# TRANSLATIONAL INSIGHTS INTO MECHANISMS AND THERAPY OF ORGAN DYSFUNCTION IN SEPSIS AND TRAUMA, 2nd Edition

EDITED BY : Lukas Martin, Christoph Thiemermann, Pietro Ghezzi, Peter Radermacher and Timothy Robert Billiar 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 978-2-8325-5947-5 DOI 10.3389/978-2-8325-5947-5

#### 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

# TRANSLATIONAL INSIGHTS INTO MECHANISMS AND THERAPY OF ORGAN DYSFUNCTION IN SEPSIS AND TRAUMA, 2nd Edition

Topic Editors:

Lukas Martin, University Hospital RWTH Aachen, Germany Christoph Thiemermann, Queen Mary University of London, United Kingdom Pietro Ghezzi, Brighton and Sussex Medical School, United Kingdom Peter Radermacher, University of Ulm, Germany Timothy Robert Billiar, University of Pittsburgh, United States

Publisher's note: This is a 2nd edition due to article retractions.

Citation: Martin, L., Thiemermann, C., Ghezzi, P., Radermacher, P., Billiar, T. R., eds. (2025). Translational Insights into Mechanisms and Therapy of Organ Dysfunction in Sepsis and Trauma, 2nd Edition. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-8325-5947-5

# Table of Contents

*06 Editorial: Translational Insights Into Mechanisms and Therapy of Organ Dysfunction in Sepsis and Trauma*

Peter Radermacher, Timothy R. Billiar, Pietro Ghezzi, Lukas Martin and Christoph Thiemermann


Yanwei Cheng, Tony N. Marion, Xue Cao, Wanting Wang and Yu Cao

*22 Linagliptin Attenuates the Cardiac Dysfunction Associated With Experimental Sepsis in Mice With Pre-existing Type 2 Diabetes by Inhibiting NF-*k*B*

Sura Al Zoubi, Jianmin Chen, Catherine Murphy, Lukas Martin, Fausto Chiazza, Debora Collotta, Muhammad M. Yaqoob, Massimo Collino and Christoph Thiemermann

*35 Effects of Changes in the Levels of Damage-Associated Molecular Patterns Following Continuous Veno–Venous Hemofiltration Therapy on Outcomes in Acute Kidney Injury Patients With Sepsis*

Jie Wu, Jianan Ren, Qinjie Liu, Qiongyuan Hu, Xiuwen Wu, Gefei Wang, Zhiwu Hong, Huajian Ren and Jieshou Li

*50 Forebrain Cholinergic Signaling Regulates Innate Immune Responses and Inflammation*

Kurt R. Lehner, Harold A. Silverman, Meghan E. Addorisio, Ashbeel Roy, Mohammed A. Al-Onaizi, Yaakov Levine, Peder S. Olofsson, Sangeeta S. Chavan, Robert Gros, Neil M. Nathanson, Yousef Al-Abed, Christine N. Metz, Vania F. Prado, Marco A. M. Prado, Kevin J. Tracey and Valentin A. Pavlov

*61 TLR3 Ligand PolyI:C Prevents Acute Pancreatitis Through the Interferon-*b*/Interferon-*a*/*b *Receptor Signaling Pathway in a Caerulein-Induced Pancreatitis Mouse Model*

Chaohao Huang, Shengchuan Chen, Tan Zhang, Dapei Li, Zhonglin Huang, Jian Huang, Yanghua Qin, Bicheng Chen, Genhong Cheng, Feng Ma and Mengtao Zhou


Tomasz Skirecki, Susanne Drechsler, Grazyna Hoser, Mohammad Jafarmadar, Katarzyna Siennicka, Zygmunt Pojda, Jerzy Kawiak and Marcin F. Osuchowski

*101 Functional Annotation of Genetic Loci Associated With Sepsis Prioritizes Immune and Endothelial Cell Pathways*

Kieu T. T. Le, Vasiliki Matzaraki, Mihai G. Netea, Cisca Wijmenga, Jill Moser and Vinod Kumar


Kristen N. Morrow, Craig M. Coopersmith and Mandy L. Ford

*133 Endogenous Uteroglobin as Intrinsic Anti-inflammatory Signal Modulates Monocyte and Macrophage Subsets Distribution Upon Sepsis Induced Lung Injury*

Andrea Janicova, Nils Becker, Baolin Xu, Sebastian Wutzler, Jan Tilmann Vollrath, Frank Hildebrand, Sabrina Ehnert, Ingo Marzi, Philipp Störmann and Borna Relja

*147 Leukocyte-Released Mediators in Response to Both Bacterial and Fungal Infections Trigger IFN Pathways, Independent of IL-1 and TNF-*a*, in Endothelial Cells*

Kieu T. T. Le, Xiaojing Chu, Martin Jaeger, Josée A. Plantinga, Vasiliki Matzaraki, Sebo Withoff, Leo A. B. Joosten, Mihai G. Netea, Cisca Wijmenga, Yang Li, Jill Moser and Vinod Kumar


Ravikanth Velagapudi, Saraswathi Subramaniyan, Chao Xiong, Fiona Porkka, Ramona M. Rodriguiz, William C. Wetsel and Niccolò Terrando

*188 Diurnal Variation in Systemic Acute Inflammation and Clinical Outcomes Following Severe Blunt Trauma*

Akram M. Zaaqoq, Rami A. Namas, Othman Abdul-Malak, Khalid Almahmoud, Derek Barclay, Jinling Yin, Ruben Zamora, Matthew R. Rosengart, Timothy R. Billiar and Yoram Vodovotz


Martin Wepler, Jonathan M. Preuss, Tamara Merz, Clair Hartmann, Ulrich Wachter, Oscar McCook, Josef Vogt, Sandra Kress, Michael Gröger, Marina Fink, Angelika Scheuerle, Peter Möller, Enrico Calzia, Ute Burret, Peter Radermacher, Jan P. Tuckermann and Sabine Vettorazzi

*235 Short-Chain Alcohols Upregulate GILZ Gene Expression and Attenuate LPS-Induced Septic Immune Response*

Hang Pong Ng, Scott Jennings, Steve Nelson and Guoshun Wang


Sen Yang, Ge Kuang, Liangke Zhang, Shengwang Wu, Zizuo Zhao, Bin Wang, Xinru Yin, Xia Gong and Jingyuan Wan

*270 Mesenchymal Stem Cell-Derived Extracellular Vesicles: A Novel Cell-Free Therapy for Sepsis*

Yanwei Cheng, Xue Cao and Lijie Qin

*278 Non-invasive Assessment of Mitochondrial Oxygen Metabolism in the Critically Ill Patient Using the Protoporphyrin IX-Triplet State Lifetime Technique—A Feasibility Study*

Charles Neu, Philipp Baumbach, Alina K. Plooij, Kornel Skitek, Juliane Götze, Christian von Loeffelholz, Christiane Schmidt-Winter and Sina M. Coldewey


Zhiyong Peng

# Editorial: Translational Insights Into Mechanisms and Therapy of Organ Dysfunction in Sepsis and Trauma

Peter Radermacher <sup>1</sup> \*, Timothy R. Billiar 2,3, Pietro Ghezzi <sup>4</sup> , Lukas Martin<sup>5</sup> and Christoph Thiemermann<sup>6</sup>

1 Institute for Anesthesiological Pathophysiology and Process Engineering, Ulm University Hospital, Ulm, Germany, <sup>2</sup> Department of Surgery, University of Pittsburgh, Pittsburgh, PA, United States, <sup>3</sup> UPMC International and Commercial Services Division, Pittsburgh, PA, United States, <sup>4</sup> Department of Clinical Medicine, Brighton and Sussex Medical School, Brighton, United Kingdom, <sup>5</sup> Department of Intensive and Intermediate Care, Medical Faculty, University Hospital RWTH, Aachen, Germany, <sup>6</sup> Centre for Translational Medicine and Therapeutics, William Harvey Research Institute, Queen Mary University of London, London, United Kingdom

Keywords: sepsis, trauma, multiple organ failure, SIRS, animal models, translational studies, clinical trials

**Editorial on the Research Topic**

#### Edited by:

Rudolf Lucas, Augusta University, United States

#### Reviewed by:

Cassiano Felippe GonçAlves-De-Albuquerque, Rio de Janeiro State Federal University, Brazil

\*Correspondence: Peter Radermacher peter.radermacher@uni-ulm.de

#### Specialty section:

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

Received: 26 June 2020 Accepted: 23 July 2020 Published: 25 September 2020

#### Citation:

Radermacher P, Billiar TR, Ghezzi P, Martin L and Thiemermann C (2020) Editorial: Translational Insights Into Mechanisms and Therapy of Organ Dysfunction in Sepsis and Trauma. Front. Immunol. 11:1987. doi: 10.3389/fimmu.2020.01987

#### **Translational Insights Into Mechanisms and Therapy of Organ Dysfunction in Sepsis and Trauma**

Multiple organ dysfunction or even failure after sepsis or trauma is due to a dysregulated host response. Currently, besides (surgical) source control (e.g., control of bleeding or drainage of abscesses) and administration of antimicrobial drugs, therapeutic approaches are limited to supportive care. Advances in our understanding of the key pathophysiological pathways involved in the excessive inflammation triggered by trauma, sepsis and/or ischemia-reperfusion have had limited impact. The 28 article in this Research Topic focus on the molecular mechanisms behind (hyper) inflammation after sepsis or trauma, with special emphasis on preclinical and translational studies that target potential organ-protective and/or -resuscitative therapeutic strategies. Most studies report rodent models of trauma and elective surgery (three articles), non-microbial hyper-inflammation induced with endotoxin exposure (LPS; seven articles) and chemical pancreatitis (one article), and cecal ligation and puncture-induced sepsis (six articles). Additional papers summarize investigations of human material (six articles) or fully-resuscitated large animal models (two articles). These article are complimented by four reviews and a commentary.

#### RODENT MODELS OF TRAUMA AND (ELECTIVE) SURGERY-RELATED TISSUE INJURY

Velagapudi et al. investigate the pathophysiology of post-traumatic/surgical delirium. The authors investigated the effect of murine limb trauma on post-operative behavior (as assessed using the 5- "choice serial reaction time task" and motor activity) and neuroinflammation and blood-brain barrier integrity (by using light-sheet microscopy). Post-surgery behavior showed impairment, in parallel with reduced microglial ramification and overall cell volume impaired hippocampal astrocytic-tight junctions. Wall et al. characterized the mechanisms underlying post-traumatic cardiac dysfunction in mice undergoing soft tissue trauma and bone fracture

**6**

followed by hemorrhage. Despite resuscitation, stroke volume remained at 25% of baseline values, which was associated with leukocyte infiltration and ultrastructural sarcomere and mitochondriadisorganization, leading to increased serum levels of heart fatty acid-binding protein and troponin I. The authors concluded that mitochondria-driven apoptosis may be a target to prevent irreversible post-traumatic cardiac injury. In a rat model of abdominal surgery, Bangash et al. further clarified the anti-inflammatory properties of the β2-agonist dopexamine. Ileal intravital microscopy demonstrated that attenuated leukocyteendothelial adhesion in post-capillary venules while neither arteriolar diameter, functional capillary density nor systemic hemodynamics nor lactic acidosis were affected compared to control animals. Interestingly, despite comparable effects on the intestinal microcirculation, high dose dopexamine only, in contrast to the pure β2-agonist salbutamol, also attenuated surgery-related increase in serum creatinine.

#### NON-MICROBIAL SYSTEMIC HYPERINFLAMMATION

Although clearly distinct from polymicrobial sepsis, endotoxin (LPS) challenge and chemically-induced pancreatitis are often used to model hyperinflammation, and indeed, sepsis due to their reproducibility and the subsequent organ failure that mimics that of sepsis. Two studies addressed the potential role of the glucocorticoid receptor (GR) in LPS-induced organ dysfunction. Wepler et al. compared wildtype (GR+/+) mice with animals with impaired GR dimerization (GRdim/dim) under conditions of full intensive care support ("lung-protective" mechanical ventilation, crystalloids, and norepinephrine). GRdim/dim mice presented with more severe shock and aggravated acute lung injury, which coincide with increased tissue osteopontin expression. Ng et al. provided evidence for the anti-inflammatory mechanisms of acute, binge-drinking-like alcohol intoxication: alcohol exposure attenuated the inflammatory response to LPS in vivo, ultimately improving survival, which was associated with increased gene expression of the Glucocorticoid-Induced Leucine Zipper (GILZ), a key molecule acting via non-canonical GR activation. Inhibition of ethanol metabolism enhanced this effect, and interestingly, the higher molecular-weight shortchain alcohols propanol and isopropanol were even more potent than ethanol.

Dayang et al. provided new evidence relating to the complex organ- and organ- and microvascular bed-specific LPS-induced expression of the endothelial adhesion molecules E-selectin and VCAM-1. The authors showed in LPS-challenged mice that the renal microvascular endothelium expressed various Eselectin and VCAM-1 subpopulations, i.e., E-selectin−/VCAM-1 <sup>−</sup>, E-selectin+/VCAM-1−, E-selectin+/VCAM-1+, and Eselectin−/VCAM-1+. The formation of subpopulations was a common response of endothelial cells to LPS challenge. FACS analysis demonstrated that the <sup>+</sup>/<sup>+</sup> subpopulation expressed the highest cytokine and chemokine response, which was mainly TLR4-mediated. Activation of NF-κB and p38 MAPK were key signaling events in the formation of this <sup>+</sup>/+subpopulation. The translational value of these murine data was highlighted by the recapitulation of these effects in LPS-exposed HUVEC and human lung microvascular endothelial cells.

The role of the forebrain cholinergic system for the systemic immune response was addressed by Lehner et al. The centralacting cholinergic agonist galanthine suppressed the systemic TNFα response to endotoxemia in mice, and this effect was suppressed by both local genetic ablation of acetylcholine release and vagotomy. In contrast, local and selective activation of the M1 muscarinic acetylcholine receptor (M1 mAChR) with the allosteric agonist benzyl quinolone carboxylic acid also suppressed the serum TNFα levels, thereby ultimately improving survival; these beneficial effects were again suppressed in M1 mAChR-ko mice. Overall, these data suggest a bidirectional relationship between brain cholinergic signaling and the systemic inflammatory response.

Two studies focussed on possible therapeutic interventions. Deng et al. investigated the potential of targeting citrullinated histone H3 (CitH3) to neutralize neutrophil extracellular traps. When administered simultaneously with a lethal dose of LPS, anti-CitH3 monoclonal antibodies prevented HUVEC damage in vitro and attenuated ALI in vivo via inhibition of the inflammatory response, thereby ultimately reducing mortality. The therapeutic potential of the anti-inflammatory xanthone glucoside mangiferin was studied by Yang et al. in mice with LPS/galactosamine-induced acute liver failure. Pre-treatment with magiferin inhibited hepatic TNF-α production, decreased serum aminotransferase activities, and improved survival. This beneficial effect was abolished by Kupffer cell deletion, and at least in part, related to increased heme oxygenase-1 expression.

Models of caerulein-induced acute pancreatitis in vivo and taurocholate-induced pancreatic acinar cell line in vitro were used to study the roll of the Toll-like receptor 3 (TLR3) ligand polyI:C by Huang et al. PolyI:C is a double-stranded RNA mimic that can act as an immune stimulant by triggering type I interferon (IFN) production and downstream IFNα/β receptor (IFNAR)-dependent signaling, and pre-treatment inhibited chemotaxis and ROS production. This was abolished in IFN-β- and IFNAR-ko mice.

#### CECAL LIGATION AND PUNCTURE (CLP)-INDUCED SEPSIS

CLP is one of the most frequently used models of polymicrobial sepsis. While three studies in this special issue were designed to further character molecular pathways of sepsis pathophysiology, others evaluated the efficacy of specific therapeutic interventions. Most of the studies have the merit of integrating the "Minimum Quality Threshold in Pre-Clinical Sepsis Studies" (MQTiPSS) guidelines, thus allowing for easier inter-laboratory comparability.

Li et al. addressed the role of the switch of macrophage phenotype from the pro-inflammatory M1 state to the M2 repair (i.e., resolution) state in septic AKI. Coculture of human M1 macrophages with proximal tubular (HK-2) cells revealed that Colony Stimulating Factor 2 (Csf2) was the most up-regulated protein, and post-CLP injection of a Csf2 antibody attenuated kidney M1-M2 macrophage transition via p-STAT5 signaling, suppressed tubular proliferation, and increased mortality.

The role of AMP-activated protein kinase (AMPK) was investigated in murine CLP-induced sepsis by Kikuchi et al. Mice with genetic deletion of the catalytic α1 isoform (AMPKα1) had higher systemic cytokine levels, liver and lung inflammation and injury, and in particular, more severe hepatic mitochondrial damage and mortality. These deleterious effects were even more pronounced in male animals, suggesting that AMPK-dependent liver metabolic (dys) function is sex-dependent.

Skiretzki et al. further characterized the translational value of an established, "humanized" murine model of CLP-induced sepsis. Animals were stratified as "predicted-to-die (P-DIE)" or "predicted-to-survive (P-SUR)," and blood, bone marrow and spleen were collected for cytokine and chemokines as well as CD-80 and HLA-DR expression. TNFα, IL-6, IL-10, IL-8/KC-, and MCP-1-levels were several-fold higher in the P-DIE group. The study nicely showed that humanized mice reflect human immune responses.

A possible therapeutic potential of the Bruton's tyrosine kinase (BTK) inhibitors ibrutinib and acalabrutinib for septic cardiomyopathy and AKI was investigated by O'Riordan et al. When administered early (1 h) after the CLP-procedure, the BTK inhibitors attenuated sepsis-induced systolic dysfunction and reduced systemic and tissue NF-κB and NLRP3 inflammasome activation. This is clinically relevant as both BTK inhibitors are already approved for the treatment of chronic lymphatic leukemia (CLL), and acalabrutinib has recently been reported to attenuate systemic inflammation in a small cohort of patients with COVID-19 (Roschewski et al.). However, despite promising data on ibrutinib during lipoteichoic acid-induced pulmonary inflammation as well as and ceftriaxone-treated pneumococcal pneumonia (1), a word of caution needs to be mentioned concerning tyrosine kinase inhibitors in general: the RESONATE-17 study on 144 CLL patients treated with ibrutinib reported a 30% incidence of severe infection including pneumonia (2), and in un-resuscitated CLP-induced murine sepsis, the tyrosine kinase inhibitor dasatinib (administered as a pre- and post-treatment) dose-dependently showed a "friend or foe" profile: while low doses attenuated sepsis-related organ dysfunction, and ultimately improved survival, high doses had the opposite effect (3).

Another approved drug, the oral antidiabetic linagliptin, was compared with the selective IkB kinase (IKK) inhibitor (IKK-16), with respect to its effects on septic cardiomyopathy as well as kidney and liver injury in murine high fat diet-induced type 2 diabetes mellitus (T2DM) (Al Zoubi et al.). This study is of particular importance, since it sheds light on one of the most frequently occurring chronic co-morbidities observed in patients with sepsis and/or after trauma, which is well-known to worsen the patient's outcome. In good agreement with these clinical observations, T2DM aggravated sepsis-related (organ) dysfunction associated with sepsis, which was comparably mitigated by both tested drugs. This effect of linagliptin was associated with reductions of NF-κB activation, iNOS expression and serum levels of pro-inflammatory cytokines.

Finally, Janicova et al. characterized the role of uteroglobin in a murine double-hit model of blunt chest trauma and subsequent (24 h) CLP-induced sepsis focussing on dynamic changes in monocyte and macrophage subsets. This doublehit model has the possibility of replicating the frequently occurring similar sequence in critically ill patients, i.e., an initial polytrauma is complicated by consecutive sepsis, and furthermore, allows investigating common "final routes" of these two pathological entities. Both pro-inflammatory monocytes and macrophages significantly increased in blood, lungs and bronchoalveolar lavage fluid (BALF) at the expense of tissue repairing macrophages, which coincided with higher levels of TNF-α, MCP-1, and RAGE. Uteroglobin neutralization further aggravated this pro-inflammatory response, and subsequently, lung damage, thereby highlighting intrinsic anti-inflammatory signaling role of this protein.

#### LARGE ANIMAL AND HUMAN STUDIES

Despite the substantial demand for personnel, infrastructure and consumables as well as the often lacking test kits for the species involved, large animal (in particular porcine) models have the advantages (i) to allow for integration of standard ICU procedures into the experimental design, and (ii) to study an immunological environment that is closer to the human situation than any other species (except for non-human primates). Horst et al. made use of these facts for the characterization of time-dependent dynamics of the inflammatory response in the blood, the femur fracture hematoma, and muscle samples of polytraumatized ("PT": lung contusion, liver laceration, femur fracture, and controlled hemorrhage) swine in comparison to animals with femur fracture alone (monotrauma). The severity of the polytrauma substantially inhibited the local generation of proinflammatory and angiogenetic mediators, suggesting that these systemic trauma-related changes of the local immunologic milieu might explain the delayed bone repair in PT patients. The same porcine PT model was used by Lackner et al. to address the role of midkine levels for post-traumatic cardiac dysfunction. The porcine data were complemented by (i) the measurement of midkine blood levels in PT patients (ISS ≥ 16), and (ii) the functional assessment of human cardiomyocytes incubated with midkine. In PT patients, midkine levels were several-fold higher than in healthy volunteers, while in vitro midkine exposure of cardiomyocytes altered cell Ca2<sup>+</sup> handling and reduced maximum mitochondrial O<sup>2</sup> consumption.

Sophisticated genome/transcriptome analyses were used by Le et al. in two studies (Le, Matzaraki et al.; Le, Chu et al.) that aimed at further characterizing the major determinants of leukocyte and/or endothelium responses to bacterial sepsis. In the first study (Le, Matzaraki et al.) expression quantitative trait loci (eQTL) data from the largest whole-blood eQTL database, cytokine QTLs from pathogen-stimulated peripheral blood mononuclear cells (PBMC), blood transcriptome data from pneumoniae-derived sepsis patients, and transcriptome data from pathogen-stimulated PBMC were integrated and found increased adherence-junction gene expression. The authors concluded that their approach provided evidence for genetically determined variability in both leucocyte and endothelial responses, which may contribute to explaining sepsis heterogeneity. In their second study (Le, Chu et al.), the authors' group investigated transcriptomic responses of human leukocytes and endothelial cells to Gram negative-bacteria, Gram positive-bacteria, and fungi. While whole pathogen lysates strongly activated leukocytes but not endothelial cells, the mutual leukocyte response to pathogens resulted in endothelial activation. Exposure of endothelial cells to leukocyte mediators activated endothelial cells at both transcription and protein levels, and revealed IL-1, TNF-α, and IFN as important drivers of endothelial activation.

Another study was dedicated to investigate the potential role of circadian rhythm and/or diurnal variations on the systemic inflammatory response, and ultimately outcome in patients with trauma (Zaaqoq et al.). From a total of nearly 500 blunt trauma survivors (ISS > 20) a subgroup of patients injured during daytime ("mDay") was compared to patients injured at nighttime ("mNight") in order to investigate the impact on 32 inflammatory mediators using Dynamic Network Analysis (DyNA) and Dynamic Bayesian Network (DyBN) inference. Both hospital and ICU length of stay were longer in the mNight patients, which coincided with higher IL-17A but lower MIP-1α, IL-7, IL-15, GM-CSF, and sST2 levels; DyBN yielded cortisol and sST2 as the major upstream determinants upstream of TGF-β1, chemokines, and Th17/protective mediators in both groups, with IL-6 being an additional downstream node in the mNight group only.

In patients with septic AKI undergoing continuous veno– venous hemofiltration (CVVH), Wu et al. investigated the impact of damage-associated molecular pattern (DAMP) (mitochondrial DNA, mtDNA; nuclear DNA, nDNA; heat shock protein 70, HSP70; high mobility group box 1, HMGB1) removal on mortality. Both HSP70 and HMGB1 clearance rates were good predictors of mortality, and high HSP70 clearance coincided with the level of HLA-DR expression. Based on these findings the authors cautioned the use of CVVH in AKI patients with sepsis merely for the removal of inflammatory mediators in the absence of any other indication for this therapeutic measure.

Finally, Neu et al. present a study on the feasibility of a direct, non-invasive, bedside measurement of skin mitochondrial oxygen tension (mitoPO2) with a new commercially available device that makes use of the "in vivo protoporphyrin IX-triplet state lifetime technique (PpIX-TSLT)." In 40 patients a three-step measurement was performed after enriching the clavipectoral triangle with 5 aminolevulinic acid, which comprised assessment of baseline values, evaluation of the effect of a local pressure to transitorily stop microcirculatory blood flow, and a second control value. The recorded data allowed calculating average and maximum mitoVO<sup>2</sup> at these three time points. The technique was reproducible, easy, and safe; potential tissue edema as assessed by bioimpedance resulted in lower mitochondrial oxygen tension.

# REVIEWS AND COMMENTARIES

Two articles reviewed potential new pathways (and consecutive therapeutic targets) of sepsis-related immune paralysis. Cheng et al. discuss the possible role of Parkinson disease protein 7 (Park 7), a well-established regulator of reactive oxygen species (ROS) release through interaction with p47phox, a subunit of NADPH oxidase. Among their various functions, ROS initiate TLR to activate macrophages. Consequently, Park 7 may be a novel therapeutic target to reverse sepsis-related immunosuppression. Morrow et al. discuss septic immune paralysis in the light of the markedly decreased T cell formation of pro-inflammatory cytokines. Since targeting lymphocyte survival did not find its way into clinical practice, cytokines with a more global immune effect may represent alternative therapeutic targets. The authors discuss the possible impact of the interleukin (IL)-17, IL-27, and IL-33 based on data from patient serum and murine models of peritonitis and pneumonia.

Denning et al. review the complex interplay of host response to pathogens via interaction of pathogen-associated molecular patterns (PAMPs) and pattern recognition receptors (PRRs) and damage-associated molecular patterns (DAMPs), and Neutrophil extracellular traps (NETs). Examples of DAMPs are extracellular cold-inducible RNA-binding protein (eCIRP), high mobility group box 1 (HMGB1), histones, and adenosine triphosphate (ATP). DAMPs are released by inflammasome activation or passively from dead cells. NETs are mixtures of extracellular DNA with histones, myeloperoxidase, and elastase, all released during inflammation. Although NETs clearly contribute to pathogen clearance, their excessive formation causes tissue damage.

Based on their multiple and ubiquitous properties, Cheng et al. discuss the potential of mesenchymal stem cells (MSCs) in the treatment of sepsis. Highlighting the possible undesired side effects, the authors also emphasize the potential of extracellular vesicles (EVs) derived from MSCs (MSC-EVs). These MSC-EVs appear to exert a therapeutic benefit similar to MSCs in protecting against sepsis-induced organ dysfunction by delivering RNAs and proteins to target cells, while at the same time being devoid of major MSC side effects. In addition, MSC-EVs provide some practical advantages over their parent MSCs.

Finally, Kesselmeier and Scherag comment on the recently popularized "adaptive clinical trials" discussing a recent article in this journal by Talisa et al. (4). The authors clearly acknowledge that adaptive trials allow modifying design elements during trial conduct, thereby possibly reducing resources, duration and sample size while simultaneously enhancing the chance of proof for an effective treatment. Nevertheless, the authors discuss the four issues response adaptive randomization (RAR), adaptive enrichment, seamless, and platform designs, which need to be taken into account.

In summary, the Research Topic entitled provides a broad and detailed overview of new molecular and mechanistic avenues for the management of sepsis- and trauma-related multiple organ dysfunction, and represents a major contribution to translational research in the field.

# 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 the Deutsche Forschungsgemeinschaft (DFG): CRC 1149, MA 7082/3-1.

4. Talisa VB, Yende S, Seymour CW, Angus DC. Arguing for adaptive clinical trials in sepsis. Front Immunol. (2018) 9:1502.doi: 10.3389/fimmu.2018.01502

**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 Radermacher, Billiar, Ghezzi, Martin and Thiemermann. 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.

# Commentary: Arguing for Adaptive Clinical Trials in Sepsis

Miriam Kesselmeier <sup>1</sup> and André Scherag1,2 \*

<sup>1</sup> Research Group Clinical Epidemiology, Center for Sepsis Control and Care, Jena University Hospital, Jena, Germany, 2 Institute of Medical Statistics, Computer and Data Sciences, Jena University Hospital, Jena, Germany

Keywords: sepsis, adaptive clinical trials, Bayesian statistics, platform trials, response adaptive randomization

#### **A Commentary on**

#### **Arguing for Adaptive Clinical Trials in Sepsis**

by Talisa VB, Yende S, Seymour CW, Angus DC. Front. Immunol. (2018) 9:1502. doi: 10.3389/fimmu.2018.01502

#### Edited by:

Peter Radermacher, Universität Ulm, Germany

#### Reviewed by:

Benjamin Mayer, Universität Ulm, Germany James Russell, University of British Columbia, Canada Derek C. Angus, University of Pittsburgh, United States

> \*Correspondence: André Scherag andre.scherag@med.uni-jena.de

#### Specialty section:

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

Received: 10 September 2018 Accepted: 10 October 2018 Published: 30 October 2018

#### Citation:

Kesselmeier M and Scherag A (2018) Commentary: Arguing for Adaptive Clinical Trials in Sepsis. Front. Immunol. 9:2507. doi: 10.3389/fimmu.2018.02507 Adaptive trial designs provide the opportunity to modify design elements during trial conduct. This approach possibly reduces resources, trial duration as well as sample size and may increase the probability of identifying an effective treatment (1). Currently, they are more frequently applied by industrial funders as indicated at clinicaltrials.gov. In a mini-review, Talisa et al. (2) proposed such designs to face the challenges of treatment development in sepsis and underlined their advantages. Finally, they pointed to some challenges in their application. For example, both extensive simulations for the selection of an adequate trial design and statistical models reflecting the trial structure are essential. In general, we completely agree that adaptive designs can be beneficial but they are no magic solutions for all challenges. Moreover, we want to stress that "safety assessments" are not a special topic of adaptive trial designs—data safety monitoring committees will also be in charge here but may require some additional education. With this comment, we would like to initiate a more differentiated view of methodological, statistical and practical challenges related to such designs. We comment on four topics: response adaptive randomization (RAR), adaptive enrichment, seamless, and platform designs.

RAR may be regarded as a kind of futility/inefficiency monitoring because fewer patients are assigned to trial arms in which less efficiency is observed (3). This potentially reduces the total sample size without a considerable loss of precision (1). However, one similarly achieves this aim with adaptive designs without RAR including interim monitoring for futility/inefficiency whose results are easier to comprehend (3). This idea can be extended to flexible multi-arm multi-stage sequential trial designs and, depending on the expected number of effective treatments, applied instead of RAR (4). However, when applying RAR, one must consider further issues. (i) The decision rules for adapting the allocation must be planned upfront of study initiation. Otherwise, introduction of statistical, operational and investigator-driven bias is possible which is difficult to identify and eliminate afterwards (1). (ii) One must pay special attention to the randomization weights update to avoid instable estimates (especially in the trial beginning), extremely unbalanced allocation proportions and erroneous dropping of trial arms (1, 3–6). (iii) Timing plays an important role. First, the outcome might be time-associated, introducing bias if not accounted for (3). Secondly, RAR requires a short-term outcome as otherwise there will be almost no randomization adaptation (1, 7). Short-term outcomes in clinical sepsis research, however, have been questioned (8).

**11**

In adaptive enrichment designs, the eligibility criteria are adapted during trial conduct to identify the probably most benefitting patient (sub-) population or the most efficacious dose. This is ideally based on profound understanding of the underlying biology (9). Though, one must decide on the enrichment's primary aim: noise reduction, identification of high-risk patients or of patients most likely showing a positive treatment response (10). Biomarkers are thought to serve as objective indications of biomedical state observed from outside the patient. Roughly one may distinguish prognostic and predictive biomarkers (10, 11). Prognostic enrichment would imply that prognostic biomarker results can be used to increase the selective enrollment of patients having a disease-related endpoint. Predictive enrichment would imply that predictive biomarker results can be used to increase the likelihood of responding to a treatment. Both types of enrichment are of interest for sepsis research but we are facing a "Chicken Egg Problem": For predictive enrichment there is no biomarker to monitor the response to a sepsis treatment given that we are (still) searching for an effective sepsis treatment.

The idea of seamless designs was introduced in drug development in order to move from phase II to phase III without a recruitment stop while using all the information in a joint final analysis (12). Typically, only a limited number of study arms will enter the phase III stage. The optimal time for the transition from phase II to III or the adaptive changes such as the decision on which arm(s) to drop now become crucial determinants of efficiency (13).

Finally, Talisa et al. (2) talked about platform trials. Generally, the platform idea is basically an IT infrastructure that should lead to a more efficient trial conduct by identifying and selecting patients eligible across many trials in parallel. Incorporation of adaptive design features into the platform design will not only require the introduction of Bayesian statistical models but also

#### REFERENCES


a big move in terms of trial logistics. Both topics ask for major changes in the way we teach biostatistics in academic medicine, in the way we recruit patients at our trial centers and on the side of some regulators. We want to make sure that clinicians comprehend the results of Bayesian statistical models, see the use of valid—i.e., cleaned—information for adaptations and finally ensure that robust results get accepted by regulators, e.g., as done with the agreement of the U.S. Food and Drug Administration on an adaptive, seamless phase II/III trial with Bayesian interim analysis (14).

In sum, we agree that changes in clinical sepsis research should also consider changes in the way we run clinical trials. In the short-term, however, we see organizational issues to be addressed first, e.g., the implementation of a central IT platform and a better integration of clinical research into the routine infrastructure [e.g., (15)]. Regarding the major methodological changes discussed here, costs and benefits need to be carefully checked. Bias is introduced easily and robustness of findings cannot be expected on the base of just a few observations. Frustrating as clinical sepsis trial results were in the past, we nevertheless learned that some of our ideas were likely too simple, including the way that we ran clinical trials.

#### AUTHOR CONTRIBUTIONS

Both authors contributed equally to the work, and approved it for publication.

#### FUNDING

This work was supported by the Integrated Research and Treatment Center, Center for Sepsis Control and Care (CSCC), at the Jena University Hospital funded by the German Ministry of Education and Research (BMBF No. 01EO1502). AS also received funding by BMBF No. 01ZZ1803C.


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

Copyright © 2018 Kesselmeier and Scherag. 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.

# Park 7: A Novel Therapeutic Target for Macrophages in Sepsis-Induced Immunosuppression

Yanwei Cheng1,2, Tony N. Marion3,4, Xue Cao2,3, Wanting Wang<sup>1</sup> and Yu Cao1,2 \*

*<sup>1</sup> West China Hospital Emergency Department, State Key Laboratory of Biotherapy, West China Hospital, Sichuan University, and Collaborative Innovation Center of Biotherapy, Chengdu, China, <sup>2</sup> Disaster Medicine Center, Sichuan University, Chengdu, China, <sup>3</sup> Department of Rheumatology and Immunology, West China Hospital, Sichuan University, Chengdu, China, <sup>4</sup> Department of Microbiology, Immunology, and Biochemistry, The University of Tennessee Health Science Center, Memphis, TN, United States*

Sepsis remains a serious and life-threatening condition with high morbidity and mortality due to uncontrolled inflammation together with immunosuppression with few therapeutic options. Macrophages are recognized to play essential roles throughout all phases of sepsis and affect both immune homeostasis and inflammatory processes, and macrophage dysfunction is considered to be one of the major causes for sepsis-induced immunosuppression. Currently, Parkinson disease protein 7 (Park 7) is known to play an important role in regulating the production of reactive oxygen species (ROS) through interaction with p47phox, a subunit of NADPH oxidase. ROS are key mediators in initiating toll-like receptor (TLR) signaling pathways to activate macrophages. Emerging evidence has strongly implicated Park 7 as an antagonist for sepsis-induced immunosuppression, which suggests that Park 7 may be a novel therapeutic target for reversing immunosuppression compromised by sepsis. Here, we review the main characteristics of sepsis-induced immunosuppression caused by macrophages and provide a detailed mechanism for how Park 7 antagonizes sepsis-induced immunosuppression initiated by the macrophage inflammatory response. Finally, we further discuss the most promising approach to develop innovative drugs that target Park 7 in patients whose initial presentation is at the late stage of sepsis.

Keywords: Park 7, sepsis-induced immunosuppression, inflammation, macrophages, ROS, p47phox, NADPH, crystal structure

# INTRODUCTION

Sepsis is a common clinical disease with high morbidity and mortality. Annually, ∼30 million (1) people are affected by sepsis and more than 6–8 million (2) of those affected die. Despite significant advances in treatment, sepsis is still a major clinical problem and remains the leading cause of death in the critically ill patient population (3, 4) with an associated severe cost burden (5). In 2013, sepsis was responsible for more than \$23 billion (6) of hospital costs in the USA alone. Thus, sepsis has been described as"the quintessential medical disorder of the twenty-first century." On 26 May 2017, the World Health Organization listed sepsis as a global health priority by adopting a resolution to improve the prevention, diagnosis and management of this deadly disease (7).

#### Edited by:

*Lukas Martin, Uniklinik RWTH Aachen, Germany*

#### Reviewed by:

*Johannes Ehler, University of Rostock, Germany Daniel Remick, Boston University, United States*

> \*Correspondence: *Yu Cao yuyuer@126.com*

#### Specialty section:

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

Received: *11 August 2018* Accepted: *25 October 2018* Published: *13 November 2018*

#### Citation:

*Cheng Y, Marion TN, Cao X, Wang W and Cao Y (2018) Park 7: A Novel Therapeutic Target for Macrophages in Sepsis-Induced Immunosuppression. Front. Immunol. 9:2632. doi: 10.3389/fimmu.2018.02632*

**14**

In the recent "sepsis-3" consensus (8), sepsis is defined as a life-threatening, multiorgan dysfunction caused by a dysregulated host response to infection, which is primarily caused by Gram-negative bacteria. However, a global study of 14,000 critically ill patients found that 47% of isolates were Gram-positive, indicating that more patients currently become septic from Gram-positive infections (9). Even after an inciting infection has been resolved, septic patients continue to mount an excessive inflammatory response (10) that leads to tissue damage and organ failure. Key advances have made earlier recognition and treatment of sepsis feasible with the result that some patients can restore immune homeostasis, completely clear infection, and achieve complete recovery (11). Otherwise, patients progress into late stage sepsis and suffer from severe immunosuppression characterized by an impaired activation of the immune response and a hypo-inflammatory response (12), resulting in more difficult recovery and poor long-term outcomes with risk of cognitive and physical impairments, even an increased incidence of delayed death due to the lack of effective treatment for sepsis-induced immunosuppression (13). At present, immunosuppression in septic patients constitutes an important focus of research. Thus far, various interrelated, nonmutually exclusive mechanisms have been proposed to explain sepsis-induced immunosuppression, including cellular apoptosis (14), autophagy (15, 16), regulation by the central nervous system (17, 18), metabolic reprogramming (19), epigenetic regulation (20–22), and endotoxin tolerance (23–25). The immunopathogenesis of sepsis-induced immunosuppression is a very complex process that involves both innate and adaptive immune cells. In fact, it is at least partially caused by the dysfuction of macrophages.

#### MACROPHAGES AND SEPSIS-INDUCED IMMUNOSUPPRESSION

Macrophages play essential roles throughout all phases of sepsis with their ubiquitous presence and comprehensive effects on immune homeostasis and inflammatory process. After infection, macrophage is activated through Tolllike receptor (TLR) that recognizes pathogen-associated molecular patterns (PAMPs) of the invading pathogen, such as lipopolysaccharide (LPS) in Gram-negative bacteria and lipoteichoic acid (LTA)/peptidoglycan (PGN) in Gram-positive bacteria (26). In the early stage of sepsis, macrophages undergo M1 differentiation and promote host defense by eliminating invading pathogens or damaged tissues and releasing massive amounts of pro-inflammatory cytokines such as tumor necrosis factor alpha (TNF-a), interleukin-1 (IL-1), interleukin-6 (IL-6) and interleukin-8 (IL-8) (27). However, macrophages may be excessively activated during the early phase and produce excessive pro-inflammatory cytokines (28), which have been identified as one of the major causes for the high mortality rate in the early stage of sepsis (29). If macrophagemediated pro-inflammatory responses cannot be adequately regulated, a cytokine storm may emerge (30) with the proinflammatory response becoming pathogenic and eventually immunosuppressive in late stage sepsis (31–33). As activated proinflammatory macrophages undergo apoptosis and/or polarize to the M2 phenotype that dampens the pro-inflammatory response, they may contribute to immunosuppression. Due to the cytokine storm, a large number of apoptosis-inducing factors are generated and released, including TNF-a, high mobility group box-1 protein (HMGB1) (34), thereby inducing and promoting macrophage apoptosis (35). Previous studies (36, 37) have determined the presence of an excessive level of macrophages apoptosis in human autopsies and animal models of sepsis. However, escaped M1 macrophages from apoptosis convert into M2 macrophages, showing downregulated inflammatory cytokines but upregulated anti-inflammatory cytokines (38). Certain cytokines (i.e., TNF-a, IL-13, IL-4, IL-10 etc.) can stimulate the polarization of macrophages toward M2 phenotype (39–41). Porta et al. (42) found that LPS-tolerant macrophages have the same characteristics as M2 macrophages. When a gram-negative infection persists, long-term accumulation of LPS can reprogram inflammatory responses (43) from activation to suppression leading to decreased production of inflammatory cytokines (44). The affected host may present a LPS-tolerant state, and macrophages also display the phenomenon of LPStolerance (45–47). In addition, M2 phenotype macrophages also accelerate T cell apoptosis and suppress Th1 cell responses (48). Collectively, this "dysfunctional" macrophage plays a key role in the pathogenesis of sepsis-induced immunosuppression because their pro-inflammatory cytokine secretions to support effective immune reactivity against primary or secondary pathogens is compromised. Therefore, modulating homeostasis of proand anti-inflammatory responses and functional stabilities of macrophages can be of great benefits for sepsis-induced immunosuppression.

#### REACTIVE OXYGEN SPECIES (ROS) AND MACROPHAGES

In addition to its cytotoxic function, reactive oxygen species (ROS) can initiate multiple signal transduction cascades to modulate macrophage function and are critical to the regulation of immune responses against pathogens (49). Previous studies have shown that ROS have an established role in regulating TLR signaling pathways, such as TLR/NF-κB and TLR/MARKs pathways (50–52). In LPS-tolerant macrophages, LPS tolerance blunts the TLR4 signaling, inhibiting the activation of the NF-κB signaling pathway downstream of TLR4, resulting in reduced production of inflammatory cytokines in response to LPS challenge (53–55). ROS can modulate the production of pro-inflammatory cytokines from LPS-tolerant macrophages by activating TLR4/NF-κB and TLR/MARKs pathways (49) mainly by accelerating the phosphorylation of IκBα and MAPK phosphatases (56, 57), respectively. In addition, it has been reported that TLR2-deficient macrophages lacked the response to Gram-positive LTA and PGN (58, 59), which can interact with TLR2, leading to NF-κB activation and induction of proinflammatory mediators in macrophages (59, 60). Rajamani (61) also demonstarted that high glucose mediated ROS could

induce TLR-2 activation and downstream NF-κB signaling mediating increased inflammation during diabetic retinopathy. TLR4/NF-κB pathway also plays a central role in the regulation of macrophage polarization (48). M1 macrophage polarization is related to the activation of the TLR4/NF-κB pathway (62), whereas M2 macrophage polarization is associated with the down-regulatation of NF-κB pathway (63). A recent study has confirmed that the p50 subunit of NF-κB inhibits the NF-κB pathway and M1 polarization (42). Kuchler et al. (64) reported that impaired ROS formation contributed to an M2 phenotype shift of macrophages in sepsis by inhibiting NF-κB signaling. Consequently, increased ROS formation may reduce the M2 polarization of macrophages and protect against sepsis-induced immunosuppression.

The TLR4/MARKs pathway is also involved in regulating the LPS/pro-inflammatory cytokines-induced autophagy (65). Autophagy can induce cell death but can also be a cytoprotective process. Deficient autophagy suppresses the immune response in sepsis and increases mortality (15, 16, 66). Macrophage autophagy is considered an important part of the host immune defense, eliminating intracellular pathogens through heterophagy. It has been reported that ROS can influence the MAPK pathways to activate macrophage autophagy. In hepatoma cells, migration inhibitory factor, produced by many cells including macrophages, induced autophagy via ROS generation (67). Likewise, autophagy also participates in regulating functions of macrophages and affects their ability to defend and clear pathogens through activating NF-κB pathway (68) and enhancing phagocytic capacity of macrophages (69). All of this suggests that ROS can activate macrophages to improve bactericidal and autophagy and increase production of proinflammatory cytokines, thereby helping to maintain immune homeostasis. Thus, a novel approach to improve ROS production in macrophages may be a useful therapy for sepsis-induced immunosuppression.

#### PARKINSON DISEASE 7

Parkinson disease 7 (Park 7), also known as DJ-1 (70), is highly conserved in almost all organisms and is ubiquitously expressed in all tissues and organs (71). Park 7 was initially discovered as a novel oncogene product (72) and is considered as a major causal factor for the early onset of Parkinson's disease (73). In the past two decades, Park 7 has been intensely studied in many diseases including cancer (74), neurodegenerative disorderes (75) and stroke (76). Among these diseases, Park 7 not only serves as a reliable predictor of auxiliary diagnosis, but also is a useful therapeutic target. Park 7 is a multi-functional protein with transcriptional regulation, protein chaperone, protease, and antioxidative stress functions (77). At present, increasing evidence has demonstrated that Park 7 plays

important functions in protecting neurons (78), astrocytes (79), cardiomyocytes (80, 81), and renal proximal tubule cells (82) against oxidative stress-induced cell injury. In addition, Park 7 played an important role in restoring impaired autophagy and ameliorated phenylephrine-induced cardiac hypertrophy in a repression of cardiac hypertrophy model (83). Oxidative stress is strongly related to inflammation and is thought to be involved in the processes of many diseases, including sepsis (84). Recently, accumulating lines of evidence for Park 7 in activating the inflammatory response through modulating ROS regulating oxidative stress have also been reported (53, 85). As an antioxidant, Park 7 helps to limit to cell and tissue injury in a number of diseases by removing accumulated ROS (82, 86–89). However, studies had shown that Park 7 surprisingly seems to be required for high intracellular ROS production (85, 90). Therefore, Park 7 plays a dual role in buffering cellular ROS levels: functions as a scavenger in high ROS levels, whereas helps ROS production when essential ROS are required. In view of the hypo-inflammation characteristics of sepsis-induced immunosuppression and the critical role of Park 7 in modulating ROS production and initiating an inflammatory response, recently it has been reported that Park 7 can protect against sepsis-induced immunosuppression.

# PARK 7 PROTECTS AGAINST SEPSIS-INDUCED IMMUNOSUPPRESSION (FIGURE 1)

In a park 7 knock-out (KO) mouse injected with LPS, Liu et al. (53) found that park 7 KO mice present immunosuppression phenotypes similar to the late stage of sepsis but not acute inflammation state, suggesting that park7 KO mice could serve as an animal model of sepsis-induced immunosuppression. In this model, Park 7 absence led to macrophage paralysis that resulted in increased abdominal bacterial burdens, reduced local and systemic inflammation, and impaired pro-inflammatory cytokines induction, eventually leading to high susceptibility to LPS. Neutrophil paralysis, similar to macrophage paralysis described above, was described in experimental studies of patients and sepsis animal models and was associated with decreased production of ROS in neutrophils (91, 92). In a liver fibrosis model, Park 7 deficiency inhibited ROS production in macrophages (93). Similarly, Liu et al. also observed greatly reduced ROS production in macrophages from park 7 KO mice (53). Macrophages with Park 7 deficiency showed downregulation of NF-κB and MAPK signaling pathways downstream of TLR suggesting that Park 7 deficiency can reduce the ROS production to limit TLR signaling and impair the activation of macrophages. Restoration of Park 7 expression with an inducible Park7 transgene restored the production of ROS in Park 7 KO macrophages to subsequently restore TLR signaling, pro-inflammatory cytokine production, bactericidal function, and eventually improve survival of the Park 7 KO mice in the late stage of sepsis. However, immunosuppressive IL-10 was not simultaneously enhanced after restoration of Park7 expression. During the late stage of sepsis, Park 7 may also enhance the macrophage functions by restoring impaired macrophages autophay through increased ROS and TLR/MARK signaling. Macrophage autophagy can affect cell death via complex pathways involving crosstalk with apoptosis, which may also partly attenuate immunosuppression (94). Moreover, Park 7 may contribute to the M1 macrophages polarization and inhibit the M2 macrophages polarization by the increased ROS.

Although there are many sources of ROS within macrophages, NADPH oxidase-derived ROS are critical in host defense. When macrophages are stimulated by an extracellular stimulus such as hormones, cytokines, and other inflammatory factors, the six proenzyme subunits of NADPH oxidase (95), including p22phox , gp91phox, GTPase Rac, p40phox, p47phox, and p67phox, form the holoenzyme complex that catalyzes the transfer of NADPH electrons to oxygen molecules to produce ROS (96). Key to the assembly process of the holoenzyme complex is p47phox (97, 98). After macrophages are stimulated extracellularly, p47phox , which resides in the cytosol during the resting state (95), is phosphorylated and translocated to the plasma membrane together with the remaining proenzyme subunits and activation of NADPH oxidase (99, 100). Consistent with Liu's study (53), by interacting with p47phox and modulating phosphorylation and membrane translocation of p47phox, Park7 promoted NADPH oxidase assembly and induced the production of ROS in macrophages. This mechanism supports the hypothesis that Park 7-targeted therapy maybe useful in the future in the treatment of sepsis-induced immunosuppression.

#### IS PARK 7 A POTENTIAL TARGET FOR DRUG TREATMENT IN THE FUTURE?

In this decade, many reports have shown the therapeutic potency of Park 7 and Park 7-targeting molecules/compounds in treating several neurodegenerative disorders (101–103). Can Park 7 be a potential target for drug treatment for sepsis-induced immunosuppression in the future? Structure-based drug design (SBDD) (104), as a valuable pharmaceutical lead discovery tool, opens up new opportunities for drug design for the patient with sepsis-induced immunosuppression. A typical example is the successful design of many valuable drugs by SBDD based on the crystal structure of Class B G-protein-coupled receptors (105). As noted above, the interaction of Park 7 and p47phox is a decisive factor in activating macrophages to ameliorate sepsisinduced immunosuppression, suggesting that the interaction between Park 7 and p47phox may be an ideal target for drug design. Single crystal structures of Park 7 and p47phox have been determined. Human Park 7 consists of 189 amino acids from N-terminus to C-terminus, which folds into a helix-strand-helix sandwich structure (106). The C-terminal domain (CTD) of Park 7 physically interacts with p47phox in vitro (53). In addition, the C106 and L166 residues in the CTD of Park 7 are important for its functions (107, 108), suggesting the two residues might play a key role in Park 7 interacting with p47phox. However, the details of the interaction depend on the crystal structure of the Park7 p47phox complex. Therefore, determing Park7-p47phox complex structure should be an urgent issue for future research.

With regard to a potential drug treatment based on Park 7 in the future it might be important to discuss three relevant points here. (1) It would be necessary to detect/diagnose the immune status of the patient in sepsis-induced immunosuppression. (2) In line with this it would be crucial to find the right timing to start drug treatment to overcome sepsis-induced immunosuppression. (3) Considering the complexity of the host response during sepsis and the variety of pathophysiological pathways involved, it is unlikely that the current "one-target" and "one-size-fits-all" approach will ever be successful. To date, absolute lymphocyte count and decreased expression of HLA-DR by monocytes seem to be the most robust markers for patient stratification in multicenter clinical trials (109–112). Measurement of soluble mediators such as IL-6, IL-10, and TNF-a can also help detect immune status. However, a convenient, faster detection protocol and other effective drugs are extremely necessary. These are interesting issues that are worth pursuing in the future.

#### CONCLUSION

In summary, macrophages, as one of the most important cells of the innate immune system, play an important role in inflammatory and immune processes. In the early stage of sepsis, macrophages usually have a pro-inflammatory phenotype, whereas the excessive inflammatory macrophage response can lead to macrophages apoptosis and change macrophage

#### REFERENCES


polarization contributing to the immunosuppression. ROS have the capacity to initiate many TLR signaling pathways and in turn modulate macrophage functions and are produced by the activation of NADPH oxidase. Park 7 has been extensively studied in many diseases and can serve as an effective therapeutic target. For research on sepsis in the late stage, Park7 KO mice can be an ideal model. The interaction of Park7 and p47phox can activate NADPH oxidase and subsequently increase ROS in macrophages to initiate TRL signaling to in turn, reinforce macrophage functions to protect against sepsis– induced immunosuppression. In light of this understanding, the Park 7/p47phox/ROS axis may become an effective therapeutic target for sepsis induced immunosuppression.

#### AUTHOR CONTRIBUTIONS

YaC wrote the first draft of this article. XC and WW designed the figure. YC and TM critically revised the manuscript for important intellectual content. All authors approved the final version.

#### FUNDING

The present work was supported by the National Natural Science Foundation of China (Grant Nos. 81471836 and 81772037), the Chengdu Science and Technology Huimin Project (Grant No. 2016-HM0M2-00099-SF).


bacteria on NOSII and TNFalpha in macrophages: role of TLRs in synergy between the two. Br J Pharmacol. (2006) 148:1067–75. doi: 10.1038/sj.bjp. 0706815


against neurodegeneration in Parkinson's disease and stroke model rats. Mol Neurodegener. (2011) 6:48. doi: 10.1186/1750-1326-6-48


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

Copyright © 2018 Cheng, Marion, Cao, Wang and Cao. 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.

# Linagliptin Attenuates the Cardiac Dysfunction Associated With Experimental Sepsis in Mice With Pre-existing Type 2 Diabetes by Inhibiting NF-κB

Sura Al Zoubi <sup>1</sup> \*, Jianmin Chen<sup>1</sup> , Catherine Murphy <sup>1</sup> , Lukas Martin<sup>1</sup> , Fausto Chiazza<sup>2</sup> , Debora Collotta<sup>2</sup> , Muhammad M. Yaqoob<sup>1</sup> , Massimo Collino2† and Christoph Thiemermann<sup>1</sup> \* †

#### Edited by:

Rudolf Lucas, Augusta University, United States

#### Reviewed by:

Hugo Caire Castro-Faria-Neto, Fundação Oswaldo Cruz (Fiocruz), Brazil Markus Bosmann, Boston University, United States

#### \*Correspondence:

Sura Al Zoubi s.y.y.alzoubi@qmul.ac.uk Christoph Thiemermann c.thiemermann@qmul.ac.uk

†These authors have contributed equally to this work and are joint last author

#### Specialty section:

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

Received: 04 September 2018 Accepted: 04 December 2018 Published: 18 December 2018

#### Citation:

Al Zoubi S, Chen J, Murphy C, Martin L, Chiazza F, Collotta D, Yaqoob MM, Collino M and Thiemermann C (2018) Linagliptin Attenuates the Cardiac Dysfunction Associated With Experimental Sepsis in Mice With Pre-existing Type 2 Diabetes by Inhibiting NF-κB. Front. Immunol. 9:2996. doi: 10.3389/fimmu.2018.02996 <sup>1</sup> Centre for Translational Medicine and Therapeutics, William Harvey Research Institute, Queen Mary University of London, London, United Kingdom, <sup>2</sup> Department of Drug Science and Technology, University of Turin, Turin, Italy

The mortality rate of patients who develop sepsis-related cardiac dysfunction is high. Many disease conditions (e.g., diabetes) increase the susceptibility to infections and subsequently sepsis. Activation of the NF-κB pathway plays a crucial role in the pathophysiology of sepsis-associated cardiac dysfunction and diabetic cardiomyopathy. The effect of diabetes on outcomes in patients with sepsis is still highly controversial. We here hypothesized that type 2 diabetes (T2DM) augments the cardiac (organ) dysfunction associated with sepsis, and that inhibition of the NF-κB pathway with linagliptin attenuates the cardiac (organ) dysfunction in mice with T2DM/sepsis. To investigate this, 10-week old male C57BL/6 mice were randomized to receive normal chow or high fat diet (HFD), 60% of calories derived from fat). After 12 weeks, mice were subjected to sham surgery or cecal ligation and puncture (CLP) for 24 h. At 1 hour after surgery, mice were treated with linagliptin (10 mg/kg, i.v.), IKK-16 (1 mg/kg, i.v.), or vehicle (2% DMSO, 3 ml/kg, i.v.). Mice also received analgesia, fluids and antibiotics at 6 and 18 h after surgery. Mice that received HFD showed a significant increase in body weight, impairment in glucose tolerance, reduction in ejection fraction (%EF), and increase in alanine aminotransferase (ALT). Mice on HFD subjected to CLP showed further reduction in %EF, increase in ALT, developed acute kidney dysfunction and lung injury. They also showed significant increase in NF-κB pathway, iNOS expression, and serum inflammatory cytokines compared to sham surgery group. Treatment of HFD-CLP mice with linagliptin or IKK-16 resulted in significant reductions in (i) cardiac, liver, kidney, and lung injury associated with CLP-sepsis, (ii) NF-κB activation and iNOS expression in the heart, and (iii) serum inflammatory cytokine levels compared to HFD-CLP mice treated with vehicle. Our data show that pre-existing type 2 diabetes phenotype worsens the organ dysfunction/injury associated with CLP-sepsis in mice. Most notably, inhibition of NF-κB reduces the organ dysfunction/injury associated with sepsis in mice with pre-existing T2DM.

Keywords: sepsis, septic cardiomyopathy, NF-κB, IKK-16, DPP-4, linagliptin, type 2 diabetes mellitus, mice

#### Al Zoubi et al. Diabetes, Sepsis, and Organ Dysfunction

# INTRODUCTION

Sepsis is a dysregulated body response to infection that results in life-threatening organ dysfunction (1). The cardiovascular system is one of the important systems affected by sepsis. Most septic patients, and all patients with septic shock, develop sepsisrelated cardiovascular dysfunction (2, 3), which is a key driver of in-hospital mortality in these patients (4). Both the incidence of sepsis and sepsis-related mortality increase with age due to the presence of significant comorbidities including chronic kidney disease and type 2 diabetes in the elderly (5). The prevalence of diabetes is increasing worldwide (6). Patients with diabetes are at high risk of developing diabetic cardiomyopathy, left ventricular (LV) hypertrophy, ischemic cardiac injury, and heart failure (6– 9). Diabetic patients are more susceptible to both common and rare infections and have a higher incidence of sepsis than patients that do not suffer from the disease (10, 11). However, the effect of diabetes on outcome in patients with infections is controversial with some studies showing increased hospitalization, organ dysfunction/injury, and mortality in diabetic patients with e.g., pneumonia (12–14), while other studies report either no effect (15–17) or even protective effects (18–21). Hence, this study was undertaken to investigate the effects of pre-existing type 2 diabetes mellitus (T2DM) caused by high fat diet (HFD) on cardiac dysfunction in mice with sepsis.

Dipeptidyl peptidase-4 (DPP-4) inhibitors (gliptins) are among the most recently approved drugs for the treatment of hyperglycemia in patients with T2DM. Gliptins mediate their anti-diabetic effects by primarily inhibiting degradation of endogenous glucagon-like peptide 1 (GLP-1) and glucosedependent insulinotropic peptide (GIP), resulting in prolongation of postprandial insulin secretion. Given the numerous and varied substrates enzymatically cleaved or bound by DPP-4, DPP-4 inhibitors may have the potential to exhibit a broader range of salutary pleiotropic effects in the heart and vasculature, increasing the concentration of peptides with potential vasoactive and cardioprotective effects, which may be independent of GLP-1 and its receptor. DPP-4 signaling cascade has been recently demonstrated to be involved also in the pathologic features of sepsis, mainly due to a selective cross-talk between DPP-4 and nuclear factor- kappa B (NF-κB) signaling pathways (22–25). In sepsis, microbial components activate different antigen presenting cells (APCs) after binding to their Toll like receptor (TLR) 4 and 2. Exposed caveolin-1 from the activated APC interacts with DDP-4 in T cells which results in strong NF-κB activation in both the APCs and the T cells (23). Adenosine deaminase (ADA) is another activator of DPP-4. The activity of ADA in the serum is increased during inflammatory diseases (e.g., sepsis) as a result of increased macrophages activity (26) and the interaction of ADA with DPP-4 leads to NF-κB activation in T cells (22). The expression of DPP-4 is increased in diabetic patients (27–29).

Activation of NF-κB plays a crucial role in the pathophysiology of both septic (30–32) and diabetic cardiomyopathy (33). In sepsis, activation of NF-κB is secondary to activation of TLR 2 and 4 by e.g., by wall fragments of Gramnegative (e.g., lipopolysaccharide; LPS) or Gram-positive bacteria (e.g., peptidoglycan; PepG) and/or pro-inflammatory cytokines including tumor-necrosis factor-α (TNF-α) or interleukin-1 (IL-1). In addition to pro-inflammatory cytokines, the activation of NF-κB in diabetes is also driven by free fatty acids (34) (which activate TLR4) and advanced glycation end products (which activate RAGE) (35). We have recently demonstrated that inhibition of the activation of NF-κB by a selective IκB kinase (IKK) inhibitor (IKK-16) attenuates the cardiac dysfunction caused by sepsis in mice without co-morbidities (32) and in mice with pre-existing chronic kidney disease (36). However, the potential protective effects of the impairment of the crosstalk between DDP-4 and NF-κB activation in sepsis-induced multiorgan dysfunction have never been tested in animal models of diet-induced diabetes, which is known to be characterized by an increase in baseline NF-κB activity. Here we investigate (a) the role of NF-κB activation in the cardiac dysfunction caused by HFD with or without sepsis, (b) the effect of linagliptin treatment on cardiac performance in the model of sepsis and T2DM. To investigate the relative contribution of NF-κB inhibition in the observed effects of linagliptin, IKK-16, a potent and selective IKK inhibitor, was used as a comparative pharmacological tool.

# MATERIALS AND METHODS

#### Ethical Statement

The animal protocols followed in this study were approved by the local Animal Use and Care Committee in accordance with the derivatives of both, the Home Office Guidance in the Operation of Animals (Scientific Procedure Act 1986) published by Her Majesty's Stationary Office, the Guide for the Care and Use of Laboratory Animals of the National Research Council (home office project license, PPL: 70/8350) and by the local ethical committee (DGSAF 0021573-P-12/11/2013) and are in keeping with the European Directive 2010/63/EU as well as the 2011 Guide for the Care and Use of Laboratory Animals.

#### Animals

This study was conducted on 56 male C57BL/6 mice aged 10 weeks (Charles River, Kent, UK) weighing 25–30 g, receiving a standard diet and water ad libitum (before starting the experiments). Mice were housed 5 per cage in a temperaturecontrolled room with a 12-h light/dark cycle.

# High Fat Diet Model of Type 2 Diabetes Mellitus

In this model of diet-induced obesity and insulin resistance, mice (Charles River UK, Kent) were randomized to receive standard chow diet [LabDiet <sup>R</sup> , St. Louis (5053: protein provides 25%, fat 13%, and carbohydrate 62% of the total calories)] or high fat diet [TestDiet <sup>R</sup> , St. Louis (58Y1: Blue diet; protein provides 18.1%, fat 61.6%, and carbohydrate 20.3% of the total calories)] with free access to water for 12 weeks. Body weight, food intake, and calories intake were measured weekly through the experiment to monitor feeding behavior.

#### Oral Glucose Tolerance Test (OGTT)

Mice were fasted for 6 h with free access to water. At the end of the 6 h fasting, the body weight and fasting blood glucose were measured using tail vein blood. Mice then received a bolus dose of glucose (2 g/kg, dissolved in drinking water) via oral gavage. Blood glucose levels were then measured at 15, 30, 45, 60, 90, and 120 min post glucose administration using blood glucose meter Accu-Chek <sup>R</sup> (Accu-Chek Compact System; Roche Diagnostics, Basel, Switzerland).

#### Insulin Tolerance Test (ITT)

Mice were fasted for 4 h with free access to water. At the end of the 4 h fasting, the body weight and fasting blood glucose were measured using tail vein blood. Mice then received a dose of insulin aspart (NovoRapid <sup>R</sup> ) (0.75 U/kg, i.p.). Blood glucose level was then measured at 15, 30, 45, 60, 90, and 120 min post insulin administration using blood glucose meter Accu-Chek <sup>R</sup> (Accu-Chek Compact System; Roche Diagnostics, Basel, Switzerland).

#### Measuring Fasting Plasma Insulin

Mice were fasted for 6 h with free access to water. At the end of the 6 h fasting period, blood samples were collected from the tail vein. Fasting plasma insulin levels were then measured using human insulin ELISA kit following the manufacturer's instructions (Abcam <sup>R</sup> , Cambridge, UK).

#### Assessment of Baseline Kidney Function

During the last week of the experiment, mice were housed in the metabolic cages to collect urine. They were housed (one mouse per cage) for 24 h with free access to food and water. Urine biochemistry (creatinine and sodium levels) was assessed blindly by IDEXX the commercial veterinary testing laboratory (IDEXX Ltd; West Sussex, UK). Urine albumin was measured using a mouse albumin ELISA kit following manufacturer instructions (Cambridge Bioscience <sup>R</sup> , Cambridge, UK). Then creatinine clearance (CrCl) and urine albumin to creatinine ratio (ACR) were calculated using the following equations:

$$\text{CrCl} = \frac{\text{Urine Creatinine}}{\text{Serum Creatinine}} \times \frac{\text{Urine Volume}}{\text{Time}}$$

**Equation 1:** Creatinine clearance (ml/min) is calculated using four measurements (a) urine creatinine (µmol/L), (b) serum creatinine (µmol/L), (c) urine volume (ml), and (d) time (minutes).

$$\text{ACR} = \begin{array}{c} \text{Urine} \quad \text{Albumin} \\ \hline \text{Urine} \quad \text{Creatinine} \end{array}$$

**Equation 2:** Urine albumin to creatinine ratio is calculated using 2 (a) urine albumin (µg/L) and (b) urine creatinine (mg/L).

#### Assessment of Cardiac Function in vivo (Echocardiography)

Echocardiography was conducted in vivo at baseline (before sepsis challenge) then at 24 h after CLP to measure cardiac function using a 30 MHz RMV707B scan head and a Vevo-770 imaging system (VisualSonics, Toronto, Ontario, Canada).

Animals were anesthetized using 3% isoflurane delivered with 0.4 l/min oxygen in the anesthesia chamber. After being sedated, mice were then transferred to the echo table and taped from the limbs in a supine position onto the metal ECG leads on the Echo platform. Anesthesia was maintained during the whole imaging process using 0.5–2% isoflurane delivered with 0.4 l/min oxygen via nosecone under spontaneous breathing. The handling platform was warmed to 40◦C in order to keep the core body temperature of the mice. After being placed on the platform, the fur on the chest was then removed carefully using Veet <sup>R</sup> hair removing cream. A pre-warmed echo gel is then applied to the chest to start the measurement. At least 10 min were left for the animals to stabilize before any measurement was taken. The body temperature was monitored using a rectal probe and the heart rate was obtained from ECG tracing during the whole procedure.

Measurements from both two-dimensional (brightness mode, B-mode) and one-dimensional (motion mode, M-mode) were obtained. Measurements of the left ventricle internal dimension (LVID) in both systole and diastole from the M-mode at the level of the papillary muscles were used to calculate the percentage ejection fraction (% EF), fractional shortening (% FS), and the measurements of LV end-systolic and end-diastolic areas from the B-mode were used to calculate the percentage functional area change (% FAC) using the following equations:

$$\% \text{ EF} = \frac{\text{(LVID; d}^3 - \text{LVID; s}^3)}{\text{LVID; d}^3} \times 100$$

**Equation 3:** Ejection fraction (%) is calculated using 2 measurements (a) left ventricle internal dimension during diastole (LVID;d, mm) and (b) left ventricle internal dimension during systole (LVID;s, mm).

$$\% \text{ FS} = \frac{\text{(LVID; d} - \text{LVID; s)}}{\text{LVID; d}} \times 100$$

**Equation 4:** Fractional shortening (%) is calculated using 2 measurements (a) left ventricle internal dimension during diastole (LVID;d, mm) and (b) left ventricle internal dimension during systole (LVID;s, mm).

$$\% \text{ FAC} = \frac{(\text{LVarea; d} - \text{LVarea; s})}{\text{LVarea; d}} \times 100$$

**Equation 5:** Fractional area change (%) is calculated using 2 measurements (a) left ventricle end-diastolic area (LV area;d, mm<sup>2</sup> ) and (b) left ventricle end-systolic area (LV area;s, mm<sup>2</sup> ).

#### Model of Cecal Ligation and Puncture (CLP) Induced Polymicrobial Sepsis

At 12 weeks after starting the HFD, mice fed either chow or HFD were randomized to undergo either sham operation or CLP surgery. Before surgery, mice were anesthetized using (1.5 ml/kg, i.p.) of 2:1 ketamine (100 mg/ml): xylazine (20 mg/ml) solution. To obtain adequate analgesia, buprenorphine (0.05 mg/kg, i.p.) was administered just before starting the surgery. The fur of the abdomen was removed using Veet <sup>R</sup>

hair removing cream and the area cleaned with 70% ethanol. A 1.5 cm midline incision of the abdomen was made and the caecum was exposed. The caecum then was ligated below the ileocecal valve and two punctures were made one at each end using an 18-G needle. A small amount of fecal matter was then squeezed from both punctures before the caecum was returned to its anatomical position and the cut in the abdomen was then sutured. Each mouse then received a resuscitation fluid (1 ml 0.9% NaCl, s.c.). Mice were left on a homeothermic blanket to recover then placed back into fresh clean cages. At 1 h after CLP surgery, mice from the HFD group were randomized to receive linagliptin (10 mg/kg, i.v), IKK-16 (1 mg/kg i.v.) or vehicle (2% DMSO; 3 ml/kg, i.v.). At 6 and 18 h after surgery, antibiotic (imipenem/cilastatin, 20 mg/kg) dissolved in resuscitation fluid (15 ml/kg 0.9% NaCl, s.c.) was administered along with analgesia (buprenorphine, 0.05 mg/kg, i.p.). At 24 h, mice were anesthetized for assessment of cardiac function in vivo. As a terminal procedure, mice were anesthetized using high dose isoflurane (3% delivered in 0.9 l/min O2) before being sacrificed. Blood samples were collected by cardiac puncture and vital organs were collected and snap frozen using liquid nitrogen then stored for further analysis at −80◦C freezer. Mice that underwent sham surgery were anesthetized and handled in the same manner as CLP mice during surgery. However, in animals undergoing sham surgery, the cecum (although exposed) was not subjected to perforation. At 1 h after surgery, sham mice were treated with vehicle (2% DMSO, 3 ml/kg, i.v.) and they were also treated with antibiotic (imipenem/cilastatin, 20 mg/kg) dissolved in resuscitation fluid (15 ml/kg 0.9% NaCl, s.c.) along with analgesia (buprenorphine, 0.05 mg/kg, i.p.) at 6 and 18 h after surgery. Surgery, treatments and assessment of cardiac function were performed by different member of the research team to minimize subjective errors. The investigator assessing the cardiac function was blinded as to the intervention that was used in the study.

# Western Blot Analysis

Immunoblot analyses of heart biopsies were carried out using a semi-quantitative western blotting. We measured the degree of phosphorylation of IKK, IκBα, and Akt, nuclear translocation of p65 NF-κB subunit to the nucleus and inducible nitric oxide synthase (iNOS) expression. For sample handling, blood was withdrawn from mice at the time of euthanasia, heart tissue was then washed with saline solution prior to homogenisation. For sample preparation, a piece of heart tissue was taken and diluted (1:10) with homogenisation buffer (HB) at 4◦C to obtain a whole tissue lysate, that contains endothelium, cardiomyocytes and leucocytes, for protein extraction. Cytosolic and nuclear extracts from hearts were prepared as previously described (37). Succinctly, hearts were homogenized at 10% (wt/vol) with a Potter Elvehjem homogenizer (Wheaton, Millville, NJ) using a homogenization buffer containing 20 mM HEPES (pH 7.9), 1 mM MgCl2, 0.5 mM EDTA, 1% Nonidet P-40, 1 mM EGTA, 1 mM DTT, 0.5 mM PMSF, 1 µl/ml of PIC. Homogenates were centrifuged at 1,300 g for 5 min at 4◦C. Supernatants were removed and centrifuged at 16,000 g at 4◦C for 40 min to obtain supernatant containing the cytosolic fraction. The pelleted nuclei were resuspended in extraction buffer (1/3 volume of the homogenation buffer) containing 20 mM HEPES (pH 7.9), 1.5 mM MgCl2, 300 mM NaCl, 0.2 mM EDTA, 20% glycerol, 1 mM EGTA, 1 mM DTT, 0.5 mM PMSF, 1 µl/ml of PIC, and incubated in ice for 30 min, followed by centrifugation at 16,000 g for 20 min at 4◦C. The resulting supernatants containing nuclear proteins were carefully removed. Both cytosolic and nuclear proteins were measured using bicinchoninic acid (BCA) protein assay following manufacturer's directions (Therma Fisher Scientific, Rockford, IL). Proteins were separated by gel electrophoresis using sodium dodecyl sulfate polyacrylamide (SDS-PAGE) then transferred into a Polyvinylidene difluoride (PVDF) membrane. The membrane then was blocked 1 h in 5% dry milk solution in TBS-tween). Incubation of the membrane was conducted overnight at 4◦C with primary antibody in 5% blocking solution then incubated the next day with the appropriated HRP-conjugated secondary antibody at room temperature for 30 min and then detected with enhanced chemilumescent (ECL) detection system and quantified by densitometry using BioRad Image Lab SoftwareTM 6.0. Results were normalized with respect to densitometric value of mouse anti-tubulin for cytosolic and total proteins, and mouse antihistone H3 for nuclear proteins.

# Cytokine Measurements

Serum cytokines levels (TNF-α, IL-6, KC, and IL-10) were measured using a bead-based immunoassay method. Serum samples were prepared and handled following manufacturer instructions (Biolegend <sup>R</sup> , San Diego, USA). Data was obtained using a LSR Fortessa (Biociences <sup>R</sup> , Berkshire, UK) and analyzed using the LegendplexTM 7.1.0.0 software.

# Measuring Myeloperoxidase (MPO) Activity in the Lung

MPO was extracted from the tissues according to the methods described by Barone et al. (38) with slight modifications to measure neutrophil accumulation in the lungs. For samples preparation, a piece of lung was diluted (1:20) with 5 mM potassium phosphate buffer to homogenize the sample (at 4 ◦C). For measurements of MPO activity, the homogenate was centrifuged (at 13,000 RPM, 30 min, 4◦C). The resulted supernatant was discarded. A 5-time dilution with 0.5% hexadecyl-trimethyl-ammonium bromide in 50 mM potassium phosphate buffer was used to suspend and homogenize the pellet. The resulted solution was then frozen and thawed three times followed by 10 s sonication at room temperature and then incubated at 4◦C for 30 min then centrifuged (at 12,500 RPM, 15 min, 4◦C). MPO activity was measured in the supernatant by mixing 100 µl of the supernatant with 0.167 mg/ml o-dianiside dihydrochloride and 0.0005% hydrogen peroxide in 2.9 ml 50 mM potassium phosphate buffer. UVvisible spectrophotometer was used to measure the change in absorbance at 460 nm for 1.5 min. MPO activity was presented as the quantity of the enzyme degraded 1 µmol of peroxide/min at 25◦C and expressed as µU/gram of the lung tissue.

# Measuring N-acetyl-β-D-glucosaminidase (NAG) Activity in the Lung

NAG activity was analyzed to measure macrophage accumulation in the lung. For samples preparation, a piece of lung was diluted with 0.01 M phosphate buffer saline and homogenized (at 4◦C). The resulted solutions were then frozen and thawed three times followed by 10 s sonication at room temperature to break the cells. For measurements of NAG activity, the homogenate was centrifuged (at 5,000 RPM, 30 min, 4◦C). The resulted supernatants were then used to measure NAG activity using a NAGase ELISA kit following manufacturer instructions (Elabscience <sup>R</sup> , Houston, Texas, USA).

#### Statistical Analysis

Data was analyzed using GraphPad Prism 7.0 (GraphPad Software, San Diego, California, USA). Values stated in the text and figures are presented as a mean ± standard error of the mean (SEM) of n observations, where n is the number of animals used. Data was tested for normality using D'Agostino-Pearon normality test and then assessed using One-way ANOVA test followed by Bonferroni's post-hoc test or unpaired Student t-test where appropriate. P-values of less than 0.05 were considered to be statically significant.

#### Materials

Unless otherwise stated, all materials, reagents, and solutions were purchased from Sigma-Aldrich Ltd (Poole, Dorset, UK).

# RESULTS

#### Diabetic Phenotype and Characterization of Organ Dysfunction in Mice With Experimental T2DM Caused by HFD

When compared to chow fed mice, mice fed a HFD showed a significant increase in fasting blood glucose, impairment in glucose tolerance after being challenged with an oral dose of glucose, impairment in insulin tolerance after challenge with i.p. insulin, an increase in fasting plasma insulin as well as an increase in total cholesterol. Mice fed a HFD also showed significant increases in (i) total body weight secondary to an increase in fat mass, but not lean mass; (ii) an increase in the serum levels of alanine aminotransferase (ALT) indicating the development of liver injury; (iii) increase in urine albumin to creatinine ratio (ACR) as well as a (iv) a decrease in creatinine clearance (CrCl) indicating the development of diabetic nephropathy (proteinuria) and glomerular dysfunction (P < 0.05). Mice also showed mild cardiomyopathy as evidence by a small, but significant, reduction in ejection fraction (EF%), fractional shortening (FS%), and fractional area change (FAC%) (P < 0.05; **Table 1**).

#### Pre-existing T2DM Augments the Multiple Organ Dysfunction and Systemic Inflammation Associated With Sepsis in Mice

Subjecting mice on chow diet to CLP resulted in a small (but not significant) effect on systolic cardiac function compared to TABLE 1 | Baseline data for both chow and HFD groups before interventions (CLP or sham surgeries).


Mice fed a HFD were compared to age-matched mice fed a chow diet. Data is presented as mean ± SEM for n number of observations. Data was analyzed by unpaired t-test. \*P < 0.05 vs. the chow-fed group. \$: mean food intake of 18 mice in each group during the study period 12 weeks (n = 12). AUC, area under the curve; OGTT, oral glucose tolerance test; ITT, insulin tolerance test

sham surgery (P > 0.05; **Figure 1**). However, mice fed a HFD and subjected to CLP exhibited a large and significant reduction in systolic cardiac function compared to sham surgery (P < 0.05; **Figure 1**).

To understand the signaling mechanism associated with cardiac dysfunction, we investigated the effect of HFD on the activation of key signaling pathways of inflammation and cell survival including pathways leading to the activation of NF-κB with or without sepsis. Mice fed a HFD exhibited significant increases in the phosphorylation of IKKα/β on Ser178/<sup>180</sup> , the phosphorylation of IκBα on Ser32/36, the translocation of p65 NF-κB to the nucleus, the expression of iNOS, and a significant decrease in the phosphorylation of Akt on Ser<sup>473</sup> (P < 0.05; **Figures 2**, **3**). Exposing mice on chow diet to CLP resulted in a similar degree of activation of NF-κB and, indeed, iNOS expression (in the heart) as observed in mice with HFD alone. CLP-sepsis in mice on HFD resulted in a further increase in the phosphorylation of IKKα/β and IκBα, the translocation of p65, and iNOS expression (P < 0.05; **Figure 2**) with no significant effect on Akt phosphorylation (P > 0.05; **Figure 3**).

We also studied the effect of HFD on the systematic synthesis of key, NF-κB-dependent cytokines including TNF-α, IL-6, KC, and IL-10. When compared to mice on chow diet, mice on HFD for 12 weeks showed no significant changes in cytokines levels (P > 0.05; **Figure 2**). When compared to mice on regular chow

randomized to undergo CLP or sham surgery. At 1 h after CLP, mice on HFD were randomized to receive linagliptin (10 mg/kg, i.v.), IKK-16 (1 mg/kg, i.v.), or vehicle (2% DMSO, 3 ml/kg). Cardiac function was assessed at 24 h after CLP or sham surgery in mice fed a chow diet or a HFD. (A) Representative M-mode echocardiograms; percentage % (B) EF, (C) FS, and (D) FAC. Data was analyzed using one-way ANOVA followed by Bonferroni's post-hoc test and expressed as mean ± SEM. \*P < 0.05 compared to HFD+CLP group. (n = 8 for sham surgery groups and 10 for CLP groups).

diet subjected to sham surgery, mice on chow diet subjected to CLP (in the presence of antibiotics) showed no significant changes in cytokines levels in the serum (P > 0.05; **Figure 4**). CLP-sepsis in mice on HFD resulted in large increases in the serum levels of TNF-α, IL-6, KC, and IL-10 when compared to mice on HFD subjected to sham surgery and mice on chow diet subjected to CLP (P < 0.05; **Figure 4**).

Markers for lung inflammation were also measured to study the effect of pre-existing diabetes on lung injury associated with sepsis. When compared to micev on chow diet, mice on HFD showed no significantv changes in MPO or NAG activities in the lungs after sham surgery (P > 0.05; **Figure 5**). Whenv compared to mice on regular chow diet subjected to sham surgery, mice on chow diet subjected to CLP showed a significant increase in NAG activity in the lung (P < 0.05; **Figure 5**) with no change in MPO activity (P > 0.05; **Figure 6**). Subjecting mice on HFD to CLP surgery resulted in big increases in both MPO and NAG activities when compared to mice on HFD subjected to sham surgery and mice on chow diet subjected to the same CLP (P < 0.05; **Figure 5**).

Our CLP-sepsis model with moderate severity had (in the presence of antibiotics and fluid resuscitation) no effect on serum creatinine, urea, or ALT compared to sham surgery (P > 0.05; **Figure 6**). However, challenge of mice on HFD with CLP resulted (despite the presence of antibiotics and fluid resuscitation) in significant increases in serum creatinine, urea, and ALT levels (P < 0.05; **Figure 6**). These results indicate that pre-existing T2DM increases the severity of both renal dysfunction and hepatocellular injury.

#### Effect of Linagliptin Post Treatment on the Multiple Organ Dysfunction Associated With Sepsis in Mice With Pre-existing T2DM

When compared to mice on HFD subjected to sham surgery, mice on HFD subjected to CLP and treated with vehicle

Bonferroni's post-hoc test and expressed as mean ± SEM.\*P < 0.05 compared to HFD+CLP group, \$P < 0.05. (n = 4–6 per group).

developed significant systolic cardiac dysfunction. Treatment of mice on HFD with linagliptin, at 1 h after CLP, attenuated the systolic cardiac dysfunction caused by CLP (P < 0.05; **Figure 1**).

Having found that linagliptin attenuates the cardiac dysfunction associated with CLP-sepsis in diabetic mice, we then investigated the potential mechanism(s) of this beneficial effect. Treatment of mice on HFD with linagliptin, at 1 h after CLP, also resulted in significant reduction in IKKα/β and IκBα phosphorylation, p65 translocation, and iNOS expression when compared to mice on HFD subjected to CLP and treated with vehicle (P < 0.05; **Figure 2**). Moreover, linagliptin treatment of HFD/CLP mice restored the degree of Akt phosphorylation to almost that seen in sham mice (P < 0.05; **Figure 3**).

Systemic inflammation was also attenuated after linagliptin treatment. When compared to mice on HFD challenged CLP and treated with vehicle, delayed treatment with linagliptin at 1 h after CLP resulted in significant reduction in IL-6, KC, and IL-10 synthesis (P < 0.05; **Figure 4**) and in a reduction in TNF-α that did not, however, reached statistical significance when compared to mice treated with vehicle (P > 0.05, **Figure 4**).

i.v.), IKK-16 (1 mg/kg i.v.), or vehicle (2% DMSO, 3 ml/kg). At 24 h, heart samples were collected and signaling events were assessed. Densitometry analysis of the bands is expressed as relative optical density (O.D.) of phosphorylated Akt on Ser<sup>473</sup> corrected for the corresponding total Akt content and normalized using the related sham band. Data was analyzed using one-way ANOVA followed by Bonferroni's post hoc test and expressed as mean ± SEM.\*P < 0.05 compared to HFD+CLP group, \$P < 0.05. (n = 4–6 per group).

Accordingly, a significant reduction in markers of lung inflammation, specifically the MPO and NAG activities, was observed in the lungs of mice exposed to linagliptin posttreatment when compared to mice on HFD subjected to CLP and treated with vehicle only (P < 0.05, **Figure 5**).

When compared to mice on HFD subjected to CLP and treated with vehicle, mice on HFD and treated with linagliptin, at 1 h after CLP, showed significant reduction in serum creatinine, urea, and ALT levels indicating that linagliptin attenuated the renal dysfunction and liver injury caused by CLP (P < 0.05; **Figure 6**).

#### Effect of IKK-16 Post Treatment on the Multiple Organ Dysfunction Associated With Sepsis in Mice With Pre-existing T2DM

In order to investigate whether inhibition of NF-κB was the main reason of the attenuated organ injury caused by sepsis in diabetic mice after linagliptin treatment, we have investigated the effects of the specific IKK-inhibitor IKK-16 in these animals. When compared to mice on HFD subjected to sham surgery, mice on HFD subjected to CLP and treated with vehicle developed significant systolic cardiac dysfunction. Delayed treatment of HFD mice with IKK-16 at 1 h after CLP attenuated the systolic cardiac dysfunction (P < 0.05; **Figure 1**) caused by sepsis.

The restoration of cardiac function afforded by IKK-16 in diabetic CLP-mice was accompanied by significant reduction in IKKα/β and IκBα phosphorylation, p65 translocation, and iNOS expression in mice treated with IKK-16 when compared to mice on HFD challenged with CLP and treated with vehicle (P < 0.05; **Figure 2**). Moreover, IKK-16 treatment restored the Akt phosphorylation (P < 0.05; **Figure 3**) caused by HFD with or without sepsis.

When compared to mice on HFD subjected to CLP and treated with vehicle, delayed treatment with IKK-16 at 1 h after CLP resulted in significant reduction of both the systemic levels TNF-α, IL-6, KC, and IL-10 (P < 0.05; **Figure 4**) and local (lung) MPO and NAG activities (P < 0.05, **Figure 5**). They also showed significant reduction of serum creatinine, urea, and ALT levels indicating that IKK-16 reduced both the renal dysfunction and the liver injury caused by sepsis in mice fed with a HFD (P < 0.05; **Figure 6**).

Interestingly, no statistically significant differences between the two post-treatment groups were recorded for any of the measured markers.

# DISCUSSION

Although the mortality rate among septic patients has declined due to improved supportive care for patients in the ICU (39), the incidence of sepsis has increased as a result of the aging of the population (40) which is associated with the presence of significant comorbidities such as T2DM (5). Patients with T2DM are more likely to develop infections and subsequently sepsis (10, 11). The cardiovascular system is one of the most important systems affected by sepsis and the development of cardiovascular dysfunction in sepsis has been linked to several pathophysiological driver including inflammatory cytokines and NO (32, 36). Many studies of the pathophysiology of sepsis have shown beneficial effects in pre-clinical models of sepsis. However, clinical studies that tested the efficacy of drugs targeted at identical aspects of the pathophysiology (often by using identical interventions) have failed to improve survival in septic patients (as a result of the limitations in both animal models and experimental designs) (41).

In this study, a clinically relevant model of T2DM caused by prolonged administration of HFD (for 12 weeks) was established in C57BL/6 male mice. As the consumption of a western diet is a key driver underlying the development of T2DM, our model of feeding a HFD for longer periods recapitulates the main cause of T2DM in humans and is considered to be one of the most clinically, relevant model of T2DM to date. Indeed,

(C) KC, and (D) IL-10. Data was analyzed using one-way ANOVA followed by Bonferroni's post-hoc test and expressed as mean ± SEM. \*P < 0.05 compared to

feeding of mice with HFD resulted, within 12 weeks, in the development of a diabetic phenotype (significant weight gain, impaired glucose tolerance, increased fasting blood glucose and increased fasting plasma insulin) and diabetic cardiomyopathy (reduction in %EF) as a result of NF-κB activation in the heart (see below).

HFD+CLP group. (n = 4 per sham surgery group and n = 6–8 per CLP group).

A "two-hit" model of pre-existing T2DM (secondary to HFD administration) followed by a mild CLP surgery (which did not cause significant organ dysfunction in young and healthy mice, but which we have reported to cause substantial organ dysfunctions in old mice or mice with CKD) was used to study (a) the effect of pre-existing T2DM on the cardiac dysfunction associated with sepsis and (b) to test novel therapeutic interventions aimed at reducing cardiac dysfunction in T2DM/sepsis mice. We show here, for the first time, that pre-existing T2DM augments the cardiac dysfunction associated with sepsis. T2DM alone resulted in a small degree of NF-κB activation and iNOS expression in the heart. However, sepsis (second hit) in diabetic mice resulted in a dramatic increase in the serum concentrations of proinflammatory cytokines and a further increase in both NF-κB activation and iNOS expression in the heart. Diabetes also resulted in reduction in the activation (phosphorylation) of the Akt pro-survival pathway, while sepsis resulted in further reduction of Akt phosphorylation in the heart. These two findings suggest that the cardiac dysfunction associated with T2DM/sepsis is most likely a result of the activation of the NF-κB pro-inflammatory signaling pathway (with subsequent increase in iNOS expression and serum inflammatory cytokines levels) and the concomitant inhibition of Akt pro-survival pathways.

FIGURE 5 | Effects of CLP and linagliptin or IKK-16 post treatment on neutrophil/macrophage infiltration in the lung in mice with pre-existing T2DM. Mice from chow and HFD groups were randomized to undergo CLP or sham surgery. At 1 h after CLP, mice on HFD were randomized to receive linagliptin (10 mg/kg, i.v.), IKK-16 (1 mg/kg, i.v.), or vehicle (2% DMSO, 3 ml/kg). At 24 h after CLP, lung samples were collected and neutrophil and macrophages infiltration were measured as the (A) MPO and (B) NAG activities. Data was analyzed using one-way ANOVA followed by Bonferroni's post-hoc test and expressed as mean ± SEM for n number of observations. \*P < 0.05 compared HFD+CLP group. \$P < 0.05 (n = 6 per group).

FIGURE 6 | Effects of CLP and linagliptin or IKK-16 post treatment on renal dysfenction and hepatocelleular injury in mice with pre-existing T2DM. Mice from chow and HFD groups were randomized to undergo CLP or sham surgery. At 1 h after CLP, mice on HFD were randomized to receive linagliptin (10 mg/kg, i.v.), IKK-16 (1 mg/kg, i.v.), or vehicle (2% DMSO, 3 ml/kg). At 24 h after CLP, blood samples were collected and serum (A) urea, (B) creatinine, and (C) ALT levels were measured. Data was analyzed using one-way ANOVA followed by Bonferroni's post-hoc test and expressed as mean ± SEM. \*P < 0.05 compared to HFD+CLP group. (n = 8 per sham surgery group, n = 10 per CLP groups).

The DPP-4 inhibitors gliptins (e.g., linagliptin and sitagliptin) have been used as anti-diabetic drugs that exert their catalytic effect by increasing the incretin levels. However, recent evidence indicates that DPP-4 inhibitors, as well as glucagon like peptide-1 (GLP-1) receptors agonists (e.g., liraglutide), also have antiinflammatory effects. The (off-target) non-catalytic effects of DPP-4 inhibitors have recently been discussed in the literature as a potential new therapeutic strategy for the treatment of diseases associated with local or systemic inflammation. Indeed, some preclinical studies suggest that inhibition of DPP-4 by different gliptins results in less cardiac dysfunction in a murine model of HFD-induced fibrosis and inflammation (42) and in a rat model of heart failure (43) by inhibiting NF-κB and by reducing the formation of pro-inflammatory cytokines. The effect of some gliptins and GLP-1 receptor agonists on survival and inflammation was also studied in animal models of endotoxaemia. Rodents subjected to endotoxaemia and treated with linagliptin or liraglutide (or their respective vehicles) showed an increase in survival rate (44, 45), and decreased formation of reactive oxygen species (ROS) (46). Treatment of cardiomyocytes with sitagliptin decreased the inflammatory response triggered by LPS (47). In contrast, pre-treatment with sitagliptin had no effect on survival in endotoxaemic animals (44, 45). Furthermore, treatment of endotoxaemic mice with vildagliptin ameliorated the degree of pulmonary fibrosis (48). However, the effect of DPP-4 inhibitors on the cardiac (organ) dysfunction associated with sepsis (in the absence or presence of diabetes) has not yet been investigated.

Based on previous work in animals with sepsis (without T2DM), we have hypothesized that the activation of NF-κB (proinflammatory) and the inhibition of Akt (pro-survival) pathways are the reasons for the cardiac dysfunction in T2DM/sepsis and, hence, studied the effect of linagliptin (repurposing of linagliptin) on NF-κB inhibition and Akt activation and their impact on cardiac performance. Indeed, the results of this study highlighted, for the first time, that inhibition of DPP-4 using linagliptin (at 1 h after CLP) attenuates the cardiac dysfunction associated with T2DM/CLP-sepsis and this was associated with, or occurred as a result of, an inhibition of NFκB activation and preservation of Akt pathway activation in the heart. Treatment with linagliptin also resulted in attenuation of the multiple organ dysfunction associated with T2DM/CLPsepsis.

To confirm the potential key role of the activation of NF-κB in the pathophysiology of septic cardiomyopathy in animals with T2DM, we investigated the effect of NF-κB pathway inhibition using a selective IKK inhibitor (IKK-16) on cardiac (organ) dysfunction associated with sepsis. Treatment with IKK-16 ameliorated the cardiac dysfunction in mice with T2DM/sepsis. This enhanced cardiac function was a result of the decreased NF-κB activation (and iNOS expression) and inflammatory cytokines synthesis and the restored Akt phosphorylation. This restoration of Akt phosphorylation can be a result of heat shock protein 90 (Hsp 90) binding to the α and β subunits of IKK as well as to endothelial nitric oxide synthase (eNOS) (49, 50). The interaction of Hsp90 and eNOS creates a complex with Akt which allows eNOS and Akt to function on the same domain of Hsp90 (51). This interaction is increased when IKK is inhibited, resulting in increased Akt-eNOS pathway activation (52). Many other studies showed that pharmacological interventions that inhibit NF-κB reduced the multiple organ dysfunction associated with sepsis (32, 36, 53). Indeed, in this study a single dose of IKK-16 at 1 h after CLP resulted in attenuation of the multiple organ dysfunction associated with T2DM/sepsis.

Although we cannot exclude that other effects associated with DPP-4 inhibition (that do not involve NF-κB inhibition) may have contributed to the observed beneficial effects of linagliptin, our data demonstrating that the magnitude of the effect of the inhibition of NF-κB with IKK-16 is similar to the one observed with linagliptin supports the view that inhibition of the activation of NF-κB is at the heart of the observed beneficial effects of both linagliptin and IKK-16. Our data also indicate (and indeed support the view of other publications) that linagliptin may be "repurposed" for the use in sepsis and/or other conditions that are associated with local or systemic inflammation driven by the excessive activation of NF-κB.

# CONCLUSIONS AND LIMITATIONS

Our data shows that a pre-existing, diabetic phenotype worsens the cardiac (organ) dysfunction associated with CLP-sepsis in mice. It also shows that activation of the NF-κB pathway is a key driver of cardiac dysfunction in mice with T2DM/sepsis. Most notably, it shows, for the first time, that inhibition of NFκB using linagliptin or IKK-16 attenuates this cardiac (organ) dysfunction even in mice with pre-existing T2DM. Therefore, targeting NF-κB activation may be a potential strategy to treat the excessive inflammation and cardiac (organ) dysfunction in patients with T2DM and sepsis. However, our study was conducted in relatively young mice (22-week old) and treatment of septic mice was introduced relatively early in the disease course (at 1 h of the induction of sepsis) in mice treated with antibiotics and fluids to mimic "standard care" in humans with sepsis. Clearly, more studies are needed to elucidate how late after the onset of sepsis linagliptin or IKK-16 can be administered to attenuate the cardiac (organ) dysfunction, especially in older mice (ideally 12 to 18-month old). The latter studies are often limited by either the availability of mice of a relevant age or cost of these animals (and often both). This is of particular importance as most cases of sepsis occur in elderly and they are usually diagnosed later in the disease when patients either have already developed multiple organ dysfunctions or, at least, significant cardiovascular abnormalities. Further studies using other DPP-4 inhibitors and/or GLP-1 receptor agonists are needed to investigate whether any of the observed beneficial effects of linagliptin are secondary to the increased incretin levels or are, indeed, a direct effect of DPP-4 inhibition and to determine whether the inhibition of NF-κB reported with linagliptin in our study is, indeed, a unique class-specific effect or an off-target effect of linagliptin.

