13 different strains of H. influenzae bound to CEACAM1, but none of those adhesins was recognized by CEACAM3 (32). Accordingly, bacteria seem to optimize their adhesins to discriminate between these closely related CEACAMs and to avoid recognition by CEACAM3.

If there is such a directed evolution on the side of bacterial CEACAM-binding adhesins, is there a discernable adaptation on the host side? The increasing availability of human and other primate genomic information has now allowed an unraveling of the evolutionary context of CEACAM3. It has been recognized before that homologs of several human CEACAMs are absent from rodents. More specifically, a CEACAM3 ortholog could not be identified even in Old World primates, such as baboon or African green monkey, by homology searches based on the extracellular IgV-like domain (92). Large-scale genome comparisons between closely related primate species have revealed a high degree of non-synonymous vs. synonymous nucleotide changes in ortholog genes of CEACAM family members (93, 94). Interestingly, the bacterial receptors within the CEACAM family (CEACAM1, CEACAM3, CEA, and CEACAM6) show an exceptionally strong signature of positive selection, suggesting that they belong to the fastest-evolving human genes. In particular, the CEACAM3 gene appears to be a recent evolutionary invention. Though ITAM-sequence containing CEACAM-related genes have been described for various mammals (95–97), a CEACAM3 gene with its specific exon/intron structure and the characteristic ITAM-like sequence seems to occur only after the emergence of Old World monkeys (around 35 million years ago) (98). Indeed, a CEACAM3 ortholog has only been detected at the syntenic locus in the genomes of baboon, macaque, orangutan, gorilla, chimpanzee, and humans (94). Due to uncertainties in the assembly of some genomes such as tarsier, lemurs, and lories, gene synteny is not a valid criterion to rule out the existence of CEACAM3 orthologs in lower primates. However, CEACAM3 orthologs share several discriminative features such as protein length or the presence of the characteristic HemITAM, which does not seem to occur in any lower primate genome supporting the idea that CEACAM3 emerged late in primate evolution. Despite the absence of a CEACAM3 gene, ITAM-containing CEACAM transcripts are present in lower primates, such as transcript XR\_001153184.2 in the mouse lemur. The ITAM encoded by XR\_001153184.2 perfectly matches the ITAM consensus (YxxL−7AS–YxxI) as well as the sequence found in human CEACAM4 (80). Therefore, this transcript could originate from an ancestral primate CEACAM4 ortholog. It is plausible that CEACAM3 is the result of gene duplication and recombination between an ancestral CEACAM1 gene (providing the Ig<sup>V</sup> domain encoding exon) and an ancestral lower primate CEACAM4 gene (providing the transmembrane and the intracellular domain encoding exons) (99). Indeed, the 3′ end of the large intron 2 of the CEACAM3 gene bears similarities to intron 2 of the CEACAM4 gene, supporting the idea that such a recombination event occurred at the advent of higher primates (99). Together, the genomic evidence suggests that CEACAM3, the specialized phagocytic receptor for CEACAM-binding bacteria, emerged relatively late during primate evolution, expanding the human innate immunity arsenal.

An interesting corollary of this continuing bacteria–host co-evolution can be seen by the occurrence of CEACAM3 polymorphisms in the human population. In fact, while people outside of Africa mainly express the common CEACAM3 allele, particular ethnic groups harbor CEACAM3 variants. For example, in several groups with African ancestry, a third of the population expresses a distinct CEACAM3 allele (minor CEACAM3 allele) (94). This minor allele carries four non-synonymous single-nucleotide changes, which convert the CEACAM3-Ig<sup>V</sup> domain amino acid sequence into a near replica of the CEACAM1 Ig<sup>V</sup> domain. Modeling of the polymorphic CEACAM3 Ig<sup>V</sup> domain according to the known CEACAM3 crystal structure (7) reveals that these residues are surface exposed and three of the modified amino acids contribute to the CEACAM3 binding interface for various bacterial adhesins (94). It comes as no surprise that this minor CEACAM3 allele, which now mimics CEACAM1, shows a striking gain of function in that it recognizes additional bacterial adhesins, such as the OMP P1 protein of H. influenzae, which escape detection by the common CEACAM3 allele (94).

These recent studies have shed light on an ongoing arms race between several host-restricted CEACAM-binding pathogens and the human immune system. In a type of Red-Queen scenario, the diverse CEACAM-binding pathogens need to optimize their adhesins to escape CEACAM3 detection, while still retaining their ability to bind mucosal CEACAMs. As a consequence, human CEACAM3 rapidly evolves to keep up its function. Viewed from the angle of bacterial adhesin binding to distinct CEACAMs as well as from the perspective of human genomic variation, the phagocytic receptor CEACAM3 appears as a central element in the innate defense against this group of extremely specialized bacteria. However, one has to acknowledge that CEACAM binding preferences of bacterial adhesins as well as analyses of CEACAM3 evolution only provide circumstantial evidence for the importance of this innate immune receptor. Accordingly, a more direct test of CEACAM3 function in the context of innate immune responses within the intact organism would be desirable. Unfortunately, the standard genetic approach of employing knock-out animals, such as a "CEACAM3-knock-out" mouse, is not an option in this case. Moreover, CEACAM3-transgenic mice with granulocyte-selective expression of CEACAM3, in the absence of other CEACAM family members, do not exist currently.

In view of the increasing availability of human genomic information and corresponding medical records, it does not appear unrealistic to foresee a future, where novel insight into CEACAM3 function might come directly from combined genomic–phenotypic studies in humans (100). Indeed, the occurrence of an unexpected high proportion of loss-of-function alleles in particular human populations (101) could help to reveal plausible connections between CEACAM3 deficiency and susceptibility for particular infectious diseases. Despite the inherent limitations of research on a primate-only protein, the study of the phagocytic receptor CEACAM3 will surely continue to unearth fascinating aspects of human biology and of our constant and challenging interplay with the microbial world.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### REFERENCES


#### FUNDING

Work by the authors has been supported by the DFG (Ha 2856/10-1 and Ha 2856/11-1) and Baden-Württemberg Stiftung (BWST\_WSF-020) to CH.

#### ACKNOWLEDGMENTS

The authors are indebted to Michael Laumann (Electron Microscopy Service; University of Konstanz, Germany) for scanning electron microscopy and Nathlee Abbai (University KwaZulu-Natal, Durban, South Africa) for providing histological samples.


diffusely adhering Escherichia coli (Afa/Dr DAEC). Mol Microbiol. (2004) 52:963–83. doi: 10.1111/j.1365-2958.2004.04033.x


during engulfment of Neisseria gonorrhoeae by the neutrophil-restricted CEACAM3 (CD66d) receptor. Mol Microbiol. (2003) 49:623–37. doi: 10.1046/j.1365-2958.2003.03591.x


**Conflict of Interest:** 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 © 2020 Bonsignore, Kuiper, Adrian, Goob and Hauck. 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.

# Bone Marrow-Derived and Elicited Peritoneal Macrophages Are Not Created Equal: The Questions Asked Dictate the Cell Type Used

Cheryl M. Zajd<sup>1</sup> , Alexis M. Ziemba<sup>2</sup> , Grace M. Miralles 1†, Terry Nguyen<sup>1</sup> , Paul J. Feustel <sup>3</sup> , Stanley M. Dunn<sup>2</sup> , Ryan J. Gilbert <sup>2</sup> and Michelle R. Lennartz <sup>1</sup> \*

<sup>1</sup> Department of Regenerative and Cancer Cell Biology, Albany Medical College, Albany, NY, United States, <sup>2</sup> Department of Biomedical Engineering, Rensselaer Polytechnic Institute, Troy, NY, United States, <sup>3</sup> Department of Neuroscience and Experimental Therapeutics, Albany Medical College, Albany, NY, United States

#### Edited by:

Florence Niedergang, Centre National de la Recherche Scientifique (CNRS), France

#### Reviewed by:

Panagiotis F. Christopoulos, Oslo University Hospital, Norway Jaya Talreja, Wayne State University, United States

> \*Correspondence: Michelle R. Lennartz

lennarm@amc.edu †Present address:

Grace M. Miralles, Regeneron Pharmaceuticals, Tarrytown, NY, United States

#### Specialty section:

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

Received: 18 October 2019 Accepted: 03 February 2020 Published: 21 February 2020

#### Citation:

Zajd CM, Ziemba AM, Miralles GM, Nguyen T, Feustel PJ, Dunn SM, Gilbert RJ and Lennartz MR (2020) Bone Marrow-Derived and Elicited Peritoneal Macrophages Are Not Created Equal: The Questions Asked Dictate the Cell Type Used. Front. Immunol. 11:269. doi: 10.3389/fimmu.2020.00269 Macrophages are a heterogeneous and plastic population of cells whose phenotype changes in response to their environment. Macrophage biologists utilize peritoneal (pMAC) and bone marrow-derived macrophages (BMDM) for in vitro studies. Given that pMACs mature in vivo while BMDM are ex vivo differentiated from stem cells, it is likely that their responses differ under experimental conditions. Surprisingly little is known about how BMDM and pMACs responses compare under the same experimental conditionals. While morphologically similar with respect to forward and side scatter by flow cytometry, reports in the literature suggest that pMACs are more mature than their BMDM counterparts. Given the dearth of information comparing BMDM and pMACs, this work was undertaken to test the hypothesis that elicited pMACs are more responsive to defined conditions, including phagocytosis, respiratory burst, polarization, and cytokine and chemokine release. In all cases, our hypothesis was disproved. At steady state, BMDM are more phagocytic (both rate and extent) than elicited pMACs. In response to polarization, they upregulate chemokine and cytokine gene expression and release more cytokines. The results demonstrate that BMDM are generally more responsive and poised to respond to their environment, while pMAC responses are, in comparison, less pronounced. BMDM responses are a function of intrinsic differences, while pMAC responses reflect their differentiation in the context of the whole animal. This distinction may be important in knockout animals, where the pMAC phenotype may be influenced by the absence of the gene of interest.

Keywords: peritoneal macrophages, bone marrow-derived macrophages, polarization, cytokines, phagocytosis, flow cytometry, gene expression

# INTRODUCTION

Macrophages are innate immune cells that provide a first line defense against infection. While many studies historically utilized macrophage-like cell lines, the availability of knockout animals as well as development of molecular techniques for these notoriously difficult-to-transfect cells has resulted in the increased use of primary macrophages. The most commonly used primary

**138**

macrophages are elicited peritoneal and bone marrowderived. Sterile thioglycolate, injected into the peritoneum, recruits circulating monocytes that differentiate into small peritoneal macrophages. The small peritoneal macrophages are phenotypically indistinguishable from resident peritoneal macrophages (pMACs) (1). Myeloid progenitor cells, harvested from the bone marrow, are differentiated with macrophage colony-stimulating factor (M-CSF) or conditioned L929 media, to produce adherent bone marrow-derived macrophages (BMDM) (2). Neither pMACs nor BMDM preparations are homogeneous (1, 3, 4). pMACs have more lysosomal protease activity and don't significantly proliferate, indicative of a more mature phenotype (3, 5); BMDMs gravitate toward the M2 end of the polarization spectrum (6). Despite their intrinsic heterogeneity, thioglycolate-elicited pMACs and BMDMs are similar with respect to forward and side scatter as determined by flow cytometry. However, their differentiation environment may influence their phenotype, particularly if differentiation occurs in the context of a genetically manipulated (knockout or transgenic) animal. Given that pMACs and BMDMs are differentiated in vivo and ex vivo, respectively, and there are reported differences between the two (3, 6, 7), it is somewhat surprising that the two have not been compared with respect to the properties that define macrophages: phagocytosis, respiratory burst, polarization, and gene regulation. Despite reports that pMACs are more mature (and thus respond more robustly to stimulation), we found that BMDMs are more phagocytic (rate and amount of material ingested) and respond more robustly to polarization (surface molecule expression, gene induction/repression, and cytokine/chemokine release). These findings are consistent with the differential plasticity of pMACs and BMDMs. That is, pMACs, being differentiated in vivo, respond modestly when stimulated ex vivo while BMDMs are poised to respond rapidly and robustly to either pro-inflammatory or pro-resolving stimuli in vitro.

# RESULTS

## BMDMs and pMACs Are Similar With Respect to Size and Granularity

Bone marrow-derived macrophages were differentiated using L-cell conditioned media as the source of macrophage colony stimulating factor (M-CSF). The resultant live, singlet population is predominantly (98 ± 2%, n = 8) CD11b+F4/80<sup>+</sup> (**Figure 1A**); there is no detectable SiglecF or Ly6G. Based on forward and side scatter, BMDM have a minor population (15.8 ± 3.4%, n = 10) of large cells. As reported previously for pMACs (1), CD11b and F4/80 expression is significantly higher on large vs. small BMDM (p < 0.01, n = 10) (**Figure 1B**).

of the live singlets were CD11b+F4/80+. (B) After gating out dead cells/debris and selecting for singlets, two populations were identified: a minor (15.8 ± 3.4%) population of high forward and side scatter (large) cells and a major population that is smaller with lower side scatter. The large population had significantly higher expression of both F4/80 and CD11b (p < 0.01, n = 10, paired t-test).

Elicited peritoneal cells are predominantly macrophages although significant percentages of non-macrophage cells have been reported (8). Three days after injection of 3% sterile thioglycolate, peritoneal cells were harvested by lavage and the red cells lysed. One aliquot of cells was kept on ice while the other was plated in petri dishes; plates were washed after 4 h and the adherent cells recovered. An average of 5.0 × 10<sup>7</sup> cells were collected (range 2.0–6.2 × 10<sup>7</sup> ). Following adhesion, an average of 1.7 × 10<sup>7</sup> cells were recovered (range 0.6–2.9 × 10<sup>7</sup> ), an average recovery of 37 ± 10% (n = 15). Flow cytometry revealed a low forward scatter, moderate side scatter population in the harvested pMACs (11 ± 4.4%, n = 10) that was significantly diminished upon adhesion (2.1 ± 1.1%, n = 10, p < 0.01, paired t-test) and not present in macrophages differentiated from bone marrow (purple arrow, **Figure 2A**). This population was CD11b−SiglecF+, consistent with a minor contamination with eosinophils (8), a population that was substantively removed by selective adhesion (**Figure 2B**). Under our elicitation conditions, the (Ly6G+) neutrophil contamination is minimal, with an average of 1.2 ± 2% of the harvested cells before adhesion being CD11b+Ly6G+Ly6Clo/neg (n = 10). The majority of recovered peritoneal cells (82.7 ± 6.2 %, n = 10) are CD11b+; this percentage rose significantly (91.5 ± 2.5 %, p < 0.005, n = 10) following adhesion (**Figures 2B,C**). Like BMDMs, selected pMACs contain large (∼20%) and small macrophages (**Figures 2A,D**) (1); adhesion does not affect the relative percentages of these populations. When compared, adherent pMACs and BMDMs are similar with respect to size and granularity (**Figure 2D**, overlay).

The CD11b<sup>+</sup> peritoneal population is Ly6Clo, lacks Ly6G, and is relatively homogeneous with respect to F4/80 expression (see below). Thus, selective adhesion removes the majority of eosinophils and leaves a relatively homogeneous cell population that is >90% CD11b+F4/80+Ly6C−/lo. Note that, from this point forward, all experiments were done with post-adherent peritoneal macrophages. For simplicity, pMACs data is presented in red and BMDM in black.

#### BMDMs, but Not pMACs, Are M1 Skewed

M1 and M2 macrophages, produced in vitro by IFN ± LPS and IL13/IL4, respectively, are acknowledged to be the extremes of the pro-inflammatory-to-pro-resolving spectrum (9). Physiologically, macrophages likely assume a hybrid phenotype of cell markers and cytokine/chemokine release, with their in vivo impact dependent on the balance between M1 and M2 outputs. At baseline, BMDMs and post-adherent pMACs have similar expression levels of CD11b, F4/80, CD16/CD32, CD16.2, MHCII, and CD119 (**Figure 3A** and p value list).

#### M1 Markers

Compared to pMACs, BMDMs express significantly higher levels of Ly6C and CD64, molecules elevated on inflammatory

macrophages (10–12). The TLR2 and TLR4 pattern recognition marker receptors are more highly expressed on BMDMs (**Figure 3B** and p value list). Elevated TLR2 and TLR4 may prime macrophages for a rapid response to pathogens. While BMDMs have high Ly6C and elevated TLR2 and 4, suggestive of an M1 phenotype, their levels of MHCII expression is low and similar to pMACs. As elevated MHCII is a marker of M1 activation, its modest expression at steady state is consistent with M1 skewing but not bone fide activation. Thus, compared to in vivo differentiated pMACs, BMDM lie further toward the M1 end of the M1–M2 polarization spectrum.

significantly higher than pMACs. N.D. Not determined.

#### M2 Markers

Of the three M2 markers tested, only CD124 is substantively expressed. CD119 is low, and similar, for both cell types; CD206 was variable (neg to low) (**Figure 3A**). As CD124 is the α chain of the IL-4 receptor, its expression could make macrophages more responsive to environmental (or in vitro) IL-4, an M2 polarizing cytokine.

In summary, BMDMs and (post-adherent) pMACs are similar with respect to size and granularity, expression of macrophage markers CD11b and F4/80, and three of the four Fcγ receptors (CD16/32, CD16.2) (**Figure 3A**). Compared to pMACs, BMDMs are Ly6Chi and have elevated M1 markers TLR2, TLR4, and CD64 as well as significantly higher CD124 (**Figure 3B**). The expression of both M1 and M2 markers on BMDMs may prime them to polarize in response to either inflammatory or pro-resolving mediators. In contrast, pMACs express modest levels of CD64 and low levels of TLR2 and CD124, suggesting they may be more refractive to polarization. Functional assays were performed to compare BMDMs and pMACs with respect to phagocytosis, respiratory burst, and their response to polarizing conditions.

# BMDMs Are Significantly More Phagocytic Than Their pMAC Counterparts

#### E. coli and E. coli-IgG (Figure 4A)

While both BMDMs and pMACs are used for phagocytosis studies, their relative phagocytic capacities have not been rigorously compared. Using pHrodo <sup>R</sup> -labeled E. coli, alone or IgG-opsonized (E. coli-IgG), we compared the phagocytic capacity of BMDMs and pMACs. pHrodo <sup>R</sup> particles are nonfluorescent when extracellular but become brightly fluorescent in the acidic environment of the phagosome. The rate of E. coli phagocytosis (MFI/min, slope of the time curves, **Figure 4A**) by BMDM was 5-fold > pMACs (41/8) and 3-fold greater for E. coli-IgG (69/21) (n = 3 BMDMs and 3 pMACs) (**Figure 4D**). Using an unpaired t-test, we determined that the rate of phagocytosis was significantly higher for BMDMs compared to pMACs (p < 0.001, **Figure 4D**).

#### Zymosan and Zymosan-IgG (Figure 4B)

A similar strategy was used to quantify internalization of zymosan and IgG-opsonized zymosan (**Figure 4B**). Alexa 488 conjugated zymosan ± IgG were incubated with BMDMs or pMACs. At varying timepoints (5–60 min), cell associated zymosan was detached by vortexing, trypan blue was added to quench the fluorescence of external particles, and fluorescence quantified by flow cytometry. Zymosan is considerably larger than E. coli and is taken up through TLR2. As zymosan phagocytosis began to plateau between 15 and 30 min, we



calculated the initial rate of uptake using the 5–15 min datapoints. Like E. coli, internalization of zymosan by BMDM was significantly higher than for pMACs: 2.7-fold for zymosan and 2.3-fold for Zymosan-IgG (p < 0.001, **Figure 4D**). Notably, IgG opsonization increased the rate of, E. coli, but not and zymosan, internalization (**Figure 4B**). Noting that the rate of zymosan phagocytosis is much higher than E. coli, it may be that the zymosan system may be "max'd out" such that addition of IgG cannot increase the rate.

#### IgG Beads (Figure 4C)

To determine whether uptake mediated by FcγR, but independent of TLR, is different between BMDMs and pMACs, we coated 2µm beads with (rabbit) IgG (BIgG) and calculated the rate of target internalization using synchronized phagocytosis, calculating the phagocytic index microscopically (13). The phagocytic index (number of beads/number of cells × 100) at every timepoint was significantly different, with BMDM internalizing more targets and having a phagocytic rate (slope of the line) ∼2-fold higher than pMACs (**Figures 4C,D**).

As CD64 is the only FcγR differentially expressed on BMDMs and pMACs (**Figure 3**), we hypothesized that BIgG uptake requires CD64. To visualize internalization, BMDMs were transduced with PKC-ε-GFP, a molecule that concentrates at the phagosome during IgG-mediated phagocytosis (14). By using macrophages from FcγRIIb knockout (CD32−/−) mice, we removed the contribution of this receptor, a modification that did not substantively affect BIgG internalization (**Figure 5B**). Likewise, adding 2.4G2 (CD16/32 blocking antibody) to CD32−/<sup>−</sup> cells did not affect phagocytosis (**Figure 5C**) nor

(A), Zymosan 488 ± IgG (5:1 MOI) (B), or BIgG (20:1 target to cell ratio) (C). (A) Phagocytosis was stopped at the indicated times by dislodging bound targets, diluting the sample in cold buffer, and analyzing by flow cytometry. (B) The fluorescence of extracellular zymosan was quenched with trypan blue immediately before analysis (n = 3 animals, 1 × 10<sup>5</sup> cells collected/sample). (C) For BIgG, cells were fixed and incomplete phagosomes were detected by the addition of Alexa 488-conjugated goat anti-rabbit IgG (Invitrogen) to label the IgG on the exposed targets. The number of fully internalized targets was quantified microscopically and the phagocytic index (PI) calculated. PI = (# internal beads/# cells counted) × 100. (n = 3 animals, 30–40 cells per animal). (A–C) \*p < 0.05; \*\*p < 0.01; \*\*\*p < 0.001, unpaired t-test. (D) Composite data from 3 each pMAC and BMDM preparations reporting the rate of phagocytosis (slope of the line) and determining statistical significance using an unpaired t-test. Because internalization of zymosan plateaus by 30 min, an initial rate of phagocytosis was calculated using the 5–15 min timepoints (dashed line). BMDM are more phagocytic for all targets, regardless of the receptor used or the method used to quantify phagocytosis.

did the inclusion of the 16.2 blocking antibody 9E9 (15) (**Figure 5D**). In contrast, the addition of α-CD64 dramatically reduced internalization (**Figure 5E**) with no apparent effect on binding (**Figure 5E**, inset). These data suggest that CD64 is the major receptor for IgG-mediated phagocytosis. To determine if CD64 is necessary and sufficient for phagocytosis, we determined the rate of BIgG internalization in BMDMs from C57BL/6 and FcγR knockout mice expressing only CD64 (FcγRI only, **Figures 5G,H**) (16). The fact that BMDMs from FcγRI only mice take up BIgG at the same rate as their wildtype counterparts (**Figure 5H**) identifies FcγRI as the major receptor mediating IgG-dependent phagocytosis. The lower expression of CD64 on pMACs provides a potential explanation for the rate differences between BMDMs and pMACs.

In summary, using three targets (E. coli, zymosan, and BIgG), multiple approaches (pHrodo <sup>R</sup> , Alexa 488-zymosan, and BIgG) and two readouts (fluorescence and live imaging), we have demonstrated that BMDMs are more phagocytic than pMACs. The higher rates of BMDM phagocytosis parallels the surface expression of target receptors (TLR2, TLR4, CD64, **Figure 3**) and is consistent with the conclusion that BMDMs are M1 skewed.

# BMDMs and pMACs Mount a Similar Respiratory Burst

As M1 polarized macrophages mount a larger respiratory burst than non-polarized or M2-polarized cells (17), and BMDMs are M1 skewed, we predicted that BMDMs would have a larger respiratory burst than pMACs. To test this, macrophages were incubated with immune complexes (IC) in the presence of Amplex Red <sup>R</sup> , a membrane impermeant indicator that fluoresces when oxidized. Fluorescence measurements were taken every 5 min for 4 h. Surprisingly, there was no difference in between the curves over the first 60 min with a slight divergence at later times (**Figure 6A**). This was not a function of "maxing out" the system as three concentrations of IC were tested (the lowest shown) and, while there was a dose dependent increase in fluorescence with increasing IC, the rate of the burst (the slope of the line) in BMDMs and pMACs were not different (**Figures 6A,C**). As M1 polarization increases the respiratory burst in BMDM (**Supplemental Figure 1**), we asked if polarization would reveal a difference between pMACs and BMDM. Cells were polarized with IFN-γ (M1) or IL4/IL13 (M2) and the respiratory burst followed with time. As with unpolarized macrophages, there were no differences in the rate of the burst under either polarization condition (**Figures 6B,C**).

#### BMDMs and pMACs Respond Differently to Polarization

The plasticity of BMDMs is well-documented (18, 19) but how pMACsrespond to polarizing cytokines is less well-studied. Thus, pMACs and BMDMs were treated with M1 (IFN-γ) and M2 (IL4/IL13) polarizing cytokines and surface molecule expression, mRNA levels, and secreted cytokines were quantified; untreated BMDMs and post-adherent pMACs served as the control

in Methods. Compared to CD57BL/6 (A); phagocytosis was unaffected by removal of CD32 (B). Adding α-CD16/32 (C) or α-CD16.2 (D) to CD32−/<sup>−</sup> cells did not affect internalization. Blocking FcγRI with α-CD64 reduced internalization (E) but not target binding (inset), supporting a role for FcγRI in IgG-mediated phagocytosis. (F–H) Internalization by C57BL/6 and FcγRI only macrophages is similar. Quantitation of phagocytic rate from movies reveal that uptake by FcγRI only cells is equivalent to controls, arguing that FcγRI is necessary and sufficient for IgG-mediated phagocytosis. (H) Each dot represents data from one cell, statistical significance was tested using an unpaired t-test.

FIGURE 6 | The respiratory burst is equivalent in BMDMs and pMACs. BMDMs and pMACs, untreated (A) or polarized overnight with IFN-γ or IL4/IL13 (B), were stimulated with an empirically determined amount of immune complexes (IC) in the presence of Amplex Red®, a H2O<sup>2</sup> reporter. Fluorescence intensities were acquired every 5 min for 4 h and the relative rate of the burst determined from the slope of the line (C). Data is presented as mean ± SEM for pMACs and BMDM from 3 animals at the lowest dose of three doses of IC tested; two higher doses increased fluorescence but did not produce any difference in the burst. NSD = not significantly different.

chain PCR are presented as mean ± SEM (n = 3 BMDM and 3 pMACs). Daggers indicate significance based on cell type: ‡p < 0.05. Asterisks indicate differences based on polarization conditions: \*p < 0.05, \*\*p < 0.005. In general, BMDM responses are more robust than those of pMACs. pMACs and BMDM from n = 7 animals were analyzed.

(M0) macrophages. To ensure reproducibility, experiments were repeated 2–3 times, with each trial containing both BMDMs and pMACs.

#### Surface Molecule Expression (Figure 7)

Flow cytometry was used to quantify expression of surface molecules. After gating out debris and aggregates, fluorescence presented as Gaussian curves (**Figure 3**) allowing the comparison of mean fluorescent intensities (MFI) for the populations. To visualize how BMDM and pMACs respond to polarization, and to remove differences due to animal-to-animal variability, we analyzed the cells from each animal independently. That is, for each set of cells, we averaged the MFIs for each protein over the three conditions (no polarization, IFN-γ, and IL4/IL13) and plotted the difference from that mean (**Figure 7**); no change from the average would be plotted as "0" (i.e., MFI for each condition is the same). Thus, each line in **Figure 7** is essentially a repeated measures ANOVA, the responses of the cells from a single animal over the three conditions. M1 polarization by IFN-γ is validated by the elevated expression of M1 marker genes MHCII, CD64, and CD16.2 (**Figure 7A** and **Table 2** Polarization Main Effect) (15, 20). Similarly, IL4/IL13-dependent upregulation of CD16/32 (**Figure 7A**) and downregulation of TLR2 (**Figure 7B** and **Table 2** Polarization Main Effect) and Arg-1 mRNA (see below) confirms M2 polarization (21). Other surface markers tested include CD11b, F4/80, and Ly6C as well as TLR4 and receptors for polarization cytokines IFN-γ (CD119) and IL4 (CD124) (**Figures 7B,C**). IFN-γ- and IL4/IL13- treated macrophages and their non-treated controls were analyzed for differences due to cell type (BMDMs vs. pMACs, irrespective of treatment, Cell Type Main Effect in **Tables 2**–**4**), treatment (IFNγ, IL4/IL13, and control, irrespective of cell type, Polarization Main Effect **Tables 2**–**4**), and both (that is, do BMDMs and pMACs respond differently to polarization?, Interaction Effect **Tables 2**–**4**). By these criteria, the expression of Ly6C, TLR2, TLR4, CD16.2, and CD124 is significantly different between BMDMs and pMACs, evident from the separation of BMDM (black) and pMAC (red) lines (**Figure 7**) and the statistical significance (bold entries) of the cell type main effect (**Table 2**). CD64 expression approaches, but does not reach, statistical significance (p = 0.077, **Figure 7A**, **Table 2**).

FcγRIIb (CD32) and FcγRIII (CD16) are detected by a single antibody, 2.4G2 (designated CD16/32 in **Figure 7A**). Thus, the increase in 2.4G2 staining with M2 polarization (**Figure 7A**)

TABLE 2 | Comparison of surface expression in BMDM and pMACs in response to polarization.


BMDM and pMACs were polarized as described in Methods. Cell surface expression was quantified as Mean Fluorescent Intensity by flow cytometry as in Figure 3 and the statistical significance based on cell type (irrespective of polarization), polarization (irrespective of cell type), or cell type and polarization (interaction effect) determined. Specifically, the deviation from the mean, induced by IFN-γ , IL4/IL13, or left untreated (NT) was calculated for each animal (and graphed in Figure 7). Statistical significance was determined by 2-way ANOVA, thus removing the animal-to-animal variability and considering only the within-animal responses. p-values are presented. Statistically significant differences are indicated in bold text. n = BMDM and pMACs from 7 animals.

could be due to elevated expression of one or both receptors. To identify the receptor(s) that are upregulated upon M2 polarization, BMDMs and pMACs from 6 animals (3 BMDMs, 3 pMACs) were polarized with IFN-γ or IL4/IL13 and the α chains of CD64, 32, 16, and 16.2 were quantified by qPCR; control cells were left untreated. Consistent with the flow data, mRNA for CD64 increased in response to IFN-γ with no difference between BMDMs and pMACs (CD64 **Figure 7A**, inset). Similarly, CD16.2 was significantly higher in M1 compared to M2 cells (CD16.2 **Figure 7A**, inset). CD32, but not CD16, message was higher in IL4/IL13-treated cells with BMDMs having a significantly higher expression compared to pMACs (CD16/32, **Figure 7A**, inset). As CD16 message was not altered by polarization, we conclude that the increase in 2.42G signal in IL4/IL13 treated cells is due to upregulation of CD32. This is not surprising as FcγRIIb is an inhibitory receptor and M2 polarization reduces inflammation and promotes resolution.

The regulation of five genes (Ly6C, TLR2, TLR4, CD16.2, and CD124) is a function of both cell type and polarization conditions (Interaction Effect, **Table 2**). While these genes are known to be regulated by polarization (**Table 2**, Polarization Main Effect), this data demonstrates that pMACs and BMDM respond differently to polarization. Comparing the BMDM (black) with pMAC (red) lines (**Figure 7**), we find that the expression of many genes, (e.g., Ly6C, CD64, TLR 2, TLR4, CD16.2, CD124) are relatively unaffected by polarization for pMACs (i.e., red lines are relatively horizontal compared to black lines), leading us to conclude that pMACs are less responsive to their environment than BMDMs. Of the 11 genes tested, CD11b, F4/80, and CD119 are unaffected by polarization or cell type (**Figure 7C**, **Table 2**); Ly6G expression was low/neg.

#### Relative Gene Expression (Figure 8)

M1 and M2 polarized macrophages release pro-inflammatory and pro-resolving cytokines, respectively, to sustain or dampen immune responses. Given the responsiveness of BMDMs to IFNγ and IL4/IL13 (**Figure 7**, **Table 2**), we predicted that polarization would elicit a greater change in gene expression in BMDMs compared to pMACs. To test this, BMDMs and pMACs were treated as above and their mRNA subjected to qPCR for IL12/iNOS and IL10/Arg-1 (the canonical proteins/cytokines expressed by M1 and M2 cells, respectively) as well as IL6 and IL-1β (associated with inflammation) and TGF-β and CD206, selectively expressed by M2 cells.

Due to biological variability, mRNA expression in polarized samples was normalized to their respective (untreated) controls and significance determined using linear regression. As with surface expression (**Table 2**), the results were analyzed to assess differences due to cell type, polarization, and interaction effect (**Table 3**).

#### M1 Markers (Figure 8A, Table 3)

Not surprisingly, IL12 p40 message increased significantly (3– 5-fold) in response to IFN-γ, while IL4/IL13 had little effect (Polarization Main Effect **Table 3**); there was no significant difference between BMDM and pMAC levels of IL12 p40 mRNA (cell type effect, **Table 3**). While there was no difference in iNOS message between BMDMs and pMACs under either condition, iNOS expression trended lower in IL4/IL13 treated cells, approaching but not reaching, statistical significance (**Table 3**). This is consistent with the reported decrease in macrophage iNOS upon alternative activation (10). For IL6, there was no difference in response between the two cell types under either M1 or M2 polarizing conditions. For IL-1β, the overall change in expression was modest (<2-fold) but significantly different between BMDMs and pMACs (Cell Type Main Effect, **Table 3**) with the fold change in pMACs lower than BMDMs (**Table 3**). The low IL-1β message, coupled with no detectable protein release (by multiplex, see below) suggests that the differences in message may not be not physiologically relevant.

#### M2 Markers (Figure 8B, Table 3)

Message levels for IL10 were significantly different between BMDMs and pMACs with pMACs, but not BMDMs, showing the predicted pattern (i.e., low with IFN-γ, high in response to IL4/IL13). Statistical analysis confirmed a Polarization Main Effect (**Table 3**). Not surprisingly, Arg-1 expression was relatively low (and similar) in IFN-γ treated cells; the levels increased with IL4/IL13 (Polarization Main Effect) and the expression in BMDMs was significantly higher than in pMACs upon M2 polarization (Cell Type Main Effect). Thus, the polarization effect is significant as is the Interaction Effect (that is, BMDMs and pMACs respond differently to polarization, **Table 3**). Consistent with its designation as an M2 marker, CD206 was significantly elevated in IL4/IL13 treated cells, responding similarly in pMACs

TABLE 3 | Comparison of mRNA expression from BMDM and pMACs in response to polarization.

0.05, ‡‡p < 0.005). Asterisks denote differences based on polarization conditions: \*p < 0.05, \*\*p < 0.005).


BMDM and pMACs were polarized as described in Methods and expression of the indicated genes quantified by qPCR and reported as fold change (Figure 7). After logarithmic transformation, those data were fit with a linear model and statistical significance determined by ANOVA (α = 0.05) based on cell type alone (irrespective of polarization), polarization only (irrespective of cell type), or a differential response dependent on both cell type and polarization conditions (interaction effect). p-values are presented. Statistically significant differences are indicated in bold text. n = BMDM and pMACs from 4–6 animals.

and BMDMs. Finally, there were no significant differences in message levels for TGF-β between BMDMs and pMACs or as a function of polarization conditions, although expression trended higher in response to IL4/IL13.

**Table 3** presents a summary of the qPCR results (n = 4– 6 each, BMDM and pMAC). Comparing relative mRNA levels for BMDMs and pMACs independent of polarization revealed that, of the eight genes tested, the expression of only IL-1β varied as a function of cell type (Cell Type Main Effect, bold text); IL10 approached, but did not reach, statistical significance. Expression of Arginase-1 (Arg-1), inducible nitric oxide synthase (iNOS), IL12 p40, IL6, TGF-β, and CD206 were not significantly different between cell types. When assessing the effects of polarization independent of cell type, Arg-1 and CD206 were higher upon M2 polarization while IL10 and IL12 p40 were significantly lower (Polarization Main Effect, bold text, **Table 3**). While not surprising for the pro-inflammatory IL12, decreased IL10 in BMDMs is inconsistent with its role as an M2 cytokine. However, this pattern tracks with the protein (**Figure 9**, see below) and, while the explanation isn't clear, it should be noted that other M2 markers (e.g., Arg-1 and CD206) are elevated, validating polarization. One possible explanation is that IL10 upregulation may require an additional stimulus or more time for full expression. Finally, the expression of Arg-1 and IL10 was significantly different when both cell type and polarization are considered (Interaction Effect, bold text, **Table 3**); that is, BMDMs and pMACs respond differently to polarization. Given that these experiments tested the response of BMDMs and pMACs under identical conditions, the data support the conclusion that BMDMs are generally more responsive than pMACs to polarizing environments. This is based on the fact that, for the most part, message levels in BMDMs are generally greater than their pMAC counterparts (the white bars in **Figure 8** are often higher than the red bars).

As mRNA levels provide a snapshot in time, with the results being a function of message half-life and the time post-treatment, we collected the media after polarization and quantified the release of a cadre of cytokines and chemokines.

FIGURE 9 | BMDM and pMACs respond differently to polarization: release of cytokines and chemokines. BMDM and pMACs were treated with IFN-γ or IL4/IL13 as in Figure 7. At 24 h, the media was collected, cells and debris removed by centrifugation, and cytokine (A) and chemokine (B) concentrations in the supernatant quantified by Multiplex®. IL-1β, IL-12p70, and TNF-α were below the limits of detection. After logarithmic transformation, data were analyzed using a linear regression model and ANOVA. Data are presented as mean ± SEM (n = 4–7 animals). Statistical significance between cell types is indicated by daggers (‡p < 0.05, ‡‡p < 0.005); differences due to polarization conditions by asterisks; \*p < 0.05, \*\*p < 0.005.

TABLE 4 | Comparison of cytokine/chemokine release in BMDM and pMACs in response to polarization.


BMDM and pMACs were polarized as described in Methods, the media was collected and secretion of the indicated genes quantified by Multiplex®. The released protein (Figure 9) was either fit with a linear model (Left columns) or normalized to their respective controls, logarithmically-transformed, and fit with a linear model (Right columns) and analyzed by ANOVA (α = 0.05). Significance was based on cell type (irrespective of polarization), polarization (irrespective of cell type), or a differential response dependent on both cell type and polarization (interaction effect). Statistically significant differences based on cell type, polarization, or both are indicated in bold text. n = BMDM and pMACs from 4–7 animals. The interaction significance indicates that BMDM and pMACs respond differently to polarization. Of the genes tested, the interaction effect for IL-10, CXCL1, and CCL2 is significant (bold text).