# AUTHOR CONTRIBUTIONS

SA, JC, MC, MY, and CT conceived and designed the experiment. SA, JC, CM, LM, FC, and DC performed the experiments. SA, MC, CT analyzed the data. SA, MC, and CT contributed to the writing of the manuscript.

# FUNDING

SA is sponsored by Al-Balqa Applied University, Jordan. This work was, in part, funded by the William Harvey Research Foundation and the University of Turin (Ricerca Locale 2017 and 2018).

# ACKNOWLEDGMENTS

Part of this work was presented at the 78th scientific session of the American Diabetes Association (ADA).

# REFERENCES


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

Copyright © 2018 Al Zoubi, Chen, Murphy, Martin, Chiazza, Collotta, Yaqoob, Collino and Thiemermann. 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.

# Effects of Changes in the Levels of Damage-Associated Molecular Patterns Following Continuous Veno–Venous Hemofiltration Therapy on Outcomes in Acute Kidney Injury Patients With Sepsis

Jie Wu<sup>1</sup> , Jianan Ren<sup>1</sup> \*, Qinjie Liu<sup>1</sup> , Qiongyuan Hu<sup>1</sup> , Xiuwen Wu<sup>1</sup> \*, Gefei Wang<sup>1</sup> , Zhiwu Hong<sup>1</sup> , Huajian Ren<sup>2</sup> and Jieshou Li <sup>1</sup>

#### Edited by:

Timothy Robert Billiar, University of Pittsburgh, United States

#### Reviewed by:

Sergio Iván Valdés-Ferrer, Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán (INCMNSZ), Mexico Ben Lu, Central South University, China

#### \*Correspondence:

Jianan Ren jiananr@gmail.com Xiuwen Wu lygwxw@163.com

#### Specialty section:

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

Received: 08 August 2018 Accepted: 10 December 2018 Published: 07 January 2019

#### Citation:

Wu J, Ren J, Liu Q, Hu Q, Wu X, Wang G, Hong Z, Ren H and Li J (2019) Effects of Changes in the Levels of Damage-Associated Molecular Patterns Following Continuous Veno–Venous Hemofiltration Therapy on Outcomes in Acute Kidney Injury Patients With Sepsis. Front. Immunol. 9:3052. doi: 10.3389/fimmu.2018.03052 <sup>1</sup> Department of Surgery, Affiliated Jinling Hospital, Medical School of Nanjing University, Nanjing, China, <sup>2</sup> Department of Surgery, Jinling Hospital, Nanjing Medical University, Nanjing, China

Background: We investigated the association of damage-associated molecular pattern (DAMP) removal with mortality in sepsis patients undergoing continuous veno–venous hemofiltration (CVVH).

Methods: Circulating levels of DAMPs [mitochondrial DNA (mtDNA); nuclear DNA (nDNA); heat shock protein 70 (HSP70); and high mobility group box 1 (HMGB1)] and cytokines were measured at baseline, 6 and 12 h after initiation of CVVH. Urinary DNA levels were analyzed at baseline and end of CVVH. The expression of human leukocyte antigen (HLA)-DR was assayed at 0, 3, and 7 days after initiation of CVVH. Moreover, the effects of HSP70 and HMGB1 clearance on survival were analyzed.

Results: We evaluated 43 patients with acute kidney injury (AKI) (33 sepsis patients). Twenty-two sepsis patients (67%) and three non-sepsis patients (30%) expired (P = 0.046). Significant reductions in the levels of circulating interleukin-6 (P = 0.046) and tumor necrosis factor-α (P = 0.008) were found in the sepsis group. The levels of mtDNA were increased (ND2, P = 0.035; D-loop, P = 0.003), whereas that of HSP70 was reduced (P = 0.000) in all patients during the first 12 h. The levels of DAMPs in the plasma were markedly increased after blood passage from the inlet through the dialyzer in survivor sepsis patients. The clearance rates of HSP70 and HMGB1 were good predictors of mortality [area under the curve (AUC) = 0.937, P = 0.000; AUC = 0.90, P = 0.001, respectively]. The level of HLA-DR was increased in response to higher HSP70 clearance (P = 0.006). Survival was significantly worse in groups with higher clearance rates of HSP70 and HMGB1 than the cut-off value (log-rank test: P = 0.000 for both). Higher HSP70 clearance was a significant independent predictor of mortality (odds ratio = 1.025, 95% confidence interval [CI]: 1.012–1.039, P = 0.000). The urinary nDNA (β-globin) level before CVVH was an independent risk factor for the duration of CVVH in patients with sepsis (sRE = 0.460, 95% CI: 1.720–8.857, P = 0.005).

**35**

Conclusion: CVVH removes inflammatory factors, reduces urinary DAMPs, and removes plasma DAMPs. However, survival decreases in response to higher HSP70 clearance.

Keywords: continuous veno–venous hemofiltration, damage-associated molecular patterns, acute kidney injury, heat shock protein 70, high mobility group box 1, urinary nuclear DNA

#### INTRODUCTION

Sepsis is a life-threatening disease caused by a dysregulated host response to infection (1). In particular, acute kidney injury (AKI) is one of the most common types of organ dysfunction, appearing early in the course of sepsis. Nearly half of patients develop AKI in the intensive care unit and account for approximately half of sepsis-related deaths (2, 3). The high mortality associated with septic AKI may be at least partly explained by an incomplete understanding of its pathophysiology and a delay in diagnosis.

The levels of pro-inflammatory mediators in the serum are increased in patients with AKI, regardless of its cause (4, 5). Proand anti-inflammatory mediators, as well as damage-associated molecular patterns (DAMPs), play important roles in regulating the immunological response that mediates the severity and complications of sepsis (6, 7). DAMPs, also known as alarmins, are constitutively available endogenous molecules released in response to tissue damage and involved in the activation of the immune system. A subset of DAMPs are nuclear or cytosolic molecules, such as mitochondrial DNA (mtDNA), nuclear DNA (nDNA), high mobility group box 1 (HMGB1), and heat shock protein 70 (HSP70). In recent years, Rajaee et al. stated that the definition of interleukin (IL)-1 and IL-33 as DAMPs or cytokines remains controversial (8).

Recently, an increased level of mtDNA in the urine has emerged as a novel non-invasive biomarker for the detection of AKI (9). Moreover, our research has demonstrated the effectiveness of using urinary levels of mtDNA as evidence of renal mitochondrial injury induced by AKI after sepsis (10). However, the role of circulating DAMPs in AKI is controversial (11, 12).

Currently, continuous renal replacement therapy (CRRT) is the main support strategy for AKI, involving several types of treatment. Considerable evidence has suggested that CRRT controls azotemia and fluid balance (13, 14). It has been reported that continuous veno–venous hemofiltration (CVVH) may assist in reducing acute inflammation through the removal of proinflammatory cytokines and signaling molecules, such as the tumor necrosis factor (TNF)-α and various ILs (4, 15). However, the effectiveness of pro-inflammatory molecule clearance is controversial. Several clinical studies have shown that changes in the type of CRRT modes did not reduce mortality compared with the standard mode, even after extensive removal of proinflammatory cytokines in the new mode groups (16, 17).

Studies investigating DAMPs and CRRT are limited to a particular molecule and the relationship between levels of DAMPs in the plasma and complications at a certain point in time (18). Thus far, there is no study reporting the effect of CVVH on the levels of DAMPs and the effects of these changes on the prognosis of patients. Therefore, the present study investigated the effects of CVVH on the circulating and urinary levels of DAMPs in AKI patients and the roles of DAMP clearance on patient outcome.

#### MATERIALS AND METHODS

#### Study Population

The population of this prospective study consisted of 43 patients with AKI requiring CVVH who were admitted to the surgical intensive care unit (SICU) of Jinling Hospital (Nanjing, China). The study was approved by the Ethics Committee of Jinling Hospital, Nanjing and conducted from November 2016–August 2017. Written informed consent was provided by all enrolled patients prior to their participation in the study.

The inclusion criteria were: age >18 years; presence of AKI requiring CVVH; presence of sepsis. The exclusion criteria were: refusal to provide consent; history of chronic kidney disease; expected duration of CVVH <12 h; death within 24 h from initiation of CVVH; ongoing CRRT; and previous renal transplantation. Baseline demographic and clinical data were automatically recorded using a software (Nanjing Haitai Medical Information System, Nanjing, China) or by physicians. For comparison, 10 age- or sex- matched AKI patients without sepsis receiving CVVH were included.

#### Definitions

We defined AKI according to the Kidney Disease: Improving Global Outcomes (KDIGO) criteria (19), corresponding to stage 1 of the KDIGO classification with increased serum creatinine level ≥0.3 mg/dL (≥26.5 µmol/L) within 48 h or increased serum creatinine ≥1.5-fold compared with baseline within 7 days. Sepsis and septic shock were defined according to the Third International Consensus Definitions for Sepsis and Septic Shock (1). Indication of CVVH required at least one of the following criteria: oliguria (urine output <100 mL continued for 6 h after adequate fluid resuscitation), serum creatinine (Scr) >250 µmol/L (2.8 mg/dL), serum potassium concentration >6.5

**Abbreviations:** CVVH, Continuous veno-venous hemofiltration; DAMPs, Damage-Associated Molecular Patterns; mtDNA, mitochondrial DNA; nDNA, nuclear DNA; HSP70, Heat Shock Protein 70; HMGB1, High Mobility Group Box 1; cf-DNA, Cell-free DNA; TNF-α, tumor necrosis factor alpha; IFN-γ, interferon gamma; mHLA-DR, mononuclear human leukocyte antigen-DR; APACHE II, Acute Physiology and Chronic Health Evaluation II; SOFA, Sequential Organ Failure Assessment; ICU, Intensive care unit; NGAL, neutrophil gelatinaseassociated lipocalin; ABC, antibodies bound per cell; IQR, interquartile range; KDIGO, Kidney Disease, Improving Global Outcomes; mCLcytokines, mean of cytokines clearance rate; mCLDAMPs, mean of DAMPs clearance rate; mCDAMPs, mean of circulating DAMPs levels.

mmol/L, severe acidemia (pH < 7.2), and presence of severe fluid overload.

#### CVVH Procedure

All patients were treated with CVVH using the Aquarius hemodialysis system (Baxter International Inc., Chicago, IL, USA). Central venous catheterizations (Quinton-Mahurkar duallumen hemodialysis catheters, Kendall, Tyco Healthcare Group LP, USA) through femoral vein sites were used for vascular access. Administration of CVVH was similar for all patients under the following parameters: blood flow, 100–200 mL/minnum; filter, high flux AV600S (polysulfone, 1.4 m<sup>2</sup> , Fresenius Medical Care); replacement fluid infused at 4 L/h through the pre-dilution route. The daily net ultrafiltration volume was decided by the attending physician. Safety monitoring, including serum electrolyte balance, acid base status, and fluid balance was performed twice daily. Anticoagulation treatment was performed according to the patient's condition. In general, citrate anticoagulation was combined with lowdose heparin to maintain the activated clotting time within the extracorporeal circuit within a desired range (200–250 s). In addition, mechanical ventilation, vasopressor therapy, or any other standard treatment was used regardless of the presence of sepsis, as long as the treatment indications were met.

# Sample Collection and Processing

Collection of blood samples (12 mL) was performed from both the inlet and outlet of the dialyzer at baseline and 6 and 12 h after initiation of CVVH. Of this sample volume, 4 mL were collected in ethylenediamine tetracetate tubes, whereas the remaining 8 mL were stored in promoting coagulating tubes. Ultrafiltrate collections (10 mL) were drawn within the first 2 h. Urine samples were collected at baseline and end of CVVH. All samples were immediately transported to the laboratory and placed on ice. Blood samples were centrifuged at 3,000 × g at 4◦C for 10 min, whereas the ultrafiltrate collections and urine samples were centrifuged at 1,000 × g at 4◦C for 5 min. Supernatants were collected and stored at −80◦C until further analysis. Samples were collected in the first 12 h of CVVH.

#### Isolation and Quantification of mtDNA and nDNA

Free DNA was isolated from 200 µL plasma samples using the QIAamp DNA Blood Mini Kit (Qiagen, Valencia, CA, USA), and 1.75 mL urine samples using an urine DNA isolation kit (NorgenBiotek, Ontario, Canada), DNA was eluted in 100 µL of supplied buffer as previously described (10, 20). Quantitative polymerase chain reaction [qPCR; real-time polymerase chain reaction (PCR)] targeting mitochondrial genes (D-loop and ND2) and nuclear genes (GAPDH and β-globin) was performed to quantify cell-free DNA (cf-DNA) content. The efficiency of all reactions was >98%. All samples were analyzed in triplicate. The standard curve of the quantitative assay was produced through the serially diluted template cloned into a plasmid DNA.

#### Analysis of Cytokines and DAMPs

The different cytokines (IL-1β, IL-6, TNF-α, IFN-γ, IL-10) and HSP70 and urinary neutrophil gelatinase-associated lipocalin (NGAL) were quantified using an enzyme-linked immunosorbent assay (ELISA) kit (R&D System, USA), while the levels of HMGB1 were quantified using a different ELISA kit (SAB, USA), according to the instructions provided by the manufacturer.

Based on the mass conservation principle, the removal rate of cytokines during CVVH was calculated as follows (without considering the concentration of cytokines at baseline):

Ctr = (Ci–Co)/Ci

Clr = Ctr × Qb

Ctr, Total concentration removal rate

Ci, Concentration in the inlet plasma prior to the addition of replacement fluid

Co, Concentration in the outlet plasma

Clr, Clearance rate

Qb, Inlet blood flow rate (mL/min).

#### Quantification of Monocyte Human Leukocyte Antigen-DR (mHLA-DR)

Quantification of mHLA-DR was performed according to the description by Döcke et al. (21). In brief, whole blood was acquired at 0, 3, and 7 days after initiation of CVVH and lysed using red blood cell lysis buffer (KeyGEN BioTECH, Jiangsu, China). Subsequently, they were fixed in 4% paraformaldehyde and incubated with Anti-HLA-DR/Anti-Monocyte Stain (Becton Dickinson, San Jose, CA, USA). Samples were analyzed using a FACScan (Becton Dickinson, San Jose, CA, USA) with a fivecolor upgrade (CyTech, Fremont, CA, USA). Flow files were analyzed in CellQuest Pro (Becton Dickinson, San Jose, CA, USA). Antibodies bound per cell (ABC) were calculated by standardizing HLA-DR geometric mean fluorescence intensity (GMFI) of monocytes to BD Quantibrite-phycoerythrin (PE) beads (Becton Dickinson, San Jose, CA, USA).

#### Statistical Analyses

Results are expressed as the mean ± standard deviation (SD) or median with interquartile range (IQR), as appropriate. Comparison of continuous variables between the two groups was conducted using the Student's t-test or Mann–Whitney U-test depending on Gaussian distribution. Categorical variables were compared between the two groups using the chi-square test and Fisher's exact test, as appropriate. We analyzed all datasets using a two-way repeated-measure ANOVA to examine the effects of CVVH duration. Analyses of receiver operating characteristic (ROC) curves were conducted to test the effectiveness for the prediction of certain outcomes. The optimal cut-off value was defined as the value closest to the Youden Index. Correlation analyses were performed on multiple variables, and the degree of correlation was determined by calculation of the Spearman rank-order coefficients. The Kaplan–Meier method and binary logistic regression analysis was used to determine predictors of mortality. Linear logistic regression analysis was performed to determine independent predictors of CVVH duration. The criterion for statistical significance in all comparisons was P < 0.05. All analyses were performed using the SPSS v21.0 software (IBM, Armonk, NY, USA).

#### RESULTS

#### Demographic and Outcome Parameters of AKI Patients With or Without Sepsis

The study flowchart is shown in **Figure 1**. The samples and clinical data were collected at the specified time points as shown in **Figure 1B**. The consecutive case series included AKI patients who met the criteria for CVVH indication, excluding those with chronic kidney disease. According to the hospital records and laboratory examination, the patients were classified as sepsis-associated AKI or non-sepsis-associated AKI. A total of 43 patients were enrolled in the study. Among them, 33 patients presented with sepsis, whereas 10 AKI patients presented without sepsis (control group). The aim of this classification was to investigate differences in the clearance rate of cytokines or DAMPs during CVVH between different diseases. All patients conformed to the criteria of CVVH and the characteristics of patients are shown in **Table 1**.

AKI patients with sepsis were significantly older compared with those without sepsis (48.9 ± 13.9 vs. 37.3 ± 9.4 years, respectively; P = 0.042) and had markedly lower 24-h urinary output (280 vs. 750 mL, respectively; P = 0.009) at the time of enrollment. At baseline, clinical indices [i.e., Acute Physiology and Chronic Health Evaluation (APACHE) II score, Sequential Organ Failure Assessment (SOFA) score, cytokine levels, and DAMPs levels] were comparable between the two groups.

However, the outcomes were distinct between the two groups. The median duration of CVVH for the treatment of AKI patients with sepsis was longer than that for non-sepsis patients (8 vs. 5 days, respectively; P = 0.042). The median duration of mechanical ventilation was significantly longer in the AKI with sepsis group (12.5 vs. 0 days, respectively; P = 0.005). Moreover, in-hospital mortality was significantly higher in the AKI with sepsis group vs. the AKI without sepsis group (66.7 vs. 30.0%, respectively; P = 0.042) (**Table 1**). Of note, CVVH parameters were similar between the two groups, except for treatment duration.

#### Clinical Indicators and Cytokine Levels During CVVH

We classified the AKI patients with sepsis into two groups, namely "survivors" and "non-survivors" to evaluate clinical data trends (**Figure S1**). As expected, most of the indices of liver and kidney function (e.g., levels of liver enzymes, serum creatinine, and blood urea) were improved during the first 7 days after initiation of CVVH, irrespectively of patient survival. However, in non-survivors, the level of total bilirubin temporarily returned within the normal range but rapidly increased after termination of CVVH. In contrast, the urine output did not recover during CVVH.

FIGURE 1 | Study flow and schedule for samples collection. (A) Study flow. (B) Clinical data was obtained every day during the study period. Urinary DNA (mitochondrial DNA and nuclear DNA) and urinary NGAL were measured at baseline and end of CVVH. The level of Circulatory cytokines and DAMPs were measured from the inlet and outlet of the dialyzer at baseline, 6 and 12 h after CVVH initiation. Levels of mHLA-DR were measured at 0, 3, and 7 days after CVVH initiation. CVVH, continuous veno-venous hemofiltration; NGAL, neutrophil gelatinase-associated lipocalin; DAMPs, damage-associated molecular patterns; mHLA-DR, mononuclear human leukocyte antigen-DR.

TABLE 1 | Clinical characteristics and biochemical variables of the study population.


(Continued)

#### TABLE 1 | Continued


BMI, body mass index; APACHE, Acute Physiology and Chronic Health Evaluation; SOFA, Sepsis-related Organ Failure; CVVH, continuous veno–venous hemofiltration; BUN, blood urea nitrogen; eGFR, estimated glomerular filtration; NGAL, neutrophil gelatinase-associated lipocalin; ALT, Alanine aminotransferase; AST, Aspartate amino transferase; CRP, C-reaction protein; PCT, procalcitonin; RBC, red blood cell; PLT, platelet; APTT, activated partial thromboplastin time; PT, prothrombin time; DAMPs, Damage-Associated Molecular Patterns; mtDNA, mitochondrial DNA; nDNA, nuclear DNA; HSP70, Heat Shock Protein 70; HMGB1, high-mobility group box 1 protein; INR, International Normalized Ratio; LOS, length of stay; ICU, Intensive care unit; IQR, interquartile range; SD, standard deviation.

Normally distributed data are presented as the mean (SD) (analysis of variance); non-normally distributed data are presented as median (IQR) (nonparametric Mann–Whitney U-tests); and categorical variables are presented as n (%) (chi-square test). Data in bold reflected P values < 0.05.

The levels of cytokines were measured from the inlet and outlet of the CVVH dialyzer at baseline, 6, and 12 h after initiation of CVVH to evaluate the effect of CVVH on cytokine removal. Only the levels of IL-6 and TNF-α were decreased in the AKI with sepsis group (P = 0.046 and P = 0.008, respectively) (**Figure 2A**). There was no difference in the levels of all measured cytokines between the inlet and outlet, regardless of the presence or absence of sepsis in AKI patients (**Figure 2B**). Furthermore, the levels of circulating cytokines were similar between the surviving and non-surviving AKI patients with sepsis (**Figure 2C**).

#### Circulating and Urinary Levels of DAMPs During CVVH

The circulating levels of DAMPs, including mtDNA (ND2, D-loop), nDNA [glyceraldehyde 3-phosphate dehydrogenase (GAPDH), β-globin], HSP70, and HMGB1, were measured at the same time points as those used for the measurement of cytokine levels. All plasma samples exhibited detectable levels of DAMPs. In the first 12 h after initiation of CVVH, the levels of mtDNA were increased in all AKI patients, regardless of the presence or absence of sepsis (ND2: P = 0.035; D-loop: P = 0.003; **Figure 3A**, **Figure S2**). The level of HSP70 in the plasma was reduced in both AKI patients with sepsis (P = 0.000) and without sepsis (P = 0.001). However, this trend was particularly apparent in the former group. The levels of nDNA in the plasma, including GAPDH, β-globin, and HMGB1 were unaltered during the first 12 h (**Figure 3A**, **Figure S2**). In brief, the tendency of change in the levels of DAMPs during CVVH was similar between the sepsis and non-sepsis group.

In addition, we compared the levels of DAMPs between the inlet and outlet of the dialyzer at the same time points to determine the effect of CVVH on the rates of DAMP clearance. We did not find differences between the inlet and outlet data in the sepsis or non-sepsis group, except an increase in the level of ND2 after blood passage through the filter in non-sepsis group (inlet: 7.08 ± 0.40 log<sup>10</sup> copies/mL; outlet: 7.24 ± 0.41 log<sup>10</sup> copies/mL; P = 0.008) (**Figure 3B**). We further divided the patients with sepsis into a survivor and non-survivor group. The circulating concentrations of DAMPs, including mtDNA, nDNA, HSP70, and HMGB1, were increased in the survivor group after blood passage through the dialyzer (**Figure 3C**, **Figure S2**). Notably, the levels of HMGB1 were decreased significantly in the non-survivor group (**Figure 3C**).

In our previous study, we demonstrated the effectiveness of using the urinary levels of mtDNA as evidence of renal mitochondrial injury induced by AKI after sepsis. Thus, we measured the urinary levels of NGAL and cf-DNA at baseline and end of CVVH to evaluate renal function. The urinary levels of nDNA and mtDNA were reduced after CVVH (**Figure 4**), unlike that of NGAL (despite an observed increasing trend, P

U-tests. \*P < 0.05, \*\*P < 0.01. IFN, interferon; IL, interleukin; TNF, tumor necrosis factor; CVVH, continuous veno-venous hemofiltration; IQR, interquartile range.

= 0.257), regardless of the presence or absence of sepsis in AKI patients. Consistently, clinical indicators for the evaluation of renal function (i.e., serum creatinine, blood urea nitrogen, and estimated glomerular filtration) were also reduced following treatment with CVVH (**Figure S1**). Therefore, CVVH may (at

#### Relationship Between the Clearance Rate of DAMPs, Level of mHLA-DR, and Outcomes

To further investigate the effect of DAMP removal in AKI patients with sepsis, we calculated their clearance rate and assessed its association with the outcome and immune state. The clearance rate is similar to the meaning of the creatinine clearance rate in the kidney, which represents the net result of the production and clearance rates of the molecules. Thus, its value depends on the integrated effects of production by the cells and clearance by the filter. Higher values indicate high clearance rates vs. production rates.

In the present study, the clearance rates of β-globin, HSP70, and HMGB1 were higher in all non-surviving AKI patients with sepsis (P = 0.006, P = 0.005, and P = 0.000, respectively) (**Figure 5**). Furthermore, we performed analyses of ROC curves to determine predictors of mortality (**Table S1**), including the circulating levels of DAMPs as well as the clearance rates of cytokines and DAMPs. The clearance rates of HSP70, HMGB1, and β-globin were good predictors of mortality (**Figure 6**). Analyses of the area under the curve (AUC) of ROC curves showed that the clearance rate of β-globin was similar to that of the APACHE II score at baseline. Notably, the clearance rates of HSP70 and HMGB1 exhibited a similar prediction efficiency to that of the SOFA score at baseline. Consistent with these findings, the Kaplan–Meier survival analysis revealed that patients with higher HSP70 or HMGB1 clearance rate were associated with significantly higher risk of mortality than those with lower clearance rate (log-rank test: P = 0.000 for both) (**Figure 7**).

In addition, we measured the level of mHLA-DR—an indicator of the immune state—at 0, 3, and 7 days after initiation of CVVH to evaluate the immune state of the patients. We found

least partly) improve renal function.

Student's t-test or Mann–Whitney U-test depending on Gaussian distribution. \*P < 0.05, \*\*P < 0.01, \*\*\*P < 0.001. DAMPs, damage-associated molecular patterns; mtDNA, mitochondrial DNA; nDNA, nuclear DNA; HSP70, heat shock protein 70; HMGB1, high-mobility group box 1 protein; CVVH, continuous veno-venous hemofiltration.

no difference in the level of mHLA-DR expression between AKI patients with sepsis and those without sepsis (median ± IQR) (**Figure 8A**). However, the level of mHLA-DR was increased in AKI patients with sepsis who expired (P = 0.05) (**Figure 8B**). To ascertain the relationship between the clearance rate of DAMPs and the change in the level of mHLA-DR in AKI patients with sepsis, we classified patients into two groups based on the cutoff value of DAMPs (**Figures 8C–E**). Interestingly, the level of mHLA-DR was significantly increased in patients with higher β-globin and HSP70 clearance rates than the cut-off value (P = 0.02 and P = 0.006, respectively). Similarly, there was a tendency toward increase in the level of mHLA-DF in patients with higher HMGB1 clearance rate than the cut-off value (P = 0.074). In addition, the level of HMGB1 at the outlet was negatively related with the level of mHLA-DR (Spearman rank correlation coefficient = −0.512, P = 0.013) (**Figure 8F**).

# Multivariate Logistic Regression Analysis

We performed univariate and multivariate logistic regression analyses to examine other baseline or disease-related factors and evaluate the contribution of the clearance rates of βglobin, HSP70, or HMGB1 to mortality. These analyses revealed that only the clearance rate of HSP70 remained independently associated with high mortality after adjustment for age, APACHE II score, SOFA score, urine output, level of total bilirubin, and coagulation indicators [odds ratio (OR): 1.025; 95% confidence interval (CI): 1.012–1.039; P = 0.000] (**Table 2**).

In addition, we analyzed the contribution of the urinary levels of DAMPs to the duration of CVVH. The urinary level of βglobin was an independent factor for the duration of CVVH even after adjustment for age, APACHE II score, SOFA score, urine output, level of creatinine in the serum, estimated glomerular filtration rate, and level of blood urea (standardization regression coefficient: 0.460; 95% CI: 1.720–8.857, P = 0.005) (**Table 3**).

# DISCUSSION

This was the first study to evaluate the efficiency of CVVH for the removal of DAMPs and the effects of this removal. Our study demonstrated three key findings. Firstly, the newly identified urinary indicators of renal injury (i.e., mtDNA and

FIGURE 4 | Effects of CVVH on levels of urinary mtDNA, nDNA and NGAL. Mitochondrial DNA (ND2, D-loop), nuclear DNA (GAPDH, β-globin) and NGAL in urine were measured at baseline and end of CVVH. Error bars denote the median and interquartile range. P-values indicating differences between the patients of two groups were calculated using Mann–Whitney U-tests. mtDNA, mitochondrial DNA; nDNA nuclear DNA; NGAL, neutrophil gelatinase-associated lipocalin; CVVH, continuous veno-venous hemofiltration.

the mean of β-globin clearance rate and HMGB1 clearance rate, and HSP70 clearance rate, and baseline APCHE II, SOFA score. AUC, area under curve; CI, confidence interval; APACHE, Acute Physiology and Chronic Health Evaluation; SOFA, Sepsis-related Organ Failure; CVVH, continuous veno-venous hemofiltration; HSP70, heat shock protein 70; HMGB1, high-mobility group box 1 protein.

nDNA) recovered after CVVH and the level of urinary nDNA was an independent prognostic factor for the duration of CVVH. Secondly, the efficiency of CVVH for the removal of cytokines and DAMPs was variable. The levels of IL-6, TNFα, and HSP70 decreased within the first 12 h of CVVH in AKI patients with sepsis, whereas the levels of DAMPs at the outlet were temporarily increased after blood passage from the inlet through the dialyzer in survivor AKI patients with sepsis. Finally, the higher clearance rate of DAMPs (especially HSP70) was significantly associated with poor outcomes and immune disorders.

CRRT is undoubtedly beneficial for patients with lethal electrolyte abnormalities, renal dysfunction, or liver dysfunction (15). In our previous study, we demonstrated that the urinary levels of nDNA and mtDNA are novel biomarkers of AKI, and may be used as evidence of renal mitochondrial injury induced by AKI after sepsis (10). Our results revealed that the urinary levels of nDNA and mtDNA decreased after CVVH in all patients, and urinary nDNA (β-globin) was an independent prognostic factor for the duration of CVVH. Furthermore, we observed improvement in the indices of renal and liver function, especially the levels of creatinine in the serum, blood urea, and liver enzymes. Therefore, we infer that CVVH may (at least partly) improve kidney injury in terms of histopathology. Nevertheless, the 24-h urinary volume did not recover during CVVH in patients who eventually expired. Perez-Fernandez et al. also emphasized that a low urine output was associated with poor outcome in AKI patients with sepsis (22). Considering that the urine output was also dependent on the circulatory function, persistently low urine output may be mainly due to disease progression-induced disturbances in microcirculation.

Our previous research revealed that the concentration of cf-DNA increased in patients with severe disease (20). Accordingly, we found that the level of nDNA correlated with prognosis. In addition, in surviving AKI patients with sepsis, the concentrations of DAMPs were markedly increased after blood passage through the inlet of the dialyzer, but declined or recovered prior to the subsequent sampling time point. DAMPs were detected in the filtrate (data not shown), suggesting that clearance through CVVH was reliable. However, in nonsurvivors, the level of DAMPs remained high. These results suggest defects in the elimination of DAMPs in non-survivors.

Although studies reported that the levels of IL-6 and TNF-α decreased with time (6), there is currently no consensus to guide clinicians in terms of controlling the levels of inflammation and DAMPs in AKI patients with sepsis managed through CRRT. This may be attributed to a lack of evidence and deep insight into the effects of CRRT on the human body (23).

In this study, there was no correlation between cytokine clearance and mortality. Chung et al. also showed that highvolume hemofiltration did not reduce mortality compared with standard treatment, despite the extensive removal of proinflammatory cytokines in patients treated with high-volume hemofiltration (17). Park et al. also performed a randomized controlled trial, demonstrating that a high dose of CVVH (80 mL/kg/h) did not improve outcomes in sepsis-associated AKI patients despite its considerable effect on the removal of proinflammatory cytokines (16). Thus, CRRT-induced changes in the levels of cytokines may not be sufficient to influence clinical endpoints. Moreover, in recent years, a new concept termed "dialysis trauma" suggests that dialysis involves microcirculation (24). Our findings suggest that elimination of DAMPs worsened the outcome in AKI patients with sepsis undergoing CVVH. The net outcome is dependent on a balance of the detrimental effects vs. the protective effects of CVVH. This may explain the inadequacy of CRRT-induced changes in the levels of proinflammatory factors to influence clinical outcomes.

Several studies have investigated the roles of DAMPs in critical-care illness and sepsis (6, 25, 26). However, there are no studies addressing the clearance rate of DAMPs in CVVH and its effects on the outcomes and immune homeostasis. In this study, data revealed that there was no difference between patients with or without sepsis. Thus, we excluded the effect of

FIGURE 7 | Kaplan-Meier Survival by cut-off value of clearance rate for the AKI with septic patients. Patients are stratified by cut-off value of HSP70 clearance rate (A) and HMGB1 clearance rate (B). Patients were censored from survival analysis after discharge. HSP70, heat shock protein 70; HMGB1, high-mobility group box 1 protein.

FIGURE 8 | Association between mHLA-DR, outcome and DAMPs clearance. The levels of mHLA-DR were inspected at baseline, 3, 7 day after the CVVH initiation. Tendency of mHLA-DR changes (median ± IQR) was analyzed by repeated measure ANOVA in sepsis and non-sepsis groups (A) or survived and non-survived septic patients (B) The subjects were divided into two groups based on the cut-off value of the clearance rates of β-globin (C), HSP70 (D), and HMGB1 (E) to compare the difference of mHLA-DR fold change on day 7 after CVVH initiation, P-values indicating differences between the patients of two groups were calculated using Mann–Whitney U-tests. (F) Correlations between the levels of mHLA-DR on 7 day and the level of HMGB1 at outlet were determined using Spearman correlation test. CVVH, continuous veno-venous hemofiltration; DAMPs, damage-associated molecular patterns; mHLA-DR, mononuclear human leukocyte antigen-DR; HSP70, heat shock protein 70; HMGB1, high-mobility group box 1 protein.


APACHE, Acute Physiology and Chronic Health Evaluation; SOFA, Sepsis-related Organ Failure; eGFR, estimated glomerular filtration; CRP C, reaction protein; PLT, platelet; APTT, activated partial thromboplastin time.

#### TABLE 3 | Prediction for duration of CVVH.


sRE, standardization regression coefficient; CVVH, continuous veno–venous hemofiltration; APACHE, Acute Physiology and Chronic Health Evaluation; SOFA, Sepsis-related Organ Failure; eGFR, estimated glomerular filtration.

the disease on DAMPs and cytokines during the first 12 h of CVVH. Furthermore, we compared the difference between the non-survivors and survivors in the sepsis group. A higher level of mHLA-DR was linked to higher mortality. In our study, AKI patients with sepsis who expired had markedly higher levels of mHLA-DR than patients in other studies (27, 28). The mHLA-DR is currently the "gold standard" for the identification of immunosuppression and participates in antigen presentation and perpetuation of the inflammatory reaction (29, 30). Hence, a higher level of mHLA-DR may reflect excessive immune activation. Accordingly, we demonstrated that the clearance rates of HSP70 and HMGB1 were higher in AKI patients with sepsis who expired. The clearance rates of HSP70 and HMGB1 increase in parallel with the level of mHLA-DR, indicating that the levels of DAMPs may play an anti-inflammatory role. Consistently, several studies have demonstrated that DAMPs may induce immune suppression and regulate immune response. This has been specifically described for HMGB1 (31, 32) and HSP70 (33, 34). Scholars have demonstrated that the extracellular levels of HSP70 activate and suppress the immune response via the paired receptors sialic acid-binding immunoglobulin-like lectins Siglec-5 and Siglec-14 (34–36). Thus, excessive removal of HSP70 during CVVH may amplify the inflammatory process, disrupt immune homeostasis, and exacerbate disease.

Timmermans et al. reported that, in trauma patients, released DAMPs are associated with an acute, predominantly antiinflammatory response, and a suppressed state of the immune system (28). Schafer et al. also demonstrated that mtDNA may be a link between initial inflammation and subsequent immunosuppression in critically ill patients, probably through a Toll-like receptor-9 pathway (37). Leijte et al. (38) have shown that increased levels of plasma DAMPs were associated with immune suppression and post-operative infections in patients undergoing cytoreductive surgery and hyperthermic intraperitoneal chemotherapy. Our results showed that extensive clearance of DAMPs was associated with an extremely high level of mHLA-DR expression and higher mortality. Collectively, these results imply that DAMPs, especially HSP70 and HMGB1, may act as anti-inflammatory molecules involved in the modulation of the immune system. Excessive removal of these molecules during CVVH may exacerbate disease and amplify the inflammatory process. Consequently, patients may expire due to an exaggerated inflammatory response at the early stage of sepsis.

This study had several shortcomings. Firstly, the patient population was small. Further, studies with larger samples are warranted to confirm the present results. Although the sample size of this study was small, our novel findings are important for the management of AKI patients with sepsis in the clinical setting. Secondly, we did not compare the different modes of RRT (e.g., high dose vs. conventional dose) or the types of filtration membrane. Hence, we were unable to reach a conclusion regarding the preferred type of membrane or mode of CRRT. Nevertheless, the process of CVVH was consistent throughout the study period, avoiding potential confounders. In addition, we did not investigate sepsis patients who were not treated with CVVH. It is not possible to calculate the clearance rate in a non-CVVH group. Moreover, the pathophysiological mechanisms of sepsis in the absence of CVVH are different from those involved in the population of the present study, regardless of disease severity. Consequently, an investigation of differences between the outcomes of sepsis patients with or without CVVH would be meaningless. Thirdly, there were

confounders (e.g., early initiation of antibiotic therapy, source control, and vasopressor support) in our study during CVVH. However, we collected the samples during the first 12 h to avoid the influence of other long-term treatments on the clearance rate. Continuous monitoring of clinical data showed that disease severity was similar during the first day of CVVH. In addition, several studies have suggested that the nDNA level is linked to disease severity (39) and our data showed that the circulating levels of nDNA during the first 12 h of CVVH were almost constant. Thus, these clinical confounders did not influence the conclusions of this study. Fourthly, there was a lack of data regarding the long-term impact of changes in the levels of DAMPs on the levels of mHLA-DR and prognosis.

Despite the aforementioned limitations, our study has important implications for clinicians. Mortality among patients in the intensive care unit remains high, despite the latest advances and innovations in CRRT (17). Most of the researchers and physicians are focused on the role of CRRT in reducing the levels of cytokines. However, despite the extensive removal of pro-inflammatory cytokines in the high-volume hemofiltration (HVHF) groups, CRRT did not reduce mortality compared with standard treatment (16). Our results indicated that extensive removal of DAMPs during CVVH was associated with poor prognosis, which may neutralize the beneficial effect of this treatment. Therefore, it is imperative to develop new filtration membranes–with improved biocompatibility to reduce the stress of circulatory cells–and individualized hemofiltration strategies by monitoring the change in the levels of DAMPs. Collectively, we do not recommend the use of CVVH merely for the removal of pro-inflammatory factors in the treatment of sepsis due to the complex effects of CVVH on DAMPs. In addition, extensive elimination of DAMPs may intensify the immune imbalance in patients with sepsis. This hypothesis is consistent with the recommendations included in the guidelines of the 2016 Surviving Sepsis Campaign (23). An advanced understanding of the host response during CVVH is necessary to optimize hemofiltration, shorten the course of AKI, reduce injury to distant organs, and improve survival.

#### CONCLUSIONS

The urinary levels of mtDNA and nDNA recovered after CVVH, and the level of nDNA was an independent prognostic factor for the duration of CVVH. These findings indicated that CVVH can alleviate kidney injury. However, the efficiency of CVVH for the removal of cytokines and DAMPs was variable and complex. The circulating levels of DAMPs were rapidly and temporarily increased after blood–obtained from surviving AKI patients with sepsis–passaging through the dialyzer. The levels of HSP70 and HMGB1 decreased in AKI patients with sepsis who expired. Moreover, the higher clearance rate of DAMPs, especially HSP70 and HMGB1, was significantly associated with immune disorders and poor prognosis. Thus, we do not recommend the use of CVVH in AKI patients with sepsis merely for the removal of inflammatory mediators without other definitive indications for CVVH.

# DATA AVAILABILITY

The datasets analyzed during the present study are available from the corresponding author on reasonable request.

# AUTHOR'S NOTE

All authors are employees of the Department of Surgery, Jinling Hospital, 305 East Zhongshan Road, Nanjing, 210002, China. An abstract of these data has been accepted as an oral presentation at the 38th Annual Meeting of the Surgical Infection Society.

# AUTHOR CONTRIBUTIONS

JW, JR, and XW conceived the study and interpreted the results. JR designed the study. JW, QL, QH, and ZH acquired the data. JW wrote the manuscript. All authors contributed to the study design and manuscript preparation, and read and approved the final manuscript.

# FUNDING

The study was supported by the National Natural Science Foundation of China (81571881, 81772052) and Key Project of Jiangsu Social Development (BE2016752). Youth Project of the Natural Science Foundation of Jiangsu Province (BK20150559). The funding bodies did not play a role in the design, analysis, or interpretation of data in this study.

# ACKNOWLEDGMENTS

We thank The Charlesworth Group for editorial assistance.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.03052/full#supplementary-material

Figure S1 | Tendency of change in clinical data under CVVH in survived and non-survived septic AKI patients. Twenty-four clinical indexes, including blood routine, liver function, renal function, electrolyte concentration, and coagulation function, etc. were collected every day during the first 7 days after CVVH initiation to form the line chart. The values at "After" denotes the indexes were acquired within 1 day of termination of CVVH treatment. <sup>a</sup>Tendency of change was significant in total patients (P < 0.05). #Tendency was significant different between survived and non-survived sepsis patients (P < 0.05). <sup>1</sup>The difference between before and after CVVH were significant distinct between survived and non-survived groups (P < 0.05). <sup>∗</sup>The values were significant different between survived and non-survived groups after finished CVVH treatment (P < 0.05). BUN, blood urea nitrogen; eGFR, estimated glomerular filtration; ALT, Alanine aminotransferase; AST, Aspartate amino transferase; CRP, C-reaction protein; PCT, procalcitonin; RBC, red blood cell; PLT, platelet; APTT, activated partial thromboplastin time; PT, prothrombin time; INR, International Normalized Ratio.

Figure S2 | Effects of CVVH on D-loop (mtDNA) and GAPDH (nDNA) level and the clearance rate of them. Mitochondrial DNA (D-loop), nuclear DNA (GAPDH) were measured at baseline, 6, 12 h of CVVH at inlet and outlet. (A) Tendency of their levels during the first 12 h was analyzed by repeated measure ANOVA in sepsis and non-sepsis groups, respectively. Error bars of the line chart denote the mean with SD. (B) Box plots shown the levels of DAMPs at inlet and outlet in sepsis and non-sepsis groups. (C) Comparisons of mean levels of mtDNA and nDNA in survived or non-survived septic patients. (D) Box plots shown the levels of clearance rate (median ± IQR) in survived or non-survived septic patients. <sup>∗</sup>P < 0.05, ∗∗P < 0.01, P < 0.001. DAMPs, damage-associated molecular patterns;

#### REFERENCES


mtDNA, mitochondrial DNA; nDNA, nuclear DNA; CVVH, continuous veno-venous hemofiltration; IQR, interquartile range.

Table S1 | Receiver Operating Characteristic curves (ROC) of the clearance of DAMPs and cytokines to predict hospital mortality. Fold front denotes the significance of ROC analysis was <0.05. The cut-off values and corresponding sensitivity and specificity were calculated based on Youden index. IFN, interferon; IL, interleukin; TNF, tumor necrosis factor; DAMPs, Damage-Associated Molecular Patterns; mtDNA, mitochondrial DNA; nDNA, nuclear DNA; HSP70, Heat Shock Protein 70; HMGB1, high-mobility group box 1.


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

Copyright © 2019 Wu, Ren, Liu, Hu, Wu, Wang, Hong, Ren and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(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.

# Forebrain Cholinergic Signaling Regulates Innate Immune Responses and Inflammation

#### Edited by:

*Pietro Ghezzi, Brighton and Sussex Medical School, United Kingdom*

#### Reviewed by:

*Bruno Bonaz, Centre Hospitalier Universitaire de Grenoble, France Egle Solito, Queen Mary University of London, United Kingdom*

#### \*Correspondence:

*Valentin A. Pavlov vpavlov@northwell.edu*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *14 December 2018* Accepted: *05 March 2019* Published: *02 April 2019*

#### Citation:

*Lehner KR, Silverman HA, Addorisio ME, Roy A, Al-Onaizi MA, Levine Y, Olofsson PS, Chavan SS, Gros R, Nathanson NM, Al-Abed Y, Metz CN, Prado VF, Prado MAM, Tracey KJ and Pavlov VA (2019) Forebrain Cholinergic Signaling Regulates Innate Immune Responses and Inflammation. Front. Immunol. 10:585. doi: 10.3389/fimmu.2019.00585* Kurt R. Lehner <sup>1</sup> , Harold A. Silverman1,2, Meghan E. Addorisio<sup>2</sup> , Ashbeel Roy 3,4 , Mohammed A. Al-Onaizi 3,5, Yaakov Levine<sup>6</sup> , Peder S. Olofsson2,7, Sangeeta S. Chavan1,2 , Robert Gros 3,4,8, Neil M. Nathanson<sup>9</sup> , Yousef Al-Abed2,10, Christine N. Metz 1,2 , Vania F. Prado3,4,11,12, Marco A. M. Prado3,4,11,12, Kevin J. Tracey 1,2† and Valentin A. Pavlov 1,2 \* †

*<sup>1</sup> Zucker School of Medicine at Hofstra/Northwell, Hempstead, NY, United States, <sup>2</sup> Center for Biomedical Science and Bioelectronic Medicine, The Feinstein Institute for Medical Research, Northwell Health, Manhasset, NY, United States, <sup>3</sup> Schulich School of Medicine and Dentistry, Robarts Research Institute, University of Western Ontario, London, ON, Canada, <sup>4</sup> Department of Physiology and Pharmacology, Schulich School of Medicine and Dentistry, University of Western Ontario, London, ON, Canada, <sup>5</sup> Department of Anatomy, Faculty of Medicine, Kuwait University, Kuwait City, Kuwait, <sup>6</sup> SetPoint Medical Corporation, Valencia, CA, United States, <sup>7</sup> Department of Medicine, Center for Bioelectronic Medicine, Center for Molecular Medicine, Karolinska Institutet, Karolinska University Hospital, Stockholm, Sweden, <sup>8</sup> Department of Medicine, Schulich School of Medicine and Dentistry, University of Western Ontario, London, ON, Canada, <sup>9</sup> Department of Pharmacology, University of Washington, Seattle, WA, United States, <sup>10</sup> Department of Medicinal Chemistry, Center for Molecular Innovation, The Feinstein Institute for Medical Research, Northwell Health, Manhasset, NY, United States, <sup>11</sup> Department of Anatomy and Cell Biology, Schulich School of Medicine and Dentistry, University of Western Ontario, London, ON, Canada, <sup>12</sup> Graduate Program in Neuroscience, Schulich School of Medicine and Dentistry, University of Western Ontario, London, ON, Canada*

The brain regulates physiological functions integral to survival. However, the insight into brain neuronal regulation of peripheral immune function and the neuromediator systems and pathways involved remains limited. Here, utilizing selective genetic and pharmacological approaches, we studied the role of forebrain cholinergic signaling in the regulation of peripheral immune function and inflammation. Forebrain-selective genetic ablation of acetylcholine release and vagotomy abolished the suppression of serum TNF by the centrally-acting cholinergic drug galantamine in murine endotoxemia. Selective stimulation of acetylcholine action on the M1 muscarinic acetylcholine receptor (M1 mAChR) by central administration of the positive allosteric modulator benzyl quinolone carboxylic acid (BQCA) suppressed serum TNF (TNFα) levels in murine endotoxemia. This effect was recapitulated by peripheral administration of the compound. BQCA also improved survival in murine endotoxemia and these effects were abolished in M1 mAChR knockout (KO) mice. Selective optogenetic stimulation of basal forebrain cholinergic neurons innervating brain regions with abundant M1 mAChR localization reduced serum TNF in endotoxemic mice. These findings reveal that forebrain cholinergic neurons regulate innate immune responses and inflammation, suggesting the

**50**

possibility that in diseases associated with cholinergic dysfunction, including Alzheimer's disease this anti-inflammatory regulation can be impaired. These results also suggest novel anti-inflammatory approaches based on targeting forebrain cholinergic signaling in sepsis and other disorders characterized by immune dysregulation.

Keywords: forebrain cholinergic, cytokines, inflammation, vagus nerve, endotoxemia, sepsis, neural regulation

# INTRODUCTION

The nervous system regulates and coordinates physiological functions and defense mechanisms. A major defense mechanism against pathogen invasion and tissue injury is provided by the innate immune system through inflammation (1). However, dysregulated immune responses and aberrant inflammation are implicated in the etiology of sepsis, inflammatory bowel disease, and many other life-threatening and debilitating disorders (1–6). While accumulating evidence reveals that the nervous system and specifically the vagus nerve regulate immune function and inflammation, the role of brain pathways in this context remains poorly understood (2, 7, 8). The brain regulation of peripheral inflammation and the mediating role of the vagus nerve have been indicated in studies with the experimental anti-inflammatory compound CNI-1493 (Semapimod). Administration of this molecule in the brain suppresses serum TNF (TNF-α) in murine endotoxemia and this effect is abrogated by surgical transection of the vagus nerve (vagotomy) (9). CNI-1493 binds to muscarinic acetylcholine receptors (mAChRs) (10) and administration of compounds that mimic the action of acetylcholine on mAChRs in the brain also suppresses circulating TNF and other proinflammatory cytokines (10–12). In addition, galantamine, an acetylcholinesterase inhibitor, which increases acetylcholine levels suppresses peripheral pro-inflammatory cytokine levels acting through a brain mAChR-mediated mechanism (13– 15). These cholinergic effects in the brain are linked with activation of the vagus nerve-based inflammatory reflex (11, 13, 14, 16, 17), a physiological immunoregulatory circuit (18, 19) with recently demonstrated utility in treating human inflammatory diseases (20, 21). These studies have implicated brain cholinergic signaling in the regulation of pro-inflammatory cytokine release and inflammation. The cholinergic system in the brain has a diverse topographic neuronal organization and projection patterns (22, 23), and specific insight into the role of cholinergic pathways and receptors in the brain in peripheral immunoregulation is presently lacking.

A major collection of cholinergic neurons is localized in the basal forebrain (22, 24). These neurons project to forebrain regions with abundant expression of M1 mAChRs, including neocortical areas and the hippocampus (23), and regulate neuroplasticity, cognition, and other processes (22, 23, 25). Here, utilizing mice with selective genetic forebrain ablation of acetylcholine release, positive allosteric M1 mAChR modulation, and optogenetic stimulation of basal forebrain cholinergic neurons, we indicate a role for forebrain cholinergic signaling via M1 mAChRs in the physiological regulation of inflammation.

#### RESULTS

#### Acetylcholine in Forebrain Mediates Cholinergic Suppression of Peripheral Pro-inflammatory Cytokine Release via a Vagus Nerve-Dependent Signaling

To investigate the role of forebrain neuronal acetylcholine in regulating peripheral inflammation, we utilized mice with selective forebrain deprivation of acetylcholine release and galantamine, a centrally-acting cholinergic drug (an acetylcholinesterase inhibitor) with anti-inflammatory properties (13, 14). A major molecular determinant of acetylcholine release is the vesicular acetylcholine transporter (VAChT), which loads acetylcholine in vesicles prior to its release in the synaptic cleft (26, 27). We used mice with Cre-loxP-based forebrain VAChT ablation, a genetic manipulation that provides a noninvasive means of selective elimination of forebrain acetylcholine release and cholinergic activity without loss of neurons (27). This is important as cholinergic neurons can also secrete GABA with ACh and targeting VAChT allows for selective manipulation of ACh release (28–31). As shown in **Figure 1A**, no VAChT immunoreactivity in the forebrain (hippocampus, CA3 area shown) was detected in these VAChTNkx2.1−Cre−flox/flox (VAChT−/−) mice as compared with VAChTflox/flox (VAChT+/+) control mice. This observation was consistent with the previously reported lack of VAChT immunoreactivity in several forebrain areas of VAChT−/<sup>−</sup> mice and no significant alterations in VAChT protein expression in brainstem regions as a result of genetic deletion (27). The selective forebrain VAChT ablation was not associated with differences in VAChT protein levels (immunofluorescence) in the spinal cord and peripheral neuronal varicosities in the heart (**Figures 1B,C**). These observations indicated that VAChT ablation was limited to the basal forebrain cholinergic system. Furthermore, no differences were observed between VAChT−/<sup>−</sup> and VAChT+/<sup>+</sup> mice on 24 h heart rate recording (**Figure 1D**). There were no differences between the two groups of mice in heart rate responses to atropine (mAChR blocker) and propranolol (beta adrenergic receptor blocker) i.p. administrations (**Figures 1E–G**). In addition, sensitivity to post-handling stress was similar between genotypes, because heart rate responses to saline administration did not differ between the two groups of mice (**Figure S1**).

The effects of acetylcholine in the brain can be modulated (enhanced) through inhibiting its degradation using centrallyacting acetylcholinesterase inhibitors, including galantamine. Administration (i.p.) of galantamine in VAChT+/<sup>+</sup> mice prior to endotoxin significantly reduced serum TNF levels as compared to vehicle administration (P = 0.017) (**Figure 2A**). However,

galantamine failed to significantly alter serum TNF levels in VAChT−/<sup>−</sup> mice (**Figure 2A**). In addition, serum TNF levels in VAChT−/<sup>−</sup> mice were significantly higher than in VAChT+/<sup>+</sup> mice during endotoxemia (P = 0.009) (**Figure 2A**), suggesting that physiological cholinergic transmission in the forebrain regulates peripheral innate immune responses. Galantamine was previously shown to stimulate vagus nerve activity (17, 32). We next examined the role of the vagus nerve in mediating galantamine forebrain-triggered anti-inflammatory effect in endotoxemic C57BL/6 mice. Galantamine administration significantly decreased serum TNF levels as compared with vehicle during endotoxemia in sham-operated (control) mice (P = 0.019) (**Figure 2B**). This effect was markedly diminished in mice with cervical unilateral vagotomy (**Figure 2B**), thus indicating a role of the efferent vagus nerve. Collectively, these data indicate that acetylcholine derived from cholinergic neurons in forebrain plays a role in mediating the suppressive effect of galantamine on peripheral TNF levels in a vagus nerve-dependent manner during murine endotoxemia.

#### Stimulation of Acetylcholine Action on the M1 mAChR by Allosteric Modulation Suppresses Lethal Peripheral Inflammation

Basal forebrain cholinergic neurons project to regions with high expression of M1 mAChR, including several neocortical areas and the hippocampus. Acetylcholine released from these neurons stimulates the postsynaptically located M1 mAChR that processes cholinergic transmission (33). Positive allosteric modulation of the M1 mAChR is a selective approach of increasing endogenous acetylcholine activity at the receptor (34, 35). To study the effect of acetylcholine on brain M1 mAChR in the regulation of

VAChT+/<sup>+</sup> control mice (\**P* = 0.017, \*\**P* = 0.007, two-way ANOVA, Sidak's multiple comparisons test), *n* = 6–7 per group. (B) Vagotomy (VgX) abolishes the suppression of serum TNF in endotoxemic mice (\**P* = 0.019, two-way ANOVA, Sidak's multiple comparisons test), *n* = 6–8 per group. See Methods for details.

peripheral inflammation, we used benzyl quinolone carboxylic acid (BQCA), a positive allosteric modulator of M1 mAChR (34, 35). BQCA has previously been shown to selectively increase (up to 129-fold) the functional affinity of endogenous acetylcholine for M1 mAChR (35). Intracerebroventricular (i.c.v.) injection of BQCA (5 µg/kg) resulted in significant suppression of serum TNF as compared with vehicle administration following endotoxin challenge (P = 0.018) (**Figure 3A**). This observation indicated that selective activation of acetylcholine action on the M1 mAChR in the brain suppresses peripheral inflammation in murine endotoxemia. To facilitate subsequent studies with BQCA, we examined whether peripheral administration of BQCA, which is known to cross the blood brain barrier (34, 35), recapitulates anti-inflammatory effects. Intraperitoneal (i.p.) treatment of mice with BQCA (1–20 mg/kg) resulted in a dose-dependent decrease in serum TNF following endotoxin as compared to vehicle injection (P < 0.001, when 20 mg/kg BQCA was used) (**Figure 3B**). I.p. administration of BQCA also dosedependently improved the survival rate in this lethal murine

Peripheral i.p. administration of BQCA 1–20 (1–20 mg/kg) in endotoxemic mice suppresses serum TNF levels (\**P* = 0.027, \*\*\**P* = 0.0005, \*\*\*\**P* < 0.0001, one-way ANOVA, Dunnett's multiple comparisons test), *n* = 7–8 per group. (C) Peripheral i.p. administration of BQCA 20 (20 mg/kg) in endotoxemic mice improves survival in endotoxemia (\**P* = 0.035, Log-rank test), *n* = 11–12 per group.

endotoxemia model as compared to vehicle administration (P = 0.035 when 20 mg/kg BQCA was used) (**Figure 3C**).

BQCA treatment (20 mg/kg, i.p.) did not significantly alter serum TNF in M1 mAChR KO mice (as compared to vehicle treatment), while in wild type (WT) mice this drug effect was significant (P = 0.028) (**Figure 4A**). Previous studies have shown that activation of brain mAChR-mediated cholinergic signaling results in suppression of TNF in spleen, a major

during endotoxemia (unpaired two-tailed Student's *t* test), *n* = 7, 8. (D) Peripheral (i.p.) administration of BQCA (20 mg/kg, i.p.) improves survival in endotoxemic WT mice and does not alter the survival rate in M1 mAChR KO mice during endotoxemia (\**P* = 0.028, Log-rank test), *n* = 15–18 per group. See Methods for details.

source of pro-inflammatory cytokines and an organ target of the vagus-nerve-based inflammatory reflex (36–38). BQCA (i.p.) administration 1 h prior to endotoxin significantly suppressed splenic TNF in WT mice (P = 0.047) and did not alter splenic TNF in M1 mAChR KO mice (**Figure 4B**). BQCA also failed to significantly alter splenic TNF in endotoxemic WT mice with unilateral cervical vagotomy (**Figure 4C**). In addition, BQCA (20 mg/kg, i.p.) injected 1 h prior to endotoxin was sufficient to significantly improve survival in WT mice (P = 0.028), but failed to significantly alter survival in endotoxemic M1 mAChR KO mice (**Figure 4D**). Together these results show that enhancement of acetylcholine activity on the M1 mAChR is sufficient to suppress inflammation in murine endotoxemia and that the efferent vagus nerve is necessary for this anti-inflammatory effect.

#### Selective Optogenetic Stimulation of Basal Forebrain Medial Septum Cholinergic Neurons Suppresses Serum TNF Levels

We next examined whether the anti-inflammatory effect of acetylcholine acting on M1 mAChR (achieved by allosteric modulation with BQCA) could be replicated by direct stimulation of basal forebrain cholinergic neurons in vivo. The medial septum (medial septal nucleus) is a major nucleus in the basal forebrain cholinergic system (22, 23). The medial septum also plays an important role as a relay of afferent vagus nerve signaling in the forebrain as recently demonstrated (39). Accordingly, we stimulated basal forebrain medial septum cholinergic neurons using a selective optogenetic approach in transgenic mice. These mice express channelrhodopsin-2 coupled to a yellow fluorescent protein (ChR2-YFP) under the control of the choline acetyltransferase (ChAT) promoter. Immunofluorescent staining of brain slices confirmed the neuronal colocalization of ChAT and ChR2-YFP in the medial septum (**Figure 5A**) and the abundant expression of ChR2-YFP in the medial septum and the adjacent vertical limb of the diagonal band of Broca (**Figure S2**). Photoactivation of medial septum neurons by laser light (473 nm) significantly suppressed serum TNF levels compared with sham stimulation (P = 0.039) (**Figure 5B**) in mice with confirmed (by microscopic histochemical evaluation) location of the fiber tip in the medial septum (**Figure S3**; **Supplementary Methods**). Laser light exposure of medial septum neurons in C57BL/6 mice not expressing ChR2-YFP on cholinergic neurons (non-carriers) was performed to control for possible confounding effects of heat and other non-thermal effects of light. This manipulation failed to alter serum TNF levels (**Figure 5C**) thus confirming the specific cholinergic nature of the mechanism. As ChAT-ChR2-YFP mice have been shown to overexpress VAChT (40, 41) and present increased cholinergic tone, LPS was administered

to ChAT-ChR2-YFP and non-carrier mice not subjected to anesthesia and surgical manipulation. No differences in serum TNF levels upon LPS administration were observed between the two groups of mice (**Figure 5D**), indicating that the genetic modification by itself did not alter the inflammatory response. These results show that selective activation of basal forebrain medial septum cholinergic neurons decreases peripheral inflammation in endotoxemia.

# DISCUSSION

Here, we show a role for forebrain cholinergic signaling and the M1 mAChR in the brain neuronal regulation of inflammation through vagus nerve-mediated signaling.

Genetic ablation of VAChT, which eliminates synaptic acetylcholine release from basal forebrain cholinergic neurons (27), abolished the anti-inflammatory effect of the centrallyacting acetylcholinesterase inhibitor galantamine. These findings point to forebrain acetylcholine as a major mediator of galantamine anti-inflammatory effects in endotoxemia. The selective forebrain cholinergic deficit (in VAChT−/<sup>−</sup> mice) also resulted in increased circulatory TNF levels in endotoxemic mice, thus suggesting a tonic anti-inflammatory role of acetylcholine in forebrain.

Forebrain areas, including the cortex (neocortex) and the hippocampus have abundant (predominantly post-synaptic) localizations of the M1 mAChR, which plays a major role in processing the effects of acetylcholine released from basal forebrain cholinergic neurons (22). Positive allosteric modulation of the M1 mAChR by BQCA is a selective approach of increasing the activity of the endogenous acetylcholine on the receptor (34, 35). The anti-inflammatory effects of the centrally-acting BQCA demonstrated here indicate a role of acetylcholine acting on the M1 mAChR in controlling peripheral inflammatory responses. Unilateral vagotomy attenuated the anti-inflammatory effects of galantamine and BQCA, thus indicating a brain-to periphery mediating role for the vagus nerve. The vagus nerve is an important neuroanatomical component of the inflammatory reflex (18) in which sensory and motor vagus nerve signaling regulates cytokine production by immune cells in the spleen, and alleviates inflammation (2, 19). Importantly, recent studies demonstrated that afferent vagus nerve signaling reaches the forebrain (hippocampus and cortex), and identified basal forebrain cholinergic nuclei, including the medial septum as major relay components (39, 42, 43). Together with these findings, our observations suggest a role for the medial septum in a forebrain regulatory hub of the inflammatory reflex. This regulation possibly involves other brain regions through multisynaptic pathways with brainstem nuclei providing peripheral vagus nerve projections. These brain networks remain to be further elucidated.

In addition to positive allosteric modulation of acetylcholine action on M1 mAChR, suppression of serum TNF levels during endotoxemia was also achieved by selective optogenetic stimulation of a subset of basal forebrain cholinergic neurons in the medial septum. Although optogenetic stimulation was performed in animals under isoflurane anesthesia and there is some evidence that relatively high doses of isoflurane can increase blood-brain barrier permeability (44), the proper sham stimulation pointed to the specificity of the effect. Medial septum cholinergic neurons innervate the hippocampus and the parahippocampal region, which are interconnected with several neocortical and subcortical areas (23, 24, 39). Multisynaptic connections between the hippocampus and the hypothalamus have also been described. Although we stereotactically targeted optogenetic stimulation of the medial septum, a spread of stimulation to adjacent cholinergic neurons in the diagonal band of Broca cannot be excluded. Cholinergic neurons in the diagonal band of Broca project to cortical areas and the amygdala (22–24). These neural networks involving cholinergic circuits suggest that multiple points of regulation can be further explored in studying the brain control of peripheral inflammation. Some of these networks have been previously associated with the regulation of autonomic responses, including modulation of vagus nerve activity (2). These previous observations are consistent with our findings that signaling through the vagus nerve mediates brain cholinergic modulation of peripheral inflammation. Our findings have clinical relevance. Galantamine is in clinical use for alleviation of cognitive deterioration in patients with Alzheimer's disease. We have also recently shown that treatment with galantamine alleviates inflammation and insulin resistance in patients with the metabolic syndrome (45). BQCA and other centrally-acting positive allosteric modulators of the M1 mAChR have also been preclinically developed in the search of efficient treatments of Alzheimer's disease and other neurological conditions (34, 35). Our results suggest considering these two types of therapeutics as anti-inflammatory agents. While optogenetics provide a very selective approach for spatiotemporal neuronal control in studying neural circuitries (46), this technology currently has limited therapeutic implications (47– 49). However, medial septum deep brain (electrical) stimulation has been successfully used for improving spatial working memory and cognition in experimental settings of traumatic brain injury (50, 51). Future insights into brain networks triggered by medial septum neuronal stimulation and possibly other regions may further inform implications of deep brain stimulation and noninvasive approaches, including transcranial magnetic stimulation and transcranial direct current stimulation (52–54) in therapeutic anti-inflammatory strategies.

Basal forebrain cholinergic signaling has a documented role in attention, learning and memory, and degeneration of basal forebrain cholinergic neurons, which results in diminished release of acetylcholine in the forebrain, is one of the hallmarks of Alzheimer's disease (23). Intriguingly, increased peripheral TNF and other pro-inflammatory cytokine levels have been found in Alzheimer's disease (55), and peripheral inflammation has been linked to exacerbation of brain pathogenesis and neurodegeneration, affecting cholinergic neurons (56, 57). This may be of importance given that cholinergic neurotransmission is dysfunctional in different types of dementia (58), suggesting a potential mechanism by which patients with Alzheimer's disease and other types of dementia may be more susceptible to distinct types of infection (59).

Peripheral inflammation in sepsis, liver disease, and other inflammatory conditions also is linked with brain inflammation (2, 60) and deterioration in brain function and delirium within the scope of characteristic encephalopathies (2, 61–64). Dysregulation in cholinergic signaling has a documented role in brain derangements, including delirium and encephalopathies in sepsis and other inflammatory conditions (60, 65–67). Together with these previously published findings, our observations suggest a bidirectional relationship between brain cholinergic signaling and peripheral inflammation. In this context targeted pharmacological or device-generated modulation of forebrain cholinergic signaling may have broader therapeutic implications.

#### MATERIALS AND METHODS

#### Animals

Male mice were used in all experiments. BALB/c mice (24– 28 g) were purchased from Taconic. Mice with genetic deletion of VAChT in the forebrain (VAChT**–**/**–** mice) were generated as previously described (27). Briefly, VAChTflox/flox mouse line (crossed for five generations with C57BL/6J) was crossed with the Nkx2.1-Cre mouse line (C57BL/6J-Tg(Nkx2-1-cre)2Sand/J) (Jackson Laboratory, stock No: 008661). Control VAChT+/<sup>+</sup> mice were VAChTflox/flox littermates. ChAT-ChR2-YFP BAC transgenic mice (Jackson Laboratory, stock No: 014546) (68) and non-carrier wild type mice (on the C57BL/6 background) were used in experiments with laser light exposure of the medial septum neurons. M1 mAChR (Chrm1−/−) KO mice (69) were generated on the C57Bl/6 background as previously described (70). M1 mAChR KO and wild type (WT) littermates were received from Charles River. Animals were allowed to acclimate for at least 2 weeks prior to the corresponding experiment. All animals were housed in standard conditions (a 12 h light–dark cycle) with access to regular chow and water.

#### Chemicals

Galantamine (Galantamine hydrobromide) (purity ≥95% by HPLC) was purchased from Calbiochem. Galantamine working solutions were prepared in sterile saline. Benzyl quinolone carboxylic acid (BQCA) (purity: ≥97% (HPLC) was purchased from Enzo Life Sciences. BQCA working solutions were prepared in betacyclodextrin and contained 5% betacyclodextrin in sterile saline (35). Atropine (purity: ≥99% (TLC) and propranolol (purity: ≥98% (TLC) were purchased from Sigma and working solutions prepared in sterile saline.

#### Endotoxemia and Drug Treatment

Endotoxemia in mice was induced by administering LPS (endotoxin, Sigma L4130 O111:B4 for consistency with our previous studies (13, 36, 71, 72) in doses as indicated, injected i.p.) Groups of animals were treated i.p. with galantamine (3 mg/kg) or different doses of BQCA or vehicle (sterile saline as a vehicle in galantamine treatments or 5% betacyclodextrin containing sterile saline as a vehicle in BQCA treatments) 1 h prior to endotoxin administration. Vehicle or BQCA (5 µg/kg) [based on the information available about the effect of the compound on M1 mAChR (34, 35)] was also administered i.c.v. 1 h prior to endotoxin (8 mg/kg, i.p.). In experiments with optogenetic stimulation LPS was injected i.p. 1 h after the end of stimulation or sham stimulation. Animals were euthanized by CO<sup>2</sup> asphyxiation 1.5 h after endotoxin administration, and blood was collected via cardiac puncture (from the right ventricle) for cytokine (TNF) determination. In other sets of survival experiments, groups of mice were treated with BQCA or vehicle (i.p.) 1 h prior to endotoxin (8 mg/kg, i.p.) injection. Mice were monitored for survival twice daily for the first 5 days, and then daily for the remainder of the 14 day experiments.

#### Vagotomy

To avoid potential high lethality associated with bilateral cervical vagotomy, a unilateral cervical vagotomy was performed. Mice were anesthetized by isoflurane inhalation and the right cervical vagus nerve was exposed, ligated with a 4–0 silk suture, and divided. In sham-operated animals, the cervical vagus nerve was visualized, but was neither isolated from the surrounding tissues nor transected. All animals were permitted to recover for 7 days following the surgical procedure and before their inclusion in endotoxemia experiments.

#### Brain Surgical Manipulations and Optogenetic Stimulation

In experiments with optogenetic stimulation, mice were anesthetized with isoflurane and placed on a stereotactic frame (Kopf Instruments). Body temperature was maintained between 36.5 and 37◦C using a feedback controlled rectal thermometer and heating pad. Using aseptic technique, the scalp was incised to expose the skull and the overlying connective tissue was removed. A small (∼500µm) craniotomy was performed on the desired locations for fiber insertion. Mice expressing lightactivated cation ChR2 tagged with a fluorescent protein (ChR2- YFP) under the control of the choline acetyltransferase (ChAT) promoter were used in optogenetic experiments. The optic fiber was inserted slowly over 3 min, targeting the dorsal cholinergic neurons in the medial septum at a location 0.8 mm anterior to Bregma and 0.7 mm lateral to a depth of 3.5 mm at a 10 degree angle below dura, which was opened with a 27 gauge needle. Laser light was delivered via a 200µm diameter fiber (Thorlabs) inserted into the craniotomy. For ChR2 activation, a 473 nm laser (Optoengine, LLC) was used at a power of 10 mW at the tip. The laser was controlled by a waveform generator (Agilent). Stimulation was performed by using square pulses at 20 Hz for 10 min. Sham stimulation was carried out the same way with the exception that following optic fiber insertion, no laser stimulation was performed. Following optogenetic stimulation or sham stimulation, the fiber was removed over 3 min, the craniotomy was covered using paraffin wax, and the animal recovered for 1 h on a heating pad at 37◦C prior to LPS injection. Laser light exposure of medial septum neurons or sham laser light exposure was performed following the same protocol in non-carrier mice.

# Brain Surgical Manipulations and I.C.V. Drug Administration

Craniotomies were performed as described above at a location 0.6 mm posterior and 1.2 mm lateral to Bregma targeting the right lateral ventricle. A 10 µL Hamilton microsyringe in a microinjector pump (UMP3-1, World Precision Instruments) was lowered into the right lateral ventricle over 3 min to a depth of 2.1 mm. Following 5 min of equilibration, BQCA (5 ug/kg) or vehicle was delivered over 3 min. Following injection, the syringe remained in place for 5 min to prevent backflow. Then the needle was slowly removed over 3 min. The craniotomy was covered using paraffin wax and the animal recovered for 1 h on a heating pad at 37◦ C prior to LPS injection.

# Immunohistochemistry and Immunofluorescent Microscopy

Expression of ChR2-EYFP in cholinergic neurons of the medial septum was confirmed with immunofluorescent staining for choline acetyltransferase (ChAT) and ChR2-YFP. Mice underwent intracardiac perfusion with 1 × PBS followed by 4% paraformaldehyde. Brains were harvested and cryoprotected with subsequent incubations of 15% and 30% sucrose. They were then stored in O.C.T. at −20◦C. 20µm sections were collected on gelatin subbed slides (SouthernBiotech) using a cryostat (Leica Microsystems). Slides were rinsed in 1 × PBS followed by washing with 0.25% Triton X-100/1 × PBS (PBT). Blocking was performed with a solution of 10% MeOH, 0.1% bovine serum albumin (BSA), 3% normal donkey serum (NDS), and 0.05% hydrogen peroxide. Slides were then washed in PBT. Cholinergic neurons were visualized using goat anti-ChAT (Millipore, AB144 1:200, dilution) as a primary antibody and specific ChR2-EYFP expression was visualized using rabbit anti-GFP Alexa Fluor 488 (ThermoFisher Scientific, A21311, 1:400 dilution) in a solution of PBT containing 1% NDS. Donkey anti-goat Alexafluor 555 (Thermo Fisher Scientific, A21432, 1:200 dilution) in a solution of 1% NDS/PBT was used to visualize Goat anti-ChAT, and slides were mounted and coverslipped with DAPI-Fluoromount G (Southern BioTech). Images were taken using an Olympus FluoView FV300 Confocal Laser Scanning Microscope.

In experiments with VAChT−/<sup>−</sup> and VAChT+/<sup>+</sup> animals, mice were anesthetized with ketamine (100 mg/kg) and xylazine (25 mg/kg) in 0.9% sodium chloride, and then sacrificed by trans-cardial perfusion. Brains were harvested and placed in 4% paraformaldehyde in 1 × PBS overnight at 4◦C. The brains were isolated and 40µm sections of the tissue were obtained using a vibratome**.** Brain sections were incubated in a blocking solution of 1 × PBS/0.4% Triton X-100 containing 0.1% glycine (wt/vol), 0.1% lysine (wt/vol), 1% BSA (wt/vol), and 1% normal donkey serum (wt/vol). Sections were incubated with anti-VAChT primary antibody (catalog no. 139103; Synaptic Systems) overnight. Sections were then incubated with the Alexa Fluor 488 anti-rabbit secondary antibody (1:1,000; Life Technologies) for 1 h. The nuclei were labeled with Hoechst. Images were acquired using the Zeiss LSM 510 Meta confocal system as previously described (27).

#### Serum Isolation and Cytokine Determination

The blood was allowed to clot for 80 min following collection. It was then centrifuged at 5,000 rpm (1,500 × g) for 10 min, and the supernatant (serum) was collected and stored at −20◦C until analysis. Serum TNF was quantified by using ELISA per the manufacturer's instructions (eBioscience).

#### Electrocardiography (ECG)

Electrocardiograms were recorded using radiotelemeters. The radio frequency transmitters were implanted subcutaneously under anesthesia and ECG recordings were initiated following a minimum recovery period of 7 days post-implantation. Heart rate was continuously measured in awake, freely moving mice over 24 h to obtain baseline recordings. To determine the effect of autonomic blockade, heart rate was recorded for 60 min following administration of atropine (1 mg/kg, i.p.), propranolol (1 mg/kg, i.p.), or atropine + propranolol. All data were collected using the Dataquest A.R.T. software (Transoma Medical). Experiments were performed as previously described (73, 74).

#### Statistical Analysis

GraphPad Prism 6.0 software was used for all statistical analysis. Values are presented as mean ± SEM. One-way or two-way ANOVA, followed by appropriate post-hoc tests for multiple comparisons, and a two-tailed two-sample equal variance Student's t-test were performed to determine statistical significance. The statistical significance of differences between groups of animals in survival experiments was analyzed by Log-rank test. P values equal to or below 0.05 were considered significant.

#### REFERENCES


# DATA AVAILABILITY

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

#### ETHICS STATEMENT

All animal experiments were performed in accordance with the National Institutes of Health Guidelines under protocols approved by the Institutional Animal Care and Use Committee and the Institutional Biosafety Committee of the Feinstein Institute for Medical Research, Northwell Health, Manhasset, NY and the Institutional Animal Care and Use Committee at the University of Western Ontario (Protocols 2018-103 and 2018-104).

#### AUTHOR CONTRIBUTIONS

KRL, MAMP, KJT, and VAP designed research. KRL, MEA, HAS, MA-O, AR, YL, PSO, SSC, RG, VFP, and VAP performed research. NMN, YA-A, RG, VFP, and MAMP contributed reagents, analytic tools and knockout and transgenic mice. KRL, HAS, RG, YA-A, CNM, VFP, MAMP, KJT, and VAP analyzed and interpreted data. KRL, KJT, and VAP wrote the manuscript. NMN, CNM, PSO, YA-A, and MAMP provided additional comments to finalize the paper.

#### ACKNOWLEDGMENTS

This work was supported by the following grants from the National Institute of General Medical Sciences, NIH: RO1GM089807 (to VAP and KJT), and RO1GM057226 (to KJT) and CIHR (MP 93651, 12600, 89919) (to VFP and MAMP).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00585/full#supplementary-material


response during endotoxemia. Proc Natl Acad Sci USA. (2006) 103:5219–23. doi: 10.1073/pnas.0600506103


**Conflict of Interest Statement:** VAP, SSC, and KJT are inventors on patents broadly related to the topic of this paper and have assigned their rights to the Feinstein Institute for Medical Research. YL was employed by SetPoint Medical Corporation (Valencia, CA 91355, USA).

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

Copyright © 2019 Lehner, Silverman, Addorisio, Roy, Al-Onaizi, Levine, Olofsson, Chavan, Gros, Nathanson, Al-Abed, Metz, Prado, Prado, Tracey and Pavlov. 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.

# TLR3 Ligand PolyI:C Prevents Acute Pancreatitis Through the Interferon-β/Interferon-α/β Receptor Signaling Pathway in a Caerulein-Induced Pancreatitis Mouse Model

Chaohao Huang1,2,3, Shengchuan Chen1,2,3, Tan Zhang1,2,3, Dapei Li 2,3, Zhonglin Huang2,3 , Jian Huang<sup>4</sup> , Yanghua Qin<sup>5</sup> , Bicheng Chen<sup>1</sup> , Genhong Cheng2,3, Feng Ma1,2,3 \* and Mengtao Zhou<sup>1</sup> \*

#### Edited by:

Timothy Robert Billiar, University of Pittsburgh, United States

#### Reviewed by:

Hansjörg Hauser, Helmholtz Association of German Research Centers (HZ), Germany Kaifeng Shao, Harvard Medical School, United States

#### \*Correspondence:

Mengtao Zhou zhoumengtao@wmu.edu.cn Feng Ma mf@ism.cams.cn

#### Specialty section:

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

Received: 23 October 2018 Accepted: 16 April 2019 Published: 03 May 2019

#### Citation:

Huang C, Chen S, Zhang T, Li D, Huang Z, Huang J, Qin Y, Chen B, Cheng G, Ma F and Zhou M (2019) TLR3 Ligand PolyI:C Prevents Acute Pancreatitis Through the Interferon-β/Interferon-α/β Receptor Signaling Pathway in a Caerulein-Induced Pancreatitis Mouse Model. Front. Immunol. 10:980. doi: 10.3389/fimmu.2019.00980 <sup>1</sup> Key Laboratory of Diagnosis and Treatment of Severe Hepato-Pancreatic Diseases of Zhejiang Province, Department of Surgery, First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China, <sup>2</sup> Suzhou Institute of Systems Medicine, Peking Union Medical College, Chinese Academy of Medical Sciences, Suzhou, China, <sup>3</sup> Center for Systems Medicine, Institute of Basic Medical Sciences, Peking Union Medical College, Chinese Academy of Medical Sciences, Beijing, China, <sup>4</sup> Department of Emergency, First Affiliated Hospital of Soochow University, Suzhou, China, <sup>5</sup> Department of Laboratory Diagnosis, Changhai Hospital, The Second Military Medical University, Shanghai, China

Acute pancreatitis (AP) is a common and devastating inflammatory disorder of the pancreas. However, there are still no effective treatments available for the disease. Therefore, it is important to discover new therapeutic targets and strategies for better treatment and prognosis of AP patients. Toll-like receptor 3 (TLR3) ligand polyI:C is a double-stranded RNA mimic that can be used as an immune stimulant. Our current study indicates that polyI:C exerted excellent anti-inflammatory effects in a caerulein-induced AP mouse model and taurocholate-induced pancreatic acinar cell line injury model. We found that polyI:C triggers type I interferon (IFN) production and downstream IFN-α/β receptor (IFNAR)-dependent signaling, which play key roles in protecting the pancreas from inflammatory injury. Knockout of IFN-β and IFNAR in mice abolished the preventive effects of polyI:C on caerulein-induced AP symptoms, which include pancreatic edema, neutrophil infiltration, the accumulation of reactive oxygen species (ROS), and inflammatory gene expression. Treating pancreatic acinar 266-6 cells with an IFNAR inhibitor, which blocks the interaction between type I IFN and IFNAR, diminishes the downregulation of oxidative stress by polyI:C. Additionally, a subsequent transcriptome analysis on the role of polyI:C in treating pancreatitis suggested that chemotaxis of neutrophils and the production of ROS were inhibited by polyI:C in the pancreases damaged by caerulein injection. Thus, polyI:C may act as a type I IFN inducer to alleviate AP, and it has the potential to be a promising therapeutic agent used at the early stages of AP.

Keywords: acute pancreatitis, TLR3 ligands, polyI:C, reactive oxygen species, type I interferon, IFN-β, neutrophil infiltration

#### HIGHLIGHTS


#### INTRODUCTION

Acute pancreatitis (AP) is an inflammatory condition of the pancreas, that frequently leads to systemic inflammatory response syndrome (SIRS), multiple organ dysfunction syndromes (MODS), and even death without early intervention (1). The mortality of individuals with AP-MODS exceeds 20%, and the quality of life of those who experience the devastating inflammatory disease is significantly worse than that of the general population (2). Research from many groups including our own have shown that pancreatitis involves pancreas edema, inflammatory cell infiltration, and high levels of serum amylase and lipase (3, 4). Meanwhile, reactive oxygen species (ROS) generated by injured pancreatic acinar cells and infiltrated immune cells are key factors in the progression of pancreatitis (5). Therefore, novel therapeutic strategies or pharmaceutical interventions to decrease the accumulation of ROS, limit local inflammatory damage, and accelerate the recovery of the injured pancreas are needed urgently.

Previous studies on AP have suggested that damage of pancreatic acinar cells results in the upregulation of inflammatory cytokines and chemokines (6). They initiate the inflammatory response and the recruitment of inflammatory cells, which leads to pancreatitis (7). Activated inflammatory cells contribute to subsequent pancreatic injury through the generation of ROS (8). As the ROS accumulate, so does the demand for the anti-superoxide response (9). Heme Oxygenase-1 (HO-1), a well-known anti-oxidant molecule, is always highly expressed during the inflammatory response (10). High levels of HO-1 are an indicator of severe tissue damage and oxidative stress, which require a stronger anti-oxidant reaction (10). Thus, HO-1 could be used as an indicator for measuring the severity of damage and inflammation in AP.

PolyI:C is synthetic double-stranded RNA, which is used as a viral RNA mimic to induce type I IFN and trigger antiviral immunity-based pathways in the host. PolyI:C is considered to be a pathogen-associated molecular pattern (PAMP) because it interacts with Toll-like receptor 3 (TLR3) and activates TLR3 dependent downstream signaling (11, 12). Pattern recognition receptors (PRRs) such as TLR9 and NLRP3 are required for the development of inflammation in AP, and their antagonism could provide a new therapeutic strategy for treating AP (13). Moreover, TLR3 ligand polyI:C treatment significantly decreases the mortality and liver injury caused by injection of lipopolysaccharide (LPS) in the presence of D-galactosamine (D-GalN) in C57BL/6 mice (14), which has driven us to test the anti-inflammatory role of polyI:C during AP progression. As a downstream product of the TLR3-TRIF-TBK1-IRF3 signaling axis, type I IFN has been demonstrated as an immune mediator that also exerts an anti-inflammatory function as well as its antiviral activity (15). IFN-α and IFN-β are key members of the type I IFN family in combating virus infection and regulating immune function, and IFN-β is the initial response of type I IFN produced at the early stages of infection or following PAMP stimulation (16, 17). Subsequently, more IFNβ and IFN-α is produced at the later stages of infection via a positive feedback loop (16–18). It was demonstrated that IFN-β counteracts the overexpressed ICAM-1 in cultured brain-derived microvascular endothelial cells (BMEC) incubated with TNF-α, which leads to a reduction in the adhesion of leukocytes to blood vessels and thus a reduced inflammatory reaction (19). Many other similar observations support the conclusion that IFNβ participates in attenuating the acute inflammatory response because of its regulatory effect (20, 21). It is interesting to investigate whether polyI:C has the potential to limit detrimental and pathological immune responses which lead to tissue damage in a caerulein-induced pancreatitis mouse model, via the IFNβ/IFNAR signaling pathway.

In this study, we have found that polyI:C pretreatment prevents caerulein-induced pancreas edema, neutrophil infiltration, the accumulation of ROS, and inflammatory gene expression in the AP mice models. PolyI:C-triggered IFN-β production and downstream IFNAR signaling activation are required for the suppressive effect of polyI:C in the caeruleininduced AP model. Our study has not only demonstrated the protective role of polyI:C in limiting AP, but has also suggested a potential application of TLR3 ligands in the treatment of AP.

#### MATERIALS AND METHODS

#### Mice and Reagents

Ifnb−/<sup>−</sup> mice and Ifnar1−/<sup>−</sup> mice were gifted from Genhong Cheng Laboratory (University of California, CA, USA). Tlr4−/<sup>−</sup> mice were purchased from Model Animal Research Center (Nanjing, China). Wild-type (WT) C57BL/6 mice were acquired from Vital River Laboratory Animal Technology (Beijing, China). All the mice were maintained in the specific pathogenfree (SPF) environment at Suzhou Institute of Systems Medicine (ISM) under a controlled temperature (25◦C) and a 12-h day/night cycle. Male 8–10-week-old mice were used in all the experimental AP mice models. All mice experiments were undertaken in accordance with the US National Institutes of Health Guide for the Care and Use of Laboratory Animals, with the approval of the Scientific Investigation Board of ISM, Suzhou. Antibodies against HO-1 (#70081) and GAPDH (#5174) were from Cell Signaling Technology (Danvers, MA). NQO1 antibody (#ab2346) and KEAP1 antibody (#ab119403) were from Abcam (Cambridge, MA). Caerulein and IFNAR inhibitor were from MCE (Monmouth Junction, NJ). L-Arginine was from Sigma-Aldrich (St. Louis, MO). PolyI:C was from Thermo Fisher Scientific (Waltham, MA). Recombinant mouse IFN-β was from R&D Systems (Minneapolis, MN). Taurocholate was from SolarBio (Beijing, China).

(Continued)

FIGURE 1 | (C) Histology scores of pancreatitis were evaluated and compared after observing five separate fields. (D) Activities of the serum amylase (left) and lipase (right) from Saline, Caerulein-AP, and PolyI:C+Caerulein-AP WT mice were compared via enzymatic methods. (E) mRNA expression levels of Il1b, Cxcl1, and Cxcl2 genes in the pancreatic tissue from Saline, Caerulein-AP, and PolyI:C+Caerulein-AP WT mice were detected by RT-qPCR and normalized to Rpl32. (F) Neutrophil infiltrations in the pancreases from WT caerulein-induced experimental AP mouse models were measured and compared by MPO staining. Top panel: 100X magnification; bottom panel: 200X magnification. (G) MPO<sup>+</sup> cells were counted and compared after observing five separate fields. (H) Pancreases neutrophils from the Saline, Caerulein-AP, and PolyI:C+Caerulein-AP WT mice were analyzed by flow cytometry. CD11b+Ly6G<sup>+</sup> cells were considered as neutrophils. (I) The percentage of neutrophils from indicated groups were calculated and compared. Data of (B,F,H) are representative of three independent experiments. Data of (C,G) are shown as mean ± SD (n = 5) from one representative experiment. Data of (D,E,I) are shown as mean ± SEM from at least three independent experiments. \*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001, one-way ANOVA test.

# Experimental AP Mice Models

Before the induction of experimental AP, the mice fasted for 8 h and were intraperitoneally injected with 100 µl polyI:C (10 mg/kg) or saline as a control. After 1 or 8 h of polyI:C/saline injection, the mice were intraperitoneally administered 200 µl caerulein (200 µg/kg) 10 times or L-Arginine (2.5 g/kg) twice, 1 h apart. Mice were sacrificed at 24 h post the initial induction of the AP, and the serum and pancreas were collected for further analysis. The pancreases were placed in 4% paraformaldehyde for histological analysis or stored in RNAlater (QIAGEN, Düsseldorf, Germany) for RNA extraction. In addition, fresh pancreases were isolated for analyzing inflammatory immune cell infiltration via flow cytometry or snap-frozen for protein extraction for western blotting analysis.

#### Cell Culture and Flow Cytometry Analysis

The mouse pancreatic acinar 266-6 cell line was purchased from ATCC (Manassas, VA) and cultured in RPMI-1640 medium supplemented with 10% FBS (Gibco, ThermoFisher Scientific) 100 IU/ml penicillin and 100µg/ml streptomycin under the conditions of 37◦C and 5% CO2. To induce pancreatitis in vitro, the 266-6 cells were stimulated with 0.5 mM taurocholate in the presence of polyI:C (1µg/ml) or IFN-β (200 U/ml). Twentyfour hours later, cells were collected for the following Western blot assay, DCFH-DA staining, and flow cytometry analysis. To analyze the infiltrated innate immune cells during caeruleininduced AP, pancreases of mice were immediately harvested and incubated with 1 mg/ml collagenase D (ThermoFisher Scientific, Waltham, MA), and minced into small pieces on ice. Single pancreatic cell suspensions were first stained with Fc blocking antibody, then immune-labeled with fluorochrome-conjugated antibodies in PBS supplemented with 2% heat-inactivated FBS (Gibco, Thermo Fisher Scientific); isotype controls were also included. Antibodies Alexa Fluor 647-conjugated anti-CD11b, PE-conjugated anti-F4/80, and PerCP/Cy5.5-conjugated anti-Ly6G were purchased from BioLegend (San Diego, CA). Flow cytometry analysis was performed on a Life Launch Attune NxT Flow Cytometer (ThermoFisher Scientific, Waltham, MA) after gating the living cells. Data were analyzed using FlowJo software (version 10.0).

#### Pancreas Histological Examination and Neutrophil Immunohistochemistry

Pancreatic tissues were collected 24 h after the induction of AP by caerulein and were fixed in 4% paraformaldehyde in PBS. Paraffin embedded tissues from each mouse were sectioned at 5µm and these were followed by H&E staining. Immunohistochemistry for neutrophil marker MPO (#ab9535, Abcam, San Francisco, CA) was performed on saline or caerulein-induced AP tissues. Briefly, 5µm thick paraffin sections of formalin-fixed paraffin-embedded pancreatic tissue were fixed in dimethylbenzene, quenched with 3% H2O2, and blocked with goat serum. After three washes with PBS, the sections were treated with anti-MPO primary antibody (1:50) overnight. Then, the sections were incubated with horseradish-peroxidase–conjugated secondary antibody for 1 h. Finally, color was developed using DAB as peroxidase substrate, and the slides were counterstained with hematoxylin for bright field microscopy. The degree of pancreatic injury was evaluated by light microscopy in 200X magnification over five separate fields. The severity of pancreatitis was scored mainly based on the description in the previous study (22), and is listed in **Supplementary Table 2**.

# Enzymatic Method Measurement of Serum Amylase and Lipase

Blood from experimental AP mouse models was centrifuged at 4,000 rpm, 1,500 g for 10 min at 4◦C to separate the serum. Serum samples were diluted to the appropriate concentration and incubated with corresponding reagents in kit, then amylase and lipase activities were measured by commercial α-Amylase Assay Kit (C016-1) and Lipase Activity Kit (A054-2), respectively, according to the manufacturer's instructions (Nanjing Jiancheng Bioengineering Institute, Nanjing, China).

#### Intracellular ROS Detection

Intracellular ROS intensity was measured by Reactive Oxygen Species Assay Kit (Beyotime, Shanghai) as described in the manufacturer's instructions. DCFH-DA fluorescence was measured by using Life Launch Attune NxT Flow Cytometer (Thermo Fisher Scientific, Waltham, MA, USA). Images were taken using Nikon Eclipse TI fluorescence microscope (Nikon Corporation, Tokyo). Data were analyzed using FlowJo software (version 10).

#### Protein Extraction and Western Blotting

Snap-frozen pancreatic tissues were homogenized and resuspended in the buffer containing 4% sodium dodecyl sulfate (SDS) and 100 mM Tris-HCl. For immunoblot analysis, pancreatic tissue and 266-6 cells were collected in Triton lysis buffer (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, and 5% glycerol) containing complete protease

FIGURE 2 | Knockout of Ifnb abolishes the protective effects of polyI:C on the caerulein-induced AP mouse model. (A) Histopathological examination of the effect of polyI:C on Ifnb−/<sup>−</sup> caerulein-induced experimental AP mouse models. Top panel: gross observation; middle panel: H&E staining, 100X magnification; bottom panel: H&E staining, 200X magnification. (B) Histology scores of pancreatitis were evaluated and compared after observing five separate fields. (C) Activities of the serum amylase (left) and lipase (right) from Saline, Caerulein-AP, and PolyI:C+Caerulein-AP Ifnb−/<sup>−</sup> mice were compared via enzymatic methods.

(Continued)

FIGURE 2 | (D) mRNA expression levels of Il1b, Cxcl1, and Cxcl2 genes in the pancreatic tissue from Saline, Caerulein-AP, and PolyI:C+Caerulein-AP Ifnb−/<sup>−</sup> mice were detected by RT-qPCR and normalized to Rpl32. (E) Neutrophil infiltrations in the pancreases from Ifnb−/<sup>−</sup> caerulein-induced experimental AP mouse models were measured and compared by MPO staining. Top panel: 100X magnification; bottom panel: 200X magnification. (F) MPO<sup>+</sup> cells were counted and compared after observing five separate fields. (G) Pancreases neutrophils from the Saline, Caerulein-AP, and PolyI:C+Caerulein-AP Ifnb−/<sup>−</sup> mice were analyzed by flow cytometry. CD11b+Ly6G<sup>+</sup> cells were considered as neutrophils. (H) The percentage of neutrophils from indicated groups were calculated and compared. Data of (A,E,G) are representative of three independent experiments. Data of (B,F) are shown as mean ± SD (n = 5) from one representative experiment. Data of (C,D,H) are shown as mean ± SEM from at least three independent experiments. \*p < 0.05, \*\*p < 0.01, n.s., not significant, one-way ANOVA test.

inhibitors (Roche). Protein concentrations of the extracts were measured with a BCA assay (ThermoFisher Scientific) and equalized with the lysis buffer. Equal amounts of the extracts were loaded and subjected to SDS-PAGE, transferred onto PVDF membranes (Millipore), and then blotted with enhanced chemiluminescence (Pierce) or Odyssey Imaging Systems (LI-COR Biosciences).

#### RNA Isolation and Quantitative PCR (qPCR)

RNA was isolated from pancreatic tissue which was stored in RNAlater or 266-6 cells using TRIzol (ThermoFisher Scientific) according to the manufacturer's instructions. Following RNA concentration, quantitation of 1 µg of RNA was used to make cDNA using PrimeScriptTM RT Reagent Kit for RT-PCR (Takara Shuzo Co., Tokyo) according to the manufacturer's instructions. Real time PCR analysis was performed using cDNA in the Roche 480 instrument using SYBR from Toneker Biotech (Suzhou, Jiangsu). The relative mRNA expression level of genes was normalized to the internal control ribosomal protein gene Rpl32 by using 2−11Ct cycle threshold method (23). The primer sequences for qPCR were from the primer bank (24), and sequences are listed in **Supplementary Table 1**.

#### RNA-Sequencing Data Acquisition, Quality Control, and Processing

Total pancreas RNA was extracted from caerulein-induced experimental AP mice models. RNA concentration was quantified using a Qubit 2.0 Fluorometer (Thermo Fisher). The quality of extracted RNA was evaluated using an Agilent Technology 2100 Bioanalyzer. RNA libraries were constructed using a TruSeq Stranded mRNA Sample Prep Kit (Illumina) according to the manufacturer's guidelines. The quantity and quality of the libraries were also assessed by Qubit and Agilent 2100 Bioanalyzer, respectively; their molar concentration was validated by qPCR for library pooling. Libraries were sequenced on the HiSeq X10 using the paired end 2<sup>∗</sup> 150 bp, dual-index format. For RNA-Seq data analysis, Trimmomatic was used to remove Illumina sequencing adapters within raw reads of every sample, trim low quality bases of both read ends (with parameters LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15) and drop one read if its length is <36 bp. Secondly, the clean reads were mapped to mouse mm 10 reference genome with STAR, and the alignment bam files were used as htseq-count (command of python package HTSeq) input to get read counts of genes. Finally, DESeq2 was used to identify DEGs (p ≤ 0.05, FC ≥ 2) based on raw read counts. For DEGs, Ingenuity Pathway Analysis (IPA) and GO biological process were performed by Fisher's exact test, the enrichment p-values of which were corrected by the Bonferroni method. RNA-Seq data have been deposited to Gene Expression Omnibus (GSE119844).

#### Statistical Analysis

The data represent the mean of at least three independent experiments, and error bars represent standard error or standard deviation of the mean. Statistical analysis was performed by unpaired 2-tailed student t test or one-way analysis of variance (ANOVA) followed by Tukey's multiple-comparison tests using GraphPad Prism (version 5; GraphPad Sofware Inc.). p < 0.05 was considered as a statistically significant difference.

#### RESULTS

#### PolyI:C Prevents AP in WT Mice

The protective effect of polyI:C on pancreatitis was evaluated in the caerulein-AP experimental mouse model at 24 h after the first injection of caerulein (**Figure 1A**). Administration with caerulein led to dramatic pathological changes including pancreas edema, inflammatory cell infiltration, and tissue necrosis. However, polyI:C pretreatment prevents these pancreatitis symptoms induced by caerulein injection (**Figures 1B,C**). Consistently, polyI:C pretreatment inhibited the elevation of serum amylase and lipase in the caerulein-AP mouse model (**Figure 1D**). PolyI:C also suppressed the induction of inflammatory cytokine and chemokine gene expression including IL-1β, CXCL1, and CXCL2 in the injured pancreases from caerulein-AP mice (**Figure 1E**). Neutrophils were recruited into the injured pancreases during pancreatitis progression, whereas notably fewer MPO<sup>+</sup> cells were observed in the injured pancreases from caerulein-AP mice pretreated with polyI:C. This suggests polyI:C pretreatment inhibited neutrophil infiltration, a key cause of AP (**Figures 1F,G**). These immunohistochemistry results were verified by flow cytometry assay. Significantly fewer infiltrated CD11b+Ly6G<sup>+</sup> cells (neutrophils) were detected in the pancreases from polyI:C-pretreated AP models (**Figures 1H,I**), while the infiltrated CD11b+F4/80<sup>+</sup> cells (macrophages) were not affected (**Supplementary Figure 1**). These data indicate that polyI:C pretreatment specifically inhibits neutrophil infiltration rather than other immune cells such as macrophages.

Additionally, we checked whether polyI:C pretreatment affects pancreas homeostasis in a non-pathogenic salineinjected mouse model instead of in the AP model (**Supplementary Figure 2A**). No significant differences to pancreas edema, serum amylase and lipase levels, inflammatory cytokine and chemokine gene expression or neutrophil and macrophage infiltration were observed between the saline-pretreated and polyI:C-pretreated groups

(Continued)

FIGURE 3 | (D) mRNA expression levels of Il1b, Cxcl1, and Cxcl2 genes in the pancreatic tissue from Saline, Caerulein-AP, and PolyI:C+Caerulein-AP Ifnar1−/<sup>−</sup> mice were detected by RT-qPCR and normalized to Rpl32. (E) Neutrophil infiltrations in the pancreases from Ifnar1−/<sup>−</sup> caerulein-induced experimental AP mouse models were measured and compared by MPO staining. Top panel: 100X magnification; bottom panel: 200X magnification. (F) MPO<sup>+</sup> cells were counted and compared after observing five separate fields. (G) Pancreases neutrophils from the Saline, Caerulein-AP, and PolyI:C+Caerulein-AP Ifnar1−/<sup>−</sup> mice were analyzed by flow cytometry. CD11b+Ly6G<sup>+</sup> cells were considered as neutrophils. (H) The percentage of neutrophils from indicated groups was calculated and compared. Data of (A,E,G) are representative of three independent experiments. Data of (B,F) are shown as mean ± SD (n = 5) from one representative experiment. Data of (C,D,H) are shown as mean ± SEM from at least three independent experiments. \*p < 0.05, \*\*p < 0.01, n.s., not significant, one-way ANOVA test.

(**Supplementary Figures 2B–G**). These results suggest that polyI:C administration to pancreases is potentially safe.

We also tested the protective effect of polyI:C in L-arginineinduced AP mouse models (**Supplementary Figure 3A**). PolyI:C pretreatment significantly reduced the pancreatic injury caused by L-Arginine injection (**Supplementary Figure 3B**). Consistently, induction of serum amylase and lipase by L-Arginine was also attenuated in the polyI:C-pretreated group (**Supplementary Figure 3C**).

To verify the preventive or therapeutic effect of polyI:C in AP, we tested multiple time windows of polyI:C injection. PolyI:C administration 1 h prior to the induction of AP by caerulein effectively inhibited pancreatic injury (**Supplementary Figures 4A,B**), and also suppressed serum amylase and lipase induction by caerulein (**Supplementary Figure 4C**). However, polyI:C administration after induction of AP by caerulein in WT mice did not show a significant protective effect (data not shown).

Thus, we have found that the TLR3 ligand polyI:C effectively and safely prevents caerulein-induced AP and L-Arginineinduced AP in WT mice.

#### IFN-β/IFNAR Signaling Is Required for the Preventive Effect of PolyI:C on AP

To confirm whether polyI:C-triggered type I IFN production mediated the protective effect of polyI:C in AP mice, we pretreated Ifnb−/<sup>−</sup> mice with polyI:C in the caerulein-induced AP mouse model. PolyI:C did not alleviate the AP symptoms such as pancreas edema, inflammatory cell infiltration, and tissue necrosis in the Ifnb−/<sup>−</sup> mice (**Figures 2A,B**). Consistently, polyI:C did not inhibit the elevation of serum amylase and lipase in the Ifnb−/<sup>−</sup> experimental AP mice (**Figure 2C**). In addition, we found even higher inflammatory cytokine and chemokine gene expression (such as IL-1β and CXCL2) in the polyI:Cpretreated Ifnb−/<sup>−</sup> AP mice, whereas polyI:C inhibited those inflammatory genes significantly in the WT AP mice (**Figures 1E**, **2D**). The number of infiltrated neutrophils in Ifnb−/<sup>−</sup> AP mice was indistinguishable between the saline and polyI:C pretreated groups (**Figures 2E–H**). These results indicate that polyI:Ctriggered IFN-β production potentially meditates the preventive effect of polyI:C on AP.

Next, we used the Ifnar1−/<sup>−</sup> mice to confirm the above conclusion. The protective function of polyI:C was not seen in Ifnar1−/<sup>−</sup> AP mice. Similar to the phenotypes observed in the Ifnb−/<sup>−</sup> AP mice, there was no significant changes in pancreas histopathological results (**Figures 3A,B**), serum amylase and lipase activity levels (**Figure 3C**), pancreas inflammatory genes (**Figure 3D**), and neutrophil infiltration (**Figures 3E–H**) between the saline and polyI:C pretreated groups. In summary, these experiments suggest that IFNβ/IFNAR signaling is necessary for the preventive effect of polyI:C on caerulein-induced AP.

#### PolyI:C Inhibits Oxidative Stress in AP in an IFN-β/IFNAR-Dependent Manner

The accumulation of ROS drives persistent tissue damage, resulting in the death of acinar cells, edema formation, and infiltration of inflammatory cells into the pancreas (25). Therefore, we tested whether polyI:C could inhibit the generation of ROS in oxidative injury-induced pancreatitis. As shown in **Figure 4A**, polyI:C treatment significantly suppressed the induction of anti-oxidant protein HO-1 by taurocholate in the 266-6 cells, however it did not suppress other anti-oxidant proteins such as KEAP1 and NQO1 (**Supplementary Figure 5**). Further experiments confirmed the role of the IFN-β/IFNAR signaling pathway in modulating ROS generation. PolyI:C or IFN-β treatment downregulated the induction of HO-1 protein by taurocholate, and blocking IFN-β and IFNAR interaction using an IFNAR inhibitor reversed the effect of polyI:C and IFN-β on HO-1 induction (**Figure 4B**).

HO-1 is not only an anti-oxidant protein for reducing ROS but can also be used as an indicator for measuring the severity of damage and inflammation in AP. Downregulation of HO-1 by polyI:C suggests the inhibition of oxidative stress. As we expected, polyI:C or IFN-β treatment significantly reduced the ROS level in taurocholate-stimulated 266-6 cells according to the fluorescence microscopy and flow cytometry results (**Figures 4C,D**).

Furthermore, we confirmed that polyI:C inhibited the generation of ROS in the pancreases of AP mice via the IFN-β/IFNAR signaling pathway. Homogenates of pancreases resected from experimental AP models with or without polyI:C in WT, Ifnb−/−, and Ifnar1−/<sup>−</sup> mice were analyzed by immunoblotting for HO-1 protein levels. PolyI:C pretreatment suppressed HO-1 induction in the WT AP mice (**Figure 4E**). However, even higher HO-1 protein levels were detected in the polyI:C-pretreated Ifnb−/<sup>−</sup> and Ifnar−/<sup>−</sup> AP mice (**Figures 4F,G**). These data suggest that IFN-β and IFNAR downstream signaling inhibits oxidative stress during AP progression, which could be indicated by higher HO-1 production. Our results suggest that polyI:C inhibits ROS generation independently of HO-1. Without IFN-β/IFNAR signaling, polyI:C cannot facilitate the clearance of ROS, and thus, the pancreas requires more HO-1 anti-oxidant protein. TLR4 was reported to be suppressed by the TLR3 ligand polyI:C and thus, alleviates liver injury induced by LPS and D-GalN (14).

FIGURE 4 | (E–G) HO-1 protein expression levels in the pancreases from Saline, Caerulein-AP, and PolyI:C+Caerulein-A WT (E), Ifnb−/<sup>−</sup> (F), and Ifnar1−/<sup>−</sup> mice (G) were measured by western blotting; a non-specific (NS) band was shown as a loading control. Data of (A–C,E–G) are representative of three independent experiments. Data of (D), right panel are shown as mean ± SEM (n = 3). \*p < 0.05, one-way ANOVA test.

However, in the caerulein-induced AP model, polyI:C protected the pancreas of AP mice from injury and inhibited the induction of amylase and lipase, although it was not as effective as in WT mice (**Supplementary Figure 6**).

By analyzing ROS levels and the amount of the oxidative stress indicator protein in vitro and in vivo, we have confirmed that polyI:C inhibits oxidative stress during AP progression in an IFN-β/IFNAR-dependent manner.

#### PolyI:C Inhibits Multiple Genes Which Positively Correlated With AP Progression

To explore the mechanism of action of polyI:C in preventing AP, we performed RNA-sequencing and analyzed the pancreas' transcriptomes from the saline-pretreated, AP, and polyI:Cpretreated AP mice. We focused on the genes that were significantly induced during AP and suppressed by polyI:C pretreatment. Among them were the inflammatory cytokine gene Il1β, the inflammatory chemokine genes Cxcl1 and Cxcl2, and other genes, such as Ccl2, Ccr2, C5a1, Mrc1, Ccr5, Hck, Tyrobp, Procr, and Fgr, related to neutrophil recruitment, neutrophil movement, the immune response of neutrophils, and ROS production (**Figure 5A**). We verified some of the genes we were interested in and further analyzed all the differently expressed genes using IPA software (**Figures 5B**, **6**). Key pathways and related genes that facilitated AP progression and were suppressed by polyI:C are shown in heatmaps (**Figure 6**).

In summary, our study has shown that polyI:C triggers IFN-β production by activating the master transcription factor IRF3, and IFN-β inhibits neutrophil infiltration, thus, alleviating AP symptoms such as edema, release of amylase and lipase, generation of ROS, and the induction of inflammatory genes by activating downstream IFNAR signaling (**Figure 7**).

#### DISCUSSION

It has been well-established that type I IFN has a protective effect against viral infections by activating IFNAR downstream signaling and inducing IFN-stimulated genes (ISGs) (26). However, it is debated whether type I IFN plays beneficial or detrimental roles in inflammatory diseases (22). Unlike other TLRs such as TLR2, TLR4, and TLR9 usually promote inflammation (24), and TLR3 seems to act as the "peacemaker" in the maintenance of a healthy internal environment. It has been shown that TLR3 or TLR7 sense viral infection in the gut and trigger the production of IFN-β that dampens DSSinduced experimental colitis (27). Here, we have described an anti-inflammatory role of TLR3 in AP. We have found that pretreatment with TLR3 ligand polyI:C protects the caeruleininduced AP mice in an IFN-β/IFNAR-dependent manner.

Compared to the recombinant type I IFN, viral dsRNA mimic polyI:C is much more stable, easier to deliver, and has a lower cost. The use of TLR3 agonists as immunotherapeutic agents has been employed in cancer therapy to induce tumor cell apoptosis in type I IFN-dependent and independent pathways (28, 29). Pretreatment of polyI:C is very effective in preventing caeruleininduced AP and L-arginine-induced AP. In addition, polyI:C is safe for use in mice pancreases according to the non-pathogenic saline-injected mouse model. Given the efficacy and safety of polyI:C in protecting AP mice, it has shown the great potential of the application of TLR3 agonists to prevent AP.

In the AP models we used in this study, IFN-β was found to be responsible for most of the preventive effects of polyI:C on AP, whereas IFN-α seems to play a minor role in this process. Type I IFN is secreted from macrophages during AP progression, and IFN-β is predominantly secreted from macrophages, endothelial cells, and epithelial cells (30). Large amounts of IFN-α protein is produced by plasmacytoid dendritic cells under certain conditions (26). IFN-β is primarily produced during AP and induces the secondary wave of IFN-α via the well-established type I IFN positive feedback regulation loop (16– 18). Most IFN-α genes are ISGs (31). Both IFN-β and IFN-α activate downstream signaling via the same receptor, IFNAR. However, the affinities of IFN-β and IFN-α for the receptor are different. IFN-β and IFN-α also induce the expression of different genes, which are expressed in macrophages stimulated with the same concentration of IFN-β and IFN-α (32, 33). In this study, we identified that IFN-β-induced genes play key roles in preventing AP.

Between 8 h and at least 1-h pretreatment of polyI:C before infection with caerulein or L-arginine is required for effective prevention of AP in the experimental mouse models. PolyI:C did not exert a good therapeutic effect on AP mice if polyI:C is injected post AP induction, since it takes over 6 h for polyI:C to induce the production of type I IFN. However, the level of serum amylase in human AP patients usually reaches its peak at around 48 h and returns to normal levels over the next 5–7 days (34), which is a much wider time window to administer polyI:C to human AP patients. Therefore, it is possible to use TLR3 agonists such as polyI:C as novel AP therapeutic agents if we treat AP patients with polyI:C at the early stages of disease.

It has previously been reported that polyI:C treatment significantly decreases mortality and liver injury caused by injection of LPS in the presence of D-GalN (14). PolyI:C also shows therapeutic effect against cerebral ischemia/reperfusion injury through the downregulation of TLR4 signaling (35). However, our results show that polyI:C protects Tlr4−/<sup>−</sup> AP mice from pancreatic injury and inhibits the induction of amylase and lipase, although it is most effective in WT mice. In addition to IFN-β and downstream IFNAR signaling, it is possible that other polyI:C-responsive genes such as TLR4 also partially mediate the preventive effect of polyI:C on AP. Our future studies will aim to identify the ISGs stimulated by polyI:C that prevent AP.

FIGURE 5 | PolyI:C suppresses multiple genes related to the progression of AP. (A) Gene expression profiles of the pancreases from Saline, Caerulein-AP, and PolyI:C+Caerulein-AP WT mice were detected by RNA-Seq. Top polyI:C-suppressed and caerulein-inducible DEGs (differentially expressed genes) are shown in the heatmap. (B) Samples are treated as described in (A), and mRNA expression level of the genes related to the progression of AP such as Ccl2, Ccr2, C5ar1, Mrc1, Ccr5, Hck, Tyrobp, Procr, and Fgr were verified by qPCR. Data of (A) are representative of three independent experiments. Data of (B) are shown as mean ± SEM (n = 3). \*p < 0.05, \*\*p < 0.01, one-way ANOVA test.

FIGURE 6 | PolyI:C suppresses ROS production and neutrophil functions. (A) Gene expression profile of the pancreases from Saline, Caerulein-AP, and PolyI:C+Caerulein-AP WT mice was detected by RNA-Seq. DEGs were analyzed by the IPA and GO, and clustered into the pathways related to the disease and bio functions. (B) Genes related to the six pathways on ROS production and neutrophil functions are shown.

PolyI:C treatment inhibited the induction of the inflammatory cytokine IL-1β as well as the chemokine genes including CXCL1 and CXCL2 in pancreases from AP mice. The infiltration of neutrophils but not of macrophages to the injured pancreases of AP mice was suppressed. In addition, a lower production of ROS was detected in both the acinar cell line and the injured pancreases of AP mice pretreated with polyI:C. This is consistent with the reduced induction of anti-oxidant protein HO-1 during AP progression. HO-1 usually acts as a protective effector to clear the elevated ROS in inflammatory cells (36). Suppressed generation of ROS by polyI:C leads to a weak induction of HO-1, which suggests polyI:C-triggered type I IFN production protects acinar cells and mice pancreases in a HO-1-independent manner. Pretreatment with polyI:C enhanced the induction of HO-1 protein production in pancreases from Ifnb−/<sup>−</sup> and Ifnar1−/<sup>−</sup> AP mice. This is consistent with the results that polyI:C did not protect the Ifnb−/<sup>−</sup> and Ifnar1−/<sup>−</sup> AP mice. ROS drives cytochrome C release, leading to ATP-dependent caspase activation and apoptosis which occurs in many cell types (37). Thus, there is a need to identify strategies to clear accumulated ROS during oxidative stress in AP. For example, type III IFN member IFN-λ regulated the activation of AKT in a non-translational manner independent of the STAT pathway to diminish ROS production (10). We have outlined a novel HO-1 independent pathway to clear ROS using type I IFN during AP progression.

AP is characterized by the activation of exocrine zymogen granules, which contain digestive enzymes (38), the infiltration of macrophage and neutrophils (39), and necrosis or apoptosis of pancreatic cells (40). Despite damaged primary pancreatic acinar cells, neutrophils isolated from patients with pancreatitis produce ROS to a greater extent (8, 41). Crosstalk among damaged cells, neutrophils, and ROS appears to synergistically promote the process of pancreatitis. According to the transcriptomic profiles of polyI:C-induced genes, we found that these genes were involved in the recruitment of inflammatory cells, production of ROS, and regulation of the inflammatory response. PolyI:C attenuated the positive feedback regulatory loop between ROS and proinflammatory cytokines in AP.

Inflammation occurring in the pancreas triggers multiple processes by recruiting inflammatory neutrophils and increasing ROS production, which leads to a homeostatic imbalance and sequentially promotes increasingly severe inflammatory injury. Regarding the protective role of polyI:C in preventing AP, TLR3 agonists are promising therapeutic agents to safely and effectively prevent AP at the early stages of pancreatitis.

#### ETHICS STATEMENT

All animal procedures were approved by the Institutional Animal Care and Use Committee of Suzhou Institute of Systems Medicine.

# AUTHOR CONTRIBUTIONS

FM and MZ conceived the idea. FM and CH designed the experiments. CH, SC, TZ, DL, ZH, JH, YQ, and BC finished all the experiments and data analysis. FM and CH wrote the manuscript. All authors contributed to the interpretation of the experiments, critically reviewed the manuscript, and gave the final approval of the manuscript submission.

#### FUNDING

This work was supported by the NFSC (81570583, 81770630, 81471606, 31670883, 31771560, 31800760, and 31870912), National Thousand Youths Talents Program to FM, CAMS Initiative for Innovative Medicine (2016-I2M-1-005), Non-profit Central Research Institute Fund of CAMS (2016ZX310189, 2016ZX310194, and 2017NL31004), NSF of Jiangsu Province (BK20170408), and the Shanghai Pujiang Program (16PJD001), Key Laboratory of Diagnosis and Treatment of Severe Hepato-Pancreatic Diseases of Zhejiang Province, Province and Ministry (WKJ-ZJ-1706), New Century Talents Project to MZ, Zhejiang Medical Support Discipline-General Surgery and the Technology

#### REFERENCES


Innovation Team of Diagnosis and Innovative Research Groups of the General Surgery of Wenzhou (No. C20150003).

#### ACKNOWLEDGMENTS

We appreciate the technical support from the RNA technology platform of Suzhou Institute of Systems Medicine.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00980/full#supplementary-material


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

Copyright © 2019 Huang, Chen, Zhang, Li, Huang, Huang, Qin, Chen, Cheng, Ma and Zhou. 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.

# Identification of LPS-Activated Endothelial Subpopulations With Distinct Inflammatory Phenotypes and Regulatory Signaling Mechanisms

Erna-Zulaikha Dayang<sup>1</sup> , Josée Plantinga<sup>1</sup> , Bram ter Ellen<sup>1</sup> , Matijs van Meurs 1,2 , Grietje Molema1† and Jill Moser 1,2 \* †

*<sup>1</sup> Medical Biology Section, Department of Pathology and Medical Biology, University Medical Center Groningen, University of Groningen, Groningen, Netherlands, <sup>2</sup> Department of Critical Care, University Medical Center Groningen, University of Groningen, Groningen, Netherlands*

#### Edited by:

*Timothy Robert Billiar, University of Pittsburgh, United States*

#### Reviewed by:

*Michael Hickey, Monash University, Australia Jaya Talreja, Wayne State University, United States*

\*Correspondence:

*Jill Moser j.moser@umcg.nl*

*†These authors share senior authorship*

#### Specialty section:

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

Received: *28 January 2019* Accepted: *08 May 2019* Published: *24 May 2019*

#### Citation:

*Dayang E-Z, Plantinga J, ter Ellen B, van Meurs M, Molema G and Moser J (2019) Identification of LPS-Activated Endothelial Subpopulations With Distinct Inflammatory Phenotypes and Regulatory Signaling Mechanisms. Front. Immunol. 10:1169. doi: 10.3389/fimmu.2019.01169* Sepsis is a life-threatening condition caused by a dysregulated host response to infection. Endothelial cells (EC) are actively involved in sepsis-associated (micro)vascular disturbances and subsequent organ dysfunction. Lipopolysaccharide (LPS), a Gram-negative bacterial product, can activate EC leading to the expression of pro-inflammatory molecules. This process is molecularly regulated by specific receptors and distinct, yet poorly understood intracellular signaling pathways. LPS-induced expression of endothelial adhesion molecules E-selectin and VCAM-1 in mice was previously shown to be organ- and microvascular-specific. Here we report that also within renal microvascular beds the endothelium expresses different extents of E-selectin and VCAM-1. This heterogeneity was recapitulated *in vitro* in LPS-activated human umbilical vein EC (HUVEC). Within 2 h after LPS exposure, four distinct HUVEC subpopulations were visible by flow cytometric analysis detecting E-selectin and VCAM-1 protein. These encompassed E-selectin−/VCAM-1<sup>−</sup> (–/–), E-selectin+/VCAM-1<sup>−</sup> (E-sel+), E-selectin+/VCAM-1<sup>+</sup> (+/+), and E-selectin−/VCAM-1<sup>+</sup> (VCAM-1+) subpopulations. The formation of subpopulations was a common response of endothelial cells to LPS challenge. Using fluorescence-activated cell sorting (FACS) we demonstrated that the +/+ subpopulation also expressed the highest levels of inflammatory cytokines and chemokines. The differences in responsiveness of EC subpopulations could not be explained by differential expression of LPS receptors TLR4 and RIG-I. Functional studies, however, demonstrated that the formation of the E-sel+ subpopulation was mainly TLR4-mediated, while the formation of the +/+ subpopulation was mediated by both TLR4 and RIG-I. Pharmacological blockade of NF-κB and p38 MAPK furthermore revealed a prominent role of their signaling cascades in E-sel+ and +/+ subpopulation formation. In contrast, the VCAM-1+ subpopulation was not controlled by any of these signaling pathways. Noteworthy is the existence of a "quiescent" subpopulation that was devoid of the two adhesion molecules and did not express cytokines or chemokines despite LPS exposure. Summarizing, our findings suggest that LPS activates different signaling mechanisms in EC that drive heterogeneous expression of EC inflammatory molecules. Further characterization of the signaling pathways involved will enhance our understanding of endothelial heterogeneous responses to sepsis related stimuli and enable the future design of effective therapeutic strategies to interfere in these processes to counteract sepsis-associated organ dysfunction.

Keywords: endothelial cells, HUVEC, lipopolysaccharide (LPS), adhesion molecules, endothelial heterogeneity, intracellular signaling

#### INTRODUCTION

Sepsis is defined as a life-threatening organ dysfunction caused by a dysregulated host response to infection (1). Despite diagnostic advances in early recognition, sepsis often escalates to multiple organ dysfunction syndrome (MODS), leading to poor outcome. Sepsis affects around 30 million people yearly worldwide, resulting in 6 million deaths (2). Although several mechanisms have been suggested to contribute to the pathophysiology of sepsis, it is still incompletely understood, thereby hindering the development of successful treatment options for patients with sepsis and sepsis-induced multiple organ dysfunction syndrome.

Endothelial cells (EC) line all blood vessels and are one of the first cells to respond to microbial product such as Lipopolysaccharide (LPS) (3). During sepsis, EC respond to LPS via the activation of pattern recognition receptors (PRRs), producing pro-inflammatory cytokines, chemokines, and adhesion molecules (4). Adhesion molecules such as Eselectin, VCAM-1, and ICAM-1 facilitate endothelial cell leukocyte interactions, resulting in leukocyte recruitment into inflamed tissues leading to subsequent impairment of organ function (5). Therapeutic strategies designed to attenuate EC production of these inflammation associated molecules to prevent leukocyte adhesion and transmigration, may lead to diminished organ failure in patients with sepsis.

Previous studies from our laboratory have shown that in mice, LPS or TNF-α exposure results in organ and microvascular compartment specific expression of adhesion molecules (6–8). Within the kidney, at the microvascular compartment level, TNF-α induced expression of E-selectin is abundant in glomeruli, but low in the arterioles. In contrast, VCAM-1 is highly expressed in the arterioles, while expressed to a lesser extent in the glomeruli (6, 7). Our knowledge on inter- and intra-organ endothelial heterogeneity has expanded in recent years, yet how these heterogeneous responses are molecularly controlled following LPS challenge still remains elusive.

Toll-Like Receptor 4 (TLR4) (9) and the more recently identified retinoic acid inducible gene-I (RIG-I) (10) are pattern recognition receptors which have been shown to drive endothelial responses to LPS. The recognition of LPS via these receptors results in the induction of endothelial adhesion molecules, cytokines, and chemokines via several downstream signaling pathways, that include NF-κB (10–12), and p38 MAPK (12, 13). Although both NF-κB and p38 MAPK have been implicated in controlling LPS-mediated inflammatory responses, it is unknown whether one, the other, both, or other signaling pathways regulate individual EC responses at a cellular level. Likewise, and of interest in relation to the microvascular compartment specific responses of the endothelium to LPS challenge in vivo, whether all EC are equipped with similar downstream signaling machinery remains to be elucidated.

Here, we investigate the molecular control of the early stages of EC activation by LPS in relation to the previously reported heterogeneous EC responses found in vivo. We focused on LPS induced expression patterns of E-selectin and VCAM-1 within mouse renal microvascular compartments in vivo and their patterns in HUVEC in vitro using immunofluorescent staining and flow cytometry, respectively. Based on the outcome we next investigated in vitro whether different stages of cell division could explain the observed heterogeneous subpopulations formed, and whether these subpopulations behaved differently because of differential expression of LPS signaling machinery components. Additionally, we used pharmacological tools to examine the activity of different kinase signaling pathways to explain heterogeneous subpopulation formation in HUVEC upon LPS exposure.

#### MATERIALS AND METHODS

#### Mice

Male C57BL/6 mice purchased from Envigo (Horst, The Netherlands) and housed in a specific pathogen-free facility, maintained on chow and water ad libitum, and housed in temperature-controlled chambers (24◦C) with a 12 h light/dark cycle. Mice were challenged with intraperitoneal (i.p) injection of 1 mg/kg LPS [E. coli, serotype O26:B6 (15,000 EU/g), Sigma-Aldrich, St. Louis, MO, USA] as described elsewhere (14). Control mice were i.p administered with the same volume of 0.9% NaCl. Mice were terminated under isoflurane anesthesia 4 h after LPS challenge. Blood was subsequently drawn via cardiac puncture and the organs harvested, snap-frozen on liquid nitrogen and stored at −80◦C until further analysis. All experiments were performed in compliance with the animal ethics committee of the University of Groningen.

#### Endothelial Cell Culture and Stimulation

Human umbilical vein endothelial cells (HUVEC) and human lung microvascular endothelial cells (HMVEC-L) were purchased from Lonza (Lonza, Breda, the Netherlands) and cultured in EBM-2 medium supplemented with EGM-2 MV Single Quot Kit Supplements and Growth factors (Lonza) at 37◦C with 5% CO2/95% air conditions until passage 5 at the UMCG Endothelial Cell Facility. Additionally, HUVEC were isolated from umbilical cords and cultured on 1% gelatin coated plates with RPMI 1640 medium (Lonza) supplemented with 20% heat inactivated FCS, 2 mM L-glutamine, 5 U/ml heparin, 50µg/ml endothelial growth factor, and antibiotics (100IE/penicillin and 50µg/ml streptomycin). HUVEC and HMVEC-L were seeded in 6 or 12 well plates. HUVEC were 70% confluent on the day of siRNA transfection while they were confluent when treated with LPS (E. coli, serotype O26:B6; E. coli, serotype O111:B4, Sigma Aldrich, St. Louis, MO) at 1µg/mL for 4 h unless indicated otherwise.

#### Cryosection Immunofluorescence Staining

Five-micrometer cryosections from snap-frozen mouse kidneys were fixed in acetone for 10 min. The sections were blocked with 0.00125% H20<sup>2</sup> (Merck, Darmstadt, Germany) in demineralized H20 for 10 min, and subsequently blocked with 3% bovine serum albumin (BSA) in phosphate buffered saline (PBS) (w/v) for 30 min. Kidney sections were incubated at room temperature (RT) for 1 h with primary rat-anti-mouse E-selectin antibody (10µg/mL, clone MES-1, kindly provided by Dr. D Brown, UCB Celltech, Brussels, Belgium) diluted in 5% fetal calf serum (FCS) (Sigma, St. Louis, Missouri, USA) in PBS. After washing, sections were incubated with rabbit anti-rat IgG (Vector Laboratories Inc, Burlingame, CA, USA) in 1% normal mouse serum (NMS) (Sanquin, Amsterdam, NL)/5% FCS in PBS for 45 min. The sections were then washed and incubated with anti-rabbit HRP polymer (Dako, Carpentaria, CA, USA) for 30 min. After washing, the sections were exposed to Alexa Fluor <sup>R</sup> 555 Tyramide reagent prepared according to the manufacturer's instructions (#B40955, Thermo Fisher Scientific, Carlsbad, CA, USA). For the sequential second primary antibody incubation, the sections were washed and blocked again with 3% (w/v) BSA in PBS for 30 min. The second primary antibody, rat anti-mouse VCAM-1 (hybridoma supernatant of clone M/K-1.9, ATCC, Manassas VA, USA) was added to the sections for 1 h at RT. After washing, the sections were incubated with 20µg/mL of goat anti-rat antibody conjugated with Alexa Fluor <sup>R</sup> 488 (Thermo Fisher Scientific) / DAPI (1.5µg/mL; Thermo Fisher Scientific) in 5% FCS in PBS for 45 min, washed, and mounted with Aqua/polymount (Polysciences Inc, Warrington, PA, USA). Isotype controls included rat IgG2a and rat IgG1 (10µg/mL; both from Antigenix America Inc., Huntington Station, NY, USA). Images were taken using a Leica SP8 confocal laser scanning microscope (Leica Microsystems Ltd., Germany) with objective and numerical apertures of 40X and 1.3, respectively. Image cubes were recorded with appropriate filters using Leica application suite X software. Peak emission for Alexa Fluor 488 (VCAM-1) was at 550 nm and for Alexa Fluor 555 (E-selectin) at 610 nm. All images were captured with equal exposure times and then analyzed using Imaris image analysis software (Bitplane AG, Zurich, Switzerland).

For CD31/E-selectin respectively CD31/VCAM-1 double staining, the kidney cryosections were blocked with peroxidase and 3% BSA as described above, incubated with rat-anti mouse CD31 antibody (0.15µg/mL, # 550274 BD Pharmingen, Franklin Lakes, NJ, USA) diluted in 5% FCS in PBS for 1 h, and subsequently stained with rat anti-mouse antibody E-selectin, or rat anti-mouse VCAM-1 according to the above-described protocols. The images were taken with a Leica DM4000B fluorescence microscope equipped with a Leica DFC345FX digital camera (Leica Microsystems Ltd., Germany) and Leica LAS V4.5 Image Software at 100X magnification with equal exposure times.

### Flow Cytometry

We employed flow cytometry to determine the expression of endothelial adhesion molecules. HUVEC or HMVEC-L were briefly washed with sterile PBS, trypsinized with trypsin-EDTA (0.025%) and washed with ice-cold wash buffer (5% FCS in PBS). Cells were then transferred to FACS tubes, washed and resuspended in 3% v/v PE-conjugated mouse anti-human Eselectin (Cat. No #322606, BioLegend, San Diego, CA, USA) and APC-conjugated mouse anti-human VCAM-1 (Cat. No #305810, BioLegend) antibodies in wash buffer for 30 min on ice in the dark. For intracellular staining, HUVEC were fixed with Fixation Reagent A (#GAS004, Thermo Fisher Scientific) for 15 min, washed, and subsequently incubated with perm buffer Reagent B (Thermo Fisher Scientific) and 3% (v/v) Brilliant Violet 421-conjugated anti-Ki67 (#330505, Biolegend, San Diego, CA, USA) antibody in wash buffer for 30 min, on ice in the dark. The cells were then washed and resuspended in wash buffer and analyzed using a MACSQuant Analyzer 10 flow cytometer (Miltenyi Biotech, San Diego, CA, USA). Data analysis was performed using Kaluza Flow analysis software (v.2.1) (Beckman Coulter, Brea, CA, USA) or FlowJo software (v.10) (Ashland, OR, USA). Isotype control antibodies mouse IgG1κ-BV421 (Cat No #400157, Biolegend), IgG2aκ-PE (Cat No #2-4724-42, eBioscience, San Diego, CA, USA), and mouse IgG1 (Cat No # IC002A, R&D System, Minneapolis, MN, USA) were used to correct for signals from non-specific binding.

# Fluorescence-Activated Cell Sorting (FACS)

HUVEC treated with LPS for 4 h were trypsinized**,** washed and subsequently incubated with PE-conjugated mouse anti-human E-selectin and APC-conjugated mouse anti-human VCAM-1 antibodies as described above. After the staining procedure, the cells were sorted into 4 subpopulations based on the staining pattern of E-selectin and VCAM-1 using a MoFlo Astrios FACS machine (Beckman Coulter, Brea, CA, USA): "quiescent" E-sel−/VCAM-1<sup>−</sup> (-/-), E-sel+/VCAM−1- (E-sel+), E-sel+/VCAM-1<sup>+</sup> (+/+), and E-sel−/VCAM-1<sup>+</sup> (VCAM-1+) subpopulations (see **Figure 4A)**. After sorting, the purity of each collected population of cells was verified by flow cytometry before subsequent analyses were performed (data not shown).

Trypsinization per se did not alter detection of membrane expression of the adhesion molecules in flow cytometry, as detachment of endothelial cells with versene, a gentle nonenzymatic cell dissociation reagent, showed similar patterns of endothelial surface protein expression (data not shown).

#### Gene Expression Analysis by RT-qPCR

Total RNA from the cells was isolated using the RNeasy <sup>R</sup> mini kit (Qiagen, Venlo, The Netherlands) according to the manufacturer's protocols. RNA concentration (OD 260) and purity (OD260/OD280) was determined using a NanoDrop <sup>R</sup> ND-1000 UV-Vis spectrophotometer (NanoDrop Technologies, Rockland, ME, USA). Samples with an OD260/OD280 ratio of ≥1.8 were included in the analysis. cDNA synthesis was performed as previously described (15). Quantitative (q)PCR was performed in a ViiA 7 PCR System (Applied Biosystems Nieuwerkerk aan den IJssel, The Netherlands) using the following assay-on-Demand primers (Applied Biosystems,): GAPDH (assay ID Hs99999905\_m1), Eselectin (assay ID Hs00174057\_m1), VCAM-1 (assay ID Hs00365486\_m1), TLR4 (assay ID Hs00152939\_m1), RIG-I (assay ID Hs00204833\_m1), MAVS (assay ID Hs00920075), IL-6 (assay ID Hs00174131\_m1), IL-8 (assay ID Hs00174103\_m1), MCP-1 (assay ID Hs00234140\_m1), CXCL10 (assay ID Hs01124251\_g1), and CXCL6 (assay ID Hs00605742\_g1). Duplicate analyses were performed for each sample and the obtained threshold cycle values (CT) averaged. All genes were normalized to the expression of housekeeping gene GAPDH, yielding the 1C<sup>T</sup> value. The relative mRNA level was calculated by 2−1CT .

#### siRNA–Mediated Gene Silencing

TLR4, RIG-I, and MAVS were knocked down in HUVEC using FlexiTube siRNA (Qiagen). AllStars negative control siRNA (Qiagen) was used as a negative control for all RNA interference experiments. Transient transfection was performed using Lipofectamine 2000 (Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. Knockdown of these genes did not diminish endothelial cell viability as described previously (10).

#### Pharmacological Inhibition of Signaling Pathways

BAY11-7082 (BAY) (Alexis Biochemicals, San Diego, CA, USA) was dissolved in DMSO as a 20mM stock solution, while LY2228820 (LY) (MedChemExpress, Monmouth Junction, NJ, USA) was dissolved in DMSO as a 10 mM stock solution according to manufacturers' instructions. Stocks were stored at −80◦C until needed. Prior to the experiment, BAY and LY stock solutions were diluted in HUVEC culture medium. HUVEC were pre-treated with 10µM of BAY 30 min before, and/or 10µM of LY 1 h before LPS stimulation. The viability and morphology of HUVEC were assessed microscopically before and after pretreatment with either BAY or LY, and they were found to be of normal cobble-stone morphology throughout the experiments.

#### Statistical Analysis

Statistical analysis of results was performed by two-tailed unpaired Student t-test, or one-way ANOVA followed by Bonferroni post hoc analysis to compare multiple replicate means. All statistical data were analyzed using GraphPad Prism software v.7 (GraphPad Prism Software Inc., San Diego, CA, USA). Differences were considered significant when p < 0.05.

#### RESULTS

#### Distinct Expression Patterns of E-Selectin and VCAM-1 Are Found Within Kidney Microvascular Compartments of LPS-Treated Mice

Heterogeneous expression of E-selectin and VCAM-1 between different renal microvascular compartments has been previously reported. Intravascular heterogeneity in expression of endothelial adhesion molecules within specific renal microvascular beds, on the other hand, is not extensively described. To investigate the expression patterns of E-selectin and VCAM-1 in the different kidney microvascular segments, we performed immunofluorescent double staining to detect these two adhesion molecules at the same time. In control kidney, E-selectin protein was absent from all microvascular compartments, while arterioles expressed basal VCAM-1 protein **Figure 1**, confirming previous data from our laboratory (7, 16). Following LPS challenge, E-selectin and VCAM-1 became visible in all kidney microvascular segments. However, E-selectin was mainly expressed in the glomeruli, whereas the arterioles predominantly expressed VCAM-1, with scattered co-expression of E-selectin. The peritubular capillaries and post-capillary venules mainly co-expressed E-selectin and VCAM-1, while occasional single E-selectin or VCAM-1 positive cells were found **Figure 1**. Double immunostaining of CD31, a pan-endothelial marker and E-selectin respectively VCAM-1 confirmed their expression by endothelial cells in the microvascular segments **(Supplementary Figure 1).** These data demonstrate that EC in different microvascular compartments, and also within the same microvascular compartment, exert a heterogeneous phenotype which can be identified by different patterns of E-selectin and VCAM-1 protein expression following LPS challenge.

#### LPS Stimulation of Endothelial Cells in vitro Uncover Cell Subpopulations Based on E-Selectin and VCAM-1 Expression

To investigate whether EC in vitro can also express E-selectin and VCAM-1 in varying patterns, we stimulated HUVEC with 1µg/mL LPS for 4 h, after which we flow cytometrically determined E-selectin and VCAM-1 protein levels. In control cells, E-selectin and VCAM-1 expression levels were low **(Figure 2A)**. After 4 h of LPS stimulation, the expression of both E-selectin and VCAM-1 was increased, as evidenced by the shift in the Mean Fluorescent Intensity (MFI) relative to non-stimulated controls **(Figure 2A).** At the cell population level, 4 h of LPS exposure surprisingly revealed the formation of EC subpopulations **Figure 2B**; **Supplementary Figure 2A** which bear similarities with the subpopulations observed in the microvasculature in the kidney. A large subpopulation of EC (approximately 50%) expressed both E-selectin and VCAM-1, while at the same time a significant subset of cells (approximately 20%) remained "quiescent" despite exposure to LPS **Figure 2B**; **Supplementary Figure 2A**. Moreover, two additional EC subpopulations were identified that expressed either E-selectin or VCAM-1 **Figure 2B**; **Supplementary Figure 2A**. Similar

results were found when exposing HUVEC isolated from single donors to 4 h LPS **(Figure 2B)**, and also human lung microvascular endothelial cells (HMVEC-L) exposed to LPS revealed the formation of these four subpopulations **Figure 2B**. Endothelial subpopulations identified by heterogeneous expression of E-selectin and VCAM-1 and combinations thereof after LPS exposure is therefore not restricted to HUVEC and not donor related but a common response of endothelial cells.

We next studied the kinetics of E-selectin and VCAM-1 expression in HUVEC. Upon LPS exposure, some EC started to express E-selectin as early as 1 h after start of activation by LPS, while at 2 h, VCAM-1 expression became apparent **Figure 2C**; **Supplementary Figure 2B**. Moreover, at this early timepoint, four EC subpopulations already had started to form, i.c., "quiescent" E-selectin−/VCAM-1<sup>−</sup> (-/-), E-selectin+/VCAM-1<sup>−</sup> (E-sel+), E-selectin+/VCAM-1<sup>+</sup> (+/+), and E-selectin−/VCAM-1<sup>+</sup> (VCAM-1+). These expression profiles were retained until 6 h of LPS exposure, after which the E-selectin positive subpopulation diminished in cell number, while the VCAM-1+ subset and quiescent subpopulations increased. At 24 h, most EC had returned to a quiescent state, with only a small VCAM-1+ subset (7% of all cells) still being present **(Figure 2C)**. A quiescent endothelial cell subpopulation was constantly present at all times **Figure 2C**; **Supplementary Figure 2B**. Furthermore, these patterns of Eselectin and VCAM-1 expression were similar when exposed to different LPS concentrations, except for the lowest LPS concentration (0.1µg/mL) which did not seem to activate the endothelial cells (**Supplementary Figure 2C**). Taken together, our data demonstrate that activation of endothelial cells by LPS leads to E-selectin and VCAM-1 (co)expression patterns that are dynamically changing depending on the time of exposure to LPS and which are not affected by the concentration of LPS used.

#### Cell Division Is Not a Major Factor in Controlling the Ability of EC to Express E-Selectin and/or VCAM-1 When Exposed to LPS

HeLa cells were reported to repress NF-κB activity following inflammatory stimulation while dividing as cell division appeared to have a higher cellular priority (17). Whether this also occurs in dividing endothelial cells is currently unknown. We hypothesized that the quiescent EC subpopulation was unable to express Eselectin and VCAM-1 because these cells are actively dividing. Staining for Ki67, a marker of actively dividing cells, however, revealed that the total number of Ki67 cells in the HUVEC population as a whole was only approximately 7% **Figure 3**. This is much less than the 25% of cells in the quiescent cell population after LPS stimulation. Moreover, when gating the Ki67 positive cells we found that they were present in all four flow cytometric quadrants **Figure 3**. Hence, cell division per se does not seem to be a major factor influencing the ability of EC to express E-selectin and/or VCAM-1 and is therefore not a likely factor

contributing to the "quiescence status" of EC that do not express the adhesion molecules upon exposure to LPS.

#### Endothelial Subpopulations Have Distinct mRNA Expression Levels of LPS Signaling Pathway Components and Inflammatory Genes

LPS induced adhesion molecule expression in EC is known to be mediated by at least two pattern recognition receptors, TLR4 and RIG-I (10, 18). We therefore hypothesized that the subpopulation distribution of EC after LPS exposure identified by E-selectin and VCAM-1 patterns was due to different subpopulations expressing LPS signaling components to different extents. To investigate this, we incubated HUVEC with LPS for 4 h and then separated the quiescent -/-, E-sel+, +/+ and VCAM-1+ subpopulations by fluorescence activated cell sorting (FACS). After sorting and confirmation of subpopulation purity (data not shown), we first determined the E-selectin and VCAM-1 mRNA expression levels **Figure 4B**. E-selectin and VCAM-1 mRNA levels were close to absent in the quiescent -/ subpopulation, while E-selectin mRNA was high in E-sel+ and +/+ subpopulations, and VCAM-1 mRNA was high in +/+ and VCAM-1+ subpopulations thereby corroborating the protein data. We proceeded to determine the mRNA expression levels of TLR4 and RIG-I within the different EC subpopulations. TLR4 mRNA levels were similar in all EC subpopulations including the quiescent one **Figure 4B**. RIG-I mRNA levels were highest in the E-sel+/VCAM-1<sup>+</sup> cell population and expressed at a slightly lower level in the other EC subpopulations

**Figure 4B**, while mRNA levels of MAVS, a downstream protein adaptor of RIG-I, were similar in all four EC subpopulations **Figure 4B**. To further investigate the inflammatory status of the four subpopulations, we analyzed the expression levels of a series of cytokines/chemokines known to be produced by endothelial cells in inflammatory conditions. While both the quiescent and the VCAM-1+ subpopulations expressed the lowest levels of these genes, the +/+ subpopulation exerted the most pronounced inflammatory phenotype **Figure 4C**. These results demonstrate that upon LPS exposure, there are distinct endothelial subpopulations with varying LPS signaling components and inflammatory phenotypes within a HUVEC cell population.

#### TLR4 and RIG-I Differentially Regulate LPS-Mediated E-Selectin/VCAM-1 Expressing Subpopulations

Since both receptors are expressed in all EC subpopulations, we proceeded to examine the effect of TLR4 and RIG-I siRNA knockdown on the formation of E-selectin and VCAM-1 expressing subpopulations in HUVEC upon LPS exposure. Knockdown of both TLR4 and RIG-I resulted in diminished protein levels of both E-selectin and VCAM-1 when analyzing the HUVEC population as a whole (data not shown) corroborating previous findings (10). While only TLR4 knockdown resulted in inhibition of the E-sel+ subpopulation formation, knockdown of either TLR4 or RIG-I resulted in a major inhibition of +/+ subpopulation formation **Figure 5A**. As a consequence, the -/- "quiescent" subpopulation increased **Figure 5A**. In contrast to the inhibitory effect observed on the E-sel+ or +/+ subpopulations, knockdown of TLR4 or RIG-I did not inhibit the VCAM-1+ subpopulation from forming **Figure 5A**. Similar to the effects of RIG-I knockdown, MAVS knockdown inhibited the formation of the +/+ subpopulation with no effect on E-sel+ or VCAM-1+ subpopulation formation **Figure 5B**. These results suggest that TLR4 predominantly controls the formation of E-sel+ cells and that both TLR4 and RIG-I control the formation of the E-sel+/VCAM-1+ population.

#### LPS-Mediated E-Selectin/VCAM-1 Expressing HUVEC Subpopulations Are Regulated by Different Downstream Signaling Mechanisms

We previously found that the absence of RIG-I or MAVS inhibited nuclear translocation of the NF-κB p65 subunit (10). Moreover, the NF-κB p65 subunit translocated from the cytoplasm to the nucleus in only 26% of EC 30 min after LPS stimulation (**Supplementary Figure 3**). Apart from NF-κB signaling, activation of p38 MAPK signaling in LPS exposed endothelial cells has previously been reported (19, 20). We therefore investigated whether one or both of these two signaling pathways were affecting the formation of one or more subpopulations employing pharmacological tools and if so, to what extent the subpopulations were affected.

When examining the subpopulation distribution after 2 h exposure to LPS we found that p38 MAPK inhibition using LY exerted a major inhibitory effect on the formation of the E-sel+ subpopulation **Figure 6A**. In contrast, NF-κB inhibition using BAY did not significantly influence the formation of any of the subpopulations at this early time point. At 4 h LPS exposure however, blockade of NF-κB and p38 MAPK had inhibited the formation of the +/+ subpopulation (**Figure 6B**). In addition, the combination of NF-κB and p38 MAPK inhibition strongly diminished the formation of the E-selectin+/VCAM-1+ expressing subpopulation **Figure 6B**. Formation of the VCAM-1+ subpopulation was not affected by inhibition of NF-κB and/or p38 MAPK. Also, inhibition of both NFκB and p38 MAPK did not fully block the formation of the subpopulations, suggesting that other signaling pathways contribute to their formation.

# DISCUSSION

Sepsis is a life-threatening condition which is characterized by progressive host dysregulation, following a known or suspected infection (1). Sepsis pathophysiology is still not fully understood which has hampered the development of therapeutic options

counteracting organ failure in patients with sepsis. Due to their location, endothelial cells are among the first cells to respond to systemic pathogens or bacterial products (3). A hallmark of sepsis-related organ dysfunction is (micro)vascular dysfunction which include endothelial activation. This results in immune cell infiltration which can be detrimental for organ function (5). The current study aimed to explore the molecular control of early LPS mediated endothelial activation in relation to

previously reported heterogeneous EC responses found in vivo. Our findings reveal that in the kidneys of mice challenged with LPS, EC between different microvascular compartments and also within the same microvascular bed, exert a heterogeneous phenotype as identified by different patterns of E-selectin and VCAM-1 protein expression. Moreover, we observed that in vitro in endothelial cell cultures, subpopulations appear shortly after the start of LPS exposure that exert a phenotypic Eselectin/VCAM-1 heterogeneity similar to that observed in the mouse kidney. These endothelial subpopulations have distinct inflammatory phenotypes, with the E-selectin/VCAM-1 double positive subset showing the highest level of activation. The formation of the subpopulations is differentially regulated by distinct signaling mechanisms, both at the level of TLR4 and RIG-I, and at the level of the NF-κB and p38 MAPK pathways. Strikingly, we found a quiescent population of cells that was not only devoid of E-selectin and VCAM-1 expression but was also lacking the expression of other proinflammatory cytokines and chemokines despite exposure to LPS.

It is well known that EC acquire organ- and tissue-specific identities to support the unique requirements of various organs in the body (21). The differential expression of E-selectin and VCAM-1 in different renal microvascular compartments of LPS-treated mice reported in this study corroborates previous observations from our group (6, 7). In addition, we revealed that

within the same microvascular compartment, endothelial cells heterogenically express E-selectin and VCAM-1 following LPS challenge. While phenotypic differences between adjacent EC conditioned in the same environment were previously shown for endothelial barrier antigen in capillaries of the brain of rats (22) and for Tie-2 in mouse xenografts associated tumor neovessels (23), heterogeneous expression of cell adhesion molecules within individual microvascular segments has to our knowledge not been reported before. Interestingly, endothelial cells in vitro exhibited a similar phenotypic heterogeneity in E-selectin and VCAM-1 (co)expression patterns. The four subpopulations of EC, each with a different composition of E-selectin / VCAM-1 (co)expression, did not only appear in multi-donor HUVEC, but also in single-donor HUVEC and in human lung microvascular endothelial cells (HMVEC-L) exposed to LPS. From this we conclude that the observed heterogenous responsiveness is not a specific attribute of HUVEC but rather a common response of endothelial cells exposed to LPS. Of notice, HMVEC-L displayed a different composition of EC subpopulations, which could be attributed to a specific lung-microvascular phenotypical characteristic to support dedicated biological functions in the lungs. The E-sel+/VCAM-1<sup>+</sup> subpopulation also expressed the highest levels of inflammatory mediators IL-6, IL-8, MCP-1, CXCL6, and CXCL10, implying that this subpopulation has a more generalized higher activation status than the other three subpopulations. This finding is in line with a previous study that showed that high IL-8-secreting EC produced higher levels of endothelial adhesion molecules, chemokines, and cytokines compared to their low IL-8 secreting EC counterparts (24). Why CXCL6, a chemoattractant for neutrophils, was specifically expressed in the VCAM-1+ subpopulation remains to be clarified. Follow up studies will determine the full nature and extent of activation of the cells in these subpopulations.

Intriguingly, we found a subpopulation of EC that did not express E-selectin or VCAM-1 despite prolonged exposure to LPS. Gene expression analysis of this sorted subpopulation of cells showed that these cells had a broader "quiescent" phenotype since they also did not express inflammatory cytokines and chemokines upon LPS exposure. One possible explanation for these findings could have been that these cells lack LPSsignaling machinery, but we showed that that was not the case as this population was expressing both TLR4 and RIG-I, two of the main LPS-signaling molecules in endothelial cells. Another explanation for these findings might have been that the cells in the "quiescent" phenotype subpopulation were dividing and thereby subjected to repression of NF-κB–driven inflammation as was previously reported in HeLa cells (17). However, we found only 7% of the total HUVEC population to be proliferating, while the quiescent population represented around 20% of cells. Moreover, Ki67 positive, proliferating cells were also able to express E-selectin, which confirmed previous observations (25, 26) and VCAM-1. Hence, cell proliferation does not appear to be a major contributing factor leading to the "quiescent" status of these cells. Since exaggerated and prolonged inflammatory responses can be detrimental to cells, a negative feedback loop mechanism exists that regulates the magnitude and duration of inflammation (27). An example of such is the activation of the zinc finger protein A20 which limits inflammation downstream of the NF-κB pathway. A20 mRNA levels were, however, comparable in all EC subpopulations (data not shown) including the quiescent EC subpopulation which suggests that an A20-dependent inhibitory feedback mechanism is not controlling the quiescent status of these E-selectin−/ VCAM-1<sup>−</sup> cells. Why these cells remain quiescent despite LPS exposure is currently unknown. We will investigate this further since understanding the molecular mechanisms associated with this quiescent phenotype may eventually be exploited for therapeutic strategies to inhibit endothelial activation in the setting of sepsis.

We explored the possibility that the heterogeneous responses by the EC subpopulations were attributed to intrinsic cellular differences in the functionality of the LPS signaling components. Using siRNA-based knock down we investigated the role of TLR4 and RIG-I in this process. While TLR4 mediated both E-sel+ and E-sel+/VCAM-1<sup>+</sup> subpopulation formation, RIG-I mainly had a role in E-sel+/VCAM-1<sup>+</sup> subpopulation formation. Knocking down MAVS, a RIG-I adaptor protein, did not

affect the formation of E-sel+ and VCAM-1+ subpopulations. Pharmacological inhibition studies showed that NF-κB and p38 MAPK were the predominant controlling pathways for the formation of the E-sel+ and E-sel+/VCAM-1<sup>+</sup> subpopulations (see **Figure 7**). The fact that the VCAM-1+ population at early time points of LPS stimulation was not affected by any of the pathways investigated raises the question how this subpopulation is being formed when endothelial cells are exposed to LPS. One way to establish a clear role of LPS signaling in VCAM-1+ subpopulation formation is by performing knockdown experiments and examining the activation status when the VCAM-1+ subpopulation is at its peak (16 h after LPS stimulation, **Supplementary Figure 2B**). At 16 h of LPS exposure, it is highly likely that secondary or tertiary responses have become part of the equation following the release of pro-inflammatory cytokines, such as IL-6 and IL-8. Exposing naïve HUVEC to cytokines (e.g., IL-6) has been shown to increase E-selectin, VCAM-1, and ICAM-1 (28). Considering the pharmacological importance of the findings reported here, we will investigate this issue in more detail in HUVEC subpopulations using kinase activity platform technology.

In conclusion, renal endothelial cells exert a heterogeneous pattern of E-selectin and VCAM-1 expression following LPS challenge, in different as well as within the same microvascular segments. Such a heterogeneous response was recapitulated in vitro when HUVEC and human lung microvascular endothelial cells were exposed to LPS. The here identified and described endothelial subpopulations have distinct inflammatory phenotypes and are regulated by different signaling mechanisms. At this point, it is not known whether signaling differences contributing to the heterogeneous expression of E-selectin and VCAM-1 has biological relevance, nor what the pharmacological ramifications are. Such biological relevance should be investigated in in vivo settings, as the molecular signatures of endothelial cells differ depending on their specialized functions and microenvironment both in homeostasis and during inflammatory responses (21, 29). Follow up studies will aim to further investigate the underlying

# REFERENCES


molecular pathways in in vitro and in vivo as a lead for future therapy choices to attenuate activation of microvascular endothelial cells in the organs of critically ill patients with sepsis.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the animal ethics committee of the University of Groningen.

#### AUTHOR CONTRIBUTIONS

JM and GM shared the conceptualization. JM, GM, and ED designed the study. ED, JP, BE, and JM performed the experiments and analyzed the data. MM provided valuable input on the statistical analysis. ED, JM, and GM wrote and edited the manuscript. All authors critically revised the manuscript and approved the submitted version.

# FUNDING

This work was supported by the Skim Latihan Akademik Bumiputera (SLAB) and UNIMAS fellowship program (to ED), Ministry of Higher Education, Malaysia, and the research foundation of the Department of Critical Care, UMCG.

#### ACKNOWLEDGMENTS

We would like to thank Henk Moorlag from the UMCG Endothelial Cell Facility and Wayel Abdulahad, Theo Bijma, Johan Teunis, and Geert Mesander from the UMCG Central Flow Cytometry Unit for providing excellent technical support.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01169/full#supplementary-material

with reduced glomerular barrier function. AJP Ren Physiol. (2009) 297:F272–81. doi: 10.1152/ajprenal.00137.2009


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

Copyright © 2019 Dayang, Plantinga, ter Ellen, van Meurs, Molema and Moser. 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 Fluctuations of Leukocytes and Circulating Cytokines in Septic Humanized Mice Vary With Outcome

Tomasz Skirecki <sup>1</sup> , Susanne Drechsler <sup>2</sup> , Grazyna Hoser <sup>1</sup> , Mohammad Jafarmadar <sup>2</sup> , Katarzyna Siennicka<sup>3</sup> , Zygmunt Pojda<sup>3</sup> , Jerzy Kawiak <sup>1</sup> and Marcin F. Osuchowski <sup>2</sup> \*

*<sup>1</sup> Laboratory of Flow Cytometry, Centre of Postgraduate Medical Education, Warsaw, Poland, <sup>2</sup> Ludwig Boltzmann Institute for Experimental and Clinical Traumatology in the AUVA Research Center, Vienna, Austria, <sup>3</sup> Department of Regenerative Medicine, Maria Sklodowska-Curie Institute-Oncology Center, Warsaw, Poland*

Sepsis remains a major challenge in translational research given its heterogeneous pathophysiology and the lack of specific therapeutics. The use of humanized mouse chimeras with transplanted human hematopoietic cells may improve the clinical relevance of pre-clinical studies. However, knowledge of the human immuno-inflammatory response during sepsis in humanized mice is scarce; it is unclear how similar or divergent mouse and human-origin immuno-inflammatory responses in sepsis are. In this study, we evaluated the early outcome-dependent immuno-inflammatory response in humanized mice generated in the NSG strain after cecal ligation and puncture (CLP) sepsis. Mice were observed for 32 h post-CLP and were assigned to either predicted-to-die (P-DIE) or predicted-to-survive (P-SUR) groups for retrospective comparisons. Blood samples were collected at baseline, 6 and 24 h, whereas the bone marrow and spleen were collected between 24 and 32 h post-CLP. In comparison to P-SUR, P-DIE humanized mice had a 3-fold higher frequency of human splenic monocytes and their CD80 expression was reduced by 1.3-fold; there was no difference in the HLA-DR expression. Similarly, the expression of CD80 on the bone marrow monocytes from P-DIE mice was decreased by 32% (*p* < 0.05). Sepsis induced a generalized up-regulation of both human and murine plasma cytokines (TNFα, IL-6, IL-10, IL-8/KC, MCP-1); it was additionally aggravated in P-DIE vs. P-SUR. Human cytokines were strongly overridden by the murine ones (approx. ratio 1:9) but human TNFα was 7-fold higher than mouse TNFα. Interestingly, transplantation of human cells did not influence murine cytokine response in NSG mice, but humanized NSG mice were more susceptible to sepsis in comparison with NSG mice (79 vs. 33% mortality; *p* < 0.05). In conclusion, our results show that humanized mice reflect selected aspects of human immune responses in sepsis and therefore may be a feasible alternative in preclinical immunotherapy modeling.

Keywords: sepsis, humanized mouse, cecal ligation and puncture (CLP), peritonitis, outcome prediction, immunity

# INTRODUCTION

In more than 30 years of intensive research no new specific therapeutics for sepsis have been introduced into clinical care (1). Despite improved standards of care, the mortality rates of sepsis have not decreased (2). On the contrary, prevalence has been on the rise, reaching approximately 19 million sepsis cases per year worldwide (3), making sepsis a global medical burden (4).

#### Edited by:

*Christoph Thiemermann, Queen Mary University of London, United Kingdom*

#### Reviewed by:

*Craig Coopersmith, Emory University, United States Regina Sordi, Federal University of Santa Catarina, Brazil*

#### \*Correspondence:

*Marcin F. Osuchowski marcin.osuchowski@trauma.lbg.ac.at*

#### Specialty section:

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

Received: *08 March 2019* Accepted: *06 June 2019* Published: *26 June 2019*

#### Citation:

*Skirecki T, Drechsler S, Hoser G, Jafarmadar M, Siennicka K, Pojda Z, Kawiak J and Osuchowski MF (2019) The Fluctuations of Leukocytes and Circulating Cytokines in Septic Humanized Mice Vary With Outcome. Front. Immunol. 10:1427. doi: 10.3389/fimmu.2019.01427*

**88**

The failure of anti-sepsis therapeutics to modulate the host response has been attributed to a large part to flawed preclinical studies (5–7). Currently, a vast majority of sepsis animal studies are performed in mice (8); one of the major translational obstacles in murine models of sepsis are substantial mousehuman differences in the immune system (9). The development of humanized mice is emerging as a promising platform for at least partially overcoming those disparities. Humanized mice are generated by xenotransplantation of human hematopoietic stem cells (HSCs) into immunocompromised mice which were additionally irradiated or chemoablated to empty their bone marrow niches for the xenograft (10). Currently, one of the most commonly used mouse strains for humanization is the NSG (NOD scid gamma) strain (11); NSG mice have composite immunodeficiency given that they lack not only lymphocytes, NK cells, and macrophages but also have an impaired complement system, enabling an efficient engraftment of the human cells with a relatively low risk of developing graft-vs.-host disease.

Humanized mice have already been used to model acute infectious diseases and have proven useful in both recapitulating some unique human responses as well as suggesting therapeutic solutions (12–14). On a limited scale, humanized mice were also employed in sepsis research using the cecal ligation and puncture (CLP), the gold standard model to recapitulate septic peritonitis (15–18). CLP-induced immuno-inflammatory responses vary according to the underlying pathophysiology, i.e., in a non-marginal severity CLP protocol (i.e., neither 100% mortality/survival), mice subjected to the same insult develop an individual response and progress to either survival or death without following a predefined immune "outcome pattern" from the onset of CLP (19). Outcomes in CLP mice can be predicted (e.g., mice predicted-to-survive/die) using body temperature and/or biomarker measurements as well as clinical assessment scores (19–22). Comparisons of such dynamic signaling changes between the surviving and dying phenotypes advance our understanding of mechanisms co-responsible for sepsis lethality (and recovery) better than simplistic (but uniformly statistically significant) comparisons of healthy to septic subjects (21–23). Humanized mice are an attractive platform for the investigation of human immuno-inflammatory responses (and treatments targeting these systems) in critical care diseases including sepsis. It is therefore crucial to characterize the cellular and cytokine responses, given that the complexity of the humanized sepsis mouse model is likely exacerbated by various interspecies interactions (11). The concomitant presence of cells and mediators of both human and murine origin results in multi-directional crosstalk which likely alters the immuneinflammatory dynamics of the host in severe infections. It is largely unclear what action the transplanted human immune cells exert upon the mouse host's response to septic insult and how impactful this interaction can be. Such a knowledge gap hinders the utility of humanized mice in modeling of sepsis as well as the translation of findings generated in such models.

To partially address the above unknowns, we subjected humanized NSG mice to polymicrobial abdominal sepsis and investigated two specific questions: (a) comparison of the initial cellular and humoral inflammatory response of human and murine origin and (b) characterization of those responses depending on outcome (i.e., dying vs. surviving) in the early phase of sepsis.

# METHODS

# Mice

Mice of the NSG strain (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) and SCID strain were obtained from The Jackson Laboratories (Bar Harbour, ME, USA). Mice were bred in the animal facility of the Center of Postgraduate Medical Education (Warsaw, Poland) under pathogen-free conditions with 12/12 light cycle and fed with standard sterile diet and drinking water ad libitum. All experiments on animals were approved by the Local Ethics Committee no IV in Warsaw, Poland (92/2012) and adhered to the ARRIVE guidelines (24).

#### Generation of Humanized Mice Human Stem Cell Harvesting

To generate humanized mice, we followed the previously published protocol from our lab (17). In brief, human umbilical cord blood (UCB) probes were processed in accordance with the procedures approved by the Institutional Review Board of the Maria Sklodowska-Curie Memorial Cancer Center and Institute of Oncology (Warsaw, Poland). The UCB units from healthy donor mothers were obtained with their informed consent. Then, human HSCs were isolated from Ficoll-centrifuged mononuclear cell fraction of human UCB and stored in liquid nitrogen after isolation. CD34-positive human stem and progenitor cells were purified from thawed mononuclear cells by positive immunomagnetic separation technique using a commercially available kit (EasySep, Stemcell Technologies, Vancouver, BC, Canada). Purity of isolated cells (>90%) was tested by flow cytometry using anti-CD34 PE/anti-CD45 FITC antibodies (BD Bioscience, San Jose, CA, USA).

#### Transplantation

Three- to four-week-old NSG female mice were pretreated with i.p. injections of busulfan (25 mg/kg) (Sigma-Aldrich) for 2 consecutive days before stem cell transplantation. Female mice were chosen as they are known to present a higher engraftment rate (25). Twenty-four hours after the last busulfan dose, mice were given a tail vein injection of 10<sup>5</sup> purified human UCB CD34+ cells. Seven weeks after transplantation, 20 µl of blood was obtained via facial vein puncture (23G needle). For the flow cytometry evaluation of human chimerism, the blood samples were stained with anti-human CD45 FITC (BD) antibody. A total of 33 mice were deemed as humanized (blood chimerism over 5%) and were enrolled in the CLP study.

# CLP Sepsis Model Surgery and Treatment

In order to reproduce human peritonitis-derived sepsis, we performed the CLP surgery in 33 humanized mice according to the original protocol from Wichterman et al. (26) with modifications (22, 27). In brief, all mice received analgesia (i.e., buprenorphine, 0.05 mg/kg in 1 ml of 0.9% saline) 20 min

prior to CLP and the abdomen was shaved and disinfected with alcohol. After midline laparotomy, the cecum was exposed, ligated underneath the ileocecal valve, and a needle puncture (through-and-through) was applied. After repositioning of the cecum, the abdomen was closed with single button sutures and Histoacryl <sup>R</sup> tissue adhesive (B. Braun, Aesculap, Germany).

Two CLP runs were performed in humanized mice: (a) with a 25G needle (n = 19) and (b) with a 27G needle (n = 14). A single CLP run with a 25G needle was performed in NSG (n = 10) and SCID (n = 10) mice. Humanized mice from both runs were combined for data shown in **Figures 1**–**5** (i.e., outcome-based comparisons). Two mice that died within 6 h of CLP were excluded as death was deemed to be due to the fact that they did not recover from the anesthesia/surgical intervention itself (and not sepsis). For the comparison of NSG to SCID in **Figure 6** (i.e., comparison of sensitivity to an identical insult), humanized mice from the first run (25G needle) only were used. Based on our previous CLP protocols (21, 28), we used two needle sizes in order to introduce more longitudinal outcome variability into the study. In the current study, both sizes produced similar overall outcomes at 32 h (**Supplementary Figure 1**). The more divergent variability is advantageous given that our study focused on the outcome-based differences (i.e., surviving vs. moribund phenotype) and not a CLP phenotype produced by a specific needle size. A similar (2-needle) approach was recently used by Kim et al. (29) in testing a hydrocortisone/ascorbic acid/thiamine treatment in mouse CLP.

From 2 h post-CLP on, all mice received subcutaneous broad spectrum antibiotic therapy (25 mg/kg imipenem, Zienam <sup>R</sup> ; MSD, Lucerne, Switzerland) and fluid resuscitation (1 ml Ringer's solution) with analgesia (0.05 mg/kg buprenorphine, Bupaq <sup>R</sup> ; Richter Pharma, Austria) twice daily (approximately every 12 h) for 5 consecutive days post-CLP.

#### Monitoring and Prediction of Outcome

All CLP mice were not maintained in a large animal facility but were kept in a small animal room to enable close observation for implementation of the humane endpoints and to allow allocation of individual mice into either predicted-to-die (P-DIE) or predicted-to-survive (P-SUR) group. All mice were monitored for clinical signs of illness and their status was evaluated using a modified mouse clinical assessment scoring system [M-CASS; e.g., fur, posture, mobility, alertness, startle, and righting reflex (22)] starting 12 h post-CLP. Also, rectal temperature was monitored (Fluke Series II thermometer, Fluke USA) at least twice daily (or more often whenever a mouse deteriorated). Mice were deemed moribund and assigned as P-DIE whenever the righting reflex was absent or/and M-CASS score ≥8 and immediately euthanized. Following the previously used protocol (21, 30), two P-SUR mice were randomly selected and euthanized at 24 h to serve as a time-matched comparison for moribund mice sacrificed as P-DIE at 24 h post-CLP. Body temperature recordings served as a supportive measure in the general health assessment and euthanasia decision-making scheme (and the P-DIE vs. P-SUR allocation) given that the body temperature fluctuations were not previously validated to serve as a predictor of outcome in any humanized mice disease model. Two mice (out of 33) died of post-surgical complications within 5 h of CLP and were excluded from data analysis.

#### Flow Cytometry

Flow cytometry was used for phenotyping of engrafted human cells. Probes of blood or re-suspended cells retrieved from solid organs were incubated with inactivated mouse serum to block unspecific Fc receptors. Then, mixtures of monoclonal antibodies against human antigens were added and cells were incubated for 30 min at room temperature. The following antibodies were used: anti-CD45 AmCyan, anti-CD14 PE, anti-CD3 Pacific Blue, anti-CD4 APC, anti-CD8 PE, anti-CD33 PeCy5.5, anti-HLA-DR PE.CY7, (BD Bioscience), anti-CD80 Alexa Fluor 488 (Biolegend, San Diego, USA). After staining, erythrocytes were lysed with BD Pharm Lyse solution (BD Biosciences) for 10 min, washed with PBS with 2% Newborn Calf Serum (NCS), and re-suspended in 0.5% paraformaldehyde in PBS. Cells were acquired using the FACS Canto II flow cytometer and Diva software (BD Bioscience, San Jose, CA). Analysis of immunophenotype was carried out applying the FlowJo 10 software (TreeStar, Inc., now part of FlowJo LLC, Ashland, OR, USA). The applied antihuman antibodies were verified for cross-reactivity with cells of non-transplanted mice and no staining was present.

#### Cytokine Measurements

Facial vein blood samples were obtained at baseline, 6 and 24 h after CLP as described (31). Samples were stored at −86 ◦C until analysis. Plasma concentration of human and murine cytokines (IL-6, IL-8/KC, IL-10, TNF, MCP-1) was measured using Luminex Multiplex Immunoassay (Invitrogen, Thermo Fisher Scientific, Vienna, Austria) according to the manufacturer's protocol. The cross-reactivity rates provided by the supplier (expressed as % recovery of either human or mouse protein) for mouse targets are as follows: IL-6: 0.1%; TNFα: 0.0%; MCP-1: 0.5%; IL-10: 0.2% Gro alpha KC: 0.1%. For human targets: IL-10: 0.0%; IL-6: 0.0%; IL-8: n.a.; Gro alpha: 0.0%; MCP-1: not checked; TNFα: 0.0%. Two P-SUR mice were also sacrificed

(E) Geometric mean fluorescence (GMF) of anti-HLA-DR staining of human monocytes. (F) GMF of anti-CD80 staining of human monocytes. Analysis of splenic cells: (G) Frequency of human monocytes. (H) GMF of anti-HLA-DR staining and (I) GMF of anti-CD80 staining of human monocytes. Insets present a comparison of healthy (*n* = 7) vs. all septic mice (*n* = 20; irrespective of outcome). Groups were compared using *t*-test. \**p* < 0.05; \*\* *p* < 0.01; \*\*\*\**p* < 0.0001.

at 24 h to serve as comparisons to P-DIE that were sacrificed at the same time-point. Additionally, in **Figures 2**, **6**, the individual cytokine values measured at 24 and 32 h post-CLP were merged (for P-DIE and P-SUR, respectively) and presented as the 24 h time-point only. This was justified by an overlapping cytokine expression (in all cytokines) between those two time-points (i.e., p > 0.1 for 24 vs. 32 h difference; data not shown).

#### Statistical Analysis

Normality of all data sets was assessed using the Shapiro-Wilk test and log-transformed to eliminate skewness and/or nonparametric distribution whenever present. Data are expressed as means and 95% confidence intervals (CI) if not otherwise stated. Comparisons between P-DIE and P-SUR group were performed by t-test (with Welch correction for unequal variances if needed) given that the P-DIE group did not meet assumptions (nonrandom deaths) for repeated measures testing. The correlation strength between the variables was assessed using the Pearson's rank correlation coefficient. Receiver operator curves (ROC) were calculated for the evaluation of predictive utility of selected variables. Cut-off values were chosen using the Youden's index. The areas under the curves (AUC) with CIs were calculated for assessment of the accuracy of the test. Sensitivity and specificity were calculated for the selected cut-off values of the variables.

The Kaplan-Meier survival curves were compared using the log-rank test. p < 0.05 was considered significant. Statistica 12 (StatSoft, Inc., USA) and GraphPad Prism 5 (GraphPad, Inc., USA) softwares were used for evaluating the statistical significance and/or graphical depiction of the data.

#### RESULTS

#### Development of Humanized Mice

Humanized NSG mice were generated based on our previously used protocol (17). The busulfan myeloablation regimen was proven to be efficient and safe without any clinical pathology in the recipient mice. Development of human granulocytes, monocytes, and B and T cells was confirmed to be similar to our previously reported readouts in the spleen, bone marrow, and peripheral blood (17) (**Supplementary Table 1**).

Among circulating human CD45+ cells, CD20+ B cells, and CD33+ myeloid cells were the most frequent. Analysis of the thymus revealed the development of human CD3+ cells in the mouse host (data not shown). Chimerism of human CD45+ (hCD45+) cells in the peripheral blood was confirmed in all transplanted mice 1 week before the CLP

experiments [mean 19.6% 95CI (11.3–27.1)]. Furthermore, all humanized mice enrolled in the study had >18% of hCD45+ cells in the bone marrow (retrospective post-CLP verification; **Supplementary Table 2**).

#### Humanized Septic Mice: Prediction of Outcome

CLP surgery in humanized mice resulted in 55% mortality at 24 h and 71% at 32 h (the experiment termination timepoint) (**Figure 1**). At the 32 h time-point, all remaining mice were assigned either to the predicted-to-die (P-DIE; n = 22) or predicted-to-survive (P-SUR; n = 9) group based on the clinical assessment and body temperature readouts (see Methods) for further retrospective comparisons of the measured parameters.

We also performed a retrospective ROC analysis using body temperature recordings only (taken at 6 h post-CLP) to assess whether a prediction of (short term) outcome is possible in humanized septic mice. This analysis combined all CLP mice and the outcome was based on our clinical P-DIE/P-SUR allocation. The area under the curve reached 0.99 (95CI: 0.98–1.00) with the cut-off set at 33.2◦C (i.e., 100% sensitivity, 95% specificity for the next 32 h). Of note, all P-SUR mice (i.e., allocated at 32 h post-CLP) had body temperature of at least 33.2◦C or higher (**Supplementary Figure 2**).

#### Humanized Septic Mice: Outcome-Based Characterization of Human Monocytes From the Bone Marrow and Spleen

To investigate the cellular response of humanized mice to abdominal sepsis, we analyzed CLP-induced leukocyte changes in two compartments: (a) the bone marrow and (b) spleen. Leukocytes from P-DIE mice were harvested between 24 and 32 h after CLP (n = 11) in order to also harvest cells from mice that became moribund prior to the end of the study. Leukocytes from remaining P-SUR mice were collected at the termination of the experiment (i.e., 32 h post-CLP; n = 9).

There was a similar percentage frequency of hCD45+ leukocytes in both compartments: 34% P-DIE vs. 45% P-SUR in the bone marrow and 18 vs. 26% in the spleen (**Table 1**). In the bone marrow, hCD14+ monocytes (**Figures 2A,D**) were similar regardless of outcome. In contrast, the splenic hCD14+ monocytes were 3-fold higher in P-DIE compared to P-SUR mice (**Figure 2G**).

We also evaluated the geometric median fluorescence (GMF) of anti-HLA-DR and anti-hCD80 antibodies on human

survival curves of hNSG (*n* = 19), naïve NSG (*n* = 10), and SCID (*n* = 10). Humanized NSG mice were terminated at 32 h post-CLP for tissue collection and analysis. Survival curves were compared by Log-rank test, *p* < 0.05 for hNSG vs. NSG and vs SCID. (B) Long-term survival curve of NSG and SCID mice from the same experiment. (NSG *n* = 10, SCID *n* = 10, hNSG *n* = 19). Survival curves were compared by Log-rank test, *p* < 0.05. Murine cytokine responses in NSG (*n* = 10) and hNSG (*n* = 19) G: (C) TNF; (D) IL-6; (E) KC/IL-8; (F) IL-10; (G) MCP-1. Murine cytokine levels in humanized vs. NSG mice were compared by the paired *t*-test at each time-point separately.

monocytes (**Figures 2B,C**). The HLA-DR GMF on the bone marrow monocytes was similar, irrespective of outcome (**Figure 2E**). In contrast, the GMF of CD80 was decreased by 32% on the human monocytes from P-DIE (vs. P-SUR) mice (**Figure 2F**).

The intensity of anti-HLA-DR staining of the splenic monocytes did not differ between the groups (**Figure 2H**), however, the GMF of anti-hCD80 was 1.3-fold higher in the P-SUR compared to P-DIE mice (**Figure 2I**). Of note, the frequency of the hCD3+ T cells appeared to be (4-fold) higher in the spleen of P-SUR than P-DIE mice but this difference was not statistically significant (high individual variability).

Pearson's rank correlation coefficient revealed a moderate, but significant correlation between the frequency of hCD3+ TABLE 1 | Human immune cells in the humanized mice based on the CLP outcome.


*Statistically significant values are bolded.*

TABLE 2 | Prognostic accuracy of human and murine cytokines 6 h post-CLP for 24 h death prediction.


*AUC, area under the curve; CI, confidence interval.*

T cells in the spleen and the hCD80 GMF on hCD14+ monocytes (r = 0.62, p < 0.01).

#### Humanized Septic Mice: Comparison of the Human vs. Murine Circulating Cytokines

Repetitive small-volume blood sampling enabled us to sequentially evaluate the concentration of cytokines in septic mice without sacrificing them. First, we compared the combined kinetics of all measured circulating cytokines and chemokines of murine and human origin. Irrespective of outcome, the analyzed mediators typically peaked at 6 h and decreased at 24 h post-CLP (**Figure 3**).

Next, we assessed the total content of circulating cytokines of human and murine origin. The combined human mediators reached 24% of all measured cytokines in healthy mice, 8% at 6 h and 7% at 24 h post-CLP in septic mice (**Figure 3A**). Accordingly, the post-CLP concentration of individual human mediators was, in most cases, an order of magnitude lower than of their murine counterparts. Interestingly, human TNFα was the only cytokine whose early post-CLP peak (i.e., 6 h) was higher (by 7-fold) than the murine TNFα (**Figure 3B**).

While the post-CLP response dynamics of human and murine TNFα, IL-6, and IL-8/KC were similar (**Figures 3B,C,E**), circulating human IL-10 and MCP-1 showed a delayed increase in comparison to the murine mediators (**Figures 3D,F**).

There was a significant negative correlation between expression of CD80 on splenic human monocytes and the concentration of human systemic pro-inflammatory cytokines, while no correlation with IL-10 was observed 6 h after CLP (**Supplementary Figure 3**).

#### Humanized Septic Mice: Comparison of Humanized and Murine Circulating Cytokines Based on Outcome

Regardless of the origin species, the concentration of all circulating cytokines (except murine KC, human IL-10, and MCP-1) was typically higher in P-DIE compared to P-SUR mice at 6 h post-CLP (**Figures 4**, **5**). At 24 h, this outcome difference remained (or appeared) in IL-6, IL-8, IL-10, and MCP-1 (both origins) and disappeared in TNFα (both origins) and KC (i.e., mouse only).

Furthermore, the different species-dependent IL-10 and MCP-1 dynamic (observed in **Figures 3D,F**) was predominantly due to the changes in the P-DIE mice (IL-10: **Figures 4E,F**; MCP-1: **Figures 5C,D**); the release of human mediators was delayed to 24 h post-CLP compared to the immediate rise of the mouse cytokines at 6 h.

We also performed the ROC analysis to assess the prognostic outcome accuracy of the mediators measured at 6 h post-CLP. For murine cytokines, TNFα and IL-6 had the highest AUC of 0.93 and 0.97 (**Table 2**). Of note, AUC for human IL-8 reached 0.93, indicating a strong discriminative value for this human chemokine.

#### Humanized vs. Immunodeficient Mice: Outcome

To verify how human immune cells transplanted to immunocompromised NSG mice influence sepsis phenotype, we simultaneously performed CLP in naïve NSG and SCID mice (**Figure 6**). We also included SCID mice as this strain is less immunodeficient than NSG; its adaptive (but not innate) immunity is impaired (32). Using the same CLP protocol, we observed that the mortality of both non-humanized strains was lower by 46% compared to the humanized NSG mice (33 vs. 79%; p < 0.05) at 32 h after CLP (**Figure 6A**). Given that we did not collect tissues from the non-humanized mice, NSG and SCID animals were followed up until day 5. Compared to SCID, NSG mice were more vulnerable to the identical CLP insult (30 vs. 80% mortality; p < 0.05; **Figure 6B**).

#### Humanized vs. Immunodeficient Mice: Circulating Cytokines

To further evaluate the impact of the human cells on murine humoral inflammatory response itself, we compared the concentrations of circulating murine cytokines after an identical CLP insult in the naïve and humanized NSG mice. The pattern of the post-CLP cytokine release was generally similar regardless of the xenotransplantation of the human immune cells (**Figures 6C–G**). The trend for an enhanced MCP-1 release (**Figure 6G**) after humanization did not reach statistical significance.

# DISCUSSION

The development of humanized mice provides a unique opportunity to study responses of human immunocompetent cells residing in different tissues in a controlled setting in sepsis, however, such a model requires an extensive characterization of human and murine immuno-inflammatory components. In our study, we used clinically-relevant CLP sepsis (26) followed by antibiotic-, fluid-, and analgesic treatment to recapitulate septic patient care. The findings show for the first time that both human and murine cytokine responses have similar dynamics with regard to acute (early) post-CLP outcome. Moreover, we demonstrated that an early release of human TNFα exceeded by 10-fold the release of its murine counterpart. We also observed that the cytokine response correlated with human monocyte changes, suggesting dynamic interactions in the human immune compartment. Importantly, we revealed that the applied protocol of human HSC transplantation did not impair the ability of murine cells to mount an inflammatory response to sepsis.

We chose to focus on the response of human monocytes in the spleen and bone marrow as these are sites of robust immune processes during infection. The high frequency of monocytes among human cells in the bone marrow likely resulted from selective mobilization of other cell types to the circulation. The number of human monocytes in the spleen of P-SUR was diminished compared to P-DIE mice. CLP sepsis evokes apoptosis of human cells in the humanized mice similar to the phenomenon observed in septic patients (15, 16). Different responses of monocytes at these two sites can be attributed to the local tissue microenvironment and local production of monocytes in the bone marrow (33). We observed a significant drop of HLA-DR and CD80 expression on monocytes from humanized mice post-CLP but no difference in HLA-DR expression was apparent between P-DIE and P-SUR mice. The lack of difference in HLA-DR expression between groups of opposing outcome was surprising; HLA-DR decrease was shown to be a marker of immunosuppression and it correlated with the magnitude of inflammatory response (34). In line with our findings, however, it has been demonstrated that changes in the HLA-DR expression did not constitute an accurate marker of mortality in septic shock patients (35, 36).

CD80 is one of the co-stimulatory molecules required for T-cell activation by antigen-presenting cells. Nolan et al (29) demonstrated an increased CD80 expression on the circulating monocytes in septic patients and peritoneal macrophages from septic mice (37). Moreover, CD80−/<sup>−</sup> mice had improved survival after CLP (38) and in another study, CLP mice lacking CD80 displayed lower circulating IL-6 levels (38). The above was contrasted by our data: (a) Monocytes from P-DIE humanized mice showed a greater reduction of CD80 expression in comparison to their P-SUR littermates and (b) we observed a negative correlation between the CD80 expression and circulating cytokine concentration. There are several possible reasons for these discrepancies. First, we analyzed CD80 expression on the bone marrow and splenic monocytes, while the other studies focused on the circulating and peritoneal cells. The choice of sampling sites seems to be of vital importance as the immune response in sepsis appears to be highly compartmentalized (39, 40). Second, monocytes from humanized mice were shown to be partly immature, therefore their response to infection can differ from the mature cells (41). Notably, Gille et al. did not find expression of CD80 on monocytes from humanized mice developed on newborn NSG mice transplanted with UCB cells (41). Third, our humanized mice showed a low frequency of human T cells which interacted with monocytes, therefore, the mutual activation of these cells may differ from the normal human setting. Finally, the interspecies interactions in humanized mice are not known and it cannot be ruled out that the murine microenvironment modulates selected responses of human cells. It would be of interest to examine the expression of the interleukin-1 receptor-associated kinase-M in the monocytes of post-CLP humanized mice as this molecule is a master-regulator of CD80 surface expression (38).

We applied small-volume repetitive sampling (31) to monitor the changes of human and murine cytokines in early sepsis. Previously we have demonstrated a simultaneous release of pro-and anti-inflammatory cytokines in the early post-CLP phase (19). Both at baseline and post-CLP, the concentration of measured murine cytokines outbalanced human cytokines. Both human and murine cytokines were upregulated 6 h post-CLP, yet, it was apparent that murine circulating cytokines predominate (concentration-wise) over the human mediators in an approximate 9:1 ratio. This has never been reported before but such a magnitude of discrepancy is not surprising given that the only source of human circulating cytokines in these mice were hematopoietic cells. In contrast, despite humanization, mice maintain several other murine-origin cell types known to secrete cytokines (e.g., endothelium, hepatocytes, myocytes). Most murine cytokines are not active on human cells while some human cytokines are not species-specific (42). The negative correlation between expression of CD80 on human monocytes and concentration of human cytokines suggests that humanized mice recapitulate systemic interactions between humoral and cellular elements of human immunity. Interestingly, there was a robust early peak of human TNFα which markedly outbalanced murine TNFα. As human TNFα cross-reacts with murine receptors (43), it can be speculated that the combined high TNFα load contributes to the increased mortality of humanized mice. The RIP3-mediated induction of necroptosis was shown by Duprez et al. (44) to contribute significantly to early deaths after CLP. We hypothesize that this mechanism was responsible for a greater sensitivity to CLP of humanized mice compared to naïve NSG or SCID mice to CLP. Similarly to our results, humanized mice were shown to be more sensitive to S. aureus infection and CLP than NSG or wild-type mice (16, 45). Ye et al. (16) showed that the high mortality of humanized mice after CLP was related to the production of HMGB-1 by human myeloid cells. It cannot be ruled out that the process of xenotransplantation impairs the murine granulocytic response in humanized mice that is otherwise efficient in the naïve NSG mice. Additionally, it is apparent that humanization did not change the general profile of post-CLP cytokine response in NSG mice. Therefore, the tissues/organs were subjected to the murine inflammatory response, preserving the ability of the host to develop organ dysfunction (recapitulating clinical sequelae).

Next, we examined whether the disease severity modulated the human/murine cytokine response. The relatively high severity (i.e., 71% by 32 h post-CLP) that we employed in our study was dictated by two elements: (a) we sought to provoke robust enough post-CLP changes in both compartments (i.e., human and murine) to enable their effective comparison and (b) a naturally high sensitivity of humanized mice to infection/injury. As early as 6 h post-CLP, all human and murine cytokines (except for human IL-10 and MCP-1) were markedly elevated in the P-DIE mice compared to P-SUR. At 24 h, only the concentration of human/murine TNFα dropped to a similarly low level regardless of outcome. Clinical sepsis data regarding the relationship between TNFα concentration and sepsis mortality are conflicting. However, several studies, in line with our own observations, demonstrated that early TNFα release (i.e., measured close to the onset of sepsis) correlates with mortality (46–48). Importantly, in septic patients, the kinetics of the TNFα response largely vary depending on patient characteristics (47); our CLP model (by default) is much more controlled, and features a relatively steady response. As revealed by the ROC analysis, human cytokines that showed the best predictive value for mortality were IL-8 and IL-6, which is in line with the clinical data (49). However, murine IL-6 and TNFα were characterized by even higher AUC values. This is rational as these cytokines are also produced by nonhematopoietic cells which constitute the majority of tissues in humanized mice.

This study has several limitations. First, our humanized mice show a (human) chimeric heterogeneity. This is unavoidable as chimerism largely depends on the donor (50). However, the heterogeneity of donors partly recapitulates the genetic variability that is present in the clinical scenario. Second, we did not further investigate human monocytes to explain the post-CLP discrepancy in CD80 expression. Third, our model included only a single gender and was characterized by a relatively high CLP severity; the current study should be repeated in males and in a low-severity sepsis scenario. Also, a study in aged humanized mice should be performed as a next step to mimic demographic characteristics of septic patients. Additionally, our study failed to characterize the fluctuations beyond day 2 post-CLP; an assessment of cellular and humoral inflammatory changes in the chronic CLP phase should be attempted. Finally, the body temperature-based prediction of early CLP outcome needs to be verified prospectively given that our assessment used retrospective readouts.

Summarizing, this study characterizes for the first time an outcome-dependent evolution of cellular (i.e., bone marrow and splenic monocytes) and circulating cytokine responses in humanized mice. We demonstrate that humanized mice subjected to CLP recapitulate important inflammatory features of clinical sepsis supporting the notion that this model can be utilized in preclinical sepsis research to maximize its translation potential.

# DATA AVAILABILITY

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

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of ARRIVE guidelines. The protocol was approved by the Local Ethic Committee no IV in Warsaw, Poland.

# AUTHOR CONTRIBUTIONS

TS: design, performing experiment, analysis of the study, and writing manuscript. SD: study design, performing experiment, and correction of manuscript. GH: study design and performing experiments. MJ: sample analysis. KS: cord blood preparation. ZP: cord blood preparation and correction of manuscript. JK: study design and correction of manuscript. MO: study design, performing experiments, data analysis/interpretation, and writing manuscript.

# FUNDING

Study funded by Polish National Science Centre grant no UMO-2016/23/D/NZ6/02554.

# ACKNOWLEDGMENTS

The authors would like to thank Danuta Wasilewska for animal care and James Ferguson for proofreading the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01427/full#supplementary-material

# REFERENCES


associated with progression to severe sepsis and death? Ann Emerg Med. (2000) 35:26–34. doi: 10.1016/S0196-0644(00)70101-6


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

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

# Functional Annotation of Genetic Loci Associated With Sepsis Prioritizes Immune and Endothelial Cell Pathways

Kieu T. T. Le<sup>1</sup> , Vasiliki Matzaraki 1,2, Mihai G. Netea<sup>2</sup> , Cisca Wijmenga1,3, Jill Moser <sup>4</sup> and Vinod Kumar 1,2 \*

*<sup>1</sup> University of Groningen, University Medical Center Groningen, Genetics Department, Groningen, Netherlands, <sup>2</sup> Department of Internal Medicine and Radboud Centre for Infectious Diseases, Radboud University Medical Center, Nijmegen, Netherlands, <sup>3</sup> Department of Immunology, K.G. Jebsen Coeliac Disease Research Centre, University of Oslo, Oslo, Norway, <sup>4</sup> University of Groningen, University Medical Center Groningen, Department of Critical Care and Department of Pathology and Medical Biology, Groningen, Netherlands*

#### Edited by:

*Lukas Martin, University Hospital RWTH Aachen, Germany*

#### Reviewed by:

*Sandra Kraemer, University Hospital RWTH Aachen, Germany Daniel Scott-Algara, Institut Pasteur, France*

> \*Correspondence: *Vinod Kumar v.kumar@umcg.nl*

#### Specialty section:

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

Received: *14 May 2019* Accepted: *01 August 2019* Published: *14 August 2019*

#### Citation:

*Le KTT, Matzaraki V, Netea MG, Wijmenga C, Moser J and Kumar V (2019) Functional Annotation of Genetic Loci Associated With Sepsis Prioritizes Immune and Endothelial Cell Pathways. Front. Immunol. 10:1949. doi: 10.3389/fimmu.2019.01949* Due to limited sepsis patient cohort size and extreme heterogeneity, only one significant locus and suggestive associations at several independent loci were implicated by three genome-wide association studies. However, genes from such suggestive loci may also provide crucial information to unravel genetic mechanisms that determine sepsis heterogeneity. Therefore, in this study, we made use of integrative approaches to prioritize genes and pathways affected by sepsis associated genetic variants. By integrating expression quantitative trait loci (eQTL) results from the largest whole-blood eQTL database, cytokine QTLs from pathogen-stimulated peripheral blood mononuclear cells (PBMCs), publicly available blood transcriptome data from pneumoniae-derived sepsis patients, and transcriptome data from pathogen-stimulated PBMCs, we identified 55 potential genes affected by 39 independent loci. By performing pathway enrichment analysis at these loci we found enrichment of genes for adherences-junction pathway. Finally, we investigated the functional role of the only one GWAS significant SNP rs4957796 on sepsis survival in altering transcription factor binding affinity in monocytes and endothelial cells. We also found that transient deficiency of *FER* and *MAN2A1* affect endothelial response to stimulation, indicating that both *FER* and *MAN2A1* could be the causal genes at this locus. Taken together, our study suggests that in addition to immune pathways, genetic variants may also affect non-immune related pathways.

Keywords: sepsis GWAS, cytokine QTLs, eQTL, functional genomics, PBMC transcriptome, endothelial response, FER locus

# INTRODUCTION

Sepsis is a major global health problem primarily caused by bacterial and fungal infections. It is a life-threatening organ dysfunction characterized by a dysregulated host immune response (1). The global burden of sepsis is high, with an estimated worldwide incidence of more than 30 million cases per year leading to nearly 6 million annual deaths (2). Regretfully, current strategies using a "one-size-fits-all" treatment approach for sepsis have failed because of the extreme heterogeneity in disease outcome (3). It is becoming increasingly clear that the heterogeneity is determined

by impact of multiple risk factors including host genetic variation and pathogens (2). Therefore, identifying the critical genetic factors that affect sepsis patient outcome will help us to unravel genetic mechanisms that determine sepsis heterogeneity.

Up to now, three genome-wide association studies (GWAS) have been conducted to identify risk genes for sepsis. Two GWAS were conducted to identify associations between single nucleotide polymorphisms (SNPs) and 28 day sepsis mortality (4, 5). Another GWAS was conducted in a cohort of extremely premature infants to identify genetic loci associated with sepsis onset (6). However, only one study identified a genome-wide significant association at non-coding SNPs in the intron of Fps/Fes related tyrosine kinase (FER) gene in patients with 28 day survival of sepsis due to pneumonia (4). Although, these studies identified associations with several common polymorphisms, it is unclear how these SNPs affect sepsis outcome. Moreover, which genes and pathways in these loci affect sepsis survival remains to be studied. Identifying these specific genes and pathways is crucial to better understand the molecular mechanisms underlying sepsis heterogeneity.

System genetic approaches have been very effective for many complex human diseases, to translate genetic associations into functional understanding (7). By integrating multiple molecular phenotypes such as gene expression, protein levels, metabolites etc. with SNPs that were associated with human diseases, studies have shown that it is possible to prioritize potential causal genes affected by GWAS SNPs and obtain insights into functional pathways that affect human disease (8). Given the polygenic nature of many complex phenotypes, SNPs that are associated with suggestive significance also provide crucial biological insights. Moreover, as GWAS SNPs function in cell-type and context-dependent manner (9), integrating such context-specific molecular data with sepsis-associated SNPs may be more effective to obtain mechanistic insights into sepsis heterogeneity.