#### Cytokine/Chemokine Release (Figure 9, Table 4)

The cocktail of cytokines/chemokines released by macrophages creates the environment to which other cells (and the macrophages themselves) respond. To compare the release of chemokines/cytokines by polarized BMDMs and pMACs, cells were stimulated with IFN-γ or IL4/IL13 overnight, the media was collected, and protein release was quantified by Multiplex <sup>R</sup> . Nine chemokines/cytokines were analyzed, three of which (IL-1β, IL12 p70, and TNF-α) were below the level of detection. Under all conditions, release of IL-12p40 and IL10 was significantly higher in BMDM compared with pMACs (**Figure 9A**, **Table 4**, left, Cell Type Main Effect). Not surprisingly, IFN-γ significantly increased IL12 p40 secretion and IL4/IL13 had little effect, with the concentration of released IL12 p40 in unstimulated and IL4/IL13-treated cells being similar (**Figure 9A**). Interestingly, IFN-γ decreased IL10 release in pMACs but increased it in BMDMs (**Figure 9A**). IL10 secretion from IL4/IL13 exposed cells was similar to their respective no treatment controls. IL6 was not significantly different between the cell types nor as a function of polarization, although

IL4/IL13 for 24 h, the media collected, and protein release quantified by Multiplex®. The data are plotted as the deviation from the average for each treatment. Each line represents a single animal's cells under the three conditions. Results are reported as the within animal deviation of the measurement from the means of that animal's cells under the three conditions. This is essentially a repeated measures ANOVA, removing the animal-to-animal variance and considering the within animal response. Interactions are apparent when the pattern of the responses differs between the BMDM (black) and pMAC (red) curves (e.g., IL10, CXCL1, and CCL2). (A) cytokines and (B) chemokines. In general, BMDM responses are more robust than those of pMACs. n = 4–7 for each cell type.

levels trended higher in BMDMs (**Figure 9A**). CXCL1/KC, a neutrophil chemoattractant, was similar in resting BMDMs and pMACs but was significantly lower in IFN-γ treated cells, decreasing more in BMDMs than pMACs (**Figure 9B**, **Table 4**, left, Polarization Main Effect, bold text). Release of CXCL1 upon IL4/IL13 treatment was less than untreated cells but did not reach statistical significance (**Figure 9B**). While the reason for the lower CXCL1 concentration under proinflammatory conditions (when recruitment of neutrophils would be advantageous) is not apparent, it is possible that IFN-γ upregulates CXC**R**1, leading to depletion of its ligand (CXCL1) from the media. Indeed, CXC**R**1 is upregulated on macrophages exposed to Staph aureus (22). CCL5/RANTES, an eosinophil, basophil, and T cell chemokine, is higher for BMDMs vs. pMACs under all conditions (**Figure 9B**, **Table 4**, left, Cell Type Main Effect, bold text). CCL5 concentrations are significantly higher in response to IFN-γ compared to no treatment, with BMDM levels higher than corresponding pMACs. CCL5 release in response to IL4/IL13 is similar to control, perhaps not surprising as it is released under inflammatory conditions. Secretion of CCL2/MCP-1, a monocyte chemoattractant, can be induced under both M1 and M2 polarizing conditions (**Figure 9B**). Interestingly, pMAC levels are higher than BMDM levels for untreated and IFN-γ exposed cells but BMDMs produce significantly more CCL2 than pMACs in response to IL4/IL13 (note that release is on a log scale).

These data suggest that, as with phagocytosis, surface molecular expression, and RNA levels (but not respiratory burst), BMDMs are more responsive than pMACs. To visualize the responses of the individual BMDM and pMAC preparations, we determined the mean for each animal under the three conditions (no treatment, IFN-γ, IL4/IL13) and, for each condition, calculated the deviation of that measurement from the mean (analysis similar to that in **Figure 7**). Comparing the responses of BMDMs and pMACs (**Figure 10**, black vs. red lines) it is clear that, for the cytokines/chemokines tested, pMACs are overall less responsive than their BMDM counterparts. **Table 4**, left, summarizes the statistical analyses with regards cytokine/chemokine release. There is a significant difference in the release of IL10, IL12 p40, CCL5, and CCL2 by pMACs and BMDM (**Table 4**, left, Cell Type Main Effect, bold text), with BMDM having a more robust response (red lines are relatively horizontal while black lines show dramatic variations, **Figure 10**). If polarization is considered irrespective of cell type, IL10, IL12p40, CXCL1, and CCL2 are differentially released (**Table 4**, left, Polarization Main Effect, bold text). Of interest physiologically is the interaction effect. For IL10, CXCL1, and CCL2, BMDM and pMACs respond differently, again with greater variations in release by BMDM (**Table 4**, left, Interaction

Effect, bold text). These findings would suggest that BMDM can alter their environment to a greater extent than pMACs.

The fact that cytokine/chemokine release is basally higher in BMDMs for five of the six released cytokine/chemokines (**Figure 9**) makes a comparison of responses difficult. Thus, to determine the relative change in release as a function of polarization, we normalized the Multiplex <sup>R</sup> results for each animal to its respective non-polarized control (e.g., untreated for the same animal). From these data, we calculated a "fold change" to determine if the changes in protein release were a function of differential response to polarization or a similar magnitude of response with different baselines. Even when compensating for the lower basal levels of release by pMACs, the normalized data revealed that BMDM responses are significantly greater for IL10, CXCL1, and CCL2, trend higher for CCL5, and are not significantly different for IL12 p40 and IL6 (**Figure 11**). The interaction statistics for the normalized data (**Table 4**, right, Interaction Effect) reveal that only IL10 and CXCL1 are significantly different with BMDMs having the greater response (i.e., fold change, bold text). With respect to polarization effect, the concentrations of IL10, IL12 p40, CXCL1, CCL5, and CCL2 released are significantly different regardless of cell type. Both cell type and polarization conditions affect secretion of IL10, CXCL1, and CCL2 (**Table 4**, right, Interaction Effect, bold text).

Taken together, these results suggest that BMDMs (differentiated in vitro) are more responsive to polarizing cytokines than pMACs, making them the preferred cell type for studying macrophage plasticity. Conversely, thioglycolate elicited pMACs are differentiated in vivo with their phenotype being a function of their intrinsic properties as well as those accrued during circulation and diapedesis. The fact that expression of CD64 and TLR4 (two major signaling receptors) does not change substantively in pMACs in response to either IFN-γ and IL4/IL-13 (**Figure 7**) suggests that pMACs may survey their environment, having a relatively high threshold for stimulation. Quantitation of cytokine/chemokine release produced a similar response pattern (**Figure 9**). That is BMDM were more responsive than pMACs to polarization (red lines in **Figure 10** show less deviation than the black lines representing BMDM).

In summary: Compared to pMACs, BMDM (1) are more phagocytic (**Figure 4**), (2) significantly upregulate surface markers in upon polarization (**Table 2**, Cell Type Main Effect, red vs. black lines, **Figure 7**) and (3) release more cytokines and chemokines (**Figure 9**, Interaction Effect, **Table 4**). The relative responsiveness of BMDM compared to pMACs suggest that they are poised to respond to infection. In contrast, pMACs have a higher threshold for response, and may serve as a "second line of defense," acting when the threat is elevated and/or sustained.

#### DISCUSSION

Despite decades of study, much of how macrophages orchestrate innate immune responses along the pro-inflammatory to proresolving axis remains to be elucidated. The seminal studies on phagocytosis and respiratory burst in macrophages were done with elicited peritoneal macrophages (23–28). With the advent of cloning, macrophage biologists moved to cell lines to circumvent the difficulties in transfecting primary cells. More recently, with advances in transfection/transduction techniques as well as the realization that the changing in vivo environment can alter macrophage phenotype (that may not be recapitulated in cell lines), the pendulum has swung back to primary cells. Thioglycolate-elicited and bone marrowderived are the most commonly used primary macrophages. Elicited peritoneal macrophages can be harvested 3 days after thioglycolate injection, providing a rapid and economical source of differentiated primary cells. The main disadvantage is the relatively low number of cells recovered and the fact that the macrophages are not a pure population, with the most common contaminants being neutrophils and eosinophils (**Figure 2**). In contrast, bone marrow-derived macrophages, differentiated from progenitor cells, are a relatively pure population that can be produced in high numbers but must be differentiated ∼7 days prior to use. While both cells types are used, the rationale for using one vs. the other is often not stated. The main difference between BMDMs and pMACs is that the pMACs are differentiated in the context of the background of the mouse. Knowing that macrophage phenotype is a function of environment, we asked if there were differences between pMACs and BMDMs under controlled conditions. Searching the literature, we found few studies that compared these two commonly used cell types. To define the similarities and differences between differentiation in vivo vs. ex vivo, we directly compared elicited pMACs to BMDMs with respect to the most common metrics of macrophage function: phagocytosis, respiratory burst, and gene regulation.

Despite their similar size and granularity (**Figure 2**), BMDMs have higher basal expression of CD64 (the high affinity IgG receptor), TLR2, and TLR4 (**Figure 3**). This is notable as CD64 mediates the uptake of IgG-opsonized particles (**Figure 5**), and TLR2 and TLR4 are receptors for zymosan (29) and E. coli, respectively. Additionally, BMDMs are Ly6Chi, characteristic of inflammatory macrophages that are more phagocytic than their pro-resolving counterparts (30, 31). Higher expression of CD64, TLR2, and TLR4 would poise BMDMs to respond early in infections when pathogen numbers are low. Indeed, BMDMs are more phagocytic than pMACs when presented with E. coli, zymosan, their IgG-opsonized counterparts, or IgG-opsonized beads (**Figure 4**). While the higher phagocytosis and receptor expression by BMDMs is consistent with an M1 phenotype, MHCII expression is similar in BMDMs and pMACs, suggesting that BMDMs are M1 skewed but not M1 activated. The M1 skewing may result from our use of conditioned L cell media as the source of M-CSF for BMDM differentiation. While M-CSF is essential for progenitor differentiation, L cell media (LCM) contains other (undefined) factors that produce a phenotype slightly different from that produced in the presence of purified M-CSF. In our experience, BMDM differentiated in LCM produce a more homogenous cell population with more reproducible results. This agrees with anecdotal comments from online forums and macrophage colleagues that suggest that BMDMs "look better" and "proliferate better" when differentiated in LCM. Also, as BMDM were historically differentiated with LCM, using this media will allow comparisons between published data and new results. To our knowledge, a direct comparison of the phenotype of BMDMs differentiated in LCM vs. M-CSF has not been reported.

Many of the studies on IgG-dependent phagocytosis and intracellular signaling have used targets opsonized with rabbit IgG (including the targets used here, **Figure 4**). To our knowledge, the Fc receptor(s) utilized have not been identified [although some groups have reported that FcRIV is the major activating FcR in mice (32)]. Given that CD64 is the only FcR whose expression correlates with the increased IgG-mediated phagocytosis in BMDM (**Figures 3**, **4**), we asked if it was the major receptor for rabbit IgG. Blocking CD64 dramatically reduced phagocytosis but not target binding (**Figure 5**). More definitive was the use of live imaging to quantify of the rate of phagocytosis in macrophages expressing all the Fcγ receptors (wild type) and those expressing only FcγRI. The fact that the rate of phagocytosis was equivalent argues that FcγRI is necessary and sufficient for IgG-mediated phagocytosis (**Figure 5**). Whether other metrics of macrophages (i.e., polarization, gene expression, respiratory burst) are CD64 dependent remains to be determined.

Given the M1 skewing of the BMDMs compared to pMACs, we expected the respiratory burst to be greater in BMDMs. We utilized Amplex Red <sup>R</sup> , a readout of H2O<sup>2</sup> release that was stable over hours. Although immune complexes stimulated the respiratory burst, there was no difference between pMACs and BMDMs (**Figure 6A**). Thinking that polarization may reveal differences between the two cell types, we polarized with IFN-γ or IL4/IL13 and followed the burst with time; polarization did not produce a difference in burst between pMACs and BMDM (**Figure 6B**). The targets in these experiments were insoluble BSA-anti-BSA immune complexes that are small and difficult to count. Thus, we tested 1X, 2X, and 3X amounts of complex to ensure that there was a high enough "multiplicity of infection" to see a difference between the cell types. The 1X data is presented in **Figure 6**; the higher concentrations produced more fluorescence but no difference in the rates between BMDM and pMACs. Thus, we conclude that the higher phagocytic rates and M1 skewing of BMDMs does not correlate with increased respiratory burst, even upon IFN-γ activation (**Figure 6B**). Notably, pilot experiments did show a significant difference between pMACs and BMDMs, with pMACs having a significantly greater burst than BMDMs. However, the earlier experiments used pMACs that had not undergone selective adhesion. Selective adhesion eliminated the difference, indicating that the higher burst is likely a function of contaminating neutrophils. It is not clear why the burst would be similar if phagocytosis is greater. However, we reported that the burst is independent of phagocytosis (33), so it is likely that the extent of FcR ligation is not the rate determining step for respiratory burst.

While the first response of macrophages to pathogens is internalization and the generation of a respiratory burst, the upregulation of gene expression propagates the response, providing cytokines to which they (and bystander cells) respond. With time, the adaptive immune response is engaged, exposing macrophages to polarizing cytokines, including IFN-γ, IL4, and IL13. The question we asked was "Do BMDMs and pMACs respond similarly to polarizing conditions?" BMDMs and pMACs were tested for their response to M1 and M2 polarizing conditions (IFN-γ and IL4/IL13, respectively). Surface molecule expression (**Figure 7**), message levels (**Figure 8**), and cytokine/chemokine release (**Figures 9**–**11**) were quantified. LPS and IFN-γ are the common M1 polarization agents. We tested the effects of LPS, IFN-γ, and both on BMDM to determine our M1 polarizing conditions. BMDMs were treated overnight with LPS (100 ng/ml), IFN-γ (100 ng/ml) or both. The media was collected after 24 h and the concentration of released IL-10 and IL-12 determined by ELISA. While both LPS and IFN-γ significantly increased cytokine secretion, the combination was neither additive nor synergistic (**Supplemental Figure 1A**). To assess the effect of IFN-γ ± LPS on the respiratory burst, polarized cells were incubated with zymosan or zymosan IgG and Amplex Red <sup>R</sup> (as in **Figure 6**) and the rate of the burst calculated. Although IFN-γ significantly increased the burst LPS had no additional effect (**Supplemental Figure 1B**). Thus, IFN-γ alone was used for M1 polarization.

Due to the tremendous amount of data produced as well as the biological variability, the statistical analyses were complicated: differences due to cell type, polarization conditions, and the two combined had to be determined. **Figures 7**–**11** provide a visualization of the relative response of BMDMs (black) and pMACs (red) under each condition. Based on the observation that BMDM (black lines) have greater deviations from the mean than pMACs (red lines) (**Figures 7**, **10**) we conclude that, overall, BMDM are more responsive to their environment. This extends to phagocytosis (**Figure 4**). **Tables 2**, **4** provide the summary of statistical significance for surface molecule expression and protein secretion, respectively. Statistical significance in the "Cell Type Main Effect" column (bold text) identify molecules whose expression is different between pMACs and BMDM regardless of polarization conditions. Statistical significance in the Interaction Effect (bold text) identifies molecules whose expression is a function of both polarization and cell type. The conclusion from the polarization data (bold text, **Tables 1**–**4**) is that, for every measure of polarization, there are molecules whose expression is a function of cell type. Where there is statistical significance, BMDMs respond more robustly than pMACs.

Which brings us back to the fundamental question: Are BMDMs and pMACs interchangeable? We would argue the answer is no, and that the cell type chosen depends on the questions to be asked. If cell biological questions are asked, BMDMs would provide a greater range of response under most conditions (phagocytosis, gene regulation, secretion) and lend themselves to molecular manipulation. pMACs provide a readout of the responsiveness of the innate immune system in the background of the animal from which they are isolated, particularly informative for knockout or genetically modified animals. These findings raise a cautionary note that, while in vitro studies are informative, they do not necessarily reflect in vivo phenotype.

# MATERIALS AND METHODS

#### Buffers

Dulbecco's Modified Eagle Medium (DMEM, Gibco), ACK lysis buffer (150 mM NH4Cl, 10 mM KHCO3, 0.1 mM Na2EDTA, pH 7.2–7.4), Bone marrow macrophage differentiation media: DMEM supplemented with 20% L-cell conditioned media, 10% FBS, 0.2% sodium bicarbonate, and gentamycin (50µg/ml); Macrophage media: DMEM containing 10% fetal bovine serum and gentamycin (50µg/ml). HBSS++: Hanks' balanced salt solution containing 4 mM sodium bicarbonate, 10 mM HEPES, and 1.5 mM each CaCl<sup>2</sup> and MgCl2.

#### Reagents

Highly purified BSA (Cat # A0281) was purchased from Sigma. Interferon-γ (Cat # 315–05; Lot # 061398), IL4 (Cat # 214–14; Lot # 111249), and IL13 (Cat # 210–13; Lot # 111207) were purchased from Peprotech. anti-BSA IgG (Cat #B1520) was purchased from Sigma. Alexa 488-conjugated α-rabbit IgG (Cat # A11070) was from Invitrogen Life Technologies.

# Flow Antibodies (See Table 1)

Targets. pHrodoTM Green E. coli BioParticlesTM (Cat # P35366), Zymosan A (S. cerevisiae) BioParticlesTM, Alexa FluorTM 488 conjugated (Cat # Z23373), E. coli BioParticlesTM Opsonizing Reagent (Cat # E2870), Zymosan A BioParticlesTM Opsonizing Reagent (Cat # Z2850), and AmplexTM Red Hydrogen Peroxide/Peroxidase Assay Kit (Cat # A22188) were purchased from Life Technologies. E. coli and Zymosan were IgG-opsonized per manufacturer's instructions.

#### Immune Complexes (IC)

IgG immune complexes were formed by incubating 1 mol of highly purified BSA with 3 mol (rabbit) anti-BSA IgG (60 min, 37◦C, with rotation). Complexes were washed with PBS before use.

#### IgG-Coated Beads (BIgG)

Were prepared as described previously (14). Briefly, 2µm borosilicate microspheres (Duke Standards, Thermo Scientific USA) were coated sequentially with poly L-lysine, activated with dimethylpimelimidate · 2 HCl, washed, and incubated with highly purified BSA (overnight, 4◦C with rotation). BSA beads were blocked (1 M Tris, pH 8.0), washed, and opsonized with rabbit anti-BSA IgG.

#### Cells

Male and female C57BL/6 mice, 12–16 weeks of age, were the cell source. FcγRI-only mice (VG1505) (16) for collection of bone marrows, were provided by Regeneron Pharmaceuticals, Inc. Requests for FcγRI only mice should be sent to Regeneron, they cannot be fulfilled by the corresponding author. Animals were bred in the Albany Medical College Animal Resource Facility. All procedures were done in under NPHS guidelines using protocols approved by the Albany Medical College Institutional Animal Care and Use Committee. The sex of the animal providing the cells was recorded, we found no differences in the responses from male and female mice.

#### Bone Marrow-Derived Macrophages (BMDMs)

Bone marrow was extruded from the femurs and pelvises of euthanized mice and differentiated in bone marrow macrophage differentiation media according to published procedures (34, 35) Albanesi, 2012 #16787}. Cells were used 7–10 days after harvesting.

#### Elicited Peritoneal Macrophages (pMACs)

Elicited peritoneal macrophages were recruited and harvested according to published methods (36). Briefly, mice were injected i.p. with aged thioglycolate. After 4 days, peritoneal exudates were collected using sterile phosphate buffered saline (PBS). Red blood cells were lysed using ACK lysis buffer. For selective adhesion, cells were plated in untreated petri dishes; after 4 h, non-adherent cells removed by washing in PBS and the adherent population removed using 5 mM EDTA/PBS (15 min with agitation). Recovered cells were resuspended in HBSS++ and used within 4 h. Note: Due to the fact that BMDMs are differentiated in vitro, and pMACs are used the day of harvest harvest, the BMDMs and pMACs used in each experiment were never from the same animal.

# Phagocytosis

#### E. coli and Zymosan

For flow-based assays, 2 × 10<sup>5</sup> post-adherent pMACs or BMDMs were added to flow tubes and the tubes placed on ice. Cold pHrodoTM Green E. coli ± IgG (50:1 MOI) or Alexa 488 conjugated zymosan ± IgG (5:1 MOI) were added; total volume of the assay was 35 µL. Tubes were kept on ice to allow target binding, then transferred to 37◦ waterbath to initiate a synchronized wave of phagocytosis. At varying times (2.5–15 min for E. coli, 5–60 min for the larger zymosan), the tubes were removed from the waterbath, vortexed to dislodge bound targets, and 100 µl of ice cold HBSS++ added to stop phagocytosis. Tubes were placed on ice and read as soon as possible. To quench the fluorescence from external zymosan, 4% trypan blue was added immediately before reading. Data was collected on 60–100 × 10<sup>6</sup> cells/tube.

#### IgG-Coated Beads (BIgG)

Sychronized BIgG phagocytosis was done as previously described (36). Briefly, macrophages (5 × 10<sup>4</sup> ) were plated onto coverslips in 24 well plates and cooled in an icebath. Targets (20:1) were added and allowed to bind on ice. Plates were transferred to 37◦C waterbath and fixed at the indicated times. Incomplete phagosomes were detected by the addition of Alexa 488 α-rabbit IgG which labeled the IgG on the exposed targets. Cell number was determined by DAPI staining.

#### Real Time Imaging

Live imaging for determination of phagocytic rates was done as previously detailed (34). BMDM were virally transduced to overexpress PKC-ε-GFP, brought into focus on the stage of the spinning disk confocal microscope, BIgG were added, and images taken every 5 s for 10 min. The rate of phagocytosis was determined to be the number of frames from the first indentation of the plasma membrane through the first frame showing completely enclosed particles × 5 (seconds between frames).

# Respiratory Burst

pMACs and BMDMs (3 × 10<sup>4</sup> ) were seeded in 96 well plates and allowed to adhere. The AmplexTM solutions were prepared as per manufacturer's instructions. Cells were stimulated with an empirically determined amount of immune complexes in the presence of 50µM AmplexTM Red and 0.1 U/mL horseradish peroxidase in HBSS++. Plate was maintained at 37◦C in a BioTekTM SynergyTM 2 Multi-Mode Microplate Reader. Data was collected every 5 min over the 4 h time period, the baseline (no treatment) was subtracted from each value and the net relative fluorescence units presented.

# Macrophage Polarization

Polarization was done on adherent macrophages (1 × 10<sup>6</sup> ) with 100 ng/mL IFN-γ or IL4 + IL13 (25 ng/ml each) for 24 h. Cells thus polarized (and with a non-treated control) were used for flow cytometry, respiratory burst, quantitative polymerase chain reaction (qPCR), and cytometric bead arrays. While LPS + IFN-γ is often used together for M1 polarization (10), and synergize in cytotoxicity assays (37, 38), we found that a 24 h incubation with IFN-γ was sufficient to M1 polarize (**Tables 2**–**4**), with the addition of LPS not significantly enhancing the IFN-γ responses for either cytokine release nor phagocytic rate (**Supplemental Figure 1**). As cytotoxicity assays are done on the order of days, it may be that feedback and/or gene regulation that occurs in that timeframe contributes to synergy. Alternatively, the fact that C57BL/6 mice are more M1 skewed at steady state, perhaps IFN-γ is sufficient to upregulate M1 markers. Finally, as summarized by Jackson Labs (https://www.jax.org/news-and-insights/jax-blog/ 2016/june/there-is-no-such-thing-as-a-b6-mouse), all C57BL/6 sub-strains are not the same. Our mice, originally ordered from Jackson but bred in our facility over many years, apparently do not require inputs from both IFN-γ and TLR agonists for the readouts we are studying.

# Flow Cytometry

1 × 10<sup>6</sup> cells were used for flow analysis on polarized cells. For all analyses except CD16/CD32 staining, cells were blocked with CD16/CD32 (Mouse Fc Block, Clone 2.4G2) for 15 min on ice, then incubated with antibodies to the proteins of interest (45 min, on ice). Unstained cells were used to establish flow cytometer settings. Fluorescence minus one (FMO) controls were used for compensation. Flow cytometric data were acquired on a FacsCalibur (Becton and Dickenson, Franklin Lakes, NJ) using FlowJo and the data analyzed using FlowJo Software

(Tree Star, Ashland, OR). Antibodies used are listed in **Table 1**.

### Quantitative Polymerase Chain Reaction (qPCR)

Quantitative polymerase chain reaction was conducted per previous lab protocols (39). The primers used are listed in **Table 1**. Expression for each gene was normalized to β-actin (1Ct) and the RNA fold change determined using the 11Ct method (1Ct gene after polarization—1Ct M0 control).

# Cytokine Bead Array Polarization Assay

Following polarization, cell supernatants were collected, clarified by centrifugation, and stored at −80◦C until analyzed. Secreted cytokine/chemokines were quantified using a Bio-Plex ProTM Mouse Cytokine, Chemokine, and Growth Factor custom 9-plex assay (Control#64145335). The assay was run per manufacturer's instructions, using 50 µl of the supernatant. Concentrations were calculated from standard curves. The kit contained M1 markers: IL12 p40, IL-1β, IL6, IL12 p70, and TNF-α, the M2 marker IL10, and the three chemokines: CCL1/KC (neutrophil chemoattractant), CCL2/MCP-1 (monocyte chemoattractant protein), and CCL5/RANTES (T cell homing factor). The absolute protein release as well as the release normalized to non-polarized controls was determined.

# Statistical Analysis

The number of subjects required for each of the assays were estimated using power analysis (set at 80% and α = 0.05) using GPower 3.1 (Dusseldorf, Germany). Statistical significance was determined by a repeated measures analysis of variance (ANOVA). The repeated measures ANOVA was conducted with the within effect of polarization state nested within macrophage type and macrophage type as a between animal effect. For **Figure 6**, each antigen was quantified by flow cytometry, MFIs for each animal under each condition (no treatment, IFNγ, and IL4/IL13) were averaged and the deviation from that mean was calculated thus effectively removing the animalto-animal variability and considering only the within-animal response, analogous to a repeated measures design. Protein release (**Figure 9**) was treated similarly. Statistical significance was accepted at p < 0.05. Data are reported as p-values, with statistical significance accepted a p < 0.05. For qPCR and protein release (**Figures 7**–**10**), the data are reported as mean ± SEM. As the qPCR data were not normally distributed, a Generalized Linear Transform was used to fit the data to estimate parameters subsequently used in the ANOVA. For polarization (**Tables 2**– **4**), statistically significant difference based on cell type alone are shaded green, those based on polarization alone are shaded blue, and interaction differences (i.e., BMDM and pMACs respond differently to polarization) are shaded orange. Animal numbers ranged from 4 to 7, with n referring to the number of BMDM and pMACs tested. Statistical analysis was performed using Minitab (State College, PA).

It is recognized that the presentation of the data in **Figures 7**– **11** is unfamiliar. This is due to the complexity of the data. The data assessing how BMDMs and pMACs respond to polarization are quite complex as there are two factors: cell type and polarization state. Each factor has a minimum of two levels (cell type: BMDM and pMAC and polarization state: none, IFNgamma, and IL4/IL13). To properly assess the effect of each of these factors and their levels, both the main effect of each factor and the interaction between the two factors is assessed, as the response to polarization stimuli may vary depending on the cell type.

The statistics in **Tables 2**–**4** correspond to **Figures 8**–**11**. They include the main effects of cell type and polarization state and the interaction effects which were computed using a Generalized Linear Transform (as opposed to a linear transformation, e.g., classical multiple regression) since the relationships between cell type and polarization state was hypothesized to be non-linear. Additionally, each biological replicate was considered to be cells from one animal, and due to animal-to-animal variability, the magnitude of responses sometimes differed between animals. To assess the effect of polarization stimuli while removing the animal-to-animal variance, changes in the polarization state of cells from each animal were analyzed using a repeated measures ANOVA where each line represents the response of cells from a single animal/biological replicate (**Figures 6**, **9**).

# DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

# ETHICS STATEMENT

The animal study was reviewed and approved by Albany Medical Center Institutional Animal Care and Use Committee.

# AUTHOR CONTRIBUTIONS

CZ and AZ contributed equally to this work. They performed the majority of the experiments, collected and analyzed data, contributed to the statistical analysis, generated many of the figures, figure legends, materials and methods, their results, and contributed to the discussion. GM and TN did some experiments, analyzed results, generated the figures, figure legends, and materials and methods. PF and SD, biostatisticians, worked with CZ, AZ, and ML to determine the best statistical methods and drove the statical analyses. RG contributed to intellectual development of the project, experimental design, and writing. ML initiated the study, coordinated the experiments, did experiments, analyzed data, wrote the manuscript and worked with PF and SD on the statistics.

# FUNDING

This work was supported by The Johnathan R. Vasiliou Foundation and an Albany Medical Center Bridge Grant (ML), an NSF Graduate Research Fellowship DGE-1247271 (AZ), NIH R01 NS092754 (RG) and New York State Spinal Cord Injury Research Board (NYSSCIRB) Institutional Support Grant (C32245GG) (RG).

# ACKNOWLEDGMENTS

Regeneron Pharmaceuticals (Tarrytown, NY) supplied the FcγRI only mice used in **Figure 5**. The authors wish to thank Regeneron for their generosity and especially acknowledge Dr. Lynn Macdonald, Amanda Pasquale, and Karoline Meagher for facilitating the transfer of the mice. The authors wish to thank Ms. Rebekah Pierce and MDDR student Alexandria M. Sadasivan for doing the pilot studies that led to this work. Also, Deborah Moran and Rosemary Naftalis for their administrative assistance with the figures and text and Dr. James Drake for critical reading of the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.00269/full#supplementary-material

Supplemental Figure 1 | LPS does not synergize with IFN-γ with respect to the respiratory burst or cytokine secretion. BMDM were treated with LPS (100 ng/ml), IFN-γ (100 ng/ml) or both for 24 h; controls received neither. The media was

#### REFERENCES


collected for quantitation of IL-10 and IL-12 by ELISA (A) and the respiratory burst quantified using the Amplex Red® fluorescence assay in response to zymosan or IgG-opsonized zymosan (B). When treated with IFN-γ, BMDM and pMACs released equivalent concentrations to IL-10/IL-12 and produced the same amount of oxidized Amplex Red regardless of the presence of LPS. Thus, IFN-γ (100 ng/ml) was used for M1 polarization in these studies. Each symbol represents cells from one mouse. <sup>∗</sup>p < 0.05, ∗∗p < 0.005 compared to no treatment. p-values were determined using one-way ANOVA and Tukey's test.

for experimental metastatic B16 melanoma. J Immunol. (2012) 189:5513– 7. doi: 10.4049/jimmunol.1201511


**Conflict of Interest:** 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 © 2020 Zajd, Ziemba, Miralles, Nguyen, Feustel, Dunn, Gilbert and Lennartz. 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.

# Microglial Corpse Clearance: Lessons From Macrophages

Mar Márquez-Ropero1,2†, Eva Benito1,2,3†, Ainhoa Plaza-Zabala1,2 and Amanda Sierra1,2,3 \*

<sup>1</sup> Achucarro Basque Center for Neuroscience, Parque Científico, University of the Basque Country (UPV/EHU), Leioa, Spain, <sup>2</sup> Department of Neuroscience, University of the Basque Country (UPV/EHU), Leioa, Spain, <sup>3</sup> Ikerbasque Foundation, Bilbao, Spain

From development to aging and disease, the brain parenchyma is under the constant threat of debris accumulation, in the form of dead cells and protein aggregates. To prevent garbage buildup, the brain is equipped with efficient phagocytes: the microglia. Microglia are similar, but not identical to other tissue macrophages, and in this review, we will first summarize the differences in the origin, lineage and population maintenance of microglia and macrophages. Then, we will discuss several principles that govern macrophage phagocytosis of apoptotic cells (efferocytosis), including the existence of redundant recognition mechanisms ("find-me" and "eat-me") that lead to a tight coupling between apoptosis and phagocytosis. We will then describe that resulting from engulfment and degradation of apoptotic cargo, phagocytes undergo an epigenetic, transcriptional and metabolic rewiring that leads to trained immunity, and discuss its relevance for microglia and brain function. In summary, we will show that neuroimmunologists can learn many lessons from the well-developed field of macrophage phagocytosis biology.

#### Edited by:

Esther M. Lafuente, Complutense University of Madrid, Spain

#### Reviewed by:

Zsuzsa Szondy, University of Debrecen, Hungary Raymond B. Birge, Rutgers, The State University of New Jersey, United States

> \*Correspondence: Amanda Sierra

amanda.sierra@achucarro.org

†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: 08 January 2020 Accepted: 05 March 2020 Published: 27 March 2020

#### Citation:

Márquez-Ropero M, Benito E, Plaza-Zabala A and Sierra A (2020) Microglial Corpse Clearance: Lessons From Macrophages. Front. Immunol. 11:506. doi: 10.3389/fimmu.2020.00506 Keywords: microglia, macrophages, phagocytosis, apoptosis, efferocytosis, epigenetic, metabolism, trained immunity

# INTRODUCTION

Microglial phagocytosis of apoptotic cells (efferocytosis) is at the core of the brain regenerative response. Its relevance in maintaining brain tissue homeostasis from development to aging and neurodegenerative diseases is undisputed but, nonetheless, many fundamental questions about the biology of the process remain open: How is phagocytosis efficiency regulated at the cellular and molecular levels? How can it be manipulated? Is it a dead-end road or does it trigger changes in the phagocyte? Is phagocytosis simply a process of garbage removal or does it actively participate in the well-being of the surrounding tissue? We here try to address these questions by collecting answers from microglia's cousins, the macrophages that reside in other tissues. In the first part of this review, we will discuss similarities and differences in the identity of microglia and other macrophages, taking into account their developmental origin and the maintenance of the adult populations. In the second part, we will compare the mechanisms that mediate recognition and engulfment and their epigenetic, transcriptional, metabolic, and immunological consequences. We will conclude that phagocytosis has a tremendous potential to impact on brain physiology and pathology.

## LINEAGE AND ORIGINS OF MICROGLIA AND OTHER MACROPHAGES

Microglia are brain-resident macrophages. They interact with the brain parenchyma and carry out essential maintenance functions. They are often studied independently from other tissue-resident macrophages, probably because they are unique in some aspects, most notably in their isolation from the rest of the body through the blood brain barrier (BBB). But how different are microglia really from other tissue resident macrophages in terms of origin, lineage, and identity? A close review of the literature shows that microglia are not as coarsely distinct to other macrophages as one may think, yet there are some fine differences in how they behave in their local environment. In the next sections, we will review evidence about the origin, lineage, identity, and population dynamics of microglia compared to other tissue-resident macrophages and highlight commonalities and differences.

## Lesson 1. The Monocytic Origin of Macrophages Is More the Exception Than the Rule

Macrophages are widespread and reside in many different organs, where they fulfill different functions. Because it is such a diverse population of cells, a fundamental question is whether they have a common precursor or whether each macrophage population develops from a different precursor. Based on the observation that some macrophages are short-lived and that they are often renewed by circulating monocytes, already in 1972 van Furth et al. proposed the "mononuclear phagocyte system" theory, by which tissue-resident macrophages were assumed to derive from blood-circulating monocytes and to differentiate within the host tissue (1). However, at that time there was evidence that some macrophage populations can proliferate locally and self-renew (2, 3), which clashed with the idea that macrophages are of monocytic origin. Why would they need to divide and self-renew locally if the main source is in fact a pool of circulating monocytes? Over the last decade, tracing and parabiosis experiments, where the origin and location of specific macrophage populations can be followed over time, have demonstrated that in fact most macrophage populations originate in the embryo prenatally. And it is only some macrophage populations, such as gut (4), dermis (5), pancreas (6), and heart (7) macrophages that are replaced from circulating monocytes. In contrast, other macrophage populations, most notably microglia and skin macrophages (Langerhans cells), but also others, originate early during development, are long-lived and capable of local self-renewal with no significant involvement of circulating monocytes (8–12).

The notion that macrophages have a prenatal and not a monocytic origin caused a paradigm shift in the field and prompted questions about their precise site of origin and about the routes by which they reach target organs. The precise spatial origin of each macrophage pool is still under debate, but there are essentially two accepted sources: the yolk sac and the aortagonad mesonephros. From there, macrophages take one of two routes: they migrate directly to their target organ or they go through fetal liver and then on to their target tissue (**Figure 1**). What percentage of the mature tissue-resident macrophages in fact originate via each of these routes is still being investigated for most macrophage populations. In this respect, the origin of microglia is probably the most undisputable: they originate in the yolk sac and migrate directly to the brain as early as embryonic day E10.5 in the mouse. Other macrophages, such as Langerhans cells seem to have a mixed origin: some derive directly from the yolk sac, whereas others travel through the fetal liver before they reach their destination (9). The latter route through fetal liver seems to be the standard and more prevalent for most other macrophage populations (13, 14).

In summary, the original theory that tissue-resident macrophages universally derive from circulating monocytes has now been replaced with a more refined picture, where macrophages in fact originate from an early embryonic precursor, seed their target tissue, and self-renew locally (12). In this new picture, macrophages that derive from circulating monocytes are more the exception than the rule and microglia probably constitute the paradigmatic example of macrophages of prenatal origin with self-renewal capacity and no exchange with circulating monocytes under physiological conditions.

But if monocytes are not the precursor cells for most macrophages, what is in fact the molecular identity of their precursors and how different is it for different macrophage populations? In the next section, we will summarize what is known about the identity of the embryonic precursors that give rise to different tissue-resident macrophages.

#### Lesson 2. Microglia Do Not Require the Transcription Factor c-Myb to Develop, but Other Macrophages Do

After it was demonstrated that most macrophages originate prenatally either from the yolk sac or from the aorta-gonad mesonephros, investigators turned their attention to the precise molecular identity of the precursors. The goal was to identify markers and transcription factors that would allow us to identify and classify them.

There is a certain correlation between the physical and temporal origin of macrophages and the markers they express. The most polarizing marker, the one that probably allows the most straightforward classification, is the transcription factor c-Myb. Early (E7.5-8.5 in mice) progenitors from the yolk sac (erythromyeloid precursors or EMPs) do not express or require the transcription factor c-Myb for development and maturation. Accordingly, macrophages that stem from EMPs, such as microglia, Langerhans cells and Kupffer cells (liver macrophages), develop normally in c-Myb−/<sup>−</sup> animals (10). In contrast, late (E10.5 in mice) progenitors that originate in the fetal liver (hematopoietic stem cells, HSCs) do express and require c-Myb for proper development (15, 16). This is the case for most tissue-resident macrophages, which fail to develop in the absence of c-Myb (10). c-Myb dependence is therefore a classifying criterion for macrophage lineage, but it is not the only transcription factor involved in maturation. The overall topic of macrophage lineage markers is beyond the scope of this review and has been covered extensively before (13). For microglia in particular, key players in their development include the growth factor receptor CSF1R (Colony Stimulating Factor 1 Receptor), the chemokine receptor CX3CR1 (CX3C chemokine receptor 1), the calcium binding protein Iba-1 (ionized calcium

self-renewal (circular arrows) vs. exchange with circulating monocytes (double-sided arrows) to that particular population. Microglia and Langerhans cells for instance are mainly self-renewing, whereas cardiac and colonic macrophages do exchange significantly with circulating monocytes. Some vector graphics were obtained from Vecteezy.com with permission.

binding adaptor molecule 1), the G-protein-coupled receptor F4/80 (also known as EMR<sup>1</sup> - EGF-like module-containing mucin-like hormone receptor-like 1) and the integrin CD11b (cluster of differentiation molecule 11B, also known as Integrin Alpha M - ITGAM) (17, 18) .

The lesson that derives from these studies is that ontogeny and molecular identity of different macrophage populations go hand in hand. There is an additional component, longevity, which also shows some level of correlation with macrophage ontogeny. Early, c-Myb-negative progenitors in the yolk sac give rise to populations that tend to be long-lasting (microglia, Langerhans and Kupffer cells), and later, c-Myb-positive cells in the mesonephros and fetal liver give rise to a mix of longand shorter-lived populations. Microglia are particularly longlived and have the ability to self-renew several times over a lifetime (19). A question that derives from this notion is how the population behaves with respect to individual cells. What is the natural "life cycle" of a single microglia and how does it contribute to the overall stability of the population? The next section will examine evidence of how microglia are replenished under homeostatic conditions.

#### Lesson 3. As Long as the Blood-Brain Barrier Is Intact, Microglia Can Self-Renew Throughout a Lifetime Without a Significant Monocyte Contribution

As reviewed in the previous section, under physiological conditions, microglia have a prenatal yolk sac origin. The BBB protects and isolates the brain from exchange with blood circulation, effectively turning it into a very isolated microenvironment. Here, microglia stay stable and self-renew with virtually no contribution from circulating monocytes in physiological conditions (8). But what is the exact population dynamics of microglia and how do they behave individually? And what happens if microglia are depleted in a healthy brain, how do they replenish under these conditions?