Therefore, in this study, we used pathogen- and celltype specific gene expression levels, cytokine responses, and genotype data from population-based cohorts to integrate molecular responses with sepsis associated SNPs. We show that about 35% of the SNPs affect gene expression (eQTLs) in blood and <30% of sepsis associated SNPs affect cytokine production by peripheral blood mononuclear cells (PBMCs) in response to pathogens. Next, we show that the genome-wide significant SNP rs4957796 in the FER locus affects transcription factor binding efficiency in both monocytes and endothelial cells, and FER and (Mannosidase Alpha Class 2A Member 1 (MAN2A1) MAN2A1 could be the causal genes in this locus via regulating endothelial function.

Taken together, our study provides evidence for genetically determined variability in endothelial pathways, in addition to leukocyte responses, as one of the important factors to explain sepsis heterogeneity. Therefore, more studies on the effect of the SNPs on different pathways such as barrier function or endothelial function are needed.

# MATERIALS AND METHODS

# Identification of Proxy SNPs

Two hundred eighteen proxy SNPs from 39 independent loci were extracted from Haploreg using D′ = 1, R <sup>2</sup> ≥ 0.95 using 1,000 Genome CEU as a reference population.

#### Integration of Suggestive GWAS Loci With eQTL Data and Cytokine QTL Data

We made use of published eQTL data from eQTLgen (http:// www.eqtlgen.org) and in house-cytokine QTL from 500 FG (10). We extracted only genome-wide significant eQTL signals from eQTLgen. Briefly, cis-eQTL analysis was performed in 31,684 blood samples. Significant cis-eQTLs were defined as eQTLs that show FDR corrected P-value of <0.05 (P < 1,829 × 10−<sup>5</sup> ) (11). For cytokine QTLs from the 500 FG cohort study (cytokine QTLs were performed on 392 individuals), we extracted reported Pvalue for each SNP (10) and applied Bonferroni correction to define significant P-value threshold. As we tested 39 independent loci, the P-value threshold is 0.0012.

#### PBMC Transcriptome

We made use of in house PBMC transcriptome data (12). Briefly, PBMCs were freshly isolated from peripheral venous blood withdrawn from healthy volunteers, according to work permission on whole blood (Ethical Committee of Radboud University Nijmegen (nr 42561.091.12). PBMC were freshly isolated with Ficoll-Paque (Pharmacia Biotech), counted (BioRad cell counter), and adjusted to 5 million cells/ml in RPMI 1640 (Gibco, ThermoFisher Scientific), supplemented gentamicin 10 mg/ml, L-glutamine 10 mM, and pyruvate 10 mM. Cells were seeded into wells to settle overnight before stimulation. PBMC were stimulated with 100 µl of Streptococcus pneumonia (ATCC 49619, serotype 19F) (1 million cells/ml), Candida albicans (ATCC MYA-3573, UC 820) (1 million cells/ml), and Pseudomonas aeruginosa (1 million cells/ml). PBMCs were also stimulated with RPMI a negative control. RNA was isolated at 4 and 24 h after stimulation. RNA sequencing was performed in a Nextseq 500 platform, single-end, read length 50 bp. Reads were mapped to the human genome hg19 using STAR (version 2.3.0), Ht-seq count was used to quantify read counts and DEseq2 was used to perform statistics analysis (FDR P ≤ 0.05 and fold change ≥2) (12).

#### Gene Expression in Pneumoniae-Derived Sepsis Patients

To further validate our prioritized genes, we checked their expression levels in pneumoniae-derived sepsis patient transcriptome (N = 265 patients) (13). Briefly, we extracted only genome-wide significant differentially expressed genes (FC>1,5, and FDR correct P-value of 0.05) reported in either the discovery cohort or the validation cohort (13).

#### Electrophoresis Mobility Shift Assay (EMSA)

EMSA were performed using LightShift Chemiluminescent EMSA Kit (Thermo Scientific) according to the manual instruction. In brief, the protocols contain three main parts, including: probe biotination, nuclei extraction, and mobility shift assay on polyacrylamide gel. Probe biotination. Probes containing nucleotide sequence of 30 bp around the SNP were designed carrying either T allele or C allele at the SNP position. The probes were then labeled with biotin at the 3′ end using Pierce Biotin 3′ End DNA labeling kit (Thermo Scientific). After labeled, probes were annealed to make double stranded DNA probes. Labeling efficiency was evaluated following the recommended protocol. Nuclei extraction. Ten million cells were used to isolate the nuclei. Cells were suspended in lysis buffer (10 mM Tris-Cl pH8.0, 300 mM sucrose, 10 mM NaCl, 2 mM MgAc2, 6 mM CaCl2, and 0.2% of NP-40 (Igepal) for 5 min. Nuclei pellets were harvested and resuspended in 100 µl of Nuclear Extract Buffer (20 mM Tris-Cl (pH8.0), 420 mM NaCl, 1.5 mM MgCl2, 0.2 mMEDTA, 25% glycerol, 1 mM DTT and 1X protease inhibitor cocktail) for 10 min on ice. After centrifugation at 14.000 rpm for 15 min, supernatant containing nuclear extract was collected and protein concentration was determined using Bradford assay. Gel mobility assay. Mobility assay was performed according to the instruction. Briefly, 5– 10 µg of total proteins from the nuclei extract was used with 20 fmol of Biotin End-labeled target DNA. Unlabeled target DNA was also used as a binding competition in the presence or absence of protein from nuclei extract. Images were obtained using BioRad system.

#### Cell Culture

To mimic the context of sepsis in which inflammation involves the role of endothelial cells and blood cells, we used Primary Human Umbilical Vein Endothelial cells (HUVEC) (Lonza, The Netherlands) as endothelial cells and THP-1 (ATCC, The Netherlands) as monocytes. Pooled donor-HUVEC were purchased from Lonza (C2519A, The Netherlands). Cells were cultured in EBM-2TM medium (Lonza) supplemented with EGM-2 MV SingleQuot Kit Supplements & Growth Factors (Cat. No. 3202, Lonza) and antibiotics 100 IU/ml of penicillin (Astellas Pharma, The Netherlands) and 50µg/ml of Streptomycin (Rotexmedica GmbH, Germany). Cells were used from passage 3–7 and cultured at 37◦C, 5% CO2 and saturating humidity. THP-1 cells (ATCC, The Netherlands) were cultured in Gibco TM RPMI 1640 containing L-glutamine +/HEPES + (Cat. No. 1640 52400-025) supplemented with 10% (v/v) heat-inactivated FBS (Gibco), 1%(v/v) Pen/Strep 10.000 U/ml (Gibco). THP-1 cells were kept at 37◦C, 5% CO2 and saturating humidity. Cells were freshly passed twice a week to keep a density of 200.000–800.000 cells/ml and used up to passage 28.

#### Knock Down Experiment in HUVEC

HUVEC were seeded to reach the confluency of 70% before transfection. siFER and siMAN2A1 were delivered into HUVEC by Lipofectamine 2000 (Invitrogen). 20 pmol of siRNA sequence was transfected into one million cells according to instructed protocol. After transfection, cells were rested for 48 h before subsequent stimulation with LPS derived from Escherichia coli, serotype O26:B6 (15,000 endotoxin units/g) (Sigma-Aldrich, St.Louis) (1µg/ml) (14). Cells were lysed in Trizol (Ambion, ThermoFisher) and kept at −80◦C until RNA isolation.

# Gene Expression by RT-qPCR

Gene expression levels were measured by RT-qPCR (reverse transcriptase-quantitative PCR) using Sybrgreen platform. Briefly, total RNA was isolated by Trizol according to the instructed protocol. RNA concentration was measured by Nanodrop. RNA quality was controlled in random samples by measuring RNA Integrity Score (Agilent). 100–5,000 ng of total RNA was loaded for cDNA synthesis using ReverAid H Minus First Strand cDNA synthesis kit (ThermoScientific). Primers (refer to **Table 1**) were designed with primer3 and conditions were optimized for each primer set. Melting curves were used to access the specificity of each reaction. GAPDH was used as a housekeeping gene. qPCR was performed in a ViiA7 real-time PCR (Applied Biosystems) following the standard protocol: 15 min at 95◦C and 40 cycles of two steps: amplification (60◦C for 60 s) and denaturation (95◦C for 15 s). Gene expression levels were calculated based on the comparison of CT values between target gene(s) and the housekeeping gene (1CT). Average messenger RNA levels relative to GAPDH from the duplicate were calculated by 2−1CT Data were shown as mean ± SD. One-way ANOVA test was used to compare between conditions and control: P ≤ 0.05 (<sup>∗</sup> ); P ≤ 0.01 (∗∗); P ≤ 0.001 (∗∗∗); P ≤ 0.0001 (∗∗∗∗). GraphPad Prism software (version 6.0) was used to make graphs and determine significant differences.

# RESULTS

# Annotation of 39 Independent Loci From Three Sepsis GWAS

Two genetic studies were conducted to identify SNPs associated with sepsis survival in adult patients (28-day mortality) and one study on sepsis onset in extremely premature infants. We extracted 25 SNPs that are associated with sepsis survival with evidence for suggestive association (P < 10−<sup>5</sup> ), which includes 11 SNPs from Rautanen et al., and 14 SNPs from Scherag et al. study (4, 5). Using the same criteria we extracted 30 SNPs that are associated with sepsis onset in infants from Srinivasan et al. study [(6); **Table S1**]. Among these 55 SNPs, we filtered by locus position, for loci located within 1 Mb from each other, and selected a SNP with the lowest P-value as the representative. As a result, we found 39 independent loci from the three GWAS. We then extracted 218 proxy SNPs (R <sup>2</sup> ≥ 0.95, D′ = 1) for these 39 independent SNPs using 1000 Genome CEU as a reference population (**Table S1**). As previously reported, none of these loci were shared between the three studies. Although, this may be because of the insufficient study power, it also emphasizes the clinical heterogeneity among patients between cohorts, which could be partly determined by genetic variations. Therefore, we followed up these independent loci to prioritize potential causal candidate genes and pathways affected by them.

TABLE 1 | Primer sequences.


#### Expression QTL Mapping and Differential Expression Analyses Prioritized Potential Causal Pathways for Sepsis

To identify potential causal genes affected by sepsis-associated SNPs, we made use of expression-QTL (eQTL) analysis. For this we extracted results from the largest eQTL study (eQTLGen) that included nearly 35,000 blood samples (11). We found significant association of SNPs from 13 independent loci with expression levels of 45 unique genes (**Table 2**). Interestingly, three loci that were associated with sepsis onset in extremely premature infants affected the most number of nearby genes (**Table 2**). In particular, SNPs rs12490944, rs41461846, and rs3844280 affected 14, 10, and 5 genes, respectively.

Moreover, it is shown that differentially expressed genes in response to infectious agents are more likely to be associated with susceptibility to infectious diseases (15) and more than 90% of the lead SNPs that have eQTL effects are located within 100 kb of the eQTL genes (11). Therefore, as a second strategy to prioritize potential causal genes at sepsis-associated loci, we tested the expression levels of all genes located within a 200 kb window of all 39 loci with suggestive association (P < 9.99 × 10−<sup>5</sup> ) in stimulated peripheral blood mononuclear cells (PBMCs) transcriptome. For this, we used RNAseq data from PBMCs that were stimulated with Pseudomonas aeruginosa (P. aeruginosa), Streptococcus pneumoniae (S. pneumoniae) or Candida albicans (C. albicans) for 4 or 24 h. We found that 12 out of 45 cis-eQTL genes (26,67%) were also differentially expressed in at least 1 condition (**Figure 1A**). In addition, we also found another 10 cisgenes, which were not implicated by eQTL mapping as causal genes, to be differentially expressed in at least one of the stimulations in PBMC (**Figure 1B**). In the end, by combining these two strategies, we prioritized 55 potential causal genes for sepsis.

#### Subsets of Prioritized Genes Are Also Associated With Severity of Sepsis

Next, we tested whether some of the prioritized sepsis-associated genes show any correlation with the severity of sepsis. To perform this analysis, we made use of publicly available blood transcriptome data from pneumoniae-derived sepsis patients (13). Out of 55 prioritized genes, we found seven genes that are differentially expressed between severe and mild sepsis patients (**Figure 2**). Among them, expression of CSGALNACT1 is increased in severe patient group whereas KLHDC8B, BCS1L, and NAT6 expression levels were decreased. Interestingly, except CSGALNACT1, all the other six genes were eQTL genes for SNPs associated with sepsis onset. This observation suggests that some of the genes associated with disease onset could also be involved in determining disease severity.

There was no evidence for enrichment of these six genes for particular pathways; however, CYP27A1 and SLC11A1 are known to be involved in sepsis. CYP27A1 is one of the key enzymes involved in synthesizing bile acid in the liver. Studies have shown that CYP27A1 down regulation in sepsis reduce the amount of circulating bile acid, which may be beneficial for sepsis patients (16, 17). SLC11A1 encodes for iron channel, involved in cation metabolism and host resistance to infection. SLC11A1 was shown to be associated with active tuberculosis (18–20). It remains to be tested how these genes contribute to sepsis severity.

#### Around 23% of the Loci Affect Cytokine Production by Leukocytes in Response to Sepsis Causing Pathogens

In addition to a global screening for the effect of 39 suggestive loci on transcriptome response, we also tested their effects in regulating inflammatory cytokine responses, a prominent phenotype in sepsis. We tested if SNPs that are associated with sepsis survival or sepsis onset affect production of cytokines by leukocytes upon stimulation by intersecting our 218 SNPs with cytokine QTL from stimulated PBMCs (10). We found that 9 independent loci affect the production and secretion of six different cytokines in the context of Gram-negative bacteria, Gram- positive bacteria and fungi (**Table 2** and **Figure 3**), albeit with nominal statistical significance (P < 0.05). Only two loci, among these 9 loci, are found to be significantly associated with cytokine production in PBMCs after correcting for multiple testing (P < 0.0012) (**Table 1**). In particular, SNP rs2237499 affected IL-1β levels upon LPS (Gram-negative bacterial infection), whereas SNP rs13380717 altered IFN-γ levels in response to C. albicans hyphae infection. In summary, only around 23% of the sepsis-associated variants affected cytokine production. These results suggest that the other non-cytokine processes are also important for explaining sepsis heterogeneity.

TABLE 2 | Summary table of genes and cytokines of which the expression levels are associated with genetic variations at 39 GWAS suggestive loci. 13/39 loci could alter RNA expression level of 45 nearby genes (cis-eQTL).


#### *(Continued)*

#### TABLE 2 | Continued


*<sup>a</sup>Gene locates within 200 kb window surrounding the suggestive GWAS loci.*

*<sup>b</sup>eQTL genes of which RNA expression levels are differentially expressed in stimulated PBMCs. 9/39 loci could alter cytokine levels upon stimulation (cytokine-QTL).*

#### Sepsis Associated Genes Are Enriched for Adherence Junction Pathway

To test if genes affected by sepsis survival associated SNPs are enriched for particular biological pathways, we made use of Pascal pathway prioritization tool (21). Based on the SNP location, and the P-value of each SNP, the Pascal software will calculate gene score of nearby genes, and the probability of each gene in involving in any signaling pathways. We initially performed gene prioritization and pathway enrichment analyses for each study separately. However, because of less number of loci from each study, we were unable to see strong enrichment of any pathways. We, therefore, combined all 39 independent loci from three studies and performed enrichment analysis. Interestingly, the enrichment analysis showed significant enrichment of genes for adherences-junction pathway (**Figure 4**).

Particularly, the enrichment analysis was based on 36 genes located within 100 kb of 39 independent SNPs. Among those, there are 17 genes that overlapped with the 55 prioritized genes above (data not shown). These findings strengthen the common notion that disruption in barrier, especially vascular wall leakage is a critical process, which lead to organ dysfunction and mortality in sepsis.

#### Regulatory Function of GWAS SNP rs4957796 at FER Locus in Endothelial Cells

We showed that many of the sepsis associated SNPs affect gene expression or alter cytokine levels in response to infections in blood. However, we didn't find any association with expression or with cytokine responses for SNP rs4957796, which is the only genome-wide significant SNP from a GWAS, at FER locus (**Table 2**). This SNP is associated with the survival of pneumoniaderived septic patients. However, how the SNP contributes to the disease severity or which genes are affected by this SNP is not clearly established. Therefore, we conducted experiments in both

FIGURE 2 | Subsets of prioritized genes are also associated with severity of sepsis. Among 55 genes, there are seven genes that are DE in patients (FC > 1.5 and FDR ≤ 0.05). Heat map shows RNA expression levels of seven genes in both discovery and validation cohort. Colors represent expression levels by fold-change between two groups: severe patients SR1 vs. mild patients SR2. Blue, significantly lowly expressed in the severe group; red, significantly highly expressed in the severe group; white, non-significantly different between the severe and mild groups (13).

immune cells and endothelial cells (HUVEC), which play central roles in sepsis pathogenesis (22).

To gain further insight into the function of this SNP, we tested if the SNP could alter the binding site of transcription factors. The alteration of nucleotide composition can lead to changes in the binding of these transcription factors, hence, affecting expression levels of genes. Based on weight matrix prediction (23) this SNP is located in the binding motif of several transcription factors (**Figure S1**). Next, we tested the expression of these transcription factors both in stimulated PBMCs and endothelial cells. We found that ARID5A, E4BP4, HLF, Jundm2, and Ncx\_2 differentially expressed in PBMCs upon stimulation. On the other hand, these transcription factors (ARID5A, BBX, E4BP4, FOXL1, Jundm2, Mef2, TBP, and p300) were expressed in endothelial cells, yet the expression levels were not altered by stimulation of IL1β, TNFα, or LPS. Next, we performed electrophoresis molecular shift assay (EMSA) to validate if the SNP can alter binding affinities of transcription factors in endothelial cells (HUVEC) and monocytes (THP-1). We found that the alteration of T (the risk allele) to C allele (the alternative allele) resulted in changes in the competition of at least two transcription factors in binding to the locus (**Figure 5A**). The effects were shared between both cell types. These findings indicated that the genome-wide significant SNP at FER locus could alter the binding of transcription factors in endothelial cells as well as in monocytes to influence the expression of cis-genes. Therefore, future studies should generate large scale endothelial cell gene expression data upon relevant stimulations to establish the link between sepsis associated SNPs and cis-genes.

# Both FER and MAN2A1 Alter Endothelial Cell Responses to Stimulation

Previous studies have speculated that FER could be a potential causal gene at this locus (4). However, the expression levels

of this gene in blood of sepsis patients did not show any correlation with the severity of sepsis (13). As SNPs can alter expression levels of multiple cis-genes, we tested if the expression of other nearby genes are associated with the disease severity using the data from Davenport et al. (13) and found MAN2A1 to be differentially expressed between the two patient groups. We first stimulated endothelial cells with different infectious agents representative of Gram-negative bacterial antigen (LPS), Gram-positive bacteria (Streptococcus pneumoniae), and Fungus (Candida albicans). We observed strong activation of endothelial cells by LPS but not by other stimuli (**Figure S2**). We therefore focused only on LPS stimulation for knockdown experiments. We then performed transient knockdown experiments on both FER and MAN2A1 genes in endothelial cells using gene-specific siRNAs. Interestingly, both FER and MAN2A1 deficiency in HUVEC altered the cell response to LPS stimulation. We found that the knockdown of MAN2A1 showed stronger effect on the expression of both adhesion molecules (E-selectin and ICAM-1) and cytokine genes (IL-8) (**Figure 5B**). Although, it is still needed to establish the connection between SNP and these two genes, these preliminary results highlight the role of more than one causal gene at this locus.

# DISCUSSION

Host genetic variation is an important factor in explaining susceptibility to infectious diseases in general, and sepsis heterogeneity in particular. Up to now, three genome-wide association studies on sepsis have been conducted. However, due to limited sepsis patient cohort size and extreme heterogeneity,

only one significant locus was identified by a GWAS. Nevertheless, the suggestive associations implicated by these three studies may provide novel insight into genes and pathways that are relevant for understanding sepsis heterogeneity.

In this study, we took advantage of existing molecular data and integrative functional genomics approach to reveal potential causal genes and pathways associated with sepsis heterogeneity. Firstly, we show that <30% of the sepsis associated loci affect cytokine production in response to pathogens. Some of these cytokine-affecting SNPs may be regulated via their effect on expression levels of its nearby genes (eQTL genes). For example, a WLS gene is located in cis-region of a SNP that affects IFN-γ production in PBMCs in response to Candida conidia (**Table 1**). In NK T cells, it is shown that the WLS gene can activate IFN-γ production independent of Wnt/B-catenin pathway (24). Another SNP that is associated with IL17 and IL6 levels upon Candida albicans conidida stimulation in PBMCs is close to AOAH gene (**Table 2**). AOAH codes for acyloxyacyl hydrolase that can deacylates and inactivate LPS, a toxin presented on Gram-negative bacteria wall. Studies have shown that AOAH can drive TH17 T cell differentiation via secreting IL-6 in mice (25). Therefore, it is likely that some of these genes may affect sepsis via regulating cytokine levels in response to infections.

On the other hand, it is possible that because of the lack of sufficient statistical power in these studies, some of these associations could be false positive findings. Nevertheless, it is interesting to observe that more than 70% of the loci were not correlated with cytokine levels suggesting the role of other functional pathways in sepsis. In concordance with this we also show that, by applying PASCAL gene prioritization tool, cisgenes are enriched for adherens junction pathway. However, pathway enrichment analysis on only eQTL genes did not reveal any pathways. It may be due to the fact that genetic effects on gene expression can be very tissue and stimulation specific (7). Therefore, the expression quantitative trait analysis in healthy blood samples may not reflect the effect of sepsis-associated genetic variants. More studies are needed to investigate the effect of genetic variants on different pathways such as coagulation, blood pressure, barrier dysfunction, and vascular leakage that are pivotal for sepsis pathogenesis. Our EMSA assays on a SNP located within FER locus also suggested that some of these sepsis associated SNPs may affect more than one causal genes. Therefore, these factors need to be taken into account when we establish causal genes from association studies. Nevertheless, eQTL mapping shows that 33% suggestive sepsis-associated loci can affect expression levels of 55 potential causal genes and some of these genes are differentially regulated in patients with severe sepsis compared to mild sepsis patient group. These genes are of interest to perform further functional studies to understand their role in sepsis onset and survival.

Our study also has several limitations. When we compared the sepsis associated SNPs from all three GWAS, we found that none of the SNPs were replicated in each other's study. This could be either due to the limited sample size and/or the extreme heterogeneity among sepsis patients caused by several factors including age of patients, type of infectious agents, clinical treatments etc. Therefore, in the future, a large-scale meta-analysis on stratified groups of sepsis patients should be done to identify genetic variations determining sepsis onset, sepsis severity or sepsis mortality. Moreover, to overcome the heterogeneity of sepsis, GWAS on sepsis-associated phenotypes such as vascular leakage, hypertension, organ damage will also be informative to gain further insights into sepsis endo-phenotypes. Secondly, eQTL mapping results were extracted only from whole blood of healthy individuals in this study. Given the prominent role of endothelial and other cell types in sepsis, future studies should focus on generating tissue and context-specific gene expression data to reveal causal genes for sepsis.

To conclude, our approach in this study provides evidence for genetically determined variability in endothelial pathways, in addition to leucocyte responses, as one of the important factors to explain sepsis heterogeneity. Future challenge is therefore to exploit the impact of genetic variation on endothelial cell related processes using both experimental and clinical studies, to develop new treatment options for sepsis.

# DATA AVAILABILITY

Publicly available datasets were analyzed in this study. This data can be found here: 500 FG cytokine: https://hfgp.bbmri.nl/, eQTL: http://www.eqtlgen.org/.

# AUTHOR CONTRIBUTIONS

VK is accredited to the study conceptualization. KL, JM, and VK designed the study. KL performed experiments. KL and VM analyzed the data. MN, JM, and CW provided reagents, protocols, and facilities to conduct experiments. KL and VK prepared the manuscript. MN, CW, JM, and VK interpreted results and critically assessed the manuscript.

#### FUNDING

This work was supported by the Ph.D. fellowship by the Graduate School of Medical Science, UMC Groningen to KL, Radboud UMC Hypatia Tenure Track Grant and a Research Grant [2017] of the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) to VK and NOCI grant from NWO gravitation grant to CW and Spinoza grant of the Netherlands Organization for Scientific Research to MN.

# ACKNOWLEDGMENTS

We thank all the volunteers for donating PBMC for this study. We are grateful to authors of previous studies of whose the dataset were publically available and facilitate our studies. We thank our colleagues within the Genetics department and the EBVDT group for fruitful discussion.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01949/full#supplementary-material

#### REFERENCES


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

Copyright © 2019 Le, Matzaraki, Netea, Wijmenga, Moser and Kumar. 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.

# Midkine Is Elevated After Multiple Trauma and Acts Directly on Human Cardiomyocytes by Altering Their Functionality and Metabolism

Ina Lackner <sup>1</sup> , Birte Weber <sup>1</sup> , Meike Baur <sup>1</sup> , Melanie Haffner-Luntzer <sup>2</sup> , Tim Eiseler <sup>3</sup> , Giorgio Fois <sup>4</sup> , Florian Gebhard<sup>1</sup> , Borna Relja<sup>5</sup> , Ingo Marzi <sup>5</sup> , Roman Pfeifer <sup>6</sup> , Sascha Halvachizadeh<sup>6</sup> , Miriam Lipiski <sup>7</sup> , Nikola Cesarovic<sup>7</sup> , Hans-Christoph Pape<sup>6</sup> , Miriam Kalbitz <sup>1</sup> \* and TREAT Research Group

#### Edited by:

Lukas Martin, University Hospital RWTH Aachen, Germany

#### Reviewed by:

Juerg Hamacher, Lindenhofspital, Switzerland Sergio Iván Valdés-Ferrer, Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán (INCMNSZ), Mexico Sandra Kraemer, University Hospital RWTH Aachen, Germany

> \*Correspondence: Miriam Kalbitz miriam.kalbitz@uniklinik-ulm.de

#### Specialty section:

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

Received: 29 April 2019 Accepted: 29 July 2019 Published: 21 August 2019

#### Citation:

Lackner I, Weber B, Baur M, Haffner-Luntzer M, Eiseler T, Fois G, Gebhard F, Relja B, Marzi I, Pfeifer R, Halvachizadeh S, Lipiski M, Cesarovic N, Pape H-C, Kalbitz M and TREAT Research Group (2019) Midkine Is Elevated After Multiple Trauma and Acts Directly on Human Cardiomyocytes by Altering Their Functionality and Metabolism. Front. Immunol. 10:1920. doi: 10.3389/fimmu.2019.01920 <sup>1</sup> Department of Traumatology, Hand, Plastic, and Reconstructive Surgery, Center of Surgery, University of Ulm, Ulm, Germany, <sup>2</sup> Institute of Orthopedic Research and Biomechanics, University of Ulm, Ulm, Germany, <sup>3</sup> Department of Internal Medicine I, University of Ulm, Ulm, Germany, <sup>4</sup> Institute of General Physiology, University of Ulm, Ulm, Germany, <sup>5</sup> Department of Trauma, Hand and Reconstructive Surgery, Goethe University Frankfurt, Frankfurt, Germany, <sup>6</sup> Department of Trauma, University Hospital of Zurich, Zurich, Switzerland, <sup>7</sup> Department of Surgical Research, University Hospital of Zurich, Zurich, Switzerland

Background and Purpose: Post-traumatic cardiac dysfunction often occurs in multiply injured patients (ISS ≥ 16). Next to direct cardiac injury, post-traumatic cardiac dysfunction is mostly induced by the release of inflammatory biomarkers. One of these is the heparin-binding factor Midkine, which is elevated in humans after fracture, burn injury and traumatic spinal cord injury. Midkine is associated with cardiac pathologies but the exact role of Midkine in the development of those diseases is ambiguous. The systemic profile of Midkine after multiple trauma, its effects on cardiomyocytes and the association with post-traumatic cardiac dysfunction, remain unknown.

Experimental Approach: Midkine levels were investigated in blood plasma of multiply injured humans and pigs. Furthermore, human cardiomyocytes (iPS) were cultured in presence/absence of Midkine and analyzed regarding viability, apoptosis, calcium handling, metabolic alterations, and oxidative stress. Finally, the Midkine filtration capacity of the therapeutic blood absorption column CytoSorb ®300 was tested with recombinant Midkine or plasma from multiply injured patients.

Key Results: Midkine levels were significantly increased in blood plasma of multiply injured humans and pigs. Midkine acts on human cardiomyocytes, altering their mitochondrial respiration and calcium handling in vitro. CytoSorb®300 filtration reduced Midkine concentration ex vivo and in vitro depending on the dosage.

Conclusion and Implications: Midkine is elevated in human and porcine plasma after multiple trauma, affecting the functionality and metabolism of human cardiomyocytes in vitro. Further examinations are required to determine whether the application of CytoSorb®300 filtration in patients after multiple trauma is a promising therapeutic approach to prevent post-traumatic cardiac disfunction.

Keywords: polytrauma, cardiac dysfunction, fracture treatment, damage associated molecular pattern, toll-like receptor, toll-like receptor signaling, prevention cardiac injury, CytoSorb® 300

# INTRODUCTION

According to the World Health Organization (WHO), trauma accounts for 10% of deaths and 16% of disabilities worldwide (1). Multiple trauma in humans (Injury Severity Score, ISS ≥ 16) are characterized by a massive release of different inflammatory biomarkers, such as cytokines, and damage associated molecular patterns (DAMPs). This damage affects different organs of the body and can trigger whole-body inflammation after trauma (2, 3). A substantial release of these trauma-dependent molecules is associated with the development of the so-called systemic inflammatory response syndrome (SIRS) and the multiple organ dysfunction syndrome (MODS), which are both associated with an increased mortality (4, 5). Many of the released inflammatory cytokines and DAMPs were recently shown to be cardiodepressive by acting on cardiomyocytes (CMs), altering their calcium handling, redox balance, signaling transduction, and finally resulting in post-traumatic cardiac dysfunction (6, 7). One inflammatory cytokine is the heparin-binding growth- and differentiation factor Midkine (Mdk). Increased Mdk expression is associated with different traumatic conditions such as bone fracture, burn injury, traumatic spinal cord injury, and sepsis (8–11). Increased Mdk in human blood can persist for overall 42 days after fracture (11). Furthermore, Mdk impairs fracture healing by reducing bone formation and increasing neutrophil infiltration during the fracture healing process (12, 13). However, the trauma-dependent elevation of Mdk in multiply injured patients as well as the exact impact of Mdk on the heart after trauma remains unclear. In patients with chronic heart failure, circulating Mdk increases significantly and is regarded as a novel marker, predicting different cardiac events (14, 15). Moreover, Mdk plays a role in ischemic heart injury, myocardial infarction and cardiac hypertrophy (16–18). Nevertheless, the function of Mdk in these different pathologies is still controversial, because in some cases such as ischemic heart injury, chronic heart failure and myocardial infarct, Mdk has positive effects by improving cell survival and cardiac function, inducing angiogenesis and reducing detrimental remodeling (17, 19, 20). In contrast, Mdk reduces cellular survival and induces pathological remodeling as well as fibrosis in patients with cardiac hypertrophy (18). Consequently, the exact effect of Mdk on the heart is ambiguous since Mdk can have beneficial and detrimental effects in cardiac pathology. The function of Mdk as an inflammatory cytokine on the heart during trauma especially requires clarification. After all, Mdk might be a potential therapeutic option in cardiac diseases as well as in the treatment and prevention of post-traumatic cardiac injury (21, 22). Mdk has been shown to play an important role in active myocarditis in patients and in experimental autoimmune myocarditis in mice (23). In these instances, Mdk promotes the recruitment of polymorphonuclear neutrophils (PMNs) and the production of neutrophil extracellular traps (NETs) in cardiac tissues, resulting in impaired systolic function (23). Increased activation and recruitment of neutrophils in cardiac tissue were also observed in humans after trauma and in experimental blunt chest trauma models in rats. In addition, it is linked to increased systemic levels of extracellular histones by NETosis, leading to cardiac dysfunction (24, 25).

In this study, we investigate the Mdk elevation in blood circulation after multiple trauma in pigs and humans. We further aim to thoroughly examine the effects of Mdk on human CMs. With regards to therapeutic options for posttraumatic cardiac dysfunction, the study aims to investigate the usage of CytoSorb <sup>R</sup> 300 hemadsorption. In clinical settings, CytoSorb <sup>R</sup> 300 hemadsorption improved the outcome of patients with endotoxemia, necrotizing fasciitis, septic shock, and cardiac surgery (26–29). Furthermore, CytoSorb <sup>R</sup> hemadsorption resulted in immediate hemodynamic stabilization and increased survival rates in patients with multiple organ failure (30). CytoSorb <sup>R</sup> 300 consists of highly porous (styrene-codivinylbenzene) hemadsorbent polymer beads, which can remove substances within 10–60 kDa of molecular weight, such as complement factor 5a, cytokines DAMPs and pathogen associated molecular patterns (PAMPs), from circulating blood (26, 31). Similarly, the high-mobility group box 1 protein (HMGB1) can be removed from blood in a time dependent manner (31). Lastly, the study examines the capacity of CytoSorb <sup>R</sup> 300 to filtrate Mdk, which may be used as a therapeutic approach for preventing and handling post-traumatic cardiac dysfunction.

#### MATERIALS AND METHODS

#### Human Blood Samples

Human plasma from 11 multiply injured patients with a history of acute blunt or penetrating trauma and an ISS ≥ 16 was collected after hospital admission in the University Hospital of the Goethe-University Frankfurt with institutional ethics committee approval (312/10), in accordance with the Declaration of Helsinki and following the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) guidelines (32). All enrolled patients either signed the written informed consent form or written informed consent was obtained from the nominated legally authorized representative of the participants in accordance with ethical standards. Exclusion criteria were the patients being younger than 18 or older than 80 years, presenting severe burn injury, acute myocardial stroke, cancer or chemotherapy, immunosuppressive drug therapy, HIV, infectious Hepatitis, acute CMV infection, and/or thromboembolic events. Control blood samples were collected from healthy volunteers (n = 6, 50:50 female male, no comorbidities). Randomization of the groups was not possible during the sample collection. Blood samples were withdrawn in ethylenediaminetetraacetic acid (EDTA) tubes (Sarstedt, Nürmbrecht, Germany) directly after admission. The samples were kept on ice until centrifugation at 2,100 g for 15 min. Then, the supernatant was collected and stored at −80◦C until assay.

#### Animals

This study presents partial results obtained from a large animal porcine multiple trauma model, conducted by the TREAT research group.

The animal housing and experimental protocols were approved by the Cantonal Veterinary Department, Zurich, Switzerland, under license no. ZH 138/2017, and were in accordance with Swiss Animal Protection Law. Housing and experimental procedures also conformed to the European Directive 2010/63/EU of the European Parliament and of the Council on the Protection of vertebrate animals used for scientific purposes (Council of Europe no. 123, Strasbourg 1985) and to the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, National Research Council, National Academy of Sciences, 2011). Twentyfive male pigs weighting 50 ± 5 kg (Sus scrofa domestica) were included in the study (mean height, snout-tail length: 123,6 cm). Animals were held in a controlled environment with 21 ± 3 ◦C room temperature (50% humidity), with a light/dark cycle of 12 h. Water was available for animals ad libitum. General instrumentation, anesthesia and trauma induction were described previously by Horst et al. (33).

#### Analgesia and Anesthesia

For premedication, pigs received an intramuscular injection with ketamine (20 mg/kg body weight), azaperone (1–2 mg/kg body weight) and atropine (0.1–0.2 mg/kg body weight). Anesthesia was performed by intravenous application of propofol (2,6-diisopropylphenol) (1–2 mg/kg body weight). Anesthesia was maintained during the study period with propofol (5–10 mg/kg/h). Pain medication was ensured by sufentanyl (1 µg/kg/h) perfusion over the whole observation period.

#### Multiple Trauma in Pigs

Analgesia and Anesthesia of the animals was maintained during the whole procedure.

Pigs underwent either multiple trauma (n = 20) or shamprocedure (n = 5). Multiple trauma includes a combination of a penetrating thorax trauma, laparotomy, liver laceration, femur fracture, and hemorrhagic shock (ISS ≥ 27). Control animals underwent sham-procedure (n = 5). Femur fracture was induced by a bolt gun (Blitz-Kernen, turbocut JOBB GmbH, Germany), positioned on the mid third of the left femur. The gun was loaded with cattle-killing cartridges (9 x 17; DynamitNobel AG, Troisdorf, Germany). For introduction of blunt chest trauma, a pair of panels (steel 0.8 cm, lead 1.0 cm thickness) was placed on the right dorsal lower chest. A shock wave was induced by a bolt shot (Blitz-Kerner, turbocut JOBB GmbH, Germany), which was applied onto the panel using cattle-killing cartridges as previously described (34, 35). Midline-laparotomy was performed by exploring the right upper liver lobe. Penetrating hepatic injury was induced by cross-like incision halfway through the liver tissue. After a short period of uncontrolled bleeding (30 s), liver package was performed. Directly after the hepatic package, pressure-controlled and volume-limited hemorrhagic shock was induced by withdrawing blood until a mean arterial pressure (MAP) of 30 ± 5 mm Hg was reached. Maximal withdrawal amounts to 45% of total blood volume. The reached MAP was maintained for 60 min. At the end of the shock period, animals were resuscitated according to established trauma guidelines (ATLS <sup>R</sup> , AWMF-S3 guideline on Treatment of Patients with Severe and Multiple Injuries <sup>R</sup> ) by adjusting FiO<sup>2</sup> and an initial substitution of the withdrawn blood volume with Ringerfundin, fluid maintenance was performed by continuous infusing additional fluids (Ringerfundin, 2 ml/kg body weight/h). Moreover, pigs were rewarmed until normothermia (38.7–39.8◦C) was reached. Sham procedure (n = 5) included instrumentation and anesthesia but without trauma or hemorrhage. The multiple trauma group (n = 20) was randomized in four therapy arms: pigs received either femoral nailing without reaming (n = 5), standard reaming (n = 5), reamed irrigation and aspiration (RIA I) (n = 5) or reamed irrigation and aspiration with reduced diameter and improved control of irrigation and suction (RIA II) (n = 5). In all groups a shortened conventional tibia nail was introduced.

#### Follow-Up and Euthanasia

Hemodynamic parameters were continuously monitored for 6 h. Pigs were euthanized under deep general anesthesia with intravenous Na-Pentobarbital.

This animal model represents a clinically relevant porcine model of severe multiple trauma (pulmonary contusion, extremity injury, liver laceration) with post-traumatic observation period under ICU conditions (33).

#### Sample Collection

Serum and plasma samples were collected at baseline, 4 and 6 h after multiple trauma and kept on ice. After centrifugation (1,500 g for 12 min at 4◦C), serum and EDTA-plasma were removed and stored at −80◦C until analysis. Heart tissue samples were obtained 6 h after resuscitation. Tissue of the superficial and the luminal left ventricle was fixed with 4% formalin, followed by embedding in paraffin. Furthermore, tissue was quick-frozen in liquid nitrogen, followed by storage at −80◦C until analysis.

#### Midkine ELISA

For determination of Midkine in human and porcine plasma, as well as for the CytoSorb <sup>R</sup> 300 experiments, the human Midkine ELISA (R&D Systems, McKinley, MN, USA) was used. All procedures were performed according to manufacturers' instructions. Midkine ELISA was performed by a blinded investigator. Human plasma samples were diluted 1:4 and porcine plasma samples were diluted 1:2.

#### ips-Cardiomyocyte Cell Culture

Human cardiomyocytes (iPS) (Cellular Dynamics, Madison, WI, USA) were cultured for 10 days in maintenance medium at 37◦C and in an atmosphere of 7% CO2, according to manufacturers' recommendations.

#### Binding Analysis of FITC-Labeled Midkine

Fluorescein isothiocyanate (FITC) (Sigma Aldrich, St. Louis, MO, USA) was dissolved in DMSO. Two mg/ml Midkine (Dianova, Hamburg, Germany), dissolved in 0.1 M NaHCO<sup>3</sup> were added to 3 mg/ml fluorescein isothiocyanate (FITC) (Sigma Aldrich, St. Louis, MO, USA) solution and were incubated for 1 h at RT while continuously shaking. Unbound FITC was removed by using SnakeSkin <sup>R</sup> dialysis tube (ThermoScientific, Waltham, MA, USA). For dialysis, 1X phosphate buffered saline was used. Human CMs were seeded at a density of 6.3 × 10<sup>4</sup> cells/cm<sup>2</sup> on ibidi 12-well chamber slides (ibidi, Germany). Afterwards, cells were incubated for 30 and 60 min with 100 ng/ml FITClabeled Midkine. Cells were washed, fixed with 4% formalin and cell nuclei were counterstained using Hoechst (Sigma Aldrich, St. Louis, MO, USA). Cells were mounted with ProLong <sup>R</sup> Gold Antifade Mountant (ThermoScientific, Waltham, MA, USA). Cells were analyzed by blinded investigator by using Axio Imager M.2 microscope (Zeiss, Jena, Germany) and the Zeiss ZEN 2.3 software (Zeiss, Jena, Germany). Images were performed with 40x magnification (N.A. 0.75).

#### Cell Viability Assay

Cell viability was analyzed using Cell Titer-Glo <sup>R</sup> Luminescent Cell Viability Assay (Promega, Madison, WI, USA). Cells were seeded with a density of 6.3 × 10<sup>4</sup> cells/cm<sup>2</sup> on a 96-well plate and treated with different Midkine concentrations (0.05, 0.1, 1µg/ml) for 3 h, or with 1µg/ml for different incubation times (0.5, 1, or 3 h). All procedures were performed according to manufacturers' instructions. For all experiments n = 6.

# Troponin I ELISA

Human CMs were seeded with a density of 6.3 × 10<sup>4</sup> cells/cm<sup>2</sup> on a 24-well plate and treated for 6 h with 100 ng/ml Midkine at 37◦C and 7% CO2. Supernatant was collected and troponin I in supernatant was determined by using Human Cardiac Troponin I ELISA (Abcam, Cambridge, UK). All procedures were performed according to manufacturers' instructions. For all experiments n = 6.

# Caspase-3/7 Assay

Human cardiomyocytes were seeded with a density of 6.3 × 10<sup>4</sup> cells/cm<sup>2</sup> on a 96-well plate and treated with 100 ng/ml Midkine for 6 h at 37◦C. Caspase-3/7 activity in human cardiomyocytes was examined by using Caspase-Glo <sup>R</sup> 3/7 Assay (Promega, Madison, WI, USA). All procedures were performed according to manufacturers' instructions. For all experiments n = 6.

# Live Cell Imaging

Live cell imaging was performed using Leica Microscope SP8 and LAS X software (Leica, Wetzlar, Germany). Cells were seeded with a density of 6.3 × 10<sup>4</sup> cells/cm<sup>2</sup> on a 96-well plate and were pre-loaded with 5µM calcium indicator Fluo-3AM (Life Technologies, Carlsbad, CA, USA) and were incubated for 30 min at 37◦C and 7% CO2. After incubation with Fluo-3AM, cells were analyzed immediately. For measurements, cells were placed in special live cell imaging chamber, adjusted at 37◦C and 7% CO2. Cells were incubated with 100 ng/ml Mdk for 30 min and calcium signals were recorded and evaluated by using LAS X software. Cell culture medium was used during measurements. Live cell imaging was performed with 63x magnification (N.A. 1.2, water). Calcium peaks were determined and compared to baseline values. For all experiments n = 6.

# Calcium Measurements

For calcium measurements, human cardiomyocytes (iPS) were seeded with a density of 6.3 × 10<sup>4</sup> cells/cm<sup>2</sup> on ibidi 8 well chambers (ibidi, Germany). Before the measurements, cells were incubated with 100 ng/ml Midkine 60 min before the start of the experiments, as well as for the duration of the experiment. For measurement of changes in intracellular Ca2<sup>+</sup> concentration, cells were loaded with 5µM Fura-2 (ThermoScientific, Waltham, MA, USA) for 30 min (in presence of pharmacological compounds if needed). After incubation, cells were washed twice with bath solution (in mM: 140 NaCl; 5.4 KCl; MgCl2; 1.8 CaCl2; 5.5 Glucose; 5 Hepes; pH = 7.4). Fluorescence imaging was performed on a Cell Observer inverse microscope (Zeiss, Jena, Germany). Cells were illuminated for 90 min at a rate of 2 Hz at each excitation wavelength (340 and 380 nm). Images were acquired using MetaFluor (Molecular Devices, Ismaning, Germany). Cells were measured in bath solution using 40x magnification (N.A. 1.3) at room temperature. Fura-2 ratios were calculated with ImageJ and the data obtained were analyzed with the Matlab script PeakCaller (36). For all experiments n = 6.

#### RNA Isolation

For qPCR experiments, human CMs were seeded at a density of 6.3 × 10<sup>4</sup> cells/cm<sup>2</sup> on a 24-well plate and were treated with 100 ng/ml Midkine for 6 h at 37◦C and 7% CO2. Cells were lysed with RLY lysis buffer (Meridian Bioscience, Cincinnati, OH, USA), containing 10 µl/ml βmercaptoethanol (Sigma Aldrich, St. Louis, MO, USA). RNA isolation from cell lysates was performed by using ISOLATE II RNA Mini Kit (Meridian Bioscience, Cincinnati, OH, USA). Remaining DNA was digested by DNase I (Meridian Bioscience, Cincinnati, OH, USA) for 15 min at RT as recommended by the manufacturer.

#### Reverse Transcribed Quantitative Polymerase Chain Reaction (RT-qPCR)

The respective RNA samples were reverse transcribed in cDNA using SuperScript <sup>R</sup> IV VILO <sup>R</sup> MasterMix (Life Technologies, Carlsbad, CA, USA). For cDNA transcription, 1–5 ng/ml mRNA were used, and experiment was performed according to manufacturer's instructions. For quantitative PCR, the PowerUp <sup>R</sup> SYBR <sup>R</sup> Green Master Mix (Applied Biosystems, Waltham, MA, USA) was used. All procedures were performed according to the manufacturers' instructions. For qPCR, the QuantStudio3 (Applied Biosystems, Waltham, MA, USA) system was utilized. Five-hundred to seven-hundred ng/ml cDNA were used for quantitative PCR. Quantitative mRNA expression of human troponin I (for: 5′ -CCTCCAACTACCGCGCGCTTAT-3′ , rev: 5′ -CTGCAATTTTCTCGAGGCGG-3′ ), sarco/endoplasmic reticulum Ca2+-ATPase (SERCA2a) (for: 5′ -CTCCTTGCCCGT GATTCTCA-3′ , rev: 5′ -CCAGTATTGCAGGTTCCAGGT-3 ′ ), ryanodine receptor 1 (RyR1) (for: 5′ -GGGTTCCTGCCC GACATGAG-3′ , rev: 5′ -GCACAGGTAGCGGTTCACG-3′ ), Na+/Ca2<sup>+</sup> exchanger (NCX) (for: 5′ -GCCTGGTGGAGATGAG TGAG-3′ , rev: 5′ -ACAGGTTGGCCAAACAGGTA-3′ ), toll-like receptor 4 (TLR4) (for: 5′ -CCTGCGTGGAGGTGTGAAAT-3′ , rev: 5′ -CTGGATGGGGTTTCCTGTCAA-3′ ), toll-like receptor 9 (TLR9) (for: 5′ -AGACCTGAGGGTGGAAGTGT-3′ , rev:

FIGURE 1 | Midkine levels in blood plasma of multiply injured humans and pigs. Midkine levels (pg/ml) in shock room blood plasma from multiply injured patients compared to healthy control group (n = 10) (A). Midkine levels (pg/ml) in blood plasma of multiply injured pigs (B). The pigs' femur fracture was either treated with femoral nailing (nailing, n = 5), conventional (conv, n = 5), or with reamer irrigator aspirator 1 or 2 (RIA I or RIA II, each n = 5). Control animals received sham-procedure (n = 5). Evaluation of blood plasma at baseline (BL) and 6 h after trauma. Results are presented as mean ± SEM. Data were analyzed by one-way ANOVA followed by Dunnett's or Tukey's multiple comparison test. Results are significant \*p < 0.05.

5 ′ -CTGGATAGCACCAGTAGCGG-3′ ) and purigenic receptor subtype 7 (P2X7) (for: 5′ -CACACCAAGGTGAAGGGGAT-3′ , rev: 5′ -GGTGTAGTCTGCGGTGTCAA-3′ ) was examined and calculated by the cycle threshold method 11Ct. Respective genes were normalized to expression of the housekeeping gene glutaraldehyde-phosphate dehydrogenase (GAPDH) (forward: 5 ′ -TCTCTGCTCCTCCTGTTCGAC-3′ , reverse: 5′ -CCAA TACGACCAAATCCGTTGA-3′ ) in order to exclude variations. Quantitative mRNA expression was determined by the doublethreshold method (11CT). Results are presented as mean fold change. For all experiments n = 6.

#### Reactive Oxygen Species (ROS)

For analysis of cellular ROS, human CMs were seeded at a density of 6.3 × 10<sup>4</sup> cells/cm<sup>2</sup> on ibidi 12-well slides (ibidi, Germany). Human CMs were treated with 100 ng/ml Midkine for 6 h at 37◦C and 7% CO2. After treatment, cells were incubated for another 30 min with 5µM CellROX <sup>R</sup> Deep Red Reagent (Life Technologies, Carlsbad, CA, USA) at 37◦C and 7% CO2. Afterwards, cells were fixed with 4% formaldehyde and cell nuclei were stained with Hoechst. Cell were mounted with ProLong <sup>R</sup> Gold Antifade Mountant. Cells were investigated by blinded investigator by fluorescence microscopy using Axio Imager M.2 microscope and the Zeiss ZEN 2.3 software. Imaging was performed by using 20x magnification (N.A. 0.5). Relative amount of reactive oxygen species was determined by Zeiss ZEN 2.3 software in order to exclude variations. For all experiments n = 6.

#### Mitochondrial Respiration With Seahorse XF Analyzer

Mitochondrial respiration was analyzed by using the Seahorse XFe96 Analyzer (Agilent Technologies, Santa Clara, CA, USA). This extracellular flux analyzer makes it possible to perform highly accurate real-time measurements of cellular metabolism in living cells by simultaneously quantifying the rates of extracellular acidification (ECAR) and oxygen consumption (OCR), and measuring the glycolysis and the mitochondrial respiration of the cells. For the analysis of mitochondrial respiration, the Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies, Santa Clara, CA, USA) was used. The Seahorse XF Cell Mito Stress Test Kit is an optimized solution for assessing mitochondrial function. During the experiment, the ECAR and the OCR were continuously measured, gaining the parameter for the basal (baseline) respiration of the mitochondria. Afterwards, 2µM oligomycin, 1µM carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP), and 0.5µM antimycin A and rotenone were pneumatically injected into the media of the cells. After automatically and gently mixing, the OCR and the ECAR were measured at multiple times after each injection. After the experiment, cells were fixed with 4% formalin at 4◦C overnight. Then, cells were stained with 0.3% Janus-Green solution (Sigma Aldrich, St. Louis, MO, USA), washed and resolved with 0.5 M hydrochloric acid. Optical density was measured at 630 nm and OCR values were normalized to OD 630 nm values to exclude variations. Results were evaluated using Seahorse Wave 2.4 software (Agilent Technologies, Santa Clara, CA, USA), gaining the parameter for spare respiratory capacity of the mitochondria. For the analysis of mitochondrial respiration, cells were seeded with a density of 5 × 10<sup>5</sup> cells/cm<sup>2</sup> on Seahorse XFe96 analyzer cell culture plates (Agilent Technologies, Santa Clara, CA, USA) and incubated for 6 h with 100 ng/ml Midkine and the above- mentioned procedure was performed. For all experiments n = 6.

#### CytoSorb® 300 Experiments

For the therapeutic experiments, the CytoSorb <sup>R</sup> 300 was used (CytosorbensInc., MonmouthJunction, NJ, USA). Therefore, small columns were prepared. An excess of CytoSorb <sup>R</sup> 300 at the ratio 2:1 (CytoSorb <sup>R</sup> 300 to plasma samples) was added on the column as recommended by the manufacturers. Human shock room blood plasma samples were added on the columns and were

incubated for 6 or 3 h at RT while continuously shaking. For timedoses experiments, different Midkine concentrations (10,000, 5,000, 2,500, 2,000, 1,500, 1,000, 500, 1,000 pg/ml) diluted in PBS with 1% BSA were added on the columns and were also incubated for 6 and 3 h at RT, while continuously shaking. For all experiments n = 6.

#### Statistical Analysis

All values were expressed as means ± SEM. Data were analyzed by one-way ANOVA followed by Dunnett's or Tukey's multiple comparison test. For the statistical analysis of two groups, unpaired two-tailed students t-test was used. p ≤ 0.05 was considered statistically significant. GraphPad Prism 7.0 software was used for statistical analysis (GraphPad Software, Incorporated, San Diego, CA, USA).

#### RESULTS

#### Midkine Plasma Levels in Multiply Injured Humans and Pigs

In humans as well as in pigs, the blood plasma concentrations of Midkine increased after multiple trauma compared to the healthy controls (**Figures 1A,B**). Animals submitted to

reamed femoral nailing showed significantly higher Mdk levels when compared with pigs treated with conventional femoral nailing or with reamer irrigator aspirator treatment (RIA I/II; **Figure 1B**). This indicates that Mdk levels correlate with the invasiveness of the reaming method. In multiply injured pigs, plasma Mdk levels increased significantly after 6 h in the group with conventional reaming of the fracture compared to the control group.

Since plasma Mdk levels increased after multiple trauma, we investigated whether Mdk affects human cardiomyocytes (CMs). After 30 and 60 min the Mdk was actively absorbed into the human CMs and was primarily located around their nucleus in vitro (**Figure 2A**).

# Cell Viability, Cell Damage, and Calcium Handling of Human Cardiomyocytes

Given that Mdk is actively taken into the cells, we examined whether it then affects the cell viability of the human CMs. The cell viability of the human CMs was neither affected by different Mdk concentrations nor by different incubation times (**Figures 2B,C**). Furthermore, there were no differences in troponin I concentrations in supernatant of the humans CMs treated with Mdk compared to control cells after 6 h (**Figure 2D**). However, the calcium handling of the human CMs was altered after Mdk treatment, which is exemplified by the significant increase in their delta calcium peaks (**Figure 3A**), meaning the cells beat slower in presence of Mdk. Moreover, the frequency of calcium signals in human CMs decreased significantly in presence of Mdk, developing bradycardic conditions (**Figure 3B**), which is also demonstrated the traces of the calcium signals of the cells (**Figures 3C–F**).

#### Gene Expression of Human Cardiomyocytes

We showed that Mdk alters the calcium handling in human CMs. Next, we investigated the gene expression of specific cardiac calcium pumps as well as the expression of different receptors, which might be involved in Mdk signaling. In human CMs, the mRNA expression of SERCA2a, NCX, TLR4, TLR9, and P2X7 increased significantly in presence of Mdk compared to control (**Figures 4A,C–F**), indicating for direct effects of Mdk on gene expression of calcium handling proteins. Moreover, the effects of Mdk might be mediated via TLR-P2X7 signaling. The mRNA expression of RyR1 was unaffected (**Figure 4B**).

# Mitochondrial Respiration of Human Cardiomyocytes

In addition, we analyzed the effects of Mdk on the mitochondrial respiration of CMs **Figure 5A**. The basal respiration as well as the spare respiratory capacity of the human CMs decreased significantly after the Mdk treatment (**Figures 5B,C**), indicating detrimental effects of Mdk on mitochondrial respiration.

#### Intracellular Reactive Oxygen Species (ROS) and Caspase3/7 Activity

As Mdk alters mitochondrial respiration and ATP production of the cells, we next investigated whether Mdk also affects the redox balance of the human CMs. The amount of ROS did not change in human CMs after being treated with Mdk compared to control cells (**Figure 5D**). Although Caspase3/7 activity increased significantly in human CMs in presence of Mdk (**Figure 5E**), indicating for enhanced apoptosis in the cells.

# Filtration of Midkine by CytoSorb® 300

Because Mdk is elevated in plasma of multiply injured humans and pigs and acts on human CMs, we examined the potential of a therapeutic approach: the absorption capacity of Mdk from human blood by CytoSorb <sup>R</sup> 300. After incubation of different Mdk concentrations with CytoSorb <sup>R</sup> 300, the Mdk levels decreased between 45 and 95% within 6 h (**Figure 6A**). Especially high Mdk concentrations (10,000 pg/ml) were significantly reduced up to 95% after filtration with CytoSorb <sup>R</sup> 300 after 6 h compared to the 3 h incubation (**Figure 6A**). Moreover, Mdk levels in plasma from multiply injured patients were significantly reduced after incubation with CytoSorb <sup>R</sup> 300 (**Figure 6B**).

# DISCUSSION

Our study shows for the first time that Mdk is elevated in blood circulation after multiple trauma. This elevation is similar to other traumatic injuries, suggesting that circulating Mdk may act as a novel inflammatory marker for polytrauma (8, 9, 11). Furthermore, we demonstrated that Mdk acts directly on

reticulum Ca2+-ATPase (SERCA2a) (A), ryanodine receptor-1 (RyR1) (B), sodium-calcium exchanger (NCX) (C), purigenic P2X receptor subtype 7 (P2X7) (D), toll-like receptor 4 (TLR4) (E), toll-like receptor 9 (TLR9) (F). Results are presented as mean ± SEM. Data were analyzed by two-tailed, unpaired students t-test. For all experiments n = 6. Results are significant \*p < 0.05.

human cardiomyocytes in vitro and is actively taken up by these cells, altering their functionality without affecting their viability. We found that Mdk affects the functionality of the human CMs by altering their calcium handling. The delta calcium peaks of the human CMs increased significantly after Mdk treatment, meaning the cells became bradycardic. Moreover, the frequency of the calcium signals in human CMs decreased significantly after Mdk treatment, confirming the bradycardic effect of Mdk. The mRNA expression of the specific cardiac calcium pumps SERCA2a and NCX also increased significantly after Mdk treatment, suggesting direct effects of Mdk on calcium handling in the cells. The location of Mdk around the cell nucleus

of the human CMs confirmed the regulatory effects on cellular gene expression of the calcium handling proteins. Alterations in calcium signals as well as in mRNA expression of SERCA2a and NCX were also described previously in presence of other trauma-related inflammatory biomarkers and DAMPs as well as in different trauma models and during sepsis, nominating Mdk as a powerful cardio-depressive mediator after trauma and during sepsis (25, 37–42). However, cardiac overexpression of SERCA2a in rodents improved cardiac contractility and relaxation, which might also indicate potential protective effects of Mdk in the heart, which would require to be investigated in future studies (43, 44). We also novelly showed that the basal respiration as well as the spare respiratory capacity of the mitochondria of the human CMs decreased significantly, indicating detrimental effects of Mdk on cellular mitochondrial respiration and energy production. Nevertheless, the amount of cytosolic reactive oxygen species (ROS) was not altered in the human CMs in presence of Mdk. Mitochondrial dysfunction was also depicted previously for other trauma-related biomarkers (7, 45, 46). The detrimental and cardio-depressive effects of Mdk on the human CMs might be mediated via the toll-like receptor (TLR) 4, TLR9, and the pyrogenic receptor subtype 7 (P2X7) since the mRNA expression of these receptors was significantly upregulated. All of these receptors have been demonstrated to be involved in the DAMP-associated cardiac signaling pathways in different trauma models (47, 48). The activation of the TLRs results in increased cardiac inflammation, mediated via the nuclear factor κ B (NFκB) (47). This TLR-mediated cardiac

inflammation leads to cardiac injury and finally results in cardiac contractile dysfunction (47–49). Since the mRNA expression of the TLRs was upregulated, the protein expression of these receptors might be increased after Mdk treatment. This might lead to sensitization of the CMs for other systemic circulating DAMPs, such as HMGB1 and extracellular histones, which were elevated after polytrauma, leading to cardiomyocyte dysfunction (7, 24, 50). The P2X7 receptor was also shown to be involved in cardiac contractile dysfunction (51). Interestingly, the Caspase 3/7 activity increased in human CMs after treatment with Mdk, which was demonstrated previously in cardiac tissue in vivo in an experimental polytrauma model in pigs (41). So far, Mdk was described as an anti-apoptotic factor by decreasing caspase activity in other cells, such as neurons and HepG2 cells, which is in accordance to unaffected cell viability in the present study (52, 53). The effects of Mdk on cellular apoptosis of human CMs has not been described, so far. In contrast to other cells, human CMs seem to follow different cellular processes and various signal cascades might be involved in Caspase-3/7 activation and activity when these cells were treated with Mdk. Moreover, this phenomenon could also be time-dependent as we solely investigated the Caspase-3/7 activity after 6 h of Mdk exposure. This observation should be the subject of future studies in

order to understand the specific effects of Mdk on apoptosis of human CMs.

Therapeutic approaches treating post-traumatic cardiac dysfunction are still limited. In this study, we clearly showed that Mdk is elevated in plasma after multiple trauma and is predominantly detrimental on human CMs, causing the development of post-traumatic cardiac dysfunction. As a consequence, we investigated the efficiency of CytoSorb <sup>R</sup> 300 in filtering Mdk from human blood plasma. CytoSorb <sup>R</sup> 300 is an absorption column, composed of porous polymer beads, which is normally used in the intensive care unit (ICU) for septic patients or for patients with SIRS. The filtration potential of CytoSorb <sup>R</sup> 300 for various trauma-associated cytokines and DAMPs was already demonstrated by others (31, 54). Here, we showed for the first time that CytoSorb <sup>R</sup> 300 is able to absorb Mdk dose-dependently, filtering high Mdk concentrations (10,000 pg/ml) up to 95%. Moreover, CytoSorb <sup>R</sup> 300 filtered Mdk from human plasma obtained on admission to the emergency room, making it a very promising therapeutic approach for treatment and prevention of post-traumatic cardiac dysfunction. One huge benefit of using CytoSorb <sup>R</sup> 300 instead of single antibodies for therapy is that CytoSorb <sup>R</sup> 300 is able to filter a high amount of many miscellaneous damage- and inflammation

molecules after trauma and not only a single molecule, which is the case of antibody treatment. Furthermore, filtration of Mdk by CytoSorb <sup>R</sup> 300 might limit other negative effects of Mdk on polytrauma patients, since it was shown that Mdk acts as an inhibitor of fracture healing and that high Mdk serum levels were associated with poor outcome in septic patients. Finally, we found that systemic Mdk is higher after conventional reaming, compared to nailing without reaming and to RIA I/II. Consequently, treatment of the fracture with RIA I/II might be better for fracture outcome as well as for fracture healing after trauma (55). In addition, conventional reaming of the fracture might have other negative effects after trauma (e.g., pulmonary embolism).

One limitation of the study might be the small sample size (n = 6) to investigate different treatment approaches for the femur fracture. Consequently, more experiments are needed to find the best and the least invasive treatment approach. The same applies for a possible correlation between fracture treatment approaches and systemic Mdk levels. Because investigated groups were heterogenous, a bigger number of samples might be helpful to extrapolate the results to a clinical population. Another limitation might be that we only used small columns with Cytosorb <sup>R</sup> 300 polymer beads in our study, trying to mimic the clinical application in ICU. However, as our study was only an experimental approach, clinical studies should be performed, including more patients and larger application approaches of Cytosorb <sup>R</sup> 300. This may help to- confirm the therapeutic potential of Cytosorb <sup>R</sup> 300 for the prevention of post-traumatic cardiac dysfunction by filtering Mdk from human blood in vivo. Furthermore, it is not possible to mimic in vitro the real in vivo inflammatory conditions, which occur after trauma. The presence of many different inflammatory mediators and DAMPs and the activation of different signal cascades in the cells lead to post-traumatic cardiac dysfunction. Consequently, it is not possible to specify these detrimental effects on one single mediator like Mdk.

Taken together, in our study we observed for the first time that Mdk is elevated systemically after multiple trauma in humans and pigs, acting cardio-depressive on human CMs by impairing their calcium handling and mitochondrial respiration capacity in vitro. PlX27/TLR might be involved in mediating these detrimental effects of MdK (**Figure 7**). In the clinical setting, the hemadsorption filter Cytosorb <sup>R</sup> 300 might be a powerful tool to remove cardio-depressive mediators from patients' circulation and therefore help to improve cardiac function.

# DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the supplementary files.

# ETHICS STATEMENT

Human plasma from 11 multiply injured patients with a history of acute blunt or penetrating trauma and an ISS ≥ 16 was collected after hospital admission in the University Hospital of the Goethe-University Frankfurt with institutional ethics committee approval (312/10), in accordance with the Declaration of Helsinki and following the Strengthening the Reporting of Observational studies in Epidemiology (STROBE)-guidelines (32). All enrolled patients signed the written informed consent form themselves or written informed consent was obtained from the nominated legally authorized representative on the behalf of participants in accordance with ethical standards.

The animal housing and experimental protocols were approved by the Cantonal Veterinary Department, Zurich, Switzerland, under license no. ZH 138/2017, and were in accordance with Swiss Animal Protection Law. Housing and experimental procedures also conformed to the European Directive 2010/63/EU of the European Parliament and of the Council on the Protection of vertebrate animals used for scientific purposes (Council of Europe no. 123, Strasbourg 1985) and to the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, National Research Council, National Academy of Sciences, 2011).

# AUTHOR CONTRIBUTIONS

IL, BW, MB, TE, GF, SH, ML, and NC performed the experiments including animal studies, cell culture experiments, microscopic studies, and ELISAs. IL primarily wrote the paper. MH-L, FG, BR, IM, RP, H-CP, and MK contributed to experimental design and data analysis and coordinated the study and supervised financial support for the studies. All authors made substantial contributions to conception and design of the study, participated in drafting the article, and gave final approval of the version to be published.

# FUNDING

This work was conducted in the framework of the CRC 1149 funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—Project number 251293561. This work was also funded by the AO Grant S-16-133T: Effects of standard reaming and RIA techniques on local soft tissue and systemic homeostasis in a porcine trauma model.

#### ACKNOWLEDGMENTS

We kindly acknowledge the Department of Children and Adolescent Medicine, Division of Pediatric Endocrinology and Diabetes for the provision of the Agilent Seahorse XFe96 analyzer. Thanks to all members of the TREAT research group.

# TREAT RESEARCH GROUP

Auner B, Stormann P, Simon TP, Marx G, Haug A, Egerer L, Giensven MV, Huber-Lang M, Tolba R, Reiss K, Uhlig S, Horst K, Teuben M, Almahoud K, Kalbas Y, Luken H, Almahoud K, and Hildebrand F.

#### REFERENCES


peptides mRNA in cultured rat ventricular myocytes. J Int Med Res. (2004) 32:57–61. doi: 10.1177/147323000403200109


receptor 4 gene. Shock. (2008) 30:740–6. doi: 10.1097/SHK.0b013e3181 73f329


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

Copyright © 2019 Lackner, Weber, Baur, Haffner-Luntzer, Eiseler, Fois, Gebhard, Relja, Marzi, Pfeifer, Halvachizadeh, Lipiski, Cesarovic, Pape, Kalbitz and TREAT Research Group. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# IL-17, IL-27, and IL-33: A Novel Axis Linked to Immunological Dysfunction During Sepsis

#### Kristen N. Morrow1,2, Craig M. Coopersmith2,3 and Mandy L. Ford2,4 \*

*1 Immunology and Molecular Pathogenesis Program, Laney Graduate School, Emory University, Atlanta, GA, United States, <sup>2</sup> Department of Surgery, Emory University School of Medicine, Atlanta, GA, United States, <sup>3</sup> Emory Critical Care Center, Emory University School of Medicine, Atlanta, GA, United States, <sup>4</sup> Emory Transplant Center, Emory University School of Medicine, Atlanta, GA, United States*

Sepsis is a major cause of morbidity and mortality worldwide despite numerous attempts to identify effective therapeutics. While some sepsis deaths are attributable to tissue damage caused by inflammation, most mortality is the result of prolonged immunosuppression. *Ex vivo*, immunosuppression during sepsis is evidenced by a sharp decrease in the production of pro-inflammatory cytokines by T cells and other leukocytes and increased lymphocyte apoptosis. This allows suppressive cytokines to exert a greater inhibitory effect on lymphocytes upon antigen exposure. While some pre-clinical and clinical trials have demonstrated utility in targeting cytokines that promote lymphocyte survival, this has not led to the approval of any therapies for clinical use. As cytokines with a more global impact on the immune system are also altered by sepsis, they represent novel and potentially valuable therapeutic targets. Recent evidence links interleukin (IL)-17, IL-27, and IL-33 to alterations in the immune response during sepsis using patient serum and murine models of peritonitis and pneumonia. Elevated levels of IL-17 and IL-27 are found in the serum of pediatric and adult septic patients early after sepsis onset and have been proposed as diagnostic biomarkers. In contrast, IL-33 levels increase in patient serum during the immunosuppressive stage of sepsis and remain high for more than 5 months after recovery. All three cytokines contribute to immunological dysfunction during sepsis by disrupting the balance between type 1, 2, and 17 immune responses. This review will describe how IL-17, IL-27, and IL-33 exert these effects during sepsis and their potential as therapeutic targets.

#### Keywords: sepsis, IL-17, IL-33, critical illness, cytokine, immunological dysfunction, IL-27

#### INTRODUCTION TO SEPSIS AND THE IL-17/IL-27/IL-33 AXIS

Although it was first described centuries ago, sepsis remains a leading cause of morbidity and mortality. While the infectious agent and the organ system(s) impacted can vary between patients, sepsis is characterized by immune dysfunction linked to alterations in systemic cytokine levels and lymphocyte apoptosis (1). The immune response during sepsis was originally thought to proceed through two distinct phases through which an initially hyper-inflammatory immune response shifted toward profound immunosuppression caused by lymphocyte impairment (2). However, this only reflects the phenotype of circulating lymphocytes in some immunocompetent patients (3–6) and does not reflect the immune response in immunocompromised patients (7).

#### Edited by:

*Christoph Thiemermann, Queen Mary University of London, United Kingdom*

#### Reviewed by:

*Sian M. Henson, Queen Mary University of London, United Kingdom Marcin Filip Osuchowski, Ludwig Boltzmann Institute for Experimental and Clinical Traumatology, Austria*

> \*Correspondence: *Mandy L. Ford mandy.ford@emory.edu*

#### Specialty section:

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

Received: *15 June 2019* Accepted: *05 August 2019* Published: *22 August 2019*

#### Citation:

*Morrow KN, Coopersmith CM and Ford ML (2019) IL-17, IL-27, and IL-33: A Novel Axis Linked to Immunological Dysfunction During Sepsis. Front. Immunol. 10:1982. doi: 10.3389/fimmu.2019.01982* In addition, evidence now exists that both pro- and antiinflammatory cytokines are released shortly after sepsis onset (8, 9) and continue to be released in tandem throughout the course of the illness (10–12).

While there have been many positive animal studies demonstrating the beneficial effect of targeting cytokines during sepsis, this has not translated into improvements in clinical treatment; no clinical trials so far have led to an approved therapeutic. The reasons behind this are multifactorial and partially stem from a failure to consider the interaction between individual cytokines and the larger cytokine milieu. In addition, the cytokine milieu varies between septic patients, making it difficult to distinguish any benefit in large studies of heterogenous patients. Although the failure of cytokine-based therapy in septic patients has been disappointing, recent phase I clinical trials (such as IL-7 infusion) have demonstrated the potential benefit of immunomodulation (13). However, before additional cytokines can be considered as therapeutic targets for sepsis, further work needs to be done to define the alterations that occur across the cytokine milieu during sepsis and distinguish how individual cytokines interact and modulate the effects of one another.

Recent work on the cytokines IL-17, IL-27, and IL-33 suggest the presence of a novel cytokine axis during sepsis. IL-17 primarily acts to promote the inflammatory response in mucosal tissue. In humans, serum levels of IL-17 are predictive of the development of sepsis and mortality in poly-trauma patients (14) and mutations in the IL-17A gene are associated with increased susceptibility to infection caused by gram positive bacteria and mortality (15). The immunosuppressive cytokine IL-27 increases in the plasma of many septic patients (16–27) and has been shown to inhibit the differentiation of Th17 cells (16, 18, 19, 28– 33). These results have been recapitulated in diverse models of sepsis in mice (34–38). The blockade of the p28 subunit of IL-27 (38) or depletion of IL-27 using a soluble and recombinant IL-27Rα (34) significantly reduces mortality in the cecal ligation and puncture (CLP) model of sepsis and is associated with reduced bacterial burden in the tissues and blood. IL-33 is a member of the IL-1 family of cytokines that modulates Th2 responses and decreases the differentiation of T cells into Th17 cells (39). IL-33 signals through the cytokine receptor ST2 and plays an anti-inflammatory role during sepsis, improving survival during the early stages of sepsis but ultimately leading to long lasting immunosuppression through the induction of regulatory T cells (Tregs) (40–42). In addition to its interactions with IL-17, IL-33 has also been reported to interact with IL-27, with both modulating the activity of ILC2 cells (43–46). As the importance of ILC2 cells during sepsis has recently been described (47–49), these interactions may become increasingly significant for the development of effective therapeutics.

In this review, we will further discuss the individual and combined roles of IL-17, IL-27, and IL-33 during sepsis and how this axis might be therapeutically targeted.

# The Role of IL-17 in Sepsis Pathophysiology

The IL-17 family of cytokines is composed of the structurally similar IL-17A-F. Apart from IL-17A (classically referred to as IL-17) and IL-17F, all the cytokines in this family are encoded separately, although they share conserved sequences. The earliest studies addressing the role of IL-17A during sepsis in animal models reported that they induced significant pathology and that eliminating IL-17A resulted in significantly improved survival (50, 51). However, subsequent studies using mice deficient in the IL-17 receptor found opposite results (52), and the literature now contains numerous studies demonstrating the mixed effects of IL-17A blockade in sepsis.

In 2012, Ogiku et al. reported that mice lacking IL-17A had significantly increased mortality following CLP that correlated with higher bacteremia at 12 h (53). Similarly, a more recent paper using the CLP model concluded that IL-17 has a partially protective role during sepsis: wild type mice had significantly increased survival and IgA production after CLP when compared to IL-17−/<sup>−</sup> mice (54). Interestingly, this study found that noncanonical signaling through NF-κB was responsible for much of the IL-17A production, as mice lacking RANKL and NF-κB inducing kinase (NIK) signaling in their intestinal epithelium cells had significantly reduced IL-17A and mortality similar to IL-17−/<sup>−</sup> mice (54).

Other studies have found that the impact of IL-17A on sepsis mortality depends on the microbe that initiated the infection. Using a bacterial pneumonia model, Ritchie et al. found that the role of IL-17A in sepsis is highly dependent on the encapsulation status of the infecting bacterium (55). IL-17A was beneficial during infections caused by minimally encapsulated bacteria, but significantly increased lung pathology and mortality if the infectious organism was heavily encapsulated (55). The authors concluded that this was due to the accumulation of neutrophils unable to phagocytose the bacteria (55). In conjunction with IL-23 signaling, IL-17A increases the recruitment of neutrophils and their accumulation in the lung following CLP, partially explaining the inflammation seen in the lung following polymicrobial sepsis originating in other tissues (56). IL-17 has also been linked to the development of acute kidney injury in septic patients and animal models (57). Given these findings, it is not surprising that multiple groups have reported that the neutralization of IL-17A or IL-17F improves survival (58, 59).

As the results of these sepsis studies conflict, it is important to note that IL-17A can induce the production of other IL-17 family cytokines, especially IL-17C (60). Although it is a distinct cytokine, it plays a similar role in neutrophil recruitment and the inflammatory process to IL-17A (60). In a mouse model of pneumonia induced by Pseudomonas aeruginosa, mice lacking IL-17C had 100% survival at 48 h, whereas wild type mice had only 25% survival at this time point (60). In contrast, another recent paper reported that IL-17C induction provides protection against LPS-induced endotoxemia (61). As IL-17C has been reported promote the production of IL-17A by Th17 lymphocytes in inflammatory conditions (such as autoimmune disease) (62), the authors concluded that these effects may be due more to the promotion of IL-17A than to IL-17C alone.

#### The Role of IL-27 in Sepsis Pathophysiology

Originally thought to be pro-inflammatory, there is now consensus that IL-27 is a potent immunosuppressant. It is composed of an alpha subunit (IL-27p28, also known as IL-30) and EBI3 (shared with IL-35) (63). IL-27 binds to the IL-27 receptor alpha (IL-27Rα, also known as WSX-1) and gp130 and is primarily produced by dendritic cells (DCs), monocytes and macrophages (63). The lymphocyte populations that respond to the presence of IL-27 or one of its subunits are T cells, natural killer (NK) cells, natural killer T (NKT) cells, and DCs (64– 69). This allows IL-27 to have wide ranging effects on cells of both the innate and adaptive immune response in addition to autocrine effects.

In septic patients and in murine models of sepsis, the plasma concentration of IL-27 significantly increases (34, 35, 70, 71), briefly causing it to be considered as a potential diagnostic biomarker in adults (22–24) and children (20, 26). However, these results have not been consistently replicated in humans, limiting its current therapeutic potential. In mice, the results are more consistent and indicate a clear role for IL-27 in the pathology of sepsis and critical illness. When the p28 subunit is neutralized or the IL-27Rα is blocked, mortality is significantly decreased in both CLP and endotoxemia (34, 37, 38).

In a study by Cao et al., mice lacking the IL-27Rα were resistant to a secondary bacterial infection caused by Pseudomonas aeruginosa following CLP in a manner dependent on alveolar macrophages and neutrophils (37). Specifically, the neutrophils and alveolar macrophages in these mice had a significantly improved ability to kill P. aeruginosa upon phagocytosis (37). Similarly, Bosmann et al. observed that the oxidative burst of macrophages was improved upon the elimination of IL-27 signaling, and determined that IL-10 limits the production of IL-27p28 in vivo following CLP (38). In addition, this study found that the cells primarily responsible for the production of IL-27p28 in the CLP model of sepsis are splenic macrophages (38). However, a more recent study has found conflicting evidence that indicates a protective role for p28 during sepsis (72). In this study, the administration of the p28 subunit or its overproduction through genetic therapy led to a reduction in mortality during sepsis directly linked to the reduction in NKT cell production of inflammatory cytokines (72).

In addition to its modulation of innate cells, IL-27 has a significant impact on T cells. IL-27 can promote the differentiation of Th1 cells and it is also a potent inducer of type 1 Treg (Tr1) cells (73). While Tr1 cells produce IFN-γ, they also produce large quantities of IL-10 and have potent suppressive functions (74). In addition to the induction of this cell population, IL-27 signaling leads to an increase in co-inhibitory molecule expression on T cells following chronic antigen exposure and during cancer (75). As T cell dysfunction and exhaustion is associated with the development of immunosuppression during sepsis and ultimately worsened survival (76–79), IL-27 could be an effective therapeutic target. However, mice can produce IL-27p28 in the absence of EBI3, so it is unclear if the reported effects of IL-27 during sepsis are actually due to the full heterodimeric cytokine or merely to its alpha subunit. A group has recently reported the development of transgenic B57L/6J mice in which the IL-27p28 subunit cannot be produced independently of EBI3 (80). This animal model will be necessary to truly distinguish the effects of IL-27p28 from those of IL-27.

# The Role of IL-33 in Sepsis Pathophysiology

ST2 was an orphan receptor until 2005, when Schmitz et al. reported their discovery of IL-33 (81). A member of the IL-1 family, IL-33 is constitutively expressed by endothelial and epithelial cells in barrier tissues and is also found at high levels under inflammatory conditions in other tissues (82, 83). When T cells, mast cells, eosinophils, and ILC2s receive IL-33 signaling, the immune response shifts toward a type 2 response (81, 84).

The first paper to describe the role of IL-33 (rather than its receptor ST2) in sepsis was published by Alves-Filho et al. (40). The authors found that survival significantly increased when IL-33 was administered to mice following CLP (40). Another 2010 study found that IL-33 is protective against LPS induced endotoxemia (85). The ability of IL-33 to improve survival during sepsis is linked to the rescue of neutrophil migration to the site of infection (40), to improvements in bacterial clearance, and to a reduction in lymphocyte apoptosis (41). IL-33 also suppresses the inflammatory response by a variety of innate lymphocytes (86) and modulates the activity of ILC2 cells (47–49). In addition to direct effects on other lymphocytes, IL-33 impacts the activity of other cytokines, including IL-17 (41, 87). While IL-33 can bind to a soluble form of ST2 (sST2), the effects of IL-33 during sepsis appear to be dependent on signaling through membrane bound ST2; in one study, patients who had did not survive sepsis had higher levels of sST2 than patients that went on to survive their infections (40).

Despite being linked to improvements in survival early after sepsis onset, IL-33 signaling may not always be beneficial. IL-33 is implicated in the induction and maintenance of immunosuppression during sepsis through the induction of Tregs (42). Nascimento et al. found that this occurs through the production of IL-4 and IL-13 by ILC2s that receive IL-33 signaling (42). The IL-4 and IL-13 then drives the proliferation of IL-10 producing macrophages and ultimately an expansion in Treg numbers (42). When they examined the blood of a small number of patients who had been diagnosed with sepsis 5–10 months prior, they found that sepsis survivors had significantly higher concentrations of both IL-10 and IL-33 and higher circulating Treg numbers compared to previously healthy patients (42). While these findings need to be replicated, they suggest that the impact of IL-33 signaling may depend on the stage of disease.

# INTERACTIONS BETWEEN IL-17, IL-27, AND IL-33

The ability for lymphocytes to recognize and respond to slight changes in their environment makes the immune system very adaptable and ensures that the balance between inflammatory and immunosuppressive responses is fine-tuned. While the ability of lymphocytes to respond so readily to their surroundings is beneficial from an evolutionary point of view, it makes it significantly harder to elucidate the role of individual cytokines. The individual and combined actions of the cytokines in the IL-17, IL-27, and IL-33 axis are summarized in **Figure 1**.

# IL-17 and IL-27

IL-17 plays a harmful role in many autoimmune diseases, particularly experimental autoimmune encephalomyelitis (EAE) and rheumatoid arthritis (RA). By limiting the differentiation of naïve CD4<sup>+</sup> T cells into Th17 cells, IL-27 is able to attenuate these diseases (16, 19, 28, 31, 32, 88). Similarly, IL-27 signaling prevents the development of neurological damage during chronic Toxoplasma gondii infection (18) and reduces tissue damage during RSV infection (89). Further research has shown that STAT1 signaling (which IL-27 induces) inhibits the expression of the transcription factor RORγt, necessary for Th17 differentiation, while promoting the induction of the protein suppressor of cytokine signaling 1 (SOCS1) (16, 19, 30, 90). This leads to the suppression of IL-22 production by Th17 cells, impairing antimicrobial defenses in the epithelium (30, 33). In addition to its direct effects on T cells, IL-27 can also inhibit Th17 differentiation by inhibiting the production of the Th17-polarizing cytokines IL-1β, IL-6, and IL-23 by DCs (29). In contrast, T cells that have already committed to the Th17 lineage are not directly inhibited by IL-27 signaling (28, 91). Instead, inhibition occurs indirectly through the induction of Tr1 (29) and the expression of co-inhibitory receptors and their ligands (32, 88).

# IL-17 and IL-33

Similar to IL-27, IL-33 has been reported to attenuate EAE through the suppression of Th17 responses (92). While IL-33 attenuates sepsis mortality, it is less clear if this is due to any effect on the Th17 response. One group reported that the administration of IL-33 actually enhanced the production of IL-17 while decreasing the levels of IL-6, IL-10, and IFNγ following CLP (41). Similarly, another group found that the deletion of the IL-33 receptor ST2 led to a reduction in the frequency and number of IL-17 producing NK cells after CLP (86). However, a recent study of human patients with Staphylococcus aureus bacteremia revealed that a higher ratio of Th17 to Th1 cytokines early after sepsis onset was associated with increased mortality (49). As there was a trend toward increased Th2 cells in surviving patients, the authors did a follow up study using a mouse model of S. aureus bacteremia (49). IL-33 provided a survival benefit in this model that was dependent on functional ILC2s and EOs, suggesting that IL-33 is protective in part because it re-balances type 1, 2, and 17 responses during sepsis (49).

#### IL-27 and IL-33

While IL-27 signaling promotes type 1 immune responses and directly limits type 17 immunity, it also serves as a negative regulator of type 2 responses by interfering with IL-33 signaling. The first paper to describe this phenomenon utilized in vitro experiments which showed that IL-27 reduced type 2 cytokine production in bone marrow cells exposed to IL-33, including IL-5, IL-13, and GM-CSF (43). For IL-5, this effect was dependent on STAT1 signaling, as STAT1 knockout bone marrow cells were not impacted by the presence of IL-27 (43). Moro et al. confirmed these findings in vivo using STAT1 knockout mice, and revealed that while IL-27 reduces type 2 cytokine production by ILC2 cells, it does not affect cytokine production in Th2 cells (44). Another recent study reported that the administration of IL-27 limits IL-33 induced ILC2 accumulation and activation in the lungs, liver, spleen, and mesenteric lymph node in vivo (45). The administration of IL-27 also led to the overrepresentation of IL-27Rα <sup>−</sup>/<sup>−</sup> cells in chimeric mice (45). While not specifically addressing IL-27, another murine study found that STAT1 signaling induced by infection with respiratory syncytial virus is sufficient to reduce the production of IL-33 (46). These studies collectively show that a major function of IL-27 is to negatively regulate the type 2 immune response, specifically ILC2 cells, in a manner that is dependent on STAT1 signaling.

#### EXPLOITING THE IL-17, IL-27, AND IL-33 AXIS DURING SEPSIS

While many reviews have discussed the therapeutic potential for targeting IL-17, IL-27, and IL-33 during sepsis (93–95), none have considered the effect that treatment targeting only one of these cytokines may have on the others. In addition, the compartmentalization of the immune response during sepsis means that cytokine therapies that restore the function of circulating lymphocytes could cause excessive stimulation and ultimately programmed cell death in the more normally responsive tissue lymphocytes. To aid the specificity of these therapies, binding should be targeted to cells expressing the circulatory chemokine receptor molecule CCR7 or the integrin CD62L (required for lymphocyte extravasation into the lymphatic system).

As IL-17 can have either beneficial or detrimental roles during sepsis depending on the murine model used, it is currently unclear what course of action would be most beneficial for human patients. However, anything that significantly increases IL-17 levels for a long period of time raises the risk of auto-immune disease formation and increased tissue damage. It seems more tenable to target IL-27 and IL-33, with an eye to keeping a balance between these cytokines and IL-17.

Neutralizing IL-27 seems likely to provide a survival benefit in septic patients if administered early after disease onset. IL-27 signaling shifts the balance too far toward a type 1 regulatory response, but its neutralization would balance type 1 and type 2 responses through the increase in activity of the type 2 promoting cytokine IL-33. IL-33 signaling has been shown to improve sepsis survival in the short term in murine models, although one report suggests that IL-33 is linked to the development of immunosuppression during sepsis (42). In this study, mice

#### REFERENCES


lacking the IL-33 receptor had attenuated immunosuppression associated with a reduction in type 2 cytokines, ILC2 cells, and Tregs (42). It is currently unknown how much IL-33 signaling changes during sepsis in the absence of IL-27 signaling and therefore might lower the efficacy of IL-27 blockade in improving long term survival in sepsis patients who receive no further interventions.

Ultimately, targeting any of these cytokines in an indiscriminate fashion is unlikely to be clinically beneficial. However, understanding the complex interplay between IL-17, IL-27, and IL-33—including the timing in which cytokine augmentation or blockade may potentially be beneficial suggests this axis may potentially be manipulatable for therapeutic gain as part of a precision medicine approach toward sepsis treatment.

#### AUTHOR CONTRIBUTIONS

KM wrote the initial draft of the manuscript. The final manuscript includes equal contribution from MF and CC.

#### FUNDING

This work was supported by funding from the National Institutes of Health (GM113228, GM104323, AA027396, GM072808).

with mortality following sepsis: a prospective cohort study. Crit Care Med. (2017) 45:e493–e9. doi: 10.1097/CCM.0000000000002292


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

Copyright © 2019 Morrow, Coopersmith and Ford. 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.

# Endogenous Uteroglobin as Intrinsic Anti-inflammatory Signal Modulates Monocyte and Macrophage Subsets Distribution Upon Sepsis Induced Lung Injury

Andrea Janicova1,2,3, Nils Becker <sup>1</sup> , Baolin Xu<sup>1</sup> , Sebastian Wutzler <sup>4</sup> , Jan Tilmann Vollrath<sup>1</sup> , Frank Hildebrand<sup>5</sup> , Sabrina Ehnert <sup>6</sup> , Ingo Marzi <sup>1</sup> , Philipp Störmann1† and Borna Relja<sup>1</sup> \* †

#### Edited by:

Christoph Thiemermann, Queen Mary University of London, United Kingdom

#### Reviewed by:

Hugo Caire Castro-Faria-Neto, Oswaldo Cruz Foundation (Fiocruz), Brazil Xin Zhou, Tianjin Medical University General Hospital, China Christoph Emontzpohl, University Hospital RWTH Aachen, Germany

> \*Correspondence: Borna Relja info@bornarelja.com

†These authors have contributed equally to this work

#### Specialty section:

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

Received: 14 June 2019 Accepted: 09 September 2019 Published: 01 October 2019

#### Citation:

Janicova A, Becker N, Xu B, Wutzler S, Vollrath JT, Hildebrand F, Ehnert S, Marzi I, Störmann P and Relja B (2019) Endogenous Uteroglobin as Intrinsic Anti-inflammatory Signal Modulates Monocyte and Macrophage Subsets Distribution Upon Sepsis Induced Lung Injury. Front. Immunol. 10:2276. doi: 10.3389/fimmu.2019.02276 <sup>1</sup> Department of Trauma, Hand and Reconstructive Surgery, Goethe University, Frankfurt, Germany, <sup>2</sup> Department of Aquatic Ecotoxicology, Goethe University, Frankfurt, Germany, <sup>3</sup> Department of Radiology and Nuclear Medicine, Experimental Radiology, Otto-von-Guericke University, Magdeburg, Germany, <sup>4</sup> Orthopedic and Trauma Surgery, Helios Horst Schmidt Clinic, Wiesbaden, Germany, <sup>5</sup> Department of Trauma Surgery, RWTH University, Aachen, Germany, <sup>6</sup> Department of Trauma and Reconstructive Surgery, Siegfried Weller Research Institute, BG Trauma Center Tuebingen, Eberhard Karls University Tuebingen, Tuebingen, Germany

Sepsis is a serious clinical condition which can cause life-threatening organ dysfunction, and has limited therapeutic options. The paradigm of limiting excessive inflammation and promoting anti-inflammatory responses is a simplified concept. Yet, the absence of intrinsic anti-inflammatory signaling at the early stage of an infection can lead to an exaggerated activation of immune cells, including monocytes and macrophages. There is emerging evidence that endogenous molecules control those mechanisms. Here we aimed to identify and describe the dynamic changes in monocyte and macrophage subsets and lung damage in CL57BL/6N mice undergoing blunt chest trauma with subsequent cecal ligation and puncture. We showed that early an increase in systemic and activated Ly6C+CD11b+CD45+Ly6G<sup>−</sup> monocytes was paralleled by their increased emigration into lungs. The ratio of pro-inflammatory Ly6ChighCD11b+CD45+Ly6G<sup>−</sup> to patrolling Ly6ClowCD11b+CD45+Ly6G<sup>−</sup> monocytes significantly increased in blood, lungs and bronchoalveolar lavage fluid (BALF) suggesting an early transition to inflammatory phenotypes during early sepsis development. Similar to monocytes, the level of pro-inflammatory Ly6ChighCD45+F4/80<sup>+</sup> macrophages increased in lungs and BALF, while tissue repairing Ly6ClowCD45+F4/80<sup>+</sup> macrophages declined in BALF. Levels of inflammatory mediators TNF-α and MCP-1 in blood and RAGE in lungs and BALF were elevated, and besides their boosting of inflammation via the recruitment of cells, they may promote monocyte and macrophage polarization, respectively, toward the pro-inflammatory phenotype. Neutralization of uteroglobin increased pro-inflammatory cytokine levels, activation of inflammatory phenotypes and their recruitment to lungs; concurrent with increased pulmonary damage in septic mice. In in vitro experiments, the influence of uteroglobin on monocyte functions including migratory behavior, TGF-β1 expression, cytotoxicity and viability were proven. These results highlight an important role of endogenous uteroglobin as intrinsic anti-inflammatory signal upon sepsis-induced early lung injury, which modules the early monocyte/macrophages driven inflammation.

#### SHORT SUMMARY

Blunt chest injury is the third largest cause of death following major trauma, and ongoing excessive pro-inflammatory immune response entails high risk for the development of secondary complications, such as sepsis, with limited therapeutic options. In murine double hit trauma consisting of thoracic trauma and subsequent cecal ligation and puncture, we investigated the cytokine profile, pulmonary epithelial integrity and phenotypic shift of patrolling Ly6ClowCD11b+CD45+Ly6G<sup>−</sup> monocytes and Ly6ClowCD45+F4/80<sup>+</sup> macrophages to pro-inflammatory Ly6ChighCD11b+CD45+Ly6G<sup>−</sup> monocytes and Ly6ChighCD45+F4/80<sup>+</sup> cells in blood, lungs and bronchoalveolar lavage fluid (BALF). Pro-inflammatory mediators and phenotypes were elevated and uteroglobin neutralization led to further increase. Enhanced total protein levels in BALF suggests leakage of respiratory epithelium. In vitro, uteroglobin inhibited the migratory capacity of monocytes and the TGF-β1 expression without affecting the viability. These results highlight an important role of endogenous uteroglobin as an intrinsic anti-inflammatory signal upon sepsis-induced early lung injury, which modulates the early monocyte/macrophages driven inflammation.

Keywords: uteroglobin, CC16, chest injury, acute lung injury, CLP, sepsis, monocytes, macrophages

#### INTRODUCTION

Trauma is one of leading causes of death worldwide (1). Twenty to twenty-five percent of patients with multiple injuries suffer from severe lung contusion, whereas thoracic trauma represents the third most frequent cause of death after major trauma (2, 3). Thoracic injury contributes significantly to the development of acute respiratory distress syndrome (ARDS) (4), while infectious complications such as e.g., sepsis constitute a serious risk factors for up to 50% of mortalities occurring upon ARDS (5). Thus, the development of secondary complications is still a major contributing factor in trauma-associated mortality (6–8).

In general, traumatic injury-related tissue damage induces a release of endogenous damage-associated molecular patterns to initiate the resolution of non-pathogenic and pathogenic inflammation with subsequent tissue repair (9). This systemic inflammatory response syndrome triggers an excessive release of pro-inflammatory cytokines such as tumor necrosis factor (TNF)-α (9), interleukin (IL)-6 (9) and chemokines, such as monocyte chemoattractant protein (MCP)-1 (10, 11), which is a potent factor for monocyte and macrophage migration and infiltration (12).

Monocytes play a pivotal role in pathogen recognition and killing, and although their functional and phenotypic alterations provide a good base for prediction of complications after trauma or of poor prognosis in septic patients, exact pathophysiologic mechanisms and solid biomarkers still remain to be elucidated (13–17). Murine monocytes expressing high levels of lymphocyte antigen (Ly)6C have pro-inflammatory and anti-microbial features and have been shown to be precursors for patrolling monocytes, which survey the vasculature and contribute to the early response of inflammation and tissue repair, and which are characterized by a low expression of Ly6C (18, 19). Next to monocytes, resident alveolar macrophages initiate the inflammatory cascade and secrete pro-inflammatory mediators during acute lung injury (ALI) (20–22). While the proinflammatory M1 macrophages release e.g., nitric oxide, TNFα, interferon-γ and IL-12 and critically contribute to pathogen clearance, their apoptosis during the process of pathogen clearance simultaneously contributes to downregulation of the pro-inflammatory phase and transition of M1 to tissuerepairing M2 macrophages (23). With regards to specific roles of monocytes/macrophages during lung injury and/or sepsis, both beneficial or detrimental effects of each cell type have been reported. While an early depletion of circulating monocytes before lipopolysaccharide (LPS) administration deteriorated lung injury (24), later monocyte depletion ameliorated lung injury (25, 26). Furthermore, macrophage polarization into M1 phenotype improved organ dysfunction and reduced mortality in lethal sepsis (27), while an intratracheal administration of M2 macrophages after CLP reduced mortality (28). These results indicate that an early balance of the pro-inflammatory and the anti-inflammatory response is required for ameliorating lung injury.