One critical factor in microglial development and survival is CSFR1. Mice devoid of this receptor do not have microglia (8, 20), and mutations that affect the function of the kinase domain of the receptor cause severe neurological syndromes in mice and humans (21–24). In line with this, in 2014, Elmore et al. showed that a chemical inhibitor of CSF1R virtually ablated the microglial population in the healthy brain. Strikingly, they also showed that microglia could be repopulated upon withdrawal of the inhibitor (25). It was then critical to identify which cells microglia repopulated from. The original report suggested the existence of a "hidden" microglial progenitor expressing the intermediate filament nestin (25). However, more detailed studies have in fact demonstrated that the most likely scenario is one where microglia replenish from the few remaining cells that are not removed during CSF1R inhibition (19, 26). A few questions are left unanswered regarding how this repopulation happens, including why some cells remain resistant to CSF1R inhibition and what signaling mechanisms tell microglia that they need to stop proliferating once the repopulation is completed. In this regard, it is interesting to note that microglia seem to have a very active population dynamics at steady-state, with a balanced ratio of proliferation and apoptosis and several self-renewal cycles during a lifetime (19).

A similar scenario is observed in Langerhans cells, which also have a predominant yolk sac origin, are long-lived (9), renew during a lifetime (27) and most likely proliferate from fully differentiated cells (28). Indeed, proliferating, selfrenewing macrophages can be found in almost all tissues under physiological conditions (11, 29), and often even more so under pathological conditions, although in the latter case the contribution from circulating monocytes is most likely instrumental (30, 31). Only when the BBB integrity is compromised, most commonly after head irradiation, circulating monocytes can give rise to functional microglia [for a review see (31)]. Nonetheless, the fact that a population of cells exists that has the potential to differentiate into microglia under certain pathological conditions is highly relevant for therapeutic intervention and merits further research.

Overall, macrophages seem to be in charge of their own local population dynamics and reach a steady state of functional self-renewal with balanced division/death rates that match the functions they carry out in their host tissues.

### Lesson 4. Microglia Have Specialized in Homeostatic and Stimulus-Triggered Remodeling of Brain Circuits and Structure

The original notion that macrophages are a uniform pool of phagocytes that reside in all organs and perform immune functions has long been discarded. Instead, the current view is that macrophages are essentially as dissimilar as their target tissues and have specialized in performing tissue homeostasis functions that make sense in their local microenvironment. For instance, whereas lung macrophages are specialized in surfactant clearance (32) and adipose macrophages participate in lipid and insulin metabolism (33), spleen macrophages are critical for iron homeostasis (34, 35). In the particular case of microglia, their specialized functions include interactions with neurons, potentially participating in synaptic remodeling (36), modulating experience-triggered events, such as neurogenesis (37), and participating in memory acquisition (38, 39). Microglia are also very active in surveilling the brain parenchyma with their highly motile processes (40, 41) and this is believed to be essential for brain maintenance and protection and to be in close connection with neuronal activity (42, 43). However, microglial motility may not be so unique, as late evidence suggests that other macrophages might also perform this type of surveillance in the intersticial space (44).

It is not surprising that macrophages have specialized roles within their host tissues to aid their homeostasis and maintenance. In this context, microglia need to be in close contact with neurons and be highly sensitive to changes in the microenvironment. In the next section, we will focus on this sensitivity to environmental factors and how that may impact the epigenetic and transcriptional identity of microglia.

## Lesson 5. Microglia Have Developed a Unique Transcriptional and Epigenetic Landscape

Given the wide diversity of macrophage tissue-specific phenotypes, a very relevant question to ask is how different macrophages are at the molecular level. How different are their transcription and epigenetic profiles under physiological conditions and how do they respond to changes in their environment? With the advent of "omics" methods and, in particular, with the application of single-cell studies the field has experimented a boom. It is now clear that macrophages are as different at the transcription and epigenetic level as they are at the phenotypic level. Macrophages share a common core transcriptional identity, but each tissue-specific population expresses a unique set of transcripts that is organ-specific (45, 46). In line with this, microglia are transcriptionally distinct to other macrophage populations but, in addition, single-cell studies have shown that there are several subpopulations of microglia that could perform functionally different tasks under basal and inflammatory conditions (47, 48). Similar observations are being made in other macrophage populations, with singlecell omics techniques revealing their underlying richness and diversity (49, 50).

In the particular case of microglia, these studies have revealed that microglia are transcriptionally homogeneous across brain regions but diverge in their transcriptional profiles across their cell division state (48). Another distinct microglia subpopulation that has earned a fair bit of attention is the so-called Disease-Associated Microglia (DAM), a subgroup of microglia revealed by single-cell sequencing specifically in the brain of an Alzheimer's disease mouse model (47). However, despite the power of these approaches in revealing the heterogeneity of distinct microglia subpopulations, there is debate as to the relevance of these populations in human brain, which seem to have subpopulations of their own (see comment in https://www.alzforum.org/news/research-news/ when-it-comes-alzheimers-disease-do-human-microglia-evengive-dam). Ultimately, and under the premise that to study human disease one should focus on human microglia as much as possible, these studies reveal that microglia are in fact more diverse than anticipated and encourage further investigation aiming at dissecting their identity.

An ever-expanding source of cell diversity lies in the epigenetic landscape, which is most commonly read in the form of histone modifications (acetylations, methylations), DNA modifications (DNA methylation) and small non-coding RNAs (miRNAs, lncRNAs), and which has also been probed in different macrophage populations. The combinatorial patterns of histone 3 acetylation (at lysine 27) and methylation (at lysine 4) have been profiled in detail, revealing a core epigenetic macrophage signature, as well as a set of specific enhancer marks that are unique to each organ (46). This study also served to validate the idea that local macrophage identity is shaped by the local microenvironment. Previous experiments had shown that some macrophages populations are highly plastic and can adopt a tissue-specific identity even after ex vivo culture (51, 52). Recent evidence suggests that this phenotypic identity reshaping is accompanied by rewiring of the epigenetic landscape (46). In this context, microglia proved to be transcriptionally and epigenetically distinct to all other tissue-resident macrophages and clustered systematically separated from other populations for different epigenetic marks (46).

A follow-up question that emerges from these observations is to which extent epigenetic reprogramming is relevant for macrophage biology, and how it may affect microglia in particular. Microglia depletion studies in diseased brains have shown that after replenishment with "fresh" microglia, animals have improved cognitive scores (53–55). These studies have played with the idea that microglia in a diseased brain may have an altered epigenetic signature that is erased at the time of depletion and that the replenishing microglia are reprogrammed and therefore functionally fitter. How much of the original epigenetic identity is retained by the few remaining microglia and what role exactly cellular reprogramming plays in microglia phenotype is still unclear. Nonetheless, microglia and other tissue-resident macrophages appear to have a unique transcriptional and epigenetic signature that is highly plastic and may play a fundamental role in how they interact with their environment.

Since the molecular and phenotypic identities of any cell are in constant interplay, it is expected that different macrophages with different tissue-specific functions and phenotypes would have different transcriptional and epigenetic signatures. A more thorough characterization of the effect of environmental factors on cellular reprogramming and function will shed light on how dynamic these changes are and, most importantly, how relevant under physiological and disease situations. However, in spite of their phenotypic diversity, macrophages do share some core functions, such as phagocytic clearance of dead cells and immune signaling. Yet even those have tissue-specific nuances. The next section will address the similarities and disparities in mechanisms and dynamics of phagocytosis in microglia vs. other macrophages.

#### PHAGOCYTOSIS BY MACROPHAGES AND MICROGLIA

The efficient phagocytic removal of apoptotic debris (efferocytosis), mostly carried out by resident macrophages, vastly influences our daily physiology. It is estimated that billions of cells are removed everyday (56): aged erythrocytes and neutrophils in the spleen, liver and bone marrow (57, 58), epithelial cells in the mammary gland after the lactation period (59), spermatogenic cells in the testis (60), and the outer segment of light-exposed photoreceptors in the retina (61), among others. In the brain, microglia remove the excess newborn cells produced during embryonic and postnatal development in the cortex, cerebellum and amygdala (62–64) and in adult neurogenic niches in the hippocampus and subventricular zone (SVZ) (37, 65), as well as deceased cells during aging and neurodegenerative diseases (66). Other cell types, such as astrocytes, neuroblasts or cells of the neural crest may also act as phagocytes but generally with less efficiency and thus microglia are considered the brain "professional" phagocytes (67).

In this section, we will compare available data on the process of phagocytosis by macrophages and microglia. We will then describe that phagocytes express overlapping recognition mechanisms of apoptotic cells, and that this redundancy ensures that phagocytosis is fast and efficient. We will then describe the functional consequences of ingestion in the phagocyte, focusing on inflammatory responses, and metabolic adaptations. Ultimately, we will conclude that phagocytosis is a powerful mechanism that needs to be firmly controlled to ensure the efficient removal of target cells while preventing the demise of live cells.

#### Lesson 6. Phagocytosis Is Tightly Coupled to Apoptosis Due to Redundant Recognition Mechanisms

Phagocytosis and apoptosis are evolutionarily linked together as a mechanism that allows the elimination of excess, dysfunctional, or aged cells without imposing alterations or damage in the surrounding cells (67). Homeostatic phagocytosis is ensured by the immediate recognition of the apoptotic cell by phagocytes through a redundant plethora of released "find-me" and membrane-bound "eat-me" signals that are recognized by the phagocytes (**Table 1**). Herein we summarize some of the most relevant signals and receptors involved in apoptotic cell recognition, and we prompt the reader to comprehensive reviews on the topic for more details (67, 72).

Among "find-me" signals stand out purines (ATP, UTP, and their dephosphorylated derivatives) and the chemokine fractalkine, and the corresponding metabotropic purinergic P2Y receptors and CX3CR1 expressed in phagocytes. Another important "find-me" signal is the lipid lysophosphatidylcholine TABLE 1 | Expression of phagocytosis-related receptors by microglia.

(http://rstats.immgen.org/Skyline/skyline.html);

proliferative-region

 associated microglia (PAM)(48).

 (Down)  a human database of microglia (Mi), macrophages

 microglia (DAM) (47); a database of genes specific of a human aging signature (HuMi-aged)

 (Up) or downregulated 


Márquez-Ropero et al.

The table indicates the name of the receptor, the gene ID, its function and the expression in different cell types found in RNAseq public databases: a brain-specific

database of microglia and other immune cells

database of genes upregulated

disease-associated

(Down) in  (Up) and downregulated

 microglia (71); and a

 (70); a database of genes enriched in cerebellar

 database of microglia compared to other brain cells (68); an immunological

(non-phagocytic)

 and striatal (phagocytic)

 (Ma), and monocytes (Mo) (69); a database of genes upregulated (LPC), which binds to GPR132 (G2A) receptors in phagocytes. The best known "eat-me" signal is the lipid phosphatidylserine (PS), with its many phagocyte receptors, such as BAI1, TIM4 and stabilins. PS is also indirectly recognized by several bridgemolecules, including MFG-E8 (Milk-fat globule E8), as well as Protein S and Gas6, which bind to members of the TAM (Tyro3, Axl, Mer) family in phagocytes. In addition, other wellknown receptors classically involved in immune responses are the inflammation triggering receptors expressed on myeloid cells TREM2 and CD300B; and integrins and complement receptors such as CD11b, which recognize opsonized apoptotic cells both directly and some of them indirectly via MFG-E8. More recently, three sugar-related receptors have been involved in phagocytosis: the mannose receptor CD206 (MRC1), which is upregulated in bone marrow and intestinal phagocytes compared to non-phagocytic macrophages and predicts their phagocytic performance (73). The sialic acid binding protein CD22, a negative regulator of phagocytosis identified in an in vitro screen that is upregulated in microglia in the aging brain (74). And the beta glucan receptor Clec7a, associated to the phagocytosis of oligodendrocytes by microglia during early postnatal development (48). Each type of phagocyte expresses their own set of receptors (67) and microglia is no exception.

The most conspicuous phagocytosis receptors expressed in both mouse and human microglia are CX3CR1, P2Y6, P2Y12, stabilin 1, SIRPα, TREM2, MerTK, and CD11b, based on brain RNASeq databases (68, 69) and the immunological RNASeq database Skyline (http://rstats.immgen.org/Skyline/skyline.html) (**Table 1**). However, it is important to note that each of these receptors is dynamically expressed during aging, disease and in different brain areas (47, 48, 70, 71) (**Table 1**), and each may serve different functions. For instance, in macrophages, MerTK and Axl show opposite regulation of their expression during inflammatory conditions, and specialize in different types of phagocytosis: MerTK in homeostasis, Axl during inflammation (75). Similarly, CD206, MerTK, and TIM4 (but not other receptors) are upregulated in peripheral macrophages early upon phagocytosis (73). In microglia, the expression of these receptors is not a proxy of their phagocytic performance either, as some including CX3CR1, P2Y12, and MerTK, have higher expression in striatal than in cerebellar microglia, whereas phagocytosis seems more prominent in the cerebellum than in the striatum (71). Similarly, their expression is also not correlated with PAM (proliferative-region- associated microglia), which has been involved in phagocytosis of developing oligodendrocytes during early postnatal development (48). Moreover, the efficiency of phagocytosis does not rely on their expression alone, but also on the motility of the microglial processes, the relative distribution of microglia compared to apoptotic cells, or the lysosomal efficiency, to name a few other factors. Their expression, however, has been linked to functional changes in microglia. For instance, some of these receptors, such as CX3CR1 and P2Y12, are part of the so-called "homeostatic microglia signature," and are downregulated in DAM (47) but upregulated in aging microglia (70). Therefore, the expression of phagocytosis receptors is not invariably linked to microglial phagocytosis efficiency. A careful analysis of the subsets of receptors expressed by microglia in different regions across the lifespan and during specific disease will shed light into their functional implication in microglial phagocytosis.

Finally, as microglia is specialized in surveilling the brain parenchyma, several receptor systems involved in the recognition of apoptotic cells have also been "hitchhiked" by other brainspecific cargos. For instance, TREM2 has been deeply involved in the recognition of beta amyloid extracellular deposits in the context of Alzheimer's disease (AD) (76), and is one of the main genetic risk factors for its development (77). The complement receptor CD11b (78, 79) and the fractalkine receptor CX3CR1 (36) are used to recognize dendritic spines and are related to synaptic prunning. SIRPα recognizes the "don't-eatme" signal CD47 on myelin debris and on active synapses, inhibiting their phagocytosis (80, 81). CD22 binding to sialic acid inhibits the phagocytosis of extracellular deposits of beta amyloid, myelin and α-synuclein during aging (74). However, to which extent the phagocytosis of beta amyloid, α-synuclein, spines, and synapses, or myelin debris phenocopies the complete process of efferocytosis, including attraction, engulfment, and degradation, remains to be determined.

### Lesson 7. Phagocytosis Is Fast and Apoptosis Is Silent

A consequence of the tight coupling between apoptosis and phagocytosis is that phagocytes are not easily caught red-handed and, consequently, apoptosis is largely underestimated. Original experiments in the thymus, where T lymphocytes undergo chromosomic rearrangement in their antigen receptor genes, suggested that the negatively selected cells were removed but few apoptotic cells were found. It was only when phagocytosis was analyzed that it became evident that apoptotic T lymphocytes were quickly removed by resident macrophages (82), leading to the suggestion that phagocytosis lasts minutes (72). Similarly, taking into account the number of erythrocytes removed daily in the spleen and liver, and the number of resident macrophages in these tissues (pulp macrophages and Kupffer cells, respectively), phagocytosis has been estimated to last under 30 min (67). Microglia are comparably fast. Using as a model the adult neurogenic cascade of the hippocampus, where newborn cells naturally undergo apoptosis and are phagocytosed by microglia, the estimated average clearance time of an apoptotic cell is 90 min (37).

A corollary of this data is that phagocytosis efficiency determines the amount of apoptosis visualized. At a given time, the number of apoptotic cells found can be conceived as a black box with doors on each side: an incoming door that represents the cells that enter apoptosis de novo; and an outgoing door that represents the cells that are removed via phagocytosis. Using this analogy is easy to understand that the size of the pool of apoptotic cells depends on the relative velocities of the two processes, apoptosis and phagocytosis (**Figure 2**). Therefore, in physiological conditions apoptotic cells are difficult to observe because microglial phagocytosis is very efficient (37). In contrast, in pathologies like epilepsy, the increased number of apoptotic cells in early stages is not due

apoptosis induction, and no apoptotic cells should be observed. (B) In most cases, apoptosis induction is faster than engulfment and degradation, and only some apoptotic cells would be observed, whereas others are already cleared by microglia. (C) In a dysfunctional scenario with deficient phagocytosis, all cells undergoing apoptosis would be visible.

to apoptosis induction, but to phagocytosis impairment and accumulation of non-removed apoptotic cells (42). In conclusion, phagocytosis efficiency determines the dynamics of apoptosis during development and disease.

## Lesson 8. Phagocytosis Is Immunomodulatory at the Epigenetic, Transcriptional and Posttranslational Levels

The tight coupling between apoptosis and phagocytosis has functional relevance, as it prevents the release of intracellular contents in physiological conditions. In contrast, during trauma, uncontrolled and unexpected cell death is usually followed by infection by microorganisms, and the dead cells release damage-associated molecular patterns (DAMPs), such as DNA, RNA, nucleotides, or chromatin proteins such as HMGB1 (high mobility group box 1 protein). These signals are initially recognized by the fluidic branch of the innate immune response (i.e., the complement and the coagulation systems), followed by recognition of specific pattern recognition receptors (PRRs) in several types of leukocytes. This cascade of events initiates a complex immune response that includes chemokine and cytokine release, release of reactive oxygen species and other responses to heal the damaged tissue and kill invading microorganisms (83). Unlike cell death caused by trauma, programmed cell death during physiological events ensures that DAMPs are contained within membranous blebs (the apoptotic bodies) and do not trigger activation of PRRs (84). The efficient coupling between apoptosis and phagocytosis avoids the development of secondary necrosis and release of DAMPs as apoptosis progresses, as well as the initiation of an inflammatory response from the immune system (84). As a result, apoptotic cell removal via phagocytosis is largely anti-inflammatory or at least immunomodulatory (67, 85), although the inflammatory responses of different types of macrophages are indeed heterogeneous (73).

In microglia, the evidences showing that phagocytosis of apoptotic cells is immunomodulatory are more tenuous than in other macrophage populations. Classic in vitro experiments showed that cultured microglia exposed to apoptotic cells express dampened responses to inflammatory stimuli such as bacterial lipopolysaccharides (LPS) (86, 87). In vivo, phagocytosis blockade induced by seizures in a mouse model of epilepsy correlated with a pro-inflammatory profile in microglia (42). Similarly, restoring phagocytosis in the aging brain using an anti-CD22 therapy reduced the microglial expression of inflammatory and disease-associated genes (74). However, the molecular mechanisms linking efferocytosis and inflammation are still unclear: is inflammation triggered by recognition of surface receptors or by downstream mechanisms related to the cargo degradation? Is it regulated at the epigenetic, transcriptional or the translational levels?

In both macrophages and microglia, these effects are at least partially mediated by apoptotic cell recognition via the complement protein C1q, whose presence turns efferocytosis anti-inflammatory (88, 89). Indeed, some of the transcriptional changes associated to efferocytosis occur while macrophages are still early in the process of phagocytosis, as shown by experiments comparing macrophages containing labeled dead cells and macrophages without apparent cargo (phagocytic and not phagocytic, respectively) (73). Similarly, posttranslational modifications related to the reduced release of the major proinflammatory mediator interleukin 1 beta (IL-1β) are related to the inhibition of the NLRP3 (NLR family pyrin domain containing 3) inflammasome triggered by mere contact with apoptotic cells in the absence of effective engulfment (90).

In addition to these early changes, it is also likely that the metabolic rewiring associated with apoptotic cell degradation (91) may trigger a "metabolite storm" in the phagocyte that would further contribute to regulate its function at later time points. For Márquez-Ropero et al. Microglia and Macrophage Phagocytosis

instance, internalization of the apoptotic cell, but not surfaceto-surface interaction, triggers in macrophages another set of transcriptional anti-inflammatory changes through a chloride channel, Slc12a2, and chloride-sensing kinases (92). Timedependent transcriptional changes have in fact been observed in cultured microglia upon phagocytosis of apoptotic cells (93). While some genes are transiently regulated at 3 h after engulfment, most are regulated in the post-degradation phase at 24 h after engulfment. This second wave of transcriptional changes is likely related to epigenetic mechanisms, as in fact chromatin remodeling related genes are upregulated in the late phagocytic microglia (93). In agreement, microglial phagocytosis is related to altered epigenetic and transcriptional profiles in vivo, as shown by comparing cerebellar and striatal microglia (i.e., phagocytic and non-phagocytic, respectively) (71). While this study analyzed epigenetic changes at the whole population level, it is likely that in brain areas with high basal apoptotic cell clearance, such as the cerebellum (71), the amygdala (64) or the hippocampus (37) microglia coexist in different stages related to phagocytosis. Even in these areas, it would be expected that at any given time point there would be non-phagocytic cells as well as cells engaged in different stages of engulfment, early degradation and late post-phagocytic events. This continuum of phagocytosis states is likely reflected on the epigenomic and transcriptional profiles of microglia.

### Lesson 9. Phagocytosis Alters the Phagocyte Metabolism and Function

Along with its epigenetic, transcriptional and immunomodulatory effects, phagocytosis also promotes metabolic adaptations that influence the phagocyte cellular function. In immune cells, the main metabolic pathways are catabolic (related to degradation and energy production) and anabolic (related to synthesis). Catabolic degradation of glucose starts with cytoplasmic glycolisis, which is responsible for glucose oxidation and produces pyruvate and lactate; and continues with the tricarboxylic acid cycle (TCA or Krebs cycle), which oxidizes pyruvate to produce reduced molecules (NADH, reduced nicotinamide adenine dinucleotide). NADH is also produced from the catabolism of fatty acids in the mitochondria through the beta-oxidation pathway. Next, NADH is completely degraded through the mitochondrial electron transport chain (ECT) during mitochondrial oxidative phosphorylation, to finally produce energy as ATP. Major anabolic pathways include the pentose phosphate pathway (PPP), which generates precursors of nucleotides; and the fatty acid and cholesterol synthesis pathway. The connection between metabolism and immune responses in the field of immunometabolism is very complex (94) and here we will focus on phagocyte's metabolic changes after inflammatory and phagocytic challenges.

Inflammatory stimuli trigger different types of metabolic adaptations. In pro-inflammatory conditions, macrophages need to act against infection and microorganisms and they undergo metabolic adaptations that meet the increased energetic demands while assuring cell survival. The most important change is a metabolic shift that potentiates glycolisis and downregulates oxidative phosphorylation, allowing faster, albeit less efficient, ATP production (95). In addition, several metabolic changes lead to the production of antibactericidal agents (96), such as reactive oxygen species (ROS) through increased PPP (97) and fatty acid synthesis (98); and mitochondrial ROS (mROS), through a disrupted ETC (99). Moreover, several metabolic pathways, as TCA cycle and fatty-acid synthesis are necessary for pro-inflammatory cytokine production, via HIF1α stabilization and through the fatty-acid synthesis regulator Laccase Domain-Containing Protein 1 (FAMIN), respectively (98, 100). In addition, fatty acid and cholesterol synthesis pathways also contribute to producing inflammatory mediators such as leukotriene B4 (LTB4) and isopropenoids, which bind the nuclear receptors peroxisome proliferator-activated receptors (PPARs) and the liver X receptor LXR (PLXR), respectively (101, 102). On the other hand, metabolic changes after anti-inflammatory stimuli help macrophages to resolve inflammation. In this case, upregulation of glycolisis and a proper TCA cycle contribute to the expression of an antiinflammatory phenotype (103, 104). In addition, increased oxidative phosphorylation, together with the promotion of fattyacid oxidation reduces the expression of pro-inflammatory cytokines (105). Thus, pro- and anti-inflammatory stimulation triggers different metabolic adaptations that contribute to modulate macrophage function.

Phagocytosis-induced metabolic adaptations are less known. In macrophages, phagocytosis of apoptotic cells drives a downregulation of fatty acid oxidation and de novo cholesterol, while promoting a metabolic shift that upregulates glycolisis and reduces oxidative phosphorylation (91, 106). In fact, phagocytosis triggers mitochondrial adaptations, such as mitochondrial fission (106) and decreased mitochondrial membrane potential via mitochondrial uncoupling protein 2 (UCP2) (107), which together with increased glycolisis (91) ensure the continued uptake of corpses. In addition, increased lactate release promotes the establishment of an anti-inflammatory environment (91). However, phagocytosis-induced metabolic adaptations not only have an effect during phagocytosis, but also trigger longlasting effects. There are several examples of how phagocytosis reprograms the function of macrophages. For instance, during Drosophila early development, naïve macrophages are insensitive to tissue damage or infection. However, upon corpse uptake macrophages become capable of migrating into damaged regions, and phagocytose bacterial pathogens, through the activation of the Jun kinase-signaling pathway and increased the expression of the damage receptor Draper (108). Similarly, phagocytosis of fungal β-glucan in mammalian macrophages drives a metabolic shift that contributes to an enhanced and nonspecific protection against infections known as trained immunity (109). In this response, macrophages upregulate glycolisis, glutaminolysis, PPP, and cholesterol synthesis, and decrease oxidative phosphorylation through the activation of the dectin-1–Akt–mTOR–HIF-1α signaling pathway. Metabolic adaptations lead to the increased production of metabolites such as fumarate and mevalonate, key for driving changes in histone acetylation and long-term epigenetic changes, which

Frontiers in Immunology | www.frontiersin.org

β-glucan phagocytosis.

lead to increased levels of pro-inflammatory cytokines after subsequent exposure to inflammatory challenges (110–112). Thus, phagocytosis of apoptotic corpses in macrophages triggers metabolic adaptations that immediately modulate cell function and also drives a long-term reprogramming of the phagocyte (**Figure 3**).

In microglia, the metabolic adaptations to inflammatory stimuli are well known and similar those of macrophages,

professional phagocytes, such as microglia. Under certain circumstances, stressed neurons (red) may be recognized and executed by nearby phagocytes, including microglia, before the apoptotic program has been fully engaged. In contrast to these two cases of canonical efferocytosis, in a third scenario, dysregulation of the "eat-me" and "don't-eat-me" signalization may lead microglia or other phagocytes to directly target for execution healthy, viable neurons in a process termed phagoptosis.

although their functional impact is less explored. Proinflammatory stimuli cause upregulation of glycolisis, leading to a speedy ATP generation that is crucial for the expression of pro-inflammatory cytokines (113). Other adaptations include impaired oxidative phosphorylation accompanied by mitochondrial fission (114, 115), increased glutamine entrance to the TCA cycle, suppressed fatty-acid oxidation and synthesis, and contradictory effects on the PPP (116). On the other hand, metabolic changes induced by anti-inflammatory factors in microglia are less known and show some differences compared to macrophages. Microglia exposed to anti-inflammatory stimuli maintain active both oxidative phosphorylation and PPP, and increase fatty-acid oxidation and synthesis while reducing glycolisis (116, 117). In contrast, little is known about the microglial metabolic adaptations to phagocytosis. Cultured phagocytic microglia upregulate genes related to metabolism and chromatin remodeling (93), suggesting long-term metabolic and phenotypic changes in microglia upon phagocytosis. Therefore, phagocytosis triggers a complex remodeling of the cell's metabolism and a "metabolite storm" that is likely to affect the function of microglia.

### Lesson 10. Does Phagocytosis Execute Cell Death?

Since apoptosis and phagocytosis are so closely related, the last question that arises is: when is cell death precisely executed? in vitro experiments with apoptosis inducers clearly show that the apoptotic pathways effectively progresses to kill the target cell via proteolytic degradation of intracellular contents by caspases (118). In this conventional scenario, called-upon macrophages would simply serve to dispose of the garbage. However, the tight coupling between apoptosis and phagocytosis also presumes a second scenario, in which expeditious nearby macrophages detect the first signs of cell distress and execute the latest stages of cell death. In these two cases, canonical efferocytosis/phagocytosis is used as a homeostatic mechanism and blocking engulfment would be expected to have detrimental consequences. In a third scenario, macrophages actually select which cells die and phagocytosis causes neuronal demise, through a process named phagoptosis (119) (**Figure 4**).

Discerning between the three scenarios is complicated by the phagocytosis-related downstream transcriptional and metabolic changes, which in turn feedback into the surrounding cells, affecting their survival and proliferation. For instance, liver phagocytic macrophages release VEGF (vascular endothelial growth factor) to support the proliferation of neighboring cells (120). Similarly, the secretome of phagocytic microglia acutely inhibits proliferation of neural progenitors, allowing their longterm maintenance and the preservation of neurogenesis. These feedback mechanisms are likely related to the compensatory proliferation induced by killer caspases during apoptosis ("apoptosis-induced proliferation," AiP), observed in some cell types and organisms (121). Because of these loops between death and life processes, the identification of phagoptosis must not be procedural, i.e., phagocytosis blockade increases survival, ergo, phagocytosis must kill cells. Instead, the discrimination between canonical phagocytosis and phagoptosis should be mechanistic and based on direct observation.

In macrophage biology, evidences of phagoptosis are in fact scarce and controversial. For instance, most literature claims that neutrophils die spontaneously by apoptosis after 24 h in circulation and are subsequently phagocytosed by bone marrow macrophages (67, 122). Others in contrast claim that neutrophils die by phagoptosis because phagocytosis blockade leads to more "alive" neutrophils (119), although in fact they are senescent and have impaired capabilities/migration (123). In spite of this controversy, macrophages are doubtless capable of activating the apoptosis program by direct contact through the so-called "death receptors," such as Fas and TNFR (tumor necrosis factor alpha receptors) (124). Similarly, deletion of engulfment genes in C.elegans increases the survival of cells treated with weak apoptotic stimuli, supporting that phagocytes execute death of stressed cells (125). Microglia execute bona fide canonical phagocytosis of apoptotic newborn cells in the adult hippocampus (37). They also engage in phagoptosis during inflammatory conditions that lead to dysregulation of the "eat-me" signalization. Dysfunctional and transient expression of PS by stressed neurons leads to their recognition and execution by microglia, and their survival when microglia is not present (126). Microglia have also been reported to kill neuroprogenitor cells during cortical development, an effect that was exacerbated in mice treated with LPS (62); and Purkinje cells in the developing cerebellum, through the production of radical oxygen species (127). However, it is not clear whether in these cases cell death was executed by phagocytosis. In the end, it is likely that canonical phagocytosis and phagoptosis are two sides of a spectrum of ways to die that depends on the fine balance between the different "find-me" and "eat-me" signals in each pair of apoptotic cell/phagocyte.

In summary, phagocytosis is a powerful double-edged sword that must be kept under a tight rein, as is exemplified by two recent papers in zebrafish and Drosophila. Phagocytosis of apoptotic cells is beneficial during brain trauma, as it prevents secondary damage spread in zebrafish larvae (128). In control larvae, the initial necrotic and apoptotic cells resulting from traumatic brain injury in the optic tectum are cleared by microglia within the first 24 h. When phagocytosis is pharmacologically and genetically disturbed by targeting PS and the zebrafish ortholog of PS receptor BAI1, a larger wave of secondary cell death spread over the brain (128). In contrast, overexpression of phagocytosis receptors Six-Microns-Under (SIMU) and Draper (Drpr), homologs of Stabilin2 and MEGF10 (Multiple EGF Like Domains 10, a complement receptor), respectively, in adult Drosophila leads to phagocytosis of live neurons, motor dysfunction and a shortened lifespan (129). These recent papers highlight the profound physiological impact of microglial phagocytosis on the survival of their surrounding neurons both in health and in disease.

Given the powerful influence of phagocytosis on tissue homeostasis, it may seem striking that few pathologies have been related to its dysfunction. One may in fact speculate that the redundant apoptotic cell recognition mechanisms are in place to ensure that efferocytosis is effectively executed. Macrophage phagocytosis impairment has been reported mostly in the context of inflammatory and autoimmune diseases (72). For instance, macrophage efferocytosis is defective in atherosclerotic plaques, in chronic inflammatory lung diseases, and in lymph nodes of systemic lupus erythematosus patients. The involvement of phagocytosis in central nervous system diseases is, however, less explored. Mutations in MERTK lead to retinal diseases possibly linked to deficient phagocytosis of photoreceptors (130). Similarly, mutations in TREM2 or its bridge protein DAP12 are well known to cause defects in phagocytosis by osteoclasts, causing bone cysts and early dementia (Nasu-Hakola disease) (131). These mutations have been more recently associated to increased risk of AD (132) and result in deficient beta amyloid clearance in mouse models of AD (76) and reduced efferocytosis in cultured human microglia (133). In addition, mouse models and human biopsy samples have also shown impaired microglial efferocytosis during epilepsy caused by hyperactivity of the neuronal network (42). Phagocytosis therefore has a strong potential for impinging on the course of neurodegenerative diseases, as has been evidenced by blocking of the phagocytosis inhibitor CD22 in aging mouse brains to restore a microglial homeostatic profile and improve cognitive function (74).

In closing, we hope to provided enough evidence to support the idea that microglia are to some extent similar to other tissue macrophages and that important lessons can be learnt from them (**Figure 5**). There are remarkable similarities between microglia and macrophages in many aspects related to their origin, the establishment and maintenance of their identity, and in their dynamic epigenetic and transcriptional landscapes.

They are also comparable and at the same time unique in the regulation of their phagocytosis efficiency, the cargos they engulf, and the functional consequences of phagocytosis, including metabolic adaptations, immunomodulation and its impact on the surrounding tissue. Our aspiration is that pointing out the (dys)similarities between microglia and macrophages will help to develop novel tools to harness microglial phagocytosis in the healthy and diseased brain.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### REFERENCES


#### FUNDING

This work was supported by grants from the Spanish Ministry of Science, Innovation and Universities with FEDER funds (BFU2015-66689 and RYC-2013-12817), a Basque Government Department of Education project (PI\_2016\_1\_0011), and a Fundación Tatiana Pérez de Guzmán el Bueno project (P-048-FTPGB 2018) to AS. In addition, MM-R is recipient of a predoctoral fellowship from the University of the Basque Country UPV/EHU. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. No funding is available for open access publications from our institution or other sources.


are shaped by the local microenvironment. Cell. (2014) 159:1312–26. doi: 10.1016/j.cell.2014.11.018


and neuroprotective molecules by microglial cells. J Neuropathol Exp Neurol. (2003) 62:208–16. doi: 10.1093/jnen/62.2.208


**Conflict of Interest:** 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 © 2020 Márquez-Ropero, Benito, Plaza-Zabala and Sierra. 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.

# Phagocytosis of Apoptotic Cells in Resolution of Inflammation

Ioannis Kourtzelis1,2 \*, George Hajishengallis<sup>3</sup> and Triantafyllos Chavakis1,4

1 Institute for Clinical Chemistry and Laboratory Medicine, University Hospital and Faculty of Medicine, Technische Universität Dresden, Dresden, Germany, <sup>2</sup> Hull York Medical School, York Biomedical Research Institute, University of York, York, United Kingdom, <sup>3</sup> Laboratory of Innate Immunity and Inflammation, Penn Dental Medicine, Department of Basic and Translational Sciences, University of Pennsylvania, Philadelphia, PA, United States, <sup>4</sup> Centre for Cardiovascular Science, Queen's Medical Research Institute, University of Edinburgh, Edinburgh, United Kingdom

Efficient inflammation resolution is important not only for the termination of the inflammatory response but also for the restoration of tissue integrity. An integral process to resolution of inflammation is the phagocytosis of dying cells by macrophages, known as efferocytosis. This function is mediated by a complex and well-orchestrated network of interactions amongst specialized phagocytic receptors, bridging molecules, as well as "find-me" and "eat-me" signals. Efferocytosis serves not only as a waste disposal mechanism (clearance of the apoptotic cells) but also promotes a pro-resolving phenotype in efferocytic macrophages and thereby termination of inflammation. Alterations in cellular metabolism are critical for shaping the phenotype and function of efferocytic macrophages, thus, representing an important determinant of macrophage plasticity. Impaired efferocytosis can result in inflammation-associated pathologies or autoimmunity. The present mini review summarizes current knowledge regarding the mechanisms regulating macrophage efferocytosis during clearance of inflammation.

#### Edited by:

Esther M. Lafuente, Complutense University of Madrid, Spain

#### Reviewed by:

Amiram Ariel, University of Haifa, Israel Jeffrey Louis Curtis, University of Michigan Medical School, United States

\*Correspondence: Ioannis Kourtzelis ioannis.kourtzelis@york.ac.uk

#### Specialty section:

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

Received: 12 January 2020 Accepted: 11 March 2020 Published: 31 March 2020

#### Citation:

Kourtzelis I, Hajishengallis G and Chavakis T (2020) Phagocytosis of Apoptotic Cells in Resolution of Inflammation. Front. Immunol. 11:553. doi: 10.3389/fimmu.2020.00553 Keywords: phagocytosis, efferocytosis, DEL-1, immunometabolism, inflammation resolution, integrins

#### INTRODUCTION

Specific recognition and engulfment of "foreign" material or pathogens by host cells, designated as phagocytosis, is an essential process modulating the immune response and tissue homeostasis (1, 2). Besides phagocytosis of opsonized pathogens by phagocytes during an infection, host cells undergoing apoptosis are also cleared by macrophages; the specific phagocytosis of dying cells by macrophages is designated efferocytosis (3). The specific pathways for phagocytosis of different types of cargo point to the versatility of the phagocytic machinery (1, 2).

A central player in sterile inflammation or inflammation associated with infection are neutrophils recruited to the inflamed site (4, 5). Recruited neutrophils phagocytose and kill pathogens, produce reactive oxygen species (ROS) and pro-inflammatory factors, such as cytokines, and can either release neutrophil extracellular traps (NETs) or undergo apoptosis (4– 6). Efferocytosis is therefore of great importance in the regulation of neutrophilic inflammation (3, 7–10). Effective removal of dying neutrophils promotes not only inflammation resolution but also contributes to restoration of tissue and organ homeostasis (3). A complex network of interactions between receptors mediating phagocytosis, bridging molecules, "find-me" and "eat-me" signals, such as phosphatidylserine (PS), which is presented on the outer part of the membrane of cells undergoing apoptosis, contributes to the formation of phagocytic synapse (**Figure 1**) and the operation of the efferocytic machinery (7, 8, 11). As resolution of inflammation occurs, efferocytic macrophages acquire a resolving phenotype producing factors that dampen inflammation and promote restoration of tissue integrity, such as IL-10 or transforming growth factor β (TGFβ) (3),

as

as well as specialized pro-resolving lipid mediators (SPM), such as resolvins, lipoxins, and maresins (9, 12). SPM synthesis can further promote efferocytosis, thereby further potentiating inflammation resolution (9, 12). Defective removal of apoptotic cells resulting from impaired efferocytosis can lead to chronicity of inflammation and development of inflammatory disorders, such as atherosclerosis and autoimmune diseases (9, 13–15).

Emerging evidence suggests that macrophage function is regulated by alterations in their cellular metabolism in response to environmental cues within the inflammatory milieu (16). For instance, specific metabolic components may promote or suppress inflammatory responses in macrophages (16). Importantly, efferocytosis also promotes immunometabolic reprograming in macrophages (17) (**Figure 2**). For example, digestion of the engulfed apoptotic cargo through phagolysosomal activity results in a load of lipid components derived from the apoptotic cell membranes that is linked to enhanced fatty acid oxidation and regulates macrophage function (17–19). The mechanisms underlying apoptotic cell removal by macrophages as well as the immunometabolic alterations in efferocytic macrophages during inflammation resolution is the focus of the present review.

#### MOLECULAR CROSS-TALK BETWEEN APOPTOTIC CELLS AND MACROPHAGES DURING EFFEROCYTOSIS

The engulfment of dying cells by efferocytic macrophages requires the recognition of the former by the latter and the formation of the engulfment synapse, which is regulated by a network of "find-me," "eat-me" and bridging molecules, "don't eat-me" signals and specialized phagocytic receptors (11) (**Figure 1**). Neutrophils undergoing apoptosis during an inflammatory response, secrete molecules serving as "findme" signals that can attract phagocytes to eliminate apoptotic cell corpses (20, 21). These include the nucleotides adenosine triphosphate (ATP) and uridine triphosphate (UTP), which are recognized by the macrophage purinergic receptor P2Y2 (22), or the lipids lysophosphatidylcholine (LPC) (23) and sphingosine-1-phosphate (24), which bind to macrophage G-proteincoupled receptors G2A and S1P1-5, respectively. Furthermore, recognition of dying cells by macrophages may be facilitated by the interaction of intercellular adhesion molecule 3 (ICAM3 or CD50) on the former with CD14 on macrophages (25) as well as by the thrombospondin (TSP1)–CD36 interaction (26). Moreover, the specific recognition of apoptotic cells is ensured by the presence of "eat-me" signals. PS is the most well characterized "eat-me" signal (**Figure 1**). During apoptosis, this phospholipid is found on the outer part of the membrane and binds directly or indirectly, via bridging molecules (opsonins), to phagocytic receptors (27, 28). Calreticulin (Crt) is a membrane-associated protein that functions as an "eat-me" signal' on the surface of dying cells and is recognized by the LDL-receptor-related protein 1 (LRP1 or CD91) on phagocytes (29). The long pentraxin PTX3 may also act as an "eat-me" signal to facilitate the capture of dying neutrophils by macrophages (30).

As alluded to above, bridging molecules are often key to efficient interactions between apoptotic neutrophils and macrophages (11). Milk fat globule-EGF factor 8 protein (MFG-E8 or lactadherin) promotes efferocytosis by binding to PS on apoptotic cells and to macrophage phagocytic integrin receptors αvβ3 and αvβ5 (31, 32). MFG-E8 shares homology with developmental endothelial locus-1 (DEL-1). Besides the established anti-inflammatory role of DEL-1 as inhibitor of β2

integrin-dependent leukocyte recruitment and IL-17-mediated inflammation (8, 33–38) and its role as a regulator of bone marrow myelopoiesis (39, 40), secreted DEL-1 promotes engulfment of apoptotic cells and inflammation resolution (41). In particular, DEL-1 functions as a molecular bridge that binds concomitantly to PS on the apoptotic neutrophil surface via its C-terminal discoidin I-like domains and to αvβ3 integrin [also known as vitronectin receptor (42)] on macrophages via its N-terminal RGD motif within the second EGF-like domain (41, 43). Importantly, the compartmentalized localization of DEL-1 differentially regulates the inflammatory response. Specifically, the endothelial cell-derived molecule promotes anti-inflammatory activity by blocking β2 integrindependent leukocyte recruitment, whereas DEL-1 derived from macrophages promotes efferocytosis-dependent resolution of inflammation (33, 41) (**Figure 1**). Other molecules that act as a molecular bridge to facilitate interactions between apoptotic cargo and phagocytes include annexin A1 or lipocortin-1 (44), β2-glycoprotein-I (β2-GPI) (45), and galectin-3 (Gal-3) (46). Moreover, the bridging molecules Growth arrest-specific factor 6 (Gas6) and protein S have been implicated in PS-mediated apoptotic cell clearance (47, 48) via interacting with the receptor tyrosine kinases Tyro-3, Axl and Mer (TAM) (49–52) (**Figure 1**).

Phagocytic receptors on macrophages involved in the regulation of efferocytosis include also the PS receptors of the T-cell membrane protein (Tim) family, such as TIM-1 and TIM-4 (53, 54). The brain angiogenesis inhibitor 1 (BAI1) (55) and stabilin-2 (56) also serve as PS receptors (**Figure 1**). Additionally, CD14, the scavenger receptor CD36, and the integrin CD11b/CD18 (αMβ2) (besides the integrins αvβ5 and αvβ3 that were mentioned above) are implicated as efferocytosis receptors (8, 11, 26, 42, 57, 58).

The presence of "don't eat-me" signals further adds to the complexity of the regulation of apoptotic cell clearance. Specifically, surface expression of CD47 (also named integrin-associated protein) prevents phagocytosis by macrophages (59, 60). Binding of CD47 to the macrophage signal regulatory protein alpha (SIRPα) modulates rearrangement of actin cytoskeleton, thereby downregulating phagocytosis (59, 60). However, apoptotic cells have decreased levels of CD47 that allows their clearance by macrophages (59–61). Platelet and endothelial cell adhesion molecule 1 (PECAM-1, CD31) also exerts a "don't eat-me" function. In this regard, homotypic CD31 interaction between non-apoptotic neutrophils and macrophages may prevent phagocytic clearance (62). Furthermore, CD24 (63) and the complement receptor CD46 (64) have been described as repulsive signals that interfere with efferocytosis. Decreased presence or alterations in the distribution of "don't eat-me" signals have been associated with enhanced efferocytic activity (65, 66).

Efficient efferocytosis is critical for shaping the pro-resolving phenotype in macrophages that includes production of immunomodulators, which in turn further enhance resolution of inflammation (67–71). For instance, production of TGFβ by efferocytic macrophages is a major orchestrator of inflammation resolution. Indeed, upregulation of TGFβ owing to efferocytosis promotes downregulation of the proinflammatory mediators TNF, IL-1β and IL-8. Consistently, antibody-mediated inhibition of TGFβ restored expression of inflammatory mediators. Along the same line, administration of apoptotic cells in vivo models of inflammation triggers resolution of inflammation in a manner dependent on TGFβ upregulation (72, 73). Additionally, interleukin 13 derived from regulatory T cells acts on macrophages and promotes production of IL-10, which in turn enhances efferocytosis via activation of Rac1 GTPase and thereby inflammation resolution in atherosclerosis (74). Besides the upregulation of immune-modulating factors, such as TGFβ or IL-10,

the direct inhibition of pro-inflammatory cytokines further contributes to inflammation resolution. As an exemplar, formation of NETs that aggregate at the inflamed site leads to protease–dependent degradation of inflammatory cytokines and chemokines, thereby promoting resolution of acute neutrophilic inflammation (75).

The phenotype of efferocytic macrophages is additionally controlled by the enzyme 12/15 lipoxygenase (12/15-LO) that oxidizes polyunsaturated fatty acids and generates bioactive lipid metabolites leading to the biosynthesis of pro-resolving lipid mediators (9, 76). Specifically, apoptotic cell engulfment can be performed by resident or monocyte-derived resolution phase macrophages expressing 12/15-LO (77–80). Apoptotic cell engulfment further promotes the expression of this enzyme (81). Moreover, 12/15-LO has been implicated to function in preventing induction of autoimmunity (77).

Plasminogen and its cleavage product plasmin not only regulate the initiation but also the resolution phase of inflammation. Treatment of mice with plasminogen/plasmin resulted in recruitment of pro-resolving macrophages and in upregulation of TGFβ. Administration of plasminogen/plasmin at the peak of inflammation was associated with increased neutrophil apoptosis and efferocytosis; the pro-resolving effect of plasminogen was mediated by annexin A1 (82). In accordance, impaired efferocytosis accompanied by decreased levels of annexin A1 was observed in mice deficient in plasminogen or its receptor (83). Besides being involved in pro-resolving actions of plasminogen, annexin A1 plays a broader role in inflammation resolution (84, 85). Annexin A1 levels are increased in the resolution phase of monosodium urate crystal–induced arthritis, a model of gout. Pharmacologic or genetic inactivation of annexin A1 resulted in insufficient resolution of gout-related inflammation in mice (86). In addition, treatment of mice with annexin A1 resulted in upregulation of IL-10 and downregulation of proinflammatory mediators, while, consistently, inhibition of annexin A1 abrogated inflammation resolution induced by glucocorticoids (87, 88).

Moreover, IFN-β from macrophages was recently identified as a factor promoting resolution of inflammation. IFN-β levels were higher in the resolution phase of pneumonia and peritoneal inflammation. Activation of IFN-β signaling via STAT3 enhances apoptosis of neutrophils and their subsequent efferocytic clearance, resulting in a pro-resolving reprograming of macrophages (79).

#### METABOLIC MODULATION OF MACROPHAGE FUNCTION IN THE CONTEXT OF EFFEROCYTOSIS

The impact of cellular metabolism on macrophage function and plasticity has gained much attention recently (16, 89–91). Metabolic pathways, such as glycolysis, tricarboxylic acid (TCA) cycle, pentose phosphate pathway and fatty acid oxidation, regulate macrophage phenotype in the context of inflammatory responses (91). For instance, increased glycolytic flux has been linked to pro-inflammatory M1-like activation of macrophages, whereas oxidative phosphorylation is associated with antiinflammatory macrophage polarization (92).

Moreover, macrophage tissue specificity may be associated with differential metabolic activity. For example, resident peritoneal macrophage survival depends on the transcription factor GATA6 that is regulated by the vitamin A metabolite retinoic acid (93, 94), while the nuclear receptor liver × receptor (LXR) alpha that is activated by lipids regulates differentiation of marginal zone splenic macrophages (95). It is now established that tissue-specific resident macrophages have distinct transcriptomic profiles and phenotypes depending on the particular microenvironment (96, 97). Importantly, in this regard, the manner by which efferocytosis is regulated in resident macrophages may be dictated by the tissue microenvironment. Indeed, parabiosis-based experiments have revealed substantial heterogeneity in the utilization of bridging molecules, efferocytic receptors and transcription factors by macrophages from different tissues. For instance, the mannose receptor CD206 is upregulated in phagocytic macrophages in bone marrow and intestine but not in the spleen (98). Moreover, the profile of upregulated anti-inflammatory mediators by efferocytic macrophages is tissue-specific (98).

Regulation of macrophage metabolic activity in the context of efferocytosis (**Figure 2**) is of major importance for the outcome of inflammation resolution and tissue repair (99). Following engulfment, macrophages degrade the apoptotic material through phagolysosomal activity (100, 101), resulting in substantial metabolic load and influence on macrophage metabolism (17).

Aerobic glycolysis was recently implicated in regulation of efferocytosis and shaping an anti-inflammatory environment by efferocytic macrophages (102). Specifically, transcriptomic analysis of macrophages engaging in active phagocytosis of apoptotic cells revealed an upregulation of several members of the solute carrier (SLC) membrane transport protein family, including the glucose transporter protein type 1 (GLUT1; encoded by the gene Slc2a1) and monocarboxylate transporter 1 (encoded by Slc16a1) promoting lactate release. Enhanced glycolysis in efferocytic macrophages promoted actin polymerization and continued uptake of apoptotic cells. On the other hand, lactate from efferocytic macrophages contributed to establishment of an anti-inflammatory phenotype of the surrounding tissue (102) (**Figure 2**). Consistently, knockdown of Slc16a1 in the setting of efferocytosis resulted in reduced mRNA expression of factors linked to the resolving macrophage phenotype and resolution of inflammation, such as Tgfb1 and Il10 (102). Furthermore, specific deletion of GLUT1 in myeloid cells was associated with defective phagocytic ability of macrophages and with development of unstable lesions in a model of atherosclerosis (103). Moreover, deficiency of the glycolytic enzyme 6-phosphofructose-2-kinase and fructose-2,6-bisphosphatase (PFKFB3) in macrophages led to reduced efferocytosis capacity, thus further supporting a role for glycolysis in apoptotic cell clearance (104).

Metabolites derived from engulfed apoptotic cells serve to fine-tune the process of efferocytosis and resolution of inflammation. Apoptotic cells are a source of the amino acids

arginine and ornithine that are metabolized in macrophages to putrescine. This metabolic process enables continual rounds of efferocytosis (**Figure 2**). Putrescine enhances, via the RNA-binding protein HuR, the mRNA stabilization of the GTP-exchange factor Dbl. Dbl in turn activates the GTPase Rac1, resulting in changes in the actin cytoskeleton that facilitate further apoptotic cell engulfment. Exogenous administration of putrescine increases inflammation resolution in the context of atherosclerosis. Consistently, deficiency in myeloid cells of either the enzyme arginase 1 (Arg1), which converts arginine to ornithine, or the enzyme ornithine decarboxylase (ODC), which mediates the decarboxylation of ornithine to putrescine (**Figure 2**), is associated with efferocytic dysfunction and defective atherosclerosis resolution (105).

Several lines of evidence support also an important role of mitochondrial metabolism on the modulation of efferocytosis and efferocytosis-dependent resolution of inflammation. Metabolomic analysis of macrophages that have engulfed apoptotic material revealed metabolic reprograming of efferocytic macrophages toward fatty acid oxidation. Specifically, upregulation of pro-resolving IL-10 in efferocytic macrophages was mediated by mitochondrial beta-oxidation and induction of sirtuin1 signaling (18) (**Figure 2**). The interplay between mitochondrial activity and effective efferocytosis was also illustrated by analyzing the function of mitochondrial uncoupling protein 2 (Ucp2). Besides its function in uncoupling oxidative phosphorylation from synthesis of ATP, Ucp2 can promote efferocytosis by reducing the mitochondrial membrane potential of macrophages. In the same vein, deficiency of Ucp2 resulted in defective apoptotic cell removal and was associated with development of atherosclerosis, whereas overexpression of Ucp2 enhanced efferocytosis (19). Additionally, dynamic alterations in mitochondrial physiology may dictate the outcome of efferocytosis. In particular, a major component of mitochondrial homeostasis, mitochondrial fission, mediated by the function of the GTPase dynamin-related protein 1 (Drp1), positively regulates continuous apoptotic cell removal by macrophages. Accordingly, the absence of Drp1 was associated with higher atherosclerosis development in low-density lipoprotein receptor 1 (Ldlr1) deficient mice (106).

Lipids deriving from the apoptotic cargo are abundant postengulfment digestion products (17); lipid metabolism induces activation of the nuclear receptors peroxisome proliferatoractivated receptor (PPAR) gamma and delta, LXR alpha and beta and retinoid × receptor (RXR) alpha and beta (107) in macrophages (**Figure 2**). Activation of these transcription factors promotes upregulation of phagocytic receptors and bridging molecules and the resolving macrophage phenotype. For instance, TGFβ and IL-10, which both promote inflammation resolution, are upregulated in efferocytic macrophages in an LXRand PPAR delta–dependent manner (108, 109). Importantly, deficiency of these nuclear receptors has been linked to defective efferocytosis and development of chronic inflammation or autoimmune manifestations (108, 109). Moreover, LXR activation is involved in DEL-1–dependent efferocytosis and macrophage reprograming to a proresolving phenotype (41). LXR signaling also regulates expression of transglutaminase 2 (Tgm2) (110), which acts as a co-receptor to αvβ3-integrin and promotes formation of engulfing portals (111).

Cholesterol metabolism has been also implicated in the modulation of efferocytosis. Treatment of macrophages with the cholesterol-lowering drug lovastatin, which inhibits the rate-limiting enzyme of cholesterol synthesis 3-hydroxyl-3 methylglutaryl coenzyme A (HMG-CoA) reductase, leads to increased apoptotic cell clearance (112). Administration of another HMG-CoA reductase inhibitor, simvastatin, enhances the efferocytosis-dependent amelioration of inflammation in the context of lung fibrosis (113). Along the same line, the ATP-binding cassette transporter (ABCA1), a protein that modulates cholesterol efflux, is upregulated during efferocytosis in a manner dependent on LXR signaling (114). Furthermore, decreased hydrolysis of cholesterol esters and impaired oxysterol production, due to blockade of the enzyme lysosomal acid lipase, negatively affect LXR pathway activation and apoptotic cell removal, resulting in chronic inflammation (115). These studies point to the intimate crosstalk between cholesterol metabolism and nuclear receptor signaling involved in efferocytosis.

## CONCLUDING REMARKS

Macrophage efferocytosis is a major player in resolution of inflammation. Efferocytosis not only paves the way toward the timely termination of the inflammatory response, but also promotes restoration of tissue homeostasis. In this context, alterations in macrophage cellular metabolism are important regulators of efferocytosis. At the same time, metabolic reprograming in efferocytic macrophages induced by digested apoptotic material substantially regulates the function and plasticity of efferocytic macrophages. Given the relevance of efferocytosis as a mechanism against chronic inflammatory disease, a better mechanistic understanding of the pathways that orchestrate the mutual interaction between clearance of dying cells and metabolic alterations in macrophages is required. This knowledge will provide a scaffold for designing therapeutic approaches to improve macrophage function in inflammation resolution and harness macrophage efferocytosis for the treatment of pathologies associated with chronic inflammation or autoimmunity.

# AUTHOR CONTRIBUTIONS

IK, GH, and TC wrote the manuscript.

# FUNDING

The authors were supported by grants from the National Institutes of Health (DE024716 to GH, DE028561 and DE026152 to GH and TC), the ERC (DEMETINL to TC), and the Deutsche Forschungsgemeinschaft (SFB1181 to TC).

#### REFERENCES

fimmu-11-00553 March 30, 2020 Time: 18:30 # 6



macrophage phenotypic, and functional changes. Front Immunol. (2019) 10:1458. doi: 10.3389/fimmu.2019.01458


**Conflict of Interest:** 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 © 2020 Kourtzelis, Hajishengallis and Chavakis. 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.

# Phagocytic Integrins: Activation and Signaling

#### Alvaro Torres-Gomez 1,2 \*, Carlos Cabañas 1,2,3 \* and Esther M. Lafuente1,2 \*

<sup>1</sup> Department of Immunology, Ophthalmology and Otorhinolaryngology, School of Medicine, Universidad Complutense de Madrid, Madrid, Spain, <sup>2</sup> Instituto de Investigación Sanitaria Hospital 12 de Octubre (i+12), Madrid, Spain, <sup>3</sup> Severo Ochoa Center for Molecular Biology (CSIC-UAM), Madrid, Spain

Phagocytic integrins are endowed with the ability to engulf and dispose of particles of different natures. Evolutionarily conserved from worms to humans, they are involved in pathogen elimination and apoptotic and tumoral cell clearance. Research in the field of integrin-mediated phagocytosis has shed light on the molecular events controlling integrin activation and their effector functions. However, there are still some aspects of the regulation of the phagocytic process that need to be clarified. Here, we have revised the molecular events controlling phagocytic integrin activation and the downstream signaling driving particle engulfment, and we have focused particularly on αMβ2/CR3, αXβ2/CR4, and a brief mention of αVβ5/αVβ3integrins.

#### Edited by:

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

#### Reviewed by:

Thomas Vorup-Jensen, Aarhus University, Denmark Kaz Nagaosa, Hirosaki University, Japan

#### \*Correspondence:

Alvaro Torres-Gomez atorr01@ucm.es Carlos Cabañas ccabanas@cbm.uam.es Esther M. Lafuente melafuente@med.ucm.es

#### Specialty section:

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

Received: 30 January 2020 Accepted: 31 March 2020 Published: 30 April 2020

#### Citation:

Torres-Gomez A, Cabañas C and Lafuente EM (2020) Phagocytic Integrins: Activation and Signaling. Front. Immunol. 11:738. doi: 10.3389/fimmu.2020.00738 Keywords: phagocytosis, integrins, signaling, CR3, Mac-1, complement

# INTRODUCTION

Phagocytosis entails the engulfment and disposal of particles in sequential steps, including particle recognition, cytoskeletal remodeling, membrane protrusion, particle engulfment, and phagolysosomal digestion (1, 2). The role of integrins in phagocytosis is evolutionarily conserved and can be observed in Caenorhabditis elegans INA-1/PAT-3, which is involved in clearance of apoptotic cells (3), and Drosophila αPS3/βν, which has roles in microbial defense and apoptotic cell removal (4, 5) (**Table 1**). In mammals, the orthologues αVβ3/αVβ<sup>5</sup> are expressed in professional and non-professional phagocytes (endothelial, epithelial, fibroblast, and neuronal and mesenchymal cells) with a role in phosphatidylserine-rich apoptotic/necrotic body clearance. Professional phagocytes in mammals express complement receptors αMβ2/CR3 and αXβ2/CR4, which are involved in host defense and tissue homeostasis (45). Other integrins with reduced phagocytic capacity (α5β1, α2β1, α3β1, and α6β1) are involved in phagocytosis of fibrillar or denatured extracellular matrix components (**Table 1**).

Integrins are characterized by requiring activation to be functional. This review has focused on the main events determining β<sup>2</sup> integrin activation and downstream signaling in relation to cytoskeletal remodeling and particle engulfment, and it makes a special mention of the main differences between other phagocytic integrins, especially those involved in apoptotic cell clearance.

# INTEGRIN STRUCTURE AND ACTIVATION

Phagocytic integrins are heterodimeric (α and β subunit) receptors. Subunits are divided into ectodomains, a transmembrane helix, and short cytoplasmic tails. The α-subunit ectodomains contain Mg2+-binding metal-ion-dependent adhesive sites (MIDAS) and Adjacent to MIDAS (AdMIDAS), which binds inhibitory Ca2<sup>+</sup> or activating Mn2<sup>+</sup> (46, 47). Ligand binding can occur



either at the αI-domain (α-subunit) in αX, αM, and α<sup>2</sup> or at the α/β-chain interface in integrins without the αI domain (**Figure 1A**, **Table 1**).

Integrins are tightly regulated by conformational changes, a hallmark of which is cytoplasmic tail separation (48). Integrin conformations are described according to the state of the headpiece (open/closed; H+/H−) and leg ectodomains (extended/bent; E+/E−) (49). Resting integrins remain in an inactive/"bent" (E−H−) conformation with the lowest free energy (−4.0 kcal/mol for α5β1) with respect to fully activated integrins (50). E−H−is characterized by a closed ligand-binding site and clasped membrane proximal regions (51). In activated integrins (E+H+), the hybrid domain (β-subunit) swings away from the α-chain, and the membrane proximal regions unclasp. This correlates with the rearrangement of the MIDAS and opening of the ligand binding site (51).

Structural and mutational studies have investigated models of integrin activation to explore whether integrin extension or leg separation occurs first. Mutations and deletions of the CDloop (β-subunit terminal domain) have been proposed to keep integrins from extending and have shown no impact on αVβ<sup>3</sup> and αIIbβ<sup>3</sup> activation (52); there is little proof that mutations in this region affects β<sup>2</sup> integrins (53), strongly indicating that releasing these constraints is not enough to induce activation.

Structural studies (54) have demonstrated that αXβ<sup>2</sup> follows the "switch-blade" model of activation, where leg separation occurs first, releasing constraints of the bent conformation and opening of the ligand-binding site resulting in an intermediate/low affinity conformation E+H<sup>−</sup> (55). The E+H<sup>−</sup> conformation has a free energy between 1.6 and 0.5 kcal/mol, meaning the high affinity conformation is thermodynamically favored (50, 56). Mutations in the EGF3 repeat of the β2-subunit have also been shown to induce a high affinity conformation through destabilizing the thermodynamically favorable bent conformation and facilitating leg separation (57). It is noteworthy that an E−H<sup>+</sup> conformation has been described for αLβ<sup>2</sup> and αMβ2, allowing integrins to bind ICAM in cis, which may regulate neutrophil function (58); however, the specifics of how this activation takes place remain unknown.

Integrin activity is regulated by changes in affinity and aggregation, with the latter affecting receptor avidity. Cytoplasmic proteins bind to α- or β-subunits causing tail separation, stabilizing their high affinity conformation (48, 59). This can be triggered either through signaling from other receptors ("inside-out" signaling, **Figure 1B**), direct ligandbinding, or experimentally, using Mn2<sup>+</sup> ("outside-in" signaling, **Figure 1C**), which triggers downstream signaling pathways (60).

#### INSIDE-OUT SIGNALING

#### Rap1 as a Signaling Node

Early studies in complement-dependent phagocytosis using mutants of small GTPases pointed to Rap1 as the main regulator of αMβ<sup>2</sup> activity (61) and to it being required for β1-mediated

phagocytosis (62). Rap1 acted as a node, connecting different signaling pathways (chemokines, fMLP, PAF, and TNFα) for integrin activation (63). Rap1-GTP loading is induced by specific Guanine–Nucleotide Exchange Factors (GEFs), being Epac1 (dependent on cyclic AMP; cAMP) and CalDAG-GEFs (dependent on Ca2+/Diacylglycerol; DAG), amongst the

denote unsolved or hypothetical signaling steps.

best characterized (**Figure 1B**). Epac1 expression was found to increase during monocyte-macrophage differentiation, correlating with the acquisition of immunoregulatory functions (64), and in neutrophilic HL-60 cells, pharmacological activation of Epac1 increased Rap1-GTP and complement-dependent phagocytosis (65). RasGRP3/CalDAG-GEFIII exhibited similar effects, promoting Rap1 activation and phagocytosis (66). Mutations in CalDAG-GEF1 produced leukocyte adhesion deficiency syndrome (LADIII) with defective neutrophilendothelial adhesion (67), and mouse CalDAG-GEF1−/ − macrophages showed reduced integrin activation (68). Rap1 activation can be induced by Toll-like receptors (TLRs) (69); however, the signaling pathways remain poorly defined. In neutrophils, secreted myeloid-related proteins (MRPs) 8 and 14 bind to TLR4 causing Rap1 activation and β2-dependent adhesion (70). In macrophages, low concentrations of TLR3/4/9 agonists induced RasGRP3-dependent Rap1 activation (71). Activation of αMβ<sup>2</sup> by TLR2/TLR4 required Rac1-GTP loading, PI3K activity, and cytohesin-1 binding to the β<sup>2</sup> subunit (72).The role of cytohesin-1 is controversial, as the use of cytohesin-1 siRNAs and inhibitors results in an increase in the αMβ<sup>2</sup> affinity conformation (73).

#### Talin1 and Kindlin-3

Talin1 and Kindlin-3 are the best-characterized integrin activators. Both belong to the FERM family but interact with distinct NPXY motifs in the cytoplasmic tails of β1, β2, and β3, and they thus contribute differently to activation (74). Although Talin-binding is required for efficient β<sup>5</sup> activation during adhesion, it is dispensable for phagocytosis (75). αVβ<sup>5</sup> requires an unknown mediator that recognizes a YEMAS motif proximal to the NPXY. A candidate could be the FERM family FRMD5, as it promotes β5-Kindlin-2 interaction and induces ROCK activation during adhesion (76), yet there is no information of its relevance in phagocytosis.

Talin1 contains an N-terminal globular head with a linear FERM domain and a C-terminal rod domain organized in 13 subdomains (R1-R13), which contains a dimerization domain, an integrin binding site, three F-actin binding sites, and several Vinculin and RIAM binding sites (77, 78). The FERM domain has four subdomains (F0-F1-F2-F3), where F3 contains the primary integrin-binding site (IBS) that interacts with the membrane-proximal NPXY motif conserved in β-integrin tails (59, 79, 80). In resting leukocytes, Talin1 remains auto-inhibited due to an interaction between F2F3 and R9 subdomains, which mask the primary IBS (81). Several Talin1 activation mechanisms have been proposed. By binding to PIP5Kγ, Talin1 is recruited to the plasma membrane where the F2F3 domain binds to phosphatidylinositol-4,5-bisphosphate (PI(4,5)P2), disrupting the head–tail interaction and exposing the IBS (82, 83). Additionally, RIAM–Talin1 interaction was described as necessary for Talin1 activation and recruitment to integrin tails (**Figure 1B**) (84).

Hematopoietic cell-specific Kindlin-3 is mutated in LADIII, causing β1/β2/β<sup>3</sup> activation defects (85, 86) and preventing neutrophils adhesion to iC3b and ICAM-1 (87). Kindlin-3 binds

Torres-Gomez et al. Phagocytic Integrin Signaling

to the membrane-distal NPKF sequence in the β<sup>2</sup> subunit tail without excluding Talin1 binding (**Figure 1B**) (87). Studies of their individual contributions to activation revealed that Kindlin-3 is not sufficient to induce the high-affinity state of αLβ2, whereas Talin1 promotes full activation (88). Whether binding of Talin1 and Kindlin-3 is sequential or simultaneous and their exact contribution to integrin activation remains to be explored. The signaling events directing Kindlin-3 to integrins also remain elusive, as in T cells, Kindlin-3 localization at immune synapses depends on Rap1 and Mst-1/RapL signaling (89), whereas no such interaction has been described for phagocytic cells.

#### RIAM–Talin1 Interaction

RIAM (Rap1-Interacting Adaptor Molecule or APBB1IP) was identified as a Rap1 effector that promoted a β<sup>2</sup> and β<sup>1</sup> high affinity state, increasing T-cell adhesion and spreading (90). RIAM binds to Rap1-GTP through a central Ras-association domain (RA), to PI(4,5)P<sup>2</sup> through a Pleckstrin-Homology (PH) domain and to VASP, Profilin, and PLCγ1 via proline-rich regions (90–94). RIAM also interacts with Talin1 through its N-terminus and Talin1 has several RIAM-binding sites located at F3, R2, R3, R8, and R11 subdomains (77). Binding of RIAM to Talin1 releases Talin1 from its autoinhibition (**Figure 1B**) (95).

The Rap1-RIAM-Talin1-Integrin pathway also operates in complement-dependent phagocytosis. Studies in Talin1-silenced THP-1 cells revealed that Rap1 and Talin1 regulated each other's localization at phagocytic cups (96). Reduced RIAM expression in human monocyte-derived macrophages (MDM), neutrophilic HL-60 cells, and THP-1 cells diminished levels of high affinity αMβ<sup>2</sup> and reduced complement-dependent phagocytosis and Talin1 recruitment to phagocytic cups (65). Complementdependent phagocytosis, cell adhesion to ICAM, and ROS production were also impaired in mouse RIAM−/<sup>−</sup> macrophages and neutrophils (97). Additionally, RIAM deficiency in vivo had a profound effect on β<sup>2</sup> activity but a moderate effect on β1- or β3-dependent functions (98).

Besides RIAM, Rap1 effectors RapL and RGS14 (Regulator of G-Protein Signalling-14) have been proposed to regulate αMβ<sup>2</sup> activation by inside-out signaling (**Figure 1B**). The former is proposed to interact with αM-subunit inducing integrin tail separation and integrin activation (99); however, RapL has only been shown to interact with a GFFKR motif in α<sup>L</sup> cytoplasmic tail, and there is no direct evidence that it plays a role in αMβ<sup>2</sup> activation (100). For RGS14, the integrin activation mechanism is unknown but seems to be dependent on Talin1-binding to β<sup>2</sup> (101).

Recently, a direct interaction between Rap1-GTP and Talin1 was described at Talin1 F0 and F1 subdomains (102–105). Synergistic interaction between this region and an F1 lipidinteracting helix facilitates relocation of Talin1 and its integrinactivating function (**Figure 1B**) (105, 106). This pathway could be relevant for fast cell responses, as disruption in mice impaired platelet aggregation, neutrophil adhesion, extravasation, and phagocytosis but had no effect on macrophage adhesion and migration (104).

## OUTSIDE-IN SIGNALING

Outside-in signaling during phagocytosis initiates upon ligand interaction, stabilizing the active conformation, separating integrin tails, allowing for the binding of actin cytoskeletal linkers (Talin1 and/or Kindlin-3), and reorganizing cytoskeletal constraints, as described in the picket-fence model (2). This generates the force needed to drive membrane extension and particle engulfment/internalization (**Figure 1C**). Regulators have been described in focal complex-like formations at the phagocytic cup (107).

#### CLUSTERING AND TYROSINE KINASES

One of the earliest events in outside-in signaling could be ligand-induced clustering, a process requiring Talin1 and/or Kindlin-3 (74, 108). Kindlin-3-induced clustering is reported to activate Src family kinases (SFKs) (109, 110) by the exclusion of tyrosine phosphatases such as CD45 (68). Size exclusion of these membrane-bound phosphatases with large extracellular domains seems to be a common feature of integrin-mediated close-contact immune processes, such as Dectin-1 and FcγRIII phagocytosis and immune synapse formation (68, 111, 112). This process does not exclude SFKs but favors their activation due to removing the inhibitory effect of these phosphatases (109, 110). However, there are as of yet only indirect evidences (109, 110) that phosphatases such as CD45 are excluded during integrin-mediated phagocytosis.

SFKs appear to be exclusively involved in "outside-in" signaling, as SFK-deficient cells produced reduced ROS after integrin clustering (113), whereas ICAM-1 adhesion and complement-dependent phagocytosis were normal in preactivated SFK-deficient cells (114, 115).

A requirement for SFK activation has been described for β1, β2, and β<sup>3</sup> integrins (109, 114, 116). Hck, Fgr, and Lyn are the representative SFKs in myeloid cells. Hck co-localized with αMβ<sup>2</sup> at phagocytic cups of complement-opsonized zymosan (117, 118), and the Hck knockout phenocopied the α<sup>M</sup> knockout (119). However, in U937 macrophage-like cells, Hck and Fgr siRNA, unlike Lyn, had no effect on particle internalization (120), and genetic restitution of Fgr-deficient cells inhibited adhesion, spreading, and Syk activation (121). In contrast, the Hck−/<sup>−</sup> Fgr−/<sup>−</sup> Lyn−/<sup>−</sup> triple knockout showed no inhibition in CR3 mediated phagocytosis (122), which may point to compensatory roles of other ubiquitously expressed SFKs. Despite the research into outside-in activation of SFKs, the exact mechanism and individual contribution of each SFK have yet to be dissected.

SFK activity precedes activation of tyrosine kinases Syk and FAK family member Pyk2. Syk is necessary for phagocytosis of iC3b-opsonized beads/zymosan and localizes at phagocytic cups (107, 123), whereas Pyk2 contributes to clearance of complement-opsonized bacteria (124). Clustering of β<sup>2</sup> integrins results in Syk activation (125), which in turn triggers Pyk2 signaling (126). Pharmacological inhibition of Syk and FAK kinases points to non-redundant functions during phagocytosis and to a possible sequential activation (107).

#### PHOSPHOINOSITIDES COORDINATE GTPASES AND CYTOSKELETAL REARRANGEMENTS

Phagocytosis requires sequential enrichment of phosphoinositides (PIPs) in the inner leaflet of the plasma membrane (127). PIP enrichment recruits GEFs for small GTPases, which are sequentially activated (128), and other components of integrin adhesion complexes.

PI(4,5)P<sup>2</sup> enrichment can be induced by lipid redistribution due to particle-induced plasma membrane deformation (129) and/or by SFK or Talin1-induced PIP5Kγ activity (83, 130, 131). PI(4,5)P<sup>2</sup> enrichment strengthens Talin1 anchoring (81) and recruits different factors involved in F-actin dynamics, like the actin-depolymerizing-factor ADF/Cofilin, whose activity is inhibited by PI(4,5)P<sup>2</sup> (132), or the formin mDia (133, 134). Additionally, RIAM binds PI(4,5)P<sup>2</sup> and may recruit VASP and Profilin, which could also contribute to actin polymerization (90, 93) (**Figure 1C**).

PI(3,4)P<sup>2</sup> recruits and induces Vinculin activation through disrupting an auto-inhibitory interaction (135). This is dependent on Syk activity and, to a lesser extent, on FAK/Pyk2 and is upstream from ROCK activation (107). In focal complexes, RIAM contributes to Vinculin binding to Talin1, as RIAM-Talin1 interaction unmasks a Vinculin binding site in Talin1 (77). Afterwards, Vinculin binding to F-actin and α-actinin favors filament bundling and force generation (136, 137).

Increased PI(3,4,5)P<sup>3</sup> at CR3-phagocytic cups (138) depends on PI3K (139) and Syk (126), and both are activated downstream of Kindlin-induced clustering (140). PI(3,4,5)P<sup>3</sup> enrichment recruits Vav1/3, which are GEFs for the RhoA family GTPases (128). Complement-dependent phagocytosis requires Vav1 to activate RhoA (61, 141) but also RhoG with no participation from Cdc42 and Rac1 (142). However, expression of constitutively active Rac1 rescues the defective engulfment of Vav1-3 knockouts (143). This discrepancy could be explained by the overlapping roles of RhoG and Rac1 (144, 145) (**Figure 1C**).

In the final steps leading to engulfment, RhoA-GTP initiates the ROCK-MLCK-myosin signaling pathway and actomyosin contractility (146). RhoA is enriched at phagocytic cups, and its localization is modulated by motifs in β2-integrin tails (141). Premature activation of RhoA is inhibited by Rap-GTP through ARAP3, a dual GAP for Rho and Arf GTPases, which is recruited by PI(3,4,5)P<sup>3</sup> and PI(3,5)P<sup>2</sup> (147). Finally, mDia contributes to phagosome closure (107, 133) and particle engulfment by connecting the actin cytoskeleton to microtubules (148) (**Figure 1C**).

#### SIGNALING DURING PHAGOCYTOSIS OF APOPTOTIC CELLS

During apoptotic cell phagocytosis by mammalian αVβ5/αVβ3, a p130Cas-CrkII-Dock180-Elmo module induces Rac1 activation, which is responsible for cytoskeletal remodeling and phagosome formation (149, 150). Other known signals include the

activation of SFKs, as signals from the Mer-TK receptor recruit phosphorylated FAK to mammalian β<sup>5</sup> in a Src-dependent manner (151), and Syk and Pyk2 activation has been shown to occur for αVβ<sup>3</sup> (152, 153). There is also evidence that Rac-1 activation is dependent on RhoG and its GEF Trio (154, 155), whereas RhoA inhibits engulfment (156), and the role of Cdc42 remains unclear (157–159).

An orthologous pathway using the CED-2-CED-5-CED10 module has been described for C. elegans INA-1, which activates the Rac ortholog and requires activation of SRC-1(Src-ortholog) (3). Similarly in Drosophila, severed axon clearance requires Src42A and Shark—the Src and Syk orthologs, respectively (160, 161)—pointing to an evolutionarily conserved pathway operating in apoptotic cell removal.

#### DISCUSSION AND FUTURE PERSPECTIVES

There are still critical gaps in the knowledge of phagocytic integrin signaling, specifically concerning proximal events and their hierarchy. There are several proposed alternative Talin1-recruitment mechanisms, but their contributions and significance are yet to be established. Rap1-Talin1 interaction is evolutionarily conserved and might constitute a mechanism for short-term adhesions (105), whereas Rap1-RIAM-Talin1 contacts would have a faster recruitment of effector proteins. In this line, it is yet to be established if RIAM is required for outsidein signaling, formation, and recycling of the focal adhesion-like complexes distributed in phagocytic cups (107).

Different F-actin nucleators/elongators are described to participate in CR3-mediated phagocytosis; however, their localization, recruitment, and relative contributions are unknown. The regulation of small GTPases, which control actin dynamics, remains obscure; there is scarce evidence of

#### REFERENCES


GEF and GAP spatiotemporal localization in phagocytic cups, and it is well established that GTPases negatively regulate each other, which also raises questions on signal termination and negative-feedback loops.

Many structural and signaling proteins required for phagocytic integrin function have potential post-translational modification-dependent functions, and, although there are several candidates, little work has been undertaken to establish Ser/Thr kinase and phosphatase recruitment and localization within the phagocytic cup.

Fine-grain elucidation of the molecular mechanisms involved in integrin-mediated phagocytosis will yield invaluable information on possible control points for phagocyte functions (antigenic capture, pathogen, tumor or apoptotic body elimination, etc.). Indeed, complementopsonized immune complexes and particles may be presented directly by subcapsular sinus macrophages to naïve B cells or conveyed to dendritic cells for B-cell presentation. This process requires cooperation between antigen-presenting cell αMβ2/αXβ<sup>2</sup> and B-cell CR1, CR2, and/or Fc receptors (162– 165). Manipulation of this pathway may inform new vaccine strategies (166).

### AUTHOR CONTRIBUTIONS

AT-G and EL wrote the original draft. AT-G prepared the figures. Final writing and editing were performed by AT-G, CC, and EL.