Uteroglobin (club cell protein (CC)16), is a 15.8 kDa protein secreted primarily by non-ciliated club cells along the tracheobronchial epithelium, especially in distal respiratory and terminal bronchioles (29, 30). Next to its biomarker character to indicate the development of secondary pulmonary complications after trauma, CC16 exerts anti-inflammatory and immunosuppressive properties (31–33). Its anti-inflammatory

**Abbreviations:** Ab, antibody; ALI, acute lung injury; ARDS, acute respiratory distress syndrome; BALF, bronchoalveolar lavage fluid; CC, club cell protein; CLP, cecal ligation and puncture; HV, healthy volunteers; ISS, injury severity score; LPS, lipopolysaccharide; Ly, lymphocyte antigen; MCP, monocyte chemoattractant protein; MPP, matrix metalloproteinase; RAGE, receptor for advanced glycation endproducts; TNF, tumor necrosis factor; TP, trauma patient; TxT, thoracic trauma.

biology has been confirmed in tracheal epithelial cells, isolated human mononuclear cells and murine macrophages (34–36). Due to this, CC16 has been described as being protective in the development of chronic obstructive pulmonary disease in human (37) and mouse (33).

Considering that functional and phenotypic alterations of monocytes/macrophages play an important role in sepsis development and due to the potent anti-inflammatory biology of CC16, we hypothesize that an early upregulation of the pro-inflammatory response by local CC16 neutralization will deteriorate the dynamic changes in monocyte and macrophage subsets and early lung damage in a murine trauma model of sepsis after blunt chest trauma.

#### MATERIALS AND METHODS

#### Ethics

The in vitro study was performed in the University Hospital Frankfurt, Goethe-University, Germany, with the institutional ethical committee approval (312/10) in accordance with the Declaration of Helsinki and following STROBE-guidelines (38). In this experimental trial, twenty severely injured trauma patients (TP) with a history of acute blunt or penetrating trauma with an injury severity score (ISS) of ≥ 16 were enrolled, along with and 8 healthy volunteers. All individuals who were <18 or >80 years of age, suffering from a severe burn injury, acute myocardial infarction, cancer or chemotherapy, HIV, infectious hepatitis, acute CMV infection and/or thromboembolic events, or receiving immunosuppressive drug therapy were excluded. The ISS was calculated according to the abbreviated injury scale (39) upon arrival to the emergency department. The signed written informed consent form was obtained from all patients or their legally authorized representatives, as well as from all included healthy volunteers (HV).

Animal experiments were conducted at the Zentrale Forschungseinrichtung of the University Hospital Frankfurt in accordance with the German Federal Law in regard of protection of animals with the approval of the responsible government authority, the Veterinary Department of the Regional Council in Darmstadt, Germany (Regierungspräsidium Darmstadt, Hessen, Germany; AZ: FK 1068). All experiments were performed in accordance with the ARRIVE Guidelines (40).

#### Animals and Experimental Model

Forty male CL57BL/6N mice (25 ± 5 g, 6–8 weeks old) were included (Janvier Labs, France) (41). Before and after experimental procedures, all animals had access to water and food ad libidum. Blunt chest trauma was performed under general mask anesthesia as described before (41). Briefly, the animals were placed in a supine position and a blunt bilateral thoracic trauma (TxT) was induced by a standardized pressure wave provided directly to the chest (41). After 24 h, a median laparotomy with moderate cecal ligation and puncture (CLP) followed as described before (41). Eight animals underwent only TxT. Twenty-four animals underwent the double hit consisting of TxT and CLP. Eight animals in the sham control group underwent anesthesia without performing any surgical procedures. After 6 h, euthanasia was done to facilitate sampling.

#### Group Allocation Based on the Administration of CC16 Neutralizing Antibody

Animals were randomly assigned to different experimental groups for local antibody (Ab) application to the lungs. Administration of Uteroglobin/SCGB1A1 (CC16 Ab, 10µg/mL, LS Biosciences) or IgG Control (IgG) Antibody (10µg/mL, R&D Systems) was performed immediately after the induction of thoracic trauma. For this procedure, mice were placed in a supine position and the tongue was thoroughly kept aside. A buttoned cannula was placed at the beginning of the trachea and 50 µL of the Ab solution were carefully administered. Then, mice were kept in a reverse trendelenburg position for 30 s to ensure the Ab distribution inside the lungs.

#### Sampling and Quantification of Protein Expression Levels in Lungs, Plasma, and BALF

The vena cava was punctured by a heparinized syringe for blood withdrawal at 6 h after CLP. After centrifugation at 1,164 g for 15 min at 4◦C, the plasma was stored at –80◦C for the subsequent measurements of pro-inflammatory mediators. MCP-1 and TNFα were measured in plasma with the CBA Mouse Inflammation Kit (BD Bioscience, San Jose, CA, USA) according to the manufacturer's instructions. Briefly, 50 µL of the Capture Beads were added into polystyrene FACS tubes (BD PharmingenTM) to 50 µL of plasma. To each FACS tube, 50 µL of the Mouse Inflammation PE Detection Reagent were added and incubated at room temperature in the dark for 2 h. Subsequently, samples were washed with 1 mL of Wash Buffer and centrifuged at 200 g for 5 min. Supernatant was discarded and pellet resuspended in 300 µL of Wash buffer. Analysis was performed using a BD FACS Canto 2TM and FCAP ArrayTMSoftware (BD).

After blood withdrawal, the trachea was punctured, intubated and the lungs were flushed with 1.2 mL phosphate buffered saline (PBS) to gain the bronchoalveolar lavage fluid (BALF) for analysis. BALF samples were centrifuged at 1,164 g at 4◦C for 5 min and the supernatant was used for the detection of the receptor for advanced glycation endproducts (RAGE DuoSet <sup>R</sup> ELISA Kit; R&D Systems, Minneapollis, US). Quickly, a microplate (Sarstedt, Nümbrecht, Germany) was coated with 100 µL Capture Antibody overnight at room temperature. Following a washing step, 300 µL of Reagent Diluent was added for 1 h to block the microplate. After another washing step, samples were loaded and incubated for 2 h at room temperature. Subsequently, the microplate was washed again and incubated with 100 µL Detection Antibody for 2 h. Microplate was washed again and incubated with 100 µL Streptavidin-HRP solution for 20 min in the dark at room temperature. Following the last washing step, 100 µL Substrate Solution was added to the wells and incubated in the dark at room temperature until a color reaction occurred. Subsequently, the reaction was stopped by adding 50 µL Stop Solution. The optical density was measured with the Infinite M200 microplate reader (Tecan, Männedorf, Switzerland, 450 nm absorbance, 570 nm reference wavelength; software Magellan).

The cell pellets from BALF were resuspended in 100 µL PBS supplemented with 0.5% bovine serum albumin (FACS buffer), and 40 µL were transferred into polystyrene FACS tubes (BD PharmingenTM) for subsequent cell staining as described below.

Then the animals were perfused with 20 mL PBS via the caudal vena cava, and, subsequently, the lungs were removed. One lung lobe was snap-frozen using liquid nitrogen for later protein isolation, and the other one was used for flow cytometric analyses. For protein isolation, lung tissue was homogenized in protein lysis buffer at 4◦C, followed by centrifugation for 30 min at 4◦C at 20,000 g. Supernatants were stored at –80◦C for later analysis. Protein concentrations of pulmonary RAGE were determined using a mouse RAGE DuoSet <sup>R</sup> ELISA Kit (R&D Systems) as described above.

#### Analysis of Monocyte and Macrophage Subsets by Flow Cytometry

Lung tissue was processed as described in the Minute Single Cell Isolation protocol (Invent Biotechnologies, Minnesota, US). Briefly, 25 mg of fresh lung tissue were placed into a filter cartridge where 100 µL ice-cold Buffer A were subsequently added. The tissue was grinded with a plastic rod for 50–60 times. After adding further 400 µL Buffer A, sample was mixed by inverting the closed filter cartridge and centrifuged at 1,200 g and 4◦C for 5 min. The pellet was resuspended and centrifuged again at 400 g and 4◦C for 5 min. Subsequently, supernatant was discarded and pellet was resuspended in 100 µL FACS buffer. Forty microliter were transferred into each polystyrene FACS tubes (BD PharmingenTM) and stained for flow cytometry analysis as described below.

Thirty microliters of whole blood was transferred into each polystyrene FACS tubes (BD PharmingenTM) and stained for flow cytometry analysis as described below.

The cell pellets from BALF were resuspended in 100 µL of FACS buffer, and 40 µL was transferred into polystyrene FACS tubes (BD PharmingenTM).

Then, the samples were incubated with Pacific Blueconjugated anti-mouse Ly-6G/Ly6C antibody (Ab) (Clone RB6- 8C5; BioLegend, San Diego, California, US), APC/Fire 750 conjugated anti-mouse CD45 Ab (Clone 30-F11; BioLegend), Alexa Fluor 647-conjugated anti-mouse CD11b Ab (Clone M1/70; BioLegend), Brilliant Violett 510-conjugated antimouse F4/80 Ab (Clone BM8; BioLegend), and Phycoerythrin-Cyanine7-conjugated anti-mouse Ly6C Ab (Clone RB6-8C5; BioLegend). Control stainings with the corresponding isotype antibodies were applied for the settings. After 30 min on ice, 5 µL of 7-AAD (BD Biosciences, Franklin Lakes, USA) were added, and samples were incubated for further 15 min. Then, the samples were washed with 2 mL FACS buffer [7 min at room temperature (RT) and 423 g]. Supernatants were removed and cell pellets were homogenized in 1 mL of BD FACS Lysing Solution for an additional 10 min (RT). Then, samples were centrifuged at 400 g for 7 min and washed twice with 2 mL of FACS buffer. After removal of supernatants, cells were diluted in 80 µL FACS buffer and stored on ice until measurement. Each cell population was defined by gating the corresponding forward and side scatter scan as well as the viable cells by applying 7-AAD for gating. From each sample a minimum of 3.0 x 10<sup>4</sup> cells was measured, which were subsequently analyzed. The percentage of Ly6C<sup>+</sup> out of CD11b+Ly6G−CD45<sup>+</sup> and F4/80+CD45<sup>+</sup> viable cells was assessed by flow cytometric analyses using a BD FACS Canto 2TM and FACS DIVATM software (BD). The gating is shown in **Figure 2**.

#### Quantification of Uteroglobin in Sera From Healthy Volunteers and Trauma Patients

Collected sera from healthy volunteers and trauma patients were analyzed using human Uteroglobin Quantikine ELISA Kit (R&D Systems, Minneapolis, US) according to the manufacturer's instructions. Briefly, 100 µL of Assay Diluent were added to each well with subsequent addition of 50 µL of each sample and incubated at room temperature for 2 h. Then, each well was washed with 400 µL Wash Buffer. Subsequently, wells were incubated with 200 µl of Human Uteroglobin Conjugate for 2 h. After the next washing step, 200 µL of Substrate Solution were added into the wells for 30 min. The reaction was stopped by addition of 50 µl of Stop Solution to each well. The optical density was measured with the Infinite M200 microplate reader (Tecan, Männedorf, Switzerland, 450 nm absorbance, 570 nm reference wavelength; software Magellan).

# Isolation of CD14<sup>+</sup> Monocytes

Isolation of peripheral blood mononuclear cells was performed by a density-gradient centrifugation (Bicoll separating solution, Biochrom, Berlin, Germany) according to manufacturer's instructions. Briefly, 25 mL of Bicoll separating solution (density: 1.077 g/mL) was carefully overlaid with an equal volume of heparinized whole blood from HV and centrifuged at 800 g for 30 min. Interphase containing peripheral blood mononuclear cells was transferred to another tube and washed with PBS w/o Ca2<sup>+</sup> and Mg2<sup>+</sup> (Invitrogen, Carlsbad, California, US). The remaining red blood cells were lysed by lysis buffer (0.155 M NH4Cl, 0.01 M KHCO3, 0.1 mM EDTA) and washed with MACS buffer (0.5% BSA, 2 mM EDTA). For CD14 labeling, cell pellet was resuspended in 75 µL of MACS buffer and incubated with 25 µL magnetic CD14 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) for 15 min. After washing, CD14<sup>+</sup> monocytes were isolated by magnetic isolation with LS columns (Miltenyi Biotec) according to the manufacturer's protocol. Cell number and cell viability were determined by Türk's solution exclusion assay (Merck, Darmstadt, Germany). Only cell cultures with a purity of ≥95% were used for further experiments. The cells were cultured in RPMI 1640 medium (Seromed, Berlin, Germany), supplemented with 10% heat-inactivated fetal calf serum (Gibco, Karlsruhe, Germany), 100 IU/mL penicillin (Gibco), 10µg/mL streptomycin (Gibco) and 20 mM HEPES buffer (Sigma) at 37◦C and 5% CO2.

#### Monocyte Treatment

Ex vivo, CD14<sup>+</sup> monocytes isolated from HV were treated with sera from HV and TP that were obtained at the admission to the emergency department. Prior to the experiment, the sera were incubated with or without anti-CC16-antibody (1µg/mL; R&D Systems) for CC16 neutralization or corresponding isotype control antibody (1µg/mL; R&D Systems), respectively, for 1 h at 37◦C and 5% CO2, slightly slewing every 15 min. For monocyte stimulation, cell culture media was supplemented with 20% sera for 2 h at 37◦C and 5% CO<sup>2</sup> and used for further analysis.

#### Migration Assay

Alterations in migratory capacity were determined by CytoSelectTM Cell Migration Assay (3µm pores; Cell Biolabs, San Diego, US). 100,000 cells were plated in the upper chamber and treated as described in the Monocyte treatment section. MCP-1 (10 ng/mL; R&D Systems) was added to the lower chamber. After 3 h at 37◦C and 5% CO2, the upper chamber was removed and cells in the lower chamber were lysed and quantified using CyQuant <sup>R</sup> GR Fluorescent Dye (Cell Biolabs, San Diego, US) according to the manufacturer's instructions. Briefly, the cells containing supernatant from the feeder tray was transferred into a black-walled, clear bottom microplate. CyQuant <sup>R</sup> GR Dye was diluted 1:75 in 4x Lysis Buffer and subsequently added to each well to reach a 1x concentration. Samples were incubated at RT for 20 min. Fluorescence intensity was measured by Twinkle LB 970 Microplate Fluorometer (490 nm excitation/520 nm emission; software MikroWin 2000).

#### Measurement of TGF-β1 Expression in Monocytes by Flow Cytometry

100,000 cells per polystyrene FACS tube (BD PharmingenTM) were treated according to the Monocyte treatment section with slight change. After 1 h of treatment with sera, Brefeldin A (Invitrogen) was added to each tube to 1x concentration and monocytes were incubated for further 2 h at 37◦C and 5% CO2. Subsequently, monocytes were incubated with Phycoerythrinconjugated anti-human CD14 antibody (2 µL; Clone M5E2; BioLegend) and fixable yellow dead cell stain (2 µL; Invitrogen, Carlsbad, California, US). After 30 min at RT, cells were washed with FACS buffer and centrifuged at 400 g for 5 min. Supernatant was removed and monocytes were fixed with Fix and Perm Medium A at room temperature for 15 min. After further washing step, cells were permeabilized with Fix and Perm Medium B (both Invitrogen, Carlsbad, California, US) and incubated with PerCP/Cyanine5.5-conjugated anti-human TGF-β1 antibody (2 µL; Clone TW4-2F8; BioLegend) at room temperature for 30 min. Following the washing step, monocytes were resuspended in 50 µL FACS buffer and analyzed using BD FACS Canto 2TM and FACD DIVATM software (BD). Monocytes were gated by the corresponding forward and side scatter scan and as shown in **Figures 5A,B**. The percentage of TGF-β1 expression of viable CD14<sup>+</sup> monocytes was analyzed.

# Quantification of IL-6 and TNF-α Levels in Monocyte Supernatants

100,000 cells were seeded in flat-bottom 96-well plate (Sarstedt) and treated as described in the Monocyte treatment section. The supernatants were collected to detect the IL-6 (Diaclone, Besançon cedex, France) or TNF-α (R&D Systems, Minneapollis, US) levels using ELISA kits according to the provider's instructions. For a brief IL-6 protocol, see equivalent RAGE measurement protocol in the Sampling and quantification of protein expression levels in lungs, plasma and BALF section. The protocol for TNF-α is equivalent to the Uteroglobin Quantikine ELISA Kit described in the section Quantification of uteroglobin in sera from healthy volunteers and trauma patients.

# Cell Viability Assay

100,000 cells per well were plated in a clear bottom, blackwalled 96-well plate (BD Biosciences) and left to adhere for 30 min at 37◦C and 5% CO2. Subsequently, monocytes were treated as described in the Monocyte treatment section. For cell viability measurement, Calcein AM reagent (1µg/mL; Cayman Chemical, Michigan, US) was added to the cells and incubated at 37◦C and 5% CO<sup>2</sup> for 30 min. Fluorescence intensity was measured by Twinkle LB 970 Microplate Fluorometer (490 nm excitation/520 nm emission; software MikroWin 2000).

# LDH Assay

100,000 cells per well were plated in a flat bottom 96-well plate (Sarstedt) and let to adhere at 37◦C and 5% CO<sup>2</sup> for 30 min. Subsequently, media was replaced with phenol-free RPMI 1640 medium, supplemented with 10% heat-inactivated fetal calf serum, 100 IU/mL penicillin, 10µg/mL streptomycin and 20 mM HEPES buffer and monocytes were treated as described in Monocyte treatment section.

For cell cytotoxicity detection, 100 µL of monocyte supernatant was transferred to a fresh 96-well plate and incubated with LDH reaction mixture (Cytotoxicity Detection Kit, Roche, Mannheim, Germany) according to the manufacturer's instructions in dark at RT for 30 min. Absorbance was measured by Infinite M200 microplate reader (490 nm absorbance, 600 nm reference wavelength; software Magellan).

# Statistical Analysis

GraphPad Prism 6 (GraphPad Software Inc., San Diego, CA) was used to perform the statistical analyses. Normality of all data was analyzed by the D'Agostino-Pearson normality test. Differences between the groups were determined by non-parametric Kruskal-Wallis test which does not assume a normal distribution of the residuals followed by Dunn's post hoc test for the correction of multiple comparisons. A p-value below 0.05 was considered significant. Data are given as box-whisker plot and min to max.

# RESULTS

#### Pro-inflammatory Mediators and Lung Damage Significantly Increase in the Early Phase of Sepsis

TxT in mice increased TNF-α and MCP-1 levels in plasma, whereby the TxT+CLP group showed a significant increase compared to the control group (p < 0.05, data not shown). RAGE protein levels were significantly higher in TxT+CLP compared to control, both in BALF and lungs (p < 0.05, data not shown). Following TxT and TxT+CLP, the total protein content in bronchoalveolar lavage which is associated with the extent of lung damage significantly increased compared to control (p < 0.05, data not shown).

#### The Ratio of Inflammatory and Patrolling Monocytes Increases in Blood, Lungs, and BALF in the Early Phase of Sepsis

For detailed examination of monocyte and macrophage subset distributions, the constituent phenotypes were characterized by their specific surface protein markers. In blood, TxT alone slightly increased levels of activated Ly6C+CD11b+Ly6G−CD45<sup>+</sup> monocytes, whereas TxT+CLP resulted in a significant elevation compared to control (p < 0.05, data not shown). Regarding the subset distribution, inflammatory Ly6ChiCD11b+Ly6G−CD45<sup>+</sup> monocyte subset expanded significantly in TxT+CLP animals (p < 0.05, data not shown), and, in parallel, patrolling Ly6CloCD11b+Ly6G−CD45<sup>+</sup> monocytes showed an equivalent decrease vs. control (p < 0.05, data not shown). Thus, a significant increase in the ratio between inflammatory and patrolling monocytes was observed in early phase of sepsis (p < 0.05, data not shown).

Similarly, activated Ly6C+CD11b+Ly6G−CD45<sup>+</sup> monocytes were markedly more abundant in lungs following TxT+CLP compared to control lungs (p < 0.05, data not shown). Whereas a significant increase of the inflammatory Ly6ChiCD11b+Ly6G−CD45<sup>+</sup> phenotype was observed in TxT+CLP group compared to control group (p < 0.05, data not shown), the number of patrolling Ly6CloCD11b+Ly6G−CD45<sup>+</sup> monocytes was equally reduced (p < 0.05, data not shown). Therefore, comparable to systemic monocytes, the ratio of inflammatory to patrolling monocytes increased markedly in TxT+CLP vs. control (p < 0.05, data not shown).

Furthermore, significantly higher emigration rates of activated Ly6C+CD11b+Ly6G−CD45<sup>+</sup> monocytes to BALF were found in TxT and TxT+CLP compared to control (p < 0.05, data not shown). Regarding the inflammatory phenotypes in BALF, the inflammatory Ly6ChiCD11b+Ly6G−CD45<sup>+</sup> monocyte subset increased in TxT and TxT+CLP compared to control (p < 0.05, data not shown), whereas the patrolling Ly6CloCD11b+Ly6G−CD45<sup>+</sup> monocyte population was reduced (p < 0.05, data not shown). Concurrent with this data, an increased ratio between pro-inflammatory and patrolling monocytes was found in both, TxT and TxT+CLP groups, vs. control group (p < 0.05, data not shown).

### The Number of Pro-inflammatory Macrophages Increases in Lungs and BALF

No significant systemic changes in Ly6C+F4/80+CD45<sup>+</sup> cell counts were found after TxT and TxT+CLP compared to control (data not shown). The pro-inflammatory Ly6ChiF4/80+CD45<sup>+</sup> phenotype slightly increased in TxT with further expansion in TxT+CLP vs. control but both without significance (data not shown). A decline of patrolling Ly6CloF4/80+CD45<sup>+</sup> phenotype was observed in TxT and TxT+CLP vs. control (data not shown).

Whereas, no differences in total macrophage counts were shown in the lungs of TxT and TxT+CLP, a significant increase of pro-inflammatory Ly6ChiF4/80+CD45<sup>+</sup> macrophages compared to control was detected (p < 0.05, data not shown). The cell number of tissue repairing Ly6CloF4/80+CD45<sup>+</sup> macrophages in lungs did not change after TxT and TxT+CLP vs. control (data not shown).

Following TxT, mice displayed a slight increase of Ly6C+F4/80+CD45<sup>+</sup> macrophage counts in BALF, whereas TxT with subsequent CLP did not markedly affect the cell numbers compared to control (data not shown). The inflammatory Ly6ChiF4/80+CD45<sup>+</sup> phenotype expanded significantly in both, TxT and TxT+CLP group in comparison to control group (p < 0.05), while a significant decline of tissue repairing Ly6CloF4/80+CD45<sup>+</sup> macrophages was observed in BALF of TxT and TxT+CLP compared to control (p < 0.05, data not shown).

#### CC16 Neutralization Is Associated With an Increase of Inflammatory Markers and Lung Damage

To investigate the impact of CC16 on inflammatory changes and lung injury, CC16 was neutralized (CC16 Ab) in mice undergoing TxT and subsequent CLP. TxT+CLP induced a significant systemic increase of pro-inflammatory TNF-α and MCP-1 levels compared to control (p < 0.05) with a trend to a further increase in animals that underwent CC16 neutralization (**Figures 1A,B**). Whereas, protein concentrations of RAGE in both lungs (C) and BALF (D) were significantly increased after TxT+CLP vs. control, CC16 neutralization significantly increased RAGE in the lungs after TxT+CLP (p < 0.05, **Figures 1C,D**). With regards to lung tissue damage, total protein content, that was measured in the BALF, and was significantly increased after TxT+CLP vs. control, with a further significant increase in the TxT+CLP group after CC16 neutralization vs. reference TxT+CLP group (p < 0.05, **Figure 1E**).

#### CC16 Modulates Phenotypic Distribution of Monocytes and Macrophages

The effect of CC16 on the subset distribution of monocytes and macrophages was analyzed. The representative gating for the data analyses is shown in **Figure 2**. Total counts of activated Ly6C+CD11b+Ly6G−CD45<sup>+</sup> monocytes in blood increased significantly after TxT+CLP compared to control, while CC16 neutralization did not show any significant impact on this increase compared with the TxT+CLP group (p < 0.05, **Figure 3A**). Similarly, although inflammatory Ly6ChiCD11b+Ly6G−CD45<sup>+</sup> monocytes became significantly more abundant, and the patrolling Ly6CloCD11b+Ly6G−CD45<sup>+</sup> monocytes displayed a significant decline in TxT+CLP animals vs. control, CC16 neutralization did not affect this subset distribution (p < 0.05, **Figures 3B,C**). The ratio of proinflammatory to patrolling monocytes was significantly increased in both TxT+CLP groups vs. control (p < 0.05, **Figure 3D**).

TxT+CLP induced a significantly increased migration of Ly6C+CD11b+Ly6G−CD45<sup>+</sup> monocytes to the lungs (p < 0.05, **Figure 3E**). CC16 neutralization markedly reinforced this effect, and significantly enhanced the presence of

Ly6C+CD11b+Ly6G−CD45<sup>+</sup> monocytes in the lungs after TxT+CLP compared to the TxT+CLP reference group (p < 0.05, **Figure 3E**). However, compared to control, the significant increase of pro-inflammatory Ly6ChiCD11b+Ly6G−CD45<sup>+</sup> and a respective decrease of patrolling Ly6CloCD11b+Ly6G−CD45<sup>+</sup> monocytes were not significantly modulated by CC16 neutralization after TxT+CLP (p < 0.05, **Figures 3F,G**). Thus, the ratio between Ly6ChiCD11b+Ly6G−CD45<sup>+</sup> to Ly6CloCD11b+Ly6G−CD45<sup>+</sup> monocytes was significantly increased in both TxT+CLP and the TxT+CLP group undergoing CC16 neutralization compared to the control (**Figure 3H**).

In BALF, a significant expansion of activated Ly6C+CD11b+Ly6G−CD45<sup>+</sup> monocytes after TxT+CLP was detected (p < 0.05, **Figure 3I**). CC16 neutralization after TxT+CLP did not change this increase compared to control (p < 0.05, **Figure 3I**). The pro-inflammatory Ly6ChiCD11b+Ly6G−CD45<sup>+</sup> phenotype was significantly more abundant in TxT+CLP vs. control (p < 0.05, **Figure 3J**), while CC16 neutralization further enhanced the migration of inflammatory monocytes into the BALF showing a significant increase compared with the TxT+CLP group (p < 0.05, **Figure 3J**). The counts of patrolling Ly6CloCD11b+Ly6G−CD45<sup>+</sup> monocytes significantly declined in TxT+CLP vs. control, while CC16 neutralization did not significantly further impact this monocyte subset decrease after TxT+CLP (p < 0.05, **Figure 3K**). TxT+CLP induced a significant increase on the ratio of pro-inflammatory to patrolling monocytes compared to control (p < 0.05, **Figure 3L**), while a further significant increase after CC16 neutralization vs. TxT+CLP reference group was detected (p < 0.05, **Figure 3L**).

Systemic Ly6C+F4/80+CD45<sup>+</sup> cells were not markedly changed after TxT+CLP or intervention with aCC16 (**Figure 4A**). Neither the pro-inflammatory Ly6ChiF4/80+CD45<sup>+</sup> phenotype (**Figure 4B**) nor the patrolling Ly6CloF4/80+CD45<sup>+</sup> phenotype (**Figure 4C**) were changed.

With regards to local influence of CC16 after TxT+CLP, in lungs, Ly6C+F4/80+CD45<sup>+</sup> macrophage levels remained stable in TxT+CLP compared to control (**Figure 4D**). However, the inflammatory macrophage subset expanded significantly after TxT+CLP vs. control (p < 0.05, **Figure 4E**), whereas the antiinflammatory phenotype decreased significantly after TxT+CLP (p < 0.05, **Figure 4F**). CC16 neutralization did not have a significant impact on macrophage subset redistribution after TxT+CLP in lungs (**Figures 4D–F**).

Total macrophage counts in BALF did not change after TxT+CLP, neither did CC16 neutralization change their levels (**Figure 4G**). However, pro-inflammatory Ly6ChiF4/80+CD45<sup>+</sup> macrophages elevated significantly, while patrolling Ly6CloF4/80+CD45<sup>+</sup> macrophages declined significantly in TxT+CLP or TxT+CLP with CC16 neutralization compared to control (p < 0.05, **Figures 4H,I**). CC16 neutralization significantly increased the percentage of pro-inflammatory macrophages and reduced significantly the percentage of patrolling macrophages after TxT+CLP compared to the TxT+CLP reference group (p < 0.05, **Figures 4H,I**).

#### CC16 Inhibits Migratory Capacity and TGF-β1 Expression in CD14<sup>+</sup> Monocytes ex vivo

To examine the impact of CC16 on monocytes under septic conditions, systemic monocytes were isolated from healthy volunteers and subsequently stimulated with sera from HV or TP (with as well as without septic complications), since the last are known to contain higher levels of CC16 compared to control or trauma patients without complications (42). We have determined CC16 concentrations in samples of trauma patients and healthy volunteers. We found that CC16 was significantly increased in sera obtained from traumatized patients compared to those obtained from healthy volunteers (28.09 ± 4.60 vs. 15.18 ± 1.25 ng/mL, p < 0.05; data not shown). Regarding the migratory rate, stimulation of CD14<sup>+</sup> monocytes with sera from HV did not show any changes compared to control and the neutralization of CC16 or application of isotype antibody in these sera did not affect the migration either (**Figure 5C**). Although the migration of CD14<sup>+</sup> monocytes treated with sera from TP remained unchanged vs. control and HV, CC16 neutralization resulted in significantly higher migration rates toward MCP-1 compared to control, stimulation with sera alone or with IgG (p < 0.05, **Figure 5C**). Intracellular TGF-β1 expression showed no significant changes following treatment with sera from HV vs. control (**Figure 5D**). Stimulation of monocytes with sera from TP significantly declined TGF-β1 expression, and administration of CC16 neutralizing antibodies recovered TGF-β1 level to the baseline (p < 0.05, **Figure 5D**). Furthermore, TNF-α and IL-6 levels in the supernatants obtained from human monocytes that were treated with sera were determined. Stimulation of CD14<sup>+</sup> monocytes with sera obtained from HV and TP did not induce any significant impact on TNF-α nor IL-6 levels (data not shown). TNF-α concentration of the control was 30.06 ± 4.60 pg/mL. Following treatment with sera from HV, TNF-α level was comparable at 32.42 ± 6.90 pg/mL, whereas CC16 neutralization in those sera did not lead to a significant decrease (24.70 ± 5.306 pg/mL). Supernatants from cells that were stimulated with TP sera have shown comparable concentrations of TNF-α to those obtained after incubation with TP sera upon CC16 neutralization (22.01 ± 3.20 vs. 30.18 ± 3.66 pg/mL). Control IL-6 concentration was 46.96 ± 12.31 pg/mL. Treatment with sera from HV did not change the IL-6 level, which was 50.13 ± 19.57 pg/mL and which also stayed stable after CC16 neutralization (40.48 ± 13.14 pg/mL). Comparable results were found in supernatants from TP samples (40.97 ± 6.436 pg/mL) and the corresponding CC16 neutralized sample also (45.84 ± 6.66 pg/mL). Further, we examined the cytotoxic potential of CC16 analyzing the release of LDH. Here, the stimulation

with neither sera from HV nor TP changed the LDH release compared to untreated control (**Figure 5E**). Finally, treatment with both sera and sera with neutralized CC16 or IgG did not show significant changes in the viability of CD14<sup>+</sup> monocytes (**Figure 5F**).

#### DISCUSSION

Since blunt chest injury with ongoing excessive proinflammatory immune response entails high risk for the development of secondary complications with limited therapeutic options, the unveiling of underlying mechanisms is necessary (43, 44). Here, we discuss the dynamic changes in monocyte and macrophage subsets and uncover the potentially protective role of the anti-inflammatory endogenous CC16 in the very early phase of sepsis development following thoracic injury. We confirmed the anti-inflammatory potential of CC16 in the early phase of sepsis-induced ALI following blunt chest injury. Its local neutralization after thoracic trauma increased the immigration of pro-inflammatory cell phenotypes to the lungs, and was accompanied by increased total protein levels in BALF, indicating the loss of epithelial lung integrity, and thus lung damage. Concurrently, systemic elevation of humoral pro-inflammatory mediators was observed. This is in line with our recent study, where early increased lung infiltration with neutrophils and lung injury in this model was shown (45). However, in that study, 24 h post-CLP, the lung injury was ameliorated and the lungs have exhibited no further increase in neutrophilic infiltration after CC16 neutralization (45). Thus, CC16 may first reduce a necessary early pro-inflammatory immune response for tissue repair, and at a later time point, may contribute to the amelioration of the lung injury. Although the mechanism is still not clear, the observed lung injury could be caused by the paralleled enhanced lung infiltration with neutrophils. This assumption is supported by Lerman et al. where neutrophil extravasation and tissue infiltration in murine CLP-induced sepsis were inhibited by blocking or deletion of α3β1 and paralleled by significantly reduced mortality (46).

Following infectious or non-infectious stimuli, alveolar macrophages are, among other cells, the first to be involved in the early immune response, initiating the inflammatory cascade and secretion of pro-inflammatory mediators (20–23). We have shown that thoracic injury followed by CLP increased systemic levels of TNF-α; which is an important indicator of sepsis development (47). Neutralization of endogenous CC16 forced further increase of TNF-α, indicating anti-inflammatory capacity

of CC16 in the present model, and confirming in general its antiinflammatory character. In turn, elevated TNF-α level is also one of the factors inducing MCP-1 expression by a variety of cell types (48). Consistent with literature, we have shown that TNFα increase after thoracic trauma and CLP was paralleled by a systemic MCP-1 increase. CC16 neutralization led similarly to TNF-α to a further systemic increase of MCP-1. Whether this further MPC-1 increase is caused directly by CC16 or indirectly by an increase of TNF-α or other factors still remains to be elucidated in future studies.

Further, we have shown that expression of RAGE significantly increased in lungs and BALF of septic mice, whereas CC16 neutralization led to further significant increase in the lungs. In a clinical study, both CC16 and RAGE were identified as plausible biomarkers for ARDS in patients with severe sepsis (49), but whether RAGE positively or negatively regulates the immune response seems to differ according to the inflammatory mechanism (50, 51). Moreover, we have shown that blunt chest injury itself increased the protein levels in BALF and subsequent CLP did not cause a further increase after 6 h. This does not mean that sepsis did not have an impact on lung epithelial integrity, but lungs are at this time point affected by thorax trauma directly and abdominal-induced sepsis may take longer to affect the lung epithelial integrity than could be seen in the observation period. Upon CC16 neutralization, total protein level in BALF further increased, suggesting a positive impact of CC16 in lung epithelial injury.

Following sepsis-induced ALI after blunt chest trauma, a significant increase of monocyte counts was observed in blood and lungs, as well as in the BALF. The characterization of monocytes has uncovered significantly more pro-inflammatory monocytes compared to a marked decline of the patrolling phenotype. Since MCP-1 is the pivotal regulator of monocyte recruitment to the site of injury (12), the observed elevated MCP-1 levels in septic mice may be the key factor for the excessive infiltration of lungs with pro-inflammatory monocytes, whereby the disrupted lung epithelial integrity may contribute to the higher monocyte content in BALF as well. Upon CC16 neutralization, increased monocyte emigration to the lungs was observed, indicating the anti-migratory potential of CC16. To substantiate our assumption, we isolated monocytes from healthy volunteers and stimulated them with sera obtained from healthy volunteers or from trauma patients, since we have shown in previous studies that increased systemic concentrations of CC16 correlate with the development of secondary respiratory complications following traumatic injury in patients and is lowered in healthy individuals (31, 42, 52). CC16 neutralization in patient's sera before monocyte stimulation led to a significant increase of monocyte chemotaxis toward MCP-1 but the mechanism still remains to be elucidated. However, although CC16 was neutralized, monocytes from TP+aCC16 display elevated migration compared to controls. Serum from trauma patients contains other mediators beside CC16, which may change the migratory behavior of monocytes, thus further mediators such as IL-6 or RANTES that potentially increase the migratory capacity of monocytes are probably concurrently present in the blood from trauma patients (53–56). In a further ex vivo experiment, we have shown unaffected viability of isolated monocytes and their release of lactate dehydrogenase following treatment with sera, suggesting no cytotoxic effects of CC16.

Interestingly, stimulation of isolated human monocytes with sera obtained from trauma patients led to a significant decrease of TGF-β1 expression, whereas neutralization of CC16 recovered the levels to the baseline. Although this was unexpected, it was already reported in rodent models of lung fibrosis that CC16 contributes to diminished TGF-β1 levels, however, the mechanism still remains elusive (37, 57). Here, some studies indicate that CC16 suppresses TGF-β1 expression via MPP-9 inhibition (34, 58).

Regarding macrophage distribution, we observed elevated levels of pro-inflammatory macrophages paralleled by decline of anti-inflammatory phenotype in lungs and BALF. CC16 neutralization reinforced the observed changes, whereas lung infiltration remained unaffected. Since resident alveolar macrophages have been described to have anti-inflammatory properties in steady state and, upon infection or injury, they display a phenotypic shift and gain pro-inflammatory features (18), we hypothesized that CC16 neutralization may contribute to macrophage polarization toward the pro-inflammatory phenotype. Whether CC16 in fact suppresses an exaggerated transition to pro-inflammatory macrophages remains to be further elucidated by future studies.

#### LIMITATIONS

Showing an isolated CLP group with performed interventions would further increase the relevance of our results. However,

monocyte cytotoxicity (E) and viability (F) were evaluated. Data are represented as box-whisker plot and min to max, \*p < 0.05 vs. indicated.

since the scope of the present study was to elaborate the role of CC16 in the underlying double-hit model, this approach was not considered. In humans, it is well-known that secondary stimuli (e.g., surgeries and infections) following chest trauma contribute to the development of secondary pulmonary complications, including ALI and ARDS (44). In mice, it was already shown that pulmonary contusion primes the systemic innate immune response to the LPS challenge, increasing inflammation and worsening lung injury compared to injury or LPS application alone (59). Moreover, recently we have shown that isolated blunt chest trauma in mice was not enough to mimic human conditions since the ongoing pro-inflammatory response decreased to baseline within 24 h, and that combining blunt chest trauma with CLP led to pulmonary changes that were characteristic for ALI (41). However, the mechanism is still unknown, and it remains to be elucidated whether either direct local tissue injury and the subsequent pro-inflammatory response, or the second hit with excessive pro-inflammatory response and remote organ damage contribute more to the ALI development (41, 60). However, literature indicates that the combination of both hits contributes to the increased pro-inflammatory response following doublehit trauma (61–64). Thus, the question of whether CC16 neutralization would affect the isolated CLP animals in the same way as in the TxT+CLP group remains unanswered. Furthermore, showing CC16 levels in all experimental groups at the timing of therapy would support our findings. Yet, due to ethical reasons with regard to animal protection, such analyses were not possible. Following the principle of 3Rs (Replacement, Reduction, Refinement) the number of animals per each group was limited to 8. Following severe thoracic injury, human trauma patients mostly require mechanical ventilation, whereas mice were spontaneously breathing in our experimental settings. Thus, the impact of mechanical ventilation following chest injury could not be considered. Moreover, although mouse models are key tools for studying different pathophysiologies, the immune response between mouse and human differs and the applied treatment cannot be directly translated into human settings. We could not show the impact of CLP on lung integrity, and we consider the short observation period as a further limitation that could lead to negative results. Similarly, monocytes and recruited and resident alveolar macrophages seem to have specific functions in a time-dependent manner. Thus, the right time frame for the examination of monocyte and macrophage function is essential and has to be examined further. In flow cytometric analysis, the chosen markers did not distinguish between recruited and resident alveolar macrophages and this has to be examined in future studies also. Furthermore, although CC16 neutralization increased pro-inflammatory monocyte and macrophages phenotypes, whether CC16 directly contributes to exaggerated transition to pro-inflammatory macrophages still remains to be elucidated. A longer observation period of up to 7 days would bring clarity to the beneficial or negative effects of CC16. Additionally, the distribution of neutrophil and monocyte/macrophage subsets should be evaluated as well. Comparing this with the extent of the lung injury would clarify whether CC16 has either a negative or positive effect on outcomes. It would be interesting to know whether in case of positive effects, CC16 would improve only the lung injury and pro-inflammatory immune response or whether the survival would be improved as well. To confirm the abovediscussed potential results, recombinant CC16 therapy should be applied as well. In in vitro studies, we pooled sera from only ten trauma patients without secondary complications and 10 trauma patients who developed sepsis in a later course, and thus, a larger sample size may clarify the results. Although CC16 is known to have anti-inflammatory properties, we have shown recovered TGF-β1 protein expression levels in CD14<sup>+</sup> monocytes following treatment with a trauma patient's sera. It is possible that TGF-β1 expression is inhibited indirectly by another mechanism and this has to be evaluated in further studies as well.

#### DATA AVAILABILITY STATEMENT

All relevant datasets for this study are contained in the manuscript.

# ETHICS STATEMENT

This studies involving human participants were reviewed and approved by Institutional Ethical Committee of the University Hospital Frankfurt, Goethe-University, Germany. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Veterinary Department of the Regional Council in Darmstadt, Germany (Regierungspräsidium Darmstadt, Hessen.

# AUTHOR CONTRIBUTIONS

BR: conceptualization, supervision, and project administration. AJ, PS, NB, BX, and BR: methodology. AJ: validation, data curation, and writing—original draft preparation. BR and AJ: formal analysis. AJ, PS, NB, and BX: investigation. BR, SW, FH, and IM: resources. AJ, JV, SE, and BR: writing—review and editing. AJ and BR: visualization. BR, FH, and SW: funding acquisition.

# FUNDING

This work was supported by grants from the DFG WU 820/2-1, HI 820/5-1, and RE 3304/8-1.

# ACKNOWLEDGMENTS

We thank Katrin Jurida and Kerstin Kontradowitz for outstanding technical assistance.

# REFERENCES


(CC16) in the development of COPD. Eur Respir J. (2015) 45:1544– 56. doi: 10.1183/09031936.00134214


**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.

The reviewer CE declared a shared affiliation, with no collaboration, with one of the authors, FH, to the handling editor.

Copyright © 2019 Janicova, Becker, Xu, Wutzler, Vollrath, Hildebrand, Ehnert, Marzi, Störmann and Relja. 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.

# Leukocyte-Released Mediators in Response to Both Bacterial and Fungal Infections Trigger IFN Pathways, Independent of IL-1 and TNF-α, in Endothelial Cells

#### Edited by:

*Pietro Ghezzi, Brighton and Sussex Medical School, United Kingdom*

#### Reviewed by:

*Toshiyuki Murai, Osaka University, Japan Selinda Jane Orr, Cardiff University, United Kingdom*

#### \*Correspondence:

*Jill Moser j.moser@umcg.nl Vinod Kumar v.kumar@umcg.nl*

*†Last authorship*

#### Specialty section:

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

Received: *04 August 2019* Accepted: *07 October 2019* Published: *25 October 2019*

#### Citation:

*Le KTT, Chu X, Jaeger M, Plantinga JA, Matzaraki V, Withoff S, Joosten LAB, Netea MG, Wijmenga C, Li Y, Moser J and Kumar V (2019) Leukocyte-Released Mediators in Response to Both Bacterial and Fungal Infections Trigger IFN Pathways, Independent of IL-1 and TNF-*α*, in Endothelial Cells. Front. Immunol. 10:2508. doi: 10.3389/fimmu.2019.02508* Kieu T. T. Le<sup>1</sup> , Xiaojing Chu<sup>1</sup> , Martin Jaeger <sup>2</sup> , Josée A. Plantinga<sup>3</sup> , Vasiliki Matzaraki <sup>1</sup> , Sebo Withoff <sup>1</sup> , Leo A. B. Joosten<sup>2</sup> , Mihai G. Netea<sup>2</sup> , Cisca Wijmenga1,4, Yang Li 1† , Jill Moser 3,5 \* † and Vinod Kumar 1,2 \* †

*<sup>1</sup> Department of Genetics, University Medical Center Groningen, University of Groningen, Groningen, Netherlands, <sup>2</sup> Department of Internal Medicine and Radboud Centre for Infectious Diseases (RCI), Radboud University Medical Center, Nijmegen, Netherlands, <sup>3</sup> Department of Pathology and Medical Biology, Medical Biology Section, University Medical Center Groningen, University of Groningen, Groningen, Netherlands, <sup>4</sup> Department of Immunology, K.G. Jebsen Coeliac Disease Research Centre, University of Oslo, Oslo, Norway, <sup>5</sup> Department of Critical Care, University Medical Center Groningen, University of Groningen, Groningen, Netherlands*

In sepsis, dysregulated immune responses to infections cause damage to the host. Previous studies have attempted to capture pathogen-induced leukocyte responses. However, the impact of mediators released after pathogen-leukocyte interaction on endothelial cells, and how endothelial cell responses vary depending on the pathogen-type is lacking. Here, we comprehensively characterized the transcriptomic responses of human leukocytes and endothelial cells to Gram negative-bacteria, Gram positive-bacteria, and fungi. We showed that whole pathogen lysates induced strong activation of leukocytes but not endothelial cells. Interestingly, the common response of leukocytes to various pathogens converges on endothelial activation. By exposing endothelial cells to leukocyte-released mediators, we observed a strong activation of endothelial cells at both transcription and protein levels. By adding IL-1RA and TNF-α antibody in leukocyte-released mediators before exposing to endothelial cells, we identified specific roles for IL-1 and TNF-α in driving the most, but not all, endothelial activation. We also showed for the first time, activation of interferon response by endothelial cells in response to leukocyte-released mediators, independently from IL-1 and TNF-α pathways. Our study therefore, not only provides pathogen-dependent transcriptional changes in leukocytes and endothelial cells during infections, but also reveals a role for IFN, together with IL1 and TNFα signaling, in mediating leukocyte-endothelial interaction in infections.

Keywords: leukocyte-endothelial interaction, sepsis, bacterial infection, fungal infection, leukocyte transcriptomes, endothelial transcriptomes, interferon pathways

#### INTRODUCTION

Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to infection (1). Despite advances in early recognition, sepsis affects around 30 million people worldwide every year and has a mortality rate of 20–40% (2). Sepsis is known to be a heterogeneous syndrome with various outcomes that depend on pathogenic characteristics as well as on host susceptibility. Despite decades of research, sepsis pathophysiology remains poorly understood. However, it is known that an interaction between innate immune cells and endothelial cells is central for the pathogenesis of sepsis, with recognition of infectious pathogens as a first step toward full activation of inflammation. The activated cells interact with different blood compartments and organ cell types such as platelets, adaptive immune leukocytes, and parenchymal cells (3). Failure in properly regulating these cellular responses and interactions often leads to multiple organ failure, and even mortality in sepsis patients.

Recent studies have employed transcriptomic approaches to characterize the global response of immune cells, particularly peripheral blood mononuclear cells (PBMCs) to various pathogens. After microbial recognition, innate immune cells are prone to inflammatory and stress responses while reducing apoptosis signaling, resulting in the release of cytokines, chemokines, and damage-associated molecular pattern factors into the circulation (4). Gene expression profiles of whole blood from human volunteers challenged with a low dose of endotoxin also showed a reduction in integrin-α and integrinβ chain expression, indicating changes in the host immune system in adherence to and interaction with other cell types (5). Nevertheless, we still know relatively little about the impact of these transcriptomic alterations of immune cells on its interaction with other cell types, specifically on endothelial cells.

Endothelial cells are a monolayer of cells lining the blood vessel that is actively involved in homeostasis as well as interacting with immune cells to regulate inflammation (6). The endothelium is activated not only by inflammatory cytokines such as IL-1β and TNF-α (7), but also by endotoxins such as lipopolysaccharides (LPS) via the Toll like receptor-4 (TLR4) and RIG-I pathways (8, 9). Upon activation, the endothelium secretes cytokines (IL-6 and IL-8) and expresses adhesion molecules (E-selectin, VCAM-1, and ICAM-1) to facilitate leukocyte extravasation. The increased influx of neutrophils and monocytes into the tissue during sepsis could then indirectly lead to tissue damage by secreting exaggerated amounts of inflammatory mediators and reactive molecules (10). In the context of sepsis, the endothelium can be exposed to different types of infectious pathogens such as bacteria (Streptococcus pneumoniae) or fungi (Candida albicans). However, the impact of these pathogens on endothelial activation has not been studied systematically.

Given the ability of endothelial cells to respond to pathogens and interact with immune cells, it is important to characterize endothelial responses upon exposure to the mixture of various cytokines, chemokines and proteins secreted by activated immune cells, as well as to different types of infectious pathogens. Therefore, in this study, we applied a two-step in vitro stimulation model to comprehensively characterize: (1) the transcriptomic responses and inflammatory proteins secreted by PBMCs in response to a variety of stimulating pathogens, including Gramnegative bacteria, Gram-positive bacteria, and fungi; and (2) the transcriptomic responses of endothelial cells exposed to humoral signals from activated PBMCs that were exposed to various pathogens. Through this work, we were able to identify the role of IL-1 and TNF-α in driving most, but not all, endothelial activation. We show that, independent of IL-1 and TNF-α, interferon (IFN) pathways in endothelial cells are strongly induced by humoral signals from activated leukocytes.

Our study provides crucial insights into the role of pathways mediating leukocyte-endothelial interactions, including IL-1, TNF-α, and IFN pathways. Further studies are required to validate the function of IFN pathways in endothelial function and IFN's role in determining sepsis progression.

#### MATERIALS AND METHODS

#### PBMC Isolation

Venous blood samples were collected from healthy volunteers. All donors provided written informed consent. Ethical permission for this study was approved by the Ethical Committee of Radboud University Nijmegen (nr 42561.091.12). Blood was collected in EDTA tubes (BD vacutainer). PBMCs were quickly isolated within 3 h of collection. Blood was diluted with 1 volume of DPBS (Gibco, ThermoFisher Scientific) before adding to Ficoll-Paque (Pharmacia Biotech). Gradient centrifugation was performed for 30 min at 400 g, using no brake. After centrifugation, the layer containing PBMCs was collected using a Pasteur pipette. PBMCs were washed twice with PBS, counted (BioRad cell counter), and adjusted to reach the final concentration of 2 million cells/ml in RPMI 1640 (Gibco, ThermoFisher Scientific), supplemented with 10% heat-inactivated Fetal Cow Serum (Gibco, ThermoFisher Scientific), gentamicin 10 mg/ml, L-glutamine 10 mM, and pyruvate 10 mM. Cells were seeded into wells to settle overnight before stimulation.

#### PBMC Stimulation

To study PBMC transcriptomes upon five types of heat-killed pathogens, PBMCs were stimulated with various pathogens, including heat-killed Streptococcus pneumonia (ATCC 49619, serotype 19F) at 1 million cells/ml, heat-killed C. albicans (ATCC MYA-3573, UC 820) at 1 million cells/ml, heatkilled Aspergillus fumigatus (V05-27) at 1 million cells/ml, Mycobacterium tuberculosis (H37Rv) at 1 million cells/ml, and heat-killed Pseudomonas aeruginosa at 1 million cells/ml (11). Cells were also incubated with RPMI 1640 only as a negative control. RNA was isolated from PBMCs at 4 and 24 h after stimulation.

#### Endothelial Cell Culture and Direct Stimulation

Primary Human Umbilical Vein Endothelial Cells (HUVECs) were used to study the response of endothelial cells upon infection. Pooled donor HUVECs were purchased (Lonza, Breda, the Netherlands) and cultured in EBM-2TM medium (Lonza) supplemented with EGM-2 MV SingleQuot Kit Supplements & Growth Factors (Lonza) at 37◦C, 5% CO<sup>2</sup> and saturating humidity. Passage 3–5, confluent cells were used for all experiments.

For direct stimulation, HUVECs were stimulated with either heat-killed Streptococcus pneumonia (ATCC 49619, serotype 19F) at the concentration of 1 million cells/ml, heat-killed C. albicans (ATCC MYA-3573, UC 820) at 1 million cells/ml, LPS (Escherichia coli serotype O26:B6, Sigma, St. Louis, MO, USA) at 1,000 ng/ml, IL-1β (Biosource Netherlands, Etten-Leur, The Netherlands) at 10 ng/ml, TNF-α (Biosource Netherlands) at 10 ng/ml for 6 or 24 h.

#### Leukocyte-Endothelial Cell Interaction

To study the effect of soluble factors released by activated PBMCs on endothelial cells, PBMCs were diluted to 2 million cells/ml and stimulated with three different types of pathogens at the ratio of 2 cells:1 pathogen heat-killed Streptococcus pneumonia, heatkilled C. albicans and LPS (10 ng/ml). RPMI medium was used as the negative control. Supernatants were collected after 24 h of stimulation, aliquoted, filtered (0.45µm filter) and kept at −20◦C before either exposing to HUVECs or measuring cytokine levels with OLINK and ELISA.

The supernatants from activated PBMCs were thawed overnight at 4◦C and warmed up shortly to 37◦C. Polymyxin B (InvivoGen, Toulouse, France) was added to the supernatants at a final concentration of 100µg/ml to neutralize residual LPS (9). The supernatants were then added to HUVECs. To study the effect of IL-1 and TNF-α secreted by activated PBMCs on endothelial cells, LEAF-purified TNF-α Antibody (BioLegend, San Diego, CA, USA) and/or IL-1RA (Anakinra) were added to the supernatants at the final concentration of 4 and 300 ng/ml, respectively, incubated at 37◦C for 1 h before adding to HUVECs (https://patents.google.com/patent/US7227003). HUVECs were incubated with the supernatants from stimulated PBMCs for 6 and 24 h. At the time of harvesting, conditioned medium, and cells were collected for ELISA, RNA isolation and flow cytometry.

#### RNA Isolation

Cells were harvested and lysed in lysis buffer from the MirVanva MagMax RNA isolation kit Applied Biosystems Nieuwerkerk aan den IJssel, The Netherlands. RNA was isolated according to the manufactures instructions. RNA concentration was measured based on Optical density (OD260) using the Nanodrop machine (NanoDrop Technologies, Rockland, ME, USA). RNA integrity was determined using the Bioanalyzer (Agilent D2000). All samples had RIN score > 9.

#### RNA Sequencing and Pathway Enrichment Analysis

For PBMC sequencing, 1,000 ng of total RNA (RIN score ≥ 9) were submitted for RNA library preparation using the NEXTflex TM Rapid Directional RNA-seq kit, BioScientific. NGS libraries were enriched for polyA tail RNA. Samples were sequenced using the Illumina NextSeq 500 platform, single-end read. Samples were randomly assigned into different flows and sequenced to reach 12–15 million reads per sample. Sequencing reads were then mapped to the human genome using STAR (version 2.3.0) with a reference to Ensembl GRCh37.71. Read counts per gene was quantified by Htseq-count, Python package HTseq (version 0.5.4p3) using the default union-counting mode (The HTSeq package, http://htseq.readthedocs.io/). For endothelial RNA sequencing. 500 ng of total RNA (RIN score ≥ 9) were sent to GenomeScan, Leiden, The Netherlands for analysis. mRNA (polyA) enriched libraries were constructed, and sequenced with the NextSeq 500 platform, single end, 75 bp with 15–20 million reads/ sample. Fragments were aligned using Hisat2. Raw counts were calculated with String Tie. Pathway analysis was performed using gene set enrichment analysis on differential expressed genes using the default setting from ConsensusPathDB-human database (http://cpdb.molgen.mpg.de).

### Flow Cytometry

To determine the protein expression of adhesion molecules on the HUVEC membrane, cells were washed with PBS, detached using trypsin, washed with PBS, and re-suspended in ice-cold FACS buffer (PBS supplemented with 5% FCS). The cells were divided equally into separate FACS tubes. The cells were stained using the following antibodies: PE-conjugated anti-human E-selectin (CD62E) (Biolegend), APC-anti human VCAM-1 (CD106) (Biolegend), FITC- anti human ICAM-1 (CD54) (Biolegend), and IgG isotope controls (IgG isotope controls, Biolegend) for 30 min on ice. The cells were washed once and resuspended in FACS buffer. Samples were analyzed using a MACSQuant Analyzer 10 system (Miltenyi Biotech, San Diego, CA, USA). Multi-color compensation was calibrated using positive control cell population (LPS activated HUVECs). Data were presented as the Geometric Mean of Fluorescence Intensity (MFI).

# ELISA

Cytokine levels secreted by PBMCs and/or HUVECs were determined using the ELISA Duo kits, IL-6 (R&D), IL-1β (R&D), IL-1α (R&D), TNF-α (R&D) and IL-8 (R&D), according to the manufacturer's instructions. Data were presented as pg/ml.

# OLINK

To further quantify the levels of other secreted proteins upon various stimulations, supernatant samples were analyzed by proximity extension assay provided commercially by Proseek Multiplex analysis (Olink Bioscience, Uppsala, Sweden) using their inflammation panel (https://www.olink.com/products/ inflammation). In brief, for each marker, a pair of nucleotide probe-conjugated antibodies was incubated with the sample. Only when binding to target antigen presented in the sample, the pair of probes are in proximity, enabling the probes to anneal and amplify during Realtime PCR. Internal control was used to minimize variation within runs. The output data is an arbitrary unit of normalized log<sup>2</sup> expression scale (NPX- normalized protein expression). The NPX value is different for each protein due to the sensitivity of each of the probes. The range and estimated inversion from NPX value to absolute amount (ng/µl) can be found in Olink website. Data were shown as NPX value.

#### Statistical Methods

For RNAsequencing data, differentially expressed genes were identified by statistical analysis using the DESeq2 package from Bioconductor. A statistically significant threshold (FDR P ≤ 0.05 and fold change ≥ 2) was applied. For pathway analysis, significant threshold (FDR <= 0.05) was used to identify significant pathways. For FACS and ELISA, graphs and statistical tests were performed using GraphPad Prism software v.6 (GraphPad Prism Software Inc., San Diego, CA, USA). Differences were considered significant when p < 0.05. ELISA and FACS data were checked for normality distribution with Omnibus K2 test. One-way ANOVA with Turkey multiple comparison tests were performed to identify significant differences between conditions and control for direct stimulation.

# RESULTS

#### Pathogen-Dependent Early and Late Transcriptional Responses of PBMCs

To identify pathogen-dependent transcriptional responses in leukocytes, we first studied the global transcriptional changes of human PBMCs upon various stimuli. PBMCs were isolated from eight healthy individuals and stimulated by five types of pathogens: Pseudomonas aeruginosa (P. aeruginosa), Streptococcus pneumoniae (S. pneumonia), Mycobacterium tuberculosis (M. tuberculosis), Candida albicans (C. albicans), and Aspergilus Fumigatus (A. fumigatus) for 4 and 24 h. We performed RNA sequencing followed by differential expression analysis to identify differentially expressed (DE) genes between stimulated and un-stimulated samples (RPMI control). This identified 4,189 protein-coding genes that were significantly differentially regulated in response to at least one of the stimulations (Adjusted P ≤ 0.05, FC ≥ 2-fold). Among those DE protein-coding genes, we observed both common genes, which respond to all pathogens, and pathogen-specific genes at 4 h (**Figure 1A**) and 24 h of stimulation (**Figure 1B**). We observed that the Gram-negative bacteria P. aeruginosa altered the expression levels of more than 2,000 genes at 4 h. In contrast, we found fewer genes to be differently regulated in response to C. albicans (666 genes) and Gram-positive bacteria (956 genes) at 4 h (**Supplemental Figure S1A**). At 24 h, we found more genes being differentially regulated by different pathogens. This indicates that P. aeruginosa is one of the stronger inducers of early responses in leukocytes (**Supplemental Figure S1B**). In contrast, C. albicans induced three times more genes, indicating it is a strong immune stimulator at later time points.

Next we performed pathway enrichment analyses on pathogen-specific DE genes (**Supplemental Table S1**). Pathway enrichment analysis of P. aeruginosa-specific genes at 4 h showed significant enrichment of DE genes for several immune pathways, including cytokine responses, IFN signaling, TNF signaling, IL-1 signaling, apoptosis, and inflammasome activation. Interestingly, S. pneumonia-specific DE genes are enriched for the suppression of inflammatory pathways and TCR signaling at both 4 and 24 h. In contrast, C. albicans specific pathways are enriched for antigen presentation and initiating inflammatory responses (**Supplemental Table S1**). Overall, the distinct pathways enriched with DE genes by each pathogen highlight the induction of different inflammatory responses in PBMCs: TNF signaling, IL-1 signaling, and IFN signaling. The transcriptome response, thus reflects complex cytokine responses of PBMCs that are needed to interact with different cell types, depending on the type of infectious pathogens.

#### Common Responses of PBMCs to a Variety of Pathogens Converge on Endothelial Cells

Next we tested whether common genes that are differentially regulated in response to all pathogens are enriched for particular pathways. At 4 h of stimulation, there were 123 genes that responded to all five pathogens (**Figure 1C**). Of note, some of the pathways that were activated at the early time point (4 h) also remained active at 24 h. Among 123 common genes that are either induced or repressed at 4 h, 50% show consistent differences at 24 h. Interestingly, chemokine genes such as CCL2, CCL3, CCL7, CXCL8 (IL-8), and IL-10 are more strongly induced at 24 h, indicating the role of chemokine signaling in communicating with different cell types at later time points (**Supplemental Figure S2**). Pathway enrichment of the 123 common genes showed the enrichment of genes involved in initiating chemokine responses as well as in arranging cellcell interaction, and assembly of cell junctions. Interestingly, expression levels of the cadherin genes CDH5 and CDH6 are reduced, suggesting a repression of adherens junction interactions (**Figure 1D**). At 24 h, we found 236 DE genes shared between all pathogens (**Figure 1E**). The up-regulated genes were enriched for the interaction of immune cells with the extracellular matrix and vascular cell wall, as well as for regulation of trafficking through gap junction (**Figure 1F**). These results show that the common pathways induced in leukocytes in response to different sepsis-causing pathogens are also involved in regulating the interaction of immune cells with the cellular matrix and in interaction with endothelial cells at the vasculature. Endothelial cells are known as a non-classical innate immune cell type that recognize and respond to bacterial lipopeptides via Toll-like receptor signaling. Endothelial responses to infection produce cytokines and chemokines, alter leukocyte migration, facilitate coagulation and, ultimately, contribute to controlling infection (12). Therefore, based on the complexity of cytokine signals released by leukocytes to different types of infection and their convergent effect on endothelial cells, it is crucial to understand the impact of leukocyte humoral signals on endothelial cells and their coordination to fight against infections.

# Minimal Impact of Heat-Killed Pathogens on Endothelial Inflammatory Responses

In sepsis, endothelial barrier disruption is commonly observed in septic shock where organ function fails (13). However, not much is known about whether different types of pathogens can activate vascular endothelial cells directly. We therefore investigated if HUVECs can respond to direct stimulation by LPS, heat-killed S. pneumoniae or heat-killed C. albicans. Transcriptome profiles of HUVECs after 6 h direct stimulation

FIGURE 1 | Core transcriptional responses of PBMC to different stimulations affect EC. The expression levels (log2-fold change) of differentially expressed (DE) genes in PBMC upon stimulation at (A) 4 h and (B) 24 h. Color represents log2-fold change value. Number of shared and specific DE genes at (C) 4 h and (E) 24 h. Pathways enriched for common DE genes at (D) 4 h and top-10 pathways enriched at (F) 24 h. Orange and blue indicate pathways enriched by upregulated and suppressed genes, respectively. Data are represented as (A,B) mean expression levels from PBMC isolated from eight individuals (D,F) –log<sup>10</sup> *q*-value.

showed strong response of HUVECs to LPS in contrast to heat-killed pathogens (**Figures 2A,B**). We found 203 genes that significantly responded to LPS, whereas only 22 and 26 genes were induced by S. pneumoniae and C. albicans, respectively (**Figure 2B**). There were 16 genes in HUVECs that responded to all stimulations (**Figure 2C**), and these were enriched for alterations in semaphorin interaction. RNAseq data also showed significant differences in the expression levels upon stimulation of the inflammatory markers: IL-8 (CXCL8), Eselectin (SELE), and VCAM-1 (VCAM1). However, the small changes in RNA levels of these markers upon S. pneumonie and C.albicans stimulation did not significantly alter protein levels (**Figures 2D,E**). Therefore, direct activation of endothelial cells by pathogens has minimal effect on inducing endothelial cell inflammatory responses. We therefore hypothesized that in response to pathogens, the leukocytes produce mediators that activate endothelial cells much more strongly than direct pathogen stimulation.

#### Leukocyte-Released Mediators Significantly Induce Endothelial Cell Activation

During bloodstream infection, endothelial cells are in contact with both infectious pathogens and immune cells. RNAseq data from activated PBMCs indicated up-regulation of several cytokine pathways upon stimulation with different types of pathogens. We therefore investigated the effect of cytokines from stimulated immune cells on endothelial cells. To mimic the humoral interaction between immune cells and endothelial cells, we first stimulated PBMCs with either LPS, S. pneumoniae or C. albicans for 24 h, then harvested the supernatant, neutralized the LPS-trace by polymyxin B (9), and exposed HUVECs to the supernatant containing cytokine signals from the activated PBMCs. After 6 h of exposure, we measured the expression levels of inflammatory markers on endothelial cells. We observed strong activation of endothelial cells at protein level. Interestingly, whereas direct exposure of HUVECs to heat-killed pathogens did not induce endothelial activation, the exposure to supernatants from S. pneumonia-stimulated PBMCs (Spneu\_Sup) and C. albicans-stimulated PBMCs (C.alb\_Sup) significantly induced the expression levels of adhesion molecules (E-selectin, VCAM-1, and ICAM-1) and cytokines (IL-6 and IL-8) (**Figures 3A–C**). Endothelial cells, together with PBMCs, are the main source of IL-6. We conclude that endothelial cells become activated by heat-killed pathogens via cytokines and other inflammatory mediators released from activated PBMCs.

To further characterize the genes and pathways activated in HUVECs outside the conventional markers, we performed RNAseq to look at the transcriptome of endothelial cells exposed to PBMC supernatants (**Figures 3D,E**). We found 72 genes induced by LPS\_Sup, 180 genes induced by Spneu\_Sup and 222 genes induced by Calb\_Sup. Among these, 65 genes are shared between all supernatants, which is 90% of the LPS\_Sup responding genes, 36% of Spneu\_Sup responding genes, and 29% of Calb\_Sup responding genes (**Figure 3E**). Among the 65 shared genes, 60 are uniquely induced by supernatant and five are also commonly induced by direct stimulation (CXCL8, F3, RND1, SELE, VCAM1). Pathway enrichment for the 60 unique genes shows a strong enrichment for cytokine signaling, IFN signaling, IL-1 signaling and TNF signaling (**Figure 3F** and **Supplemental Figure S3**). Since LPS, hence LPS\_Sup, cannot represent the complexity of PBMC responses to bacteria, we also looked at the genes shared between Spneu\_Sup and Calb\_Sup. Interestingly, here we found 85 commonly responding genes, which is 47% of the S.pneu\_Sup and 38% of the Calb\_Sup responding genes. Pathway enrichment for these genes also indicated strong enrichment for cytokine signaling, particularly for interleukins, IFN, and IL-1 (**Figures 3G–I**). Altogether, this evidence suggests that the activation of endothelial cells by mediators released from PBMCs is mostly shared and independent of the type of infectious pathogens.

#### IL-1 and TNF-α Are Major Mediators, Yet There Are Contributions From Other Cytokines Secreted by PBMCs on EC Activation

Although blocking of IL-1 or TNF-α has resulted in inconsistent results due to study design, recent clinical trials in stratified patients have shown IL-1- or TNF-blocking therapy to be effective in improving sepsis survival (14–16). Since endothelial cells express receptors for IL-1 (17, 18) and TNF-α (19), we were intrigued to investigate the effect of IL-1 and TNF-α present in the PBMC-supernatant on inducing endothelial activation, and if other PBMC secreted-mediators play a role. Before adding the supernatants to HUVECs, we either neutralized TNF-α in the supernatants with a TNF-α blocking antibody or blocked the effect of IL-1α and IL-1β by adding IL-1RA (Anakinra), or both. The blocking dose efficiency was 100% for IL-1 and approximately 75% for TNF-α (**Supplemental Figure S3**). We found that TNF-α secreted by activated PBMCs was the main mediator for endothelial expression of adhesion molecules (Eselectin, VCAM-1, and ICAM-1). However, it was not the sole mediator. IL-1α and/or IL-1β also activated the expression of E-selectin, but not VCAM-1 and ICAM-1 (**Figures 4A–C**). Moreover, neutralization of both TNF-α and IL-1 in the supernatant secreted by activated PBMCs almost completely inhibited endothelial activation, indicating an additive effect of TNF-α and IL-1 on endothelial cell activation (**Figures 4A–C**). On the other hand, TNF-α had no effect on the induction of IL-6 secretion by endothelial cells. This is in contrast to IL-1 blockage, which inhibited endothelial IL-6 secretion. However, the extent to which IL-1 regulates IL-6 expression on endothelial cells depends on other mediators that are co-secreted by PBMCs in response to a specific pathogen. Neutralization of IL-1 in the PBMC supernatant reduced the amount of IL-6 secretion only in the case of C. albicans. Of note, cytokine levels secreted by endothelial cells do not completely return to baseline levels even after blocking TNF-a and IL-1, which suggests that other pathways may be involved in producing endothelial cytokines at a marginal level.

#### Up-Regulation of IFN Pathways in Endothelial Cells by Humoral Mediators From PBMCs Is Independent of IL-1 and TNF-α

As IL-1 and TNF-α are the major mediators that induce a strong response in endothelial cells, we tested the effect of blocking IL-1 and TNF-α on endothelial transcriptional responses (**Figures 4D–F**). Comparison of DE genes in HUVECs exposed to supernatants before and after TNF-α Ab and IL-1RA treatment revealed differential expression of 15, 47, and 81 genes in the context of LPS, S. pneumoniae and C. albicans, respectively (**Figure 4E**). Interestingly, these genes were shared between different supernatants and are enriched for IFN-α/β and IFN-γ pathways (**Figure 4F**). The enrichment was much stronger in the case of C. albicans stimulation suggesting C. albicans is one of the strong stimulators of IFN-inducing mediators. Notably, in the context of S. pneumoniae and C. albicans, we found the differential expression both IFN-induced genes and the upstream genes, including DDX58 (RIG-I), NLRC5, and TLR3. We also found the expression of IFN receptor genes in HUVECs at RNA levels (**Supplemental Figure S4A**). These results suggest that the IFN-α/β and IFN-γ pathways are up-regulated in endothelial cells by mediators released by leukocytes in response to sepsiscausing pathogens, and are independent on IL-1 and TNF-α.

#### Validation of an IL-1- and TNF-α-Independent Effect of IFN-α/β/γ on Endothelial Response

In view of the above data, to test if IL-1- and TNF-α-independent effect on endothelial cells was mainly driven by IFN-α/β/γ, we made use of previously published microarray gene expression data (20). We compared whether the genes induced by direct

\*\*\**p* < 0.001, \*\*\*\**p* < 0.0001.

(A–C) mean (*SD*), representative of three independent experiments (D), mean expression value of 3 biological replications (F), –log<sup>10</sup> of *q*-value. \**p* < 0.05, \*\**p* < 0.01,

activation of HUVEC by IFN-α/β/γ are also induced in our experiment after blocking IL-1 and TNF-α. We found that 62% of genes induced in HUVECs after exposure to Candidainduced supernatant with IL-1 and TNF-α blocking agents are also present in the list of differential expressed genes induced by IFN-α (**Figure 5A**). We also found expression of IFN-β and IFN-g in PBMCs in response to stimulation at RNA levels (**Supplemental Figure S4B**). Altogether, it suggests the presence of IFN-α in the mediators released by leukocytes. We also found high levels of circulatory IFN-γ in the supernatants from Candida-stimulated leukocytes (**Figure 5D**). In addition, we also confirmed the up-regulation chemokines such as CXCL11 and TRAIL which is known to be present at higher levels in the endothelium following treatment with IFN-α (20). Of notice, protein levels of these proteins are not dependent on IL-1 and TNF-α (**Figures 5B–D**).

#### DISCUSSION

Sepsis is not a homogeneous disease, but rather life-threatening organ dysfunction syndrome caused by a dysregulation of host response to infection (1). Given the heterogeneity of sepsis patients, many clinical trials targeting the hyperinflammatory response in sepsis, including corticosteroids, and anti-cytokine therapies (e.g., anti-IL1 and anti TNF-α) have yielded disappointing results (21, 22). Understanding the impact of the complex interactions between causal cell types in a pathogen-specific manner is therefore crucial to delineate the molecular basis of heterogeneity in sepsis outcome. In the present study, we applied an integrative genomics approach to not only characterize the global transcriptional response of leukocytes and endothelial cells to many sepsis-causing pathogens, but to also identify important molecular pathways induced during leukocyte-endothelium cross-talk in regulating overall immune response in sepsis.

First, to identify leukocyte responses to three classes of infectious pathogens (Gram-negative bacteria, Gram-positive bacteria, and fungi) we assessed the transcriptional response of PBMCs to P. aeruginosa, S. pneumoniae, M. tuberculosis, C. albicans, and A. fumigatus. This allowed us to identify several pathways that were induced in a pathogen-specific manner. The S. pneumoniae-specific transcriptome revealed the repression of genes involved in IFN signaling and antigen presentation, while P. aeruginosa-specific genes were enriched for the TNFα and cytokine signaling pathways. These findings suggest that pathogen-specific responses in leukocytes could influence pathophysiology found sepsis patients, partly underlying the heterogeneity observed in sepsis. However, our analysis also identified core pathways that were induced in PBMCs in response to all pathogens as reported before (4). This suggests that the common genes that correspond to core pathways such as antigen presentation, cell-cell signaling, immune regulatory pathways, are absolutely necessary to fight against all types of bacterial or fungal infections. Importantly, we found a strong enrichment of genes involved in leukocyte interaction with endothelial cells and cellular matrix. This finding suggests that the core leukocyte response to pathogens converges on endothelial cells. Interestingly, as suggested by our previous study (23), genes located nearby genetic variants that are associated with sepsis are enriched for immune and endothelial pathways. Therefore, studying leukocyte-endothelium cross-talk is critical in the context of sepsis and potentially can explain sepsis heterogeneity.

Second, by characterizing the global transcriptional response of endothelial cells to sepsis causing pathogens, we emphasize the role of leukocyte-endothelial cross-talk during infections. Although endothelial cells are not considered classical immune cells, HUVECs have been shown to express TLR4 and RIG-I, pattern recognition receptors for LPS-mediating responses to regulate cytokine responses and endothelial activation (8, 9). We indeed observed very strong activation of endothelial transcriptional pathways in response to LPS in comparison to direct stimulation of endothelial cells using other bacterial and fungal pathogens. These findings are in line with the observation of Filler et al. who showed that heat-killed C. albicans were not sufficient to induce strong expression of cytokines and adhesion molecules in endothelial cells in comparison to live and germinating C. albicans (24). Our study mapped the whole transcriptomic response to various pathogens including heatkilled S. pneumoniae and heat-killed C. albicans, which to our understanding, has not been reported before. Moreover, we showed that endothelial responses to Gram-positive bacteria and fungi are strongly affected by inflammatory signals from activated leukocytes. More than 66% of responding genes in HUVECs stimulated by PBMC humoral signals are shared across all stimuli. Although we expected to find activation of endothelial cells in response to circulatory mediators released by leukocytes, it was interesting to find a strong induction of IFN pathways together with IL-1 and TNF-α pathways. Whereas, IL-1 and TNF-α pathways have been investigated as a strategy to improve sepsis outcome, by studying the impact of humoral signals from leukocytes on endothelial cells, we observed a stronger enrichment of IFN-α/β and IFN-γ signaling than IL-1 and TNFα signaling in endothelial cells. This suggests that, together with IL-1 and TNF-α, IFN pathways can result in aberrant responses within endothelial cells. Moreover, by neutralizing IL-1 with Anakinra and TNF-α with TNF-α blocking antibody in leukocyte humoral signals, we confirmed that IL-1 and TNF-α are the major mediators involved in activating endothelial cells. Intriguingly, we also identified different downstream mechanisms regulating endothelial adhesion molecules such as E-selectin, VCAM-1, and ICAM-1. E-selectin is strongly regulated by IL-1 and TNF-α, whereas ICAM-1 and VCAM-1 are driven more by TNF-α. We also found that activated endothelial cells are a major source of IL-6 production, corroborating a previous study (25).

Interestingly, even after blocking IL-1 and TNF-α, we were able to identify strong activation of IFN-α/β and IFN-γ pathways in endothelial cells. Several studies have already identified IFNα/β pathway in leukocytes in response to bacteria (26) and fungi (27). In endothelial cells, although, IFN-α/β has been shown to promote endothelial proliferation in vitro (28), and reduce intracellular NO generation and impair fibrinolysis of HUVECs in vitro (29), the precise impact of IFN-α/β on endothelial function in the context of bloodstream infection or sepsis is not clear. Since NO production from the endothelium maintains blood pressure and blood flow (30) and reduced NO bioactivity is associated with sepsis severity (31), it will be relevant to study the effect of neutralizing IFNs to improve sepsis outcomes. In fact, it is currently being discussed whether IFN-β should be neutralized during the hyper-inflammatory phase in sepsis patients due to its contribution to pro-inflammation and/or whether it needs to be supplemented while patients are in the hypo-inflammatory phase given its ability to restore and reverse immunosuppression (32).

Further studies should investigate which inflammatory mediator(s) from activated leukocytes induced IFN-α and IFN-β signaling in endothelial cells. We observed the mRNA expression of IFN-α/β receptors (IFNAR1, IFNAR2) in endothelial cells across all the stimulatory pathogens we studied, but could not detect IFN-α or IFN-β in the medium of activated leukocytes (data not shown). Nevertheless, we observed a high amount of IFN- γ released by leukocytes in response to C. albicans and also increased RNA expression of IFN- γ receptor (IFNGR1) in endothelial cells. We also observed that C. albicans-stimulated leukocytes secrete the most potent mix of mediators for inducing endothelial activation, which suggests that C. albicans could be a good model to represent the broad impact of leukocyte signals on endothelial cells. One could use it, for example, as a model to study the interaction of leukocytes and endothelial cells with more functional assays to study the biological effect of IFN-activation on endothelial function.

Interestingly, we also observed that TLR3 (TLR-3), DDX58 (RIG-I), and NLRC5, together with IRF-1, -2, -3, and -7 were highly expressed in endothelial cells exposed exclusively to C. albicans- and S. pneumoniae-derived supernatants. RIG-I and IRF-1 are important mediators of endothelial activation in response to LPS and TNF-α as described previously (9, 33, 34). Nevertheless, what regulates the expression of these upstream molecules, either cytosolic DNA or other mediators present in the supernatants, remains elusive.

One of the limitation of our study is the use of heat-killed pathogens, which mounts, to some extent, differences in the host response to pathogens due to the exposure of ligands and the lack of dynamic interactions between the pathogens and the host cells. It has been shown that heat-killed pathogens (such as Bordetella pertussis) induce consistently high RNA expression of inflammatory cytokines (such as TNF-α, MIP-1β, IL1-α, and IL-1β) over time, whereas live bacteria induce a transient increase of those genes that was followed by gene suppression (35). Live bacteria, on the other hand, can induce inflammasome activation, altering the amount of secreted IL-1β in macrophages and dendritic cells whereas heat-killed pathogens cannot (36). In our study, although each heat-killed pathogen induced the responses in PBMCs (at RNA levels), which match with the typically known pathogenic toxins and ligands, to what extent the differences between the heat-killed and live pathogens affect the interaction between leukocytes and endothelial cells remains to be further investigated.

Future studies, if possible, should also investigate the host response to live attenuated pathogens or heat-killed pathogen lysate supplemented with bacterial RNA (36). Secondly, although we characterized the global transcription and protein expression levels during leukocyte and endothelial cells cross-talk, followup studies are needed to understand the consequences of up-regulated IFN signaling in endothelial cells during sepsis. In addition, it will be important to also study the impact of interaction between neutrophils together with leukocytes and endothelial cells in human-relevant model systems. In

#### REFERENCES

1. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA. (2016) 315:801–10. doi: 10.1001/jama.2016.0287

conclusion, our work suggests that activation of IFN pathways in endothelial cells plays an important role in the context of sepsis.

#### DATA AVAILABILITY STATEMENT

The sequencing data generated by this study was deposited to GEO, access number: GSE131590.

#### ETHICS STATEMENT

Ethical permission for this study was approved by the Ethical Committee of Radboud University Nijmegen (nr 42561.091.12). The patients/participants provided written informed consent to participate in this study.

#### AUTHOR CONTRIBUTIONS

YL, VK, and JM shared the conceptualization. KL, JM, and VK designed the study and prepared the manuscript. KL, MJ, and JP performed experiments. XC, VM, and YL analyzed the transcriptome data. LJ, MN, SW, and CW provided reagents, protocols, and facilities to conduct stimulation/blocking experiments. LJ, MN, CW, YL, JM, and VK interpreted results and critically assessed the manuscript.

# FUNDING

This work was supported by the Ph.D. fellowship from the Graduate School of Medical Science, University Medical Center Groningen to KL, a Radboud University Medical Center Hypatia Tenure Track Grant, and a Research Grant (2017) from the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) to VK and a NWO Gravitation Netherlands Organ-on-Chip Initiative (024.003.001) grant to CW. MN was supported by a Spinoza grant of the Netherlands Organization for Scientific Research.

#### ACKNOWLEDGMENTS

We thank all the volunteers for donating PBMCs for this study. We are grateful to K. McIntyre for editing the manuscript and our colleagues within the Genetics department and the EBVDT-SHOCK Research group for fruitful discussion.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02508/full#supplementary-material

2. Fleischmann C, Scherag A, Adhikari NK, Hartog CS, Tsaganos T, Schlattmann P, et al. Assessment of global incidence and mortality of hospital-treated sepsis current estimates and limitations. Am J Respir Crit Care Med. (2016) 193:259–72. doi: 10.1164/rccm.201504- 0781OC


phase III, randomized, double-blind, placebo-controlled, multicenter trial. the interleukin-1 receptor antagonist sepsis investigator group. Crit Care Med. (1997) 25:1115–24. doi: 10.1097/00003246-199707000-00010


**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 Le, Chu, Jaeger, Plantinga, Matzaraki, Withoff, Joosten, Netea, Wijmenga, Li, Moser and Kumar. 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.

# DAMPs and NETs in Sepsis

#### Naomi-Liza Denning1,2,3, Monowar Aziz 1,2 \*, Steven D. Gurien1,3 and Ping Wang1,2,3,4 \*

*<sup>1</sup> Center for Immunology and Inflammation, Feinstein Institutes for Medical Research, Manhasset, NY, United States, <sup>2</sup> Elmezzi Graduate School of Molecular Medicine, Manhasset, NY, United States, <sup>3</sup> Department of Surgery, Donald and Barbara Zucker School of Medicine at Hofstra/Northwell, Manhasset, NY, United States, <sup>4</sup> Department of Molecular Medicine, Donald and Barbara Zucker School of Medicine at Hofstra/Northwell, Manhasset, NY, United States*

Sepsis is a deadly inflammatory syndrome caused by an exaggerated immune response to infection. Much has been focused on host response to pathogens mediated through the interaction of pathogen-associated molecular patterns (PAMPs) and pattern recognition receptors (PRRs). PRRs are also activated by host nuclear, mitochondrial, and cytosolic proteins, known as damage-associated molecular patterns (DAMPs) that are released from cells during sepsis. Some well described members of the DAMP family are extracellular cold-inducible RNA-binding protein (eCIRP), high mobility group box 1 (HMGB1), histones, and adenosine triphosphate (ATP). DAMPs are released from the cell through inflammasome activation or passively following cell death. Similarly, neutrophil extracellular traps (NETs) are released from neutrophils during inflammation. NETs are webs of extracellular DNA decorated with histones, myeloperoxidase, and elastase. Although NETs contribute to pathogen clearance, excessive NET formation promotes inflammation and tissue damage in sepsis. Here, we review DAMPs and NETs and their crosstalk in sepsis with respect to their sources, activation, release, and function. A clear grasp of DAMPs, NETs and their interaction is crucial for the understanding of the pathophysiology of sepsis and for the development of novel sepsis therapeutics.

#### Edited by:

*Timothy Robert Billiar, University of Pittsburgh, United States*

#### Reviewed by:

*Markus Bosmann, Boston University, United States Michael Thomas Lotze, University of Pittsburgh Cancer Institute, United States*

#### \*Correspondence:

*Monowar Aziz maziz1@northwell.edu Ping Wang pwang@northwell.edu*

#### Specialty section:

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

Received: *14 May 2019* Accepted: *11 October 2019* Published: *30 October 2019*

#### Citation:

*Denning N-L, Aziz M, Gurien SD and Wang P (2019) DAMPs and NETs in Sepsis. Front. Immunol. 10:2536. doi: 10.3389/fimmu.2019.02536* Keywords: DAMPs (damage-associated molecular patterns), NETs (neutrophil extracellular traps), sepsis, HMGB1 (high-mobility group box 1), CIRP, cold-inducible RNA-binding protein, histone, neutrophils

#### INTRODUCTION

Sepsis is common and deadly; 30–50% of patients suffering an in-hospital mortality have sepsis. In the United States, sepsis affects 1.7 million adults annually resulting in more than 250,000 deaths (1, 2). It is estimated that, worldwide, sepsis impacts 30 million people per year and leads to 6 million deaths (3). Until recently, sepsis was defined as the systemic inflammatory response syndrome (SIRS)—hypo or hyperthermia (>38◦C or <36◦C), increased heart rate and respiratory rate and increased or decreased white blood cell count- in the presence of an infection. Sepsis with organ dysfunction was severe sepsis and fluid-refractory hypotension was septic shock (2). New guidelines, called Sepsis-3, established new definitions of sepsis, defining sepsis as "life threatening organ dysfunction caused by dysregulated host response to infection" (2). Organ dysfunction, as recommended by Sepsis-3, is defined clinically as changes of 2 points or more on the Sequential [Sepsis-related] Organ Failure Assessment (SOFA). The most severe subset of sepsis—septic shock- is defined as "sepsis in which underlying circulatory and cellular metabolism abnormalities are profound enough to substantially increase mortality" (2).

Sepsis arises from the body's exaggerated immune response to infection (4). Based on the "germ theory" of disease (5), it was initially thought that the inflammation, organ injury, and death that follows an infection were solely due to the body's response to microbial products,

**160**

such as pathogen-associated molecular patterns (PAMPs) (6). PAMPs are recognized by pattern recognizing receptors (PRRs) expressed on immune-reactive cells (7). Numerous studies have been published to demonstrate the role of PAMPs and PRRs in activating the immune system in sepsis (4, 6). During the last several decades, subsequent studies have identified damage-associated molecular patterns (DAMPs). DAMPs are host nuclear or cytoplasmic non-microbial molecules which, when released from the cell following tissue injury, serve as potent activators of the immune system initiating and perpetuating a non-infectious inflammatory response to cause systemic inflammation, organ injury, and death (8–10). Like PAMPs, DAMPs are also recognized by PRRs and utilize the same signal transduction machinery to activate the immune system (6, 11). Clinically, sepsis severity has been shown to correlate with DAMPs; studies have shown that increased serum levels of DAMPs including high mobility group box 1 (HMGB1), extracellular cold-inducible RNA-binding protein (eCIRP), and H3 correspond with increased with disease severity (12–14). This review describes several well-known DAMPs, details the mechanisms of their release and actions, and describes therapeutic strategies that target DAMPs in sepsis.

Neutrophils are the most abundant leukocytes in the body and serve as the first line of defense against infection (15). The effector function of neutrophils is mediated through phagocytosis, reactive oxygen species (ROS), and protease dependent killing of ingested pathogens. In addition, activated neutrophils release neutrophil extracellular traps (NETs)—webs of DNA and antimicrobial proteins designed to kill pathogens (16, 17). The discovery of NETs provided new insights into neutrophil effector function. However, numerous studies have also revealed the detrimental role of NETs in sepsis (18). Homeostasis in regards to NETs requires the interplay between their beneficial bactericidal properties and the hyperstimulation of immune cells by the DNA and proteins contained within NETs that results in inflammation and tissue injury in sepsis.

A number of review articles have been published demonstrating the individual role of DAMPs or NETs in sepsis (6, 19, 20). In sepsis, DAMP mediated signaling fuels pro-inflammatory cytokine and chemokine production by macrophages and other immune cells. This, in turn, leads to excessive neutrophil infiltration into the tissue. Activated neutrophils produce reactive oxygen species (ROS), inducible nitric oxide synthase (iNOS), and NETs which contain noxious molecules, leading to tissue inflammation and injury in sepsis. In this review, we focus on DAMPs, NETs, and explore their interplay during sepsis (**Figure 1**). We also discuss some of the therapeutic interventions targeting both DAMPs and NETs in experimental sepsis (**Table 1**).

#### DAMPs

DAMPs were first proposed as part of the "Danger Theory" by Polly Matzinger in the mid 1990's as an initial explanation for the robust inflammatory response elicited in response to sterile inflammation, which could not be explained solely by the self vs. non-self-hypothesis of the time (8). Intracellularly, DAMPs are hidden from view of the innate immune system. After tissue injury, caused by either sterile or infectious insults, they are released extracellularly to activate the immune system and resultant pro-inflammatory cascades (34). As discussed above, DAMPs are thus defined as endogenous molecules that can initiate and potentiate a non-infectious inflammatory response (8). In addition to their role in sepsis, as is discussed in the rest of this article, the release of DAMPs is critical to the development of sterile inflammation including inflammation that occurs secondary to organ ischemia and reperfusion injuries (35–37), non-infectious inflammatory liver diseases such as nonalcoholic fatty liver disease (38), or the sterile inflammation associated with aging (39).

Allowing the evolution of the Danger Theory from an abstract concept to a concrete entity, probably the first DAMP identified was HMGB1 (40, 41). Other DAMPs include histones, ATP, uric acid, DNA, mitochondrial DNA, and IL-33 (42). Recently, eCIRP has been identified as a newly discovered DAMP (43, 44). Although numerous endogenous molecules have been identified as inflammation-causing DAMPs, here we briefly review a selective group of DAMPs which have been strongly implicated in sepsis.

#### HMGB1

HMGB1 is a highly conserved protein expressed in all mammalian cells (21). HMGB1 as a DAMP causing sterile inflammation was discovered in 1999 (41). HMGB1 can be released actively via cytoplasmic vesicles or passively from necrotic cells. Active release is mediated by several pathways; JAK/STAT-1 mediated acetylation is responsible for the initial HMGB1 translocation from the nucleus to the cytoplasm, while extracellular release is partially mediated by double-stranded RNA-activated protein kinase R (PKR)/inflammasome-mediated pyroptosis (45). While passive release after necrotic cell death is rapid, active HMGB1 release is much slower. HMGB1 levels reach a plateau approximately 16–32 h after the onset of endotoxemia (46). HMGB1 related signaling is modulated by the redox state of its three cysteines (numbers 23, 45, and 106) (47, 48). Once released into the extracellular space, HMGB1 activates innate immune cells to propagate pro-inflammatory signaling cascades (49). HMGB1 induces recruitment of neutrophils to the site of tissue injury (50). HMGB1 binds to other PAMPs, including DNA (51), LPS (52), and lipoteichoic acid (53), potentiating their inflammatory responses. HMGB1 has been shown to bind to numerous cell surface receptors, including but not limited to receptor for advanced glycation end products (RAGE), TLR2, TLR4, TLR9, and triggering receptor expressed in myeloid cells 1 (TREM-1) (49, 54). After binding to these receptors, it activates macrophages and endothelial cells, stimulating the production of proinflammatory chemokines, cytokines, and endothelial adhesion molecules (49). HMGB1 is elevated in patients with sepsis (12, 55), and dozens of studies have demonstrated that targeting HMBG1 improves outcomes in sepsis (24, 25, 56, 57).

#### eCIRP

Extracellular CIRP is an 172-amino acid RNA chaperone protein (26, 58–60) that was previously identified as a DAMP in 2013 (43). It is a cold shock protein, originally recognized as a protein that suppresses mitosis and promotes cell differentiation in the setting of hypothermia (61). It is upregulated by hypothermia, hypoxia, and oxidative stress, such as UV irradiation. In addition to passive release during necrotic cell death, in times of cellular stress (like the aforementioned hypothermia, hypoxia, or oxidative stress), CIRP can translocate from the nucleus to cytoplasmic stress granules; from these, it is released to the extracellular space (62). After eCIRP binding to its receptor, the TLR4-myeloid differentiation factor 2 (MD2) receptor complex (43), activation proceeds through the TLR4/MyD88/NF-κB pathway (63) to stimulate the release of pro-inflammatory cytokines TNF-α and HMGB1 from macrophages (43). Furthermore, during sepsis, hemorrhage or ischemia-reperfusion (I/R) injury, CIRP is released extracellularly and leads to organ injury (36, 43). Elevated plasma levels of eCIRP have been independently correlated with a poor prognosis in patients with sepsis (13).

eCIRP as a DAMP has been demonstrated in several cell types including macrophages, lymphocytes, and neutrophils in the context of cellular activation, cytokine and chemokine production and neutrophil extracellular trap (NET) formation (44). eCIRP has also been shown to stimulate the Nlrp3 inflammasome, cause endoplasmic reticulum (ER) stress, and induce pyroptosis in lung endothelial cells (EC) (64, 65). eCIRP is associated with acute lung injury (ALI). Healthy mice injected with recombinant murine (rm) CIRP develop ALI via macrophage, neutrophil, and EC activation, and cytokine production in the lungs (65). Beneficial outcomes have been seen in CIRP−/<sup>−</sup> mice or CIRP inhibition in murine models of renal, intestinal, and hepatic I/R injury (36, 66, 67). CIRP−/<sup>−</sup> mice are protected from sepsis and ALI (64, 65). In an animal models of adult or neonatal sepsis, treatment with a polyclonal anti-CIRP antibody or a CIRP-derived inhibitory peptide prolonged survival and attenuated organ injury (43, 68, 69).

#### HISTONES

Histones are highly basic proteins that are located mainly in the nucleus. In humans, histone H2A, H2B, H3, and H4 form a complex with DNA, called a nucleosome. The nucleosome regulates gene transcription and facilitates efficient higher-order chromatin compaction (22). However, histones



*<sup>a</sup>Rodent, <sup>b</sup>Human, <sup>c</sup>Non-human primates.*

play proinflammatory functions upon their release from the nucleus into the extracellular environment (23). Histone release from cells can occur passively after cellular necrosis or as part of an active process such via NETosis (70). In 2009, Xu et al. demonstrated that histones were cytotoxic when added to cultured endothelial cells (71). In vivo, intravenous injection of histones in mice was lethal, whilst anti-histone antibodies were found to reduce mortality in murine models of LPS endotoxemia, TNF-α, or cecal ligation, and puncture experimental models of murine sepsis (71). Xu subsequently demonstrated that the injection of sublethal doses of histones resulted in high levels of the cytokines TNF-α, IL-6, and IL-10, a phenomenon which did not occur when TLR4−/<sup>−</sup> mice were used. Conversely, TLR2−/<sup>−</sup> mice maintained their hyperinflammatory profiles after histone injection (72). However, using specific TLR-transfected HEK cells, histones signaling was transduced via both TLR4 and TLR2 (72). Histones have also been shown to bind to TLRs in cardiomyocytes where they alter levels of regulatory proteins and potentiate sepsis-induced cardiomyopathy (27). The impact of histones has also been investigated in human sepsis. Ex-vivo administration of serum from septic patients directly induced cardiomyocyte death; this effect was abolished by anti-histone antibody (73). Histone levels in septic patients are significantly increased and, like in murine models, appear to cause cellular injury in a TLR4 dependent method (14).

#### CELL FREE DNA

In the extracellular space, deoxyribonucleic acid (DNA) can serve as a DAMP. Apoptosis, necroptosis, NETosis, and pyroptosis can all contribute to the release of nuclear contents into the extracellular space (74). Cell free DNA in plasma is elevated in patients with severe sepsis or septic shock when compared to patients without these diagnoses (28), and increased levels of cell free DNA in the plasma of septic patients has been linked to increased mortality during sepsis (75).

Viral, bacterial, and even host cell free DNA can all function as a DAMP and initiate pro-inflammatory cascades (74, 76). Additionally, mitochondrial DNA (mtDNA) has been proven to be a DAMP; it is released into the circulation during trauma or sepsis (77, 78). mtDNA has been shown to cause TNF-α secretion by mouse splenocytes and IL-1β release from bone marrow-derived macrophages (79). In addition to promoting the release of proinflammatory cytokines, DNA has been shown to prolong the lifespan of neutrophils. Neutrophils stimulated with either purified bacterial or mitochondrial DNA demonstrated increased viability compared to controls (78). Excessive neutrophil accumulation in tissues has been associated with poor outcomes in sepsis (80).