#### FUNDING

This work has been supported by Ministerio Español de Economía y Competitividad (MINECO) grants: SAF2016-77096- R (EL and CC). AT-G was supported by an FPU predoctoral fellowship from MINECO.

protein signaling to promote Candida albicans clearance. Cell Host Microbe. (2011) 10:603–15. doi: 10.1016/j.chom.2011.10.009


**Conflict of Interest:** 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 © 2020 Torres-Gomez, Cabañas and Lafuente. 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.

# Cell-in-Cell Structures in the Liver: A Tale of Four E's

#### Scott P. Davies<sup>1</sup> , Lauren V. Terry<sup>1</sup> , Alex L. Wilkinson<sup>1</sup> and Zania Stamataki1,2 \*

<sup>1</sup> Centre for Liver and Gastrointestinal Research, Institute of Immunology and Immunotherapy, University of Birmingham, Birmingham, United Kingdom, <sup>2</sup> NIHR Birmingham Liver Biomedical Research Centre, University Hospitals Birmingham NHS Foundation Trust, Birmingham, United Kingdom

The liver is our largest internal organ and it plays major roles in drug detoxification and immunity, where the ingestion of extracellular material through phagocytosis is a critical pathway. Phagocytosis is the deliberate endocytosis of large particles, microbes, dead cells or cell debris and can lead to cell-in-cell structures. Various types of cell endocytosis have been recently described for hepatic epithelia (hepatocytes), which are non-professional phagocytes. Given that up to 80% of the liver comprises hepatocytes, the biological impact of cell-in-cell structures in the liver can have profound effects in liver regeneration, inflammation and cancer. This review brings together the latest reports on four types of endocytosis in the liver -efferocytosis, entosis, emperipolesis and enclysis, with a focus on hepatocyte biology.

#### Edited by:

Florence Niedergang, Centre National de la Recherche Scientifique (CNRS), France

#### Reviewed by:

Luis Enrique Munoz, University of Erlangen-Nuremberg, Germany Jean-pierre Couty, Institut National de la Santé et de la Recherche Médicale (INSERM), France

#### \*Correspondence:

Zania Stamataki z.stamataki@bham.ac.uk

#### Specialty section:

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

Received: 03 February 2020 Accepted: 23 March 2020 Published: 13 May 2020

#### Citation:

Davies SP, Terry LV, Wilkinson AL and Stamataki Z (2020) Cell-in-Cell Structures in the Liver: A Tale of Four E's. Front. Immunol. 11:650. doi: 10.3389/fimmu.2020.00650 Keywords: cell-in-cell, liver, efferocytosis, entosis, emperipolesis, enclysis, cancer, regeneration

#### INTRODUCTION

Cell-in-cell (CIC) structures are formed when a whole cell resides inside the cytoplasm of another, and they have been observed for decades in various contexts. The best characterized CIC mechanism is known as efferocytosis, the clearance of dead or dying cells by professional and non-professional phagocytes (1–9). Yet, CIC structures, in which the internalized cell remains viable, have been observed for over a century (10). Recent work has provided evidence for the role of hepatocytes, the principal parenchymal cell within the liver, in several of these processes: efferocytosis (1), live cell internalization events including suicidal emperipolesis (11), entosis (12) and enclysis (13) (**Table 1**). Although the immediate consequences of dead and live cell capture have been investigated, the biological implications and impact on clinical outcomes remain to be elucidated.

The liver receives 75% of its blood supply from the gastrointestinal tract via the hepatic portal vein (14). As such, it is persistently challenged by toxic substances and microbial- or food-derived antigens. Not only must the liver function to detoxify and neutralize harmful products it is exposed to, it must also maintain an immunotolerising environment so as not to initiate inappropriate immune responses to commensal microbes and food antigens. Nonetheless, the liver must retain the ability to mount a rapid immune response in the case of infection. The role of the liver in immunity is well-established (15), and the cells residing within it are finely tuned to maintain the balance between immunotolerance and immunogenicity. If this balance is perturbed and tolerance is breached, liver disease can develop due to hepatocyte damage during inflammation.

Chronic liver diseases follow a common pathway of progression independently of etiology. Repeated liver injury results in fibrosis, cirrhosis and ultimately, end-stage disease leading to liver failure, the only viable treatment option for these patients is liver transplantation, which is associated with significant pitfalls including organ shortage and graft rejection. Since liver disease

#### TABLE 1| The mechanism of cell-in-cell structures.


The best-characterized process is efferocytosis, where multiple receptors have been identified, however, only one receptor is known for hepatocytes (ASGR1). Receptors unique to entosis, emperipolesis or enclysis have not been identified.

continues to increase worldwide (16) there is an unmet clinical need to develop novel therapies that will alleviate chronic inflammation, prevent fibrosis or boost liver immunity in the context of viral infection and primary or metastatic liver cancer. Hepatocytes constitute an attractive target for therapy in these patients, because (i) they are uniquely found in the liver, (ii) they drive regeneration in injury, (iii) they are the focus of infection or malignancy in hepatocellular carcinoma, (iv) they are a natural destination for drug absorption, and (v) unlike targeting immune cells, hepatocyte-directed therapies are unlikely to cause systemic immunosuppression or autoimmunity. We propose that targeting CIC structures has the potential to lead to clinical benefit for patients with liver diseases.

#### EFFEROCYTOSIS

The capture and deletion of dying cells by efferocytosis (from effere, Latin for "to take to the grave," "to bury"), a specialized form of phagocytosis, is a crucial process for the liver with important biological impact (1). The liver is inundated with infiltrating immune cells that are destined to die by apoptosis and be digested by liver cells (17). The frequent turnover of hepatocytes, associated with the detoxification of waste products, further contributes to the dead cell burden faced by the liver. Failure to clear these cell corpses can spell disastrous immune consequences, including premature inflammatory responses and an increased risk of autoimmune disease (8).

To prevent the build-up of cellular debris, the cellular composition of the liver is uniquely prepared, it is frequented by monocyte-derived macrophages and possesses a specialized resident macrophage population known as Kupffer cells, which arise following signals from liver-resident cells (9, 18). Aside from these "professional phagocytes," liver- parenchymal and non-parenchymal cells can also capture and delete dying cells. These "non-professional" populations include hepatic sinusoidal endothelial cells (HSECs), biliary epithelial cells (BECs), stellate cells and hepatocytes (**Figure 1**) (3, 4, 6, 7). As such, the liver is universally prepared to rapidly clear cell corpses, thus maintaining its immune tolerance.

An astute adaptation of the liver to manage the persisting need to clear dying cells is for its principal cell type, the hepatocyte, to be adept at efferocytosis (1). Hepatocytes are epithelia tasked with drug detoxification and can undergo necrotic cell death in the process, thus neighboring hepatocytes are most likely to make first contact with a dying cell. Hepatocyte efferocytosis was first described in 1952, when Rosin and colleagues observed the presence of erythrocytes within the cytoplasm of hepatocytes (2). This was later ratified by Dini et al., who showed that hepatocytes could also engulf apoptotic cells (3). The same investigation suggested a role for asialoglycoprotein receptor 1 (ASGR1) in the recognition of these cells. Experiments conducted in our group has further confirmed hepatocyte ability to engulf necrotic cells in health and in cancer (1, 13). Other cells which have this capability generally require alternative, more improvisational molecular mechanisms to capture necrotic cells, compared to those known for apoptotic cell capture (5, 20).

Hepatocytes can both remove and replenish areas of necrotic sheets associated with disease-related hepatotoxicity, and drive regeneration during injury (21–23).

In contrast to the liver's professional phagocyte populations, the mechanisms by which hepatocytes clear dead cells are poorly understood. There are few candidate receptors, in addition to ASGR1, by which hepatocytes may recognize and capture dying cells. The consequences of efferocytosis for the hepatocyte have also not been widely explored. The hepatocyte would be granted nutrients from the lysosomal digestion of captured cells. Hepatocytes may also acquire increased genetic diversity at the cellular level through efferocytosis. Efferosomes may physically impede cytokinesis, causing the engulfing hepatocyte to become multinucleate, as seen in breast cancer cells with CIC structures (12). Increased genetic diversity amongst hepatocytes has been shown to increase the ability of the liver to adapt and regenerate in response to a wider variety of insults (24, 25). Efferocytosis may be a mechanism which accelerates this phenomenon, although this may also increase the risk of contracting mutations associated with the onset of hepatocellular carcinoma (HCC). Although hepatocyte multinucleation is both frequent and tolerated in the liver, particularly in older individuals (26, 27), chronic efferocytosis resulting from disease-associated necrosis may promote the acquisition of oncogenic mutations. As the onset of HCC is rarely spontaneous and frequently associated with chronic liver disease, increased hepatocyte efferocytosis may represent a risk factor for its onset.

Dysregulation of efferocytosis in the liver can lead to disease development. This has been exemplified in the case of macrophage clearance, knockout mice lacking hepatic macrophages that express the dead cell scavenger receptor, MerTK, showed exasperated damage when treated with acetaminophen (APAP) (28). More recently it was demonstrated that carbon tetrachloride-treated glycoprotein NMB (gpnmb) KO mice, whose macrophages lack the ability to process internalized cells, showed greater activation of pro-fibrotic myofibroblasts (29). It is also likely that the dysregulation of hepatocyte efferocytosis may contribute to the pathogenesis of other chronic liver diseases. Autoantibodies against ASGR1 have

been detected in patients with autoimmune hepatitis (30, 31). Additionally, ethanol-treated rat hepatocytes were shown to be defective in ASGR1-mediated efferocytosis (32).

The effects of aging and the accompanying immune paresis must be considered in liver homeostasis, specifically regarding the clearance of apoptotic and necrotic cells. In both aging and chronic liver disease, there is an accumulation of senescent cells, which produce senescence-associated secretory phenotype (SASP) factors. SASP factors include pro-inflammatory cytokines and growth factors, that have been noted to alter the local microenvironment and induce paracrine senescence and in turn, immuno senescence (33–35). One characteristic of immune senescence is the reduced capacity of a cell to phagocytose, which may contribute to persistent inflammation in older individuals, termed "inflammageing" and lead to defective clearance and resolution of inflammation (36, 37).

Whilst little is understood about hepatocyte efferocytosis in terms of aging, several in vivo studies have shown an age-associated decline in macrophage efferocytosis in other tissue types. For example, one study observed that peritoneal macrophages from aged (24-month old) mice had an impaired ability to efferocytose apoptotic Jurkat cells, compared to 2 month old, young mice (38). This result was similarly observed by Linehan et al., whom proceeded to transplant young (8 to 12-week-old mice) peritoneal macrophages into aged (15 to 20 month-old mice) peritoneal space (39). The transplanted, young macrophages in fact exhibited a diminished ability to efferocytose post-transplantation, suggesting that the microenvironment lead to alterations in the efferocytic ability. Additionally, there was a decline in the ability of alveolar macrophages to efferocytose neutrophils in aged mice, which may contribute to lung damage (40). Based on links drawn between diminished efferocytic capacity and old age, it is logical to infer that hepatocytes could be subjected to similar pressures from aging and this warrants further investigation.

Further understanding into the mechanisms of hepatocyte efferocytosis will likely provide opportunities for promoting dead cell clearance and thus preventing immature inflammatory responses in the liver.

#### ENTOSIS

For over a century, CIC structures in which viable cells are internalized into other cells have been reported (10, 41, 42). Live cells have been shown to invade or be engulfed by host cells of non-phagocytic origin. Unlike with efferocytosis, which consistently targets cell corpses for lysosomal degradation, these cells can remain viable within vacuole-like structures for long periods and succumb to variable outcomes depending on the context. Although the molecular mechanisms for most examples of live CIC formation generally remain poorly understood, several processes are well-described in the literature. One of these is known as entosis (εντóς, inside, into, within) (**Figure 2**) (41, 43).

In 2007, Overholtzer and colleagues reported that extracellular matrix detachment of cancer cells could promote CIC formation via contractile forces associated with adherens junction formation. This process involved junctional proteins, E-cadherin and β-catenin, and was dependent on actomyosin contractility mediated by Rho-associated coiled-coil-containing protein kinase (ROCK) activity in the target cell specifically (12). This finding, coupled with time-lapse microscopy of CIC formation, was strongly suggestive of target cell invasion as opposed to engulfment and has since been confirmed in several studies (44, 45).

The plasma membrane is the primary site for initiating CIC formation. Plasma membrane blebbing and polarized actin dynamics have been suggested as drivers of entotic invasion (45), with a recent study demonstrating the requirement for the myocardin-related transcription factor-serum response factor (MRTF-SRF) pathway and subsequent sustained ezrindependent plasma membrane blebbing (44). Furthermore, in addition to the requirement for adherens junctions (12, 46, 47), studies have shown that the composition of the plasma membrane play a role in entosis. Both liposomes and cholesterol were shown to inhibit CIC formation, presumably by hindering myosin light chain phosphorylation and thus actomyosin contractility (48).

The fate of the internalized cell is variable, most succumb to non-apoptotic cell death and lysosomal degradation, although some target cells occasionally undergo division or release (12, 43, 49–51). Thus, the biological consequences of entosis and the impact on tumor biology remain controversial (52). Since degradation of target cells by neighboring cancer cells has the potential to limit tumor growth, then perhaps entosis represents an intrinsic tumor suppressor mechanism, by which metastatic cancer cells that become detached from matrix are eliminated. Yet, adherent epithelial cells can also undergo entosis, a process driven by mitosis and negatively regulated by cell cycle protein Cdc42 (46). Furthermore, tumor cell cannibalism could promote host cell survival by providing nutrients to those which lack vascular access (53). In support of this, Overholtzer's group later demonstrated that entosis is induced in adherent cells by glucose starvation, in a manner requiring activity of target cell AMPactivated protein kinase (AMPK) (54). The ability of cancer cells to adapt to starvation by performing entosis and enabling nutrient recovery would confer metabolic advantage of malignant cells, thereby promoting progression of more aggressive tumors. Indeed, it has been proposed that there is direct competition between cancer cells, dictated by mechanical deformability and subsequent entosis, thus ensuring the survival of the most adapted tumor cells (55). These findings highlight the importance of the tumor microenvironment in regulating intracellular signaling pathways that mediate entosis and tumor survival.

The clinical impact of entosis in hepatocellular carcinoma has not been investigated. Similarly to observations made in other epithelial cells, we reported recently that hepatomas cultured in 2D were also capable of engulfing their neighbors (13). The vesicle that housed the internalized cell was enriched in E-cadherin, suggesting that this was another example of entosis (**Figure 2**). It is not yet clear if non-neoplastic hepatocytes perform entosis. Regardless, liver cancers may benefit from entosis as a source of adaptation and nutrition. Given that there is no effective therapy and the incidence of hepatocellular carcinoma is increasing in the West (56), targeting entosis may prove to be of clinical value.

#### EMPERIPOLESIS

Emperipolesis is a term coined by Humble et al. (57) and used to describe the movement of live cells following internalization ("inside-round-about wandering") (**Figure 3**) (57). It has been proposed that whilst CIC and emperipolesis should be used generically to describe the process of cell movement associated with CIC structures, cannibalism and entosis should be used to refer to mechanisms of CIC formation specifically (10, 41). Cell-in-cell structures, or emperipolesis, have long been observed by histopathologists in several types of chronic liver disease. Emperipolesis is increased in autoimmune hepatitis (58, 59) and chronic viral infection (60, 61), suggesting a potential role in liver injury or T cell clearance (62, 63). The precise physiological and pathophysiological role of emperipolesis, however, remains elusive.

The first demonstration of a physiological role for emperipolesis in the liver was reported by Bertolino and

colleagues in 2011. They defined a mechanistically distinct type of emperipolesis known as suicidal emperipolesis, in which autoreactive CD8<sup>+</sup> T lymphocytes actively invade hepatocytes and undergo lysosomal degradation (11, 64). Inhibition of this process by wortmannin led to intrahepatic accumulation of autoreactive cells and a breach of tolerance. Wortmannin-treated mice developed immune-mediated hepatitis 3 days post-infusion with autoreactive CD8<sup>+</sup> T cells, as determined by raised alanine aminotransferase levels and histological liver damage. The authors therefore proposed this as a mechanism of extrathymic regulation for maintaining immune tolerance within the liver.

There is also evidence for a pathophysiological role of CIC structures. Emperitosis, another form of emperipolesis, was the name initially given to natural killer (NK) cell invasion of tumor cells and subsequent programmed cell death. Like entosis, emperitosis also requires cadherins, Rho/ROCK proteins and ezrin (65, 66). In contrast to entosis, NK cells succumb to caspase-3-mediated apoptosis, which was attributed to granzyme B accumulation within the vacuole (65). This process has also been extended to human cytotoxic regulatory T cell line, HOZOT, which actively penetrate cancer cell lines but not cells of non-neoplastic origin (67). It is therefore conceivable that emperitosis of cytotoxic immune cells serves as one of the many mechanisms employed by cancer cells to evade immune surveillance. Furthermore, a recent study showing that internalization of anti-fibrotic NK cells in HBV cirrhotic patients is transforming growth factor-β-dependent and may represent a novel mechanism of fibrogenesis (68). Further work is required to fully elucidate the molecular mechanisms of suicidal emperipolesis, which may allow therapeutic targeting in the context of liver transplantation, autoimmune disease and viral hepatitis. Nevertheless, the evidence that this process is distinct from other CIC mechanisms is compelling, and is one example of the complex pathways which can underlie CIC formation.

#### ENCLYSIS

We have recently reported a distinct cell capture process within the liver termed enclysis (Eγκλε ωεγ´ - (ε´v-) + κλε ω, to enclose,

to confine, to keep in captivity), in which live CD4<sup>+</sup> T cells are captured by hepatocytes (**Figure 4**) (13). This process occurred in vitro, in primary human hepatocytes and in hepatoma cells (Huh-7 and HepG2 cells), and ex vivo within patient liver samples. T cells were also found to reside within hepatocytes in vivo as shown in 30 µm-thick sections from cirrhotic patients.

Whilst intercellular adhesion molecule-1 (ICAM-1) facilitated early T cell adhesion to hepatocytes, the ligands for ICAM-1 are not distinct to CD4<sup>+</sup> T cells and therefore this adhesion molecule does not explain enclysis specificity. Interestingly, adhesion molecule and junctional protein, β-catenin, selectively associated with the enclytic vesicle, in contrast to the efferosome (phagosome containing dead cell) which showed no β-catenin localization. Despite both entosis and enclysis involving formation of membrane blebs (13, 44, 45), enclysis was distinguished from entosis by the lack of E-cadherin association with the enclytic vesicle. Notably, instances of entosis were observed between Huh-7 hepatoma cells, where a clear localization of E-cadherin was apparent. The lack of requirement for the RhoA/ROCK pathway, similar to suicidal emperipolesis, provides a further distinction of enclysis and entosis. Enclysis resembles macropinocytosis, in that there are significant membrane alterations during cell capture events including ruffling, blebs and lamellipodia formation (13, 69), which is in contrast to emperipolesis where these membrane protrusions are absent (63). Furthermore, the wortmannin-insensitivity of enclysis further defines this process as mechanistically distinct, compared with emperipolesis which is abrogated by wortmannin treatment (11).

Whilst CD4<sup>+</sup> T cells were specifically targeted over CD8<sup>+</sup> T cells and CD20<sup>+</sup> B cells, Tregs were three times more likely to be engulfed than non-Treg cells. Vesicles containing Tregs readily acidified with cells undergoing degradation via the lysosomal pathway, unlike non-Tregs, which survived for long periods and remained connected to the extracellular space via the endocytic pathway. Moreover, FOXP3<sup>+</sup> Tregs were more frequently found within hepatocytes than Tbet<sup>+</sup> effector cells, in both donor livers surplus to clinical requirement and liver explants from end-stage disease patients (13). Thus, we propose enclysis as a novel immunomodulatory pathway within the liver that could offer therapeutic opportunities to toggle inflammation. But why would hepatocytes possess the ability to target Tregs for degradation when an integral function of the liver is to maintain immunotolerance? Although this seems counter-intuitive, given the previous studies which have evidenced a role for the liver in maintaining peripheral immune tolerance (70), it is conceivable that enclysis could act as a biological switch, preventing the liver from becoming "immunoblind". The ability of hepatocytes to control local T cell populations and modulate ratios of regulatory and effector cells may represent an intrinsic mechanism by which the liver can rapidly respond to its local inflammatory environment. The stimuli and endogenous regulators of enclysis, however, are yet to be defined.

Identification of selective modulators of enclysis may offer opportunities for therapeutic intervention. On one hand, inhibition of Treg cell capture and/or degradation may be successful in situations where it is desirable to enrich local Treg populations and dampen inflammation. Indeed, research surrounding Treg cell-based therapy is ongoing (71–75), and combination with pharmacological inhibitors of enclysis may show promise for several indications, including chronic inflammation and to promote immunotolerance following organ transplantation. Alternatively, in the context of cancer, boosting Treg sequestration or modulating release of effector T cell subsets may be beneficial to enhance tumor immunogenicity (76).

#### IMPLICATIONS AND FUTURE PERSPECTIVES

The impact of cell-in-cell structures on the host cell biology has only recently been investigated. Phagocytosed cells that enter the phagocyte as apoptotic cells or cellular debris, and also engulfed live cells that may subsequently die inside endosomes, can present an added source of nutrients. However, CIC may have longerlasting implications on the host cell.

Consequences of viable cell internalization include eventual death of host or target cells, target cell division or release, or prevention of host cell division which can cause multinucleation, polyploidy and aneuploidy (19, 77). This has implications for cancer metastatic potential (78), and links between aneuploidy and genomic instability (loss of tumor suppressor genes) have been established (19, 77). A recent study has shown that p53 mutations in lung adenocarcinoma patients are associated with increased incidence of cell-in-cell structures, and that mutant p53 expression promotes entotic engulfment, tumorigenesis and disease recurrence (51). Whilst host cells lacking p53 had perturbed cell division and subsequent death, mutant p53 cells underwent aberrant cell division, multinucleation, and tripolar mitosis. Thus, p53 expression facilitated pro-tumorigenic entotic engulfment and abnormal mitosis, which consequently contributed to genomic instability.

Cell-in-cell structures in patients are indicative of worse clinical grade and poor prognosis (51, 58, 79). In the context of the liver, ploidy changes and multinucleation in hepatocytes are important considerations for liver regeneration (25–27, 80–82) and associate with various pathological processes (83). In a study where oxidative stress was shown to promote polyploidy in nonalcoholic fatty liver disease, the authors suggested that hepatocyte multinucleation preceded the onset of hepatocellular carcinoma (84). In the absence of cancer, it is now understood that the polyploid state in mice may restrict hepatocyte proliferation and liver regeneration (81).

It is important to consider the biological impact of cellin-cell structures in liver diseases. Hepatocytes have evolved to eliminate apoptotic and necrotic cells efficiently to prevent inflammation, and this is also true for other CIC processes, the mechanisms of cell death in the liver have been described previously (85, 86). Failure to eliminate necrotic or autoreactive cells would exacerbate liver injury and increase the incidence of fibrosis. Fibrosis (liver scarring) is the consequence of various chronic liver diseases caused by viral, autoimmune, metabolic or cholestatic liver injury, and can lead to cirrhosis and end-stage disease requiring a transplant. The precise mechanism of bile acid hepatotoxicity has not been fully elucidated.

Non-alcoholic fatty liver disease (NAFLD) is of increasing concern at a global scale, and up to 25% of patients can progress to non-alcoholic steatohepatitis (NASH). Increased liver enzymes denote hepatocellular damage [ALT, AST, and others, reviewed in (86)]. The molecular mechanisms controlling hepatocellular injury have begun to emerge in recent studies that revealed a role for the transcription regulator TAZ in preventing hepatocyte death, inflammation and fibrosis (87, 88). Further, hepatocyte Notch activation was linked directly to NASH-related fibrosis (89). The role of efferocytosis in the clearance of apoptotic cells and the prevention of necrotic cell injury and fibrosis in NASH has been reviewed recently (90).

The pro- or anti-inflammatory impact of enclysis in NASH remains to be established, however, NASH liver explants show measurable CD4<sup>+</sup> T cells inside hepatocytes, including FOXP3<sup>+</sup>

and Tbet<sup>+</sup> T cells (13). Of note, Ma et al. showed that in NAFLD, dysregulation of lipid metabolism causes a selective loss of intrahepatic CD4<sup>+</sup> but not CD8<sup>+</sup> T cells, leading to impaired tumor surveillance and accelerated carcinogenesis (91). The mechanism of CD4<sup>+</sup> T cell elimination in this context has not been described, however, it was shown that T cells died by apoptosis following linoleic acid exposure from lipidladen hepatocytes.

#### CONCLUSION

The engulfment of live, apoptotic and necrotic cells by hepatocytes has important implications for their biology in health, inflammation and cancer. These range from nutrient acquisition that can promote cancer cell survival in poorly vascularized tumors, to changes in ploidy that can affect liver regeneration and cancer aggressiveness. It is therefore important to understand the molecular mechanisms that govern these processes so that they can be targeted specifically for patient benefit. **Figure 5** summarizes our current knowledge of cell-incell structures linked to hepatocyte biology.

Increasing the clearance or necrotic cells is an important goal to prevent inflammation and liver failure, including in catastrophic drug-induced liver injury such as paracetamol (acetaminophen) toxicity. Modulation of T cell capture by suicidal emperipolesis (CD8<sup>+</sup> T cells) or enclysis (Treg cells) has the potential to influence liver tolerance and toggle inflammation

#### REFERENCES


in conditions such as autoimmune hepatitis, viral infection or liver cancer, where the unmet clinical needs are profound. We propose that understanding CIC structure mechanisms will enable specific therapeutic targeting and has the potential to provide new therapeutic targets for liver diseases and liver cancer.

#### AUTHOR CONTRIBUTIONS

LT and ZS prepared the figures. SD, LT, AW, and ZS wrote the review.

#### FUNDING

SD was funded by an NC3R trainee fellowship. LT was funded by an MRC iCASE fellowship, and AW was funded by a Wellcome Trust Ph.D. studentship in Mechanisms of Inflammatory Diseases. This work was initiated thanks to support from a small seed award by Wellcome Trust Institutional Support Fund to ZS, who has subsequently received funding from the Medical Research Foundation (MRC), Wellcome Trust, GMG, The Birmingham Children's Hospital Research Foundation, and a Royal Society Dorothy Hodgkin Fellowship. This manuscript presents independent research supported by the NIHR Birmingham Biomedical Research Centre at the University Hospitals Birmingham NHS Foundation Trust and the University of Birmingham.



emperipolesis in HBV cirrhotic patients. Sci Rep. (2017) 7:44544. doi: 10. 1038/srep44544


**Disclaimer:** The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.

**Conflict of Interest:** 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 © 2020 Davies, Terry, Wilkinson and Stamataki. 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.

# Phagocytosis: Our Current Understanding of a Universal Biological Process

Eileen Uribe-Querol <sup>1</sup> and Carlos Rosales <sup>2</sup> \*

<sup>1</sup> División de Estudios de Posgrado e Investigación, Facultad de Odontología, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>2</sup> Departamento de Inmunología, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, Mexico City, Mexico

Phagocytosis is a cellular process for ingesting and eliminating particles larger than 0.5µm in diameter, including microorganisms, foreign substances, and apoptotic cells. Phagocytosis is found in many types of cells and it is, in consequence an essential process for tissue homeostasis. However, only specialized cells termed professional phagocytes accomplish phagocytosis with high efficiency. Macrophages, neutrophils, monocytes, dendritic cells, and osteoclasts are among these dedicated cells. These professional phagocytes express several phagocytic receptors that activate signaling pathways resulting in phagocytosis. The process of phagocytosis involves several phases: i) detection of the particle to be ingested, ii) activation of the internalization process, iii) formation of a specialized vacuole called phagosome, and iv) maturation of the phagosome to transform it into a phagolysosome. In this review, we present a general view of our current understanding on cells, phagocytic receptors and phases involved in phagocytosis.

Keywords: immunoglobulin, antibody, phagocytosis, neutrophil, ERK, complement, integrin

# INTRODUCTION

Phagocytosis is a basic process for nutrition in unicellular organisms, and it is also found in almost all cell types of multicellular organisms. However, only a specialized group of cells called professional phagocytes (1) accomplish phagocytosis with high efficiency. Macrophages, neutrophils, monocytes, dendritic cells, and osteoclasts are among these dedicated cells. Professional phagocytes are responsible of removing microorganisms and of presenting antigens to lymphocytes in order to activate an adaptive immune response. Fibroblasts, epithelial cells, and endothelial cells can also accomplish phagocytosis with low-efficiency and are thus described as non-professional phagocytes. These cells cannot ingest microorganisms, but are important in eliminating dead cells and maintaining homeostasis (2). Phagocytosis is the process of sensing and taking in particles larger than 0.5µm. The particle is internalized into a distinctive organelle, the phagosome. This phagosome subsequently changes the structure of its membrane and the composition of its contents in a process known as phagosome maturation (3). The phagosome next fuses with

#### Edited by:

Uday Kishore, Brunel University London, United Kingdom

#### Reviewed by:

Sergio Grinstein, Hospital for Sick Children, Canada Tina Sørensen Dalgaard, Aarhus University, Denmark

> \*Correspondence: Carlos Rosales carosal@unam.mx

#### Specialty section:

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

Received: 05 February 2020 Accepted: 04 May 2020 Published: 02 June 2020

#### Citation:

Uribe-Querol E and Rosales C (2020) Phagocytosis: Our Current Understanding of a Universal Biological Process. Front. Immunol. 11:1066. doi: 10.3389/fimmu.2020.01066 lysosomes to become a phagolysosome. This new organelle contains enzymes that can degrade the ingested particle (4).

Phagocytes can identify several diverse particles that could potentially be ingested, including apoptotic cells and microbial pathogens. Discrete receptors mediate this recognition by sensing the particle as a target and then initiating signaling pathways that favor phagocytosis. Plasma membrane receptors of phagocytes are divided into non-opsonic or opsonic receptors. Non-opsonic receptors directly identify distinct molecular patterns on the particle to be ingested. These receptors include C-type lectins, such as Dectin-1 (5), Dectin-2, Mincle, or DC-SIGN (6); lectin-like recognition molecules, such as CD33; and scavenger receptors (7). Although, the toll-like receptors (TLRs) (8) can also detect molecular patterns on pathogens, they are not phagocytic receptors. Nevertheless, TLRs can cooperate with phagocytic receptors to make phagocytosis more efficient (9). Opsonic receptors detect host-derived proteins bound to target particles. These proteins known as opsonins include antibodies, fibronectin, complement, milk fat globulin (lactadherin), and mannose-binding lectin (10). Opsonins label particles as targets of phagocytosis. Fc receptors (FcR) and the complement receptors (CR) are the best characterized opsonic receptors. FcRs bind to the Fc portion of IgG (11, 12) or IgA antibodies (13). Complement receptors bind to activated complement components, such as iC3b, deposited on the particle (14).

Upon binding to the particle, phagocytic receptors initiate signaling pathways leading to remodeling of the actin cytoskeleton and lipids in the membrane, that result in the membrane extending to cover the particle (15). Then, the membrane closes at the distal end creating the phagosome. Thus, the particle gets internalized inside the phagosome. During membrane extension, the phagocytic receptors bind to the target in a sequential order and help completing the formation of the phagosome (16, 17). Next, this early phagosome undergoes sequential fusion and fission events with endocytic vesicles to create a late phagosome (18). This late phagosome then fuses with lysosomes and becomes a phagolysosome. The process to change a phagosome into a potent anti-microbial phagolysosome is known as phagosome maturation (3).

The process of phagocytosis involves several phases: (i) detection of the particle to be ingested, (ii) activation of the internalization process, (iii) formation of a specialized vacuole called phagosome, and (iv) phagosome maturation. In this review, we present the main phagocytic receptors and a general view of our current understanding on phagocytosis.

#### DETECTION OF THE TARGET PARTICLE

The first phase in phagocytosis is the detection of the target particle. Detection is mediated by dedicated receptors on phagocytic cells. Receptors directly recognizing pathogenassociated molecular patterns (PAMPs) are the patternrecognition receptors (PRRs). Some of these PRRs can initiate phagocytosis and thus constitute the non-opsonic receptors for phagocytosis. Other PRRs, for example TLRs, can bind to PAMPs but not induce phagocytosis. These receptors however, can prepare (prime) the cell for phagocytosis. Foreign particles can also be detected indirectly by opsonic receptors. The receptors for antibody and complement are the best described opsonic receptors.

#### Non-opsonic Receptors Receptors for Microorganisms

Some receptors that directly bind PAMPs and can induce phagocytosis include Dectin-1, Mincle, MCL, and DC-SIGN (**Table 1**). All these molecules are members of the family of C-type lectin receptors (6). Dectin-1 (dendritic cell-associated C-type lectin-1) recognizes yeast polysaccharides (19), and it has been shown to be a bona fide phagocytic receptor. When expressed on non-phagocytic heterologous cells, Dectin-1 allowed the cells to perform phagocytosis (19–21). In vivo, it is also possible that Dectin-1 cooperates with other phagocytic receptors in particular cells. For example, in neutrophils, Dectin-1 has been reported to connect to the phagocytic receptor Mac-1 (CD11b/CD18, CR3) (33). Mincle (macrophage-inducible C-type lectin) is a receptor for trehalose dimycolate (TDM), which is present on the cell wall of some mycobacterium (22). MCL (macrophage C-type lectin, Dectin-3) is another receptor for TDM that also binds α-mannans. Both, Mincle and MCL are considered bona fide phagocytic receptors, because when individually expressed in 293T cells, they induce internalization of beads covered with antibodies against each receptor (23). In myeloid cells, Mincle and MCL seem to cooperate for enhanced phagocytosis by forming heterodimers on the cell membrane (23). DC-SIGN (dendritic cell-specific ICAM-3-grabbing nonintegrin) is another receptor that can bind multiple microbial pathogens, including viruses, fungi, and bacteria (6), through recognition of fucosylated glycans and mannose-rich glycans (24). DC-SIGN was shown to be a phagocytic receptor by expressing it in non-phagocytic human myeloid K562 cells or in epithelial HeLa cells. K562 cells were then capable of internalizing Mycobacterium tuberculosis mannose-capped lipoarabinomannan (ManLAM)-coated beads (25), while HeLa cells could bind and internalize Escherichia coli bacteria (26). DC-SIGNR is another C-type lectin receptor with high homology to DC-SIGN, and capable of binding mannose-rich ligands (34). Therefore, DC-SIGNR is also very likely a phagocytic receptor. Other C-type lectin domain-containing proteins have been implicated in phagocytosis long before Dectin-1 and other C-type lectin receptors (6). The macrophage mannose receptor (CD206) presents several C-type lectin carbohydrate recognition domains, which detect α-mannan on many microorganisms (**Table 1**). The mannose receptor was also shown to be a bona fide phagocytic receptor when expressed in non-phagocytic COS-1 cells. Transfected COS-1 cells were then able to mediate internalization of zymosan (27).

Other PAMP receptors are also involved in phagocytosis, but it is still not clear whether they can induce phagocytosis on their own, or they do it indirectly by just bringing the particle close to the phagocyte (35). It is also possible that these receptors just prime the phagocyte, while other receptors mediate phagocytosis (35). CD14, scavenger receptor A (SR-A), CD36, and MARCO are among these receptors (**Table 1**). CD14 is a TABLE 1 | Human non-opsonic phagocytic receptors and their ligands.


receptor for lipopolysaccharide (LPS)-binding protein (28). SR-A recognizes LPS on Gram-negative bacteria (29), and on Neisseria meningitidis (30). CD36 detects Plasmodium falciparum-infected erythrocytes (31), and MARCO (macrophage receptor with collagenous structure) is involved in recognition of several bacteria (32).

#### Receptors for Apoptotic Cells

In multicellular organisms many cells die constantly by apoptosis for maintaining homeostasis. These apoptotic cells are eliminated by phagocytosis. Detection of apoptotic cells requires particular receptors for molecules that only appear on the membrane of dying cells. These molecules include lysophosphatidylcholine, and phosphatidyl serine (PS) (36). These molecules deliver to phagocytes an "eat me" signal (37). Receptors directly recognizing PS include TIM-1, TIM-4 (38), stabilin-2 (39), and BAI-1 (brain-specific angiogenesis inhibitor 1) (40) (**Table 2**). The integrin αvβ3 can also bind PS after other receptors, for example lactadherin, connect PS to the integrin (41). The integrin αVβ5 (42), CD36 (45), and CD14 (44, 46) are also receptors for apoptotic cells (**Table 2**). Some normal cells, for example activated B and T lymphocytes, may express significant levels of PS on their surface. These cells avoid phagocytosis by expressing at the same time molecules that serve as "don't eat me" signals (2). One such molecule is CD47, a ligand to the receptor SIRPα (signal regulatory protein α), which is expressed on phagocytes (47). Upon engagement, SIRPα delivers an inhibitory signal for actin assembly (47). The signaling events from these receptors to activate phagocytosis are just beginning to be elucidated. Since phagocytosis of apoptotic cells is central to homeostasis (48), determining the phagocytosis mechanisms of all these receptors for apoptotic cells will be an active area of future research.

#### Opsonic Receptors

Foreign particles can also be labeled for phagocytosis by opsonins, which are host-derived proteins that bind specific TABLE 2 | Receptors for apoptotic cells.


\*TIM, T cell immunoglobulin mucin; BAI-1, brain-specific angiogenesis inhibitor 1; MFG, milk fat globule.

TABLE 3 | Human opsonic phagocytic receptors and their ligands.


receptors on phagocytic cells. Important opsonins promoting efficient phagocytosis include antibody (IgG) molecules and complement components. These opsonins and their receptors are the best studied so far (**Table 3**).

#### Fcγ Receptors

Fcγ receptors (FcγR) are glycoproteins that specifically bind the Fc part of IgG molecules (12, 54). When FcγR engage IgG molecules in multivalent antigen-antibody complexes, they get clustered on the membrane of the cell, and then trigger phagocytosis as well as other cellular responses (11, 55) (**Figure 1**).

Three types of FcγR are expressed on human cells, FcγRI (CD64), FcγRII (CD32), and FcγRIII (CD16) (56) (**Figure 1**). FcγRI has three Ig-like domains, and displays high affinity for IgG molecules. In contrast, FcγRII and FcγRIII have two Iglike domains, and display low-affinity for IgG molecules. Thus, they can only bind multimeric immune complexes (57). FcγRI is expressed together with a dimer of the common Fc receptor gamma (FcRγ) chain. Each FcRγ chain contains tyrosine residues within an immunoreceptor tyrosine-based activation motif (ITAM; consensus sequence: YxxI/Lx(6−12)YxxI/L) (58, 59). The clustering of activating FcγRs results in the phosphorylation of tyrosine residues in the ITAM sequence present within

receptors presents only two Ig-like domains. Activating receptors contain an ITAM (immunoreceptor tyrosine-based activation motif) sequence within the α subunit (for FcγRIIa) or within the accessory γ and ζ chains (for FcγRI and FcγRIIIa). FcγRIIIa has a homodimer of γ chains in macrophages, natural killer (NK) cells, and dendritic cells, whereas it has a heterodimer of γ/ζ chains and an extra β chain in basophils and mast cells. FcγRIIIb is also an activating receptor, which is bound to the cell membrane via a glycosylphosphatidylinositol (GPI) anchor. In contrast, FcγRIIb is an inhibitory receptor containing an ITIM (immunoreceptor tyrosine-based inhibition motif) sequence.

the cytoplasmic domain of the receptor (as is the case with FcγRIIa and FcγRIIc), or in an associated FcR common γchain (as with FcγRI and FcγRIIIa) (11, 12, 57). These tyrosine residues are phosphorylated upon activation and are essential for receptor signaling. FcγRII presents two isoforms: FcγRIIa expressed mainly in phagocytic cells and FcγRIIb expressed mainly in B lymphocytes (56). FcγRIIa does not associate with FcRγ chains, but has an ITAM motif in its cytoplasmic tail. FcγRIIb also does not associate with FcRγ chains, but in contrast, has an immunoreceptor tyrosine-based inhibition motif (ITIM; consensus sequence: S/I/V/LxYxxI/V/L) in its cytoplasmic tail involved in negative signaling (60). Phosphorylated tyrosine residues within the ITIM recruit phosphatases that downmodulate signals coming from ITAM-containing activated receptors (60, 61). FcγRIIb functions as a negative regulator of cell functions, such as phagocytosis (62, 63). FcγRIII presents two isoforms: FcγRIIIa expressed in macrophages, natural killer (NK) cells, basophils, mast cells and dendritic cells, and FcγRIIIb expressed exclusively on neutrophils (57) (**Figure 1**). FcγRIIIa is a receptor with a transmembrane portion and a cytoplasmic tail, associated with a dimer of FcRγ chains, while FcγRIIIb is a glycosylphosphatidylinositol (GPI)-linked receptor, lacking a cytoplasmic tail and no known associated subunits (64) (**Figure 1**).