Viral, bacteria, host cell free DNA, and mtDNA can all act via the TLR9 receptor (74), which is located intracellularly in endosomes (81). It is important to recognize the spatial relationship of DNA that acts as an immunomodulatory molecule and the TLR9 receptor. TLR9's intracellular location requires that nuclear DNA molecules that are released into the extracellular space by NETosis, apoptosis and other forms of cell death need to be translocated intracellularly in recipient cells in order to activate the TLR9 receptor (74). Besides TLR9, intracellular DNA can trigger other alarmin sensors such as cyclic guanosine monophosphate-adenosine monophosphate synthase (cGAS), absent in melanoma 2 (AIM2), interferon-inducible protein 16 (IFI16), and stimulator of interferon genes (STING), all of which lead to the initiation of immune responses (74).

#### ATP

ATP is a nucleotide that, in times of homeostasis, is generated mainly within mitochondria during the tricarboxylic acid cycle and from the respiratory chain. ATP is also produced in the cytoplasm during glycolysis (82). ATP is released actively from dying cells during apoptosis, and passively during necroptosis and cellular necrosis (38, 83). Although some extracellular ATP is beneficial, as it functions as a chemoattractant recruiting phagocytic cells to the site of tissue damage, extracellular ATP is also detrimental, binding to ionotropic P2X receptors (P2XR) (84). P2XR channel opening results in increases in intracellular calcium, which activates the p38 MAPK pathway, activating the inflammasome with the associated caspase-1 activation and release of pro-inflammatory cytokines IL-1β and IL-18 (84–86). Elevated ATP levels in the plasma of septic patients interfere with neutrophil function and signaling, resulting in an excessive and uncoordinated neutrophil activation (87). Excessive extracellular ATP has also been associated with T cell suppression in sepsis (88). Reduction in the extracellular levels of ATP has proven to be an effective method of attenuating sepsis severity in some murine models of sepsis. Removal of extracellular ATP to decrease activation of the P2X7 receptor by CD39 has been shown to attenuate sepsis-induced liver injury (89). Treatment with apyrase, an ATP hydrolase that removed extracellular ATP, protected mice against a lethal LPS challenge and resulted in a reduction of serum cytokines (90).

#### MOLECULES THAT MAY OR MAY NOT BE DAMPs

Several endogenous molecules located intracellularly or on the cell surface are released into the circulation and serve as diagnostic and prognostic markers in various inflammatory diseases (4, 29). These molecules include components of the extracellular matrix (ECM) like collagen, fibrinogen, and laminin and shredded cell surface receptors, such as soluble ST-2(30), a member of the interleukin 1 receptor family, sMD2(91), sTREM-1(92), microRNAs (93), exosomes (94), and vesicles (95). However, it is not clear whether these and similar molecules should be classified as DAMPs (**Figure 2**). DAMPs are frequently released from cells following necrosis, pyroptosis or apoptosis, however the ECM, shredded receptors, exosomes, micro-vesicles are released into the extracellular environment without cell lysis. Conversely, mtDNA and cell-free DNA are classified as DAMPs and are released in both suicidal and vital NETosis, meaning a molecule can be classified as a DAMP without cell lysis first occurring. Many DAMPs undergo structural modification (96, 97) e.g., oxidation, reduction, acetylation, phosphorylation, or cleavage after release into the circulation. Conversely, it is not known whether the shredded receptors or exosomal molecules undergo post release modification in the extracellular milieu. Extracellularly, DAMPs play largely pro-inflammatory roles, while the secreted proteins, cleaved receptors, exosomes and vesicles are not always pro-inflammatory and are not necessarily responsible for excessive inflammation (98). Cell surface proteins that are shed have diverse functions and include chemokines, cytokines, adhesion molecules, growth factors, and their receptors (99).

The shedding process of these proteins regulates the density of cell surface receptors, the release of factors that serve as agonists, and the release of soluble receptors that can function as antagonists (100). Cleaved receptors such as sTREM-1 acts as a decoy receptor, sequestering TREM-1-ligands and dampening TREM-1 activation (101, 102). Soluble ST-2 serves as an antagonist for IL-33 to control excessive innate immune response (103). Exosomes, macrovesicles, and microparticles are enriched in pro- and anti-inflammatory molecules, therefore they may play dual roles in sepsis. LPS-challenged macrophages have been shown to release histone-coated microvesicles to cause inflammation (104). Exosomes released from alveolar macrophages during hemorrhagic shock have been shown to promote necroptosis (105). By contrast, exosomes filled with anti-inflammatory molecule milk fat globule-EGF-factor-8 (MFG-E8) were shown to be beneficial in reducing markers of inflammation in sepsis and improving survival (106). Cleaved receptors or exosomes often directly serve as chemoattractants (107), but the ability of DAMPs to directly serve as a chemokine are not as well studied.

Excess production and release of ECMs may cause tissue fibrosis, abnormal cell proliferation, migration and inflammation (108). Receptor protein cleavage occurs due to the actions of matrix metalloproteinases (MMPs), disintegrins, and metalloproteinases (ADAMs) which are upregulated during inflammation (109). The exosomes and microvesicles are released from the cells through pore formation in the plasma membrane by caspase-mediated GSDMD or by a budding out process (110). The release of excess amount of exosomes and microvesicles are correlated with an increased release of DAMPs, allowing the possibility that exosomes and microvesicles may be a mechanism of DAMP release in sepsis (111). Exosomes and microvesicles may also serve as a means to maintain cell to cell communication; they have the ability to enter into adjacent cells and modulate function. Extracellular microRNAs levels are increased in various inflammatory conditions and may serve as diagnostic markers (112). Studies have shown that extracellular microRNA plays a pro-inflammatory role following its re-entry into macrophages and activation of the endosomal TLR7 receptor to produce TNF-α and IL-6 (113).

More studies on these molecules will help elucidate their pathophysiological role in sepsis and other inflammatory conditions. This information will aid in clarification of these molecules as DAMPs or non-DAMPs.

#### NETs

Neutrophils are phagocytic cells; they predominantly defend against pathogens either by engulfing the offending cell and destroying it via oxidant- or protease-dependent mechanisms

or by the secretion of anti-microbial peptides (114). This classical understanding of neutrophil function was found to be incomplete after the discovery of a third effector function of neutrophils in 2004, the release of NETs (17). NETs are web-like chromatin based structures that are released into the extracellular environment to aid in pathogen clearance, but they have also been implicated in excessive inflammation with resultant tissue damage, potentiation of autoimmunity, and promotion of vascular thrombosis (16). NETosis is a form of cellular death in which neutrophils decondense their nuclear chromatin and DNA into the cytoplasm. Chromatin and DNA mix with granule-derived antimicrobial peptides and are extruded into the extracellular space (115). NETosis can be induced in many ways; one of the most well-described is phorbol myristate acetate (PMA), a protein kinase C (PKC) activator (116).

NETs contain proteins from azurophilic granules e.g., neutrophil elastase (NE), myeloperoxidase (MPO) and cathepsin G; proteins from secondary and tertiary granules e.g., lactoferrin, and gelatinase; and nuclear proteins e.g., histones H1, H2A, H2B, H3, and H4 (117). Detection of NETs has proved challenging due their fragile structure, timing of NET formation and turnover, and ubiquitous presence of DNase I. Several tools to assay NETosis have been reported: these include microscopy (118), flow cytometry (119, 120), ImageStream <sup>R</sup> (121), and ELISA (122). The ability to detect NETs precisely is paramount to studying the disease pathophysiology associated with NETosis.

# MECHANISM OF NET FORMATION

The first reported descriptions of NETs demonstrated that neutrophils stimulated with PMA, IL-8 or LPS released NETs (17). Subsequent studies have revealed a wide range of stimuli including bacteria, virus, fungi, yeast, parasites, and concanavalin A are capable of inducing NET formation (20). In addition, NETs are upregulated in various cancers, including pancreatic cancer, through receptor for advanced glycation end products (RAGE) dependent and neutrophil autophagy mediated pathways (123). The induction of NETosis by various DAMPs will be discussed in the later part of this article.

Two forms of NETosis have been described: suicidal NETosis, in which NET formation only occurs via neutrophil cell death and was described above, and vital NETosis where NETs are released without cell death (124). In suicidal NETosis, NADPH-dependent ROS production is a prerequisite. This leads to increased calcium influx and peptidyl arginine deaminase 4 (PAD4) activation, leading to chromatin decondensation. Elastase and MPO are also transported from the granules to the nucleus to cleave linker histone H1 and modify the core histones. MPO also intensifies chromatin decondensation, through the synthesis of hypochlorous acid. Finally, chromatin is released outside the cell through membrane pores and cellular lysis through the activation of a pore forming protein GSDMD (125, 126).

Denning et al. DAMPs, NETs, and Sepsis

First described in 2012, vital NETosis results in the release of NETs without a loss in the integrity of the nuclear or plasma membrane (127). As such, neutrophils are able to survive the process and are still capable of normal neutrophil functions including phagocytosis. Unlike suicidal NETosis, vital NETosis does not require generation of ROS or activation of the Raf/MERK/ERK pathway (126). In contrast to the several hour time frame required for stimulation of suicidal NETosis, vital NETosis occurs quickly, usually within 5 to 60 min after neutrophil are stimulated (128). In vital NETosis, after neutrophil stimulation, typically via TLR or complement receptor for C3 protein ligand binding, the nuclear membrane morphology changes to allow vesicle budding. These vesicles, containing nuclear DNA, move through the cytoplasm to coalesce with the plasma membrane and are released extracellularly (118, 124, 126).

Besides the aforementioned types of NETosis, in 2009 it was reported that neutrophils are able to undergo vital NETosis using mitochondrial DNA (129). GM-CSF primed neutrophils, when activated via TLR-4 or complement factor 5a receptor stimulation, generated NETs containing solely mitochondrial DNA. NETosis facilitated release of mitochondrial DNA seems to be ROS-mediated (129). In vivo, NETs containing mitochondrial DNA have been found in the serum of individuals after trauma (130) and associated orthopedic surgery (131).

Several other mechanisms of NET formation have been reported. Carestia et al. demonstrated that activated platelets are able to amplify the amount of NETs released from neutrophils (132). This process seemed to depend on interaction between glycoprotein Ib (CD42) in platelets with β2 integrin (CD18) in neutrophils, as well as the release of von Willebrand Factor. Platelet triggered NETosis did not rely on NADPH oxidase or ROS generation, but was reduced when inhibitors of ERK, PI3K, or Src kinases were used (132). NET formation has been shown to depend on the activation of cell-cycle proteins CDK4/6; Cdk6−/<sup>−</sup> neutrophils and mice showed impaired NET formation to several stimuli including PMA and C. albicans (133). The lipoxin pathway has been shown to reduce lung inflammation and acute lung injury after both infectious and sterile inflammation (134). Lefrancais et al. demonstrated that this pathway, through Fpr2 receptor signaling, is a potent modulator of NET formation. After intratracheal injection of methicillin-resistant Staphylococcus aureus (MRSA), Fpr2−/<sup>−</sup> mice produced excessive NETs compared to wild type mice (135). Additional studies are needed focusing on the pathways behind these types of NET formation to determine the type of NETosissuicidal or vital.

#### PHENOTYPIC AND FUNCTION DIVERSITIES OF NEUTROPHILS AND NET FORMATION

Neutrophils exhibit phenotypic and functional heterogeneity (136). Neutrophil heterogeneity has tremendous impact on NET formation. Neutrophils from diabetic patients are more likely to undergo NETosis than neutrophils from euglycemic patients (31). Neutrophils from pediatric patients with systemic lupus erythematosus also undergo increased NETosis as compared to their healthy counterparts (137). ICAM-1 (CD54) is mainly expressed on the endothelial cell surface (138). Following simulation of neutrophils with PAMPs or DAMPs, ICAM-1 expression in the neutrophils is dramatically increased (139– 141). The ICAM-1<sup>+</sup> neutrophils produce higher levels of NETs, probably because of increased ROS (140). However, the involvement of ICAM-1 or its ligand Mac-1 in the increased levels of NETs in these cells has not been elucidated. The relationship seems to be circular, with NETs inducing ICAM-1 in neutrophils and ICAM-1<sup>+</sup> neutrophils producing increased quantities of NETs (142). ICAM-1<sup>+</sup> neutrophils are found in increased concentrations in blood and lungs of humans and mice under inflammatory conditions (143–146).

It is still not clearly known which type of neutrophilscirculating or tissue resident-produce increased levels of NETs. Using density gradient centrifugation, circulating neutrophils can be separated into two layers- high density neutrophils (HDN) and low-density neutrophils (LDN) which co-localize with peripheral blood mononuclear cells (147). LDN are a heterogeneous population containing both immature and mature neutrophils and their functions differ depending on the inflammatory stimulus (148, 149). Interestingly, it has been demonstrated that LDNs have an increased proinflammatory profile as compared to other neutrophils with increased secretion of proinflammatory cytokines (150, 151) and an increased capacity to generate NETs (149, 152, 153).

Since the ROS pathway is essential for suicidal NETosis (125), it is logical that neutrophils that produce increased levels of ROS may produce excessive NETs. Although evidence is conflicting (154), Zhang et al. found that aged neutrophils (CXCR4+) produced both increased levels of ROS and increased amounts of NETs (155). It is also evident that human neutrophils are more prone to produce NETs compared to murine neutrophils (156, 157), indicating the role of specific surface markers in NETs production between these species. Overall, neutrophil heterogeneity may play a pivotal role in NET formation.

#### INCREASED NET FORMATION IN SEPSIS

NETs are vital to pathogen clearance, but simultaneously NETs induce collateral damage to host tissues in sepsis (16). In 2007, Clark et al. described an interaction between platelets and neutrophils in sepsis, resulting in NET formation and enhanced bacterial trapping in blood vessels (158). Activation of TLR4 receptors on platelets lead to the binding of the platelets to neutrophils in the blood. These neutrophils were then activated and produced NETs. These results were recapitulated using the plasma from severely septic patients (158).

Sepsis often results in acute lung injury (ALI) (159). Lefrancais et al. demonstrated abundant NET formation in both murine models of severe bacterial pneumonia and ALI (135). Furthermore, when comparing NET levels in samples from critically ill human subjects they found higher levels of NETs in subjects with infectious etiology of acute respiratory distress syndrome (ARDS) as opposed to patients with cardiacinduced respiratory dysfunction. In addition, among patients with microbiologically confirmed pneumonia, plasma NET levels were higher in patients with ARDS than in patients without. Finally, there was a correlation between the severity of ARDS, mortality, and the serum level of NETs (135).

In a clinical study, the levels of neutrophil-derived circulating free DNA (cf-DNA/NETs) have been shown to directly correlate with multiple organ dysfunction score, sepsis-related organ failure assessment, leukocyte counts, and MPO levels (160). A 2018 study of 55 critically ill patients demonstrated rapid and sustained increases in the circulating levels of MPO-DNA complex in the serum, indicating NET formation in the early stages of sepsis. In this study, MPO-DNA complex levels were also correlated with the severity of organ dysfunction and 28-day mortality rates (161).

In opposition to these findings, impaired NET formation in neonates has been associated with relative immunodeficiency of human newborns (162). Czaikoski et al. found increased bacterial burden in the blood and decreased survival in a murine model of CLP in mice treated with DNase to prevent NET formation, however these effects were ameliorated by treatment with DNase plus antibiotics (163). Given that there are both hyper and hypodynamic phases of sepsis, the levels of NETosis at various stages in sepsis may impact the outcomes. This idea is supported by work done by Mai et al. (164). They found that when given early after induction of sepsis by CLP, DNase increased pro-inflammatory cytokines and worsened renal and pulmonary damage. However, when given at a later timepoint after CLP, DNase administration reduced IL-6 levels, increased levels of anti-inflammatory IL-10, and reduced organ damage and bacterial dissemination. It also increased survival after CLP (164).

Several studies have demonstrated that severe sepsis alters the neutrophil phenotype and hinders NETosis ex vivo (165, 166). However, it is not clear from these studies whether in vivo NET formation is impaired during sepsis. Further investigation will need to be done in this area.

#### DETRIMENTAL EFFECTS OF NETs IN SEPSIS

During sepsis, neutrophil-endothelial interaction is increased to promote neutrophil infiltration into tissues (167). Neutrophilendothelial cell (EC) interaction leads to increased NET formation; this increased NET formation is partially dependent on IL-8 released from activated EC (168). Prolonged co-culture of neutrophils with EC resulted in EC damage; this damage is attributed to NETs as co-incubation with either NAPDH oxidase inhibitors or DNase ameliorated this damage (168).

Recent studies demonstrated the crucial role of NETs in the pathogenesis of disseminated intravascular coagulation and intravascular thrombosis, both of which increase morbidity and mortality in sepsis (169–173). McDonald et al. found profound platelet aggregation, thrombin activation, and fibrin clot formation within NETs, implicating the NET–platelet– thrombin axis in the promotion of intravascular coagulation in sepsis. Inhibition of NETs during sepsis by DNase infusion reduced intravascular coagulation, improved microvascular perfusion, and reduced organ damage (172).

NETs have been detected in bronchoalveolar lavage samples from septic humans or canines with ARDS, indicating that, even after transmigration, neutrophils are capable of undergoing NETosis (174, 175). A recent study utilizing samples from different models of ALI in mice and from patients with ALI revealed increased levels of NETs and histones H3 and H4 in the bronchoalveolar lavage fluids (BALF) (176). Administration of the extracellular histones contained in NETs resulted in damage to alveolar epithelial cells and increased severity of ALI (176).

In addition to the damage inflicted by the DNA released during NETosis, enzymes released during NETosis also have a detrimental effect on the surrounding tissues. Neutrophil elastase, a key component of chromatin degranulation, has been show to increase permeability of alveolar epithelial cells by altering the actin cytoskeleton (177) and its inhibition has been demonstrated to be beneficial in animal models of inflammation and associated ALI (178, 179). Serine proteases released during NETosis have been shown to degrade surfactants which are vital in the clearance of inflammatory cells and residual inflammation after ALI (18). These findings clearly demonstrate that excessive NETs play detrimental role in sepsis.

#### THERAPEUTIC STRATEGIES TARGETING NETs IN SEPSIS

Therapeutic strategies aimed at NETs primarily target the DNA component- DNase is the most frequent treatment modality. DNase treatment reduced NETs, improving lung injury and survival in a murine model of pneumonia (135). Cl-Amidine, a PAD4 inhibitor, had no effect on the level of neutrophil-DNA complexes or the degree of lung inflammation in a murine pneumonia model (135) but Biron et al. found that Cl-Amidine prevented H3 citrullination, NET formation, and improved survival in a murine model of CLP-induced polymicrobial sepsis (180). Similarly, PAD4−/<sup>−</sup> mice demonstrated decreased NETs and lung injury in the pneumonia model (135). However, these benefits were offset by an increased bacterial load and increased systemic inflammation. Therefore, Lefrancais et al. developed a mouse with a partial PAD4 deficiency (PAD4+/−) which demonstrated an improved survival curve (135). These findings support the notion that a there is a thin line for the amount of NETosis required to both prevent lung injury and maintain microbial control.

Chloroquine has also been effective as an early upstream inhibitor of NETs, decreasing NETosis and the associated hypercoagubility and improving survival in murine models of pancreatic adenocarcinoma (181) and acute pancreatitis (32). Activated protein C (APC) is a multifunctional protease with anti-inflammatory, anticoagulant, and cytoprotective properties (182). A recent study demonstrated that APC binds human leukocytes and prevents activated platelet supernatant or PMA from inducing NETosis. Additionally, they found that pretreatment of neutrophils with APC prior to induction of NETosis inhibited platelet adhesion to NETs (182). It should be noted however, that activated protein C has failed to have any impact on survival in large scale human clinical trials of patients with severe sepsis (183, 184). Li et al. demonstrated that antibodies neutralizing serum citrullinated Histone 3 could improve survival after a murine CLP model (185). These studies demonstrate that abrogating excessive NET formation can lead to beneficial outcomes in sepsis.

The early inhibitors of NETs such as chloroquine, PAD4 inhibitors, and APC are specifically targeted for controlling NET formation. By contrast, late inhibitors of NETs, such as DNase and anti-histone antibodies, can target extracellular DNA or histones regardless of their source. These molecules are also considered as DAMPs and can be released by a number of immune cells, in addition to their release from neutrophils. Therefore, the molecules/drugs that specifically control intracellular NET formation could be used as a more specific therapeutic regimen against NETs.

#### CROSSTALK BETWEEN DAMPs AND NETs IN SEPSIS AND INFLAMMATION

Although the extracellular release of DAMPs and NET formation are both a byproduct of sepsis, there is increasing evidence of linkage between the two. The major components of NETs, i.e., DNA, histones, and granule proteins- are recognized as DAMPs that can trigger inflammation, inducing cell death and organ failure. Extracellular histones are elevated in patients with coagulopathy and multiple organ failure (186) and are believed to be a major mediator of death in sepsis (71). Cell free DNA has been shown to be cytotoxic and results in coagulopathy and disseminated intravascular coagulation (DIC) (33) Additionally, inhibition of NETosis via PAD4 deficiency or inhibition results in a reduction in the release of DNA and improves outcomes in sepsis (187–189).

Concomitantly, various DAMPs have been shown to induce NETosis. Tadie et al. demonstrated that HMGB-1 is able to induce NETosis via TLR4 signaling (190). Incubation of neutrophils with HMGB-1 resulted in increased extracellular DNA, histone 3, and histone 3 citrullination. Exposure of neutrophils isolated from wild type and RAGE KO mice to HMGB1 resulted in significant NET formation, whereas neutrophils from TLR4 KO mice demonstrated a diminished ability to form NETs. Finally, HMGB1 acted synergistically with LPS, as neutrophils from the bronchoalveolar lavage (BAL) of mice exposed to both LPS and HMGB1 displayed greater ability to produced NETs compared to neutrophils isolated from the BALs of mice that received LPS alone. This increase was hindered by a neutralizing antibody to HMGB1 (190).

eCIRP has also been shown to activate NETosis through a TLR4/NF-κβ dependent mechanism (140). Mice subjected to polymicrobial sepsis via cecal ligation and puncture demonstrated increased levels of ICAM-1<sup>+</sup> neutrophils in both the blood and the lungs. In contrast, mice genetically deficient in CIRP displayed diminished levels of ICAM-1<sup>+</sup> neutrophils. In vitro, treatment of neutrophils with recombinant murine CIRP (rmCIRP) increased levels of ICAM-1<sup>+</sup> neutrophils, and this increase was inhibited by both a neutralizing antibody to TLR4 or an NF-κβ inhibitor. ICAM-1<sup>+</sup> neutrophils displayed increased levels of NETosis (140).

Unlike eCIRP and HMGB1, mitochondrial DNA (mtDNA) seems to generate NETosis through a TLR9 dependent pathway. mtDNA induced NADPH oxidase-independent NET formation in polymorphonuclear neutrophils of healthy volunteers. NETosis was completely inhibited by treatment with a TLR9 inhibitor (130). Liu et al. further identified that mtDNA also activates NETosis via the STING pathway (191). Neutrophils treated with mtDNA demonstrated increased NETosis in a manner which displayed significant increases of AKT and ERK1/2 phosphorylation and increased expression of Rac2 and PAD4. They further confirmed that both TLR9 and STING pathways are important in mtDA-induced NETosis via examination of the lungs of mice intravenously injected with mtDNA (191). Lungs displayed decreased NET formation in TLR9 KO and STING KO mice compared to wild type mice. Additionally, in vitro stimulation of BMDN from TLR9−/<sup>−</sup> and STING−/<sup>−</sup> mice displayed decreased percentages of NETs after treatment with mtDNA as compared to WT mice (191). Further confirming that mtDNA-induced NETosis proceeds through the Raf/MEK/ERK and p38 MAPK pathways, TLR9−/<sup>−</sup> and STING−/<sup>−</sup> neutrophils exhibited decreased phosphorylation of ERK 1/2 and p38 MAPK, as well as decreased levels of PAD4 and Rac2 after stimulation with mtDNA than WT neutrophils did. Inhibitors of these downstream mediators resulted in decreased mtDNA-induced NET formation in WT neutrophils (191).

Oxidized low-density lipoproteins (oxLDL) are upregulated in sepsis and intestinal inflammation (192) and have been recognized as a DAMP (193). In vitro treatment of PMNs with oxLDL resulted in increased NET formation in a dose dependent manner. oxLDL stimulation of NETosis seems to depend on TLR2 and 6; blocking of neutrophils with a TLR4 antibody had no effect on NET formation, while blocking with anti-TLR2 or TLR6 antibodies modestly reduced NETosis. However, the combination of anti-TRL2 and anti-TLR6 antibody treatment of PMNs prior to oxLDL stimulation resulted in a significant reduction in the formation of NETs (194). Additionally, confirming the role of the PKC pathway in oxLDL-induced NETOsis, inhibition of PKC or IRAK was able to reduce NET formation in normal neutrophils. Inhibition of downstream mediators in the pathway, ERK1/2 and p38 MAPK, also reduced oxLDL-induced NET formation (194).

# FUTURE DIRECTIONS AND CONCLUSIONS

In this review article, we discussed DAMPs and NETs in sepsis, with a focus on their interaction and therapeutic strategies for amelioration of sepsis-associated morbidity and mortality. Future studies on the interaction between the two entities would add value to the study of innate immunology and could be expanded to other inflammatory conditions in addition to sepsis. Moreover, future emphasize should also be focused on pinpointing the relationship between PAMPs and NETs and developing new therapeutic tools to target their interplay. DAMPs are released by several cell types, while NETs are specific to neutrophils. Recently, extracellular traps (ETosis) has been described in macrophages (195) and eosinophils (196). Future studies on DAMPs and ETosis would be interesting. Immune cells in sepsis are very plastic with several phenotypic polarizations—more investigation is needed into the role of immune cell plasticity on DAMP release. Similarly, future studies on how DAMPs skew immune cell polarization and the subsequent impact on sepsis would be revealing. In conclusion, we have provided a literature review of the role of DAMPs,

#### REFERENCES


NETs, and their interaction in sepsis to increase and update our understanding in this area of research.

#### AUTHOR CONTRIBUTIONS

N-LD and MA did literature review and wrote the manuscript. SG helped in writing the extracellular DNA section and reviewing the manuscript. N-LD prepared the table and MA prepared the figures. PW reviewed, edited the manuscript, and conceived the original idea of the project.

# FUNDING

This study was supported by the National Institutes of Health (NIH) grant R35GM118337 (PW) and R01GM129633 (MA).

contribution to cellular injury and inflammation. Crit Care. (2014) 18:543. doi: 10.1186/s13054-014-0543-8


impairs their mitochondrial function via P2Y11 receptors. J Biol Chem. (2019) 294:6283–93. doi: 10.1074/jbc.RA118.007188


the lung irrespective of inflammatory disease. Clin Exp Immunol. (2009) 155:559–66. doi: 10.1111/j.1365-2249.2008.03791.x


mortality. Mediators Inflamm. (2018) 2018:4065362. doi: 10.1155/2018/40 65362


**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 Denning, Aziz, Gurien and Wang. 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.

# Orthopedic Surgery Triggers Attention Deficits in a Delirium-Like Mouse Model

Ravikanth Velagapudi 1†, Saraswathi Subramaniyan1†, Chao Xiong1†‡, Fiona Porkka<sup>2</sup> , Ramona M. Rodriguiz <sup>2</sup> , William C. Wetsel 2,3 and Niccolò Terrando<sup>1</sup> \*

<sup>1</sup> Department of Anesthesiology, Center for Translational Pain Medicine, Duke University Medical Center, Durham, NC, United States, <sup>2</sup> Department of Psychiatry and Behavioral Sciences, Mouse Behavioral and Neuroendocrine Analysis Core Facility, Duke University Medical Center, Durham, NC, United States, <sup>3</sup> Departments of Neurobiology and Cell Biology, Duke University Medical Center, Durham, NC, United States

#### Edited by:

Massimo Gadina, National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS), United States

#### Reviewed by:

Paul S. Garcia, Columbia University, United States Pietro Ghezzi, Brighton and Sussex Medical School, United Kingdom

> \*Correspondence: Niccolò Terrando niccolo.terrando@duke.edu

†These authors share first authorship

#### ‡Present address:

Chao Xiong, State Key Laboratory of Cardiovascular Disease, Department of Anesthesiology, National Center for Cardiovascular Disease, Fuwai Hospital, Chinese Academy of Medical Sciences, Peking Union Medical College, Beijing, China

#### Specialty section:

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

Received: 20 March 2019 Accepted: 30 October 2019 Published: 19 November 2019

#### Citation:

Velagapudi R, Subramaniyan S, Xiong C, Porkka F, Rodriguiz RM, Wetsel WC and Terrando N (2019) Orthopedic Surgery Triggers Attention Deficits in a Delirium-Like Mouse Model. Front. Immunol. 10:2675. doi: 10.3389/fimmu.2019.02675 Postoperative delirium is a frequent and debilitating complication, especially amongst high risk procedures such as orthopedic surgery, and its pathogenesis remains unclear. Inattention is often reported in the clinical diagnosis of delirium, however limited attempts have been made to study this cognitive domain in preclinical models. Here we implemented the 5-choice serial reaction time task (5-CSRTT) to evaluate attention in a clinically relevant mouse model following orthopedic surgery. The 5-CSRTT showed a time-dependent impairment in the number of responses made by the mice acutely after orthopedic surgery, with maximum impairment at 24 h and returning to pre-surgical performance by day 5. Similarly, the latency to the response was also delayed during this time period but returned to pre-surgical levels within several days. While correct responses decreased following surgery, the accuracy of the response (e.g., selection of the correct nose-poke) remained relatively unchanged. In a separate cohort we evaluated neuroinflammation and blood-brain barrier (BBB) dysfunction using clarified brain tissue with light-sheet microscopy. CLARITY revealed significant changes in microglial morphology and impaired astrocytic-tight junction interactions using high-resolution 3D reconstructions of the neurovascular unit. Deposition of IgG, fibrinogen, and autophagy markers (TFEB and LAMP1) were also altered in the hippocampus 24 h after surgery. Together, these results provide translational evidence for the role of peripheral surgery contributing to delirium-like behavior and disrupted neuroimmunity in adult mice.

Keywords: attention, delirium, blood-brain barrier, microglia, neuroinflammation, surgery

#### INTRODUCTION

Postoperative delirium is a common complication characterized by acute cognitive impairments featuring disorganized thinking, fluctuating levels of consciousness, altered arousal levels, and inattention (1). A diagnosis of delirium consistently correlates with a poor prognosis (i.e., a 5-fold increased risk for mortality at 6 months), persistent functional decline, increased nursing staff time per patient, increased hospital stays, and higher rates of nursing home placement (2, 3). This condition contributes to escalating healthcare costs previously estimated >\$150 billion each year (4). Delirium is frequently seen in medical, intensive, emergency, nursing, and palliative care settings amongst any age group; however, it is the most common surgical complication in older adults (5).

Orthopedic surgery is a common intervention across every age group that can frequently contribute to cognitive impairments, emotional disturbances, and pain in a significant proportion of patients (6). As many as 50% of patients suffer from postoperative delirium after orthopedic surgery (7) and delirium is a wellappreciated predictor for mortality and adverse outcomes in patients with underlying dementia (8, 9). We have developed a mouse model of tibial fracture and repair to study the impact of orthopedic surgery on the central nervous system (CNS) and to interrogate the pathogenesis of perioperative neurocognitive disorders (10). We and others have described some of the behavioral changes affecting the hippocampus and other brain areas after surgery, although no formal evaluation of attention has been conducted to translate preclinical findings to human delirium assessment. The 5-choice serial reaction time task (5-CSRTT) is a well-established attention paradigm used to assess various aspects of attentional control in rodents, including selective, divided, and sustained attention or vigilance (11). The 5-CSRTT has been commonly used to elucidate attention impairments in several neuropsychiatric conditions including attention deficit/hyperactivity disorder (ADHD) and schizophrenia (12).

Neuroinflammation has emerged as an active driver in the pathogenesis of multiple neurological conditions, including neuropsychiatric, neurodegenerative, and perioperative disorders (13, 14). Following surgical procedures such as orthopedic surgery, pro-inflammatory cytokines have been detected in the cerebrospinal fluid of patients developing postoperative delirium (15, 16). The exact roles of these cellular processes in postoperative delirium remains elusive and their active involvement after anesthesia and surgery is generating significant interest. Recent advances in tissue clarification techniques have enabled complex 3D features in intact specimens to be observed, especially in brain (17, 18). CLARITY provides a unique approach to preserve cellular integrity through the formation of a tissue-hydrogel mesh that enables deeper optical imaging. This technology can elucidate complex 3D structures and cellular architectures of relevance to both tissue physiology and pathology, including the diagnosis of clinical conditions (19–21).

In the present study, we hypothesized that orthopedic surgery may acutely impair attention processes as assessed with the 5- CSRTT, and this delirium-like behavior may correlate with CNS inflammation assessed by microglial morphology, glial fibrillary acidic protein (GFAP) expression, and blood-brain barrier (BBB) permeability. We also implemented a CLARITY protocol combined with light sheet microscopy to enable visualization and 3D rendering of pathological features associated with postoperative complications focusing on neuroinflammation and endothelial dysfunction.

#### MATERIALS AND METHODS

#### Animals

A total of 61 inbred C57BL6 (male, 12 weeks old) were purchased from The Jackson Laboratory (Bar Harbor, ME) and housed in the Mouse Behavioral and Neuroendocrine Analysis Core Facility. Animals were matched for weight and housed 3–5 per under controlled temperature and humidity with a 14:10 h light:dark cycle (lights on at 0600 h). Mice were fed standard rodent chow (Prolab RMH3500, Autoclavable; LabDiet, St. Louis, MO) and had access ad libitum access to food and water. Thirty animals were used for histological evaluation. Open field motor activity was assessed in 16 mice and was performed 24 h following fracture surgery (8 sham controls and 8 fracture animals). The remaining 15 mice were assigned to the 5-CSRTT study. One week before beginning the 5-CSRTT testing, animals were handled daily and food-restricted to achieve and be maintained at 85–90% normal body weight. Food restriction continued through the duration of the study. Once 5-CSRTT begun, mice were always fed 2–3 h following completion of training or testing. All experiments were conducted under an approved protocol from the Institutional Animal Care and Use Committee at Duke University Medical Center and under the guidelines described in the National Science Foundation "Guide for the Care and Use of Laboratory Animals" (2011). Duke University is an AAALAC certified institution.

#### Surgery

Tibial facture was performed as described (10) with minor modifications. Mice were anesthetized with isoflurane using the SomnoSuite apparatus, a low-flow digital anesthesia system (Kent Scientific Corporation, Torrington, CT). Heart rate, blood oxygen saturation, and core body temperature (36.5 ± 0.6◦C) were monitored throughout the procedure using pulse oximetry and a homeothermic pad system (Kent Scientific Corporation). Muscles were disassociated following an incision on the left hindpaw. A 0.38-mm stainless steel pin was inserted into the tibia intramedullary canal, followed by osteotomy, and the incision was sutured with 6-0 Prolene.

# Open Field Assessment of Activity Following Tibial Fracture Surgery

Motor activity was assessed in a 5-min open field test within 24 h of surgery for animals subjected to tibia fracture with sham controls matched for age and weight. Testing began when mice were singly placed into a 60 × 40 × 24 cm polyfoam box indirectly illuminated at 375 lux with a single camera suspended overhead interfaced to a computer running Ethovision 11.5 software (Noldus Information Technology Inc., Leesburg, VA). Activity of the mice was assessed with 3-point tracking and the total distance moved (cm) was scored for each animal.

# 5-Choice Serial Reaction Time Task (5-CSRTT)

Testing occurred in 5-CSRTT chambers (24 × 18.5 cm; MedAssociates, St. Albans, VT). Each operant chamber had five 1.24 cm<sup>2</sup> nose-poke apertures, illuminated with LED lights, with infrared diodes to register nose pokes into the aperture. A single food magazine with an LED light and infrared diodes was positioned on the opposite wall. The food magazine dispersed single food rewards (20 mg chocolate Rodent Purified Diet pellets; BioServ, Flemington, NJ). After surgery, perforated plastic mesh canvas (Darice Inc., Strongsville, OH) with 0.01 cm<sup>2</sup>

openings were placed on bottoms of each test chamber to ensure animal mobility. Mice were trained and tested in the same chamber during 1,000 and 1,500 h. Animals were trained with a modified procedure (11, 22). For the first week of training (phase 1), all nose poke apertures were illuminated. If mouse poked into any hole, the food magazine was illuminated and a food reward dispensed. The mice had 5 s to retrieve and consume the food reward. After a 5 s inter-trial interval (ITI) the next trial began with the illumination of the nose poke apertures. This procedure was repeated until 20 rewards were earned or a maximum of 1 h had elapsed. Once mice earned 20 rewards and showed consistent performance for 3 days, the next phase was initiated. In phase 2, a single nose poke aperture was lit for a 5 s stimulus duration (SD). Animals could respond during the 5 s illuminated period or during the 5 s limited hold (LH) period which followed if the mouse didn't make an immediate selection while the aperture was lit. If the mouse selected the correct aperture during the SD or LH periods, the food magazine was illuminated and a food reward dispensed. Mice were given 5 s to retrieve the pellet. If a mouse failed to respond during the SD or LH periods or poked into an incorrect hole, the trial was immediately terminated and a 5 s time-out (TO) was initiated before proceeding to the next ITI. Mice were trained daily with the 5 s SD and 5 s LH until the animal achieved correct responses on at least 80 percent of the trials (i.e., 32 trials of 40 total trials/day) over 3-consecutive days. Of the 15 mice which began the study, 14 were capable of being trained to nose poke for food reward in phase 1 of training. Upon meeting criterion, the SD was decreased by 0.5 s each day. If the animal was successful in achieving 80% correct responses at that presentation time, the SD was reduced the following day by 0.5 s until the mouse reached a 2 s SD. In all cases, the LH was set to 5 s. When the mice exhibited at least 80% correct responses out of 40 trials for 3 consecutive days at the 2s SD, the animals were subjected to tibial fracture surgery within 24 h. Of the 14 mice that begun training with decreased SD, only 8 animals were capable of reaching the criterion for tibial fracture surgery. Surgery was done between 0700 and 1,200 h. Animals were given ∼25–28 h to recover following surgery before 5-CSRTT testing and this proceeded on day 1 post-surgery between 1,300 and 1,600 h. Testing continued daily with 40 trials per day with the 2 s SD and 5 s LH for 7 days until animals reached pre-surgical performance levels.

For all phases of training and testing, task performance was assessed by three primary measures. Correct responses were defined as the number of trials in which the mice nose-poked into the correct aperture during the SD or LH divided by the total number of trials administered (40 trials) and expressed as a percentage. The accuracy rate was calculated as the total number of trials in which the mouse made a correct response divided by the number of trials in which the animal made a nose-poke response and was expressed as a percentage. Finally, the omission rate was calculated as the number of trials in which the mouse failed to respond divided by the total number of trials (40 trials) and expressed as a percentage. In addition, the latency to nosepoke (s) was recorded for each trial. On trials when a mouse failed to nose-poke, a maximum latency of 7 s was recorded (2 s SD + 5 s LH).

# Tissue Collection for Brain Samples

Under isoflurane anesthesia, mice were perfused transcardially with 30–50 mL PBS (Gibco, #10010-023), followed by 20–30 mL 4% paraformaldehyde (PFA) (#158127, Sigma-Aldrich, St. Louis, MO) in PBS, pH 7.4. Brains were harvested and post-fixed in 4% PFA at 4◦C for 24 h. Brains were sliced into 1 mm thick coronal slices according to specific stereotaxic coordinates (Bregma 1.4 to 0.4; −1.2 to −2.2; −2.2 to −3.2 mm) with a vibratome. Slices were incubated in 1x PBS at 4◦C for 24 h before beginning polymerization.

# Tissue Polymerization and Active Tissue Clearing

Each 1 mm thick slice was incubated in 1 mL X-CLARITYTM Hydrogel Solution (#C1310X; Logos Biosystems, Annandale, VA) at 4◦C for 24 h. After incubation, polymerization was performed using the X-CLARITYTM Polymerization System (Logos Biosystems) at 37◦C for 3 h. Active tissue clearing occurred using the X-CLARITYTM Tissue Clearing System (Logos Biosystems) and electrophoretic tissue clearing (ETC) solution (#C13001; Logos Biosystems) with the following setting: 0.9 A, 37◦C, for 3 h. After clearing, slices were washed 3x with PBS- 0.2% TritonTM-X100 (#T8787; Sigma-Aldrich) solution overnight on a shaker at room temperature.

#### Tissue Labeling

After tissue clearing, each slice was incubated with primary antibody diluted in 1 mL blocking buffer (10% normal donkey serum in PBS-0.2% TritonTM-X100 solution) at 4◦C for 3 days. Slices were washed 3x in PBS-0.1% TritonTM-X100 solution at 4 ◦C overnight. Slices were incubated in secondary antibody and DAPI diluted in 1 mL blocking buffer at 4◦C for 3 days. Slices were washed 3x in blocking buffer at 4◦C overnight and were stored in PBS at 4◦C before mounting. Details of the primary and secondary antibodies can be found in **Table 1**.

#### Imaging Setup

Slices were embedded with 1% agarose (#BP160-100, ThermoFisher Scientific, Waltham, MA) and placed in 1 mL syringes (#300013; BD Biosciences, San Jose, CA) without


tips and incubated with mounting solution (#C13101; Logos Biosystems) at 4◦C overnight followed by 1 h incubation at room temperature before imaging. Slices were imaged with a light sheet fluorescence microscope (LSFM, Z1; Zeiss, Germany) at the Duke University Light Microscopy Core Facility. During imaging, slices were immersed with mounting solution in a CLARITY optimized sample chamber. Slices were rotated to be parallel with the glass of the detection lens. All z-stacks and tile scans were acquired with 16-bit depth and 1,920 × 1,920 pixels using 5x or 20x objectives. Refractive index of the detection objective was set as 1.46. For laser excitations, the 405/488/561/640 filter set was used. Laser intensities were set between 0.5 and 2%. The exposure time for each frame was 29.97 ms. The step interval between frames was 0.659 µm.

#### Image Analyses

Tile scans were stitched together and presented as high resolution 3D rendering and corresponding movies using Arivis 4D viewer (Washington, DC). The 3D rendering of single z-stack was performed with Imaris software 8.21 (Bitplane USA, Concord, MA). Immunoreactivity intensity for each channel was calculated automatically with Imaris. For automated detection of glia cells the "surface" function of Imaris was used. Volume size of each glial cell was automatically calculated after removing those segmented at the edges of the image frame. The 3D algorithm-based surface rendering and quantification of fluorescence intensity for Iba-1, IgG, GFAP, and claudin-5 were performed with Imaris at 100% rendering quality. Each channel was analyzed separately. 3D surface rendering detected immunostained cells and their processes. The channel mean intensity filter was applied and minimum thresholds were used for all the experimental groups. Voxel-based surface rendering was applied to each channel. The preset parameters remained constant throughout the subsequent analysis of Iba-1, IgG, GFAP, and claudin-5 immunoreactivities. The quantification of microglia and astrocyte number (GFAP co-labeled with DAPI) was facilitated using the Spot tools of Imaris module. All the adjustments during image processing in Imaris and Arivis were performed under identical conditions for the surgical and control groups.

#### Dextran Imaging Studies to Detect Blood-Brain Barrier (BBB) Permeability

Mice were deeply anesthetized with isoflurane followed by transcardial infusion of TMR (Tetramethyl Rhodamine) 70kD dextran tracer (ThermoFischer Scientific) that was diluted in sterile PBS into 2 mM stocks and stored from light at −20◦C. After 5 min, mice were perfused using 20 mL of 0.1 M PBS. The brains were harvested immediately and embedded in optimal cutting temperature (OCT) compound on dry ice (Tissue-Tek, USA). The blocks were cut into 10µm sections using a cryostat (Microm HM550; Thermo Scientific, Waltham, MA) and the sections transferred to slides placed at room temperature. Once dried they were stored at −80◦C. On the day of staining, the slides were thawed at RT for 10 min and the sections fixed with 4% PFA for 10 min at room temperature followed by washing in PBS for 5 min. Subsequently, sections were blocked using blocking buffer made of 10% normal donkey serum (#D9663; Sigma-Aldrich) with 0.3% Triton X-100 in PBS for 1 h at RT. The slides were then incubated overnight at 4◦C with CD31 primary antibody (1:200, Goat CD31, R&D Systems, USA), diluted in the blocking buffer. The next day, the slides were washed with blocking buffer for 15 min followed by secondary antibody incubation in the above buffer for 1 h at room temperature. For CD31 immunostaining, donkey anti-goat Alexa 633 was used at a dilution of 1:500 (ThermoFischer Scientific). For staining of nuclei, DAPI (4,6-diamidino-2-phenylindole dihydrochloride, 1:1,000; Sigma-Aldrich) in PBS was incubated for 20 min. The sections were mounted, dehydrated and cover-slipped. Images were acquired using Zeiss 880 inverted confocal Airy scan. For the quantification, z-stack images were imported into Imaris x64 (version 9, Bitplane AG, Zurich, Switzerland) for 3D surface rendering and volumetric analysis. After display adjustment, the DAPI filter was removed to view the images in blend mode with rendering quality set to 100%. A volume filter was then applied to remove non-specific staining and minimum thresholds were used for both control and surgery groups. To quantify dextran volume, a "Filament Trace" was created from a masked surface channel of CD31. Later, by removing the CD31 filter the dextran volume was quantified using same threshold parameters for both control and surgery groups.

#### Statistics

The data were expressed as mean ± S.E.M and sample sizes (n) are provided in the figure legends. The behavioral data were analyzed with SPSS 25 (IBM SPSS Statistics, Chicago, IL) using repeated measures ANOVA (RMANOVA) where test day was used as the within-subject variable for 5-CSRTT. For open field testing, a comparison of distance moved for surgical animals compared to shams was examined with an independent samples t-test. Statistical analyses for the imaging data were made using Prism GraphPad 8.0 (GraphPad Software, San Diego, CA). Bonferroni corrected pair-wise comparisons were used for post-hoc analyses. In all cases, a P < 0.05 was considered statistically significant.

# RESULTS

#### 5-CSRTT

Attention impairment is a salient component of delirium (1). Here, we assessed attention using the 5-CSRTT task before and following orthopedic surgery (**Figure 1A**). Several measures of performance were evaluated, including the percent correct responses, accuracy, omitted responses, and the latency for the correct responses. A RMANOVA for the percent correct responses revealed a significant effect across days [F (7,49) = 21.716, p < 0.001; **Figure 1B**]. Mice were significantly impaired over the first 3 days post-tibial fracture (p ≤ 0.043–0.001). For percent accuracy the RMANOVA revealed a significant time effect [F(7,49) = 4.612, p < 0.001; **Figure 1C**], however, Bonferroni corrected post-hoc comparisons failed to find a significant reduction in accuracy on the first day following surgery compared to baseline (p = 0.346). Interestingly,

FIGURE 1 | Effects of tibial fracture on attention processes using the 5-choice serial reaction time task. (A) Schematic of the experimental design. (B) Tibial fracture significantly impaired overall performance over the first 3 days post-surgery as assessed by the percent correct responses out of a total of 40 trials/day. Percent accuracy (C) showed small but non-significant changes following surgery. Latency (s) to correct responding (D) was prolonged on day 1 post-surgery compared to baseline prior to surgery. (E) The percent omission errors increased significantly for 2 days post-surgery before returning to baseline. Baseline 5-CSRTT performance represents the mean performance on the last 4 prior to surgery. n = 8 mice; results are presented as means ± SEMs; (A–C) \*p < 0.05, \*\*p < 0.01 Bonferroni post hoc for post-surgical time point compared to pre-surgical "baseline".

the mean overall latency to respond to the correct nosepoke aperture (**Figure 1D**) was highly significant [F(7,49) = 4.164, p < 0.001]. Despite this small, albeit significant delay across test days, the Bonferroni comparisons failed to find a significant difference between day 1 post-surgery compared to baseline (p = 0.599). Tibial fracture was sufficient to induce an increase in omission errors [F(7,49) = 16.932, p < 0.001; **Figure 1E**] over the first 2 days post-surgery. Bonferroni analyses revealed significant increases in omission errors on days 1 and 2 post-surgery compared to the baseline omission rate (p ≤ 0.026–0.001). As a control, we evaluated general locomotor activity 1 day after surgery and found no significant differences, including changes in body weight or food intake, that may have confounded the 5-CSRTT results (**Supplementary Figure 1**). Taken together, these results indicate that attention processes as evaluated with the 5-CSRTT task are significantly impaired at least on the first day post-surgery, and this deficiency is not the result of any motor impairment.

# Surgery Alters Microglial Morphology in Clarified Hippocampus

Next we evaluated postoperative neuroinflammation using CLARITY. One mm thick brain slices were processed and cleared using electrophoresis (active clearing). Ortho-image and 3D reconstruction of the whole 1,000µm z-stack was performed to test imaging depth and antibody penetration, and it was found to be homogeneous across clarified slices (**Supplementary Figure 2**, **Supplementary Video 1**). Orthopedic surgery activates microglial cells in the hippocampus, with morphological changes observed at 24 h post-surgery by standard histology (23, 24). Thus, we have focused our analyses at this time point using CLARITY. Surgery significantly affected microglial morphology, with an overall reduction in their ramifications (**Figure 2A**). Retraction of thin-ramified processes is associated with microglial activation in response to changes in the CNS microenvironment, such as pro-inflammatory cytokines and infiltrating immune cells (25). Larger, more complex cells indicative of a surveillance state, were clearly identifiable in control slices but were absent 24 h after surgery. Using surface area 3D reconstruction in Imaris we confirmed the decrease in the average volume of microglial cells, with a significant overall loss of area coverage following surgery (p < 0.05, **Figure 2B**). Further studies in surgery group failed to reveal any gross changes in the number of microglia compared to control (**Figures 2C,D**). Consistent with altered microglial morphology in the surgery group, we observed an increased translocation of transcription factor EB (TFEB), a master regulator of the autophagy-lysosome system along with the LAMP-1 expression that coordinates autophagy and lysosomal biogenesis (p < 0.001, **Figures 2E,F**). These results suggest that higher activity of lysosomal biogenesis in activated microglia is closely linked to the inflammation in the CNS.

# Time-Dependent BBB Opening After Surgery

BBB disruption is a key hallmark of many neurodegenerative conditions and it often correlates with disease progression (26). Here we used CLARITY to evaluate IgG deposition in whole hippocampal tissue 24 h after surgery (p < 0.001, **Figures 3A,B**). Surgery increased IgG deposition as compared to naive control mice (**Supplementary Video 2**). We confirmed these results using another marker for BBB disruption, fibrinogen, which was also up-regulated in the vessels of operated mice (p < 0.001, **Figures 3C,D**). Parenthetically, fibrinogen is a blood coagulation protein deposited in the brain during neurodegeneration that drives microglial activation (via CD11b receptor binding) and cognitive dysfunction (27). Interestingly, experiments using the 70 kD dextran tracer perfused at 24 h after surgery revealed no BBB leakage compared to control (**Figures 3E,F**), suggesting the barrier opens transiently within the first 24 h after surgery.

# Astrocytic and Endothelial Dysfunction After Surgery

Astrocytes are critical to the formation and maintenance of the BBB (28). At 24 h after surgery we found that orthopedic surgery induced significant changes in the expression of the intermediate filament GFAP in clarified hippocampal slices (**Figure 4A**). We evaluated this increase by using surface rendering in Imaris, which revealed changes both in the number of GFAPpositive astrocytes (p < 0.05, **Figure 4B**) as well as coverage area of these enlarged cells (p < 0.001, **Figure 4C**). Next, we focused on the expression of aquaporin-4 (AQP4), a water channel that is highly localized in the end feet of astrocytes that are in contact with BBB. Alterations in the expression of AQP4 has been linked to neurodegenerative diseases (29). AQP4 levels were reduced in the hippocampus of surgical mice compared to naïve controls (p < 0.001, **Figures 4D,E**). Finally, using CLARITY we measured alterations in claudin-5 (cld-5), one of several intracellular tight junction proteins that form BBB (**Figures 4F,G**). Surgery caused a significant disruption of cld-5 compared to control 24 h after surgery. Interestingly, the loss of cld-5 coverage was especially notable in the granular cell layer, as well as other areas of the hippocampus. Collectively, these results suggest that orthopedic surgery triggers astrogliosis and endothelial impairments in the hippocampus.

# DISCUSSION

This study provides new evidence for the impact of peripheral surgery on a cognitive domain of major relevance to postoperative delirium. We also explored the contribution of postoperative neuroinflammation and endothelial dysfunction using CLARITY in this clinically-relevant model of orthopedic surgery.

Inattention is one of the core features of delirium and a mandatory component for diagnosis in compliance with DSM-5 criteria (30). However, delirium is a very heterogeneous syndrome characterized by multiple features and subtypes, which often make its characterization challenging and under diagnosed in hospital settings (31). Other cardinal features of delirium include an acute onset with fluctuating course, disorganized thinking, and altered level of consciousness, which can be difficult to define both in humans and preclinical models. In rodents, we and others have previously reported on different behavioral deficits affected by orthopedic surgery. These include changes in fear conditioning (23, 24, 32, 33), novel object recognition memory/open field (33, 34), zero maze (35), Y-maze (32, 36), and Morris water maze (37). While spatial and working memory tasks have overlapping mechanisms with attention (38), the 5- CSRTT allowed us to examine and manipulate more discrete aspects of attention. The 5-CSRTT is an appetitive task and is one of the most robust assays to study visuospatial attention in rodents, which is analogous the commonly used continuous performance task used measure of attention in both clinical and research settings (39). Our results using the 5-CSRTT after surgery demonstrate the most profound impairment occurs

the microglia in control and surgery groups. Tibial fracture increased the expression of both LAMP1 and TFEB puncta in the microglia. The results are presented as means ± SEMs, \*p < 0.05, \*\*\*p < 0.001 Student's t-test (n = 3 for autophagy markers).

24 h after surgery with full recovery by day 5 post-procedure. This time-course is very similar to the clinical manifestation of delirium and it provides a valid endpoint to test possible interventions. Notably, discrete aspects of 5-CSRTT performance are centrally mediated by cholinergic and other transmitter manipulations (40). Acetylcholine (ACh) has been strongly linked to selective attention processing across species, including humans (41), and cholinergic dysfunction has been proposed as a key feature of delirium (42, 43). Recent data from elective surgical patients suggest that acetylcholinesterase activity, the enzyme responsible for degrading ACh, is higher in patients with delirium compared to non-delirious controls, supporting

Student's t-test (n = 4).

the cholinergic hypothesis for postoperative delirium (44). Similar findings in rodents suggest that cholinergic depletion predisposes to acute cognitive impairments and microgliosis after systemic LPS challenge (45). Here, we described changes in the expression of autophagy markers in microglia, which may relate to the systemic inflammatory response and monocytic infiltration following orthopedic surgery (46–48). Whether autophagy is induced by this peripheral response or centrally as a protective mechanism for microglia to resolve neuroinflammation needs further elucidation. Gliosis and endothelial dysfunction have been recently observed in critically ill patients with delirium (49). Using CLARITY we have implemented a staining protocol to improve quantification of these changes from standard immunofluorescence and are now better able to evaluate morphological changes in these highly complex structures. Vascular patterns of BBB disruption can be observed in the hippocampus at 24 h. Interestingly, injection of 70 kDa dextran at this time point revealed no significant parenchymal leak,

suggesting the barrier opens transiently. These changes can be detected using other markers, such as IgG and fibrinogen. This is consistent with other work where laparotomy triggered changes in selective tight junctions, including claudin-1, occluding, ZO-1, and 10 kDa dextran but not 70 kDa (50). Thus, it is possible for peripheral cytokines to enter the CNS and trigger delirium-like behaviors after surgery. Circulating plasma proteins, including pro-inflammatory makers like C-reactive protein and IL-6, have been recently correlated with patients developing delirium after non-cardiac surgery (51). Importantly, we previously found IL-6 to be upregulated both in the periphery as well as in the CNS (23), and administration of a selective monoclonal antibody was shown to prevent postoperative neuroinflammation and cognitive decline in this model (52). Selective blockade of pro-inflammatory cytokines may have translational implication for delirium treatment, although side effects associated with excessive immunosuppression and/or impaired wound healing will need to be evaluated carefully. Since, pro-inflammatory cytokines are detected in the cerebrospinal fluid (CSF) of patients with delirium (15, 53), blocking these mediators in the CSF without affecting the peripheral immune response may reveal alternative approaches to prevent surgery-induced neuroinflammation. Finally, many of these cytokines are already upregulated at baseline in patients at-risk for delirium. In fact, aging and chronic stress elevates IL-6 amongst other mediators (54, 55) and these cytokines can further prime microglia cells and possibly impair neuro-immune circuits of relevance to delirium-like behavior. Indeed surgery also triggers acute plasma cortisol, which was found elevated in patients with delirium after cardiac surgery (56). However, recent trials using steroids have not shown significant effects on delirium outcomes after surgery and critical illness (57, 58). In this context, delirium may be considered a surrogate marker for neuroinflammation in selective brain areas that may contribute to the different clinical manifestations of this disorder (59), thus requiring more selective disease-modifying therapies to treat and possibly prevent.

Preclinical models to study postoperative delirium have been notably limited. Systemic infective challenges (e.g., lipopolysaccharide or LPS) have been used to examine certain aspects of behavioral change. Culley et al. (60) have evaluated the effects of LPS using the attention set-shifting task (AST) and found selective impairments in attention/executive function in aged rats for up to 72 h post-surgery. Parenthetically, AST involves a series of discriminations based on stimulus dimensions such as digging medium or shape cue in a task analogous to the Wisconsin Card Sorting Task. Without affecting initial discrimination learning, LPS impaired reversal-learning at 24 but not at 48 h and extra-dimensional shift discrimination at 72 h. Transient changes in working memory have been similarly described in other delirium models of neurodegeneration after LPS exposure using the Y maze, T maze alternation tasks, and novel object recognition memory (61). Peng et al. (62) have evaluated certain aspects of delirium in a mouse model of abdominal surgery using composite Z scores from a battery of tests including the buried food test, open field test and Y maze. They found that anesthesia and surgery impaired both natural and learned behavior of the mice with acute onset and fluctuating course, modeling one of the cardinal features of delirium. Further, other studies have described behavioral changes after anesthesia exposures that can affect neuronal circuits related to recognition memory, arousal, and motoric behaviors (63, 64). In this study we have not included an anesthesia exposure group for several reasons. First, we previously reported that isoflurane exposure in this exact model does not trigger significant neuroinflammation, or behavioral deficits (23). We have confirmed this by assessing sham and surgical mice in the open field without observing significant deficits. Second, recent work by Hambrecht-Wiedbusch et al. showed no effects of general anesthesia (including ketamine exposure) focusing on attention in healthy rats using the 5-CSRTT (65). Last, we allowed the mice to have a full 24 h recovery before resuming testing 25 h later in order to avoid any possible effects of isoflurane exposure on food intake or nausea; although studies have shown that a single exposure of isoflurane in C57BL/6 mice does not have significant effects on food intake for up to 9 days post exposure (66).

This study presents limitations and challenges to this emerging field of preclinical research. Inattention is one of many features of delirium; in fact, its fluctuating course was not addressed by these experiments and may require other correlates, such as electrophysiology, to fully ascertain in rodents. Further, not all patients experience delirium. However, our data indicate that mice have different degrees of attentional impairment so this may provide opportunities to identify resiliency traits that may impact on perioperative cognitive outcomes. Finally, we have not addressed the contribution of advanced age and other risk factors into this model, which will be objective of future work as well as more mechanistic experiments to further identify the contribution of neuroinflammation and endothelial dysfunction as key drivers of delirium pathogenesis. Defining more specific targets at the neurovascular unit may provide much needed therapies to resolve neuroinflammation and inform future clinical trials.

In conclusion we have provided translational evidence using a well-established surgical model to study attention processes and features relevant to the pathophysiology of postoperative delirium. The utility of these endpoints can be translated into trials evaluating biomarkers of neuro-glia and endothelial dysfunction after surgery, which may contribute to a better understanding on the genesis of postoperative delirium.

# 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.

# AUTHOR CONTRIBUTIONS

CX optimized the CLARITY protocol. RV, CX, and SS analyzed light-sheet images. RV performed dextran experiments and immunostainings. FP and RR conducted and analyzed behavioral experiments. RR, WW, and NT designed research and provided new analytical tools. NT drafted the manuscript. All authors read and approved the final manuscript.

#### FUNDING

This study was supported by NIH grants R01AG057525, and R21AG055877-01A1, the Duke Institute for Brain Sciences, and a DREAM Innovation Grant from Duke Anesthesiology to purchase the X-CLARITY setup (to NT). Some of the behavioral experiments were conducted with equipment and software funded with a North Carolina Biotechnology Center grant (to WW). The Zeiss light sheet was purchased with an NIH S10 grant 1S10OD020010-01A1 (PI Lisa Cameron).

#### ACKNOWLEDGMENTS

NT attended the hands-on CLARITY training workshops managed and developed by Kristin Overton, Ph.D. and Karl Deisseroth, M.D., Ph.D. at Stanford University. We thank Ping Wang, Ph.D. (Terrando lab, Duke University) for providing the image of the agar mount; Benjamin Carlson, Ph.D. (Light Microscopy Core Facility, Duke University) for excellent technical support with the light sheet imaging; and Christopher Means (Behavioral Core Facility at Duke University) for assistance in behavioral testing.

#### REFERENCES


# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02675/full#supplementary-material

Supplementary Figure 1 | Schema for 5-choice serial reaction time task at the time of tibial fracture. (A) A trial is initiated once a mouse is placed into a darkened 5-CSRTT chamber. Testing begins when one of the five nose-poke apertures is randomly illuminated. The nose poke aperture is lit for 2 s and the mouse has 5 s to respond after the light is extinguished. If the mouse responds with a nose-poke in the lit aperture ("correct response"), the food magazine is immediately illuminated and a 20 mg chocolate food pellet ("reward") is dispensed. The mouse has 5 s to consume the food reward, followed by a 5 s inter-trial interval (ITI) before the commencing of the next trial. If the mouse responds with a nose-poke into an unlit aperture ("incorrect response") or fails to respond at all ("omission"), the mouse is given a 5 s time-out (TO), where all lights are extinguished, and the 5 s ITI. (B) Open field activity in mice 24-h following tibial fracture. No significant differences were found between sham (n = 10) and mice subjected to tibia fracture (n = 10) for distance (cm) traveled during a 5 min open field tested 24 h post-surgery. (C,D) represent body weight and food intake in the mice exposed to the 5-CSRTT task. No significant changes were observed throughout the duration of the study. Results are presented as means ± S.E.M.

Supplementary Figure 2 | Schema of the clearing process and transparency of 1 mm coronal brain slices. (A) The transparency and agar mount for 1 mm mouse brain slice after tissue clearing. (B) Imaging depth was verified using Iba-1 (green)/DAPI (blue) immunostaining and (C) staining quality evaluated at every 200µm z-stack intervals up to 1,000µm (20X objective, stack size 1 mm). n = 5 slices. Scale bars: 4 mm (A), 200µm (B), 100µm (C).

Supplementary Video 1 | Iba1+ microglia in the hippocampus. Cleared brain slices were stained with Iba1 (Green) and DAPI (Blue). The video was generated using Arivis 4D viewer.

Supplementary Video 2 | IgG deposition in the hippocampus 24 h after surgery. Endothelial permeability was assessed by IgG staining (Green) followed by DAPI counterstain (Blue). The video was generated using Arivis 4D viewer.

a case of effect modification. Am J Geriatr Psychiatry. (2016) 25:308– 15. doi: 10.1016/j.jagp.2016.10.008


rat model of the Metabolic Syndrome. Anesthesiology. (2013) 118:1098– 105. doi: 10.1097/ALN.0b013e318286d0c9


and delirium in the steroids in cardiac surgery trial: a randomized, double-blind, placebo-controlled substudy. Anesthesiology. (2017) 126:223– 33. doi: 10.1097/ALN.0000000000001433


**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 Velagapudi, Subramaniyan, Xiong, Porkka, Rodriguiz, Wetsel and Terrando. 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.

# Diurnal Variation in Systemic Acute Inflammation and Clinical Outcomes Following Severe Blunt Trauma

Akram M. Zaaqoq1,2†, Rami A. Namas 1,3†, Othman Abdul-Malak <sup>1</sup> , Khalid Almahmoud<sup>1</sup> , Derek Barclay <sup>1</sup> , Jinling Yin<sup>1</sup> , Ruben Zamora1,3, Matthew R. Rosengart 1,2 , Timothy R. Billiar 1,3 and Yoram Vodovotz 1,3 \*

*<sup>1</sup> Department of Surgery, University of Pittsburgh, Pittsburgh, PA, United States, <sup>2</sup> Department of Critical Care Medicine, MedStar Washington Hospital Center, Washington, DC, United States, <sup>3</sup> Center for Inflammation and Regeneration Modeling, McGowan Institute for Regenerative Medicine, University of Pittsburgh, Pittsburgh, PA, United States*

#### Edited by:

*Valentin A. Pavlov, Northwell Health, United States*

#### Reviewed by:

*Ioannis P. Androulakis, Rutgers, The State University of New Jersey, United States Charles Wade, University of Texas Health Science Center at Houston, United States*

#### \*Correspondence:

*Yoram Vodovotz vodovotzy@upmc.edu*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *13 September 2019* Accepted: *04 November 2019* Published: *20 November 2019*

#### Citation:

*Zaaqoq AM, Namas RA, Abdul-Malak O, Almahmoud K, Barclay D, Yin J, Zamora R, Rosengart MR, Billiar TR and Vodovotz Y (2019) Diurnal Variation in Systemic Acute Inflammation and Clinical Outcomes Following Severe Blunt Trauma. Front. Immunol. 10:2699. doi: 10.3389/fimmu.2019.02699* Animal studies suggest that the time of day is a determinant of the immunological response to both injury and infection. We hypothesized that due to this diurnal variation, time of injury could affect the systemic inflammatory response and outcomes post-trauma and tested this hypothesis by examining the dynamics of circulating inflammatory mediators in blunt trauma patients injured during daytime vs. nighttime. From a cohort of 472 blunt trauma survivors, two stringently matched sub-cohorts of moderately/severely injured patients [injury severity score (ISS) >20] were identified. Fifteen propensity-matched, daytime-inured ("mDay") patients (age 43.6 ± 5.2, M/F 11/4, ISS 22.9 ± 0.7) presented during the shortest local annual period (8:00 am−5:00 pm), and 15 propensity-matched "mNight" patients (age 43 ± 4.3, M/F 11/4, ISS 24.5 ± 2.5) presented during the shortest night period (10:00 pm−5:00 am). Serial blood samples were obtained (3 samples within the first 24 h and daily from days 1–7) from all patients. Thirty-two plasma inflammatory mediators were assayed. Two-way Analysis of Variance (ANOVA) was used to compare groups. Dynamic Network Analysis (DyNA) and Dynamic Bayesian Network (DyBN) inference were utilized to infer dynamic interrelationships among inflammatory mediators. Both total hospital and intensive care unit length of stay were significantly prolonged in the mNight group. Circulating IL-17A was elevated significantly in the mNight group from 24 h to 7 days post-injury. Circulating MIP-1α, IL-7, IL-15, GM-CSF, and sST2 were elevated in the mDay group. DyNA demonstrated elevated network complexity in the mNight vs. the mDay group. DyBN suggested that cortisol and sST2 were central nodes upstream of TGF-β1, chemokines, and Th17/protective mediators in both groups, with IL-6 being an additional downstream node in the mNight group only. Our results suggest that time of injury affects clinical outcomes in severely injured patients in a manner associated with an altered systemic inflammation program, possibly implying a role for diurnal or circadian variation in the response to traumatic injury.

Keywords: circadian rhythm, blunt trauma, chemokines, nervous system, acute inflammation

# INTRODUCTION

Despite advances in critical care over the past 40 years, severe blunt trauma is still associated with significant long-term morbidity and mortality (1, 2). Systemic acute inflammation is thought to be a key driver of post-injury critical illness (3). Although properly regulated inflammation is crucial for promoting adequate tissue healing and recovery, an overly exuberant or insufficient response may result in immune dysregulation as well as secondary tissue and organ damage that can be complicated by prolonged hospitalization (4–6).

The timing of the light-dark cycle results in a shift of the circadian rhythms, functioning to synchronize and coordinate organ systems in response to environmental light dynamics (7, 8). The immune system is under direct circadian control by systemic cues and molecular clocks within immune cells (9). These oscillations may also help to promote tissue recovery and the clearance of potentially harmful cellular elements from the circulation (8, 9). The suprachiasmatic nucleus (SCN) of the anterior hypothalamus contains specialized neurons that receive photo input through the retinohypothalamic tract (RHT) and non-photo cues by disparate neural inputs (9, 10). Subsequently, the nervous system regulates the inflammatory response through neuronal and neuroendocrine pathways (7).

At the molecular level, there are multiple sets of transcription factors that result in autoregulatory transcription-translation feedback loops of core clock genes, such as BMAL1 and CLOCK, which in turn control the output of circadian physiology and behavior (11). Recent studies have suggested that this complex system impacts inflammatory responses in the context of sepsis (12–14).

To address whether the time of injury alters trauma-induced systemic inflammation and, thereby, trauma-related outcomes, we retrospectively analyzed data from a large cohort of blunt trauma patients who survived up to discharge. A granular, temporal sampling of the early systemic inflammatory response of both the overall cohort as well as in propensity-matched subcohorts of patients injured during the day vs. the nighttime, combined with data-driven computational methods, allowed us to define differential dynamic inflammation networks as a function of time of injury. Our analyses revealed that early, persistent changes in post-injury inflammation manifest in unique biomarker patterns associated with the time of injury. Also, these patterns, which are independent of the mechanism of injury, injury severity, age, or gender suggest a diurnal, and perhaps also circadian, control of post-trauma inflammation that impacts clinical outcomes.

# MATERIALS AND METHODS

The study protocol was reviewed and approved by the University of Pittsburgh Institutional Review Board (IRB), and has been detailed previously (15, 16). Written informed consent was taken from each patient or the next of kin in line with our locally agreed protocols with Institutional Review Board regulations. Patients eligible for enrollment in the study were at least 18 years of age, admitted to the intensive care unit (ICU) within

24 h of injury, and, per the treating physician, were expected to live more than 24 h. Reasons for ineligibility were isolated head injury, penetrating trauma (due to our focus on blunt trauma), and pregnancy. Blunt trauma patients were enrolled in the study from 2004 to 2012. Laboratory results and basic demographic data were recorded in the database directly from the electronic medical record. Three plasma samples, starting with the initial blood draw upon arrival to the emergency department (ED), were assayed within the first 24 h following trauma and then daily between 4:00 and 5:00 a.m. from days 1 to 7 post-injury. The blood samples were centrifuged, and plasma aliquots were stored in cryoprecipitate tubes at −80◦C for subsequent analysis of inflammatory mediators.

#### Study Design

A retrospective study involving a cohort of 472 blunt trauma survivors (330 males and 142 females, age 48.4 ± 0.9, ISS 19.6 ± 0.5) who were admitted to the emergency department of the Presbyterian University Hospital (Level 1 trauma center) (16) (**Figure 1**). Exclusion criteria included patients with evidence of alcohol intoxication upon admission due to the potential impact of alcohol on systemic inflammation. From this cohort, "Day"


TABLE 1 | Demographics, mechanism of injury, co-morbid conditions, and clinical outcomes of the total trauma cohort compared to the Day and Night cohorts.

*Values are expressed as mean* ± *SEM. One-Way ANOVA or Fisher exact test were used as appropriate with statistical significance set at P* < *0.05.*

patients were identified based on the time of presentation to Presbyterian University Hospital during the shortest daylight in Pittsburgh, PA, USA throughout the year (8:00 a.m.−5:00 p.m.) and "Night" patients presented during the shortest night period (10:00 p.m.−5:00 a.m.) (**Figure 1**). The overall demographics, mechanism of injury, clinical data, and co-morbidities of the 174 (121 males and 53 females, age 52.8 ± 1.4, ISS 17.3 ± 0.6) patients in the Day group vs. 33 (22 males and 11 females, age 53.3 ± 3.3, ISS 16.7 ± 1.7) patients in the Night group are shown in **Table 1**. Initially, we sought to avoid the confounding impact related to the type of mechanism of injury by selecting patients based on the predominant mechanism of injury in both cohorts, i.e., motor vehicle accidents (MVA). Given that the Night group exhibited statistically significantly higher rates of head and extremity injuries when compared to the Day group (see **Table 1**) as revealed by the abbreviated injury scale (AIS), we next performed a one-to-one propensity matching based on age, sex, and ISS > 20, which yielded two sub-cohorts of 15 Day patients (mDay) matched to 15 Night patients (mNight) (**Figure 1**).

#### Clinical Data Collection

Clinical data, including injury severity score (ISS), abbreviated injury scale (AIS) score, Marshall Multiple Organ Dysfunction (MOD) score, ICU LOS, hospital LOS, and days on mechanical ventilation were collected from the hospital inpatient electronic and trauma registry database. ISS (17) and AIS scores (18) were calculated for each patient by a single trauma surgeon after attending radiology evaluations were finalized. The ISS is based on an anatomical scoring system that provides an overall score for patients with multiple injuries (17). Each injury is assigned an AIS score, allocated to one of six body regions: head, face, chest, abdomen, extremities (including pelvis), and external. We focused our studies on trauma patients with ISS > 20, which would be considered moderate/severe (19).

As an index of organ dysfunction, the MOD score (20) (ranging from 0 to 24) was calculated. In brief, six variables were obtained from the electronic trauma data registry including (a) the respiratory system (PO2/FIO<sup>2</sup> ratio); (b) the renal system (serum creatinine concentration); (c) the hepatic system (serum bilirubin concentration); (d) the hematologic system (platelet count); (e) the central nervous system (Glasgow Coma Scale); and (f) the cardiovascular system- the pressure-adjusted heart rate (PAR).

#### Analysis of Inflammatory Mediators

Blood samples were collected into citrated tubes via indwelling catheters within 24 h of admission and daily for 7 days postinjury. The blood samples were centrifuged, and plasma aliquots were stored in cryoprecipitate tubes at −80◦C for subsequent analysis of inflammatory mediators. The human inflammatory MILLIPLEXTM MAP Human Cytokine/Chemokine Panel-Premixed 26 Plex (Millipore Corporation, Billerica, MA) and LuminexTM 100 IS (Luminex, Austin, TX) was used to measure plasma levels of interleukin (IL)-1β, IL-1 receptor antagonist (IL-1Ra), IL-2, soluble IL-2 receptor-α (sIL-2Rα), IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-13, IL-15, IL-17A, interferon (IFN)-γ, IFN-γ inducible protein (IP)-10 (CXCL10), monokine induced by gamma interferon (MIG; CXCL9), macrophage inflammatory protein (MIP)-1α (CCL3), MIP-1β (CCL4), monocyte chemotactic protein (MCP)-1 (CCL2), granulocytemacrophage colony stimulating factor (GM-CSF), Eotaxin, and tumor necrosis factor-alpha (TNF-α). The human Th17 MILLIPLEX Panel kit (Millipore Corporation, Billerica, MA) was used to measure IL-9, IL-21, IL-22, IL-23, IL-17E/25, and IL-33. The LuminexTM system was used in accordance to the manufacturer's instructions. Soluble ST2 (sST2) was measured by a sandwich ELISA assay (R&D Systems, Minneapolis, MN). NO<sup>−</sup> 2 /NO<sup>−</sup> <sup>3</sup> was measured using the nitrate reductase/Griess assay (Cayman Chemical Co., Ann Arbor, MI). Serum cortisol and transforming growth factor (TGF)-β1 were assayed using commercially available enzyme-linked immunosorbent assays (ELISA) kits (R&D Systems, Minneapolis, MN) according to the manufacturer's instructions. In brief, active vs. latent TGF-β1 were assayed as follows: To collect platelet-poor plasma, plasma was collected on ice using EDTA as an anticoagulant and centrifuged for 15 min at 1,000 × g within 30 min of collection followed by an additional centrifugation step of the plasma at 10,000 × g for 10 min at 2–8◦C for complete platelet removal. To activate latent TGF-β1 to immunoreactive TGF-β1 detectable by the Quantikine <sup>R</sup> TGF-β1 immunoassay, 20 µL of 1 N HCl was added to each 40 µL of plasma, mixed and incubated for 10 min at room temperature. Next, the acidified samples were neutralized by adding 20 µL of 1.2 N NaOH/0.5 M HEPES, mixed and then diluted with calibrator diluent prior to the assay. To perform the assay, 50 µL of diluent RD1-73 were added to each well-followed by addition of 50 µL of recombinant human TGF-β1 standard, control, or activated sample per well and incubated for 2 h at room temperature. Following 4 aspirations/wash steps, 100 µL of TGF-β1 conjugate was added to each well and incubated for an additional 2 h at room temperature, washed, and then 100 µL substrate solution was added to each well and incubated for 30 min at room temperature. Finally, 100 µL of stop solution was added to each well and the optical density (wavelength set at 450 nm) was determined.

#### Statistical Analysis

All data are expressed as mean ± SEM. Group-time interaction of plasma inflammatory mediators' levels was determined by Two-way analysis of variance (ANOVA) which was confirmed by non-parametric Mann-Whitney U test to compare the P-values generated by the Twoway ANOVA in the case where values were not normally distributed, all using SigmaPlotTM 11 software (Systat Software, Inc., San Jose, CA). Fisher's exact test was performed for categorical data using Graphpad PRISM (GraphPad Software, Inc., La Jolla, CA). The correlation between different inflammatory mediators was determined by Spearman's correlation using the actual values of these inflammatory mediators. P < 0.05 was considered statistically significant for all analyses.

#### Dynamic Bayesian Network Inference

Dynamic Bayesian Network (DyBN) inference was carried out to define the most likely single-network structure that best characterizes the dynamic interactions among systemic inflammatory mediators across time, in the process suggesting likely feedback structures that define central nodes. The networks might also suggest possible mechanisms by which the progression of the inflammatory response differs within a given experimental group. This analysis was carried out using MATLABTM (The MathWorks, Inc., Natick, MA), using an algorithm adapted from Grzegorczyk and Husmeier (21) and revised by our group (22). In this analysis, inflammatory mediators were represented at multiple time points within the same network structure. In this approach, time was modeled discretely as in a discrete Markov chain. Each mediator was given a time index subscript indicating the time slice to which it belonged. Additional temporal dependencies were represented in a DyBN by edges between time slices. Each node in the network was associated with a conditional probability distribution of a variable that is conditioned upon its parents (upstream nodes). This particular network structure was used to assess the dominant inflammatory mediators and the probable interaction among various mediators, including possible feedback (21).

#### Data-Driven Modeling: Dynamic Network Analysis (DyNA)

DyNA was carried out as described previously (15, 23, 24). The goal of this analysis was to gain insights into dynamic changes in network connectivity of the post-traumatic inflammatory response for both day and night over time. The mathematical formation of this method is essential to calculate the correlation between the variables by which we can examine their dependence. To do so, inflammatory mediator networks were created in adjacent 8-h time periods (0–8, 8– 16, and 16–24 h) using MATLAB <sup>R</sup> (The MathWorks, Inc., Natick, MA). Connections in the network were created if the correlation coefficient between two nodes (inflammatory mediators) was greater or equal to a threshold of 0.7. For the network density calculation, to account for network sizes (number of significantly altered nodes) in the adjacent 8-h time periods detailed above, we utilized the following formula: [a minor revision of the one reported by Assenov et al. (25)].

> Total number of edges ∗ Number of total nodes Maximum possible edges among total nodes

# RESULTS

#### Overview of Demographics and Outcomes in a Large Cohort of Blunt Trauma Patients

Over the 8-year study period, 472 blunt trauma patients admitted to the ICU were enrolled in the study, as previously described (15, 16). The majority of the 472 trauma patients were males (70.6 %), with a mean age of 48.4 ± 0.9 years and a mean ISS of 19.6 ± 0.5. These patients sustained blunt trauma in the form of MVA (57%), falls (21.6%), motorcycle accidents (13.8%), and others (7.6%). The AIS analysis revealed that the Night cohort exhibited statistically significantly higher rates of head (1.69 ± 0.3 vs. 0.89 ± 0.1; P < 0.003) and extremity (1 ± 0.2 vs. 1.5 ± 0.09; P = 0.017) injuries when compared to the day group (**Table 1**). The average ICU LOS was 7.01 ± 0.36 d, the mean hospital LOS was 12.7 ± 0.4 d, and the mean number of days on a mechanical ventilator was 3.1 ± 0.3 d.

#### Characteristics of Day and Night Injury Cohorts: Demographics, Outcomes, and Propensity Matching

A total of 174 patients met our definition of being injured during the day and 33 patients injured during the night as defined in the Materials and Methods (**Table 1**). Overall, males were predominant in both the Day and Night cohorts (68.9 and 66.7%, respectively), with no statistical difference in mean age (52.8 ± 1.4 vs. 53.5 ± 3.3; P = 0.9) between the two cohorts. Also, there was no statistically significant difference in ISS (17.3 ± 0.6 vs. 16.7 ± 1.7; P = 0.5), ICU LOS (5.9 ± 0.5 vs. 6.7 ± 1.3; P = 0.8), hospital LOS (11.6 ± 0.7 vs. 12.1 ± 1.7; P = 0.7), days on mechanical ventilation (2.7 ± 0.5 vs. 2.2 ± 0.7; P = 0.7), the prevalence of nosocomial infection (NI: 39/173 [22.5%] vs. 8/33 [24.2%]; P = 0.8), or the average Marshall MOD score across days 1 to 7 (1.3 vs. 1.6; P = 0.9) between the two cohorts.

A total of 15 patients in the Night cohort (matched Night [mNight]) were matched with 15 patients in Day cohort

(matched Day [mDay]), according to age, gender, and ISS >20. Overall, males were predominant in both mDay and mNight sub-cohorts (73.3% in both sub-cohorts), with no statistical difference in mean age (43.6 ± 5.2 vs. 43 ± 4.3; P = 0.9; respectively) between the two sub-cohorts. Moreover, ISS was not statistically significantly different between both groups (22.9 ± 0.7 vs. 24.5 ± 2.5; P = 0.9; respectively; **Figure 2A**). Importantly, there were no statistically significant differences in any of the body regions between both subcohorts based on AIS body regions (the components of the ISS; see Materials and Methods) (**Figure 2B**).

TABLE 2 | Demographics, mechanism of injury, co-morbid conditions, and clinical outcomes of the stringently matched (m)Day and mNight sub-cohorts.


*Values are expressed as mean* ± *SEM. One-Way ANOVA or Fisher exact test were used as appropriate with statistical significance set at P* < *0.05.*

#### Greater Requirement for Surgical Interventions Needed for Trauma Management, as Well as Worse Clinical Outcomes, in Stringently Matched Night vs. Day Sub-cohorts

The mNight group required more surgical interventions than the mDay group. These interventions were mainly either reductions/fixations for simple or compound fractures: 5/15 (33.3%) of the mNight patients had orthopedic intervention vs. 2/15 (13.3%) in the mDay sub-cohort (P = 0.02). There was no difference in the rate of exploratory laparotomies between the sub-cohorts. In addition, there was no statistically significant difference in the rates of NI (5/15 [33.3%] vs. 2/15 [13.3%]; P = 0.2) and average Marshall MOD score (1.3 vs. 0.9; P = 0.4) between the two sub-cohorts. Interestingly, though there was no statistically significant difference in the requirement for mechanical ventilation (0.8 ± 0.3, 2.3 ± 1.1; P = 0.9), there was a statistically significantly longer ICU LOS (2.6 ± 0.3 vs. 8.5 ± 2.3; P = 0.04) and total hospital LOS (7.6 ± 0.9 vs. 15.5 ± 3; P = 0.02) in mNight sub-cohort when compared to the mDay group (**Table 2**).

# Divergent Systemic Inflammatory Responses in Stringently Matched Day vs. Night vs. Sub-cohorts

Since trauma and subsequent organ dysfunction elicit a systemic inflammatory response which is regulated in part by diurnal and circadian rhythms, we hypothesized that the dynamics of inflammatory mediators could differ according to the time of injury and day/night cycle. We observed that IL-17A (P = 0.008; **Figure 3A**) and sST2 (P = 0.022; **Figure 3B**) were significantly higher in the mNight sub-cohort compared to the mDay sub-cohort. In contrast, IL-7 (P < 0.001; **Figure 3C**), IL-15 (P = 0.002; **Figure 3D**), GM-CSF (P = 0.001; **Figure 3E**), and MIP-1α (P = 0.047; **Figure 3F**) were significantly lower in the mNight sub-cohort. Notably, the circulating levels of the ligand for ST-2, IL-33 (26, 27), which we previously reported as being elevated in trauma patients (28), were not significantly different in the mDay vs. mNight cohorts (**Figure 3G**). Though cortisol is a key mediator whose levels vary with time of day in healthy individuals (29, 30), no differences in circulating cortisol levels were observed between mDay and mNight patients (**Figure 3H**). Also, there was no statistically significant difference in active, latent, and total TGF-β1 (**Figure 4**) between the mDay and mNight subcohorts.

#### Different Dynamic Inflammatory Networks Inferred in Day vs. Night Patients

We next sought to gain insights into the systemic inflammatory programs present in mDay and mNight sub-cohorts by examining dynamic network connectivity among inflammatory mediators using Dynamic Network Analysis (DyNA) (15, 23, 24). In addition to determining which networks were present at specific time intervals, we also assessed the total degree of connectivity at each of these intervals. **Figure 5** shows the detailed DyNA results for Day and Night in three different time periods following presentation (0–8, 8–16, and 16–24 h). We focused especially on DyNA connectivity among nodes with ≥ 4 connections. In the mDay group, these highly connected sub-networks initially involved MCP-1/IL-10/sIL-2Rα/IL-5/IL-7/IFN-γ/IL-6, and IL-1RA (0–8 h), and then MCP-1/IL-1β/IL-2/IL-5/IL-13/IL-10/IL-15/IFN-γ over the period of 8–16 h (**Figure 5**).

In contrast, network analysis of the mNight sub-cohort data revealed interactions among IL-4/IL-5/IL-6/IL-8/IL-13/IL-17A/sIL-2Rα/IL-1Rα/MIP-1α over the first 0–8 h. During the 8–16 h time period, the network interactions included sIL-1Rα/IL-8/IL-6/Eotaxin and IL-10/IP-10/MCP-1/MIP-1β. The overall degree of network complexity was lower in the mDay group over the first 0–8 h period, and then increased over time approaching the degree of

course of active TGF-β1. (B) Time course of latent TGF-β1. (C) Time course of total (latent + active) TGF-β1. Mean circulating levels of TGF-β1 in both the mDay (*n* = 15) and mNight (*n* = 15) sub-cohorts. The indicated inflammatory mediators were assessed in serial plasma samples obtained at the indicated time points. Values are mean ± SEM (pg/ml). None of the levels were significantly different between mDay and mNight patients.

inflammatory network complexity of mNight patients by 16–24 h (**Figure 6**).

Finally, we sought to define potential feedback structures in the dynamic networks of inflammation associated with the first 24 h post-injury during the day vs. night, and to this end, we employed DyBN inference as in previous studies (31, 32). Based on this analysis, we inferred a central motif involving dynamic interactions between cortisol and sST2 in both sub-cohorts, affecting the levels of MIG/CXCL9, IP-10/CXCL10, MCP-1/CCL2, TGF-β1 latent, TGF-β1 total, IL-22, IL-23, and IL-17E/IL-25 (**Figure 7**). Notably,

the matched (m)Day and mNight sub-cohorts (*n* = 15 each). DyNA was carried out using data on all inflammatory mediators assessed, as described in the *Materials and Methods*. This analysis suggested a higher dynamic network complexity/connectivity in the mNight as compared to the mDay group.

mNight sub-cohorts. The mNight group exhibited relatively a higher network density at 0–8 h time point which decreased over time but remained higher than the mDay group.

the only differentiating feature in these networks was the presence of IL-6 as a downstream node in mNight but not mDay patients.

#### DISCUSSION

The interactions of earth's rotation, sleep-wake cycle, the individual suprachiasmatic nucleus of the hypothalamus, and cellular core clock genes led to the evolution of daily circadian rhythms. Numerous physiological systems are under that circadian control, including inflammatory and immune responses (8, 33–35). Previous animal studies confirmed that murine responses to various pathogens as well as proinflammatory cytokines are under circadian control (36–40). Moreover, a recent study demonstrated that circadian rhythm was disrupted in trauma patients who went on to become septic (41). However, our study is the first to test directly for an association between diurnal variability and blunt traumainduced clinical outcomes and systemic inflammatory responses in human subjects. In this study, we found worse clinical outcomes and more complex dynamic networks of systemic inflammation in trauma patients injured during the night as compared to during the day.

Trauma triggers a robust inflammatory response, which is important for an effective resolution of injury (16, 24, 42). However, the dysregulated immune response can impair recovery and complicate the clinical outcome (43, 44). In parallel, key components of the immune system exhibit circadian rhythmicity, which is dependent on the rest-activity diurnal phase (45). Indeed, diurnal variation was demonstrated previously in circulating levels of human peripheral blood mononuclear cell (PBMC) subsets and serum cytokine and cytokine receptors including IL-2, IL-10, GM-CSF, IL-1β, IL-6, TNF-α, MCP-1/JE, CCR2/CD192, IFN-γ, and IFN receptors (46–48). Thus, we hypothesized that the timing of the insult relative to the circadian oscillation of the immune system would trigger differential immune/inflammatory response associated with divergent clinical outcomes.

In the present study, we focused on the potential role of time of injury in propensity-matched, moderately/severely injured patients. A key observation was the association between nighttime injury and adverse clinical outcomes, namely longer intensive care unit and hospital length of stay. While controversial factors related to the quality of medical care during nighttime (49–51) or other variables such as hospital volume and socioeconomic factors cannot be ruled out in our study, our data suggest that patients injured during the night require additional and more extensive procedures rather than experiencing worse outcomes with the same degree of intervention. Moreover, the use of stringent propensity matching further reduces the likelihood of artifactual reasons underlying our findings.

Our dynamic network analyses suggest potential mechanisms underlying the differential inflammatory responses in daytimevs. nighttime-injured patients. While neither cortisol levels nor levels of TGF-β1 and several other mediators were statistically significantly different in mDay vs. mNight, several other pathways were inferred to be activated differentially in patients injured at night vs. during the day. The lack of statistically significant differences in circulating cortisol levels in mNight vs. mDay patients suggests that trauma induces derangements in the known diurnal variation of this mediator. Indeed, cortisol

dysregulation and the role of adrenal insufficiency have been appreciated as key aspects of trauma-induced critical illness (52, 53). Importantly, cortisol is a key regulator of systemic inflammation, and we observed differences in circulating levels of multiple cytokines and chemokines. Within the panoply of inflammatory and immune pathways known to be by a traumatic injury (54, 55), there has been a recent focus on type 17 immunity (56). Notably, circulating IL-17A levels were higher in mNight patients as compared to mDay patients, suggesting the possibility that a dysregulated IL-17A response may in part underlie the worse clinical outcomes in mNight patients. Moreover, circadian rhythms are known to impact Th17 development and function (57). We have previously utilized correlation analyses (IL-17A vs. GM-CSF and IL-17A vs. IL-10 to suggest the presence of pathogenic and non-pathogenic Th17 cells) (15). In a similar analysis of the mDay vs. mNight cohorts, no significant correlations were observed (**Supplementary Figure 1**). However, there was a trend suggesting the possible prevalence of nonpathogenic Th17 cells in the mDay sub-cohort (r = 0.19, P = 0.06; **Supplementary Figure 1B**) and a similar trend suggesting the prevalence of pathogenic Th17 cells in the mNight subcohort (r = 0.19, P = 0.051; **Supplementary Figure 1C**). Further studies with larger cohorts may thus be warranted to define the role, if any, of circadian alterations on Th17 (and other IL-17A-producing) cells.

We gleaned additional information by examining the dynamic evolution of networks of systemic inflammation using two different algorithms: DyNA (23) and DyBN (22). DyNA suggested a generally higher degree of inflammatory activation in mNight patients as compared to mDay patients suggesting an early activation of immune pathways compared to the mDay group. Moreover, the inflammatory responses of mDay patients were more "Th2-like" (involving IL-5, IL-10, and IL-13 early on), whereas the networks of mNight patients involved both Th2 responses (involving IL-4, IL-5, and IL-13) as well as Th17 responses (IL-17A and also IL-6, though the latter is involved in multiple other pathways). As discussed above, we speculate that the presence of type 17 immunity in the mNight group may indicate the presence of overly exuberant inflammation. We further speculate that the presence of Th2 responses in both mDay and mNight patients may underlie the absence of significant differences in nosocomial infections in patients injured at night vs. the day, though this may be a function of the limited data set in propensity-matched patients (see below).

We recognize that there are several limitations to the current study. First, this is a single-center study and thus may not be generalizable to other centers that adopt alternative management practices or challenged by different demographics or injury characteristics. These issues warrant multi-centric studies to validate the results suggested by the current one. Moreover, the number of inflammatory biomarkers analyzed was limited to the number of analytes we could measure using commercially available LuminexTM bead sets. Further future studies examining a larger panel of inflammatory biomarkers are suggested. Finally, we note that DyNA lacks mechanistic insight; however, it can be used to understand abstract key features and interactions of the trauma-induced inflammatory response.

In conclusion, we report for the first time on a potential impact of time of injury on blunt trauma outcomes and the dynamics of systemic inflammation in humans. Our findings may have larger implications for a growing body of evidence implicating the need to consider the time of day when designing therapeutic approaches, especially in the context of diseases that are strongly impacted by immune/inflammatory interactions (12, 58–60). The present study adds to this field by demonstrating that clinical outcomes of trauma patients are likewise impacted by time of injury along with other factors such as injury severity, genotype, and the character of systemic inflammation and immune dysregulation that ensues following severe injury. Given the complexity of the intertwined inflammatory, immune, and physiologic interactions in the context of circadian rhythms, it is likely that systems and computational biology approaches (59, 61) will be necessary to help define novel therapeutic control points in the context of traumatic injury (54).

#### 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 University of Pittsburgh Institutional Review Board (IRB). The patients/participants provided their written informed consent to participate in this study.

#### REFERENCES


#### AUTHOR'S NOTE

This work was presented in part at the 37th Annual Conference on Shock.

#### AUTHOR CONTRIBUTIONS

AZ and RN participated in study design, data collection and interpretation, and writing. OA-M and KA participated in data collection. DB and JY participated in analysis of inflammatory mediators. RZ participated in computational and statistical analysis, data interpretation, and writing. MR and TB participated in data interpretation and writing. YV participated in study design, data interpretation, and writing.

#### FUNDING

This work was supported by the Office of the Assistant Secretary of Defense for Health Affairs, through the Defense Medical Research and Development Program under Award Nos. W81XWH-18-2-0051 and W81XWH-15-PRORP-OCRCA.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02699/full#supplementary-material

Supplementary Figure 1 | Correlation analyses aimed at defining Th17 cell subsets computationally. Spearman correlations were carried out on mDay vs. mNight patient cytokine data to suggest the presence of pathogenic (IL-17A vs. GM-CSF; A,C) and non-pathogenic (IL-17A vs. IL-10; B,D) Th17 cells. This analysis suggested the possible predominance of pathogenic Th17 cells in the mNight group and non-pathogenic Th17 cells in the mDay.


**Disclaimer:** Opinions, interpretations, conclusions, and recommendations are those of the authors and not necessarily endorsed by the Department of Defense.

**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 Zaaqoq, Namas, Abdul-Malak, Almahmoud, Barclay, Yin, Zamora, Rosengart, Billiar and Vodovotz. 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.