#### Complement Receptors

Complement receptors (CRs) bind activated complement molecules deposited on microorganisms or cells (65, 66). Complement receptors belong to three groups of molecules: (i) CR1 and CR2, which are formed by short consensus repeat (SCR) elements, (ii) CR3 and CR4, which belong to the β2 integrin family (66), and (iii) CRIg, which belongs to the immunoglobulin Ig-superfamily (14) (**Figure 2**). The integrin αMβ2 (also known as CD11b/CD18, CR3, or Mac-1) binds the complement component iC3b, and is the most efficient phagocytic receptor among complement receptors (66–68).

#### Phagocytic Receptors Cooperation

For efficient recognition of the target particle and initiation of phagocytosis, numerous receptors on the phagocyte membrane must interact with several IgG molecules on the opsonized particle. For this, receptors must have good mobility of the membrane (69) so that they can aggregate and get activated. However, free diffusion is not easy for most phagocytic receptors, because they are among other (usually bigger) transmembrane glycoproteins that cover the cell surface. Phagocytic receptors are very short molecules compared to these longer glycoproteins; hence short receptors are obscured among a layer of large glycoproteins (the glycocalyx), such as mucins, hyaluronan, and the membrane phosphatases CD45 and CD148 (70). In addition, many large glycoproteins are tied to the cytoskeleton, and can interfere with the lateral diffusion of receptors on the cell membrane (15, 17).

Interactions of Fcγ receptors with possible targets can be enhanced by cooperation with other receptors that can remove Syk, spleen tyrosine kinase.

larger glycoproteins from the area of the membrane in contact with the target particle. The result is that Fcγ receptors can then diffuse more freely on the membrane and engage more IgG molecules (16) (**Figure 3**). Removal of large glycoproteins from the membrane area of contact with the target particle is achieved by activated integrins. Integrins, for example CR3, increase their affinity for their ligand after they receive an inside-out signal (71, 72) from other receptors such as Fc receptors (73), TLRs (74), or CD44 (75). Inside-out signaling leads to activation of integrins (66, 76) via the small GTPase Rap1 (77). Activated integrins extend their conformation and create a diffusion barrier that keeps larger glycoproteins, for example the phosphatase CD45, away from phagocytic receptors (16) (**Figure 3**). Also, extended integrins can engage more distant ligands on the particle (78) and create a progressive wave of large molecules migrating in front of the bound Fcγ receptors, which aggregate in microclusters to mediate a strong adhesion between the phagocyte membrane and the particle to be ingested (17). Thus, during phagocytosis integrins cooperate with Fcγ receptors by promoting adhesion to the opsonized particle (79). Interestingly, this type of cooperation was implied by earlier studies showing that in neutrophils FcγRIIIb associates with Mac-1 integrins (80, 81).

#### ACTIVATION OF THE INTERNALIZATION PROCESS

When a particle is recognized by phagocytic receptors, various signaling pathways are activated to initiate phagocytosis. Reorganization of the actin cytoskeleton and changes in the membrane take place resulting in a depression of the membrane area touching the particle, the phagocytic cup. Then, pseudopods are formed around the particle until the membrane completely covers the particle to form a new phagosome inside the cell. The signaling mechanisms to activate phagocytosis are best-known for Fc receptors and for complement receptors (10, 67, 82– 84). For other phagocytic receptors, signaling pathways are just beginning to be investigated.

# Fcγ Receptor Signaling

Fcγ receptors get activated when they bind to antibody molecules covering the target particle and get clustered on the phagocyte membrane. Upon clustering of Fcγ receptors, they co-localize with Src-family kinases (such as Lyn, Lck, and Hck). These kinases phosphorylate tyrosines within the ITAM. Then, Syk (spleen tyrosine kinase) binds to the phosphorylated ITAMs and gets activated (67, 85). Activated Syk, in turn, can phosphorylate multiple substrates and initiate different pathways that connect to distinct cellular responses such as phagocytosis (67, 85, 86) and transcriptional activation (86) (**Figure 4**). Important Syk substrates involved in phagocytosis are the adaptor molecule LAT (linker for activation of T cells), phosphatidylinositol 3-kinase (PI 3-K), and phospholipase Cγ (PLCγ) (87, 88) (**Figure 4**). Phosphorylation of LAT induces docking of additional adaptor molecules such as Grb2 and Gab2 (Grb2-associated binder 2) (89). Phosphorylated (active) PI 3-K generates the lipid phosphatidylinositol-3,4,5-trisphosphate (PIP3) at the phagocytic cup (90, 91). This lipid also regulates activation of the GTPase Rac, and contractile proteins such as myosin. Active Rac is important in actin remodeling and activation of other signaling molecules such as JNK and the nuclear factor NF-κB (**Figure 4**). Activated PLCγ produces inositoltrisphosphate (IP3), and diacylglycerol (DAG). These second messengers cause calcium release and activation of protein kinase C (PKC), respectively (92). PKC leads to activation of extracellular signal-regulated kinases (ERK and p38) (93). The Guanine nucleotide exchange factor (GEF) Vav activates GTPases of the Rho and Rac family (94), which are involved in regulation of the actin polymerization that drives pseudopod extension (**Figure 4**).

# Complement Receptor Signaling

Among complement receptors, CR3 (integrin Mac-1) is the most efficient phagocytic receptor (66, 67). From very early studies, it has been realized that CR3 on macrophages initiates a different type of phagocytosis from the one mediated by antibody Fcγ receptors. CR3-mediated phagocytosis is characterized by "sinking" of the target particle into the cell membrane without generation of pseudopods around the particle (95). Also, the usage of cytoskeleton components for particle internalization is different between FcγR- and CR-mediated phagocytosis. During FcγR-mediated phagocytosis the actin cytoskeleton is used, whereas during CR-mediated phagocytosis the actin and microtubule cytoskeletons are involved (96, 97). In complement phagocytosis F actin remodeling depends on activation of the GTPase Rho, but not on the GTPases Rac or Cdc42 (98, 99).

FIGURE 4 | FcγR signaling for phagocytosis. FcγRIIa crosslinking by immunoglobulin (IgG) bound to a particle, induces activation of Src family kinases (SFK), which phosphorylate tyrosine residues in the ITAM sequence within the cytoplasmic tail of the receptor. Then, spleen tyrosine kinase (Syk) associates with phosphorylated ITAMs and leads to phosphorylation and activation of a signaling complex formed by the scaffold protein LAT (linker for activation of T cells) interacting with various proteins. One of these proteins is phospholipase C gamma (PLCγ), which produces inositoltrisphosphate (IP3), and diacylglycerol (DAG). These second messengers cause calcium release and activation of protein kinase C (PKC), respectively. PKC leads to activation of extracellular signal-regulated kinases (ERK and p38). The guanine nucleotide exchange factor Vav activates the GTPase Rac, which is involved in regulation of the actin cytoskeleton. Rac is also involved in activation of transcription factors such as NF-κB and JNK. The enzyme phosphatidylinositol 3-kinase (PI3K), which is recruited and activated by Syk, generates the lipid phosphatidylinositol-3,4,5-trisphosphate (PIP3) at the phagocytic cup. This lipid also regulates Rac activation, and contractile proteins such as myosin. P represents a phosphate group. ER, endoplasmic reticulum. IP3R, receptor (calcium channel) for inositoltrisphosphate.

Active Rho in turn, promotes actin polymerization via two mechanisms (**Figure 5**). First, Rho stimulates Rho kinase, which phosphorylates and activates myosin II (100). Myosin then leads to activation of the Arp2/3 complex, which promotes actin assembly at the phagocytic cup (100). Second, Rho can induce accumulation of mDia1 (mammalian diaphanousrelated formin 1) and polymerized actin in the phagocytic cup (101). Also, mDia1 binds directly to the microtubule-associated protein CLIP-170 at the phagocytic cup (102) and provides a link to the microtubule cytoskeleton required for CR-mediated phagocytosis (96, 97) (**Figure 5**).

#### PHAGOSOME FORMATION

Phagocytosis initiates when phagocytic receptors engage ligands on the particle to be ingested. Then, receptors activate signaling pathways that change the membrane composition and control the actin cytoskeleton, resulting in the formation of membrane protrusions for covering the particle. Finally, these membrane protrusions fuse at the distal creating a new vesicle that pinches out from the plasma membrane. This new vesicle containing the ingested particle is the phagosome.

During phagosome formation the membrane changes its lipid composition. These changes have been revealed by elegant fluorescence imaging techniques (3, 103), and involve the formation and degradation of different lipid molecules on the phagosome membrane in an orderly fashion. During Fcγ receptor-mediated phagocytosis, phosphatidylinositol-4,5 bisphosphate [PI(4,5)P2] initially accumulates at the phagocytic cup but then it declines rapidly (91). The decline in PI(4,5)P<sup>2</sup> is important for particle internalization, probably by facilitating actin disassembly (104). The decline in PI(4,5)P<sup>2</sup> is caused by the action of PI 3-K, which phosphorylates it to produce PI(3,4,5)P<sup>3</sup> at the phagocytic cup (105). Reduction of PI(4,5)P<sup>2</sup> in the membrane is also mediated by the action of PLCγ, which produces diacylglycerol (DAG) (91). DAG in turn, induces activation of PKCε for enhanced phagocytosis (92).

Together with the changes in lipid composition, the plasma membrane also changes by remodeling the actin cytoskeleton in order to generate the membrane protrusions that will cover the target particle. Important steps for pseudopodia formation are recognized. First, the cortical cytoskeleton gets disrupted. Second, pseudopodia are formed by F-actin polymerization. Third, at the base of the phagocytic cup, actin gets depolymerized while the membrane phagosome is sealed at the distal end to form the phagosome (15). When phagocytosis is initiated, the membrane-associated cortical cytoskeleton is altered by the action of coronins (F-actin debranching proteins) (106), and cofilin (107) and gelsolin (108) (F-actin-severing proteins). Coronin 1 concentrates at the nascent phagosome and debranches F-actin leaving linear fibers that can be severed by cofilin and gelsolin. The activity of these enzymes is controlled by their binding to phosphoinositides, such as PI(4,5)P2, resulting in their association with or separation from actin filaments (108, 109). Next, nucleation of new actin filaments, mediated by the actin-nucleating activity of the Arp2/3 protein complex, leads to pseudopodia formation. During FcγR-mediated phagocytosis, the GTPase Cdc42 and the lipid PI(4,5)P<sup>2</sup> activate the proteins WASP (Wiskott-Aldrich syndrome protein) and N-WASP (110), which induce activation of Arp2/3 complex at the nascent phagocytic cup (111, 112). Different from this, during CRmediated phagocytosis, actin polymerization is regulated by the GTPase Rho (113). Rho leads to activation of the Arp2/3 complex, via Rho kinase and myosin II (100). The Arp2/3 complex then produces branched actin-filament assembly at the phagocytic cup (100, 114). Rho also promotes accumulation of mDia1, which produces long straight actin filaments at the phagocytic cup (101, 114) (**Figure 5**). Together, these changes help extend membrane protrusions that completely cover the target particle.

The final step for phagosome formation involves fusion of the membrane protrusions at the distal end to close the phagosome. Just before the phagosome is completed, F-actin disappears from the phagocytic cup. It is thought that removal of actin filaments from the phagocytic cup may facilitate curving of the membrane around the particle (115). The mechanism for removing F-actin involves termination of actin polymerization and depolymerization of existing filaments. Both steps seem to be controlled by PI 3-K. Inhibition of this enzyme blocks actin depolymerization at the phagocytic cup and stops pseudopod extension (90). Activation of GTPases is necessary for stimulating the Arp2/3 complex during phagocytosis for actin polymerization (116). But, PI(3,4,5)P3, the product of PI 3-K can stimulate Rho-family GAPs (GTPase activating proteins), which cause deactivation of GTPases and in consequence prevents actin polymerization. In support of this model, it was found that inhibition of PI 3-K led to an increase of activated GTPases at the phagocytic cup (94, 116). In addition, the activity of PI-3K decreases the levels of PI(4,5)P2. This phospholipid activates the Arp2/3 complex, via WASP and N-WASP (110). Thus, its disappearance at the phagocytic cup (111, 112) promotes pseudopod extension (90).

It seems that myosins, actin-binding proteins (117, 118) use their contractile activity to facilitate phagosome formation. In macrophages, it was shown that class II, and IXb myosins were concentrated at the base of phagocytic cups, while myosin Ic increased at the site of phagocytic cup closure, and myosin V appeared after phagosome closure (119). During pseudopod extension, a tight ring of actin filaments moves from the bottom toward the top of the phagocytic cup squeezing the particle to be ingested (120). This contractile activity is dependent of myosin light-chain kinase (MLCK). Thus, myosin II activated by MLCK is required for the contractile activity of phagocytic cups (121). It seems that the squeezing action of the phagocytic cups pushes extra-particle fluid out of the phagosomes. Myosin X is also recruited to phagocytic cups in a PI 3-K-dependent manner, and seems to be important for pseudopod spreading during phagocytosis (122). At the same time, myosin Ic, a subclass of myosin I, concentrates at the tip of the phagocytic cup, implicating it in generating the contraction force that closes the opening of phagocytic cups in a purse-string-like manner (123). Myosin IX also appears in phagocytic cups similarly to myosin II (119, 123). Thus, it is believed that myosin IX is involved in the contractile activity of phagocytic cups. However, it is also possible that myosin IX functions as a signaling molecule for the reorganization of the actin cytoskeleton. This idea is based on the fact that class IX myosins contain a GTPase-activation-protein (GAP) domain that activates the GTPase Rho (124) involved in actin remodeling. Finally, myosin V appears on fully internalized phagosomes. Because class V myosins are involved in vesicular transport in other cell types (125), it is possible that myosin V is responsible for phagosome movement rather than formation of phagosomes (120). Video microscopy experiments have shown that newly formed phagosomes remain within the periphery of the cells for a while, hence it is likely that myosin V mediates the short-range slow movement of newly formed phagosomes (126). Consequently, the described roles of myosins during phagosome formation are: myosin II is involved in phagocytic cup squeezing, myosin X and myosin Ic are responsible for pseudopod extension and phagocytic-cup closing, respectively, myosin IX may activate Rho to direct actin remodeling, and myosin V controls the short-range movement of new phagosomes.

# PHAGOSOME MATURATION

Once internalized the new phagosome transforms its membrane composition and its contents, to become a new vesicle, the phagolysosome, that can degrade the particle ingested. This transformation is known as phagosome maturation, and consists of successive fusion and fission interactions between the new phagosome and early endosomes, late endosomes, and finally lysosomes (4, 127).

### Early Phagosome

The new phagosome combines with early endosomes (3) in a process that involves membrane fusion events regulated by the small GTPase Rab5 (128, 129). Rab5 recruits the molecule EEA1 (early endosome antigen 1), promoting the fusion of the new phagosome with early endosomes (130). EEA1 functions as a bridge between early endosomes and endocytic vesicles (131), and promotes recruitment of other proteins, such as Rab7 (132, 133). Although, the new phagosome combines with several endosomes it does not increase in size because at the same time vesicles, named recycling endosomes, are removed from the phagosome (**Figure 6**).

#### Late Phagosome

As phagosome maturation proceeds, Rab5 is lost, and Rab7 appears on the membrane (133). Then, Rab7 mediates the fusion of the phagosome with late endosomes (134). At the same time, there is an accumulation of V-ATPase molecules on the phagosome membrane. This V-ATPase is responsible for the acidification (pH 5.5–6.0) of the phagosome interior by translocating protons (H+) into the lumen of the phagosome

(135, 136) (**Figure 6**). Also, lysosomal-associated membrane proteins (LAMPs) and luminal proteases (cathepsins and hydrolases) are incorporated from fusion with late endosomes (4, 127) (**Figure 6**).

EEA1, early endosome antigen 1; LAMP, lysosomal-associated membrane protein; NADPH, nicotinamide adenine dinucleotide phosphate oxidase.

#### Phagolysosome

At the last stage of phagosome maturation, phagosomes fuse with lysosomes to become phagolysosomes (3). The phagolysosome is the fundamental microbicidal organelle, equipped with sophisticated mechanisms for degrading microorganisms. First, phagolysosomes are very acidic (pH as low as 4.5) due to the accumulation of many V-ATPase molecules on their membrane (136). The phagolysosome membrane also presents the NADPH oxidase complex, that is responsible for producing reactive oxygen species (ROS), such as superoxide (O2−) (137, 138). Superoxide dismutates to H2O2, which can in turn react with Cl<sup>−</sup> ions to form hypochlorous acid, a very potent microbicidal substance. This last reaction is catalyzed by the enzyme myeloperoxidase (139). In addition, the phagolysosome contains several hydrolytic enzymes, such as cathepsins, proteases, lysozymes, and lipases, which contribute to degrade ingested microorganisms (135) (**Figure 6**).

# PHAGOCYTOSIS-ASSOCIATED RESPONSES

Phagocytosis is not an isolated cell response. It usually occurs together with other cell responses, including formation of reactive oxygen species (ROS) (140, 141), secretion of proinflammatory mediators (142), degranulation of anti-microbial molecules (143, 144), and production of cytokines (142). Cell responses associated to phagocytosis can be controlled by parallel signaling pathways triggered by the same phagocytic receptors. For instance, antibody-dependent phagocytosis in monocytes is controlled by PKC, independently of PI 3-K and ERK (145). However, in the same monocytes, antibody stimulation induces cytokine production via PI 3-K and ERK (145). Phagocytosis and associated cell responses can also be controlled by partially overlapping signaling pathways. For instance, antibodydependent phagocytosis, in macrophages involves the signaling molecules Syk, PI 3-K, PKC, and ERK, but it is independent of an increase in cytosolic calcium concentration (146, 147). In contrast, in neutrophils production of ROS also involves Syk, PI 3-K, PKC, and ERK, but it is dependent on cytosolic calcium (148). Also, in macrophages different PKC isoforms seem to be required either for phagocytosis, or for production of ROS. The isoforms PKCδ and PKCε are involved in regulation of phagocytosis, while PKCα is involved in regulation of ROS production (92). These observations suggest that particular Fcγ receptors can trigger diverse signaling pathways for specific cell responses (55). In support of this idea, in neutrophils in was found that FcγRIIa and FcγRIIIb signal differently for phagocytosis (149), and also for neutrophil extracellular trap (NET) formation (150).

# PHAGOCYTOSIS EFFICIENCY

Most phagocytes have relatively low levels of phagocytosis at resting conditions. However, during inflammation, phagocytes are exposed to a variety of activating stimuli, which increase phagocytosis efficiency. These stimuli include bacterial products, cytokines, and inflammatory mediators. The signaling induced by these stimuli leads to increased stimulation of molecules involved in phagocytosis. For example, leukotriene B4 increases Syk activation and in consequence antibody-dependent phagocytosis (151). Similarly, the activity of PI 3-K and/or ERK, which are essential enzymes for efficient phagocytosis (83), can be enhanced by the bacterial peptide fMLF (152), granulocyte colony-stimulating factor (153), leukotrienes (154), and cytokines such as interleukin 8 (IL-8) (155).

Phagocytosis efficiency can also be regulated by cell differentiation. For example, monocytes have a lower phagocytic capacity than neutrophils and macrophages, but can enhance their phagocytic capacity upon cell differentiation (1, 156). The capacity of monocytes to phagocytize diverse targets changes with their state of differentiation. IgG-opsonized particles are phagocytized better by mature macrophages than by undifferentiated monocytes (83). Similarly, the efficiency of complement-mediated phagocytosis depends on monocyte differentiation (157, 158). How the process of monocyte-tomacrophage differentiation enhances phagocytic capacity is still unknown. It is possible that during cell differentiation the molecular machinery for phagocytosis gets rearranged. In support of this idea, it was found that in monocytes phagocytosis signaling requires PKC, but it does not use PI 3-K and ERK (145). However, during monocyte-to-macrophage differentiation the enzymes PI 3-K and ERK are recruited in an orderly fashion for efficient phagocytosis (159). Similarly, PLA2 is also implicated in regulation of phagocytosis. During phagocytosis, various PLA2 isoforms participate in releasing arachidonic acid from membrane triglyceride lipids. In monocytes, a calcium-independent PLA2, under PKC control is involved in phagocytosis (160, 161), while in macrophages, a calciumdependent PLA2, under ERK and p38MAPK control is involved (162). Thus, during monocyte-to-macrophage differentiation important signaling enzymes are reorganized in order to achieve enhanced phagocytosis.

#### CONCLUSION

Phagocytosis is a fundamental process for the ingestion and elimination of microbial pathogens and apoptotic cells. All

#### REFERENCES


types of cells can perform phagocytosis, but specialized cells called professional phagocytes do it much more efficiently. Phagocytosis is vital, not only for eliminating microbial pathogens, but also for tissue homeostasis. Because there are different types of phagocytic cells and they can ingest a vast number of different targets, it is evident that phagocytosis involves diverse mechanisms. We have presented the main steps of phagocytosis as performed by professional phagocytes and in response mainly to Fcγ receptors. For other phagocytic receptors, we are just beginning to describe the signaling pathways they use to activate phagocytosis. Today, we have a better understanding on the process of phagosome maturation, but there are still many gaps in our knowledge of the signaling pathways regulating this process. Similarly, the resolution of the phagolysosome, after degradation of the ingested particle, is a topic that requires further research. Many important questions remain unsolved. For example, how different phagocytic receptors on the same cell work together? and what is the role different phagocytes in tissue homeostasis? An improved understanding of phagocytosis is essential for future therapeutics related to infections and inflammation.

#### AUTHOR CONTRIBUTIONS

EU-Q prepared the reference list, made the figures and reviewed the manuscript. CR conceived the issues which formed the content of the manuscript and wrote the manuscript.

#### FUNDING

Research in the authors' laboratory was supported by grant 254434 from Consejo Nacional de Ciencia y Tecnología, Mexico.


plasmalemmal subdomain during Fcγ receptor-mediated phagocytosis. J Cell Biol. (2001) 153:1369–80. doi: 10.1083/jcb.153.7.1369


**Conflict of Interest:** 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 © 2020 Uribe-Querol and Rosales. 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.

# Physical Constraints and Forces Involved in Phagocytosis

#### Valentin Jaumouillé\* and Clare M. Waterman

Cell and Developmental Biology Center, National Heart Lung and Blood Institute, National Institutes of Health, Bethesda, MD, United States

Phagocytosis is a specialized process that enables cellular ingestion and clearance of microbes, dead cells and tissue debris that are too large for other endocytic routes. As such, it is an essential component of tissue homeostasis and the innate immune response, and also provides a link to the adaptive immune response. However, ingestion of large particulate materials represents a monumental task for phagocytic cells. It requires profound reorganization of the cell morphology around the target in a controlled manner, which is limited by biophysical constraints. Experimental and theoretical studies have identified critical aspects associated with the interconnected biophysical properties of the receptors, the membrane, and the actin cytoskeleton that can determine the success of large particle internalization. In this review, we will discuss the major physical constraints involved in the formation of a phagosome. Focusing on two of the most-studied types of phagocytic receptors, the Fcγ receptors and the complement receptor 3 (αMβ2 integrin), we will describe the complex molecular mechanisms employed by phagocytes to overcome these physical constraints.

#### Edited by:

Carlos Rosales, National Autonomous University of Mexico, Mexico

#### Reviewed by:

Spencer Freeman, Hospital for Sick Children, Canada Roberto Botelho, Ryerson University, Canada Silvia C. Finnemann, Fordham University, United States

> \*Correspondence: Valentin Jaumouillé valentin.jaumouille@nih.gov

#### Specialty section:

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

Received: 21 February 2020 Accepted: 06 May 2020 Published: 12 June 2020

#### Citation:

Jaumouillé V and Waterman CM (2020) Physical Constraints and Forces Involved in Phagocytosis. Front. Immunol. 11:1097. doi: 10.3389/fimmu.2020.01097 Keywords: phagocytosis, cell mechanics, actin dynamics, membrane, Fc receptors, integrins, receptor diffusion

# INTRODUCTION

Internalization of large particulate material by phagocytosis is a fundamental and well-conserved cellular mechanism of eukaryotic organisms. It enables multiple essential functions from unicellular organisms to arthropods to mammals: uptake of microbes as nutrients by single cell organisms like amoebae, removal of dead cells during tissue development or cell turnover, and clearance of microbes as a first line of defense against infection (1). Seminal work by Korn and Weisman showed that amoeba ingested multiple small particles together within the same vacuole through macropinocytosis, whereas larger particles ≥0.5µm were phagocytosed individually and appeared tightly surrounded by a membrane derived from the plasma membrane (2). In addition, like macropinocytosis, phagocytosis is characterized by its reliance on the actin cytoskeleton, as inhibition of actin polymerization drastically reduces internalization of large particles (3–5).

While most cells can endocytose small molecules or molecular complexes, the capacity to phagocytose larger particles is not equally shared. In mammals, phagocytosis of micron-sized microbes is the prerogative of specialized innate immune cells, namely neutrophils, macrophages, monocytes and dendritic cells, also often referred as "professional phagocytes." Physical characteristics of the particulate material, such as its shape, size and mechanical properties vary for each target and affect the success of internalization (6–9). However, the versatility and engulfment capacity of professional phagocytes is remarkable. For instance, a professional phagocyte can engulf particles substantially larger than themselves, such as IgG-coated microspheres up to 20µm in diameter for bone-marrow derived macrophages that measure about 14µm in suspension, or 11µm IgG-coated microspheres for 4µm human neutrophils (10, 11). How can they achieve such a feat?

# GENERAL PRINCIPLES OF INTERNALIZATION BY PHAGOCYTOSIS Internalization of Large Particles Through Zipper and Trigger Mechanisms

Two fundamentally distinct mechanisms have been proposed for the internalization of large particulate material: the trigger mechanism where discreet signaling elicits formation of actinbased plasma membrane protrusions that non-specifically surround nearby material, and the zipper mechanism where sequential engagement of cell surface receptors to ligands on the target particle leads to a complete wrapping of the particle by the plasma membrane (**Figure 1**). The trigger mechanism is typified by intracellular pathogens like Shigella and Salmonella, which induce their uptake into phagocytes and non-phagocytic cells by injecting effectors via a syringe-like type III secretion system, without relying on adhesion to a specific receptor (12). Those effectors hijack the host cell signaling and actin polymerization machinery to trigger the formation of large ruffles that lead to the internalization of the bacteria in a mechanism that resembles macropinocytosis (13, 14). This was demonstrated by Galán et al., who showed that internalization of a non-invasive strain into epithelial cells could be triggered by the addition of wild type Salmonella (15). In contrast, other pathogens like Yersinia and Listeria employ a zipper mechanism to invade non-phagocytic cells, which requires binding of each invasive bacterium to the host cell receptors β1 integrins and E-cadherin, respectively (12, 16, 17). This illustrates that micron-sized particles like bacteria can be internalized by mechanistically distinct processes defined as trigger and zipper mechanisms. Because the trigger mechanism is limited to a very small number of specific examples, this review will focus on the zipper mechanism which has been demonstrated to mediate phagocytosis across multiple cell and receptor types and for a wide range of target particles.

# Evidence Supporting the Zipper Mechanism for Phagocytosis

A series of foundational studies from Samuel Silverstein's lab demonstrated that phagocytosis occurs through a zipper mechanism (18–20). In a first study, macrophages were exposed to red blood cells (RBC) coated with F(ab')<sup>2</sup> fragments, which do not bind FcγRs and were not internalized. When IgGopsonized bacteria were added, those were internalized, while the F(ab')2-coated RBCs remained on the surface, demonstrating that internalization of RBCs could not be induced by another uptake event, ruling out the trigger model. In contrast, addition of an IgG that bound the F(ab')<sup>2</sup> fragments, providing a ligand for FcγRs, led to the internalization of the RBCs, demonstrating that particle internalization required direct surface receptor engagement, in agreement with the zipper model (18). Next, IgG or complementopsonized RBCs were added to macrophages in conditions allowing binding but preventing internalization. Upon switching to permissive conditions, phagocytosis was prevented if receptors were blocked or if the opsonins were removed on the unengaged surface of the particle (19). This suggested a requirement for circumferential engagement of receptors, which was further demonstrated using lymphocytes coated with IgGs, either uniformly or on only one arc of their circumference. Remarkably, the latter were not internalized unless another IgG that bound their entire surface was added (20). These experiments demonstrated that the initial engagement of phagocytic receptors was not sufficient for particle internalization, but further recruitment of receptors was required to sequentially engage the entire surface of the particle, like a zipper, to drive engulfment. These results were confirmed more recently with asymmetrically IgG-coated "Janus" particles, which were internalized with a lower efficiency than particles evenly coated with the same amount of IgG (21). Contrary to a trigger mechanism, where particles can be captured by ruffles without direct surface-tosurface binding, the zipper model implies a very close interaction between the particle and the phagocyte surface. Experiments using a frustrated phagocytosis model demonstrated that the surface of contact with the macrophage was so tight it excluded molecules as small as 50 kDa (22). Together, these studies demonstrated that phagocytosis occurs through a zipper mechanism, which requires receptor recruitment to tightly engage the entire surface of the target particle.

Given the evidence supporting the zipper model, we will focus on the essential physical constraints associated with the uptake of large particulate material through a zipper mechanism, and the molecular mechanisms employed by professional phagocytes to overcome these constraints. Detailed discussions of the molecular mechanisms underpinning the trigger model can be found in recent reviews (23, 24). In addition, recognition of the surface molecules of phagocytic targets involves a plethora of receptors, which elicit distinct signaling pathways, which have been reviewed elsewhere (25–27). Here we will focus on the mechanisms described for two of the best-studied pathways in mammalian professional phagocytes: Fc-mediated phagocytosis, which involves binding of Immunoglobulin g (IgG) to Fc γ receptors (FcγR), and complement-mediated phagocytosis, which involves binding of the complement molecule iC3b to αMβ2 or αXβ2 integrins, also named complement receptors (CR) 3 and 4, respectively.

# OVERVIEW: PHYSICAL ORCHESTRATION OF PHAGOCYTOSIS

Uptake of large particles represents a physical challenge for the cell. However, while numerous physical constraints could be proposed intuitively, mathematical modeling combined with biophysical measurements and quantitative imaging has helped decipher which physical constraints are likely to be the most critical for phagocytosis. In the following part of this review, we will focus on five physical constraints that appear to be decisive for phagocytosis: (1) cell-surface receptors binding to ligands on the target particle, (2) generation of a protrusive

adhesion to host cell receptors along the entire surface of the particle. Converging evidence demonstrates that phagocytosis occurs through a zipper mechanism.

force to overcome cortical tension to initiate phagocytic cup formation, (3) tangential coupling of the protrusion along the particle surface to advance the phagocytic cup, (4) membrane surface area availability, and (5) membrane fission to close the phagosome and internalize the target particle (**Figure 2**).

#### RECEPTOR BINDING: ROLE OF RECEPTOR AFFINITY, DIFFUSION, AND ACCESSIBILITY

Receptor binding is the first essential aspect of the zipper model. However, it is determined by several parameters: the affinity of the receptors for the ligands, the lateral diffusion of the receptors in the plasma membrane, and the accessibility of the ligands. These parameters are not necessarily fixed and can be dynamically regulated by complex molecular mechanisms.

### Receptor Properties Are Essential Determinants of the Zipper Mechanism

Binding to receptors is imperative for internalization by a zipper mechanism. The dependence of receptor binding on receptor affinity, diffusion and ligand density has been formalized in mathematical modeling (28, 29). In addition, one model suggests that in the absence of actin polymerization to drive protrusion of the phagocytic cup, a passive zipper based on receptor diffusion and random membrane fluctuations could be sufficient to mediate internalization of small particles (30). In that model, internalization is slow, with highly variable phagocytic cups, and requires a low surface tension. Consistent with this model, inhibition of actin polymerization by cytochalasin D does not prevent internalization in 60 min of small IgG-opsonized beads by FcγR-transfected fibroblasts or bone-marrow derived macrophages (BMDM) (30, 31). This suggests that receptor affinity and diffusion are critical for phagocytosis, and in certain circumstances are sufficient to drive internalization.

# Conformational Changes Can Regulate Receptor Affinity

The capacity of a receptor to bind a ligand at equilibrium is defined as its affinity. The FcγR and integrin-based phagocytic receptors have very different properties in terms of regulation by receptor-ligand affinity. The different isoforms of FcγRs have different affinities for the various IgG isotypes (32). However, structural studies showed that binding to FcγRs is not associated with conformational changes, suggesting that their affinity is constant (33). In contrast, α/β integrin heterodimers can switch between three major conformations, which have vastly different affinities for their ligand [**Figure 3**; (34)]. In the absence of stimulus, β2 integrins largely adopt a bent conformation that is associated with a low affinity for their ligand. Engagement of selectins, TLRs, cytokine receptors or immunoreceptors induces inside-out signaling, which involves activation of the GTPase Rap1. This leads to a kindlin- and talin-mediated unfolding of the heterodimer extracellular domains into an extended-closed conformation. Binding of the αI domain to an immobilized ligand enables the actin cytoskeleton to exert a pulling force on the integrin β2 chain through talin, which separates the α and β chains, and rearranges the ligand binding site and the adoption of the extended-open conformation (35, 36). For αLβ2-ICAM-1 interactions, the extended-closed conformation shows an increase of affinity of only 10-fold over

the bent conformation, and the force-mediated opening of αLβ2 increases the affinity of the extended-open conformation for its ligand by over 5,000-fold (37). However, the affinity of iC3b for the various αMβ2 conformations has not been determined as precisely. Nevertheless, the application of a pulling force on the iC3b-αMβ2 bond has been shown to increase the bond lifetime, a phenomenon called catch-bond (38, 39). These observations support the model of a force-based change of the αMβ2 conformation, which drastically increases its affinity for iC3b, and thus likely has important implications for increasing phagocytic efficiency.

# Receptor Diffusion Is Dynamically Regulated by the Actin Cytoskeleton

separation of the phagosome from the plasma membrane.

In addition to affinity, the ability of receptors to find and bind to their ligand is dependent on their lateral diffusion within the plasma membrane (28, 29). However, super resolution microscopy suggested that FcγRs are not evenly distributed at the nanometer scale (40), implying that constraints on their distribution must exist. Indeed, single molecule tracking studies showed that the diffusion of FcγRs along the plane of the plasma membrane is not free, but is heterogeneous and restricted by the membrane-associated actin cytoskeleton (41).

Numerous studies have now shown that the cortical actin cytoskeleton locally constrains the diffusion of both proteins and lipids of the plasma membrane. Together these studies lead to the model of a diffusion constrained by a "fence" formed by the network of actin filaments in the cortex, which form "corrals" that are connected to "pickets" comprised of transmembrane proteins linked, directly or not, to actin filaments (42). Ultra-fast single molecule tracking shows that diffusion within actin corrals is free, but movement between corrals is limited (43). According to the fence and picket model, friction against the pickets would impede the diffusion of molecules within the membrane, including lipids and proteins associated with the outer leaflet of the plasma membrane. It should be noted that while the cortical actin cytoskeleton is dynamic and constantly reorganizes, this occurs slowly compared to the diffusion of mobile membrane proteins and lipids (42, 44). CD44 is an abundant transmembrane protein, that associates with actin filaments through ezrin, and appears to play a major role as a picket in macrophages by restricting the diffusion of FcγRs (45). Interestingly, CD44 also binds hyaluronan, which forms a pericellular coat that curtails FcγR diffusion. This implies that CD44 is a major picket protein in macrophages that restricts FcγR diffusion, and is constrained by two fences, the intracellular actin cytoskeleton and the extracellular hyaluronan network.

The fence that restricts FcγR diffusion is dynamically regulated. Prior to their engagement, FcγR diffusion and clustering can be modulated by tyrosine kinase-mediated reorganization of the actin cytoskeleton, in response to environmental cues sensed through integrins, toll-like receptors (TLR) or cytokine receptors (41, 46–49). Monte Carlo simulations and experimental evidence suggest that a decrease of receptor confinement and receptor clustering affect receptor engagement, implying that environmental cues can prime phagocyte responsiveness via the organization of the cortical actin cytoskeleton (41, 45–47). This effect is rather complex however, as receptor engagement also depends on the density of the ligand at the surface of the target, which can vary greatly in physiological conditions, and the affinity of the receptors for the ligand, which depends on the IgG isotype. On the other hand, diffusion at the surface of bacteria is very limited (50, 51). During phagocytosis, as polymerization increases actin

an extended-open conformation associated with a high affinity for its ligands. The current model suggests that integrins are maintained in an autoinhibited bent conformation by the interaction of the cytosolic domains of the α and the β chains. The switch from a bent to an extended conformation requires binding of Talin to the cytosolic domain of the β chain. The pulling force generated by the actin cytoskeleton through Talin induces the open conformation, when the integrin is attached to an immobile ligand. The ligand iC3b binds through its TED domain to the αI domain of αM. Some evidence suggest that the C345C domain could associate with the βI domain of β2, in addition to the TED-αI association. Contrary to integrins, the structure of FcγRIIA shows no conformational change upon binding to an immunoglobulin G. Average heights estimated from the membrane surface for these receptors in each conformation and for the RO isoform of CD45 are shown. For the extended conformations of αMβ2, the indicated height corresponds to the height of the β-propeller. Talin was not schematized to scale, but its estimated length at a focal adhesion is indicated.

density around the cup, the diffusion of un-engaged FcγRs and even lipids become more restricted within the cup (41, 52). Thus, dynamic regulation of receptor diffusion could favor directional actin polymerization by allowing formation of new signaling clusters at the edge of the phagocytic cup where the FcγRs remain mobile, rather than within the cup where un-engaged receptors are restricted.

In contrast to FcγRs, diffusion properties of αMβ2 integrins are not firmly established. However, by analogy with studies done on αLβ2, they are expected to depend on the integrin conformation. Super resolution microscopy suggests that αLβ2 forms nanoclusters in the absence of stimulation, implying complex diffusional properties (53). Also, whereas the majority αMβ2 or αLβ2 appear to be immobile in resting cells, insideout activation leads to a marked increase of the mobile fraction (54, 55). This is somewhat surprising since, as for other integrins, β2 inside-out activation requires binding of its cytosolic domain to talin (56–58). However, the lower mobility observed at rest could be due to binding of bent β2 integrins to ICAM-1 on the phagocyte's own membrane (59). In addition, consistent with cytoskeleton association through talin, extended-open αLβ2 integrins are relatively immobile (55). The respective contributions of the regulation of integrin affinity by conformational change or avidity by clustering has been a matter of debate. However, it is now apparent that the increase of affinity upon integrin conformational change is so large that, in the case of surface-associated ligands such as iC3b, the force-mediated switch to an extended-open conformation is predominant over diffusion and clustering for αMβ2 engagement.