# Trauma Severity and Its Impact on Local Inflammation in Extremity Injury—Insights From a Combined Trauma Model in Pigs

Klemens Horst 1,2 \*, Johannes Greven1,2, Hannah Lüken<sup>1</sup> , Qiao Zhi <sup>2</sup> , Roman Pfeifer <sup>3</sup> , Tim P. Simon<sup>4</sup> , Borna Relja5,6, Ingo Marzi <sup>5</sup> , Hans-Christoph Pape<sup>3</sup> and Frank Hildebrand<sup>1</sup>

<sup>1</sup> Department of Orthopedic Trauma, University Hospital Aachen, Aachen, Germany, <sup>2</sup> Orthopedic Trauma Research Laboratory, University Hospital Aachen, Aachen, Germany, <sup>3</sup> Department of Orthopaedic Trauma Surgery, University Hospital Zurich, Zurich, Switzerland, <sup>4</sup> Department of Intensive Care and Intermediate Care, RWTH Aachen University, Aachen, Germany, <sup>5</sup> Department of Trauma-, Hand- and Reconstructive Surgery, University Hospital Frankfurt, Frankfurt, Germany, <sup>6</sup> Experimental Radiology, Department of Radiology and Nuclear Medicine, Otto von Guericke University Magdeburg, Magdeburg, Germany

#### Edited by:

Christoph Thiemermann, Queen Mary University of London, United Kingdom

#### Reviewed by:

Greg Gaski, Indiana University, United States Jordi Lopez Tremoleda, Queen Mary University of London, United Kingdom Martijn van Griensven, cBITE, MERLN Institute, Maastricht University, Netherlands

> \*Correspondence: Klemens Horst khorst@ukaachen.de

#### Specialty section:

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

Received: 28 November 2018 Accepted: 10 December 2019 Published: 09 January 2020

#### Citation:

Horst K, Greven J, Lüken H, Zhi Q, Pfeifer R, Simon TP, Relja B, Marzi I, Pape H-C and Hildebrand F (2020) Trauma Severity and Its Impact on Local Inflammation in Extremity Injury—Insights From a Combined Trauma Model in Pigs. Front. Immunol. 10:3028. doi: 10.3389/fimmu.2019.03028 Background: Extremity fracture is frequently seen in multiple traumatized patients. Local post-traumatic inflammatory reactions as well as local and systemic interactions have been described in previous studies. However, trauma severity and its impact on the local immunologic reaction remains unclear. Therefore, fracture-associated local inflammation was investigated in a porcine model of isolated and combined trauma to gain information about the early inflammatory stages.

Material and Methods: Polytrauma (PT) consisted of lung contusion, liver laceration, femur fracture, and controlled hemorrhage. Monotrauma (MT) consisted of femur fracture only. The fracture was operatively stabilized and animals were monitored under ICU-standard for 72 h. Blood, fracture hematoma (FH) as well as muscle samples were collected throughout the experimental period. Levels of local and systemic pro- and anti-inflammatory as well as angiogenetic cytokines were measured by ELISA.

Results: Both groups showed a significant decrease in pro-inflammatory IL-6 in FH over time. However, concentrations in MT were significantly higher than in PT. The IL-8 concentrations initially decreased in FH, but recovered by the end of the observation period. These dynamics were only statistically significant in MT. Furthermore, concentrations measured in muscle tissue showed inverse kinetics compared to those in FH. The IL-10 did not present statistical resilient dynamics over time, although a slight increase in FH was seen by the end of the observation time in the MT group.

Conclusions: Time-dependent dynamics of the local inflammatory response were observed. Trauma severity showed a significant impact, with lower values in pro- as well as angiogenetic mediators. Fracture repair could be altered by these trauma-related changes of the local immunologic milieu, which might serve as a possible explanation for the higher rates of delayed or non-union bone repair in polytraumatised patients.

Keywords: polytrauma, pigs, local inflammation, extremity, hematoma, muscle, fixation

#### INTRODUCTION

Trauma severity directly affects the pattern of injuries. Beside injuries to the head, thorax and abdomen, extremity injuries are common, and present in the majority of multiple traumatized patients (1, 2). However, fracture incidence also increased in the non-polytraumatised patient population during the past decade (3). The severity of extremity injury and its negative impacts on long-term outcome are well-documented (4–6). While pain and limited range of motion are frequently seen in isolated trauma (4, 5), previous clinical and experimental studies linked multiple trauma to significantly longer fracture healing times and higher incidences of non-unions in comparison to isolated fractures (7–9). Overwhelming local and systemic inflammatory responses with an associated negative influence on downstream processes of bone repair are a potential pathomechanism for this impaired fracture healing (10–13). Despite knowledge about the connectivity between the systemic and local inflammatory responses, information on the impact of trauma severity on systemic and local immunologic interactions and responses is scarce (14). Against this background, the purpose of this study was to investigate and compare systemic and local inflammatory responses in isolated and combined trauma. Within an established long-term porcine model of combined trauma (femur fracture, chest-, and abdominal injury, and hemorrhagic shock) (15), post-traumatic immunologic responses were analyzed and compared to those gained from a group with isolated femur fracture. Early kinetics of systemic and local (fracture hematoma and surrounding muscle tissue) immunologic response around the fracture zone were investigated during a 72 h clinically realistic study period.

#### MATERIALS AND METHODS

#### Animal Care

Official permission to perform the study was granted from the governmental animal care and use office (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Recklinghausen, Germany, AZ: 84.02.04.2014A265). All experimental protocols were approved by the governmental animal care and use office and performed in accordance with the German legislation governing animal studies, following The Principles of Laboratory Animal Care (16). The data presented in this paper were collected in the context of a larger study (15) for the benefit of the principles of the 3Rs (Replacement, Refinement, and Reduction) (17).

In total 24 male pigs (German Landrace, Sus scrofa) weighing 30 ± 5 kg, aged 3 months were used. After arrival from a diseasefree barrier breeding facility all animals underwent clinical examination by a veterinarian. Thereafter all animals were housed for 7 days before experiments started. Polytrauma (PT) was induced to 12 animals while 12 animals received isolated femur fracture and were defined as monotrauma (MT). Animals were housed in ventilated rooms and allowed to acclimatize to their surroundings for a minimum of 7 days before start of the experiment. All sections of this report adhere to the ARRIVE Guidelines for reporting animal research (18).

#### Sample Size and Power Calculation

A sample size calculation was performed for the primary study (15). The chosen sample sizes of 12 in the two groups (MT, PT) show comparable effect sizes as observed in a previous published study on hypothermia in a porcine trauma model (19) and will provide at least 80% power at a significance level of 5%. As all physiological, morphological and inflammatory outcomes characterizing the long-term evolution of severe multiple trauma are equally important to describe the intermodal animal model, no distinction between primary, and secondary outcome was made.

#### General Instrumentation and Anesthesia

The model was previously described in detail elsewhere (15). In brief: premedication was induced by an intramuscular injection of azaperone (4 mg kg−<sup>1</sup> ). During the 12-h fasting period animals had free access to water. Anesthesia was induced by propofol (3 mg kg−<sup>1</sup> ), followed by orotracheal intubation. Volume-controlled, lung protective mechanical ventilation was applied, and vital parameters were continuously monitored and documented as previously described (20).

During the entire study period, general anesthesia was maintained with propofol and sufentanil (40–90 µg Kg−<sup>1</sup> /h). Continuous crystalloid infusion (Sterofundin ISO <sup>R</sup> ; 2 ml/kg BW/h) preserved animals from dehydration (15).

Administration of fluids and anesthesia was done by a central venous catheter which was placed in the external jugular vein. Furthermore, this was used to monitor the central venous pressure. The right femoral vein was instrumented via a three-lumen haemodialysis catheter to perform hemorrhage. Continuous monitoring of blood pressure, e.g., mean arterial pressure (MAP) was performed via an arterial line, that was placed in the femoral artery. Reference for intravascular pressure measurements was the mid-chest level and at end of expiration. Suprapubic catheter was applied. Finally, random allocation to either the PT group or the MT group was performed (15).

#### Induction of Multiple Trauma and Hemorrhage

Trauma was induced as previously described (15) and after achieving stable baseline conditions (at least 120 min after instrumentation). During the 90-min period of shock animals were not prevented from hypothermia to simulate the clinical situation (in humans) after trauma and transport to the hospital.

A bolt gun machine (Blitz-Kerner, turbocut JOBB GmbH, Germany) was used to induce femur fracture in mono- as well as multiple trauma. Therefore the bolt hit a custom-made punch positioned on the mid third of the femur. Cattle-killing cartridges (9 × 17; DynamitNobel AG, Troisdorf, Germany) were used. The PT group received blunt chest trauma, induced by a bolt shot fired on a pair of panels that was placed on the right dorsal, lower chest (20, 21). Lungs were inflated when the bolt shot was applied. Moreover, a midline-laparotomy was performed and the right upper liver lobe was explored in PT. A penetrating hepatic injury was induced by a crosswise incision (4.5 x 4.5 cm) halfway through the liver tissue (22, 23). Liver packing was carried out with five sterile packs of 10 × 10-cm gauze after a short period of uncontrolled bleeding (30 s). Thereafter, hemorrhagic shock was induced by withdrawal of blood until a MAP of 40 ± 5 mm Hg was reached, with a maximum withdrawal of 45% of the total blood volume. The MAP was maintained for 90 min. The ISS (Injury Severity Score) was calculated as 27 points in PT. One investigator (KH) induced trauma, and the period of shock was monitored by two experienced clinicians (KH, TPS) (15).

Animals were resuscitated at the end of the shock period in accordance with established trauma guidelines (ATLS <sup>R</sup> , AWMF-S3 guideline on Treatment of Patients with Severe and Multiple Injuries <sup>R</sup> ) by adjusting FiO<sup>2</sup> to baseline values and re-infusing the withdrawn blood and additional fluids (Sterofundin ISO <sup>R</sup> ; 2 ml kg/BW/h) in PT (24). Furthermore, animals were rewarmed until normothermia (38.7–39.8◦C) wasreached using a forced-air warming system (24).

According to established trauma guidelines, operative stabilization of the femur fracture was performed after surgical disinfection and sterile draping at the end of the resuscitation (25). Fluoroscopy (Ziehm Vision, ZiehmImaging, Germany) was used to guide reduction and operation of the femur fracture. According to the clinical situation were internal as well as external stabilization is used for fracture treatment, an intramedullary nail (T2 System, Stryker) was applied to six animals in each group while the remaining six animals received external fixation (Radiolucent Fixator, Orthofix). Surgery was performed by one experienced trauma surgeon. Before surgery and then every 24 h until the end of the experiment antibiotics (Ceftriaxon <sup>R</sup> 2 g, i.v.) were administered (15).

#### Data Collection

Following parameters were measured every 30 min by blood gas analysis (BGA) for a period of 5.5 h after trauma: pH, lactate (LAC), pCO2, pO2, hemoglobin (Hb), and base excess (BE). From then on, BGA was performed every 6 h until observation period came to an end. Time points of whole blood sampling are paralleled by data on physiologic responses (MAP and heart rate; HR) as well as BGA results. Results demonstrating severe signs of shock were published earlier (15).

Blood samples used in this study were obtained after resuscitation and operative treatment (3.5 h) and after 24, 48, and 72 h (15). Samples were kept on ice. Subsequently, after centrifugation at 2,000 × g for 15 min at 4◦C, serum samples were stored at −80◦C until analysis of IL-6, IL-8, and IL-10 concentrations (Quantikine <sup>R</sup> porcine ELISA kit; R&D systems, USA), according to the manufacturer's instructions. Muscle tissue was obtained by biopsy after resuscitation and operative treatment (3.5 h) and after 24, 48, and 72 h; samples were frozen in liquid nitrogen. For protein analysis, 100 mg of frozen muscle tissue were thawed in 300 µL of lysis and extraction buffer and immediately homogenized in an Eppendorf tube on ice with a T10 basic ULTRA-TURRAX <sup>R</sup> (IKA, Germany). Protein concentrations were measured with commercially available ELISA kits. Fracture hematoma was extracted under sterile conditions by puncturing the fracture zone at 3,5, 24, 48, and 72 h. Hematoma was collected in an EDTA monovette <sup>R</sup> (SARSTEDT AG & Co, Germany) and diluted with Sterofundin <sup>R</sup> 1:1. After centrifugation, serum was removed and stored at −80◦C for further analysis. Referring to higher concentrations, all fracture hematoma samples were diluted once more (IL-6 1:10, IL-8: 1:4, IL-10: 1:4).

#### Statistical Analysis

Statistics were performed with SPSS (Version 21.0.0.0) using Mann-Whitney-U, Wilcoxon rank sum and Friedman tests (including Chi<sup>2</sup> -Test) [illustrated as mean (SEM)]. For all comparisons, the significance level was set at 5%. Graphics were created using SPSS.

#### RESULTS

#### Physiological Response

In contrast to previously reported data from the PT group (15), MT did not present with comparable shock parameters. The mean arterial pressure (MAP) was significantly higher (p < 0.001) in MT (69 ± 2.3 mmHg) than in PT (43 ± 1.9 mmHg) 90 min after trauma induction. Additionally, heart rate was significantly lower (p < 0.001) in MT (85 ± 7b/min) compared to PT (170 ± 11b/min) at this time. Furthermore pH (MT 7.51 ± 0.01 vs. PT 7.43 ± 0.01, p < 0.001), Lactate (MT 1.2 ± 0.2 mmol vs. PT 4.4 ± 0.4 mmol, p < 0.001), Base Excess (MT 4.8 ± 0.5 mmol vs. PT 0.4 ± 0.6 mmol, p < 0.001) did prove severe haemorrhagic shock only in the PT group. In regard to the reported time points during further clinical course only Lactate was slightly increased after 3.5 h (MT 1.01 ± 0.12 mmol vs. PT 1.38 ± 0.11 mmol). Otherwise there were no statistically significant differences found between the groups. Due to interrupted warming during the trauma phase body temperature was 36.7 ± 0.3◦C in MT and 36.7 ± 0.2◦C in PT after 90 min. These values were not statistically significant (p = 0.887). After rewarming, animals presented with physiological body temperature (MT 3.5 h: 38.2 ± 0.2◦C; D1: 38.7 ± 0.1◦C; D2: 38.7 ± 0.3◦C and D3: 38.8 ± 0.1◦C, p < 0.001 resp. PT 3.5 h: 37.9 ± 0.1◦C; D1: 38.7 ± 0.1◦C; D2: 38.8 ± 0.1◦C and D3: 38.9 ± 0.1◦C, p < 0.001). Although temperature changed statistically significant during the clinical course in both groups, there were no differences between the groups.

#### Interleukin-6

According to the post-traumatic phase, a decrease in serum concentrations was observed in PT, while concentrations in MT remained stable on a low level (**Table 1**). In both groups, a statistically significant decrease in local IL-6 concentrations in muscle tissue as well as fracture haematoma were observed over time (**Table 1** and **Figure 1**). Although fracture haematoma concentrations were higher compared to serum concentrations in both groups, local concentrations of IL-6 in fracture haematoma were significantly lower in PT than in MT (**Table 1**). In contrast to PT, haematoma concentrations in MT showed statistically significantly higher levels compared to muscle tissue concentrations (**Table 1**).

#### Interleukin-8

Serum IL-8 showed a slight increase over time in PT, while in MT, there was a decrease in systemic concentrations. However, TABLE 1 | Systemic and local concentrations of IL-6 pg/ml; <sup>a</sup>p < 0.05 compared to serum concentrations, <sup>b</sup>p < 0.05 compared to muscle concentrations, <sup>c</sup>p < 0.05 compared to PT.


Values are given in mean (SEM), <sup>F</sup> Friedman-Test and Chi<sup>2</sup> -Test, <sup>W</sup> Wilcoxon-Test.

this finding in MT was statistically not significant (**Table 2**). Interestingly, concentrations in muscle tissue showed opposite trends compared to the systemic ones. Initially increasing concentrations decreased by the end of the observation time (**Table 2**). In contrast, concentrations measured in fracture haematoma presented with inverse dynamics compared to those seen in muscle tissue. IL-8 dynamics in haematoma described a v-shaped curve, which was statistically significant in MT (**Table 2** and **Figure 2**). At all measured time points, IL-8 concentrations in fracture haematoma of MT were higher compared to those in PT (**Table 2**).

#### Interleukin-10

While serum concentrations of IL-10 in PT slightly decreased over time (**Table 3**), values in MT did not present statistically significant changes over time. Although local concentrations measured in muscle tissue and fracture hematoma remained uneventful and were detectable only on a very low level, haematoma concentrations in MT showed an increase by the end of the observation period (**Table 3** and **Figure 3**). However, this finding was not statistically significant.

# DISCUSSION

Fracture healing is significantly influenced by the local inflammatory response after trauma (26–29). The impact of trauma severity may lead to a different post-traumatic response, which potentially influences the onset of fracture healing (30, 31). However, information about local inflammatory reactions regarding fracture repair are mostly gained from small animal models with either limited observation time or conditions that do not closely mimic a clinically realistic situation (32–36). As pigs respond to trauma similar to humans, we used an established long-term porcine model of isolated and multiple TABLE 2 | Systemic and local concentrations of IL-8 pg/ml; <sup>a</sup>p < 0.05 compared to serum concentrations, <sup>b</sup>p < 0.05 compared to muscle concentrations, <sup>c</sup>p < 0.05 compared to PT.


Values are given in mean (SEM), <sup>F</sup> Friedman-Test and Chi<sup>2</sup> -Test, <sup>W</sup> Wilcoxon-Test.

trauma to investigate the local and systemic inflammatory responses in regard to extremity injury and trauma impact (37, 38).

The main results might be summarized as follows:


an inflammatory milieu. However, this finding was not statistically significant.

#### The Pro-inflammatory Phase

The early post-traumatic immunologic milieu of fracture hematoma is characterized by inflammation and hypoxia (28). During this early period of acute inflammation, proinflammatory mediators such as IL-6 recruit cells needed for tissue regeneration (39). As previously reported and confirmed by others, IL-6 in fracture haematoma increases during the initial post-traumatic phase, followed by a continuous decrease during the further clinical course (20, 40, 41). While its early peak is discussed to maintain the onset of bone healing, persistent high values negatively influence osteogenic differentiation from stem cells (42–44). In regard to multiple trauma commonly associated with an advanced post-traumatic inflammatory response (45), Recknagel et al. revealed that concomitant thoracic trauma considerably enhanced the number of PMN, decreased the number of macrophages and slightly increased IL-6 expression TABLE 3 | Systemic and local concentrations of IL-10 pg/ml; <sup>a</sup>p < 0.05 compared to serum concentrations, <sup>b</sup>p < 0.05 compared to muscle concentrations, <sup>c</sup>p < 0.05 compared to PT.


Values are given in mean (SEM), <sup>F</sup> Friedman-Test and Chi<sup>2</sup> -Test, <sup>W</sup> Wilcoxon-Test.

locally at the fracture site, suggesting that post-traumatic systemic inflammation altered the finely tuned inflammatory balance during the early healing phase, leading to impaired bone healing (31, 46). Accordingly, De Benedetti et al. showed that overexpression of IL-6 resulted in severe osteopenia with reduced osteoblast and increased osteoclast numbers and activity (47). Thus, the observed time-dependent decrease in IL-6 concentrations seems to be a consistent step in the sequence of fracture repair. Heiner et al. suggested IL-6-induced upregulation of the suppressor of cytokine signaling-3 (SOCS-3) as a possible mechanism for the reduction of local IL-6 concentrations (40).

Although we found this decrease in both groups, significant differences between concentrations of IL-6 in MT and PT were observed, with higher values in MT. This dichotomy is interesting as excessive trauma is known to increase systemic cytokine concentrations (14). However, in contrast to lung contusion or haemorrhagic shock, fracture associated soft tissue trauma was found not to be the driving force leading to significant increase of cytokine concentrations (36, 48, 49). Moreover, haemorrhagic shock was discussed to reduce supply in the fracture zone (50), which might explain the observed lower cytokine concentrations in PT fracture hematoma compared to MT ones. Altered immunologic reactions after bone injury with hemorrhage compared to isolated bone injury were previously described and support our findings (51–53). An altered immunologic milieu in the early fracture hematoma also support the findings reported by Lichte et al. who demonstrated impaired bone healing and a significantly decreased number of osteoclasts, a decrease in bone quality and more cartilage islands after hemorrhagic shock in mice (53). Additionally, Wichmann et al. reported on a murine model comparing isolated tibia fracture with tibia fracture and combined hemorrhagic shock (51). The authors concluded that severe hemorrhage after closed bone fracture depresses osteoblast activity and increases osteocyte necrosis, which should compromise fracture healing under those conditions (51). In line with others, the authors discussed decreased blood supply to the fracture zone to negatively influence fracture repair (51–53). Thus, our observation of lower pro-inflammatory concentrations in the PT group suggests the absence of important pro-inflammatory pacemakers in the very early phase of fracture repair, leading to a delay in skeletogenic mesenchymal stem cell differentiation, with consecutive non- or delayed bone healing (54–56). This finding could serve as one possible explanation why polytraumatised patients suffer from bony non-union more often than patients with isolated injury (9, 30, 57). Against this background, the value of traumatic hemorrhage and its influence on the local immunologic milieu in fracture healing must not be underestimated. Comparable to the benefit of typical shock organs (58), it seems likely, that early resuscitation would also improve the perfusion at the site of the fracture zone supporting recovery to a physiological and immunological state (59, 60). In this context, Augat et al. found that a transient hemorrhagic shock situation followed by isovolumetric blood volume resuscitation resulted in improved fracture healing. The authors concluded that the positive healing response might be associated with improved revascularization of the soft callus adjacent to the fracture site (59, 60). Accordingly, Melnyk et al. described that soft tissue damage without destruction of the bone-soft tissue interface is likely to have only a limited effect on fracture healing (61).

Beside its pro-inflammatory properties, IL-8 is well-known for its angiogenetic characteristics. Accordingly, high local levels were found in fracture haematoma in a previous clinical study, which underlines the importance of IL-8 in the process of bone healing (28). Our data revealed that IL-8 kinetics in muscle tissue showed opposite trends to those seen in fracture haematoma. While concentrations in muscle tissue increased initially and decreased during the clinical course, haematoma concentrations initially decreased and recovered during the later clinical course. Comparable to the observations made in IL-6, the dynamics of IL-8 in fracture haematoma may be explained by reduced blood flow due to haemorrhagic shock in PT (62). Accordingly, Heppenstall et al. report on a rabbit model with an inhibition of fracture healing in hypovolaemia, which was attributed to impaired delivery of oxygen to the fracture site (63). According to our divergent findings in muscle tissue and fracture haematoma, Schmidt-Bleek et al. reported differences in the immunologic milieus of muscle haematoma and fracture haematoma in a sheep model (64). The authors indicate that the inflammatory processes differ due to a unique immune cell composition (64). Although the authors report on a different animal model of isolated trauma, investigating cell migration, our data also reveal differences in the immunologic post-traumatic milieus of muscle tissue and fracture haematoma within the MT, but not the PT group. In regard to the concentrations measured in muscle tissue, our findings are supported by Dragu et al. who proved alterations in the gene expression level in human muscle free flaps after ischaemia and reperfusion (65). The authors report on IL8 as one of four genes that were significantly upregulated after reperfusion of ischemic muscle tissue (65). Accordingly, Huda et al. showed that a significantly elevated concentration was measurable in blood plasma after 3–4 h of reperfusion (66). Furthermore, Kukielka et al. investigated IL-8 expression after ischemia and reperfusion in canine myocardium. The authors found that IL-8 mRNA peaked in the first 3 h of reperfusion and persisted at high levels beyond 24 h (67). Based on these findings, Kukielka et al. speculated that surface-bound chemoattractants may represent an effective mechanism of chemotactic agent presentation and neutrophil activation wherever a reduced blood flow prevents the establishment of a stable soluble chemotactic gradient (67). The observation of increased IL-8 levels are in line with our results from muscle tissue analysis. Both groups showed increased IL-8 concentrations 24 and 48 h after trauma. While values in MT were doubled, concentrations in PT increased even four times compared to initial values. As PT received haemorrhagic shock, this finding might support Kukielka's speculation on the effect of surface-bound chemoattractants in tissue with reduced blood flow. Thus, cell composition as well as interaction of immunologic key players in the early local inflammatory response after multiple trauma must be the focus of further studies.

#### The Anti-inflammatory Phase

Interleukin-10 is known as an anti-inflammatory mediator that also plays a central role in the fracture healing process (28). It influences bone resorption and enhanced bone healing (35, 68, 69), and a deficit results in osteopenia, mechanical fragility of bones, and defects in their formation (70). While some authors report increased IL-10 concentrations in fracture haematoma during the early post-traumatic phase (71, 72), we could not prove significant kinetics over time. Baker et al. compared different trauma models and proofed that polytrauma plus hemorrhage did not induce the systemic release of IL-10 (49). The authors showed that an additional hemorrhage component appears to attenuate the systemic release of IL-10 after polytrauma (49). In line with Baker et al. and Wichmann et al. proofed that a bone injury, coupled with haemorrhagic shock, produces a more severe depression of immune functions than a haemorrhagic shock alone (73). The authors concluded that bone injury appears to play a contributory role in further depressing immune functions in trauma patients who experienced major blood loss (73). These observations may further reflect that the combined insult leads to the induction of a state of immune paralysis, which also affects IL-10 concentrations within the fracture haematoma (49, 74). In contrast, Hauser et al. found significantly increased IL-10 levels in fracture haematomas in the early phase after trauma, whereas lower levels were observed in the later period (>24 h) (71). However, the authors reported on isolated injures with very heterogeneous entity and severity, which might not realistically reflect the polytraumatised situation. Additionally, Hoff et al. also reported elevated IL-10 concentrations in fracture haematoma (72). However, these values were compared to IL-10 concentrations gained from non-traumatic osteotomy in hip replacement. Thus, the expressiveness of this early "increase" might be also questioned against the background of traumatic injuries. Delayed migration of IL-10-producing cells into fracture haematoma, proved by Schmidt-Bleek et al. may be another cause for time-dependent kinetics in local IL-10 concentrations (44). This might allow a careful speculation about the observed IL-10 increase in fracture haematoma of the MT group, but not the

PT group after 72 h, representing a possible shift from a proinflammatory immunologic milieu toward an anti-inflammatory and angiogenic one (28, 44, 75). Yet, literature about local concentrations of IL-10 remains sparse, and further research is warranted. However, the absence of IL-10 in the haematoma of polytraumatised patients might be another explanation for impaired bone regeneration in this patient cohort.

#### Limitations

The purpose of our study was to gain knowledge about trauma impact and its effect on local inflammatory response around the fracture zone in a clinically relevant, large animal model of isolated vs. multiple trauma. Unfortunately, molecular mechanisms that regulate local or systemic inflammatory response could not be derived. Also, interaction of local inflammatory response to osteo- and chondrogenesis remain unlighted. Furthermore, testing of a relatively small sample size yielded relatively large standard errors for each parameter. Additionally, it would have been interesting to analyse individual immunologic responses as well as financial restrictions led to measurement of only three mediators which is regrettable in the context of a vast immunologic system whereby dozens of inflammatory mediators dynamically interact resulting in a plethora of possible phenotypes. However, research regarding this field is ongoing, and follow-up studies that concentrate on cell migration, but also on bone healing, are in preparation.

#### CONCLUSIONS

To the best of our knowledge, this is the first study that characterizes and compares chronologic data of locally active inflammatory mediators in regard to femur fracture and trauma impact. Although inference of systemically circulating mediators cannot be drawn from this study, it might be suggested that concomitant injuries, such as haemorrhagic shock, significantly influence local post-traumatic reactions in fracture/soft-tissue haematomas. Combined trauma (or "severe trauma") may cause perturbations in local and/or systemic cytokines and chemokine levels intimately involved in the early phases fracture healing, which may influence adverse outcomes such as fracture nonunion. Based on the results of this study, further studies of our group will focus on the role of inflammatory mediators in the repairing process of injured tissue and their role in the systemic process of responding to trauma.

#### REFERENCES


#### ADVANCES IN KNOWLEDGE

To our knowledge, this is the first study comparing and discussing local (fracture haematoma and muscle tissue) inflammatory response in a large animal model of isolated (MT) and combined (PT) trauma, giving chronological data of locally active inflammatory mediators in regard to extremity fracture during early post-traumatic stages up to 72 h.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the German legislation governing animal studies, following The Principles of Laboratory Animal Care (16). Official permission was granted from the governmental animal care and use office (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Recklinghausen, Germany, AZ: 84.02.04.2014A265). The protocol was approved by the governmental animal care and use office (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Recklinghausen, Germany, AZ: 84.02.04.2014A265).

#### AUTHOR CONTRIBUTIONS

KH and FH conceived the study, set up its design and coordinated the experimental and analytic phase. KH, JG, HL, QZ, and TS carried out the experiments and performed the analysis. RP participated in its design and coordination. BR, IM, and H-CP conceived the study, participated in its design and coordination, and helped to draft the manuscript. All authors listed have made a substantial, direct and intellectual contribution to the work, read and approved the final manuscript.

#### FUNDING

Project no. S-14–14P was supported by the AO Foundation.

#### ACKNOWLEDGMENTS

The authors thank Yannik Kalbas, Rafael Serve, Lukas Schimunek, Birte Weber, Lukas Egerer, Felix Hönes, and Simona Klee for their assistance in conducting the experiments. We also thank Thaddeus Stopinski for his considerable support during the entire study. Finally, we would like to thank the RWTH Aachen, Faculty of Medicine, for supporting the project via the scientific rotation program for young researchers.


of the femur. Can J Surg.(2008) 51:366–70. Available online at: http://canjsurg. ca/wp-content/uploads/2014/03/51-5-366.pdf


**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 Horst, Greven, Lüken, Zhi, Pfeifer, Simon, Relja, Marzi, Pape and Hildebrand. 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.

# Citrullinated Histone H3 as a Therapeutic Target for Endotoxic Shock in Mice

Qiufang Deng1,2†, Baihong Pan1,2†, Hasan B. Alam<sup>2</sup> \*, Yingjian Liang2,3, Zhenyu Wu1,2 , Baoling Liu<sup>2</sup> , Nirit Mor-Vaknin<sup>4</sup> , Xiuzhen Duan<sup>5</sup> , Aaron M. Williams <sup>2</sup> , Yuzi Tian1,2 , Justin Zhang<sup>2</sup> and Yongqing Li <sup>2</sup> \*

*<sup>1</sup> Xiangya Hospital, Central South University, Changsha, China, <sup>2</sup> Department of Surgery, University of Michigan, Ann Arbor, MI, United States, <sup>3</sup> The First Hospital, China Medical University, Shenyang, China, <sup>4</sup> Division of Infectious Diseases, Department of Internal Medicine, University of Michigan, Ann Arbor, MI, United States, <sup>5</sup> Department of Pathology, Loyola University Medical Center, Maywood, IL, United States*

#### Edited by:

*Patrizia Rovere Querini, Vita-Salute San Raffaele University, Italy*

#### Reviewed by:

*Kimberly Martinod, KU Leuven, Belgium Claudia Farias Benjamim, Federal University of Rio de Janeiro, Brazil*

#### \*Correspondence:

*Hasan B. Alam alamh@med.umich.edu Yongqing Li yqli@med.umich.edu*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *06 June 2019* Accepted: *02 December 2019* Published: *09 January 2020*

#### Citation:

*Deng Q, Pan B, Alam HB, Liang Y, Wu Z, Liu B, Mor-Vaknin N, Duan X, Williams AM, Tian Y, Zhang J and Li Y (2020) Citrullinated Histone H3 as a Therapeutic Target for Endotoxic Shock in Mice. Front. Immunol. 10:2957. doi: 10.3389/fimmu.2019.02957* Sepsis results in millions of deaths every year, with acute lung injury (ALI) being one of the leading causes of mortality in septic patients. As neutrophil extracellular traps (NETs) are abundant in sepsis, neutralizing components of NETs may be a useful strategy to improve outcomes of sepsis. Citrullinated histone H3 (CitH3) has been recently shown to be involved in the NET formation. In this study, we demonstrate that CitH3 damages human umbilical vein endothelial cells (HUVECs) and potentiates NET formation through a positive feedback mechanism. We developed a novel CitH3 monoclonal antibody to target peptidylarginine deiminase (PAD) 2 and PAD 4 generated CitH3. In a mouse model of lethal lipopolysaccharide (LPS) induced shock, neutralizing CitH3 with the newly developed anti-CitH3 monoclonal antibody attenuates inflammatory responses, ameliorates ALI, and improves survival. Our study suggests that effectively blocking circulating CitH3 might be a potential therapeutic method for the treatment of endotoxemia.

Keywords: citrullinated histone H3, endotoxic shock, neutrophil extracellular traps, new anti-CitH3 antibody, inflammation, acute lung injury, survival

# INTRODUCTION

Sepsis is defined as a life-threatening organ dysfunction caused by the dysregulation of host response secondary to infection (1). Acute lung injury (ALI) develops in nearly 40% of the septic patients and is one of the leading causes of death (2).

Excessive neutrophil extracellular traps (NETs) have been detected in sepsis and are associated with significant organ injury (3). NETs are stranded, decondensed DNA (deoxyribonucleic acid) fibers accompanied by intracellular proteins, including histones, neutrophil elastase (NE), myeloperoxidase (MPO), and other proteins coming from various neutrophil organelles (4, 5). Histone citrullination/deimination induced by peptidyl arginine deiminases (PADs) is an important posttranslational modification that facilitates chromatin decondensation during NET formation (6). Moreover, citrullinated histones are found in the extracellular space of neutrophils along with DNA as components of NETs (7).

NETs are generally regarded as a double-edged sword (8). NETs can immobilize and kill a broad range of pathogens: gram-positive and -negative bacteria, fungi, viruses, and protozoa (4, 9–14). However, they also promote tissue damage, increase thrombosis, and cause disruption of the autoimmune system (15). Strategies targeting NET formation or NET components have proven therapeutic in animal models of sepsis. PAD inhibitors, which can disrupt NET formation, have been shown to protect animals from endotoxic shock or septic shock. Administration of a pan-PAD Inhibitor, Cl-amidine, improves survival in both lethal and sub-lethal models of murine sepsis, increases bacterial clearance, and ameliorates thymus and bone marrow atrophy (16, 17). YW3-56, another pan-PAD inhibitor, has been shown to increase survival in mice with lipopolysaccharide (LPS) induced endotoxic shock (18). Administration of DNase has been demonstrated to markedly reduce cell-free DNA and improve outcomes in both Escherichia coli-induced and CLP-induced sepsis (19, 20). Neutralization of histone H4, a NET component, has been shown to significantly reduce the mortality in mouse models of cecal ligation and puncture (CLP) (21).

Citrullinated histone H3 (CitH3) has been shown to be highly involved in the process of NETosis (4, 6, 18, 22). As such, CitH3 is considered a good biomarker for the diagnosis of endotoxic shock due to its early appearance (as early as 30 min) in the blood, long half-life, high specificity, and response to treatment (23). However, the physio-pathologic role of CitH3 in sepsis has not been well-defined. Furthermore, it is unknown whether CitH3 could be considered a therapeutic target. In this study, we investigated the adverse effects of CitH3 and developed a new anti- histone H3 (citrullinated R2+R8+R17+R26) monoclonal antibody [CitH3 mAb (4 Cit)] utilizing CitH3 with four citrullines (4 Cit) as the antigen. We then evaluated this novel antibody in a murine model of endotoxic shock to explore the therapeutic value of neutralizing the circulating CitH3 protein.

#### MATERIALS AND METHODS

#### Generation of CitH3 mAb (4 Cit)

CitH3 peptide with four citrulline residues [A(Cit)TKQTA(Cit) KSTGGKAP(Cit) KQLATKAA(Cit)KSAP], referred to as CitH3 (R2+R8+R17+R26) peptide, was chemically synthesized by New England Peptide, Inc. (Gardner, MA, USA) and utilized to generate the 4 Cit monoclonal antibody in ProMab Biotechnologies, Inc. (Richmond, CA, USA) using the companyapproved animal protocol. Balb/c mice were immunized with the CitH3 (R2+R8+R17+R26) peptide. Antibody titers were then determined by enzyme-linked immunosorbent assay (ELISA). Splenocytes from the mouse with the highest antibody titer were fused with myeloma cells (SP2/0) to generate hybridomas. Anti-CitH3 (R2+R8+R17+R26) peptide-specific hybridomas were identified using ELISA against the CitH3 (R2+R8+R17+R26) peptide. One to two million viable hybridoma cells were injected into a mouse peritoneal cavity to produce ascites, which was harvested, and particles were removed through centrifugation. Protein G purification was performed to get the final CitH3 mAb (4 Cit).

#### Cell Culture and Treatment

Human umbilical vein endothelial cells (HUVECs) (Lonza, Walkersville, MD, USA) were cultured using endothelial cell growth medium (EGM) BulletKit (Lonza, Walkersville, MD, USA). HUVECs (5 × 10<sup>5</sup> cells/ml) were grown on 12 mm Transwells with 0.4µm pore polyester membrane inserts (Corning Life Sciences, Corning, NY, USA) for 3 days to develop a confluent (90%) monolayer. HUVECs were then treated for 16 h with 5µg/ml histone H3 peptide (ARTKQTARKSTGGKAP RKQLATKAARKSAP) or CitH3 peptide [A(Cit)TKQTA(Cit) KSTGGKAP(Cit)KQLATKAA(Cit)KSAP]. Chambers were then incubated in the presence of 1 mg/ml 10-kDa FITC-dextran (Thermo Scientific, Rockford, IL, USA). The fluorescence of media in the lower chambers was measured by a GloMax-multi detection system (Promega, Madison, WI, USA).

#### Mice

Male C57BL/6 mice (7–8 weeks) were purchased from The Jackson Laboratory (Bar Harbor, ME, USA) and housed for at least 3 days with food and water ad libitum before the experiment. All experiments were performed in compliance with the animal welfare and research regulations. The animal protocol for this study was approved by the University of Michigan Institutional Animal Care and Use Committee.

#### Lethal Endotoxic Shock and Antibody Treatment

LPS was injected intraperitoneally (20 mg/kg), inducing lethal endotoxic shock in the mice. Either CitH3 mAb (4 Cit) (about 20 mg/kg) or the same amount of anti-histone H3 [(citrullinated R2+R8+R17) monoclonal antibody (CitH3 mAb (3 Cit), Item number 9003062 with Batch numbers 0515031-1, 0513766-1, and 0516044-1; Cayman Chemical, Ann Arbor, MI, USA)] was administered via tail vein injection. Mouse receiving immunoglobulin G (IgG) only (20 mg/kg) or LPS followed by IgG served as controls (n = 9/group). Survival was monitored for 10 days. Kaplan–Meier curves were used to compare the survival rates.

In another cohort, mice were also randomly divided into four groups: (1) IgG only (20 mg/kg), (2) LPS (20 mg/kg) + IgG (20 mg/kg), (3) LPS + CitH3 mAb (4 Cit) (20 mg/kg), and (4) LPS + CitH3 mAb (3 Cit) (20 mg/kg) (n = 3/group). Animals were sacrificed 12 h after treatment (n = 3), and organs were collected and stored in −80◦C for further use. Blood samples were at room temperature (RT) for 1 h to allow for clotting and separation of serum. Serum was collected by centrifugation of the clotted blood at 3,000 × g at 4◦C for 20 min, and then stored immediately at −80◦C.

#### Western Blotting for Antibody Validation

One-half microgram of five different peptides [H3, AceH3, CitH3 (R2+R8+R17+R26), CitH3 (R26), and MetH3] or 3 ng of CitH3 protein was subjected to SDS-polyacrylamide gel electrophoresis and was transferred onto a nitrocellulose membrane. Membranes were then probed with the same concentration (2µg/ml) of CitH3 mAb (4 Cit) or CitH3 mAb (3 Cit). Donkey antimouse 800 CW antibodies (LI-COR, Lincoln, NE, USA) were

**Abbreviations:** CitH3 mAb (4 Cit), anti-histone H3 (citrullinated R2+R8+R17+R26) monoclonal antibody; CitH3 mAb (3 Cit), anti-histone H3 (citrullinated R2+R8+R17) monoclonal antibody.

Deng et al. CitH3 Antibody Treatment in Endotoxic Shock

used as the secondary detection antibodies (1: 5,000 dilution). Finally, the membranes were exposed to 800 channel Odyssey Imaging System (LI-COR, Lincoln, NE, USA). Immunoblot signal intensity was analyzed using Image Studio Lite (LI-COR, Lincoln, NE, USA).

#### CitH3 ELISA

A "sandwich" ELISA, which has been developed by our laboratory and described previously (23), was used. In brief, 0.5 µg/well CitH3 mAb (4 Cit) or CitH3 mAb (3 Cit) was coated in 96 well plates (Corning Life Sciences, Corning, NY, USA) at 4◦C overnight and then blocked with 100 µl of protein-free blocking buffer (Thermo Scientific, Rockford, IL, USA) at 4◦C overnight. The wells were then incubated with CitH3 (R2+R8+R17+R26) peptide or mouse serum (1:1 diluted in blocking buffer) at RT for 2 h, followed by rabbit anti-CitH3 polyclonal antibody (1:3,000 diluted, Abcam, Cambridge, MA, USA) incubation for 2 h at RT. Next, 96-well plates were probed with donkey antirabbit horseradish peroxidase (HRP) conjugate IgG (1:50,000 diluted, Jackson ImmunoResearch, West Grove, PA, USA). 3,3′ ,5,5′ -Tetramethylbenzidine (TMB, Thermo Fisher Scientific, Waltham, MA, USA) was utilized to develop the plate for 30 min at RT in the dark before adding stop solution (R&D Systems Inc., Minneapolis, MN, USA). Absorbance was measured at 450 nm.

# Cytokines

Levels of pro-inflammatory cytokines in the serum or lung homogenates were measured by ELISA. IL-1β was measured using the Mouse IL-1β/IL-1F2 DuoSet ELISA (R&D Systems, Minneapolis, MN, USA) and TNF-α was detected using Mouse TNF-α DuoSet ELISA (R&D Systems, Minneapolis, MN, USA). The ELISA was performed blindly by an independent researcher.

# Histopathology

Twelve hours after treatment, lung samples were collected and fixed with 4% paraformaldehyde, and then dehydrated in 70% ethanol. The lung tissues were embedded in paraffin and cut into 5-µm sections. Hematoxylin–eosin staining was performed by a blinded researcher. Histological analysis of ALI was also graded by a blinded pathologist with a scale from 0 to 3 among the following domains: (1) septal mononuclear cell/lymphocyte infiltration, (2) septal hemorrhage and congestion, (3) neutrophils, (4) alveolar macrophages, (5) alveolar hemorrhage, and (6) alveolar edema (0: "absent," 1: "mild," 2: "moderate," and 3: "severe"). The total injury score was calculated by adding up the scores for all parameters.

#### Human Neutrophil Isolation and Treatment

Collection of blood samples from a healthy human volunteer was approved by the Institutional Review Board (IRB) of the University of Michigan (HUM00048623). Neutrophil isolation has been described previously (24). Briefly, human whole blood up to 60 ml was placed in 7 ml 0.25 M citrate solution and 10 ml 6% dextran solution in PBS. After incubation for 30 min, the upper phase was collected and layered on 15 ml of Histopaque-1077 (Sigma, St. Louis, MO, USA), and separated through centrifugation for 30 min at 800 × g at RT. The pellet was resuspended with 3 ml of cold PBS and layered on 12 ml of Histopaque-1119 (Sigma, St. Louis, MO, USA) and centrifuged again for 30 min at 800 × g at RT. The neutrophil layer was then transferred to a 50-ml tube and 40 ml of PBS was added. After 10 min centrifugation at 500 × g, 4◦C, the supernatant was disposed and the neutrophil fraction was suspended in RPMI 1640 supplemented with 2% BSA to make the final neutrophil concentration at 500,000 cells/ml. One milliliter cell solution was added to each 22 × 22, 1.5/2.5 glass coverslip that has been treated with 0.001% poly-L-lysine (Sigma, St. Louis, MO, USA). After cell adherence, neutrophils were treated for 2 h with 5µg/ml CitH3 peptide [A(Cit)TKQTA(Cit) KSTGGKAP(Cit)KQLATKAA(Cit)KSAP] or H3 peptide (ARTKQTARKSTGGKAPRKQLATKAARKSAP) as control. Media was removed before immunocytochemistry.

#### Immunocytochemistry

Cells were washed before fixation in 3.7% paraformaldehyde/PBS for 10 min at RT. After three washes with PBS for 5 min each, cells were blocked in 2% BSA/PBS overnight at 4◦C. Neutrophils were then probed overnight at 4◦C with anti-CitH3 monoclonal antibody (Cayman Chemical, Ann Arbor, Michigan, USA) at a dilution of 1: 1,000 and anti-MPO polyclonal antibody (Abcam, Cambridge, MA, USA) at the dilution of 1:100 in 10% normal donkey serum. Cells were washed with PBS three times for 5 min each. Next, cells were incubated at RT for 1 h with FITCconjugated donkey anti-rabbit and TRITC-conjugated donkey anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) at a dilution of 1:300 in 10% normal donkey serum. Following another three washes with PBS, the coverslips were mounted with anti-fade reagent with DAPI (Thermo Fisher Scientific, Waltham, MA, USA). Eight fields each group were selected randomly for further quantification of NETs formation. Fluorescence microscopy images were analyzed with Image J software to count the number of NETs induced by peptides per 100 neutrophils.

#### Measurement of Serum Levels of dsDNA

The PicoGreen assay kit (Invitrogen, San Diego, CA, USA) was used to detect circulating dsDNA per manufacturer's instruction.

#### Statistical Analysis

Analyses were performed with GraphPad Prism 7 (GraphPad Software Inc., La Jolla, CA, USA). Data are presented as mean ± standard error of mean (SEM). Log-rank test was used to analyze the survival curve. One-way analysis of variance (ANOVA) followed by Bonferroni's multiple comparison test was used for comparisons between three or more groups. Mann–Whitney U test (non-parametric test) was performed for comparisons between two groups. A p < 0.05 was considered statistically significant. <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

# RESULTS

# CitH3 Increases HUVEC Permeability and Induces NET Formation

CitH3 is undetectable in the serum and peritoneal fluid under normal physiologic conditions in mice; however, it is elevated in the samples obtained from endotoxic or septic mice (17, 23).

The function of CitH3 remains largely unknown. It was recently reported that human recombinant CitH3 protein could disrupt endothelial barrier in vivo, probably through opening cell–cell adheres junctions and reorganizing the actin cytoskeleton (25). This conclusion needs further evaluation since the recombinant CitH3 protein was catalyzed by PAD4 purified from E. coli, and endotoxin contamination can compromise the validity of the experimental results (26, 27).

In the present study, synthesized CitH3 peptide was used. It was found that the synthesized CitH3 peptide (5µg/ml) significantly increased (p < 0.01) dextran leakage from HUVECs compared to sham control or 1µg/ml CitH3 treatment (**Figure 1A**). HUVECs treated with 5µg/ml H3 peptide had increased dextran leakage compared to sham; however, it was significantly lower (p < 0.01) compared to CitH3 at the same concentration (5µg/ml). The result suggests that CitH3 might be more toxic to HUVECs than H3, implicating the adverse effect of CitH3 in sepsis. Since CitH3 could induce prominent leakage from endothelial cells, neutralizing CitH3 may be beneficial in endotoxic shock or sepsis.

Human neutrophils incubated with CitH3 peptide also produced NETs after a short incubation period of 2 h. To visualize NETs, cells were co-stained with MPO, CitH3, and DNA. As shown in **Figure 1C**, there were no NETs detected in neutrophils treated with H3 peptide; however, NET structures were observed following CitH3 treatment, including stretched DNA filaments along with MPO and CitH3. As such, CitH3 increases HUVEC permeability and can induce NET formation ex vivo.

# The CitH3 mAb (4 Cit) Recognizes Histone H3 Citrullinated R26 and Has Higher Binding Capacity Compared to the Antibody Against Histone H3 [Citrullinated R2+R8+R17 (CitH3 mAb (3 Cit))]

Although there are some commercially available antibodies that bind to citrulline residues in histones, their efficacy has been found to be inconsistent (28), and there is a need for more reliable anti-CitH3 antibodies. In addition, the commercially available anti-CitH3 mAb only recognizes histone H3 citrullinated R2+R8+R17 [CitH3 mAb (3 Cit)], for which PAD4 is responsible (29). However, histone H3 R26 can also be citrullinated, by PAD2 but not PAD4 (30). Therefore, utilizing the CitH3 peptide with four citrulline residues at histone 2+8+17+26, we developed

available anti-CitH3 monoclonal antibody [CitH3 mAb (3 Cit)]. (A) Half microgram of five different peptides [H3, AceH3, CitH3 (R2,8,17,26), CitH3 (R26), and MetH3] or (B) 3 ng CitH3 protein were submitted to SDS-polyacrylamide gel electrophoresis and transferred onto a nitrocellulose membrane. Then, membranes were probed with the same concentration of CitH3 mAb (4 Cit) or CitH3 mAb (3 Cit) (2µg/ml). Other immunoblotting conditions were kept the same. Signal intensity was analyzed using Image Studio Lite and presented as mean ± SEM (*n* = 4/group). (C) Half microgram of CitH3 mAb (4 Cit) or CitH3 mAb (3 Cit) was coated per well in 96-well plates and incubated for 2 h at room temperature (RT) with CitH3 peptide (left panel), and or serum (1:1 diluted, right panel) from endotoxic mice, followed by anti-CitH3 polyclonal antibody incubation and then donkey anti-rabbit HRP-conjugated IgG. 3,3′ ,5,5′ -Tetramethylbenzidine was utilized to develop the plate for 30 min at RT in the dark before adding stop solution. Absorbance was measured at 450 nm and presented as mean ± SEM (*n* = 3/group). \**p* < 0.05 compared to CitH3 mAb (3 Cit); \*\*\**p* < 0.001 compared to CitH3 mAb (3 Cit); \*\*\*\**p* < 0.0001 compared to CitH3 mAb (3 Cit); ####*p* < 0.0001 compared to H3 incubated with CitH3 mAb (4 Cit). WB, Western blot.

the new mouse anti-CitH3 monoclonal antibody, referred to as CitH3 mAb (4 Cit), with the intention to completely block the CitH3 catalyzed by both PAD2 and PAD4.

To ensure the quality of the CitH3 mAb (4 Cit), we performed both immunoblotting and ELISA to test its specificity and capability for CitH3 recognition, compared with a commercial CitH3 mAb (3 Cit) (**Figure 2**). Immunoblotting was performed under the same experimental conditions, including concentration of antibodies, incubation time, and same-time exposure. As shown in **Figure 2A**, both CitH3 mAb (4 Cit) and CitH3 mAb (3 Cit) had a high specificity for CitH3. However, no band was detected for H3 (non-modified H3), AceH3 (acetylated H3), or MetH3 (methylated H3). For CitH3 detection, CitH3 mAb (3 Cit) only showed staining for CitH3 (R2+R8+R17+R26) peptide; however, no signal was found for CitH3 (R26) peptide. The CitH3 mAb (4 Cit) specifically detected histone H3 citrullinated R26, in accordance with our expectation. Furthermore, using CitH3 mAb (4 Cit), a stronger blot signal appeared for the CitH3 (R2+R8+R17+R26) peptide [densitometry unit, CitH3 mAb (4 Cit) vs. CitH3 mAb (3 Cit): 220,413.7 ± 4,444.8 vs. 104,416.3 ± 17,285.4; p < 0.0001]. Consistent with the peptide immunoblotting results, CitH3 mAb (4 Cit) gave a higher signal (three-fold) for the CitH3 protein [**Figure 2B**, densitometry unit, CitH3 mAb (4 Cit) vs. CitH3

mAb (3 Cit): 29,689 ± 919.8 vs. 9,596 ± 5,598; p < 0.05]. Since the experimental conditions of immunoblotting can only be roughly controlled, ELISA was used to further confirm the superiority of CitH3 mAb (4 Cit) in comparison with CitH3 mAb (3 Cit). As shown in **Figure 2C**, CitH3 mAb (4 Cit) showed significantly higher optical density (OD) than the CitH3 mAb (3 Cit) in detecting CitH3 peptide at various concentrations. Most importantly, a higher concentration of CitH3 was detected using CitH3 mAb (4 Cit) compared to the CitH3 mAb (3 Cit) in a mouse model of endotoxic shock (the right panel of **Figure 2C**).

Taken together, these findings suggest that CitH3 mAb (4 Cit) not only has specificity comparable to CitH3 mAb (3 Cit) but also possesses higher binding capabilities and recognizes more citrulline residues such as histone H3 citrullinated R2+R8+R17.

### The CitH3 mAb (4 Cit) Improves Survival Compared to CitH3 mAb (3 Cit) Following LPS-Induced Endotoxic Shock

After validation of the CitH3 mAb (4 Cit), we evaluated whether the new antibody could improve survival in a mouse model of LPS-induced endotoxic shock. It has been shown previously that CitH3 appears within 30 min of endotoxic shock (18, 23). In the present study, mice were intravenously administrated anti-CitH3 mAb immediately after LPS injection. The potential nonspecific therapeutic effects of immunoglobulin were controlled in a parallel cohort by giving mouse IgG treatment. Survival was monitored for 10 days (**Figure 3A**). The CitH3 mAb (4 Cit) treatment significantly improved the survival rate of endotoxic mice compared to either LPS + IgG (44.44 vs. 0%; p = 0.0004) or LPS + CitH3 mAb (3 Cit) group (44.44 vs. 11.11%; p = 0.0039). In addition, a significant decrease was observed in the serum IL-1β and TNF-α levels with CitH3 mAb (4 Cit) treatment at the 12-h time point, compared to either the LPS + IgG group (p < 0.001 and p < 0.01, respectively) or the LPS + CitH3 mAb (3 Cit) group (p < 0.01 and p < 0.05, respectively) (**Figure 3B**).

These data revealed that the new CitH3 mAb (4 Cit) could effectively protect the mice from lethal endotoxic shock, and significantly ameliorate the pro-inflammatory effects caused by LPS administration, in comparison with CitH3 mAb (3 Cit) and mouse IgG.

# The CitH3 mAb (4 Cit) Protects Against LPS-Induced ALI

Sepsis causes end-organ dysfunction, and septic patients are particularly at the risk of developing ALI (31). Moreover, ALI is one of the leading causes of mortality in septic patients (32). Therefore, we examined pathological changes in the lung 12 h after LPS insult, using H&E staining. As shown in the left panel of **Figure 4A**, IgG injected control animals showed normal histology, while the lungs from the LPS + IgG group displayed obvious inflammatory changes: inflammatory infiltrates, pulmonary congestion, edema, alveolar hemorrhage,

harvested at 12 h after LPS injection. (A) Representative hematoxylin and eosin staining of mouse lung sections is shown, and histological analysis of acute lung injury (ALI) was graded by a blinded pathologist and presented as ALI score (mean ± SEM, *n* = 3/group). White arrows indicate inflammatory changes. (B) IL-1β and TNF-α of lung homogenates were determined by ELISA. Cytokine levels were normalized by protein concentration and were significantly lower in CitH3 mAb (4 Cit)-treated mice (mean ± SEM, *n* = 3/group). \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001. ns, non significance.

and thickening of the alveolar wall. The CitH3 mAb (4 Cit) administration markedly ameliorated the histopathology changes induced by LPS. On the contrary, the CitH3 mAb (3 Cit) treatment was unable to protect the lungs against ALI.

The severity of lung injury was further quantitatively evaluated by a pathologist, blinded to the group allocation of the samples, by calculating the ALI score (**Figure 4A**). Lung septal mononuclear cell/lymphocyte, septal hemorrhage and congestion, neutrophils, alveolar macrophages, alveolar hemorrhage, and alveolar edema were assessed (see Materials and Methods). The ALI score was significantly increased after LPS insult (LPS + IgG vs. IgG: 6.750 ± 0.351 vs. 1.050 ± 0.000; p < 0.0001); however, the score was markedly reduced more than two-fold in the CitH3 mAb (4 Cit) treatment group [LPS + CitH3 mAb (4 Cit) vs. LPS + IgG: 2.583 ± 0.483 vs. 6.750 ± 0.3512; p < 0.0001]. The effects of CitH3 mAb (3 Cit) were not as strong as the CitH3 mAb (4 Cit). Even though the mean ALI score value of the LPS + CitH3 mAb (3 Cit) group was slightly lower than the LPS + IgG group, there were no statistical difference between the groups [LPS + CitH3 mAb (3 Cit) vs. LPS + IgG: 5.457 ± 0.107 vs. 6.750 ± 0.3512; p < 0.0001].

We also measured the levels of IL-1β and TNF-α in the lung homogenate. A significant decrease in lung IL-1β [LPS + CitH3 mAb (4 Cit) vs. LPS + IgG: 130.7 ± 44.67 vs. 601.0 ± 43.47 pg/mg; p < 0.001] and TNF-α [LPS + CitH3 mAb (4 Cit) vs. LPS + IgG: 19.33 ± 1.856 vs. 44.33 ± 6.741 pg/mg; p < 0.01] after CitH3 mAb (4 Cit) treatment was observed, in accordance with the serum IL-1β and TNF-α changes. However, the CitH3 mAb (3 Cit) could not effectively decrease IL-1β [LPS + CitH3 mAb (3 Cit) vs. LPS + IgG: 780 ± 54.5 vs. 601.0 ± 43.47 pg/mg; ns] or TNF-α [LPS + CitH3 mAb (3 Cit) vs. LPS + IgG: 55.00 ± 5.774 vs. 44.33 ± 6.741 pg/mg; ns]. In these experiments, the sample size may seem low (n = 3/group) but the results reach significant difference (p < 0.05) based on our statistical analysis.

Taken together, administration of CitH3 mAb (4 Cit), but not CitH3 mAb (3 Cit), was found to protect against LPS-induced ALI and attenuate the inflammatory cytokines.

# Administration of CitH3 mAb (4 Cit) Decreases Serum Levels of dsDNA

Since CitH3 is generally considered a component of NETs, the binding of the anti-CitH3 antibody and the antigen might affect NETs in the circulation.

dsDNA levels were significantly decreased after the CitH3 mAb (4 Cit) treatment [LPS + IgG vs. LPS + CitH3 mAb (4 Cit): 9.047 ± 0.816 vs. 2.537 ± 0.3767µg/ml; p < 0.001]. However, this was not observed with the CitH3 mAb (3 Cit) [LPS + IgG vs. LPS + CitH3 mAb (3 Cit): 9.047 ± 0.816 vs. 10.72 ± 0.68µg/ml; ns]. Moreover, the CitH3 mAb (4 Cit) treatment group had less dsDNA compared with the 3 Cit mAb [LPS + CitH3 mAb (4 Cit) vs. LPS + 3 Cit mAb: 2.537 ± 0.3767 vs. 10.72 ± 0.68µg/ml; p < 0.0001].

# DISCUSSION

In the current study, we have demonstrated that exposure to CitH3 can increase endothelial cell leakage and induce the formation of NETs through a positive feedback system. We neutralized the circulating CitH3 with a newly developed CitH3 mAb (4 Cit) that strongly binds to four citrulline sites on CitH3, in comparison with the commercial CitH3 mAb (3 Cit) that binds to only three citrulline sites. We found that injection of CitH3 mAb (4 Cit) markedly improves survival following LPSinduced lethal endotoxic shock and attenuates pro-inflammatory responses, ALI, as well as NET formation. Our antibody fills the gap in the field of citrullinated histone that was reported by Neeli and Radic, "current challenges and limitations in antibody-based detection of citrullinated histones," in Frontiers in Immunology (28). In addition, we believe that CitH3 can be a potential target for directed therapies in endotoxic patients in the future.

Neutrophils are the most abundant innate immune cells in humans, and their functions include phagocytosis of pathogens, release of multiple inflammatory mediators, and induction of NET formation. Release of NETs by neutrophils was first discovered by Brinkman et al. (4), and has attracted significant interest in recent years. NETs have been found to play key roles in various diseases including sepsis, tumors, systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), gout, and others (15). NETs also play a key role in sepsis. NETs are able to trap microbes and keep them in a restricted area with high concentrations of anti-microbial agents, which are released as components of NETs (33). DNA itself has been reported to possess anti-microbial activities (34). Histones and histone-like proteins also function as microbial-combating reagents in a variety of species (35). However, unless carefully controlled, excessive NETs and components of NETs can be detrimental in sepsis. Histones and citrullinated histones are both components of NETs and have been found to be elevated in sepsis. Histones bind to endothelial cells, cause an increase in endothelial cell permeability and Ca2<sup>+</sup> influx, eventually leading to cell death (36). Histones also promote thrombosis through impairing thrombomodulin-dependent protein C activation (37, 38). In addition, extracellular histones have been found to mediate liver injury through toll-like receptor 2 (TLR2) and TLR4 (39). Unlike histones, the role of citrullinated histones in diseases remains

FIGURE 5 | The CitH3 mAb (4 Cit) decreases serum dsDNA. Mice were randomly divided into four groups: (1) IgG only (20 mg/kg), (2) LPS (20 mg/kg) + IgG (20 mg/kg), (3) LPS + 4 CitH3 mAb (20 mg/kg), and (4) LPS + CitH3 mAb (3 Cit) (20 mg/kg) (*n* = 3/group). Blood was harvested at 12 h after LPS injection. dsDNA was measured using PicoGreen assay kit and expressed as mean ± SEM (*n* = 3/group). \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001.

largely unknown. CitH3 release has been detected in blood and peritoneal fluid in sepsis (17, 18, 23); however, the effects of CitH3 production and its pathophysiologic roles in sepsis are unclear.

There are multiple ways to induce septic response in animals: (1) LPS, bacterial, or cecal slurry injection; (2) endogenic protection barrier model, such as CLP and colon ascendens stent peritonitis (CASP); and (3) extra-abdominal models of sepsis: pneumonia and urosepsis (40). Due to its stability, repeatability, and long track record (41), we selected LPS-induced endotoxic shock as a starting model to test the therapeutic efficacy of the CitH3 mAb (4 Cit). The commercial anti-CitH3 mAb (3 Cit) detected H3 citrullinated R2+R8+R16 and served as a control. Improved survival was observed in the CitH3 mAb (4 Cit)-treated mice compared to the CitH3 mAb (3 Cit) or IgG-treated endotoxic mice (**Figure 3A**). Several reasons might contribute to the better outcome after CitH3 mAb (4 Cit) administration: (1) less endothelial leakage (**Figure 1A**); (2) inhibited inflammatory responses (**Figure 3B**, **Figure 4B**); (3) attenuated ALI (**Figure 4A**); and (4) decreased formation of NETs (**Figure 5**).

Endothelial dysfunction is the main reason for multiple organ failure in sepsis, since it causes tissue edema, disarrangement of hemostasis, and vasomotor control, eventually leading to death (42). In this study, the CitH3 peptide was shown to increase permeability of HUVECs in vitro (**Figure 1A**). In addition, infusion of CitH3 protein has been shown to induce an extravasation of fluorescent-labeled albumin across mouse mesenteric micro vessels without cell death (25). These observations suggest that CitH3 might be one of the causes for vascular leakage and organ edema in sepsis.

Release of cytokines, for example, IL-1β and TNF-α, can activate and recruit effector inflammatory cells to the site of infection. However, excessive or deregulated cytokine profile is a common occurrence in sepsis, exerting harmful effects both systemically and locally. The level of TNF-α in septic patients correlates with fatal outcome (43). Persistent elevation of cytokine concentrations in patient plasma, including TNF-α and IL-1β, is associated with poor prognosis in septic patients with acute respiratory distress syndrome (ARDS) (44). The fact that CitH3 mAb (4 Cit) can decrease the levels of circulating IL-1β and TNF-α (**Figure 3B**) could attenuate the adverse effects that result from excessive cytokine release.

ALI is one of the most frequent complications to develop in septic patients and among the leading causes of deaths in these patients (32). In our experimental LPS model, survival correlated strongly with reduced lung injury. Endotoxic mice treated with the CitH3 mAb (4 Cit) had better lung histology (lower ALI score) as well as much less pro-inflammatory cytokines in lung, compared to the mice treated with IgG or the CitH3 mAb (3 Cit) (**Figure 4A**). The alleviated ALI might be one of the mechanisms responsible for the significantly better survival in the CitH3 mAb (4 Cit)-treated mice. Endotoxin/sepsis-induced inflammatory responses are important defense mechanisms in disease conditions; however, they can also contribute to the development of ALI. Endothelial cells, epithelial cells, resident alveolar macrophages, and neutrophils secrete cytokines and chemokines, leading to an increase in tissue inflammation and subsequent cellular damage. Among the cytokines, IL-1β and TNF-α are major cytokines released into the alveolar spaces of patients with ARDS (45). In our study, we also found elevated levels of IL-1β and TNF-α in lung homogenates (**Figure 4B**). IL-1β and TNF-α are known to exaggerate the inflammatory responses initiated by endotoxin, which could result in adverse consequences. We have shown that administration of CitH3 mAb (4 Cit) significantly attenuated the local concentrations of IL-1β and TNF-α in the lung, which could be protective against development of ALI.

Our results (**Figure 1C**) also suggest that CitH3 can induce NET formation through a positive feedback mechanism. There is plenty of evidence showing the detrimental role of NETs (15). NETs can damage epithelium, endothelium, and various tissues including liver and lung (36, 46–48); moreover, NETs promote thrombosis, leading to vasculature occlusion (49–52). The level of circulating dsDNA has been shown to correlate with disseminated intravascular coagulation (DIC) score and predicts DIC independently (53). In this study, the CitH3 mAb (4 Cit) efficiently inhibited this positive feedback loop, leading to less formation of CitH3/NETs (**Figure 5**). The underlying mechanism may be that the mAb (4 Cit) binds to CitH3 and forms immune complex, which are further cleared by immune system (54, 55). The DNA backbone and associated proteins may also be eliminated along with the ICs. Therefore, neutralizing CitH3 directly decreased free CitH3 in circulation, as well as blocking the positive feedback to inhibit CitH3 self-amplification, thus preventing the endothelial dysfunction caused by CitH3.

The major difference between the CitH3 mAb (4 Cit) and the CitH3 mAb (3 Cit) is thought to be secondary to their relative abilities to neutralize the CitH3 protein. The newly developed CitH3 mAb (4 Cit) recognizes four citrulline sites: H3 citrullinated R2+R8+R17+R26, while the CitH3 mAb (3 Cit) is only designed for three citrulline spots: H3 citrullinated R2+R8+R17. As demonstrated by immunoblotting and ELISA (**Figure 2**), the CitH3 mAb (4 Cit) specifically recognized H3 R26 citrullination and had a stronger capability to bind CitH3. In addition, different citrullination may correspond to different PADs. Among the five PAD isoforms that have been discovered, only PAD2 and PAD4 have been demonstrated by several lines of evidence to translocate from cytosol to nucleus and citrullinate histone H3. We reasoned that CitH3 in sepsis originated from PAD2 or PAD4 pathways, or both. It has been reported that H3 R26 is a valid target for PAD2, but not PAD4 (30). Therefore, the CitH3 mAb (4 Cit) is suspected to bind and neutralize more circulating CitH3. The one citrulline difference partly contributes to the binding ability discrepancy between these two antibodies.

Wang et al. have demonstrated in Science that PAD4 deiminates three arginine residues on H3 at Arg 2, Arg 8, and Arg 17 in vitro, and two arginine residues on H3 at Arg 8 and Arg 17 in vivo (29). It is not clear what causes the different citrullinations. Basically, an enzyme's active site binds substrates and plays a role in catalysis. It is conceivable that some proteins might bind to histone H3 in the in vivo condition and therefore prevent H3 from citrullination by PAD4.

Our interpretations regarding differential PAD2/PAD4 involvement in CitH3 production is supposed not only by the data from this study but also by previous reports by other investigators (29). These findings will need additional verification in the future, and more mechanistic experiments will have to be performed to fully understand this complex process.

In this study, we decided to use the histone H3 peptide for several reasons. First, the N-terminal tail (N-tail) of histone H3, which protrudes beyond the nucleosome DNA (56), is regulated by multiple posttranslational modifications (PTMs) (57). Focusing on citrullination of the N-tail, we synthesized Nterminal CitH3 peptide, instead of CitH3 protein, to ensure that histone H3 is citrullinated at the N-terminus of four arginine residues. Second, the commercial CitH3 (citrulline R2+R8+R17) (58, 59) antibodies, including Item 9003062 (Cayman Chemical, Ann Arbor, MI) and ab5103 (abcam, Cambridge, MA), are generated with CitH3 N-tail peptide. To compare our CitH3 mAb (4 Cit) to the commercial CitH3 mAb (3 Cit), it is more appropriate to use the N-terminal CitH3 peptide as an antigen. Third, histone H3 (citrullinated or non-citrullinated) epitope(s) can be cleaved off at the N-terminus of H3 (60). To determine the effect of the N-tail of H3 and CitH3 in vitro, it is logical to use the N-terminal H3/CitH3 peptide as the stimulus.

There are several limitations to this study. We only used a murine endotoxic shock model to test the therapeutic effects of the CitH3 mAb (4 Cit); however, further evaluation is required in different models of bacterial and polymicrobial infections. In addition, more mechanism studies are needed. For example, further exploration of CitH3 in diverse cell signaling pathways is helpful to better explain how the CitH3 mAb (4 Cit) improves outcomes in endotoxic shock.

In conclusion, we have demonstrated that CitH3 can increase endothelial cell leakage and self-amplify through positive feedback. Neutralizing circulating CitH3 with the CitH3 mAb (4 Cit) markedly increases mouse survival from LPS-induced lethal endotoxic shock, likely secondary to the CitH3 mAb (4 Cit) binding to CitH3 and specifically recognizing H3 R26 citrullination. The CitH3 mAb (4 Cit) also attenuates pro-inflammatory responses, ALI, as well as NET formation, compared to the CitH3 mAb (3 Cit). Overall, these results suggest that sufficient neutralization of CitH3 might be a promising therapeutic strategy for endotoxic shock.

#### DATA AVAILABILITY STATEMENT

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

#### ETHICS STATEMENT

The protocol for the animal experiments was approved by the University of Michigan Institutional Animal Care and Use

#### REFERENCES


Committee (PRO00008861). All experiments complied with animal welfare and research regulations.

#### AUTHOR CONTRIBUTIONS

YLi and HA designed this study. QD, BP, and YLia performed the experiments and collected and analyzed data. BL, YT, AW, and JZ provided experimental support. XD pathologically examined the lung tissues. QD wrote the manuscript, which was critically reviewed and revised by YLi, HA, NM-V, ZW, and AW. All authors read and approved the final manuscript.

# FUNDING

This work was funded by grants from Mcubed U049657 and Kickstart N022142 to YLi, SIS to AW, and UMHS-PUHSC Joint Institute U050150 to HA.


**Conflict of Interest:** YLi and HA are inventors on a patent application related to the CitH3 mAb (4Cit).

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

Copyright © 2020 Deng, Pan, Alam, Liang, Wu, Liu, Mor-Vaknin, Duan, Williams, Tian, Zhang and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(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.

# Impaired Glucocorticoid Receptor Dimerization Aggravates LPS-Induced Circulatory and Pulmonary Dysfunction

Martin Wepler 1,2 \* † , Jonathan M. Preuss 3†, Tamara Merz <sup>1</sup> , Clair Hartmann1,2 , Ulrich Wachter <sup>1</sup> , Oscar McCook <sup>1</sup> , Josef Vogt <sup>1</sup> , Sandra Kress <sup>1</sup> , Michael Gröger <sup>1</sup> , Marina Fink <sup>1</sup> , Angelika Scheuerle<sup>4</sup> , Peter Möller <sup>4</sup> , Enrico Calzia<sup>1</sup> , Ute Burret <sup>3</sup> , Peter Radermacher <sup>1</sup> , Jan P. Tuckermann<sup>3</sup> and Sabine Vettorazzi <sup>3</sup> \*

#### Edited by:

Guochang Hu, University of Illinois at Chicago, United States

#### Reviewed by:

Monowar Aziz, Feinstein Institute for Medical Research, United States Claude Libert, Flanders Institute for Biotechnology, Belgium

#### \*Correspondence:

Martin Wepler martin.wepler@uni-ulm.de Sabine Vettorazzi sabine.vettorazzi@uni-ulm.de

†These authors have contributed equally to this work

#### Specialty section:

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

Received: 12 September 2019 Accepted: 30 December 2019 Published: 23 January 2020

#### Citation:

Wepler M, Preuss JM, Merz T, Hartmann C, Wachter U, McCook O, Vogt J, Kress S, Gröger M, Fink M, Scheuerle A, Möller P, Calzia E, Burret U, Radermacher P, Tuckermann JP and Vettorazzi S (2020) Impaired Glucocorticoid Receptor Dimerization Aggravates LPS-Induced Circulatory and Pulmonary Dysfunction. Front. Immunol. 10:3152. doi: 10.3389/fimmu.2019.03152 1 Institute for Anesthesiologic Pathophysiology and Process Engineering, Ulm University, Ulm, Germany, <sup>2</sup> Department of Anesthesiology, University Hospital, Ulm, Germany, <sup>3</sup> Institute of Comparative Molecular Endocrinology (CME), Ulm University, Ulm, Germany, <sup>4</sup> Institute of Pathology, University Hospital, Ulm, Germany

Background: Sepsis, that can be modeled by LPS injections, as an acute systemic inflammation syndrome is the most common cause for acute lung injury (ALI). ALI induces acute respiratory failure leading to hypoxemia, which is often associated with multiple organ failure (MOF). During systemic inflammation, the hypothalamus-pituitary-adrenal axis (HPA) is activated and anti-inflammatory acting glucocorticoids (GCs) are released to overcome the inflammation. GCs activate the GC receptor (GR), which mediates its effects via a GR monomer or GR dimer. The detailed molecular mechanism of the GR in different inflammatory models and target genes that might be crucial for resolving inflammation is not completely identified. We previously observed that mice with attenuated GR dimerization (GRdim/dim) had a higher mortality in a non-resuscitated lipopolysaccharide (LPS)- and cecal ligation and puncture (CLP)-induced inflammation model and are refractory to exogenous GCs to ameliorate ALI during inflammation. Therefore, we hypothesized that impaired murine GR dimerization (GRdim/dim) would further impair organ function in LPS-induced systemic inflammation under human like intensive care management and investigated genes that are crucial for lung function in this setup.

Methods: Anesthetized GRdim/dim and wildtype (GR+/+) mice were challenged with LPS (10 mg·kg−<sup>1</sup> , intraperitoneal) and underwent intensive care management ("lung-protective" mechanical ventilation, crystalloids, and norepinephrine) for 6 h. Lung mechanics and gas exchange were assessed together with systemic hemodynamics, acid-base status, and mitochondrial oxygen consumption (JO2). Western blots, immunohistochemistry, and real time quantitative polymerase chain reaction were performed to analyze lung tissue and inflammatory mediators were analyzed in plasma and lung tissue.

Results: When animals were challenged with LPS and subsequently resuscitated under intensive care treatment, GRdim/dim mice had a higher mortality compared to GR+/<sup>+</sup> mice, induced by an increased need of norepinephrine to achieve hemodynamic targets.

**221**

After challenge with LPS, GRdim/dim mice also displayed an aggravated ALI shown by a more pronounced impairment of gas exchange, lung mechanics and increased osteopontin (Opn) expression in lung tissue.

Conclusion: Impairment of GR dimerization aggravates systemic hypotension and impairs lung function during LPS-induced endotoxic shock in mice. We demonstrate that the GR dimer is an important mediator of hemodynamic stability and lung function, possibly through regulation of Opn, during LPS-induced systemic inflammation.

Keywords: glucocorticoid receptor, lung function, endotoxic shock, inflammation, osteopontin

#### INTRODUCTION

Anti-inflammatory acting glucocorticoids (GCs) mediate their effects through the glucocorticoid receptor (GR). The GR is an intracellular, ligand-activated transcription factor, which regulates gene transcription as a protein dimer or monomer in several mechanisms: as a protein dimer it can bind palindromic DNA sequences (glucocorticoid response elements—GRE) or DNA half sites as a monomeric protein [GR monomer, (1)]. Further GR monomers potentially inhibit the activity of pro-inflammatory transcription factors, e.g., NF-κB (2–5), AP-1 (6–8), or IRF-3 (9–11). This trans-repression is one major mechanism of GCs anti-inflammatory effects (12). However, it was recently shown that GC mediated anti-inflammatory responses also crucially require gene activation during inflammation (13). Indeed, mice with a point mutation in the GR DNA-binding domain (GRdim) exhibit less transactivation of GC-induced genes in vivo (1, 14) and fail to resolve inflammation in allergic (15), autoimmune (rheumatoid arthritis) (16), and systemic inflammation (17, 18).

Previously, we identified a novel mechanism by which GCs interfere with the pathogenesis of murine ALI, involving increased sphingosine kinase 1 (SphK1) gene expression and sphingosine-1-phosphate (S1P) production. The SphK1–S1P axis is recognized as an important regulator of endothelial barrier integrity that prevents lung inflammation (19, 20). In our previous study, we showed that the induction of SphK1 was GR dimerization dependent and therefore mice with an impaired GR dimerization (GRdim/dim) had an impaired lung barrier function during systemic lipopolysaccharides (LPS)-induced inflammation and GC treatment (13).

However, all studies described so far were lacking simultaneous control of temperature, as well as hemodynamic, and respiratory support, which is standard in intensive care treatment. Moreover, if potential beneficial effects of GCs in the treatment of lung injury occur, it is yet not clear which mechanisms are involved. Therefore, we tested the effects of an impairment of the GR in a murine model of LPS-induced systemic inflammation when factors like temperature, hemodynamics, and respiration are controlled (intensive care management including measurements of hemodynamics, infusion of crystalloids and norepinephrine to achieve hemodynamic targets, lung-protective mechanical ventilation, determination of gas exchange) in the present study. During intensive care management additional information about metabolic and hemodynamic parameters are observed that were missing in the former studies. We report that a congenital deficiency of the GR dimer aggravates hypotension, impairs lung function, and increases mortality in LPS-challenged mice.

#### MATERIALS AND METHODS

This study was approved by the federal authorities for animal research of the Regierungspräsidium Tübingen, Baden-Wuerttemberg, Germany, and performed in adherence with the National Institutes of Health Guidelines on the Use of Laboratory Animals and the European Union "Directive 2010/63 EU on the protection of animals used for scientific purposes". GRdim/dim mice (Nr3c1tm3Gsc) (21) were bred in a mixed background (129/SvEv × C57BL/6) and housed in the animal facility at University Ulm. GR+/<sup>+</sup> littermate controls were used as wildtype mice. Animals were kept under standardized conditions and were equally distributed in terms of age and body weight.

#### Implementation of General Anesthesia and Surgery

Surgery for all animals included induction of anesthesia with sevoflurane (2.5%; sevoflurane, Abbott, Wiesbaden, HE, Germany) as described previously (22, 23), followed by intraperitoneal injection (ip) of ketamine (120 µg·g −1 ; Ketanest-S, Pfizer, New York City, NY), midazolam (1.25 µg·g −1 ; Midazolam-ratiopharm, Ratiopharm, Ulm, BW, Germany), and fentanyl (0.25 µg·g −1 ; Fentanyl-hameln, Hameln Pharma Plus GmbH, Hameln, NI, Germany). Afterwards, animals were placed on a closed-loop-system for body temperature control (22, 23). Lung-protective mechanical ventilation using a small animal ventilator (FlexiVent, Scireq, MO, Canada) was performed via a tracheostomy (22, 23). Surgical instrumentation comprised catheters in the jugular vein, the carotid artery, and the bladder (22, 23). General anesthesia was titrated to guarantee complete tolerance against noxious stimuli and was sustained by continuous intravenous administration of ketamine, midazolam, and fentanyl to reach deep sedation. Animals were mechanically ventilated with ventilator settings being FiO<sup>2</sup> 0.21%, respiratory rate 150·min−<sup>1</sup> , tidal volume of 6 mL·kg−<sup>1</sup> , and inspiratory/expiratory time ratio 1:2. Ventilation was modified to maintain an arterial PaCO<sup>2</sup> between 30 and 45 mmHg, and positive end-expiratory pressure (PEEP) was adjusted according to the arterial PaO<sup>2</sup> (PaO2/FiO2-ratio > 300 mmHg: PEEP = 3 cmH2O; PaO2/FiO2-ratio < 300 mmHg: PEEP = 5 cmH2O; PaO2/FiO2-ratio < 200 mmHg: PEEP = 8 cmH2O) (22, 23). Recruitment maneuvers (5 s hold at 18 cm H2O) were repeated hourly to avoid any impairment of thoraco-pulmonary compliance due to anesthesia- and/or supine position-induced atelectasis.

#### Induction of Systemic Inflammation

After surgical instrumentation, systemic inflammation was induced by intraperitoneal (i.p.) injection of lipopolysaccharides (LPS=lipopolysaccharide from Escherichia coli [055:B5], L2880 Sigma, 10 mg·kg−<sup>1</sup> , dissolved in 10 µl·g <sup>−</sup><sup>1</sup> phosphate buffered saline [PBS]). Mice were then resuscitated with crystalloids (30 µl·g −1 ·h −1 , Jonosteril, Braun Medical, Melsungen, HE, Germany). As soon as the mean arterial blood pressure (MAP) dropped below 55 mmHg, infusion of norepinephrine was started to reach a MAP >55 mmHg during the 6 h of resuscitation (maximum infusion rate 1.5 µg·h −1 ). If blood pressure declined despite increasing doses of norepinephrine, the experiment was terminated. GR+/<sup>+</sup> mice, which received vehicle (10 µl·g <sup>−</sup><sup>1</sup> PBS) with subsequent resuscitation, served as controls.

#### Parameters of Lung Mechanics, Hemodynamics, Gas Exchange, and Metabolism

Systemic hemodynamics, body temperature, and static thoracopulmonary compliance were recorded hourly. Blood gas tensions, acid-base status, glycaemia, and lactatemia were assessed at the end of the resuscitation period via aterial blood gas analysis (ABL800 Felx; Radiometer, Krefeld, Germany) (22, 23). At the end of the experiment, animals were exsanguinated, blood and lung tissue were taken immediately thereafter, and prepared for further analyses (22, 23). All lung tissue was utilized due to organ size. The left lung was harvested for histology and IHC, whereas the right lung served for immunoblotting, expression analysis, and cytokine and chemokine evaluation.

#### Histological Analysis of Lung Tissue

Histological analysis of lung tissue was independently performed by two experienced pathologists (AS and PM) blinded for group assignment. Similar to previous studies (23), analyzed criteria comprised thickening of alveolar membranes, dystelectasis, emphysema, and inflammatory cell (lymphocytes) infiltration. These parameters were scored from 0 (absent), 1 (hardly detectable), 2 (rare), 3 (minor), 4 (moderate), to 5 (extensive).

#### Mitochondrial Respiration

Mitochondrial respiratory capacity was determined via highresolution respirometry with a clark-electrode-based system (Oxygraph 2k, OROBOROS Instruments Corp., Innsbruck, Austria) as described previously (22). Post-mortem heart, muscle, liver, and brain biopsies were mechanically homogenized in Mir05 (respiration medium). Mir05 is composed of 0.5 mM EGTA, 3 mM MgCl2·6H2O, 60 mM Lactobionic acid, 20 mM Taurine, 10 mM KH2PO4, 20 mM HEPES, 110 mM Sucrose, 1 g·L <sup>−</sup><sup>1</sup> bovine serum albumin). 1.5–2 mg of tissue (1.5 mg: heart, 2 mg tissue: muscle, liver, and brain) were added to the Oxygraph chamber. By addition of a defined sequence of substrates and inhibitors, various states of mitochondrial function could be assessed. Complex I activity was determined after addition of 10 mM pyruvate, 10 mM glutamate, 5 mM malate, and 5 mM ADP. Ten micrometers cytochrome c was added to check for mitochondrial integrity. Maximum oxidative phosphorylation (OxPhos) was evaluated after subsequent addition of 1 mM octanoyl-carnitine and 10 mM succinate. Leak compensation was assessed after inhibition of the ATPsynthase by 2.5µM oligomycin, followed by stepwise titration of the uncoupling agent Carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone (FCCP, final concentration 1.5µM) to reach maximum respiratory activity of the electron transfer system in the uncoupled state (ETS).

#### Western Blot

The lung was dissected at the end of the experiment and frozen on dry ice. The tissue was homogenized in EDTAfree lysis buffer with tissue homogenisator (Precellys <sup>R</sup> ). Total protein concentration was determined using Pierce <sup>R</sup> BCA Protein Assay Kit (23225). Proteins were separated by SDS-PAGE and blotted on nitrocellulose membrane using the Trans-Blot Turbo system (BioRad). Osteopontin was detected with anti-osteopontin primary antibody (mouse anti-mouse, LFMb-14, Santa Cruz Biotechnology Inc., sc-73631) diluted 1:500. Vincullin was detected as loading control using anti-vincullin primary antibody (mouse anti-mouse, Santa Cruz Biotechnology Inc., sc-73614) diluted 1:1.000. Primary antibody incubation was done overnight at 4◦C. Secondary antibody (rabbit antimouse, HRP coupled polyclonal Ig, Dako, P0161) was diluted 1:10.000 and incubated 1 h at room temperature. Membrane blocking and dilutions were done with 5% BSA. Except for the last washing step which was done with 1x TBS, 0.1% TBST was used. Blots were developed using immobilion forte WBLUF0500 (Merck Millipore) and ImageLab software (version 5.2). Osteopontin abundancy was quantified using ImageJ (version 1.52a) by determination of mean signal intensity of osteopontin normalized to mean intensity of vincullin.

#### Immunohistochemistry

The left lung was formalin-fixed and embedded in paraffin for immunohistochemistry analysis. Immunohistochemistry (IHC) for extravascular albumin content (anti-albumin rabbit polyclonal #16475-1-AP, Proteintech, USA) was performed as described previously (24). Primary antibodies were detected by secondary anti-rabbit antibody conjugated to AP (Alkaline Phosphatase-conjugated antibody; Jackson, ImmunoResearch, West Grove, Pa) and visualized with a red chromogen (Darko REAL Detection System Chromogen Red), and Mayers hematoxylin (Sigma, Taufkirchen, Germany). Visualization was performed using the Zeiss Axio Imager A1 microscope (Zeiss, Jena, TH, Germany). Four distinct 800.000 mm<sup>2</sup> regions were quantified for intensity of signal by using the Axio Vision 4.8 software. Results are presented as mean densitometric sum red (24, 25).

# Cell Culture

The primary bone marrow-derived macrophages (BMDMs) were isolated from humerus, femur and tibia of 8–13 weeks old GR+/<sup>+</sup> and GRdim/dim mice as described previously (18). Briefly, cells were cultured until day 7 in DMEM (D5671, sigma) supplemented with 10% fetal bovine serum (FBS, F7524, sigma), 30% L929-cell conditioned medium, 1% Penicillin / Streptomycin (P0781, sigma), 1% L-Glutamine (G7513, sigma), 1% Sodium Pyruvate (S8636, sigma) at 37◦C and 5% CO2. All BMDMs were treated with PBS as control and LPS (100 ng/ml, L6529, sigma) for the indicated time points. Osteopontin Elisa (R&D System) was performed with the supernatant of GR+/<sup>+</sup> and GRdim/dim BMDMs.

#### Analysis of Relative mRNA Levels

For quantitative real-time PCR analysis (qRT-PCR), RNA was extracted from lungs by homogenization with tissue homogenisator (Precellys <sup>R</sup> ) in Trizol (invitrogen) following the manufacturer's instructions. RNA quality was checked using the nanodrop (thermofisher). DNaseI-treated RNA (1 µg) was used to generate cDNA by oligo(dT) priming. qRT-PCR was performed with the ViiATM 7 Realtime PCR System (Life technologies) using a Platinum SYBR Green (Invitrogen) and analyzed with the QuantStudio Realtime-PCR software using the 11CT method. β-Actin and Ribosomal protein L (Rpl) served as housekeeping genes. The specific primers were obtained from Sigma with the sequences listed in **Table 1**.

# Measurements of Cytokine and Chemokine Concentrations

Bio-Plex Pro Mouse Cytokine 23-plex Assay (Group I) (Biorad) was used to measure 23 cytokines, chemokines and growth factors simultaneously in the plasma. The Bio-Plex Assay was conducted according to the manufacturer's protocol. The assay was performed with Bio-Plex 200 machine (Biorad) and analyzed with the Bio-Plex Manager TM 6.1 software (Biorad).

#### Statistical Analysis

Unless stated otherwise, all data are presented as median (25th and 75th percentile). Data sets were analyzed using nonparametric statistics, i.e., Mann–Whitney U-test (one factor, two independent samples) or Kruskal–Wallis test with posthoc Dunn's comparison testing (one factor, four independent


samples). P < 0.05 were considered statistically significant. Quantitative graphical presentations and statistical analyses were accomplished by using GraphPad Prism 7 (GraphPad Software Inc., La Jolla, Calif).

# RESULTS

#### GR Dimerization Mediates Stability of Hemodynamics, Acid-Base Status, and Mitochondrial Respiration After LPS Challenge

To examine the effects of an impaired GR dimerization during resuscitation in LPS-induced systemic inflammation, hemodynamics, metabolic parameters, and mitochondrial respiration were investigated. Hemodynamic stability was defined as MAP >55 mmHg and preserved via infusion of crystalloids and, if necessary, norepinephrine (maximum infusion rate 1.5 µg·h −1 ). Therefore, in some mice (GR+/<sup>+</sup> PBS n = 0, GRdim/dim PBS n = 1, GR+/<sup>+</sup> LPS n = 1, GRdim/dim LPS n = 5), the experiment had to be terminated due to hemodynamic instability despite increasing norepinephrine doses, which lead to MAP values below 55 mmHg. After LPS-challenge, GRdim/dim mice had a trend toward a lower MAP when compared to GR+/<sup>+</sup> mice (**Table 2**). In line with the compromised systemic hemodynamics, norepinephrine requirements in GRdim/dim mice challenged with LPS were significantly higher when compared to LPS-challenged GR+/<sup>+</sup> animals (**Figure 1** and **Table 2**). In GRdim/dim mice, a challenge with LPS led to an increase of lactate levels at the end of the resuscitation phase, and, consequently, to a more pronounced decrease in base excess (BE) compared to GR+/<sup>+</sup> mice, suggesting an aggravation of metabolic acidosis (**Table 2**). Because hyperlactatemia may be linked to disturbances in mitochondrial respiration (26, 27), important metabolic organs like muscle, heart, liver and brain were investigated for mitochondrial respiration in the current study. In LPS-challenged GRdim/dim mice, oxygen flux during maximal coupled mitochondrial respiration (OxPhos) was decreased in heart tissue when compared to vehicle treated GRdim/dim mice, whereas in GR+/<sup>+</sup> animals OxPhos was not changed between PBS and LPS-challenge (**Figure 2A**), suggesting a disturbed mitochondrial function in the heart of GRdim/dim animals. In the liver tissue, oxygen flux during maximal mitochondrial respiration and during inhibited ATP synthase in the coupled state (LEAK) was higher in LPSchallenged than in vehicle challenged GR+/<sup>+</sup> mice (**Figure 2B**), suggesting an increased mitochondrial respiratory capacity in these animals. However, in the muscle and brain mitochondrial respiration was not changed neither in GRdim/dim nor in GR+/<sup>+</sup> animals during PBS- or LPS-challenge under resuscitation (**Figures 2C,D**). In summary, the impaired dimerization of the GR results in higher noradrenaline requirements and therefore compromised systemic hemodynamics as well as a more pronounced lactic acidosis and altered mitochondrial respiration during LPS-challenge.

TABLE 2 | Hemodynamic and metabolic measurements as well as parameters of lung function in GRdim/dim and GR+/<sup>+</sup> mice intraperitoneally challenged with lipopolysaccharide (LPS) or vehicle (phosphate buffered saline, PBS) at the end of the experiment.


#P < 0.05 vs. GRdim PBS. PBS = phosphate buffered saline (vehicle), 10 µl·g −1 . LPS = Lipopolysaccharide from Escherichia coli (055:B5), 10 mg·kg−<sup>1</sup> . Data is shown as median (25th and 75th percentile).

FIGURE 1 | Doses of norepinephrine in mechanically ventilated GRdim/dim and GR+/<sup>+</sup> mice intraperitoneally challenged with lipopolysaccharides (LPS) or vehicle (PBS) and resuscitated (crystalloids, norepinephrine) for 6 h. Norepinephrine was titrated intravenously during resuscitation to keep systemic mean arterial blood pressure above 55 mmHg. LPS = lipopolysaccharide from Escherichia coli [055:B5], 10 mg·kg−<sup>1</sup> , dissolved in 10 µl·g <sup>−</sup><sup>1</sup> phosphate buffered saline (PBS). GR+/<sup>+</sup> mice challenged with PBS: n = 8, GRdim/dim mice challenged with PBS: n = 7, GR+/<sup>+</sup> mice challenged with LPS: n = 9, GRdim/dim mice challenged with LPS: n = 11. Data is presented as median (25th and 75th percentile and minimum/maximum).

#### GRdim/dim Mice Have Higher Mortality After LPS Challenge With Subsequent Resuscitation

In LPS-challenged mice with subsequent resuscitation for a maximum of 6 h (lung-protective mechanical ventilation, hemodynamic measurements, crystalloid and norepinephrine infusion to keep hemodynamic stability), mortality was significantly higher in GRdim/dim mice when compared to GR+/<sup>+</sup> animals (**Supplementary Figure 1**).

#### No Differences in Most Plasma Cytokines 6 h After LPS Challenge and Resuscitation

After LPS-challenge, the concentration of Il-1α in plasma increased in GRdim/dim, but not in GR+/<sup>+</sup> mice. In contrast, concentrations of Il-2, Eotaxin, and Ifnγ only increased in GR+/<sup>+</sup> mice, but not in GRdim/dim mice, however, all are expressed on a low level (**Table 3**). Pro-inflammatory plasma cytokines like Il1-β (**Figure 3A**), Il-6 (**Figure 3B**), and Tnfα (**Figure 3C**) were significantly induced in both genotypes after LPS challenge, demonstrating that the intensive care management and resuscitation does not increase basal level of these inflammatory cytokines.

FIGURE 2 | Mitochondrial respiration (JO2) in tissue from heart (A), liver (B), muscle (C), and brain (D) in GRdim/dim and GR+/<sup>+</sup> mice intraperitoneal challenged with lipopolysaccharide (LPS) or vehicle (PBS). \*p < 0.05 vs. GR+/<sup>+</sup> PBS. #p < 0.05 vs. GRdim/dim PBS. Overall p-values from Kruskal–Wallis-Test are shown above accordingly. LPS = lipopolysaccharide from Escherichia coli [055:B5], 10 mg·kg−<sup>1</sup> , dissolved in 10 µl·g <sup>−</sup><sup>1</sup> phosphate buffered saline [PBS]. GR+/<sup>+</sup> mice challenged with PBS: n = 7–8, GRdim/dim mice challenged with PBS: n = 4–5, GR+/<sup>+</sup> mice challenged with LPS: n = 3–8, GRdim/dim mice challenged with LPS: n = 6–11. Data is presented as median (25th and 75th percentile and minimum/maximum).

#### Lung Function Is Impaired in GRdim/dim Mice During Intensive Care Treatment

Lung compliance as a marker for lung mechanics was reduced in PBS-treated GRdim/dim mice in comparison to PBStreated GR+/<sup>+</sup> animals (**Figure 4A**). A significantly lower lung compliance was observed in GRdim/dim mice compared to GR+/<sup>+</sup> controls during LPS-challenge (**Figure 4A**). In line with a lower lung compliance, GRdim/dim mice had a reduced Horovitz-Index (partial pressure of oxygen in the arterial blood, divided by the inspiratory concentration of oxygen) as a marker for systemic oxygenation (**Table 2**). However, GR+/<sup>+</sup> animals did not show any significant changes in lung compliance or Horovitz-Index after LPS-challenge in comparison to PBS-challenged GR+/<sup>+</sup> animals (**Figure 4A** and **Table 2**), suggesting a more severe lung dysfunction in response to LPS-challenge in GR dimerization impaired mice. The histological evaluation of lung tissue revealed no significant differences in the total score between the corresponding groups, however the GRdim/dim animals had a slight elevated total score (**Table 4**). Immunohistochemistry (IHC) for albumin extravasation showed a higher expression in lung tissue after LPS-challenge in GRdim/dim mice compared to GR+/<sup>+</sup> mice, whereas no difference was observed in extravascular albumin expression between GRdim/dim and GR+/<sup>+</sup> mice after vehicle treatment (**Figure 4B**). In previous studies, increased albumin extravasation and vascular leakage was accompanied by a reduced expression of Sphk1 dependent on the GR dimerization during inflammation (13). In the present study, LPS-challenged GR+/<sup>+</sup> mice showed increased Sphk1 expression TABLE 3 | Concentrations of cytokines in plasma of GRdim/dim and GR+/<sup>+</sup> mice intrapoeritoneal challenged with lipopolysaccharide (LPS) or vehicle (phosphate buffered saline, PBS) measured at the end of the experiment.


\*P < 0.05 vs. GR+/<sup>+</sup> PBS. #P < 0.05 vs. GRdim PBS. PBS = phosphate buffered saline (vehicle), 10 µl·g −1 . LPS = Lipopolysaccharide from Escherichia coli (055:B5), 10 mg·kg−<sup>1</sup> . Data is shown as median (25th and 75th percentile). Bold values indicate significant differences.

that correlated with lower albumin expression compared to GRdim/dim challenged LPS mice having increased extravascular albumin expression and significantly reduced Sphk1 expression (**Figures 4B,C**). Taken together, lung function was impaired in GRdim/dim mice upon LPS-challenge during intensive care treatment. In PBS-treated animals a so far not described slight basal difference in lung function was observed in GRdim/dim mice. These data revealed that the dimerization of the GR is crucial for the lung compliance in an inflammatory setting under intensive care management.