### Access to the Receptors Limits Target Binding

In addition to diffusion driving a searching behavior, in order for receptors to reach their ligands on the target surface they must be accessible. However, at the surface of phagocytes, the glycocalyx forms a thick layer composed of large highly glycosylated membrane proteins, which can interfere with receptor binding. For instance, because the ectodomain of CD45RO, the main CD45 isoform expressed by macrophages and neutrophils, extends about 22 nm above the membrane, it could sterically block binding of IgG to FcγRs, which form a 11.5 nm complex [**Figure 3**; (60, 61)]. This size difference led to the idea that engagement of immunoreceptors would bring the two surfaces so close it would locally prevent CD45 diffusion into this site because of its larger size, excluding CD45 by "kinetic segregation" (62, 63). Consistent with this, CD45 appears to be excluded from FcγR engagement sites, depending on the length of CD45 ectodomain and the size of the antigen associated with the IgG (64, 65). The segregation of CD45 has important implications because its cytosolic domain is a phosphatase that regulates FcγR signaling. The observation that short antigens induce higher tyrosine phosphorylation of FcγRs and particle internalization than longer antigens, independent of receptor density, confirms the notion that steric constraints can regulate receptor signaling (64). These findings also suggest that the mechanism of CD45 exclusion induced by liquid-liquid phase separation of signaling clusters, as shown for the T cell receptor in a reconstituted system, might not be sufficient to segregate CD45 and FcγRs in macrophages (66). In addition, the close apposition of the two surfaces could facilitate engagement of nearby receptors by kinetic segregation, facilitating the formation of clusters, as proposed for the T cell receptors (62). Thus, CD45 and FcγRs engagement are mutually exclusive by steric constraints. Therefore, while the presence of large surface protein like CD45 can sterically preclude FcγR binding to IgG, CD45 local exclusion can promote FcγR signaling. In addition, a recent report showed in neutrophils that αMβ2 in its bent conformation can bind FcγRs in cis, impeding IgG access to FcγRs (67). This inhibition can be lifted by αMβ2 inside-out activation through cytokines or perhaps the engagement of FcγRs that remain available, facilitating further FcγR binding (54, 67). Thus, the occlusion of FcγRs appears to be a general mechanism that can be tuned dynamically to regulate FcγR binding.

In contrast to FcγRs, whether binding to αMβ2 enables glycocalyx exclusion on phagocytes remains to be explored. Analysis of αLβ2 height in T cells by iPALM showed that the β-propeller domain stands ≈23 nm above the membrane when β2 integrins are activated, to which the length of the αI domain and the ligand should be added (36, 68). This height of αLβ2 measured on cells is in agreement with the heights of αMβ2 and αXβ2 seen in structures obtained by electron microscopy. Although the actin cytoskeleton pulling on ligandbound αLβ2 generates a tilt of the β chain that reduces its height by 4 nm, it still remains close to the height of CD45 (68, 69). This is in stark contrast with the much larger tilt observed for fibronectin-bound αVβ3 integrins on the surface of fibroblasts, which brings the integrin headpiece within a few nanometers from the plasma membrane (70). Consistent with this, binding of integrins to the extracellular matrix (ECM) appears to exclude the glycocalyx in breast cancer cells (71). Similarly, activation of integrins by FcγR signaling promotes the segregation of CD45 and facilitates further engagement of FcγRs (54, 65). Together, these observations suggest that the height of αVβ3 and possibly other integrins can be reduced enough to exclude CD45, while β2 integrins might remain too tall, suggesting that size-dependent kinetic segregation may not operate in the case of β2 integrinmediated processes on leucocytes.

### GENERATION OF PROTRUSIONS BY THE ACTIN CYTOSKELETON

While internalization is possible through receptor binding solely driven by passive diffusion and random membrane fluctuations, this remains slow and highly inefficient for large particles, unless work is produced to promote cell surface deformation (30). Consistent with this notion, multiple models suggest that a protrusive force is required to deform the cell around the target particle to initiate formation of the phagocytic cup (11, 29, 72). Compelling evidence indicate that this protrusive force is generated by the actin cytoskeleton. However, to date, structural information regarding the actin organization within the cup remains limited. Thus, we will look at how the general principles involved in actin-based protrusions, largely learned from studies of cell migration, apply to phagocytosis.

# Actin-Based Protrusion Is a General Feature of Phagocytosis

Actin polymerization facilitates phagocytosis (3–5). Particle binding to FcγRs or many other receptors is associated with the formation of thin actin-filled membrane protrusions, usually called pseudopods, which extend around the targets (25, 73, 74). In contrast, early studies by electron microscopy suggested that αMβ2-mediated phagocytosis occurred by sinking of the particle into the cell body (73, 74), however, thin protrusions surrounding iC3b-opsonized particles have since been observed by electron microscopy (31, 75, 76). Moreover, three-dimensional live cell microscopy revealed the formation of actin-based membrane protrusions in all the observed phagocytic events of iC3b-opsonized particles (77). Thus, formation of actin-based protrusions that extend along the target appears to be a defining feature of phagocytosis, independent of the receptor.

#### Phagosome Formation Is Driven by an Actin-Based Protrusive Force

Formation of protrusions around large particles implies substantial morphological rearrangements. Modeling predicts that as the phagocytic cup grows around larger particles, deforming the cell costs more and more energy (30, 78). Internalization can be reached with a model that combines a repulsive force that pushes the leading edge forward, such as by actin polymerization, and an attractive force that anchors the cytoskeleton tangentially to the membrane engaged by the particle, which guides the protrusion around the particle [**Figure 4**; (11, 72)]. In contrast, a cytoskeletal expansion (gel swelling) model required unlikely parameters and failed to replicate the cup morphology observed experimentally (11).

The cortical actin cytoskeleton not only restricts receptor diffusion but the tension within the network, termed cortical tension, acts as a barrier to cell deformation. Measurements of the cortical tension during FcγR-mediated phagocytosis show that it rises when the surface area increases (from ≈ 33 to 500 pN/µm for a neutrophil engulfing a large particle) and counterbalances the protrusion of the phagocytic cup in a manner that effectively pulls the target particle inward, without requiring direct pulling of the particle by molecular motors (11, 72). Moreover, modeling predicts that the growth rate of the cup size is determined by the balance between the actin-generated protrusive force and the restoring force provided by the cortical tension, creating a bottleneck at the widest point of the particle (29). Consequently, phagocytosis can stall before the protrusion reaches the widest

tension (blue arrows). The rate of deformation of the cell is determined by the ratio of the surface tension and the cytoplasm viscosity, while the surface tension effectively propels the particle inward. The in-line tension of the plasma membrane is compensated by the flattening of surface folds of the membrane and the exocytosis of intracellular vesicles.

point of a spherical particle, but always succeeds once it passes the widest point, implying that there is no requirement for a purse-string mechanism to complete particle envelopment. In addition, the disassembly of the actin cytoskeleton observed at the base of the phagocytic cup after a few minutes of cup formation could locally reduce the cortical tension and therefore facilitate internalization (79). Finally, the energy cost for bending the membrane (≈10−<sup>18</sup> J) is negligible compared to the work exerted against the cortical tension (≈10−<sup>14</sup> J), consistent with the observations that cell surface tension is predominantly due to the actin-based cortical tension (78, 80, 81). These observations imply that the deformation of the cell around the target is driven by actin-generated protrusive forces, while a specific mechanism to bend the lipid bilayer such as BAR-domain containing proteins is not required. Taken together, these observations and models suggest that the major role of actin polymerization is to overcome cortical tension in order to form a protrusion around the particle, rather than pulling the particle inward.

# Comparison of the Actin Organization in Migrating Cells and at the Phagocytic Cup

The broad thin protrusions that advance over the surface of the particle during phagocytosis share many common features with the broad thin protrusions that advance over the extracellular matrix (ECM) at the leading edge of a migrating cell, which is known as the lamellipodium. The lamellipodium is composed of a branched actin network nucleated by the Arp2/3 complex and stabilized by adhesions to the ECM substrate (82–84). The lamellipodium is followed by a less dynamic and thicker region called the lamella, where actin cross-linking proteins and non-muscle myosin II motors organize the actin network into contractile bundles (85–87). Similar to lamellipodia, the Arp2/3 complex is implicated in both FcγR and αMβ2-mediated phagocytosis (31, 77, 88). Live cell SIM-TIRF microscopy of filamentous actin (F-actin) during the formation of a frustrated phagocytic cup upon engagement of αMβ2 suggests the formation of a branched actin network, with Arp2/3 localized at the leading edge, similar to a lamellipodium (77). The branched

actin network of the lamellipodium generates high protrusive forces, ranging from 2 to 10 kPa in migrating cells, which would be well-suited to overcome the increasing surface tension during phagosome formation and advancement (89). Moreover, the structure of a branched network self-adapts to the load generated by membrane tension, which increases the F-actin density and resistance under higher loads (90, 91). These observation suggest that the actin-based protrusions formed during phagocytosis are similar to a lamellipodium, except that the cup shape is imposed by the geometry of the engaged particle (**Figure 5**).

#### Regulating Factors of Actin Dynamics at the Lamellipodium and the Phagocytic Cup

How is Arp2/3 activated during phagocytosis? In lamellipodia, directed actin polymerization involves the recruitment and activation of Arp2/3 at the leading edge, which requires its interaction with the VCA (verprolin, connecting, acidic) domain of a nucleation promoting factor (NPF). In macrophages and neutrophils, the Wiskott-Aldrich syndrome protein (WASP) is an abundantly expressed NPF and is involved in FcγRmediated phagocytosis (92–95). WASP activation requires binding to the GTP-bound Rho-family GTPase Cdc42 and to phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] (95, 96), which are both localized at the tip of protrusions during FcγRmediated phagocytosis (97, 98). Cdc42 is required for FcγRmediated phagocytosis and its recruitment is favored by the adaptor Nck (92, 99–101). The GTPases Rac1 and Rac2 are also activated during FcγR-mediated phagocytosis, though more at the base of the cup, and can activate Arp2/3 through the NPF WAVE2 (98, 102). Rac dominant negative and Rac2 silencing suggested that Rac family proteins are required for FcγR but not for αMβ2-mediated phagocytosis (100, 101). However, rac1 rac2 double-knockout macrophages are defective in both FcγR and αMβ2-mediated phagocytosis (75) and RhoG, a Rac-related GTPase, has been implicated in both pathways (100). Thus, Arp2/3 could be activated at the edge of the phagocytic cup by various Rho family GTPases depending on the engaged receptors.

Actin assembly and motion exhibit a stereotypical organization in protrusive cellular structures. Fluorescent speckle microscopy shows that the lamellipodium is characterized by assembly at the leading edge followed by disassembly a few microns back in a process known as treadmilling (86). In addition to Arp2/3, which localizes in the first micrometer of the leading edge and has a shorter lifetime than actin, actin dynamics are strongly affected by capping proteins, which block incorporation of new monomers at the filament barbed end within ≈ 0.5µm of the edge (103). In contrast, Ena/VASP proteins prevent capping of actin filaments and can affect polymerization by recruiting G-actin-profilin complexes and by reducing branching (104). In macrophages, knockout of the gene coding for the capping protein CapG reduces FcγR and αMβ2 mediated phagocytosis (105). Furthermore, VASP is strongly recruited at the FcγR phagocytic cup, independently of the classic Rho GTPases, and inhibition of Ena/VASP proteins impairs uptake (106). Thus, actin polymerization is regulated by the combined activities of Arp2/3, capping proteins and elongation factors that may mediate treadmilling at the phagocytic cup.

#### Mechanism of Actin Depolymerization at the Phagocytic Cup

As micron-size particles are too large to pass through the mesh of a branched actin network in the cortical cytoskeleton, F-actin disassembly and clearance at the base of the phagocytic cup appears to be essential for particle internalization (79). Remarkably, actin disassembly occurs even upon forced activation of Rac at the phagosome, but coincides with and requires PI(4,5)P<sup>2</sup> hydrolysis by phospholipase C, downstream of PI3K (79). This suggests that actin clearance does not simply require inhibition of Rho GTPase signaling, but active regulation of actin network disassembly. The proteins ADF (actin depolymerization factor), cofilin and gelsolin can sever actin filaments into shorter polymers and accelerate disassembly of the slow growing ends of the filaments. Yet, as severing also creates a new fast growing end, it increases the rate of filament turnover but does not necessarily lead to a reduction in F-actin concentration (107). Aip1/Wdr1 binds to cofilin and causes net depolymerization (108). In addition, the Arp2/3 complex can be inhibited directly by several proteins, including coronins and arpin (109, 110), and coronins can synergize with cofilin to sever ADP F-actin (111). Several studies have reported a role for cofilin in FcγR and αMβ2-mediated phagocytosis (112–114). As cofilin is inhibited by PI(4,5)P<sup>2</sup> (115), it is likely to be inactive at the tip of the protrusions, but become activated and induce actin depolymerization as soon as PI(4,5)P<sup>2</sup> is hydrolyzed at the base of the cup. On the other hand, gelsolin, which is activated by Ca2+, enhances FcγR but not αMβ2-mediated phagocytosis in neutrophils, and is dispensable in macrophages (105, 116). The role of coronin-1 in phagocytosis has been a matter of debate (117–120). However, arpin is recruited to the forming phagosome, reduces F-actin density and enhances uptake by FcγRs, consistent with its activity as an Arp2/3 regulator (121). Thus, actin depolymerization at the base of the cup could be promoted by cofilin, activated upon PI(4,5)P<sup>2</sup> hydrolysis, in conjunction with Arp2/3 inhibition by arpin.

#### Does Myosin II Play a Role in Phagosome Formation?

While non-muscle myosin II has been localized at the phagosome, its contribution to internalization remains unclear (122). Myosin IIA, the predominant isoform expressed in leukocytes, assembles into 320 nm long bipolar filaments with an average of 14 myosin heads at each end of the filament to form a contractile unit (123, 124). These bipolar filaments pull actin filaments into antiparallel bundles, such as dorsal arcs in migrating cells, or concentric arcs at the immunological synapse (125, 126). The polarity of actin at the leading edge and the directionality of myosin motors result in myosin pulling actin in the opposite direction of the leading edge protrusion to drive actin retrograde flow (86, 127, 128).

Despite this putative negative effect on protrusion formation, several studies have suggested a role of myosin II in particle

uptake during FcγR and αMβ2-mediated phagocytosis (129– 131), whereas in other studies, inhibition of myosin II motor activity had no effect on particle internalization (31, 77). Furthermore, while myosin II-driven actin arcs are clearly apparent by SIM-TIRF microscopy at the immunological synapse, no such actin structures are visible in αMβ2-mediated phagocytic cups (77, 126). In addition, traction stresses at αMβ2 phagocytic cups are similar to those measured upon myosin II inhibition in migrating cells (77, 132). Likewise, myosin IIdependent contractility is only observed at the FcγR phagocytic cup after the cell surface increases by over 225% (133). Thus, it seems unlikely that myosin II plays a role in the advancement of the phagocytic cup, but it could participate in other aspects of phagocytosis. Indeed, the conflicting effects of myosin II in phagocytosis might be explained by experiments that combined tracking the displacement of IgG-opsonized beads with realtime measurements of cortical tension, which suggested that particles were not directly pulled by molecular motors, but that the increase of cortical tension effectively propelled particles inward (11). As myosin II activity increases cortical tension (80), it would impede protrusion around the particle, but facilitate particle inward movement (29). The effect of myosin II inhibition might thus be variable for different phagocytes since they exhibit distinct cortical tensions (134). Interestingly, myosin II also promotes actin disassembly at the rear of migrating cells (135) and in the cytokinetic furrow of dividing cells (136, 137). Therefore, a contribution of myosin II in actin clearance at the base of the cup is worth considering.

# COUPLING THE PROTRUDING CUP TO THE PARTICLE SURFACE

The actin cytoskeleton is capable of generating appropriate forces to deform the phagocytic cell around the particulate target. However, quantitative imaging combined with modeling suggests that the actin-based pushing forces must be directed tangential to the particle surface to guide the protrusion around the particle instead of pushing it (11, 72). This implies that actin polymerization should not emanate from the phagocytic receptors, but the growing network should be anchored to the target surface tangential to the direction of the actin polymerization by molecular linkages to the phagocytic receptors or plasma membrane molecules localized at the particle interface. This concept has been previously established for mesenchymal cell migration, where experiments show that coupling of directed actin polymerization oriented tangential to the ECM surface to engaged integrins near the cell leading edge determines cell displacement along the ECM (138, 139). Furthermore, this model is consistent with the case of enteropathogenic Escherichia coli (EPEC), which employs a type III secretion system to inject its own receptor, Tir (140). Tir activates host cell actin polymerization perpendicular to the bacterium-cell interface, which does not result in phagocytosis, but instead leads to the formation of a broad protrusion called a pedestal that elevates the bacterium and makes it surf along the host cell surface (141). This illustrates that the directionality of actin polymerization relative to the target is critical to achieve internalization. Here we will discuss the molecular mechanism that could mediate actin cytoskeleton coupling to the particle surface during phagocytosis.

# The Molecular Clutch Model in Cell Migration

Because protrusion of the phagocytic cup is analogous to the lamellipodium and utilizes a similar machinery, we can extend the analogy with cell migration to learn about the mechanism of coupling the protruding phagocytic cup to the particle surface. In mesenchymal cell migration, coupling of the actin cytoskeleton to the ECM substrate is mediated by an integrinand talin-based "molecular clutch." At the leading edge of the lamellipodium, directional incorporation of actin monomers into the actin network generates a pushing force against the plasma membrane. The plasma membrane provides a resistive force that is large enough that unconstrained actin assembly cannot deform it, and instead the force of actin assembly results in pushing the entire actin network back from the membrane in a process termed retrograde actin flow (138, 142). To instead utilize the pushing force of actin polymerization to drive forward protrusion of the plasma membrane, a resistance force must anchor the actin network to the substrate to prevent it from sliding back. Thus, actin assembly could either drive retrograde flow or forward protrusion, depending on whether the actin is anchored to the substrate or not. Consistent with this notion, the forward movement of the leading edge is inversely proportional to the F-actin retrograde flow rate in lamellipodia of migrating cells (138, 143). Based on these observations, Mitchison and Kirschner proposed that a "molecular clutch" connects the retrograde moving actin cytoskeleton to ECM-bound transmembrane receptors in order to propel the cell forward [**Figure 6**; (144)]. This molecular clutch is composed of focal adhesion (FA) proteins, which transmit actin-generated forces to integrin cytoplasmic tails, creating traction stresses onto the to ECMbound integrin in the same direction as the retrograde flow (132, 139). While several proteins can bind both integrin tails and actin filaments directly, talin is required for cell spreading and lamellipodium stabilization (145). Thus, talin is a molecular clutch protein that couple integrins to the actin cytoskeleton in the lamellipodium of migrating cells.

Vinculin is an important talin binding partner that is thought to regulate the strength of the molecular clutch. Although talin is sufficient to link ligand-bound integrins to the actin cytoskeleton, it is a weak and labile bond (146), and vinculin is thought to reinforce the talin-actin linkage. Indeed, the linkage of ligand-bound integrins to actin retrograde flow generates tension across talin that stretches the molecule, revealing binding sites for the recruitment of vinculin in a force-dependent manner (147–149). As vinculin binds to talin and actin filaments, it reduces slippage of the molecular clutch, slowing down the Factin retrograde flow and increasing traction onto the ECM (150, 151). Vinculin recruitment can also occur in a forceindependent manner when vinculin has an open conformation, which can be promoted by its phosphorylation, or when the adaptor protein paxillin is tyrosine phosphorylated by Focal Adhesion Kinase (FAK) (152–154). Mechanical loading also induces the maturation of small nascent adhesions into larger and stronger vinculin-, zyxin-, and tyrosine phosphorylation-rich FAs, in a RhoA and mDia1 dependent manner, while ROCK and myosin II motor activity are dispensable (155, 156). However, myosin II can also stimulate FA maturation by promoting actin filament bundling (155). Thus, vinculin can be recruited to reinforce the talin molecular clutch in myosin II dependent and independent manners.

# A Molecular Clutch Is Involved in αMβ2-Mediated Phagocytosis

Although the notion of a talin-mediated molecular clutch driving cell migration is well-accepted, what is the evidence that coupling of the actin cytoskeleton to ligand-bound receptors by a molecular clutch promotes phagosome formation? Similar to integrin-mediated migration, talin is required for αMβ2-mediated phagocytosis (56, 57). Moreover, whereas the rod domain of talin, which binds actin filaments, is dispensable for target particle binding, it is required for efficient internalization (57, 77). Vinculin and paxillin are also recruited to the phagosome, and vinculin recruitment is promoted by the Syk, FAK/Pyk2, and Src family tyrosine kinases (74, 77). Since αMβ2-mediated phagocytosis involves RhoA and mDia1, they might also contribute to adhesion maturation (100, 101, 157). Furthermore, traction force microscopy shows that αMβ2 integrins are mechanically coupled to the actin cytoskeleton within the phagocytic cup through talin and vinculin, generating a pulling force tangential to the target surface that drives protrusion of the phagocytic cup (77). Thus, a talin/vinculin-based molecular clutch promotes phagosome formation by coupling actingenerated forces to αMβ2 integrins engaged to iC3b on the particle surface.

#### Mechanical Coupling of the Actin Cytoskeleton Enables Mechanosensing

The talin/vinculin-based molecular cutch is known to be sensitive to the mechanical loading that occurs in response to the stiffness of the integrin-engaged substrate, enabling the regulation of cellular functions, such as cell adhesion, migration and transcriptional regulation (158). During phagocytosis, a major consequence of the mechanical coupling of αMβ2 integrins to actin-generated forces is an increased protrusion speed of the phagocytic cup edge in a manner dependent on the stiffness of the target particle. Consequently, actin-αMβ2 coupling is associated with more efficient uptake of stiff IC3b-opsonized targets, whereas soft targets are poorly internalized. Interestingly, the elastic modulus of Gram-negative and Gram-positive bacteria is in the range of 20–200 MPa (159–162), which is much higher than mammalian cells, which range from 0.2 to 20 kPa (163). Moreover, cells become stiffer during apoptosis, which promotes their internalization independent of the "don't eat me" signal mediated by CD47 (164–166). This implies that the talin/vinculin molecular clutch could contribute to target discrimination for integrin-mediated phagocytosis based on their mechanical properties.

Interestingly, internalization of IgG-opsonized particles is also stiffness sensitive, implying mechano-sensitivity mediated by FcγRs (6, 166). Mechanosensing requires that a force is applied to the target, however no protein is known to couple FcγRs to the actin cytoskeleton. How does mechanosensing occur during FcγR-mediated phagocytosis? One possibility is that FcγRs are not directly connected to the actin cytoskeleton, but the friction generated by cytoskeleton-membrane contacts such as WASP-Arp2/3 interactions, while the actin network flow relative to the membrane could effectively pull on IgG-bound FcγRs (167). Such a "loose clutch" has been proposed for the sweeping of T cell receptors along with actin flow at the immunological synapse, and is consistent with the centripetal motion of FcγR clusters when IgGs are associated with a fluid surface (168, 169). Alternatively, integrins could be responsible for mechanosensing during FcγR-mediated phagocytosis. Engagement of FcγRs activates integrins, and FA proteins such as talin, vinculin, paxillin, and α-Actinin are recruited to the FcγR phagosome (74, 170). High resolution microscopy showed that upon binding to IgG, podosomes are formed several micrometers back from the edge of the phagocytic cup (171, 172). Formation of podosomes instead of FA is typically promoted by Src family kinases, which are activated by FcγRs, combined with low contractility (173–175). The role of integrins in FcγR-mediated phagocytosis had been initially discounted since silencing of talin impaired internalization of iC3b-opsonized, but not IgIopsonized RBCs (57). However, the combination of β1 and β2 integrin blocking antibodies, or the over-expression of talin head domain, which uncouples integrins from the actin cytoskeleton, reduces FcγR-mediated uptake (65). Importantly, contrary to other integrin adhesions, podosomes exert a pushing force normal to the surface, which could cause the indentations observed during phagocytosis of soft IgG-opsonized particles (176, 177). Therefore, by generating a normal pushing force, podosomes might not contribute directly to the leading edge protrusion, but could enable stiffness sensing of IgG-opsonized particle stiffness.

# Receptor-Independent Anchorage of the Actin Cytoskeleton

The postulated need for anchorage of the cytoskeleton to drive FcγR-mediated phagocytosis could also be supported by proteins that link actin to the membrane instead of to the receptors themselves (11, 72). In particular, myosin I is a class of small monomeric motors that bind actin filaments through their head domain and membranes through the TH1 domain of their tail (178). Myosin Ig localizes within the protrusions formed along IgG-opsonized particles, whereas myosin Ie and If precede Factin at the edge of the protrusions (179, 180). The membrane binding tail of myosin Ig is involved in facilitating internalization, consistent with a role in anchoring the cytoskeleton to the membrane (180). Myosin Ie and If however appear to stimulate F-actin turnover within the FcγR-mediated phagocytic cup (179). Finally, the actin cytoskeleton could also be anchored to the plasma membrane through ezrin-radixin-moesin (ERM) family proteins, which can bind to plasma membrane-associated molecules including PI(4,5)P2, EBP50, ICAM, or CD44, as suggested by the localization of ezrin to the forming phagosome (181). Thus, in addition to a molecular clutch, membrane-actin tethering proteins could participate in cytoskeleton anchorage during phagocytosis.

# PROVIDING ENOUGH MEMBRANE TO ENVELOP THE TARGET

In addition to the mechanical constraints involved in deforming the cell, biophysical properties of the membrane are also critical for phagocytosis. The zipper mechanism implies that the target particle becomes entirely enveloped by a membrane. However, as the plasma membrane is essentially inextensible (182), models suggest that the membrane surface area available represents an absolute limit on the internalization of large particles (78). Consistent with this, experiments comparing uptake of beads of various sizes or the extent of spreading during frustrated phagocytosis showed that macrophages reach their limit at a fixed surface area, well before all FcγRs are occupied (10). Indeed, enveloping large particles or multiple smaller particles requires substantial membrane surface, yet neutrophils and macrophages can engulf particles larger than their initial diameter (10, 11). The forming phagosome, at least initially, is derived from invaginations in the plasma membrane (2, 183). However, the plasma membrane is poorly elastic and does not expand more than 2–4% before rupturing (182). This implies that phagocytes need to mobilize extra membrane to their surfaces in order to engulf large or numerous targets.

#### The Different Sources of Membrane

Two types of "membrane reservoirs" appear to act as sources for phagosome formation: folds in the plasma membrane or intracellular vesicles, which upon mobilization and fusion with the plasma membrane increase its surface area. In support of the first model, macrophages, and neutrophils present a very rough surface as they constantly form ruffles, filopodia, and other membrane protrusions or invaginations. Early observations by scanning electron microscopy revealed that the surface of macrophages becomes smoother after phagocytosis, suggesting that membrane folds have been flattened out to provide more membrane surface to the phagosome (184). In support of the second model, Hirsch and Cohn showed that neutrophils degranulate during phagocytosis and suggested that granules might fuse with the phagosome, which was later observed directly by video microscopy (185, 186). Macrophage phagocytic capacity is reduced upon artificial expansion of the surface area of lysosomes or the depolymerization of microtubules, which are required for intracellular organelle movement, suggesting that mobilization of intracellular compartments contributes to the membrane reservoir (10). Different intracellular compartments appear to contribute to the formation of the phagosome, including recycling endosomes and late endosomes, which fuse with the plasma membrane at the forming phagosome in a SNARE protein-dependent manner (187–189). On the other hand, whereas the association of ER proteins with the phagosome suggested that the ER could contribute as a membrane reservoir, multiple experiments suggest that the ER membrane does not fuse with the plasma membrane but is recruited through the interaction of STIM-1 with ORAI, leading to peri-phagosomal Ca2<sup>+</sup> signaling (183, 190). Thus, membrane appears to be provided by endocytic vesicles and granules, but not by the ER during phagosome formation.

# The Role of Membrane Tension

The regulation of membrane reservoir mobilization during phagocytosis remains poorly understood but is likely to involve membrane tension. In cells, membrane tension is defined as the membrane capacity to resist deformation and results from the combination of membrane in-plane tension, membrane bending stiffness and membrane attachment to the actin cortex (191). Experiments using the frustrated phagocytosis model suggest that the mobilization of each membrane reservoir occurs sequentially (192). First the flattening of membrane folds can provide 20 to 40 % of surface area, then exocytosis at the phagocytic cup occurs once the membrane tension reaches its maximum (192). Similarly, during internalization of large particles, the membrane tension measured outside of the phagocytic cup rises from ≈30 to 45 pN. More generally, an increase of plasma membrane tension by a hypotonic shock in macrophages and other cells results in exocytosis (192–194). This suggest that the increase of membrane tension observed during phagocytosis could be the signal that induces exocytosis of membrane reservoirs. Importantly, because the membrane is an inelastic fluid, it is generally assumed that stresses (in-plane tension) equilibrate very rapidly across the entire plasma membrane (193). So, how does focal exocytosis occur at the forming phagosome? Since the membrane tension is affected by the membrane attachment to the actin cytoskeleton, the dramatic actin reorganization at the forming phagosome may in fact locally increase the membrane tension, as has been observed at the leading edge of fast migrating cells (195). The increased membrane tension, along with the aforementioned clearance of F-actin from the base of the cup, may govern the mobilization and local fusion of intracellular membrane reservoirs during phagocytosis. While the signaling pathway(s) orchestrating these events are incompletely understood, it is known to involve the phosphoinositide 3 kinase (PI3K), which is required for internalization of particle larger than 3µm (8, 192, 196). Thus, membrane tension can govern the mobilization of intracellular membrane reservoirs during phagocytosis, through a signaling pathway that is incompletely understood.

#### MEMBRANE FISSION DURING PHAGOSOME CLOSURE

Phagosome closure is the final essential step of particle internalization but arguably the least understood. Phagosome formation occurs when the edges of the advancing phagocytic cup reach a point of contact and merge, leading to the fission of the membrane that releases the phagosome from the plasma membrane. This process is one of the most difficult aspects of phagocytosis to study because it is challenging to identify fully wrapped but unclosed phagosomes (197). Membrane fission requires bringing sites of a continuous membrane to <3 nm apart to induce merging of the outer leaflet to form an hemifission neck, followed by merging of the inner leaflet to allow separation (198). Interestingly, the observation that when two macrophages try to engulf the same target they do not fuse together at the contact site, suggests that phagosome closure involves a molecular mechanism distinct from previously described cell fusion mechanisms (122). When physical constraints are minimal, several mechanisms can elicit membrane fission without energy consumption in vitro. However, given the membrane tension measured during phagocytosis and the physical barrier created by the extracellular domains of surface proteins, which can impede contact between lipid bilayers, phagosome closure is likely to involve an active mechanism to drive membrane fission (199). Current evidence suggests the role of two possibly synergistic mechanisms: membrane constriction by mechanochemical proteins and membrane pushing by the actin cytoskeleton.

#### Mechanochemical Proteins Involved in Membrane Constriction

Dynamin is the first and best characterized protein known to induce membrane fission and is involved in various endocytosis and organelle division pathways (200, 201). While it is not a molecular motor in the classical sense, dynamin assembles into a ring-shaped polymer that has contractile properties through its enzymatically driven hydrolysis of GTP. Constriction of the ring from a 20 nm inner diameter to 3.7 nm is achieved by a GTP-dependent conformational change by twisting, while the fission event is promoted by membrane tension (202, 203). Dynamin-2 is ubiquitously expressed and is recruited to FcγR and αMβ2-mediated phagosomes concomitantly with F-actin (204, 205). More importantly, dynamin-2 has been visualized at the phagosome closure site in a TIRFbased assay, and inhibition of dynamin activity reduces internalization. Interestingly, dynamin inhibition inhibits protrusion formation, suggesting that dynamin cross-talks with actin dynamics (205). Thus, dynamin-2 might interact with actin filaments at the edges of the phagocytic cup and induce membrane fission when the edges converge into a closure site.

In addition to dynamin-2, myosin Ic is a molecular motor recruited to the phagosome at the late stage of FcγR-mediated phagocytosis (122). Interestingly, when one IgG-opsonized RBC is phagocytosed simultaneously by two macrophages, myosin Ic localizes to the meeting point of the opposing phagocytic cups, suggesting a role in phagosome closure (122). There is no clear evidence so far that myosin I family proteins could mediate constriction. However, as a membrane-actin tether, myosin I can increase membrane tension, which could facilitate dynamin-mediated membrane fission (206). Unlike myosin Ic, myosin II has not been localized to the meeting point (122). Because myosin II plays an important role in the formation of the constriction ring during cytokinesis, it has been proposed that myosin II could assist phagosome closure by a pursestring mechanism. However, it should be noted that myosin II is required to maintain cortical tension during cytokinesis through actin bundling, whereas its motor activity is dispensable for cell division in culture and in vivo (207). Furthermore, the 320 nm length of myosin II contractile units makes a role in membrane fission, which occurs at a much smaller scale, highly unlikely (123). Thus, myosin I proteins are the myosins that are the most likely to contribute to membrane fission during phagocytosis.

#### Does Actin-Mediated Pushing Force Participate in Membrane Fission?

Actin-based protrusive forces could also facilitate membrane fission during phagosome closure. In yeast, a major role of Arp2/3-mediated actin polymerization is to support clathrinmediated endocytosis (208). The current model suggests that actin polymerization pushes against the plasma membrane in the area of membrane bending, where WASP is localized. Myosin I and the dynamin protein Vps1 form a ring around the clathrincoated pit, which can tether the F-actin network to the forming endosome (209, 210). This tethering could enable F-actin retrograde flow to overcome the membrane tension and turgor pressure to pull the forming endosome away from the surface, consistent with evidence that myosin I primarily contributes to endosome inward movement (211). Furthermore, pushing force generated by actin polymerization could also facilitate membrane fission during phagocytosis by bringing the lipid bilayers closer at the closure site. Interestingly, a burst of actin polymerization is often visible at the point of closure and appears to push the phagosome away from the surface during αMβ2-mediated phagocytosis (77). Furthermore, a normal stress of about 150 Pa is observed on fully wrapped, IgG-opsonized soft particles, indicating that a pushing force, presumably generated by actin polymerization, propels the particle inward (177). Taken together these observations support the idea that actin polymerization, tethered to the membrane by dynamin-2, myosin I, and/or a talin-based molecular clutch, can promote membrane fission during phagosome closure.

# CONCLUDING REMARKS

Since phagocytosis is a cellular process broadly employed across eukaryotes, it seems likely that phagocytes employ fundamentally shared molecular mechanisms to overcome the physical constraints imposed by the internalization of large particulate material by cells. The comparison of phagocytosis with other general cellular processes, such as cell migration, shape change, cell division, or endocytosis, is very helpful to understand the molecular mechanisms that are at play. However, it also highlights the fact that we still only have pieces of the puzzle, which are largely consistent with the general framework, yet it will take major effort to complete the details of the molecular mechanisms involved and to understand how they are coordinated to enable uptake. It should be noted that while we have learned a lot from studies on the canonical receptors FcγR and αMβ2 with model particles like microspheres and red blood cells, we can speculate that the general concepts exposed here will apply broadly to phagocytosis in physiological conditions, with a number of nuances and additional physical constraints.

For instance, microbes not only vary in size but can exhibit an array of diverse shapes, which present distinct physical constraints. It has long been observed that elongated bacteria like Legionella pneumophila and Borrelia burgdorferi, hyphal fungi like Candida albicans, and parasites like Leishmania and Trypanosoma cruzi are engulfed by so-called "coiling phagocytosis," in which phagocyte membrane protrusions wrap around the complex morphology of these microbes (212–214). In many cases the molecular mechanisms have only been partially explored. Yet, commonalities with the concepts described in this review are evident from the role of receptor binding and the formation of actin-based protrusions, which involve Arp2/3 and formin activation and signaling similar to those described for lamellipodia and filopodia formation (215). However, uptake of elongated or spiral shaped microbes or synthetic particles is generally less effective than that of spherical particles, and several biophysical models have suggested increased mechanical constraints linked to the uptake of complex shapes (7, 9, 28). For instance, prolate spheres or rod particles, a very common shape for microbes, can bind to phagocytes efficiently, but their internalization is markedly reduced compared to that of spherical particles (216). Remarkably, elongated particles are internalized more efficiently when their initial contact with the phagocyte occurs at the pole rather than the side (7). This phenomenon can be recapitulated in a two-stage model that combine passive diffusion-based receptor binding followed by a stage that actively promotes further receptor engagement (28). Thus, even for complex shapes, the same scheme of receptor binding, actin-based protrusion and coupling between the actin and particle-engaged receptors seems to apply. However, the mechanical burden associated with the formation of more geometrically complex phagocytic cups can become unsurmountable for the phagocyte. Consequently, elongated microbes can at least partially escape killing by phagocytosis thanks to their morphology, as observed for Candida albicans (217).

The diversity of phagocytic targets is managed by the expression by phagocytes of a plethora of different phagocytic receptors, which are able to recognize various opsonins, pathogen-associated molecular patterns and "eat me signals" (27, 218). While the engagement of these receptors likely involves similar concepts to those described here for FcγR and αMβ2, such as receptor lateral diffusion, formation of signaling clusters and activation of actin polymerization, specific properties of these receptors could vary greatly, and in most cases, remain largely under-characterized. For instance, the lateral diffusion of the scavenger receptor CD36 appears to be restricted by the cortical actin cytoskeleton, but displays anisotropic trajectories very distinct from those observed for FcγR or β2 integrins (41, 55, 219). Also, CD45 can inhibit signaling by the Ctype lectin receptor Dectin-1, and appears to be excluded from the Dectin-1-mediated phagocytic cup (220). This is consistent with the presumed small size of Dectin-1 and the kinetic segregation model established for immunoreceptors. However, it is clear that the dimensions of phagocytic receptors vary dramatically, suggesting that some phagocytic events are likely to occur independent of a size-based segregation mechanism to regulate signaling (221). Thus, it is likely that the physical constraints and concepts presented here are broadly shared across the different phagocytic pathways, yet the details of their implications could vary and should be examined specifically for each individual case.

Finally, while this review is focused on the mechanics of internalization mechanisms, killing, processing and disposing of internalized material can represent tremendous constraints and limit phagocytosis capacity. In particular, cell turnover, and tissue homeostasis probably represent the largest burden on phagocytes, as it has been estimated that in humans, 200–300 billion cells are replaced every day (222). Given the scale of this task, it is not surprising that phagocytosis of dead cells, also called efferocytosis, must be shared between many cells, including professional and non-professional phagocytes, such as Sertoli cells and retinal pigmented epithelial cells. Moreover, in addition to the membrane surface area initially available on the phagocyte, the phagocytic capacity over time can be limited by the rate of degradation of the internalized material, which involves activation of appropriate enzymatic and metabolic pathways (223). It is therefore conceivable that phagocytes gather information regarding the physical properties of the ingested material, including their size and stiffness, in order to regulate their processing programs (25). How sensing of physical and molecular cues is integrated to regulate the broad range of phagocyte functions remains largely unknown and will be an exciting problem for the coming years.

#### AUTHOR CONTRIBUTIONS

VJ and CW wrote the manuscript. VJ prepared the figures.

#### FUNDING

VJ and CW are supported by the NHLBI Division of Intramural Research.

# REFERENCES


localized to cellular regions of dynamic actin filament assembly. J Cell Biol. (1997) 138:375–84. doi: 10.1083/jcb.138.2.375


internalization by macrophages. J Control Release. (2010) 147:408– 12. doi: 10.1016/j.jconrel.2010.07.116


**Conflict of Interest:** 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 © 2020 Jaumouillé and Waterman. 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.

# Microglia Inhibition Delays Retinal Degeneration Due to MerTK Phagocytosis Receptor Deficiency

Deborah S. Lew† , Francesca Mazzoni † and Silvia C. Finnemann\*

*Department of Biological Sciences, Center for Cancer, Genetic Diseases and Gene Regulation, Fordham University, Bronx, NY, United States*

#### Edited by:

*Carlos Rosales, National Autonomous University of Mexico, Mexico*

#### Reviewed by:

*Lian Zhao, National Eye Institute (NEI), United States Hemant Khanna, University of Massachusetts Medical School, United States*

> \*Correspondence: *Silvia C. Finnemann finnemann@fordham.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: *01 February 2020* Accepted: *05 June 2020* Published: *16 July 2020*

#### Citation:

*Lew DS, Mazzoni F and Finnemann SC (2020) Microglia Inhibition Delays Retinal Degeneration Due to MerTK Phagocytosis Receptor Deficiency. Front. Immunol. 11:1463. doi: 10.3389/fimmu.2020.01463* Retinitis Pigmentosa (RP) is a group of inherited retinal diseases characterized by progressive loss of rod followed by cone photoreceptors. An especially early onset form of RP with blindness in teenage years is caused by mutations in *mertk*, the gene encoding the clearance phagocytosis receptor Mer tyrosine kinase (MerTK). The cause for blindness in mutant MerTK-associated RP (mutMerTK-RP) is the failure of retinal pigment epithelial cells in diurnal phagocytosis of spent photoreceptor outer segment debris. However, the early onset and very fast progression of degeneration in mutMerTK-RP remains unexplained. Here, we explored the role of microglia in the Royal College of Surgeons (RCS) rat model of mutMerTK-RP. We found elevated levels of inflammatory cytokines and CD68 microglia activation marker, and more ionized calcium-binding adapter molecule 1 (Iba-1) positive microglia in RCS retina when compared to wild-type retina as early as postnatal day 14 (P14). Strikingly, renewal of photoreceptor outer segments in P14 wild-type rat retina is still immature with low levels of RPE phagocytosis implying that at this early age lack of this process in RCS rats is unlikely to distress photoreceptors. Although the total number of Iba-1 positive retinal microglia remains constant from P14 to P30, we observed increasing numbers of microglia in the outer retina from P20 implying migration to the outer retina before onset of photoreceptor cell death at ∼P25. Iba-1 and CD68 levels also increase in the retina during this time period suggesting microglia activation. To determine whether microglia affect the degenerative process, we suppressed retinal microglia *in vivo* using tamoxifen or a combination of tamoxifen and liposomal clodronate. Treatments partly prevented elevation of Iba-1 and CD68 and relocalization of microglia. Moreover, treatments led to partial but significant retention of photoreceptor viability and photoreceptor function. We conclude that loss of the phagocytosis receptor MerTK causes microglia activation and relocalization in the retina before lack of RPE phagocytosis causes overt retinal degeneration, and that microglia activities accelerate loss of photoreceptors in mutMerTK-RP. These results suggest that therapies targeting microglia may delay onset and slow the progression of this blinding disease.

Keywords: retinitis pigmentosa, MerTK, phagocytosis, microglia, retinal degeneration

# INTRODUCTION

Retinitis Pigmentosa (RP) is a heterogeneous group of inherited retinal degenerations that show rod photoreceptor defects and loss before secondary loss of cone photoreceptors resulting in blindness (1). To date, mutations in over a hundred genes mostly expressed in rod photoreceptors or the adjacent retinal pigment epithelium (RPE) are known to lead to RP (2, 3). Although RP may progress at different rates even among patients with mutations in the same gene, usually rod dysfunction leads to night blindness in late teenage age followed by rod death by apoptosis and loss of peripheral vision (tunnel vision). Eventually, secondary "bystander" death of cone photoreceptors will lead to blindness usually in middle age (1). Mutations of the mertk gene encoding the Mer receptor tyrosine kinase (MerTK) cause an exceptionally severe form of RP in human patients with childhood onset and blindness in teenage years (4–10). No therapy is available to date for mutant MerTK-associated RP (mutMerTK-RP) that will prevent or even delay progression to blindness.

Disease manifestation in mutMerTK-RP has been elucidated exploring animal models that mimic well the human disease. The Royal College of Surgeons (RCS) rat strain was recognized as model retinal degeneration in the 1960's and has since been studied extensively (11). The RCS rat genome carries a deletion in the coding sequence of the mertk gene resulting in an aberrant transcript encoding only 20 of 999 amino acids (12, 13). No MerTK protein is expressed and thus RCS rats are a natural null strain for MerTK. Acute re-expression of MerTK significantly but not completely decreases the severity of RCS rat retinal degeneration (14–16). Mice engineered to lack mertk gene activity (mertk−/<sup>−</sup> mice) fully phenocopy the RCS rat further confirming that loss of functional MerTK causes the RCS rat RP phenotype (17).

Mechanistically, MerTK functions as engulfment receptor in non-inflammatory apoptotic cell clearance phagocytosis, a specialized form of phagocytosis also known as efferocytosis. As essential aspect of the life-long and continuous process of photoreceptor outer segment renewal, clearance phagocytosis by the RPE removes spent photoreceptor outer segment tips from the outer retina in a strict diurnal rhythm (18, 19). MerTKdeficient RPE cells fail to engulf spent outer segment tips (20, 21) causing outersegment debris to build up in the outer retina where it is thought to distress photoreceptors such that they die by apoptotic cell death (22). Outer segment renewal including daily clearance phagocytosis in rodents commences after postnatal formation of photoreceptor outer segments and has been assumed to be associated with retinal maturation around eye opening, which in our RCS rats occurs at postnatal day 15 (P15) plus or minus one day. No significant debris buildup is observed before P22, yet rod photoreceptor apoptosis is substantial only shortly thereafter, at ∼P25 (22, 23). Retinal function as measured by electroretinography shows normal photoreceptor response at eye opening, and modestly and severely reduced photoreceptor responses at P22 and P34, respectively (11). Altogether, it has long been assumed that photoreceptor degeneration in mutMerTK-RP occurs only after eye opening, subsequent to postnatal retinal maturation and as consequence of defective RPE phagocytosis. Here, we investigate whether inflammatory mechanisms contribute to its remarkably early onset and rapid speed of progression.

Recent studies investigating the role of microglia in several forms of retinal degeneration suggest a dual role for these highly mobile cells. Upon acute damage, as following retinal detachment, microglia may prevent photoreceptor death acting in a protective role (24). Conversely, in chronic damage, such as seen in hereditary retinal dystrophies in mouse models, microglia may contribute to retinal cell death and accelerate retinal degeneration (25–28). In late stage human and rodent RP increased levels of activated microglia have been reported (29, 30). Likewise, in both mouse and rat mutMerTK-RP, microglia activation have been reported at P35 and P50, the late stage of the disease (31, 32). Strikingly, mice constitutively lacking the pro-inflammatory cytokine CCL3 in addition to MerTK (mertk−/<sup>−</sup> ccl3−/<sup>−</sup> double knockout mice), exhibit decreased retinal microglia activation and retain more photoreceptor cells than mertk−/<sup>−</sup> mice by P56 but only this advanced stage of the disease was tested (33). Moreover, RCS rats show activated microglia in the photoreceptor layer of the retina at P21, an early stage of disease (34). Altogether, these intriguing data suggest that microglia may play a role in mutMerTK-RP. Here, we asked if microglia activation has a primary role in the early onset of retinal degeneration due to MerTK deficiency. We found that molecular markers indicating microglia activation are already elevated at P14, an age just before eye opening when RPE phagocytosis of photoreceptor outer segment fragments has not yet reached its mature level, and that microglia relocalize to the photoreceptor layer of the retina starting at P20. We then used three different experimental paradigms to suppress early microglia activation in RCS rats. Assessment of photoreceptor retention and function showed that inactivating microglia at an early age prolonged photoreceptor survival and retinal function.

# MATERIALS AND METHODS

## Reagents

All reagents were from Millipore-Sigma (St. Louis, MO) or Thermofisher (Carlsbad, CA) unless indicated.

#### Animals and Tissue Processing

Animals were housed in a 12-h light/12-h dark light cycle with food and water ad libitum. Animals of both sexes were used. Wild-type (WT) and mertk−/<sup>−</sup> mice in the same 129T2/SvEmsJ genetic background (23), pink-eyed dystrophic RCS rats (rdy/rdy-p) and Sprague Dawley (SD) WT rats were raised to yield litters at defined age for experiments. For tissue harvest, animals were euthanized by CO<sup>2</sup> asphyxiation before immediate eye enucleation and processing. Unless indicated otherwise, all tissue harvest was done at 3–4 h after light onset to avoid variability due to circadian effects. Eyeballs were chilled, dissected and tissue fractions flash-frozen for immunoblotting. For tissue sectioning, cornea and lens were dissected before fixation of tissue in 4% paraformaldehyde in PBS for 30 min followed by sequential dehydration and embedding

#### TABLE 1 | RT-PCR primers used.


in paraffin. To generate whole mount preparations of posterior eyecups, cornea, lens and retina were dissected from eyes harvested from WT rats sacrificed 1 h after light onset. Eyecups were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer followed by radial cuts to flatten the eyecup in preparation for immunofluorescence labeling.

#### Drug Treatments

To start localized drug treatment before onset of microglia activation we administered tamoxifen (tmx) as eye drops starting at P10. Schlecht et al. previously demonstrated that tmx eye drops applied to closed eyelids of mice from P8 through P12 are sufficient to yield efficient Cre induction in the posterior retina implying that tmx reaches the posterior eye if applied to closed eyelids (35). Here, RCS rats received eye drops of 10 µl of 5 mg/ml tmx in corn oil into both eyes 3 times a day. To administer tmx systemically, rats were fed tmx-supplemented chow (500 mg/kg, Envigo, South Easton, MA, #130858) ad libitum starting at weaning (P19). Liposomal clodronate (LC, Liposoma, Amsterdam, The Netherlands) was administered at 10 µl LC/g body weight by intraperitoneal injections every 7 days starting at P13 and at 4 µl /eye by intravitreal injection once the day after eye opening (at P16 or P17). For combined tmx and LC administration, rats received tmx eye drops and the LC treatment as described above. Control siblings were manipulated identically but received corn oil-only eye drops and PBS injections. For all treatments, ERGs were recorded at P33 followed by continued treatment until sacrifice and tissue harvest at P40.

#### Electroretinogram (ERG) Recordings

The entire procedure was carried out under dim red light. RCS rats were dark-adapted overnight before intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine to induce anesthesia. Scotopic responses were recorded exactly as described previously using a UTAS-E2000 visual electrodiagnostic system (LKC Technologies, Gaithersburg, MD) (23). Stimuli were presented in order of increasing intensity as a series of white flashes of 1.5 cd-s/m2 attenuated with neutral density filters. For each flash intensity, three to six recordings were averaged. For all recordings, a-wave amplitudes were measured from the baseline to the trough of the a-wave, and b-wave amplitudes were measured from the trough of the a-wave to the peak of

# RNA Extraction and RT-PCR

the b-wave.

Two dissected neural retinas from a single animal were pooled and processed following the manufacturer's direction using the Qiagen RNeasy Plus Mini kit (Qiagen, Waltham, MA). Purity and concentration of each sample were analyzed by spectrophotometry, and 5 ng/µl RNA stocks were stored at −20◦C. RT-PCRs on 10 ng RNA were performed using the Qiagen One-Step RT-PCR kit. Primer sequences are listed in **Table 1**. Quantification of bands following product electrophoresis was performed using ImageJ.

#### Immunofluorescence Staining and Tissue Analysis

Posterior eyecup whole mount preparations were stained with rhodopsin antibody B6-30 and AlexaFluor488-conjugated secondary antibody to label phagosomes containing photoreceptor outer segment particles, and counterstained with AlexaFluor594-labeled phalloidin and DAPI nuclear dye (36). Phagosome counts were obtained using fixed value of threshold settings in ImageJ including only rhodopsin positive particles with 0.5µm diameter and above.

Seven micrometer thick microtome sections cut within 200µm from the optic nerve were deparaffinized. Epitope unmasking was performed by boiling for 10 min in 10 mM citric acid, 0.05% Tween-20, pH 6. Sections were then blocked with PBS, 1% BSA, 0.01% Triton-X100 and incubated sequentially with primary and appropriate AlexaFluorconjugated secondary antibodies. Primary antibodies used were to ionized calcium-binding adapter molecule 1 (Iba-1) (1:400, Fujifilm Wako Chemicals, Richmond, VA, #019-19741) and to CD68 (1:500, Biorad, Hercules, CA, #MCA341R). DAPI was used to counterstain nuclei.

Vectashield mounted samples were imaged using a Leica TSP5 laser scanning confocal microscopy system. X-Y stacks were collapsed to yield representative maximal projections for image quantification.

Quantification of images was performed manually aided by ImageJ software. To quantify numbers of CD68 or Iba-1 positive microglia in tissue sections, counts from 3 sections for each tissue were averaged to obtain the number of microglia per 350µm of retina, and a minimum of 3 tissues from 3 animals per sample type were analyzed. To quantify the number of photoreceptor nuclei rows in tissue sections, in each section the number of photoreceptor rows was counted in 3 regions of the image (left, central, and right) and averaged.

#### SDS PAGE and Immunoblotting

Dissected single tissues were lysed in HNTG buffer (50 mM HEPES, 150 mM NaCl, 1% Triton-X100, 10% glycerol, pH 7.5)

freshly supplemented with protease inhibitor cocktail. Proteins from cleared lysates were separated by standard SDS-PAGE and transferred to nitrocellulose membranes, Blots were blocked in 10% non-fat milk powder in TBS before incubation with primary and appropriate HRP-conjugated secondary antibodies, and enhanced chemiluminescence digital detection by a KwikQuant Imager (Kindle Biosciences, Greenwich, CT). Band densities were quantified using the KwikQuant imaging software. Primary antibodies used were to CD68 (1:1,000, Biorad #MCA341R), Iba-1 (1:2,000, Fujifilm Wako Chemicals, #016-20001), PSD95 (1:3,000, Cell Signaling, Danvers, MA, #3450), and α-tubulin (1:3,000, Cell Signaling, #9099). For all immunoblotting experiments, tissues from 2 different rats of each sample group were directly compared in 3 independent experiments. Tubulin reprobing of membranes was used to control for sample load.

Levels are shown relative to WT. Data were analyzed by Student *t*-test; \*\* indicates *p* < 0.01, \* indicates *p* < 0.05.

#### Statistical Analysis

All data were collected from at least three independent experiments. The means and standard deviations were calculated for each comparison group. Comparisons between two groups were performed using the Student's two-tailed t-test. Comparisons between three or more groups were performed using one-way or two-way ANOVA as appropriate, with Tukey's post-hoc test for comparison of two groups within multiple groups. P values below 0.05 were considered statistically significant for all experiments.

# RESULTS

### The Pro-inflammatory Cytokine CCL5 and Microglia Activation Marker Iba-1 Are Elevated Even Prior to Eye Opening in RCS Rat Retina

Cytokine secretion is one of the first indications of tissue inflammation. Once secreted these small molecules serve to attract inflammatory cells expressing specific cytokine receptors, causing migration to inflammatory sites. As reported earlier by others, inflammatory cytokines in mertk−/<sup>−</sup> mouse retina are present at P35, a late disease stage (**Supplementary Figure 1A**). Strikingly, we found levels of mRNA for CCL5 (Chemokine C-C motif ligand 5) already elevated at P14, the day of or after eye opening of our mouse strains (**Supplementary Figure 1B**). CCL5 is a pro-inflammatory cytokine whose elevation has been linked to retinal stress associated with glaucoma and agerelated macular degeneration (37, 38). To determine whether this observation was characteristic for MerTK deficient retina, we next tested CCL5 levels in RCS rat retina. Indeed, we found that CCL5 is also significantly elevated in RCS retina at P14, one day prior to eye opening in our rats (**Figure 1**). Compared to WT retina, RCS retina contained 3.4-fold higher levels of CCL5 transcripts at P14 (**Figures 1A,B**) as compared to ∼2 fold higher levels at P20 (**Figures 1C,D**), and P35 (**Figures 1E,F**), the previously recognized early and late stage of RCS retina degeneration, respectively. This was somewhat unexpected

because diurnal photoreceptor outer segment renewal is thought to commence in rodent retina only after eye opening, and thus any process occurring prior is unlikely to be a consequence of its deficiency due to the RPE phagocytosis defect. Unlike CCL5 transcripts, levels of GAPDH transcripts were the same in WT and age-matched RCS retina indicating that the change in CCL5 was not due to global change in transcription (**Figure 1**). Motivated by prior reports on microglia activation in RCS rat retina, we next compared levels of the microglia markers Iba-1 and CD68 between RCS and WT rat dissected neural retina. Resident tissue microglia upregulate basal levels of Iba-1 upon activation and this change correlates with their migration to sites of tissue injury (39, 40). CD68 is expressed only in activated retinal microglia (41). **Figure 2** shows that already at P14, Iba-1 and CD68 protein levels in RCS retina are 2.2-fold and 4.9-fold higher than levels in WT rat retina, respectively (**Figures 2A,B**). By P20, Iba-1 levels had increased to 2.6-fold of WT levels, while CD68 levels were 1.9-fold elevated over WT (**Figures 2C,D**). We noted that CD68 levels relative to tubulin loading control increased little in RCS retina from P14 to P20, while CD68 in WT retina rose 2.9-fold. By P35, Iba-1 and CD68 levels in RCS retina were 12-fold and 23-fold of WT levels, respectively (**Figures 2E,F**). Differences in microglia marker content between RCS and WT tissues were more pronounced in dissected retina samples than in whole eye samples indicating a primary role for retinal rather than choroidal microglia (**Figure 2**, compare F and G). As control for global neural tissue changes, we also measured levels of the synaptic marker PSD95 at all ages. We found no differences between RCS and WT PSD95 at P14 and P20 and only modest decline at P35, which did not reach statistical significance (**Figure 2**, panels and bars as indicated). Up to P35, tubulin levels were identical in age-matched RCS and WT rat retina confirming that overall retina cell loss is negligible by this age. All protein levels were thus quantified relative to tubulin levels in the same sample and development to account for differences in tissue yield during dissections. Altogether, these results suggested that inflammatory signaling and responses by microglia cells in RCS rat retina commence before eye opening.

#### RPE Phagocytosis in P14 Wild-Type Rat Retina Just Prior to Eye Opening Is Only Partially Active

It has long been assumed that outer segment renewal in rodents starts after eye opening. The WT rats tested in our study fully open eyes at P15. Here, we measured outer segment phagosome load of the RPE in situ as quantitative assessment of outer segment renewal. **Figure 3** shows rhodopsinpositive phagosomes in WT rat RPE and their quantification 1 h after light onset, when phagosome load is at its peak.

Compared to P11, RPE phagocytosis at P14 one day prior to eye opening is about 2-fold more active (**Figures 3A,B**). However, RPE phagosome load is still 44% lower at P14 compared to P16, one day after eye opening (**Figures 3A,B**). Notably, we found no difference in total levels of retinal or RPE marker proteins including the rod outer segment marker rhodopsin between P14 and P16 but less rhodopsin at P11 in WT rat eyes (**Figures 3C,D**). These results suggest that outer segment formation is complete by P14, 1 day before eye opening, but outer segment renewal is not yet fully active. Eye opening of our RCS rats between P15 and P16 implies that at P14 there is little to no debris accumulation that could distress photoreceptors, and that therefore cytokine elevation in P14 RCS retina is highly unlikely a consequence of lack of RPE phagocytosis.

## Microglia Migrate to Reach the Photoreceptor Layer of RCS Rat Retina as Early as P20

We next determined the localization of Iba-1 positive retinal microglia in RCS rat retina with increasing age. As expected, WT retina harbored microglia only in the inner retina regardless of age (**Figure 4A**, WT panels as indicated). In contrast, all RCS retina samples tested showed microglia in the inner as well as the outer retina starting at P20 (**Figure 4A**, RCS panels as indicated). Quantification of Iba-1 positive cells over the entire retina showed that the total number of Iba-1 positive cells in RCS rat retina did not change from P14 to P30, although it was 1.8-fold higher than cell numbers in WT retina as expected from prior studies on rats aged P21 and up (34) (**Figure 4B**). Additionally, the fraction of Iba-1 positive cells in the outer retina starting at P20 increased dramatically with age (**Figure 4C**). These results suggest that microglia numbers are already elevated in RCS rat retina before onset of retinal degeneration and prior to eye opening. Moreover, increase in outer retina microglia in RCS rats likely arises from migration of resident microglia rather than cell proliferation or infiltration.

### Suppression of Microglia by Tamoxifen Eye Drops Mitigates Photoreceptor Death

Given these unexpectedly early changes of microglia in RCS retina, we decided to continue our studies by using tamoxifen (tmx) to inactivate microglia starting in rat pups

treat breast cancer in human patients (42). In experimental animals, tmx is commonly used to control gene expression via engineered tmx-inducible promoters (43). Systemic application, via intraperitoneal injection or oral supplementation, and topical administration via creams and eye drops have been successfully used in published studies (35, 43, 44). Schlecht and colleagues showed that tmx administration as eye drops from P8 through P12 yields effects on outer retinal cells indicating that topical tmx penetrates eye tissues reaching effective concentration even in the posterior retina (35). Notably, such topical administration of tmx to the eye (45) does not affect retinal morphology even though tmx can be toxic to the retina (46). Even more recently, Wang and colleagues showed that tmx provided as dietary supplement reduced the extent of photoreceptor degeneration caused by light damage by reducing the number of activated microglia (47). The mechanism used by tmx to deplete activated microglia is still unknown, but it appears to be independent of its effects on estrogen receptors (47). To suppress microglia activity starting at preweaning age we applied tmx as eye drops in RCS rats three times a day from P10. At P33 we recorded scotopic ERGs

at P10. Tmx is an estrogen receptor agonist long used to

to assess photoreceptor functions, and at P40 we collected retina tissues to quantify levels of Iba-1 and CD68 proteins by immunoblotting. These experiments yielded a modest but significant reduction by about one third in both of these microglia markers in tmx-treated RCS rat eyes compared to control littermates that received eye drops with corn oil solvent only (**Figures 5A–C**). Moreover, tmx-treated rats at P33 showed significantly better light responses than control RCS rats at lower flash intensities eliciting rod activation only (−20dB to −8dB). At high flash intensities that excite both rods and cones there was no difference in retinal responses based on treatment (−4dB and 0dB) (**Figure 5D**). The b-wave indicative of activities of second order retinal neurons was also improved by tmx eye drop treatment (**Figure 5E**). RCS rats with ongoing retinal degeneration do not show normal retinal activity in ERG recordings with consistent increases in a- and b-wave amplitudes in response to light flashes of increasing intensity (23). Recording at P33 specifically in our study we observed flat or even moderately declining a- and b-wave amplitudes at highest flash intensities applied indicative of dysfunction of photoreceptors at this age. Nonetheless, tmx treatment modestly improved retinal function.

\**p* ≤ 0.05, \*\**p* ≤ 0.01; Student's *t*-test). (D,E) Graphs show average a-wave and b-wave amplitudes in response to light flashes of increasing intensity as indicated (mean ± s.e.m., n = 5 biological replicates, (\**p* ≤ 0.05; \*\*\**p* < 0.001, ANOVA). Open circles: RCS rats treated with oil eye drops (control); X: littermate RCS rats treated with tmx eye drops (treated).

#### Adding Tamoxifen Diet to Tamoxifen Eye Drops Treatment Further Suppresses Microglia but Is Toxic

To improve the retinal degeneration delay effects observed from tmx eye drops alone, we modified the treatment by adding tmx-supplemented diet starting at weaning at P19. This combined tmx eye drop/diet strategy reduced both Iba-1 and CD68 in treated RCS rat retina by ∼50 and 40%, respectively, compared to control siblings receiving neither topical nor systemic tmx (**Figures 6A–C**). As complement, we performed immunofluorescence microscopy analysis of retina tissue sections. We observed fewer Iba-1 positive microglia specifically in the outer nuclear layer of the retina of tmx treated RCS rats compared to control littermates (**Figures 6D–F**). Moreover, the retina of tmx-treated RCS rats showed one additional row of photoreceptor nuclei suggesting modestly improved retention of photoreceptor cells (**Figures 6D,E,G**). We also found fewer CD68 positive, activated microglia, in treated rats compared to control rats (**Figures 6H–J**). Scotopic ERGs at P33 showed an improvement of the photoreceptor response at all light flash intensities (**Figure 6K**). However, bwave amplitudes did not differ between tmx-treated and control animals (**Figure 6L**). Moreover, animals treated with tmx diet were smaller, more sensitive to the anesthetic, and overall less healthful than control siblings (data not shown). Taken together, we conclude from these observations that combining topical and systemic tmx treatments is more effective in inactivating retinal microglia than topical treatment alone. However, considerable systemic tmx toxicity renders such dual application unsuitable for further research let alone translation.

#### Tamoxifen Eye Drops Combined With Intravitreal and Systemic Injection of Liposomal Clodronate Reduces Photoreceptor Death by Suppressing Microglia Activation

As an alternate strategy to increase efficiency of microglia inhibition, we tested the addition of local and systemic administration of liposomal clodronate (LC) to our previous tmx eye drop treatment. LC is selectively taken up by phagocytic cells triggering their apoptotic pathway (48, 49). Intravitreal LC injection has been shown to successfully deplete retinal microglia in several species including rats (50–52). In a pilot

labeled with Iba-1 antibody to indicate microglia (green) and nuclei counterstain (red). The same regions of the retina were imaged to directly compare retina of control and treated RCS rats. Scale bars, 40µm. (F) Bar graph shows quantification of total retinal or photoreceptor layer microglia as indicated (mean ± s.e.m., *n* = 5 biological replicates; \**p* ≤ 0.05, Student's *t*-test). (G) Bar graph shows number of rows of photoreceptor nuclei in the outer nuclear layer (mean ± s.e.m., *n* = 5–7 biological replicates; \**p* ≤ 0.05, Student's *t*-test). (H,I) Representative micrographs of retina cross sections labeled with CD68 antibody to indicate activated microglia (green) and nuclei counterstain (red). The same regions of the retina were imaged to directly compare retina of control and treated RCS rats. Scale bars, 40µm. (J) Bar graph shows quantification of activated retinal microglia as indicated (mean ± s.e.m., *n* = 4 biological replicates; \*\**p* ≤ 0.005; Student's *t*-test). (K,L) Graphs *(Continued)* FIGURE 6 | show average a-wave and b-wave amplitudes in response to light flashes of increasing intensity as indicated (mean ± s.e.m., *n* = 4–9 biological replicates, \**p* ≤ 0.05; \*\**p* < 0.005, ANOVA). Open circles: RCS rats treated with oil eye drops and fed standard chow (control); X: littermate RCS rats treated with tmx eye drops and fed tmx chow (treated).

experiment, we confirmed that a one-time intravitreal injection one day after eye opening reduces the number of Iba-1 positive retinal microglia in RCS rats (**Supplementary Figure 2**). To increase the efficacy of our anti microglia treatment, we then combined inactivation with depletion of microglia by applying tmx eye drops starting at P10 as in our previous experiments and adding a one-time intravitreal LC injection the day after eye opening. To suppress replenishment of retinal microglia we additionally applied LC intraperitoneally every 7 days starting at P13. Indeed, **Figure 7** shows that this combination treatment was more effective than tmx eye drops alone or even tmx eye drops with the toxic tmx food. Immunoblotting showed reduction of retinal Iba-1 levels in treated animals by about 60% and of CD68 by about 50% compared to protein levels in sibling controls, indicating a greater reduction of microglia and their activation in RCS rats treated with the combined treatment as compared to rats treated with tmx alone (**Figures 7A–C**). Moreover, ∼ 30% fewer microglia localized to the photoreceptor layer of the retina (the outer nuclear layer) suggesting impaired microglia migration in response to tmx/LC treatment and the number of total Iba-1 positive retinal microglia was similarly reduced (**Figures 7D–F**). The combination treatment directly benefitted photoreceptor viability preserving more than two additional rows of photoreceptor nuclei in LC/tmx treated retina by P40 compared to control sibling retina (**Figures 7D,E,G**). Matching the immunoblotting results, immunofluorescence microscopy of CD68 showed a significant reduction of about 40% in activated microglia in retina from treated rats (**Figures 7H–J**). Finally, scotopic ERGs revealed significantly higher a-wave and b-wave amplitudes in treated RCS rats confirming improved retinal functionality (**Figures 7K,L**). Importantly, the combination treatment did not show obvious systemic toxicity. Altogether, these results suggest that a combination treatment of tmx eye drops and LC injections is more efficient in suppressing microglia than tmx eye drops alone. Our results imply that reducing microglia activity delays degeneration of photoreceptors in mutMerTK-RP.

#### DISCUSSION

In this study, we investigated the age of onset and effects of early treatment of inflammatory changes in animal models lacking the phagocytosis receptor MerTK. The MerTK-deficient mice and rats we studied exhibit early onset fast progressing retinal degeneration like human patients with mutMerTK-RP. We found pro-inflammatory cytokines, elevated levels of retinal microglia and microglia activation markers prior to eye opening. It has long been assumed (but never directly tested as far as we are aware) that eye opening marks the beginning of life-long diurnal outer segment renewal including MerTKdependent RPE phagocytosis. Indeed, our results confirm that there is only limited RPE phagocytosis of spent outer segment fragments before eye opening in WT rat retina whose RPE cells require MerTK for daily engulfment of spent photoreceptor outer segment tips. These results suggest that inflammatory activation prior to eye opening in MerTK-deficient retina is highly unlikely a mere consequence of the accumulation of outer segment debris in the subretinal space due to phagocytosis-defective RPE.

In both mouse and rat models of MerTK deficiency, we found expression of the chemoattractant cytokine CCL5 (and others) significantly and robustly elevated at P14 prior to eye opening and before photoreceptor debris buildup due to lack of MerTK-mediated RPE phagocytosis. The cells producing the cytokines remain to be specified. RPE in vivo may produce CCL5 and other cytokines in culture for instance in response to viral infection (53). Given that we studied dissected retina samples from which the RPE and posterior eyecup was removed, it is however likely that retinal cells not RPE or choroidal cells generate these transcripts. The robust and consistent level of CCL5 mRNA in our retina extracts renders contaminating RPE unlikely as major source.

Likely as response to elevated cytokines, we observed microglia relocalization to the outer retina as early as P20. In elegant experiments, Kohno et al. recently discriminated between retinal microglia and invading macrophages in mertk−/<sup>−</sup> mice by Cx3cr1 and Ccr2 promotor labeling, respectively. They found invading macrophages in mertk−/<sup>−</sup> retina only at ages of 6 weeks and above (32). This suggests that retinal microglia are the Iba1-positive cells population we observed in our MerTKdeficient RCS rats at ages up to P40. In agreement with previous studies that mostly focused on RCS rat retina in mid to late stage retinal degeneration at P33 and older (31, 34), we saw increased numbers of microglia in RCS rat retina as compared to WT retina at all ages tested including before eye opening. The total number of microglia did not increase, suggesting that the increased number of microglia observed in the outer retina from P21 was solely due to migration and not proliferation. In agreement, Di Pierdomenico et al. showed proliferation of retinal microglia in RCS retina at P45 but not at P33 (34).

Inflammation has been reported mostly at advanced stages of RP in retina that has already partly degenerated and in which retinal cells continue to die (31, 33, 54–56). However, precedent exists that microglia activation may occur early, preceding any retinal cell death, and contribute to degenerative processes (57). Microglia may have protective roles (24, 58) or promote photoreceptor cell death increasing the severity of retinal degeneration (29, 40, 59). To investigate the relevance of the early microglia activation for RCS rat retinal degeneration, we inhibited microglia using two experimental approaches, which we begun 4 days before eye opening. First, we inhibited microglia with tmx shown previously to inactivate retinal microglia in adult mice (47). Application of tmx as eye drops alone yielded

tissue harvest for immunoblotting and histology analyses at P40. (A) A representative immunoblot is shown comparing retina extracts from control and treated RCS rats. The same membrane was probed for Iba-1, CD68, and for tubulin as loading control. (B,C) Bar graphs show quantification of experiments as in (A) with levels of Iba-1 (B) and CD68 (C) in each sample set relative to levels of tubulin before comparing control and treated samples. Normalized Iba-1 and CD68 levels in control samples were set as 1 (mean ± SD, *n* = 4 biological replicates; \*\**p* ≤ 0.01, Student's *t*-test). (D,E) Representative micrographs of retina cross sections labeled with Iba-1 antibody to indicate microglia (green) and nuclei counterstain (red). The same regions of the retina were imaged to directly compare retina of control and treated RCS rats. Scale bars, 40µm. (F) Bar graph shows quantification of total retinal or photoreceptor layer microglia as indicated (mean ± s.e.m., n = 5 biological replicates; \**p* ≤ 0.05; \*\*\**p* ≤ 0.001, Student's *t*-test). (G) Bar graph shows number of rows of photoreceptor nuclei in the outer nuclear layer (mean ± s.e.m., *n* = 5–7 biological replicates; \*\*\**p* ≤ 0.001, Student's *t*-test). (H,I) Representative micrographs of retina cross sections labeled with CD68 antibody to indicate activated microglia (green) and nuclei counterstain (red). The same regions of the retina were imaged to directly compare retina of control and treated RCS rats. Scale bars, 40µm. (J) Bar graph shows quantification of activated retinal microglia as indicated (mean ± s.e.m., *n* = 4 biological replicates; \*\**p* ≤ 0.005; Student's *t*-test). *(Continued)* FIGURE 7 | (K,L) Graphs show average a-wave and b-wave amplitudes in response to light flashes of increasing intensity as indicated (mean ± s.e.m., *n* = 4–9 biological replicates, \*\*\**p* ≤ 0.001, ANOVA). Open circles, RCS rats treated with oil eye drops (control); X, littermate RCS rats treated with tmx eye drops and LC (treated).

a small but significant reduction in microglia activation and marginally better photoreceptor light responses at P33. To enhance efficacy and to prevent possible systemic replenishment of microglia to tmx eye drop-treated retina, we next added feeding animals tmx-supplemented diet to the tmx eye drop treatment. While this dual treatment showed improved efficacy in the retina we observed significant toxicity of the tmx diet resulting in reduced body weight and viability. Therefore, we instead complemented the tmx eye drop administration with a well-established alternative approach depleting microglia with liposomal clodronate, which poisons microglia. Combined local treatment of tmx eye drops and one-time intravitreal injection of LC with weekly systemic LC injections further increased microglia depletion compared to eye drops alone with no apparent toxicity. Taken together, these experiments demonstrate that early microglia inhibition is effective in delaying retinal degeneration due to MerTK deficiency. Moreover, the extent of microglia inhibition directly correlates with improved survival of photoreceptors and preservation of photoreceptors function. Finally, our results indicate that microglia driven processes aggravate retinal degeneration due to MerTK deficiency rather than playing a protective role.

More experiments are needed to unravel how microglia aggravate retinal degeneration due to deficiency in the engulfment receptor MerTK. In brain, MerTK deficiency impairs efferocytosis activity of microglia and macrophages (60, 61). In the retina of the rd10 mouse model of RP, microglia phagocytose living photoreceptors increasing the rate of photoreceptor loss (59). It will be important to test microglia phagocytosis in RCS retina in the future. It is important to note however that the genetic defect in the rd10 RP model is in pde6b, which encodes a photoreceptor-specific protein involved in phototransduction. In contrast, MerTK is a phagocytosis receptor expressed not by photoreceptors but by the adjacent RPE, which relies on it for daily clearance of spent outer segment tips as part of outer segment renewal and possibly by other phagocytic cells in the retina. We did not find reports showing MerTK protein expression in mouse or rat retinal microglia in situ. With MerTK antibodies recognizing rat MerTK in tissue sections unavailable, we tested MerTK localization in WT mouse retina. **Supplementary Figure 3** shows abundant labeling by a well-established anti-mouse MerTK antibody of WT mouse RPE but little labeling in neural retina and no obvious overlap with Iba-1 microglia marker. Thus, MerTK protein is unlikely to be abundant in resting retinal microglia, and its expression may be upregulated in retinal disease such as rd10 RP to enhance microglia phagocytosis of distressed photoreceptors. If the aggravating role of retinal microglia in MerTK-deficient RP also involves photoreceptor phagocytosis it will indicate that retinal microglia perform clearance phagocytosis using a MerTK-independent pathway with an alternative engulfment receptor. Indeed, numerous engulfment receptors relevant to efferocytosis are known (62). Moreover, increased staining of complement factor C1q has been reported in RCS retina in the OPL where microglia may phagocytose synaptic components (63). The same study found that elimination of retinal microglia by systemic treatment with a CSF1 receptor inhibitor had no effect on photoreceptor light responses (and worsened inner retinal responses), although drug administration via dietary supplementation only started at P15 and may not have been effective immediately in pre-weaning rats. As dead cells do not permanently accumulate in advanced stages of retinal degeneration in RCS retina clearance phagocytosis is likely to take place, although the specific phagocytic cells responsible and whether and when they enter the retina is unknown. Finally, microglia in MerTK-deficient retina may impair photoreceptor function and viability through activities that do not require them to be phagocytic. Distinguishing between these intriguing possibilities will be the subject of further investigation.

As of now, there is no cure for any form of RP. However, translational approaches including clinical trials are in development or underway (64, 65). Yet, treatments in development such as gene therapy or RPE replacement will be successful only if there is significant retention of retinal cells and architecture whose functionality and longevity can be rescued by treatment. Delaying the exceptionally early onset and rapid progression to irreversible retinal cell death that is characteristic for mutMerTK-RP will increase the window of opportunity during which advanced treatments such as gene therapy will be effective. As our results show a role for early inflammation before the phagocytic defect of the RPE harms photoreceptors, it is tempting to speculate that other forms of RP may benefit from microglia inactivation at very early stages of disease prior to overt retinal dysfunction and degeneration.

Altogether, our results show that lack of MerTK activates inflammatory mechanisms causing retinal microglia to harm photoreceptors prior to onset of fully active daily outer segment renewal. These processes are thus independent of accumulation of subretinal outer segment debris that results from engulfment failure of MerTK-deficient RPE. In addition to orchestrating debris engulfment, MerTK signaling is complex and has been shown to contribute to cell differentiation and homeostasis (66). Our results let us hypothesize that cells lacking MerTK in the mutMerTK retina elicit pro-inflammatory signaling which in turn activates microglia. Additional studies will be needed to determine whether the triggering cell type is the RPE, which may utilize MerTK signaling during normal development and differentiation. Distressed MerTK-deficient RPE cells may communicate with cells in the retina via additional cytokines or other signaling mediators causing these secondary cells to upregulate the cytokines we found elevated. Identifying the primary and secondary cell types and stimulus mechanism for such inflammatory pathways will be a subject of future studies and may identify cellular molecular targets for therapies aiming to delay onset of the aggressive forms of retinitis pigmentosa associated with defective MerTK.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

#### ETHICS STATEMENT

All procedures involving animals were performed according to the ARVO Guide for Use of Animals in Vision Research and the Guide for the Care and Use of Laboratory Animals (NIH, 8th edition) and approved by the Institutional Animal Care and Use Committee of Fordham University.

### AUTHOR CONTRIBUTIONS

This study was designed by DL, FM, and SF. DL, FM, and SF performed experiments, analyzed, and interpreted

#### REFERENCES


results. DL and FM wrote the manuscript with input from SF. All authors read, commented, and approved of the final manuscript.

#### FUNDING

This work was funded by NIH R01-EY026215 from the National Eye Institute and by a Grant-In-Aid of Research from Sigma Xi, The Scientific Research Society. SF holds the Kim B. and Stephen E. Bepler Chair in Biology.

#### ACKNOWLEDGMENTS

We thank Samantha Forbes and John Macias 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. 2020.01463/full#supplementary-material


**Conflict of Interest:** 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 © 2020 Lew, Mazzoni and Finnemann. 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.