#### Increased Osteopontin in GRdim/dim Mice Might Contribute to Disturbed Lung Function

Inflammatory cytokine mRNA expression of Il1-β (**Figure 5A**), Il-6 (**Figure 5B**), Tnfα (**Figure 5C**), and Il-10 (**Figure 5D**) in the lung revealed no significant differences between GRdim/dim and GR+/<sup>+</sup> mice after challenge with LPS under intensive care management. Therefore, the reduced lung compliance is most likely not a result of the aforementioned inflammatory mediators. Next to the expression of Sphk1, which was identified as an important regulator of lung barrier integrity, we now also aimed at other potential regulators of lung injury. Osteopontin (Opn, secreted phosphoprotein 1–Spp1) is a crucial mediator for inflammatory responses and a regulator of inflammation, especially lung inflammation. Opn neutralizing antibody could protect mice against ALI during sepsis (28). In our setting (intensive care management) Opn protein expression was significantly enhanced in lungs of GRdim/dim compared to GR+/<sup>+</sup> mice, both under PBS-treatment and LPS-challenge (**Figures 6A,B**). Opn was shown to have an impact on type-1 immunity to bacterial infections as OPN deficient mice have increased Il-10 production (29). In accordance with this, GRdim/dim mice with increased levels of Opn showed a trend toward reduced Il-10 mRNA expression in the lung compared to GR+/<sup>+</sup> mice during PBS and LPS-challenge (**Figure 5D**). This suggests that increased Opn expression in the lung of GRdim/dim mice might have an impact on Il-10 mRNA expression. However, in the plasma of GR+/<sup>+</sup> and GRdim/dim PBS- and LPS-challenged animals only a trend to induced Opn expression was observed, due to high variations in the groups (**Supplementary Figure 2A**). To assess the effect of impaired GR dimerization in macrophages and their contribution to Opn expression and inflammatory cytokines bone marrowderived macrophages (BMDM) from GR+/<sup>+</sup> and GRdim/dim animals were stimulated with LPS. LPS stimulation increased the inflammatory cytokine expression in GR+/<sup>+</sup> and GRdim/dim BMDMs, however, no genotype difference for the upregulation of Il-6 expression (**Supplementary Figure 2B**), Tnfα expression (**Supplementary Figure 2C**), and iNos expression (**Supplementary Figure 2D**) could be detected after LPS

treatment. Opn expression was not different, however a slight, but not significant, increase could be detected after LPS treatment in GR+/<sup>+</sup> and GRdim/dim BMDMs with a more pronounced trend in the GRdim/dim BMDMs (**Supplementary Figure 2E**). This observation is supported by the Opn levels in the supernatant of the LPS-treated GRdim/dim BMDMs, that showed a trend to induced Opn compared to GR+/+, however, not significantly different (**Supplementary Figure 2F**). This suggests that macrophages are not the main source of GRdim/dim dependent regulation of Opn, but they contribute to the Opn induction. Moreover, these data suggest that GR dimerization dependent regulation of Opn in vivo might depend on other cells.

In summary, there is possibly a correlation between increased Opn expression in the lung of the GRdim/dim mice that renders them more sensitive during inflammation and therefore, Opn may be a target for the reduced lung compliance.

#### DISCUSSION

In the present study, we tested the hypothesis if an impaired glucocorticoid receptor (GR) function, presented as an impaired GR dimerization (GRdim), would impair organ function in lipopolysaccharide (LPS)-induced systemic inflammation in mice undergoing intensive care treatment to compensate for LPSinduced cardiovascular depression. We found that GRdim/dim mice challenged with LPS had a significantly increased need of norepinephrine to achieve hemodynamic targets and a more

and minimum/maximum).

vehicle (PBS). GR+/<sup>+</sup> mice challenged with PBS: n = 7, GRdim/dim mice challenged with PBS: n = 8, GR+/<sup>+</sup> mice challenged with LPS: n = 9, GRdim/dim mice challenged with LPS: n = 11. Data is presented as median (25th and 75th percentile and minimum/maximum).

pronounced lactic acidosis during resuscitation measures ("lungprotective" ventilation, fluid resuscitation, and norepinephrine treatment) compared to LPS-challenged GR+/<sup>+</sup> mice. Most interestingly, GRdim/dim mice challenged with LPS presented with aggravated ALI, shown by a more pronounced impairment of lung mechanics when compared to LPS-challenged GR+/<sup>+</sup> mice. According to the results of the present study, the impaired lung function in GRdim/dim mice was most likely mediated via an increased endothelial barrier dysfunction, indicated via a reduced expression of sphingosine kinase 1 (Sphk1), which was associated with a higher albumin extravasation in lung tissue. Furthermore, the lung injury in GRdim/dim mice was accompanied by an increase in Osteopontin (Opn) levels in lung tissue, which indicated Opn as a marker of lung injury. The impaired lung



For detailed description of the score for histological tissue analysis see Methods section. \*P < 0.05 vs. GR+/<sup>+</sup> vehicle. PBS = phosphate buffered saline (vehicle), LPS = Lipopolysaccharide from Escherichia coli (055:B5), 10 mg·kg−<sup>1</sup> . Data is shown as median (interquartile range).

function in GRdim/dim mice after LPS-challenge was not mediated via the systemic inflammatory response, because we found no differences in cytokine levels between the genotypes, neither in lung tissue (**Figure 3**) nor in plasma (**Table 3**).

We previously showed that GRdim/dim mice present a higher mortality in LPS-and cecal ligation and puncture (CLP)-induced systemic inflammation without any resuscitation procedures (18). In the present study, after LPS-challenge GRdim/dim mice had (i) an increased hemodynamic instability indicated by a significantly increased need of norepinephrine, (ii) more pronounced lactic acidosis, and (iii) a reduced lung compliance, which altogether lead to an increased mortality rate despite resuscitative measures. The increased mortality of GRdim/dim mice in the current study (**Supplementary Figure 1**) confirms that the GR dimer is of importance for survival after challenge with LPS, even when animals receive fluid resuscitation and norepinephrine treatment to achieve target hemodynamics as well as "lung-protective" mechanical ventilation (22, 23).

i) **Hemodynamic instability**: In the present study, GRdim/dim mice presented with an aggravated hemodynamic instability after LPS-challenge, reflected by a significantly increased need of norepinephrine to reach hemodynamic targets (**Figure 1**). Aggravated hemodynamic instability after LPSchallenge, which also led to increased mortality, has also been reported in mice with an endothelial-specific GR deletion (GRECKO) (30). The increased hemodynamic instability in these mice was accompanied by an increased expression of inducible (iNOS) and endothelial (eNOS) nitric oxide synthase (NOS), which resulted in increased levels of nitric oxide (NO) following LPS-challenge, thus contributing to arterial hypotension. Interestingly, corticosterone levels increased after challenge with LPS, but did not differ between GRECKO and control mice (30). While an endothelial GR dysfunction induces hypotension, in other studies it is reported that a stimulation of an intact GR via GCs leads to hypertension. In addition, this GC-induced hypertension was reported to be mediated via downregulation of eNOS in rats (31). Therefore, endothelial GR dysfunction leads to NO-induced hypotension, whereas GR stimulation via GCs leads to downregulation of NO synthases, which induces hypertension. Although we did not examine any vascular specific effects of an ubiquitous impairment of GR dimerization, it is likely that the increased hemodynamic instability in GRdim/dim mice has been, at least in part, meditated via a NO-induced vasodilation after LPSchallenge in the present study as reported previously (30). The anesthesia and surgery represent a trauma for the animals and this is most likely the reason for the slight increased mortality of the GRdim/dim mice compared to GR+/<sup>+</sup> challenged with PBS. We can only speculate about the reasons for this increased instability. Due to the decreased blood pressure in the GRdim/dim mice during the ICU management, it is most likely that either the systemic vascular resistance or the cardiac output is affected. A decreased hemodynamic instability has already been described for mice with an endothelial glucocorticoid receptor knockout (30).


FIGURE 5 | Measurements of relative mRNA level of cytokines in lung tissue. Relative mRNA level of (A) interleukin 1β (Il-1β), (B) interleukin 6 (Il-6), (C) tumor necrosis factor α (Tnfα), and (D) interleukin 10 (Il-10) in lung tissue of GRdim/dim and GR+/<sup>+</sup> mice intraperitoneally challenged with lipopolysaccharide (LPS) or vehicle (PBS). LPS = lipopolysaccharide from Escherichia coli [055:B5], 10 mg·kg−<sup>1</sup> , dissolved in 10 µl·g <sup>−</sup><sup>1</sup> phosphate buffered saline (PBS). GR+/<sup>+</sup> mice challenged with PBS: n = 8, GRdim/dim mice challenged with PBS: n = 7–8, GR+/<sup>+</sup> mice challenged with LPS: n = 9–10, GRdim/dim mice challenged with LPS: n = 10–11. Data is presented as median (25th and 75th percentile and minimum/maximum).

an attenuated GR dimerization (GRdim/dim), which leads to only partial GR impairment; therefore these mice survive after birth. However, in our model of systemic inflammation via LPS-challenge, GRdim/dim mice presented with impaired lung function, here indicated by a reduced lung compliance (**Figure 4A**). This reduced lung compliance was accompanied by a clear trend toward an increased albumin extravasation in GRdim/dim mice after LPSchallenge (**Figure 4B**), indicating an impaired lung barrier function. Independent of the presence or absence of ICU treatment, we (13) and others (34–36) reported an impaired lung barrier function after LPS-challenge in previous studies. A major mediator of endothelial barrier function in the lung is sphingosine kinase 1 (SphK1). The impaired lung barrier function in our previous study could be linked to a GR dimerization-dependent SphK1 expression in myeloid cells, particularly macrophages, because ablation of the SphK1 gene in the myeloid lineage abolished GC effects on vascular leakage and inflammation (13). In addition, mice with a complete deletion of SphK1 (SphK1−/−) are highly susceptible to LPS-induced ALI and exhibit increased lung vascular leakage (37). In the current study, reduced lung compliance after LPS-challenge was accompanied by a significantly lower Sphk1 expression in GRdim/dim mice (**Figure 4C**), highlighting the contribution of Sphk1 expression to mediate a physiological lung function. In addition, there was no change in lung compliance in GR+/<sup>+</sup> mice despite LPS-challenge but a significant increase of lung Sphk1 expression compared to PBS-treated GR+/<sup>+</sup> mice, which also might suggest a beneficial impact of SphK1

in lung function. However, PBS-treated GRdim/dim mice revealed a trend toward a reduced lung Sphk1 expression, which correlates with the trend, however, not significant, to reduced lung compliance in GRdim/dim animals.

The glucocorticoid receptor is involved in modulating the host response to inflammatory stimuli; therefore, we assessed cytokine mRNA levels in lung tissue (**Figure 5**) as well as cytokine concentrations in the plasma (**Figure 3** and **Table 3**). Interestingly, we did not find any significant genotype difference in systemic inflammatory mediators in the plasma and lung mRNA expression after LPS-challenge. However, in our previous LPS-induced endotoxic shock model without intensive care management we already described that GRdim/dim mice present increased levels of inflammatory mediators at later time points (18). This could explain the difference to our present findings, where mice were observed for only 6 h after LPS-induced endotoxic shock under intensive care management.

Another inflammatory mediator involved in the pathogenesis of inflammatory diseases is osteopontin (Opn) (38, 39), especially in the lung, as patients with various pulmonary diseases revealed increased lung Opn expression (40–45). Furthermore, in experimental models of lung diseases (asthma, lung injury, lung fibrosis), Opn has a detrimental and functional role (40, 46– 50), and, moreover, Opn neutralizing antibody could protect mice against ALI during sepsis (28). We showed that mice with an impaired GR dimerization have elevated Opn protein expression in the lung, the same animals that had a significant decrease in lung compliance after LPS-induced endotoxic shock. The finding that Opn might be a critical regulator of lung compliance is supported by Opn knockout mice showing an increased lung compliance (51). Here we describe, for the first time, that Opn expression in the lung is linked with GR dimerization. Mice lacking GR dimerization had increased Opn expression in the lung, which was accompanied by impaired lung mechanics.

Il-10 is described to be negatively regulated by Opn in LPS-stimulated macrophages, higher Il-10 levels were observed during infections in Opn knockout mice (29). Our data showed a trend to reduced Il-10 expression in the lung of PBS-treated GRdim/dim mice (**Figure 5**), and this effect was enhanced in the inflammatory setting. In addition, GR+/<sup>+</sup> animals revealed a trend to reduced Il-10 expression in the lung upon inflammation, independent of Opn, suggesting an alternative regulation of Il-10 that might be mis-regulated in GRdim/dim animals. Possibly, the duration of the GRdim/dim mice in the intensive care management during systemic inflammation was too short to detect more pronounced effects concerning inflammatory cytokines, because our previous results during systemic inflammation (without intensive care management) revealed significant differences at later time points only (18).

In conclusion, impairment of GR dimerization aggravates systemic hypotension and impairs lung function during LPSinduced endotoxic shock in mice. We now demonstrate that the GR dimer is an important mediator of hemodynamic stability and lung function during LPS-induced systemic inflammation. Further studies are warranted to examine if selective activation of the GR dimer may be able to attenuate lung injury during systemic inflammation.

#### 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 the federal authorities for animal research of the Regierungspräsidium Tübingen, Baden-Wuerttemberg, Germany.

#### AUTHOR CONTRIBUTIONS

SV, MW, EC, PR, and JT conceived and designed the study. SV, MW, JP, TM, CH, UW, OM, JV, SK, MG, MF, EC, and UB performed the experiments and organ analysis. AS and PM examined the histology. SV, MW, JP, TM, UW, and OM analyzed the data and interpreted the results. SV, JP, and MW prepared the figures. SV and MW wrote the manuscript. PR and JT revised the manuscript.

#### FUNDING

PR and JT were supported by the Deutsche Forschungsgemeinschaft (Collaborative Research Center 1149, DFG Tu 220/13-1 to JT). MW was supported by the GEROK program of the Deutsche Forschungsgemeinschaft (Collaborative Research Center 1149), and the University Hospital Ulm. CH

#### REFERENCES


was supported by the Hertha-Nathorff-Fellowship of the Medical Faculty of Ulm University. SV was supported by the Deutsche Forschungsgemeinschaft (Collaborative Research Center 1149), Pulmosens GRK 2203, and the University Hospital Ulm.

#### ACKNOWLEDGMENTS

We are grateful to the staff of the animal facilities of the University of Ulm, our animal welfare officer Dr. Sibylle Ott and in particular to Birgit Widmann.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.03152/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 Wepler, Preuss, Merz, Hartmann, Wachter, McCook, Vogt, Kress, Gröger, Fink, Scheuerle, Möller, Calzia, Burret, Radermacher, Tuckermann and Vettorazzi. 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.

# Short-Chain Alcohols Upregulate GILZ Gene Expression and Attenuate LPS-Induced Septic Immune Response

#### Hang Pong Ng<sup>1</sup> , Scott Jennings <sup>1</sup> , Steve Nelson<sup>2</sup> and Guoshun Wang1,2 \*

*<sup>1</sup> Department of Microbiology, Immunology and Parasitology, Louisiana State University Health Sciences Center, New Orleans, LA, United States, <sup>2</sup> Department of Medicine, Louisiana State University Health Sciences Center, New Orleans, LA, United States*

#### Edited by:

*Lukas Martin, University Hospital RWTH Aachen, Germany*

#### Reviewed by:

*Babak Baban, Augusta University, United States Carlo Riccardi, University of Perugia, Italy Borna Relja, Otto von Guericke University, Germany Emira Ayroldi, University of Perugia, Italy*

> \*Correspondence: *Guoshun Wang gwang@lsuhsc.edu*

#### Specialty section:

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

Received: *08 July 2019* Accepted: *09 January 2020* Published: *03 February 2020*

#### Citation:

*Ng HP, Jennings S, Nelson S and Wang G (2020) Short-Chain Alcohols Upregulate GILZ Gene Expression and Attenuate LPS-Induced Septic Immune Response. Front. Immunol. 11:53. doi: 10.3389/fimmu.2020.00053* Alcohol differentially affects human health, depending on the pattern of exposure. Moderate intake provides beneficial mood modulation and an anti-inflammatory effect, while excessive consumption leads to immunosuppression and various alcohol use disorders. The mechanism underlying this bi-phasic action mode of alcohol has not been clearly defined. Our previous publication demonstrated that ethanol, in the absence of glucocorticoids (GCs), induces expression of Glucocorticoid-Induced Leucine Zipper (GILZ), a key molecule that transduces GC anti-inflammatory effect through a non-canonical activation of glucocorticoid receptor (1). Here we report that similar short-chain alcohols, such as ethanol, propanol and isopropanol, share the same property of upregulating GILZ gene expression, and blunt cell inflammatory response *in vitro*. When mice were exposed to these alcohols, GILZ gene expression in immune cells was augmented in a dose-dependent manner. Monocytes and neutrophils were most affected. The short-chain alcohols suppressed host inflammatory response to lipopolysaccharide (LPS) and significantly reduced LPS-induced mortality. Intriguingly, propanol and isopropanol displayed more potent protection than ethanol at the same dose. Inhibition of ethanol metabolism enhanced the ethanol protective effect, suggesting that it is ethanol, not its derivatives or metabolites, that induces immune suppression. Taken together, short-chain alcohols *per se* upregulate GILZ gene expression and provide immune protection against LPS toxicity, suggesting a potential measure to counter LPS septic shock in a resource limited situation.

Keywords: ethanol, propanol, isopropanol, anti-inflammation, immunosuppression, GILZ, LPS, septic shock

# INTRODUCTION

An alcohol is any organic compound in which a hydroxyl group (−OH) is bound to a carbon atom of an alkyl or substituted alkyl group. In daily life, alcohol usually refers to ethanol, also known as grain alcohol or spirits of wine. Because of its mood modulation property, ethanol is one of the most consumed recreational substances, which often leads to abuse. In medicine, ethanol and its similar short-chain alcohols (propanol and isopropanol) are commonly used as antiseptics and disinfectants. Ethanol has long been known to be anti-inflammatory and immunosuppressive. Moderate ethanol intake is associated with reductions in many adverse health conditions, including coronary artery disease, diabetes, hypertension, congestive heart failure, stroke, arthritis, and dementia (2–4). However, excessive ethanol intake has been clearly linked to organ and tissue damage (4–6) and life-threatening medical disorders (7, 8). Alcohol abuse predisposes individuals to infections by bacteria, fungi, and viruses (4, 9–12) and leads to specific defects in innate and adaptive immunity (7, 13). Acute ethanol exposure, in vitro as well as in vivo, inhibits the production of pro-inflammatory mediators, including TNF-α, IL-1, IL-6, IL-8, and MCP-1 (14), and enhances the production of anti-inflammatory mediators, such as IL-10 (15). Additionally, acute ethanol exposure reduces lipopolysaccharide (LPS)-induced inflammatory response in vivo (16, 17) and protects mice against staphylococcal enterotoxin B (18, 19). Multiple inflammatory networks, including AP-1 and NF-κB, are reported to be involved in alcohol tempering host response to LPS and SEB (20). However, the upstream signaling pathways underlying this alcohol immunosuppressive effect have not been clearly defined.

Sepsis is defined as a life-threatening organ dysfunction caused by a dysregulated host response to infection (21, 22), which frequently manifests an initial hyper-inflammatory phase, reflected by fever, shock, and respiratory failure (23). If patients survive the initial phase and sepsis persists, they enter a phase of immunosuppression (22, 24, 25). Septic shock, a subset of sepsis marked by severe circulatory, cellular, and metabolic abnormalities, is associated with a greater risk of mortality than sepsis alone (21). Septic shock caused by LPS, the major component of the cell wall of Gram-negative bacteria, is a common condition encountered clinically (26). To study the disease process, an animal model often employed is the peritoneal challenge of mice with LPS. Strikingly, there are natural mouse strains that are exceptionally resistant to LPS. For example, SPRET/Ei mice are highly resistant to LPS and Gram-negative bacterial infection (27), while C3H/HeJ and C57BL10/ScCr mouse strains are resistant to LPS, but susceptible to bacterial infection (28). Genetic analyses of both have revealed that the C3H/HeJ and C57BL10/ScCr mice are deficient in Toll-like receptor 4 (TLR4) function. In contrast, the SPRET/Ei mice highly express Glucocorticoid-Induced Leucine Zipper (GILZ), a member of the transforming growth factor-beta (TGF-β) stimulated clone-22 (TSC22) family (29) from the gene located on the X-chromosome (30). GILZ, ubiquitously expressed, is primarily regulated by glucocorticoid receptor (GR) signaling to transduce glucocorticoid (GC) effects (31–34). GILZ is known to regulate cell apoptosis, proliferation and differentiation, and to modulate host immunity and inflammation (35–39). More evidence suggesting the crucial role of GILZ in LPS resistance comes from mice receiving recombinant cell-permeable GILZ protein. The GILZ protein administration leads to increased resistance to LPS and reduced LPS-induced mortality (40). Moreover, overexpression of GILZ protects mice against lethal septic peritonitis (41). Directly related to the current alcohol study, our and others' research indicated that ethanol activates GR signaling in the absence of GCs (42, 43). This activation is through ethanol interplay with the cytoplasmic GR complex, releasing GR without GC coupling. The bare GR enters the nuclei to activate its downstream responsive genes, including GILZ (1), which contributes to ethanol inflammosuppression and immunosuppression.

In the current study, we hypothesized that if ethanol indeed prompts GR-GILZ signaling non-canonically, other short-chain alcohols should share the same effect. To test this hypothesis, we compared ethanol, propanol and isopropanol in their modulation of GILZ expression and their effect on host protection against LPS septic immune response.

# MATERIALS AND METHODS

#### Reagents

Dexamethasone, mifepristone, fomepizole, and common reagents were purchased from Sigma-Aldrich. Lipopolysaccharide (E. coli, serotype O111:B4 L2630) was from List Biological Laboratories (Campbell, CA) or from Sigma-Aldrich. Pure anhydrous ethyl alcohol or ethanol (200 proof/100%, Koptec), propanol (Sigma-Aldrich), and isopropanol (2-propanol, Sigma-Aldrich) were obtained commercially.

#### Cell Culture and Treatments

Human Mono-Mac-6 (MM6) cells were cultured in advanced RPMI-1640 (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS) (HyClone, Logan, Utah), 2 mM GlutaMax (Gibco), 100 U/ml penicillin, 100µg/ml streptomycin, 0.25µg/ml amphotericin B, OPI media supplement, and nonessential amino acids. The cells were incubated in 5% CO<sup>2</sup> at 37◦C. When alcohol and/or LPS were applied, freshly cultured MM6 cells were exposed to ethanol, propanol, or isopropanol at the level of 50 mM for 24 h in a respective alcohol presaturated incubator. LPS (1µg/ml) was added 1 h after alcohol addition, and kept in the system until cell harvest. Cells were pelleted and supernatants were collected for cytokine measurements.

#### Real-Time Quantitative PCR (RT-qPCR) and Immunofluorescent Staining to Measure GILZ Gene Expression in MM6 Cells RT-qPCR of GILZ mRNA

MM6 cells (1.5 × 10<sup>6</sup> ) were exposed to one of the shortchain alcohols (ethanol, propanol or isopropanol) at a 50-mM concentration or 1µM dexamethasone (Dex) for 24 h in the presence or absence of 5µM mifepristone. The cells were harvested and washed twice with 1x PBS. Total RNAs were extracted using the Qiagen RNeasy Kit. The cDNAs were synthesized using the Quantitect Reverse Transcriptase Kit (Qiagen). Human GILZ primers (GILZ-F: 5′ -CATGGAGGTGGCGGTCTATC-3′ and GILZ-R: 5 ′ -CACCTCCTCTCTCACAGCGT-3′ ) and Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) primers (GAPDH-F: 5 ′ -AAGGTCGGAGTCAACGGATTTGGT-3′ and GAPDH-R: 5 ′ - ACAAAGTGGTCGTTGAGGGCAATG-3′ ) were used at a final concentration of 500 nM, as published previously (1). The final reaction for each sample was brought to a total volume

of 20 µl with RT SYBR green qPCR mastermix (Qiagen). All reactions were carried out in duplicate on a CFX96 system (Bio-Rad Laboratories, Hercules, CA) for quantitative real-time PCR (qPCR) detection. The qPCR data were analyzed by the comparative Ct (11CT) method. The expression of GILZ of each treated group was compared to that of GAPDH, and normalized to the non-treatment group.

#### Immunofluorescence Staining of GILZ Protein in MM6 Cells

MM6 cells (1.5 × 10<sup>6</sup> ) were exposed to ethanol, propanol or isopropanol at a final concentration of 50 mM or dexamethasone (Dex, 1µM) for 24 h. The cells were then fixed with 4% paraformaldehyde for 1 h at room temperature. The fixed cells were permeabilized with 0.5% Triton X-100/PBS for 1 h, washed with PBS, and blocked with Blocking Buffer [PBS containing 0.1% Triton X-100, 2% donkey serum and 1% bovine serum albumin (BSA)], for 1 h. GILZ expression was detected by staining with a rabbit anti-GILZ antibody (5µg/ml; Santa Cruz Biotechnology, Dallas, Texas) for 1 h. PE-conjugated F(ab)<sup>2</sup> donkey anti-rabbit IgG (5µg/ml; Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) was used as the secondary antibody. The stained cells were analyzed by flow cytometry.

#### ELISA Measurements of Human Cytokines From MM6 Cells

MM6 cells were exposed to 50 mM ethanol, propanol, or isopropanol for 24 h. LPS (1µg/ml) was added 1 h after alcohol addition, and was kept in the system until cell harvest. After centrifuging to pellet cells, the supernatants were collected for TNF-α and IL-6 production measurement using ELISA (R & D Systems, Minneapolis, MN).

#### Animal Experiments

This animal research was approved by the LSUHSC Institutional Animal Care and Use Committee (IACUC #3578). Adult C57BL/6 mice (7–12 weeks old, mixed sex) were either purchased from The Jackson Laboratory or produced from our breeding colony.

#### Alcohol Exposure and LPS Challenge

Mice were exposed to ethanol (2 or 4 g/kg), propanol (2 g/kg), isopropanol (2 g/kg) or PBS via intraperitoneal injection. One hour later, a lethal dose of LPS (10 mg/kg, i.p.) was injected. Animals were monitored every 2 h for 36 h post LPS injection. Then, the surviving animals were continuously observed once a day for one more week before termination of the experiment. For experiments where alcohol dehydrogenase inhibitor was applied, fomepizole (10 mg/ml) was injected with the alcohol solutions. For serum cytokine measurements, a separate set of animals that received a similar procedure was reserved. The assigned animals at the indicated time points were anesthetized and bled via cardiac puncture. Serum from each animal was obtained and stored at −20◦C until use.

#### Assessment of Alcohol Influence of GILZ Expression in Peripheral Blood Leukocytes

Mice were intraperitoneally injected with dose-escalating ethanol, propanol, or isopropanol at 0, 2, and 4 g/kg, respectively. Peripheral blood was collected 16 h post-injection. After red blood cell lysis, peripheral blood leukocytes were fixed in 4% paraformaldehyde, followed by permeabilization with 0.5% Triton X-100 in PBS for 1 h. After washing with PBS, the cells were blocked with Blocking Buffer (PBS containing 0.1% Triton X-100, 2% donkey serum and 1% BSA) for 1 h. GILZ expression was detected by staining with a rabbit anti-GILZ antibody (5µg/ml; Santa Cruz Biotechnology, Dallas, Texas) for 1 h. PE-conjugated F(ab)<sup>2</sup> donkey anti-rabbit IgG (0.25 mg/ml; Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) was used as the secondary antibody. The stained samples were analyzed by flow cytometry.

To examine alcohol influence of GILZ expression in different types of leukocytes, we selected ethanol as the representative alcohol. Mice were administered (i.p.) with PBS or 4 g/kg ethanol for 8 or 16 h. Under CO<sup>2</sup> anesthesia, blood from each animal was collected via cardiac puncture. After centrifugation, the cell pellet was resuspended in Qiagen red blood cell lysis buffer, and white blood cells (WBCs) were obtained. Next, the WBCs were blocked with TruStain FcTMXPLUS (2.5µg/ml; BioLegend), and subjected to immunostaining with antibodies against CD11b-FITC (5µg/ml; Invitrogen), Ly6G-APC (4µg/ml; BD Pharmingen), CD3e-Alexa 700 (10µg/ml; BD Pharmingen), CD8-Pacific Blue (5µg/ml; BioLegend), CD4-PE-Cy5 (10µg/ml; BD Pharmingen), and CD19-PerCP Cy5.5 (10µg/ml; BD Pharmingen), Next, the cells were permeabilized and fixed using BD Cytofix/CytopermTM Fixation/Permeabilization Kit. Then, the cells were intracellularly stained with GILZ-PE antibody (5µg/ml; Invitrogen), followed by flow cytometry analysis.

#### Assessment of Phospho-IκB Levels in Peripheral Blood Leukocytes

Adult C57BL/6 mice were exposed to LPS (10 mg/kg) alone or combined with 4 g/kg ethanol for 16 h. Under CO<sup>2</sup> anesthesia, cardiac puncture was performed to obtain blood. After plasma and cell separation, the cell component was resuspended in Qiagen red blood cell lysis buffer, and the consequent white blood cells were isolated. Next, the cells were blocked with TruStain FcTMXPLUS (0.25µg/ml; BioLegend), and subjected to immunostaining with the following surface marker antibodies: CD11b-FITC (5µg/ml; Invitrogen), Ly6G-APC (4µg/ml; BD Pharmingen), CD3e-Alexa 700 (10µg/ml; BD Pharmingen), CD8-Pacific Blue (5µg/ml; BioLegend), CD4-PE-Cy5 (10µg/ml; BD Pharmingen), and CD19-PerCP Cy5.5 (10µg/ml; BD Pharmingen). Next, the cells were permeabilized and fixed using the BD Cytofix/CytopermTM Fixation/Permeabilization Kit. Then, the cells were intracellularly stained with phospho-IκB-PE antibody (1.25µg/ml; Invitrogen), followed by flow cytometry analysis.

#### ELISA Measurement of Mouse Cytokines

The collected plasma were diluted appropriately and ELISA was performed to measure the serum level of IL-6, an indicator septic

X-axis measures GILZ-PE staining, and Y-axis indicates sidescatter property. (G) Statistical data. Total 50,000 cells were acquired for each condition. GILZ-positive cells of the parental were expressed. Asterisks denote significant difference as compared to the respective control by Student's *t*-test (*P* < 0.01, *n* = 3 per condition). cytokine, using the mouse Duoset ELISA kits (R & D Systems, Minneapolis, MN).

#### Statistics

Data were statistically analyzed by Student's t-test for differences between two comparing groups. The animal survival data were compared by Log-Rank test. Results were expressed as mean ±SD. Differences with P-values smaller than or equal to 0.05 were considered statistically significant.

#### RESULTS

#### Short-Chain Alcohols Upregulate GILZ Expression in MM6 Cells

Our previous studies demonstrated that ethanol upregulates GILZ gene expression and suppresses LPS-elicited inflammatory response in human airway epithelial cells and MM6 cells (1, 42). As ethanol, propanol, and isopropanol are all short-chain alcohols with a similar molecular structure, we predicted that they share the same property in regulating GILZ expression and cell inflammatory response. To test this prediction, we exposed MM6 cells to the three alcohols, separately, at 50 mM for 24 h. As GILZ is a glucocorticoid- (GC-) responsive gene, we also stimulated the control group of cells with dexamethasone (Dex, 1µM). Furthermore, we previously found that ethanol activates the GILZ gene via a GC-independent non-canonical mechanism (1). A parallel experiment was set with addition of mifepristone (5µM) to block GR. RT-qPCR was performed to measure the GILZ mRNA levels. The results (**Figure 1A**) demonstrate that the three short-chain alcohols significantly enhanced GILZ gene expression, and mifepristone did little to blunt such an effect. In contrast, the Dex-activated GILZ expression was abolished by mifepristone, suggesting that Dex and the shortchain alcohols exploit different mechanisms to activate GILZ. To validate this finding at the protein level, we performed GILZ immunofluorescence staining and flow cytometric analysis. The results (**Figures 1B–G**) show that ethanol, propanol, and isopropanol significantly elevated GILZ protein expression in the exposed cells, compared to the no alcohol control cells. Taken together, these results indicate that short-chain alcohols are capable of upregulating GILZ expression at both transcriptional and translational levels.

#### Short-Chain Alcohols Suppress LPS-Stimulated Inflammatory Response in vitro

To examine if propanol and isopropanol behave like ethanol in suppressing cell inflammatory response to LPS, we exposed MM6 cells to 50 mM ethanol, propanol, or isopropanol, followed by LPS (1µg/ml) stimulation. Levels of TNF-α and IL-6, the two major proinflammatory cytokines in cell response to LPS, in the supernatant of each treatment were determined by ELISA. As shown in **Figure 2**, the LPS-stimulated cells produced high levels of TNF-α and IL-6, which were significantly reduced by each alcohol.

FIGURE 2 | Short-chain alcohols reduce LPS-stimulated TNF-α and IL-6 production by MM6 cells. MM6 cells (1 × 10<sup>6</sup> ) were exposed to 50 mM ethanol, propanol, or isopropanol for 24 hours. LPS exposure (1µg/ml) was added 1 h after the addition of alcohol. A PBS control group of MM6 cells was similarly set and stimulated. Culture media were collected. TNF-α (A) and IL-6 (B) levels were measured by ELISA, and are expressed here as picogram per milliliter per million cells. Asterisks denote statistically significant difference by student's *t*-test (*p* < 0.01, *n* = 4 per condition).

# Short-Chain Alcohols Enhance GILZ Expression in vivo

Short-chain alcohols (ethanol, propanol, or isopropanol) at an escalating dose for each (0, 2, or 4 g/kg) were intraperitoneally administered to adult C57BL/6 mice. Sixteen hours later, white blood cells (WBCs) from each animal were isolated, intracellularly stained for GILZ, and analyzed by flow cytometry. The data (**Figures 3A–H**) demonstrate that GILZ expression in the cells responded to each of the applied alcohols in a dosedependent manner, and was significantly higher than that of the non-alcohol treated control.

To determine which types of immune cells were altered by alcohol in GILZ expression, we similarly exposed a separate set of mice to ethanol (4 g/kg), a representative alcohol, for 8 or 16 h. Cell-surface staining with antibodies against CD11b, Ly6G, CD19, CD3, CD4, and CD8, in combination with intracellular staining of GILZ was performed. Flow cytometry using the gating strategy (**Figure S1**) revealed that GILZ expression in monocytes (CD11b+Ly6G−) was significantly reduced after ethanol exposure for 8 h. However, neutrophils (CD11b+Ly6G+)

FIGURE 4 | GILZ expression in different types of immune cells. Peripheral blood WBCs from mice that had been exposed to 4 g/kg ethanol or PBS for 8 h (A) or 16 h (B) were isolated. Immunostainings for cell surface markers (CD11b, Ly6G, CD19, CD3, CD4, and CD8) and GILZ were performed. Monocytes (CD11b+Ly6G−), neutrophils (CD11b+Ly6G+), B lymphocytes (CD19+), CD4 lymphocytes (CD3+CD4+), and CD8 lymphocytes (CD3+CD8+) were categorized. GILZ-positive cells in each cell type of the parental were compared between the ethanol and control groups. Asterisks indicate statistically significant difference by student's *t*-test (*n* ≥ 3 per group). Mono (Monocytes); Neu (Neutrophils). \**P* ≤ 0.05, and \*\**P* ≤ 0.01.

had significantly higher GILZ expression (**Figure 4A**). Moreover, 16 h ethanol exposure led to significantly higher expression of GILZ in neutrophils (**Figure 4B**). These data suggest that neutrophils are a major cell type in alcohol upregulation of GILZ expression in the current experimental setting.

# Short-Chain Alcohols Protect Mice From LPS Septic Shock

Administration of a lethal dose of LPS elicits an overwhelming inflammatory response that leads to multiple organ failure, shock, and death. As short-chain alcohols effectively suppress inflammatory response to LPS in vitro, we predicted that they should attenuate LPS-induced septic shock in vivo. To test this hypothesis, we administered adult C57BL/7 mice with ethanol (4 g/kg), propanol (2 g/kg), isopropanol (2 g/kg), or PBS control. The reason for selection of a higher dose of ethanol is that our pilot experiments indicated that ethanol at a 2 g/kg dose provided little protection against lethal LPS. One hour after alcohol exposure, the animals were challenged with a lethal dose of LPS (10 mg/kg). A survival curve for each condition was traced, and compared with that of the non-alcohol control (**Figure 5A**). The results show that without alcohol administration, almost all animals died within the time frame of 16–36 h after LPS challenge, while the short-chain alcohol exposures significantly protected the mice from LPS-induced septic shock. No further casualties were observed after 36 h until termination of the experiment a week later.

LPS-induced septic shock in mice is a well-studied and widely used model, which mimics almost all the pathological consequences that occur during sepsis (44). Mortality caused by sepsis or septic shock is associated with overproduction of inflammatory cytokines, also referred to as cytokine storm (45). IL-6, IL-1β and TNF-α are among the major cytokines responsible for sepsis disease pathogenesis. To delineate the potential mechanism underlying the alcohol LPS protection, we measured the serum level of IL-6, a representative cytokine to indicate the severity of sepsis. Adult mice were exposed to the short-chain alcohols similarly, as previously stated, and challenged with the lethal dose of LPS. Sixteen hours post LPS challenge, all live animals were bled to collect their sera for IL-6 cytokine measurement. Data (**Figure 5B**) demonstrate that ethanol (4 g/kg), propanol (2 g/kg) and isopropanol (2 g/kg) significantly reduced IL-6 serum levels. However, ethanol at the 2 g/kg dose had no significant impact on LPSinduced inflammation.

One crucial mechanism for ethanol to attenuate LPS toxicity is through suppression of NF-κB signaling (46, 47). To investigate whether this mechanism was involved in the observed alcoholprotection against LPS in our experimental model, we examined the levels of IκB phosphorylation in WBCs from the animals that had been exposed to ethanol (4 g/kg) and LPS. Cell surface marker and phospho-IκB double-immunostaining was performed, followed by flow cytometry analysis using the gating strategy (**Figure S2**). As compared to the non-alcohol treated cohort, neutrophils and CD4 cells from the alcohol-treated animals had significantly lower phospho-IκB levels (**Figure 5C**), suggesting that the observed alcohol protection against LPS septic shock is, at least partially, through the suppression of NF-κB signaling.

#### Ethanol Instead of Its Derivatives Confers LPS Protection

It was unexpected that ethanol at the 2 g/kg dose failed to suppress LPS inflammation and protect mice from lethal LPS challenge, while the same dose of propanol and isopropanol were effective. We hypothesized that this may result from a greater or faster metabolic rate of ethanol. If this hypothesis is correct, inhibition of ethanol metabolism may enhance ethanol LPS protection. Three groups of adult C57BL/6 mice were assigned, with one group administered with PBS and the alcohol dehydrogenase inhibitor fomepizole (10 mg/kg), another administered with ethanol (2 g/kg) alone, and the third group with ethanol (2 g/kg) and fomepizole (10 mg/kg) together. One hour later, all animals were challenged with the lethal dose of LPS (10 mg/kg). A Kaplan-Meier survival curve for each condition was traced (**Figure 6A**) and statistically compared by

FIGURE 5 | Short-chain alcohol protection against LPS septic shock. (A) Survival curves. C57BL/6 mice were i.p. injected with ethanol (4 g/kg), isopropanol (2 g/kg), propanol (2 g/kg), or PBS. One hour later, the animals were challenged with a lethal dose of LPS (10 mg/kg). The animals were observed for 36 h. The Kaplan-Meier survival curve for each group of animals was traced, and statistical comparisons were performed by Log-rank test. Asterisks denote significant difference in each comparison with PBS group or indicated group (*n* = 10 per group, *p* < 0.01). (B) Short-chain alcohols attenuate host inflammatory response to LPS. C57BL/6 mice were i.p. injected with ethanol (2 g/kg or 4 g/kg), isopropanol (2 g/kg), propanol (2 g/kg), or PBS. One hour later, the animals were challenged with LPS (10 mg/kg). Sixteen hours after LPS challenge, the animals were bled. Serum from each animal was obtained and measured for IL-6 levels. Asterisks denote significant difference in each comparison (*n* = 4 per group, *p* < 0.01). (C) Ethanol exposure significantly affects cellular Phospho-IκB levels in monocytes and neutrophils. Peripheral blood WBCs were isolated from the mice that had been exposed to 4 g/kg ethanol and LPS (10 mg/kg) or the control mice that had been exposed to PBS and LPS (10 mg/kg). Immunostainings for cell surface markers (CD11b, Ly6G, CD19, CD3, CD4, and CD8) and phospho-IκB were performed. Monocytes (CD11b+Ly6G−), neutrophils (CD11b+Ly6G+), B lymphocytes (CD19+), CD4 lymphocytes (CD3+CD4+), and CD8 lymphocytes (CD3+CD8+) were categorized. Percentage of the phospho-IκB-positive cells in each cell type of the gated population were compared between the ethanol and control groups. Asterisks denote statistically significant difference by student's *t*-test (*p* < 0.01, *n* ≥ 3 per group).

Log-Rank test. The results demonstrate that without fomepizole, 2 g/kg ethanol had no protective effect on LPS septic shock. However, fomepizole significantly improved the protection efficacy of ethanol at the otherwise non-protective concentration. To further investigate whether fomepizole enhances ethanol suppression of host inflammatory response to LPS, we performed a parallel experiment with the same design. Sixteen hours later after LPS challenge, the serum obtained from each animal was measured to determine the indicator cytokine IL-6 level. As displayed in **Figure 6B**, fomepizole significantly reduced the serum IL-6 levels, indicating enhancement of ethanol suppression of the host inflammatory response to LPS. Taken together, the data indicate that inhibiting ethanol metabolism facilitates ethanol protection against LPS septic shock, strongly suggesting that it is the molecular ethanol instead of its derivatives or metabolites that engenders the protective effect against LPS-induced septic immune response.

#### DISCUSSIONS

The legal blood ethanol concentration limit for driving in the United States is 0.08%, which is equivalent to 17.36 mM. However, blood ethanol levels can reach to over 87 mM in patients with acute alcohol intoxication (48, 49). A previous study from our group (50) documented that acute alcohol intoxication in mice by intraperitoneal (i.p.) injection of 20% alcohol in pyrogen-free saline at a dose of 5 g/kg gave rise to blood alcohol levels of 119.7 ± 1.3, 106.3 ± 1.5, 87.7 ± 3.6, and 48.4 ± 3.5 mM, respectively, at 45 min, 90 min, 3 h, and 6 h post alcohol administration. In the current study, the highest dose of ethanol used in vivo was 4 g/kg. Thus, the blood ethanol level should be lower than the levels referenced. Isopropanol is widely used in household applications. Deliberate or accidental ingestion of isopropanol ranks second as a cause of alcohol poisoning clinically (51). Blood isopropanol concentrations have been reported as high as 560 mg/dl (93 mM) (52). In the current paper, we used 2 g/kg isopropanol for i.p. administration. A previous publication reported that injection of mice with 2 g/kg isopropanol generates a blood alcohol concentration of 200 mg/dl (33 mM) after 30 min (19). Thus, the alcohol doses used in this study should be relevant to clinically encounterable alcohol levels.

Glucocorticoids (GCs) are steroid hormones produced by the adrenal cortex under control of the hypothalamic–pituitary– adrenal (HPA) axis in response to internal circadian clock and external stress challenge (53, 54). GCs are the most prescribed anti-inflammatory drugs. The profound effectiveness of GCs provides the rationale for their use in a wide range of autoimmune, inflammatory, and allergic diseases, such as rheumatoid arthritis, lupus erythematosus, inflammatory bowel disease, transplant rejection and asthma (55, 56). However, longterm application of these steroids results in detrimental side effects, including diabetes, immunosuppression, osteoporosis and increased risk of cardiovascular events, all of which are closely associated with the alteration of physiological metabolism by GCs (57). Thus, new anti-inflammatory drugs are urgently needed, ideally ones that maintain the effectiveness of GCs while avoiding the GC-associated detrimental effects. Bypassing GCs to activate GILZ represents a novel strategy to achieve

anti-inflammation. In the current study, we found that shortchain alcohols (ethanol, propanol and isopropanol) upregulate GILZ without GCs, which deserves further investigation to explore if the alcohols can serve as prototype compounds to search for new anti-inflammatory agents.

Our previous genome-wide gene expression analysis on human airway epithelial cells that were exposed to dose-escalating ethanol (42) revealed that a cluster of glucocorticoid-targeting genes, including TSC22D3 (GILZ), ALOX15B, SYNPO2, and PTEN, responded in a dose-dependent manner. GILZ was the most affected, upregulated 2-fold by 50 mM ethanol and 3-fold by 100 mM ethanol (42). Importantly, GILZ is an essential molecule to convey the alcohol antiinflammatory effect, as knockdown of GILZ diminishes alcohol suppression of LPS-induced inflammatory response (1). Our further research revealed that ethanol activation of the GILZ gene is through a non-canonical activation of the GR signaling pathway, which is independent of GCs (1). In the current study, we found that other similar short-chain alcohols (propanol and isopropanol) share the same property of modulating GILZ expression in vitro and in vivo, validating the novel alcohol-GR interaction. It is well studied that ethanol modulates the immune function of T cells, monocytes, macrophages, dendritic cells and neutrophils (4, 58–60), and the specific effects depend on the pattern of ethanol exposure (acute or chronic) (61). Acute ethanol decreases TLR responses and attenuates proinflammatory cytokine production (62, 63). However, chronic ethanol exposure renders monocytes and macrophages more responsive to LPS stimulation. Mechanistic studies demonstrate that the ethanol-induced LPS tolerance or sensitization is mediated through modulation of IRAK-M, IRAK1/4, Bcl-3, and NF-κB (14, 47). In contrast, the isopropanol-induced effect is conveyed through the regulation of discrete members of the NFAT family of transcription factors and AP-1 family of transcription factors (18, 19). In the current study, we demonstrate that short-chain alcohols modulate GILZ gene expression and suppress IκB phosphorylation, which adds another layer of regulation to the known mechanisms. Even though GILZ is known to interact with the key inflammatory signaling mediators NF-κB and AP-1 (32, 64, 65), the finding that short-chain alcohols exploit this mechanism for inflammosuppression and immunosuppression is novel. A recent publication reported that ethanol and other short-chain alcohols inhibit NLRP3 inflammasome activation through protein tyrosine phosphatase stimulation (66). It is noteworthy that stimulation of a functional inflammasome requires two steps. The first step is priming during which activation of NF-κB is essential to induction of several components of the inflammasome. Our data have shown that short-chain alcohols suppress IκB phosphorylation, which inevitably undermines the priming and the downstream activation of the inflammasome. It will be very interesting to investigate whether alcohol-induced protein tyrosine phosphatase stimulation subdues phospho-IκB production in future studies.

Despite intensive research, sepsis continues to be a major health problem world-wide. The incidence of sepsis in the past two decades has annually increased by 9%, to reach 240 per 100,000 people in the US (67, 68). This rate of occurrence translates into ∼750,000 cases and over 250,000 deaths each year (68). When septic shock develops, the mortality rate of patients is substantially increased. At present, there is no specific treatment for sepsis and septic shock. Clinical management basically focuses on infection containment and organ function support (69). Alcohol attenuation of LPS-induced septic shock, as we demonstrated in this report, may be employed as an emergency measure to save lives under the circumstance of no medical care available. With regard to potential use of alcohol for therapy, recent studies have proposed to use ethanol to treat traumatic brain injury in humans (51, 52). Acute ethanol gavage attenuates hemorrhage/resuscitation-induced injury (70). Thus, alcohol, the oldest drug in medicine, may find new applications, as long as the molecular mechanism for its action is clearly understood.

While this is related to but beyond the scope of the current research, alcohol activation of GR signaling and upregulation of GILZ expression may be important in explaining alcohol-associated psycho-behavioral problems. Alcohol is long known to be an emotion regulatory agent. Moderate intake relieves stresses and produces pleasure, while heavy drinking induces mood and psychological abnormalities, such as depression. GILZ over-expression is found to be associated with depression (71, 72). Our data clearly demonstrated that short-chain alcohols upregulate GILZ, which may serve as a critical mechanism for alcohol mood regulation and alcohol-precipitated depression.

There are several limitations associated with this research. First, only acute and one-dose application of alcohols was tested. Multiple doses may have a stronger potency in immunosuppression against LPS. Second, this study was to prove the principle. Alcohols were applied prior to LPS challenge. However, in a clinical setting LPS septic shock typically occurs first. Future studies will test the clinically relevant mode. Third, in this report alcohols were administered intraperitoneally. Alternatively, gavage through a feeding tube is another clinically applicable way to deliver alcohols, which will be tested in the future.

In summary, this report demonstrated that short-chain alcohols upregulate GILZ expression, suppress host immune response to LPS, and attenuate LPS-triggered septic shock. This finding implies that short-chain alcohols can be used to alleviate LPS sepsis as an emergency measure if no other medicines are available.

#### DATA AVAILABILITY STATEMENT

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

#### REFERENCES


#### ETHICS STATEMENT

The animal study was reviewed and approved by the Institutional Animal Care and Use Committee of Louisiana State University Health Sciences Center with IACUC #3578.

#### AUTHOR CONTRIBUTIONS

HN and SJ performed experiments and data analyses. SN contributed to the original concept and design of the work. GW designed and conducted experiments, performed data analyses, and did manuscript writing.

#### FUNDING

This work was supported by the research grant to GW from the National Institutes of Health (AA024549).

#### ACKNOWLEDGMENTS

The authors would like to thank Connie Porretta at LSUHSC Flow Cytometry Core Facility for her technical assistance in flow cytometry analyses. We also would like to thank Dianne Wellems for her critical reading of this manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.00053/full#supplementary-material


the alteration of discrete transcriptional pathways in activated lymphocytes. J Immunol. (2008) 181:2348–55. doi: 10.4049/jimmunol.181.4.2348


the context of inflammation resolution. J Immunol. (2015) 194:4940–50. doi: 10.4049/jimmunol.1401722


**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 Ng, Jennings, Nelson and Wang. 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.

# Hepatocyte-Specific Deletion of AMPKα1 Results in Worse Outcomes in Mice Subjected to Sepsis in a Sex-Specific Manner

Satoshi Kikuchi <sup>1</sup> , Giovanna Piraino<sup>2</sup> , Michael O'Connor <sup>2</sup> , Vivian Wolfe<sup>2</sup> , Kiana Ridings <sup>2</sup> , Patrick Lahni <sup>2</sup> and Basilia Zingarelli 2,3 \*

*<sup>1</sup> Department of Emergency Medicine, Ehime University, Toon, Japan, <sup>2</sup> Division of Critical Care Medicine, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States, <sup>3</sup> Department of Pediatrics, College of Medicine, University of Cincinnati, Cincinnati, OH, United States*

Alterations in the energy homeostasis contribute to sepsis-mediated multiple organ

#### Edited by:

*Valentin A. Pavlov, Northwell Health, United States*

#### Reviewed by:

*Charles E. McCall, Wake Forest Baptist Medical Center, United States Qun Sophia Zang, UT Southwestern Medical Center, United States*

> \*Correspondence: *Basilia Zingarelli basilia.zingarelli@cchmc.org*

#### Specialty section:

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

Received: *11 November 2019* Accepted: *27 January 2020* Published: *13 February 2020*

#### Citation:

*Kikuchi S, Piraino G, O'Connor M, Wolfe V, Ridings K, Lahni P and Zingarelli B (2020) Hepatocyte-Specific Deletion of AMPK*α*1 Results in Worse Outcomes in Mice Subjected to Sepsis in a Sex-Specific Manner. Front. Immunol. 11:210. doi: 10.3389/fimmu.2020.00210* failure. The liver plays a central role in metabolism and participates to the innate immune and inflammatory responses of sepsis. Several clinical and experimental studies have suggested that females are less susceptible to the adverse outcome of sepsis. However, underlying mechanisms of organ damage in sepsis remain largely undefined. AMP-activated protein kinase (AMPK) is an important regulator of mitochondrial quality control. The AMPK catalytic α1 isoform is abundantly expressed in the liver. Here, we determined the role of hepatocyte AMPKα1 in sepsis by using hepatocyte-specific AMPKα1 knockout mice (H-AMPKα1 KO) generated with Cre-recombinase expression under the control of the albumin promoter. Using a clinically relevant model of polymicrobial sepsis by cecal ligation and puncture (CLP), we observed that male H-AMPKα1 KO mice had higher plasma levels of tumor necrosis factor-α and interleukin-6 and exhibited a more severe liver and lung injury than male H-AMPKα1 WT mice, as evaluated by histology and neutrophil infiltration at 18 h after CLP. Plasma levels of interleukin-10 and the keratinocyte-derived chemokine were similarly elevated in both KO and WT male mice. At transmission electron microscopy analysis, male H-AMPKα1 KO mice exhibited higher liver mitochondrial damage, which was associated with a significant decrease in liver ATP levels when compared to WT mice at 18 h after sepsis. Mortality rate was significantly higher in the male H-AMPKα1 KO group (91%) when compared to WT mice (60%) at 7 days after CLP. Female H-AMPKα1 WT mice exhibited a similar degree of histological liver and lung injury, but significantly milder liver mitochondrial damage and higher autophagy when compared to male WT mice after CLP. Interestingly, H-AMPKα1 KO female mice had lower organ neutrophil infiltration, lower liver mitochondrial damage and lower levels of cytokines than WT female mice. There was no significant difference in survival rate between WT and KO mice in the female group. In conclusion, our study demonstrates that AMPKα1 is a crucial hepatoprotective enzyme during sepsis. Furthermore, our results suggest that AMPK-dependent liver metabolic functions may influence the susceptibility to multiple organ injury in a sex-dependent manner.

Keywords: AMPKα1, Cre-lox, cecal ligation and puncture, mitochondria, lung injury, liver injury, female sex

#### INTRODUCTION

Sepsis is a life-threating organ dysfunction caused by dysregulated host responses to infection (1). Sepsis is the most common cause of patient mortality in intensive care units, with a global incidence of ∼18 million cases per year and a mortality rate of 28–40% (2). Although co-morbidities contribute to the clinical variability, numerous experimental and clinical studies indicate that sex-specific differences influence the susceptibility to sepsis and the subsequent multiple organ dysfunction syndrome (MODS) and mortality (3–5).

The liver plays a central role in metabolic and immunological homeostasis (6). Clinical studies have shown that liver dysfunction and failure are serious complication in sepsis and directly contributes to disease progression and death (6, 7).

Mitochondrial dysfunction has been proposed as an important cause of sepsis-related organ failure in sepsis (8, 9). AMP-activated protein kinase (AMPK) is a crucial sensor of energy status and contributes to several metabolic processes for energy homeostasis. This kinase consists of a catalytic α-subunit (α1 and α2) and two β and γ regulating subunits, which are allosterically activated by low levels of adenosine triphosphate (ATP) and high levels of adenosine monophosphate (AMP) (10–12). A key component of the metabolic effects of AMPK is the activation of mitochondrial biogenesis leading to improved cellular energy utilization (10). AMPK also contributes to activation of autophagy, a highly conserved catabolic process that degrades and recycles dysfunctional cytoplasmic constituents, including damaged mitochondria, thus ensuring a proper process of mitochondrial turnover (10, 13).

We have recently demonstrated that specific pharmacological activators of AMPK exert beneficial effects in sepsis and reduce hepatic, cardiac and pulmonary injury in experimental models using male mice (14–16). Specifically, in the liver, these beneficial effects are associated with amelioration of mitochondrial biogenesis and function (14).

Given the potential benefit for AMPK activators to attenuate sepsis-induced liver injury, further investigation is merited to determine the biologic role of AMPK within the liver. As AMPKα1 is abundantly expressed in the liver of rodents and the predominant isoform in human hepatocytes (11, 12, 17), we sought to investigate the role of the hepatocyte AMPKα1 on sepsis-induced liver injury and mortality by employing hepatocyte-specific AMPKα1 knockout (H-AMPKα1 KO) young male mice (18). In order to characterize the sexual dimorphism of liver injury, we also used young H-AMPKα1 KO female mice. In these loss-of function studies, we demonstrated that specific AMPKα1 gene deletion in hepatocytes was associated with increased susceptibility to sepsis-induced liver and lung injury and increased mortality in male mice only. On the contrary, hepatocyte-specific AMPKα1 deletion in female mice had a surprising protective effect in liver mitochondrial structure, lung injury and systemic inflammatory response. Thus, our data suggest that hepatocyte AMPKα1 is an important modulator of the metabolic response in sepsis; however, its function is sex-dependent.

### MATERIALS AND METHODS

#### Murine Model of Polymicrobial Sepsis

The investigation conformed to the National Institutes of Health Guide for the Care and Use of Laboratory Animals (Eighth Edition, 2011) and was approved by the Institutional Animal Care and Use Committee of the Cincinnati Children's Hospital Medical Center. Specific deletion of AMPKα1 in hepatocyte was achieved by using Cre-lox technology. Mice expressing Cre recombinase under the control of the albumin promoter [B6.Cg-Tg(Alb-cre)21Mgn/J] and AMPKα1flox/flox mice (Prkaa1tm1.1Sjm/J), both on C57BL/6 genetic background, were obtained from Jackson Laboratories (Bar Harbor, Maine) and were crossed to generate hepatocyte-specific AMPKα1 (H-AMPKα1) KO mice. Mice were housed in pathogen-free conditions under a 10 h light/14 h dark cycle with free access to food and water. Both male and female mice were used at 8–12 weeks of age. Mice were anesthetized with 2.0% isoflurane in 50% oxygen and polymicrobial sepsis was induced by cecal ligation and puncture (CLP) (19). A midline laparotomy was performed. After opening the abdomen, the cecum was exteriorized, ligated and punctured twice with a 23-G needle. The cecum was then returned into the peritoneal cavity and the abdominal incision was closed. Mice were resuscitated subcutaneously with 35 ml/kg 5% dextrose solution immediately after and at 3 h after the surgical procedure. Control mice did not undergo any surgical procedure. Mice were then sacrificed at 18 h after CLP. Blood, liver, and lungs were collected for biochemical assays.

#### Survival Study

In a separate study, another cohort of mice (n = 20–23) was subjected to a milder model of CLP by puncture with 25-G needle and was used for assessing survival rate. Mice received fluid resuscitation (35 ml/kg normal saline with 5% dextrose subcutaneously) immediately after, at 3 h and every 24 h after the CLP procedure up to 72 h. To minimize pain at the surgical incision site, lidocaine hydrochloride (1%, 4 mg/kg total dose) was applied locally immediately after the procedure and every 12 h up to 48 h. Survival was monitored for 7 days.

#### Myeloperoxidase Activity

Myeloperoxidase (MPO) activity was measured as an indicator of neutrophil infiltration in lung and liver tissue (20). Tissues were homogenized in a solution containing 0.5% hexa-decyltrimethyl-ammonium bromide dissolved in 10 mM potassium phosphate buffer (pH 7.0) and centrifuged for 30 min at 4,000 × g at 4◦C. An aliquot of the supernatant was allowed to react with a solution of tetra-methyl-benzidine (1.6 mM) and hydrogen peroxide (0.1 mM). The rate of change in absorbance was measured by spectrophotometry at 650 nm. MPO activity was defined as the quantity of enzyme degrading 1 µmol of hydrogen peroxide/min at 37◦C and expressed in units per 100 mg weight of tissue.

#### Histopathologic Analysis

Frozen liver sections and paraffin-embedded lung sections were stained with hematoxylin and eosin, and evaluated by two independent observers blinded to the treatment groups. Liver injury was evaluated on the presence of necrosis, sinusoid congestion and infiltration of red blood and inflammatory cells. Lung injury was evaluated on the presence of alveolar capillary congestion, infiltration of red blood cells and inflammatory cells into the airspace, alveolar wall thickness, and hyaline membrane formation.

#### Plasma Alanine Aminotransferase (ALT)

Plasma levels of ALT was evaluated as index of liver function by an enzymatic assay kit (Sekisui Diagnostics, Charlottetown, Canada) using the protocols recommended by the manufacturer.

#### Measurement of ATP Levels

Mitochondrial function in the liver was assessed by measuring ATP levels. Homogenates were obtained from fresh livers and were deproteinized with perchloric acid using a Deproteinization Sample Preparation Kit (BioVision, San Francisco, CA). Liver ATP levels were measured using an ATP Fluorometric Assay Kit (BioVision, San Francisco, CA). ATP levels were expressed as nmol/g tissue weight.

#### Transmission Electron Microscopy

Liver samples were fixed in 3% glutaraldehyde, postfixed in 1% osmium tetroxide in sodium phosphate buffer, and cut with ultramicrotome. Samples were stained with 2% uranyl acetate and lead citrate. The sections were viewed and photographed on Hitachi H-7650 transmission electron microscope at 120 kV. The total number of mitochondria and autophagosomes, and the presence of abnormal or enlarged mitochondria with loose matrix, fragmented cristae and membranes were determined in 9 consecutive cells in four different sections for each animal by using NIH ImageJ analysis (21).

#### Cytosol Extraction and Western Blot Analysis

Livers were homogenized in a buffer containing 0.32 M sucrose, 10 mM Tris-HCl (pH 7.4), 1 mM EGTA, 2 mM EDTA, 5 mM NaN3, 10 mM β-mercaptoethanol, 20µM leupeptin, 0.15µM pepstatin A, 0.2 mM phenylmethanesulfonyl fluoride, 50 mM NaF, 1 mM sodium orthovanadate, and 0.4 nM microcystin. Samples were centrifuged at 1,000 × g for 10 min at 4◦C and the supernatants collected as cytosol extracts. Cytosol content of the light-chain (LC)3B-I and LC3B-II was determined by immunoblot analyses. Extracts were boiled in equal volumes of NuPAGE <sup>R</sup> LDS Sample Buffer (4X) and 40µg of protein loaded per lane on a 16% Tris-glycine gradient gel. Proteins were separated electrophoretically and transferred to nitrocellulose membranes. For immunoblotting, membranes were blocked with 5% non-fat dried milk in Tris-buffered saline (TBS) for 1 h and incubated with primary antibodies for 24 h. Membranes were washed in TBS with 0.1% Tween 20 and incubated with secondary peroxidase-conjugated antibody; the immunoreaction was visualized by chemiluminescence and x-ray. Membranes were also re-probed with primary antibody against GADPH to ensure equal loading for cytosol proteins. Densitometric analysis of blots was performed using Quantity One (Bio-Rad Laboratories, Des Plaines, IL, USA).

#### Plasma Levels of Cytokines

Plasma levels of tumor necrosis factor-α (TNFα), interleukin-10 (IL-10), interleukin-6 (IL-6), and keratinocyte-derived chemokine (KC) were evaluated by a commercially available multiplex array system (Milliplex, Millipore Corporation, Billerica, MA) using the protocols recommended by the manufacturer.

#### Materials

The primary antibodies directed at LC3B-I and LC3B-II were obtained from Cell Signaling (Beverly, MA); the primary antibody directed at GADPH was obtained from Abcam (Cambridge, MA); the secondary antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Unless otherwise stated, all other chemicals were obtained from Sigma-Aldrich (St. Louis, MO).

#### Statistical Analysis

Statistical analysis was performed using SigmaPlot 13.0 (Systat Software, San Jose, CA, USA). Data in figures and text are expressed means ± SEM of n observations (n = 4– 8 animals for each group). The results were examined by analysis of variance followed by the Student-Newman-Keuls's correction post hoc t-test. The Gehan-Breslow and log-rank tests were used to compare differences in survival rates (n = 20–23 animals for each group). A value of P < 0.05 was considered significant.

#### RESULTS

#### Hepatocyte-Specific Deficiency of AMPKα1 Results in Liver Damage After Sepsis in a Sex-Independent Manner

At 18 h after CLP, both male and female WT mice exhibited liver damage with modest areas of necrosis and sinusoid congestion at histological examination. However, H-AMPKα1 KO mice of both sexes showed more prominent liver damage with significant necrosis, edema and infiltration of inflammatory cells (**Figures 1A–H**). To confirm the histological findings of liver neutrophil infiltration in KO mice after sepsis we measured the activity of MPO, a neutrophil lysosomal enzyme. At 18 h after CLP, there was no increase in liver MPO activity in male or female WT mice. However, both male and female KO mice exhibited a significant elevation of MPO after sepsis when compared to sex-matched WT septic mice (**Figure 1I**). To further quantify liver injury, we measured plasma levels of ALT. Both male and female WT mice exhibited a similar degree of plasma ALT levels at 18 h after CLP. However, male and female KO mice had significant higher levels of ALT after sepsis when compared to sex-matched WT septic mice, thus confirming a more severe liver injury in a sex-independent manner (**Figure 1J**).

# Hepatocyte-Specific Deficiency of AMPKα1 Results in Liver Mitochondrial Damage After Sepsis in a Sex-Dependent Manner

At electron microscopic analysis, mild mitochondria damage was evident in both male and female H-AMPKα1 WT mice at 18 h after CLP and was characterized by a few swollen mitochondria (**Figure 2**). H-AMPKα1 WT mice of both sexes also exhibited an increased number of autophagosomes and autolysosomes with sequestrated materials when compared to basal levels of control mice. There was also a significant increase of elongated mitochondrial morphology in H-AMPKα1 WT mice of both sexes after CLP. On the contrary, male H-AMPKα1 KO mice exhibited a significantly higher structural damage of mitochondria at 18 h after CLP when compared to male WT mice. Structural damage was characterized by the presence of swollen organelles with distorted cristae, translucent matrix and disrupted membrane. Interestingly, mitochondrial damage was significantly lower in KO female mice when compared to WT female mice and KO male mice after CLP; whereas there was no difference in the number of elongated organelles between WT and KO mice of both sexes. H-AMPKα1 WT and KO mice also exhibited an increased number of autophagosomes and autolysosomes with sequestrated materials when compared to basal levels of control mice. However, WT female mice had a higher number of autophagosomes and autolysosomes when compared to WT male mice after sepsis. Interestingly, number of autophagosomes was significantly lower in female KO mice when compared to sex-matched WT mice and to male KO mice after sepsis. To determine whether hepatocyte-specific deficiency of AMPKα1 might affect energy homeostasis, we measured ATP levels in the liver. There was no change in ATP levels after sepsis in male or female WT mice when compared to baseline conditions of WT sex-matched control animals (**Figure 2L**). However, male KO experienced a significant decrease in ATP levels at 18 h after CLP when compared to baseline conditions of KO control animals. Interestingly, there were no changes in

FIGURE 2 | Transmission electron microscopy of hepatocytes in hepatocyte-specific (H-AMPKα1) wild-type (WT) and knock-out (KO) male (A–D) and female (E–H) mice at basal control condition or 18 h after cecal ligation and puncture (CLP). Arrows, damaged mitochondria presenting translucent matrix, disrupted membrane and cristae; arrow heads, elongated mitochondria; asterisk, autophagic vescicles packed with mitochondria; N, nucleus. Quantification of damaged mitochondria (I), elongated mitochondria (J), and autophagosomes (K) of hepatocytes of hepatocyte-specific (H- AMPKα1) wild-type (WT) and knock-out (KO) mice at 18 h after cecal ligation and puncture (CLP). Liver sections were examined at transmission electron microscopy. Damaged, elongated mitochondria and autophagosomes were determined by using the NIH Image J software and expressed as percentage of total number of mitochondria in nine consecutive cells. Data are means ± SEM of 3–4 mice for each group. Liver ATP content (L) of hepatocyte-specific (H-AMPKα1) wild-type (WT) and knock-out (KO) mice at 18 h after cecal ligation and puncture (CLP). Data represents the mean ± SEM of 4–6 mice for each group. \**P* < 0.05 vs. sex-matched control mice; #*P* < 0.05 vs. sex-matched WT mice; ‡*P* < 0.05 vs. male group of the same genotype.

ATP content after sepsis in female KO mice. Thus, hepatocytespecific AMPKα1 deficiency promotes mitochondrial damage in a sex-dependent manner.

#### Hepatocyte-Specific Deficiency of AMPKα1 Influences Liver Autophagy After Sepsis in a Sex-Dependent Manner

To confirm the number of autophagosomes observed at electron microscopy, we further quantified the process of autophagy in the liver by evaluating the expression of the light chain 3B protein (LC3B) (**Figure 3**). The protein is converted from a cytosolic LC3B-I form to a conjugated LC3B-II form in the autophagosomal membrane and correlates with autophagic vesicle formation (22). At 18 h after CLP, LC3B-II/LC3B-I ratio significantly increased in both male and female WT mice when compared to baseline content of sexmatched control mice. However, WT female mice had a higher ratio when compared to WT male mice after sepsis, further confirming the higher number of autophagosomes seen at electron microscopy. Interestingly, female KO mice exhibited a higher expression of LC3BI when compared to male KO mice after sepsis. Consequently, LC3B-II/LC3B-I ratio was significantly lower in female KO mice when compared to sex-matched WT mice and to male KO mice after sepsis. There was no difference in LC3B-II/LC3B-I ratio between WT and KO mice in the male group. Thus, hepatocytespecific AMPKα1 deficiency influences autophagy in a sexdependent manner.

# Hepatocyte-Specific Deficiency of AMPKα1 Results in Lung Injury After Sepsis in a Sex- Dependent Manner

To obtain insight into the role of hepatocyte AMPKα1 in the development of multiple organ failure we also evaluated lung injury by histology. At 18 h after CLP, both male and female H-AMPKα1 WT mice exhibited similar lung damage with modest infiltration of inflammatory cells and reduced alveolar space (**Figure 4**). However, male H-AMPKα1 KO mice exhibited marked lung injury characterized by reduced alveolar space, alveolar and bronchial congestion and accumulation of red and inflammatory cells when compared to male WT mice (**Figures 4A–D**). Male KO mice also exhibited higher lung MPO when compared to WT male mice (**Figure 4I**). Interestingly, the H-AMPKα1 KO female mice had lower MPO levels than WT female mice after CLP. Also, lung injury was milder (**Figures 4E–H**) and tissue MPO (**Figure 4I**) levels were significantly lower in KO female mice when compared to KO male mice after CLP. Thus, hepatocyte-specific AMPKα1 deficiency influenced the inflammatory response in the lung in a sex-dependent manner.

#### Hepatocyte-Specific Deficiency of AMPKα1 Influences Systemic Production of Cytokines in a Sex-Dependent Manner

To determine whether hepatocyte AMPKα1 influenced the systemic production of cytokines during sepsis, plasma levels

hepatocyte-specific (H-AMPKα1) wild-type (WT) and knock-out (KO) male (A–D) and female (E–H) mice at baseline conditions (Control) or 18 h after cecal ligation and puncture (CLP). Lung damage in H-AMPKα1 WT male (B) and female (F) mice with modest infiltration of inflammatory cells (*arrow heads*) and reduced alveolar space (*arrows*). Lung damage in H- AMPKα1 KO male mice (D) with reduced alveolar space, bronchial congestion (*asterisk*) and accumulation of red and inflammatory cells. Mild lung damage in H-AMPKα1 KO female mice (H) with mild infiltration of inflammatory cells only.

*(Continued)*

FIGURE 4 | Magnification ×100; scale bar = 200µm. A similar pattern was seen in *n* = 4–8 different tissue sections in each experimental group. Lung myeloperoxidase (MPO) activity (I) in male and female H- AMPKα1 WT and KO mice at 18 h after CLP. Data represents the mean ± SEM of 4–8 mice for each group. \**P* < 0.05 vs. sex-matched control mice; #*P* < 0.05 vs. sex-matched WT mice; ‡*P* < 0.05 vs. male group of the same genotype.

of TNFα, IL-6, IL-10, and KC were measured. At 18 h after CLP, both male and female WT mice exhibited a significant elevation in plasma levels of all cytokines compared with sexmatched control mice. However, male KO mice had significantly higher levels of TNFα and IL-6 after sepsis when compared to male WT septic mice (**Figures 5A,B**), while levels of IL-10 and KC were similar in both WT and KO male mice (**Figures 5C,D**). Interestingly, plasma levels of cytokines were significantly lower in KO female mice when compared to WT female mice and KO male mice after CLP (**Figure 5**). Thus, hepatocyte-specific AMPKα1 deficiency influenced the systemic inflammatory response in a sex-dependent manner.

#### Hepatocyte-Specific Deficiency of AMPKα1 Exacerbates Sepsis-Induced Mortality in Male Mice Only

To confirm the protective role of hepatocyte AMPKα1 in longterm outcomes of sepsis, we performed a model of CLP with low mortality at 48 h after CLP (<20% mortality) but delayed high mortality between 3 and 7 days after CLP. Male H-AMPKα1 KO mice experienced higher mortality (91%) than WT mice (60%, P < 0.05) at 7 days after CLP. However, this effect was not observed in female H-AMPKα1 KO mice, which had similar mortality rate as female WT mice (67% in KO mice and 76% in WT mice) (**Figure 6**).

# DISCUSSION

Being responsible of important physiological functions, such as detoxification, energy production, hormonal and immune balance, and coagulation, the liver is a critical organ for host survival following severe injury, including sepsis (6). Among sepsis patients, liver dysfunction or failure is associated with mortality rates ranging from 54 to 68%, higher than the mortality rates of sepsis patients with respiratory system dysfunction or failure (23–25). Despite these data of poor prognosis, the pathophysiology of liver dysfunction and its potential role in influencing multiple organ failure remain unclear. In the present study we have demonstrated that hepatocyte AMPKα1 is a key regulator of liver metabolic and innate immune function. Specifically, we have provided evidence that hepatocyte-specific AMPKα1 deletion worsened liver and lung injury and exacerbated the lethal effects of sepsis in young male mice. The underlying cellular mechanisms of this excessive vulnerability to injury related to increased neutrophil infiltration in liver and lung and increased systemic production

control mice; #*P* < 0.05 vs. sex-matched WT mice; ‡*P* < 0.05 vs. male group of the same genotype.

of TNFα and IL-6. However, the most intriguing observation made here was that, in young female mice with hepatocytespecific AMPKα1 deletion, the signs of liver injury did not coincide with worsening of sepsis. On the contrary, hepatocytespecific AMPKα1 deficiency in female mice was associated with protective effects in the lung and reduced systemic inflammatory response without affecting survival. Thus, our data demonstrate that AMPKα1 modulates the metabolic and inflammatory responses in the liver in a sex-dependent manner.

AMPK is the master regulator of diverse metabolic events to maintain energy homeostasis. The protein consists of a catalytic α-subunit and two regulatory β- and γ-subunits. Two isoforms have been identified of the α-subunit (1 and 2) and β-subunit (1 and 2) and three isoforms of the γ-subunit (1, 2, and 3). Both AMPKα1 and α2-containing complexes are present in the liver in equal distribution in rodent models (11, 12, 18), but AMPKα1 is the predominant form in human hepatocytes (17). In previous studies, we have demonstrated that pharmacological activation of AMPK ameliorated liver injury in young and mature male animals subjected to CLP by regulating autophagy and gene transcription of mitochondrial structural and transport proteins, and metabolic enzymes (14). In this current study, despite sharing some histological and biomarkers similarities of liver injury, male and female H- AMPKα1 KO mice exhibited quite different liver mitochondrial features, as KO female mice did not exhibit ultrastructural damage compared with WT littermates. On the contrary, H-AMPKα1 KO female mice had the least mitochondrial damage and maintained normal ATP production among all groups of septic mice. In regard to this discrepancy between histological injury and changes in mitochondrial damage, it must be noted that such conflicting results have also been reported in clinical studies. Autopsy studies revealed a discordance between histologic findings and the degree of organ dysfunction of patients who died of sepsis (26). Cell death in the heart, kidney, liver, and lung was relatively minor and did not reflect the clinical evidence of more profound organ dysfunction (26). Furthermore, examination of post mortem organs from septic patients identified abnormalities of mitochondrial structures in cardiomyocytes, but cardiomyocyte cell death was rarely observed despite the magnitude of organ dysfunction (27). Taken together, our data suggest that maintenance of metabolic stability is associated with resilience to deleterious effects of sepsis and further confirm that mitochondrial damage is the major pathophysiologic characteristic of organ failure as seen in human sepsis.

Clinical studies have reported increased autophagy in multiple organs and tissues in sepsis (28). AMPK induces autophagy

as determined by Gehan-Breslow analysis.

by inhibiting the mammalian target of rapamycin pathway (10, 13). Autophagy is a critical step in maintaining mitochondrial quality control by selective removal of damaged organelles (13). There is experimental evidence that disruption of autophagy is associated with organ dysfunction. For example, using a young rat model, Chien et al. demonstrated that liver autophagy occurred early during sepsis, but it declined at later time points, when it was associated with liver dysfunction (29). On the contrary, AMPK-mediated autophagy has been shown to attenuate mitochondrial dysfunction in endotoxin stimulated hepatocytes (30). Mitochondrial morphology is also an adaptive response to cellular metabolic demands and an elongated pattern of mitochondria has been associated to mechanisms of fission/fusion to maximize ATP production (31). However, male H-AMPKα1 WT had less autophagosomes than WT female mice. At examination of molecular events, the conversion of LC3B from its free form (LC3B-I) to its phosphatidylethanol-amine-conjugated form (LC3B-II) is a required step in autophagosome formation. Therefore, the LC3B-II/LC3B-I can be considered as a reliable marker to quantify autophagy (22). In our study, the LC3B-II/LC3B-I ratio was higher after sepsis in the liver of H-AMPKα1 WT female mice, thus suggesting a better capability to mount an autophagic event in response to cellular stress. Interestingly, female KO mice exhibited a reduction of autophagosomes when compared to WT littermates after sepsis. This event was also consistent with lower number of altered mitochondria in female KO mice when compared to male KO mice, thus suggesting that mitochondrial damage is an important requisite for upregulation of autophagy. Therefore, our data further suggest that AMPKα1 influences organ function in a sex-dependent manner also through finely tuning of autophagy according to the degree of mitochondrial impairment.

In concert with Kupffer cells, other immune competent and non-hematopoietic resident cells, hepatocytes produce a complex cytokine milieu of both pro-inflammatory and antiinflammatory cytokines, which contribute to the initiation and amplification of the systemic acute phase response (32, 33). In our study, we observed that AMPKα1 deficiency in male mice enhanced systemic production of TNFα and IL-6, thus suggesting that AMPKα1 modulates the acute phase response. Consistent with our observation, it has been demonstrated that pharmacological activation of AMPK suppressed immune stimulated inflammation in mouse and human hepatocytes (34). Interestingly, in our study the specific inactivation of AMPKα1 in the liver in male mice led to an exacerbation of lung injury, thus suggesting that liver metabolic disturbance affects distant organ function. Therefore, the results presented here, employing a mouse model for genetic specific inactivation of AMPKα1 in hepatocytes, further extend the evidence for the liver as a key driver of immune and inflammatory responses during sepsis and strongly validate AMPK as a potential therapeutic target.

Paradoxically, specific deficiency of AMPKα1 in hepatocytes in female mice led to a remarkable reduction in the systemic inflammatory response and protected against sepsis-induced lung injury. Because AMPKα1 deficiency was not accompanied with liver metabolic defects in female mice, these sex-dependent differences point to other liver compensatory mechanisms independent by AMPKα1 in young female. In this regard, it is possible that estrogen may exert a compensatory effect for AMPKα1 deletion. For example, previous studies reported that double global genetic deficiency of AMPKα1 and AMPKα2 in female mice did not affect the inflammatory process of osteoarthritis; whereas estrogen prevented articular cartilage destruction (35, 36). Since in our study male and female H-AMPKα1 WT mice exhibited similar organ injury and outcomes after sepsis, we speculate that the differential hormonal milieu between male and female may, indeed, result in a diversification of mitochondria quality control during acute stress only in condition of dysregulation of AMPKα1 and the level of expression per se of this enzyme is most likely a key element for sex-dependent switch of metabolic regulation. Further comprehensive studies and elucidations of mitochondrial bioenergetics are required to better understand the signaling pathways and molecules involved in this switch.

#### CONCLUSIONS AND LIMITATION

Our findings are based on cross-sectional data comparing young age-matched male and female mice, future longitudinal studies are needed to disentangle age and sexual maturation effects, including the impact of ovariectomy and orchiectomy, steroid hormonal levels and the hepatoprotective effects of estrogen during sepsis. We used male and female mice at the age of 8–12 weeks old, which are in their peak reproductive age (37). Therefore, their comparison with older groups of mice is necessary to establish association with levels of estrogens and decline of sexual maturation. Furthermore, a broad investigation of bioenergetic and biosynthetic mitochondrial functionalities by omics technologies should be included to identify novel sex-related mitochondrial pathomechanisms. Nevertheless, by means of cell type-specific gene knockout technology, the current study provides conclusive data that AMPKα1 expression in hepatocytes functions as a protective factor against sepsis in young male mice. However, paradoxically findings of our study point out that sex-dependent differences exist in the susceptibility to sepsis in the context of AMPKα1 deficiency. As AMPK activators, such as metformin, are widely used for diabetes type II in the clinical arena (38), further insight into the exact biological mechanism of AMPK may help to provide better clarity as to the potential pharmacological targeting of AMPK in sepsis.

#### DATA AVAILABILITY STATEMENT

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

#### REFERENCES


#### ETHICS STATEMENT

The animal study was reviewed and approved by the Institutional Animal Care and Use Committee of the Cincinnati Children's Hospital Medical Center.

#### AUTHOR CONTRIBUTIONS

SK and BZ conceived and designed the projects. SK and VW performed the animal experiments. SK, GP, PL, and MO'C performed the biochemical assays. SK, GP, and KR performed the electron microscopy analysis. SK, GP, and BZ analyzed the data and prepared graphics. SK and BZ wrote the draft of the manuscript. All authors reviewed and approved the final manuscript.

#### FUNDING

This work was supported by the National Institutes of Health grants R01 GM-067202 and GM-115973 to BZ, and in part by grant P30 DK-078392 of the Digestive Research Core Center (Integrative Morphology Core).


signalling in mouse liver and in human hepatocytes. Mol Cell Endocrinol. (2013) 375:68–78. doi: 10.1016/j.mce.2013.05.013


**Disclaimer:** The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes 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 Kikuchi, Piraino, O'Connor, Wolfe, Ridings, Lahni and Zingarelli. 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.

# Mangiferin Attenuates LPS/D-GalN-Induced Acute Liver Injury by Promoting HO-1 in Kupffer Cells

Sen Yang1†, Ge Kuang1†, Liangke Zhang<sup>1</sup> , Shengwang Wu<sup>2</sup> , Zizuo Zhao<sup>3</sup> , Bin Wang<sup>3</sup> , Xinru Yin<sup>4</sup> , Xia Gong<sup>2</sup> \* and Jingyuan Wan<sup>1</sup> \*

*<sup>1</sup> Chongqing Key Laboratory of Biochemistry and Molecular Pharmacology, Chongqing Medical University, Chongqing, China, <sup>2</sup> Department of Anatomy, Chongqing Medical University, Chongqing, China, <sup>3</sup> Department of Anesthesiology, The First Affiliated Hospital of Chongqing Medical University, Chongqing, China, <sup>4</sup> Department of Gastroenterology, Institute of Surgery Research, Daping Hospital, Third Military Medical University, Chongqing, China*

#### Edited by:

*Christoph Thiemermann, Queen Mary University of London, United Kingdom*

#### Reviewed by:

*Yongsheng Li, Army Medical University, China Enrico Calzia, University of Ulm, Germany*

#### \*Correspondence:

*Xia Gong xiagong@cqmu.edu.cn Jingyuan Wan jywan@cqmu.edu.cn*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *05 June 2019* Accepted: *04 February 2020* Published: *25 February 2020*

#### Citation:

*Yang S, Kuang G, Zhang L, Wu S, Zhao Z, Wang B, Yin X, Gong X and Wan J (2020) Mangiferin Attenuates LPS/D-GalN-Induced Acute Liver Injury by Promoting HO-1 in Kupffer Cells. Front. Immunol. 11:285. doi: 10.3389/fimmu.2020.00285* Acute liver injury and its terminal phase, hepatic failure, trigger a series of complications, including hepatic encephalopathy, systematic inflammatory response syndrome, and multiorgan failure, with relatively high morbidity and mortality. Liver transplantation is the ultimate intervention, but the shortage of donor organs has limited clinical success. Mangiferin (MF), a xanthone glucoside, has been reported to have excellent anti-inflammatory efficacy. Here, a lipopolysaccharide (LPS)/D-galactosamine (D-GalN)-induced acute liver injury mouse model was established to investigate the protective role of MF and the underlying mechanisms of action. Pretreatment with MF improved survival, decreased serum aminotransferase activities, and inhibited hepatic TNF-α production in LPS/D-GalN-challenged mice. Through Kupffer cell (KC) deletion by GdCl<sup>3</sup> and KC adoptive transfer, KCs were confirmed to be involved in these beneficial effects of MF. MF reduced LPS-mediated TNF-α production via the suppression of the TLR4/NF-κB signaling pathway *in vitro*. MF promoted HO-1 expression, but the knockdown of HO-1 prevented TNF-α inhibition, suggesting that the damage-resistance effects of HO-1 occurred via the suppression of TNF-α synthesis. When HO-1-silenced KCs were transferred to the liver with KC deletion, the protective effect of MF against LPS/D-GalN-induced acute liver injury was reduced, illustrating the role of KC-derived HO-1 in the anti-injury effects of MF. Collectively, MF attenuated acute liver injury induced by LPS/D-GalN via the inhibition of TNF-α production by promoting KCs to upregulate HO-1 expression.

Keywords: mangiferin, heme oxygenase-1 (HO-1), acute liver injury, lipopolysaccharide/D-galactosamine (LPS/D-GalN), TNF-α

# INTRODUCTION

The liver is the most important metabolic organ and plays a pivotal role in the host's defensive response owing to its ability to scavenge pathogenic microorganisms and toxins (1). However, the liver is also the main victim of such attacks, resulting in the activation of host immune cells, which incites inflammation (2). The dysregulated inflammatory response not only impairs the defensive

**258**

function of liver but also induces massive necrosis of hepatocytes, triggering acute liver injury and, ultimately, provoking fulminant hepatic failure (FHF) (3, 4). As the incidence of FHF is associated with high mortality and there is no recognized treatment, the terminal stage of FHF should be prevented. Thus, there is an urgent need for new therapeutic agents for the clinical treatment of acute liver injury (5).

Lipopolysaccharide, a component of Gram-negative bacteria, can induce acute liver injury with high lethality in D-galactosamine (D-GalN)-sensitized mice (6). This wellestablished experimental murine model, developed by Galanos et al. (7), is distinguished by a dramatic activation of Kupffer cells (KCs), the resident macrophages of the liver tissue (8), which can produce many pro-inflammatory mediators in the liver (9). Among these mediators, TNF-α, the terminal effector in the pathogenesis of acute liver injury, initiates extensive hepatocyte apoptosis and massive hepatocyte necrosis, prompting the explosion of inflammation. Considering that the model exactly resembles the pathogenesis of human clinical FHF, it is employed widely as an experimental animal model to investigate the underlying mechanism of clinical FHF and develop effective therapeutic strategies for endotoxin challenge (10).

Heme oxygenase-1 (HO-1), a cellular stress protein belonging to the heme oxygenase enzyme system, is part of the rate-limiting step in the degradation of heme to carbon monoxide (CO), iron, and biliverdin-IXa (BV); all of these are involved in various pathophysiological processes (11). It has manifold biological activities, such as anti-oxidative stress, anti-inflammatory, and anti-apoptosis activity (12, 13). HO-1, a downstream molecule in multiple signaling pathways, is induced and responds to pathological challenges from various diverse stimuli (14). Recent studies have demonstrated that expression of HO-1 is upregulated in acute liver injury and that the increase in HO-1 activity alleviated the toxic action of LPS in the liver (15). In particular, the stimulation of macrophage-derived HO-1 appears to significantly attenuate liver injury, which has caused researchers to recognize its critical role (16).

Mangiferin, a xanthone glucoside, exhibits a broad spectrum of pharmacological properties, including antioxidant, antiviral, antitumor, anti-inflammatory, and gene-regulatory effects (17, 18). A growing number of studies have reported that MF exerts potent anti-oxidative and anti-inflammatory effects against the acute inflammation of multiple organs, such as the lung and kidney and the cardiovascular system (19–21). The attenuation of liver injury by MF via multiple pathways has been previously reported, illustrating its prominent role in hepatic inflammation (22, 23). With respect to the signaling pathways concerning antiinflammation, Nrf2/HO-1 exerts its defense against injury by abating oxidative stress, mitigating the inflammatory response, and preventing cell death (24). Pan et al. (25) reported that MF alleviated LPS/D-GalN-induced acute liver injury, coupled with the activation of the Nrf2 signaling pathway. The evidence encouraged us to explore whether the underlying target of MF is closely associated with KC-derived HO-1 in LPS-D-GalNmediated fulminant hepatic damage. Overall, the study aimed to search for effective molecule targets of MF against FHF, to elucidate the protective mechanism of MF, and to demonstrate the involvement of HO-1. The study has screened for a potential antidote to LPS/D-GalN-induced acute liver injury and has provided novel insights into the target and mechanisms of MF.

#### MATERIALS AND METHODS

#### Animals

BALB/c mice, weighing 18–22 g, were obtained from the Experimental Animal Center of Chongqing Medical University. The animals were kept in individual wire-bottomed cages. All animals were fed a standard laboratory diet and water ad libitum and maintained in a controlled environment (20–25◦C, 50% ± 5% relative humidity, 12 h dark/light cycle). The animals were acclimatized to the experimental conditions for at least l week before use in experiments. All experimental procedures involving animals were approved by the Animal Care and Use Committee of Chongqing Medical University.

#### Reagents

MF (C19H18O11, FW = 422.34, purity ≥ 95%) was purchased from Nanjing ZeLang Medical Technology Co. Ltd. (Nanjing, China). LPS (Escherichia coli, 0111: B4), D-GalN, and ZnPP IX were obtained from Sigma (St. Louis, MO, USA). ALT and AST detection kits were purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). The bicinchoninic acid (BCA) protein assay kit was purchased from Pierce (Rockford, IL, USA). Trizol reagent was purchased from Invitrogen (Grand Island, NY, USA). AMV transcriptase, RNasin, SYBR green PCR Master Mix, and the Dual-Luciferase Reporter Assay kit were obtained from Promega (Madison, WI, USA). The tumor necrosis factor (TNF)-α ELISA kit was obtained from Bender MedSystems (Vienna, Austria). Rabbit HO-1 antibody was obtained from Abcam (Cambridge, MA, UK), rabbit β-actin antibody was purchased from Cell Signaling Technology (Boston, MA, USA), and the rat anti-mouse TLR4/MD-2 complex and rat IgG2a isotype control PE were obtained from eBioscience (San Diego, CA, USA). Lipofectamine 2000 was obtained from Invitrogen (Carlsbad, CA, USA).

# Kupffer Cell Isolation and Culture

Liver-resident Kupffer cells were isolated from mice. Briefly, liver samples were perfused with calcium (Ca2+)- and magnesium (Mg2+)-free Hanks' balanced salt solution (HBSS) containing 0.5 mM EGTA, 10 mM HEPES, and 4.2 mM NaHCO3, Type I collagenase (0.05%), and trypsin inhibitor (50µg/mL) for 10– 20 min. The liver samples were removed from the solution and transferred into a sterile tube, minced into small pieces by using scissors, and placed in RPMI 1640 medium supplemented with 5% fetal bovine serum (FBS) and 0.5 mg/mL collagenase at 37◦C for 15 min. Then, the liver cell suspensions were filtered through a sterile nylon mesh and placed into fresh, calcium-free medium and subsequently centrifuged at 250 rpm for 15 min. The supernatant was discarded, and the cell pellets were resuspended in RPMI with 5% FBS and separated by centrifugation on a 25–50% Percoll gradient. The Kupffer cells in the interface

were seeded in RPMI 1640 containing 10% FBS, 100µM βmercaptoethanol, 10µg/mL insulin, 100µg/mL streptomycin, and 100 U/mL penicillin at 37◦C in an atmosphere of 5% CO2.

#### Experimental Protocols

FHF was induced in mice by the injection of LPS (10 µg/kg) and D-GalN (700 mg/kg). The mice were gavaged orally with PBS (Control) or MF (30, 100, or 150 mg**/**kg, respectively) twice, at 1 and 7 h prior to LPS/D-GalN injection. ZnPP IX (45 mg/kg) was administered intraperitoneally (i.p.) 30 min before LPS/D-GalN. The survival rate was evaluated for up to 48 h after LPS/D-GalN administration. The serum and liver were collected at the indicated time point for the analysis. For KC depletion, 25 mg/kg GdCl<sup>3</sup> was intravenously (i.v.) injected into mice at 24 h before challenge or cell transfer. In the adoptive transfer experiment, 2 × 10<sup>6</sup> donor KCs were i.v.-infused into the tail vein of the recipient mouse in the 24 h before LPS challenge.

In the experiment with KCs, primary KCs were cultured in FBS-free RMPI 1640 and pretreated with PBS or MF (10−<sup>5</sup> M) for 30 min, and then PBS or LPS (100 ng/mL) was added. The supernatants and cells were collected at 30 min or 3 h after LPS for the subsequent analyses.

#### ALT and AST Measurement

Blood was collected from the retro-orbital vein of the eye of mice 6 h after the LPS/D-GalN injection. The blood samples were centrifuged at 3,000 rpm for 10 min. The serum was separated, transferred to other tubes, and frozen at −20◦C. ALT and AST activities were assayed by using commercially available kits in accordance with the manufacturer's recommendations.

#### Cytokine Analysis

Hepatic tissues and cell supernatants were assayed for TNF-α level by ELISA in accordance with the manufacturer's protocol. Briefly, samples or standards were pipetted into a microplate pre-coated with a monoclonal antibody specific for mouse TNFα. After any unbound substances had been washed away, an enzyme-linked polyclonal antibody specific for mouse TNF-α was added to each well. After any unbound antibody-enzyme reagent had been washed away, a substrate solution was added to the wells. The enzyme reaction yielded a blue product that turned yellow when the stop solution was added. The intensity of the color was in proportion to the amount of mouse TNF-α bound in the initial step, and the concentration of the sample values were then read from the standard curve.

# Histopathology Analysis

The liver tissue samples from mice were excised and fixed in 10% formaldehyde at 25◦C and then embedded in paraffin. Serial paraffin sections (4µm) were stained with hematoxylineosin for conventional morphological evaluation by using a light microscope (Olympus, Tokyo, Japan).

#### Reverse Transcription-Quantitative Polymerase Chain Reaction (RT-qPCR)

Total RNA was isolated from the liver and cell samples by using Trizol reagent in accordance with the manufacturer's protocol. First-strand complementary DNA (cDNA) was synthesized. The cDNA samples were then incubated at 90◦C for 7 min to stop the reaction. Quantitative PCR was performed with SYBR Green PCR Master Mix in accordance with the protocol. The primers were used to amplify TNF-α cDNA (sense, 5′ -TTC TCA TTC CTG CTT GTG GC-3; antisense, 5′ -GTT TGC TAC GAC GTG GGC TA-3′ ), HO-1 cDNA (sense, 5′ -GAG ATA GAG CGC AAC AAG CAG-3′ ; antisense, 5′ -CTT GAC CTC AGG TGT CAT CTC-3′ ), GAPDH cDNA (sense, 5′ -CCT GCA CCA CCA ACT GCT TA-3′ ; antisense, 5′ -TCA TGA GCC CTT CCA CAA TG-3 ′ ). The generation of specific PCR products was confirmed by melting curve analysis. Relative HO-1 mRNA expression was normalized to GAPDH expression. Data were calculated by using the 2−11Ct method as described by the manufacturer and were expressed as fold increase over the indicated controls.

# Western Blotting Analysis

The protein contents of liver and cell samples were prepared by using a Protein Extraction Kit (20 mM Tris, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM Na3VO4, 1 mM β-glycerophosphate, 1µg/mL leupeptin, and aprotinin). Protein concentrations were determined by using the BCA protein assay kit. Protein extracts (40 µg) were fractionated on 12% polyacrylamide-sodium dodecyl sulfate (SDS) gel and then transferred to nitrocellulose membranes. Non-specific binding to the membrane was blocked by incubation in 5% (w/v) fat-free milk in Tris-buffered saline (TBS) containing 0.05% Tween-20, followed by incubation with a rabbit primary polyclonal antibody at 4◦C overnight. The membrane was then treated with horseradish peroxidaseconjugated goat anti-rabbit secondary antibody. Antibody binding was visualized by using a chemiluminescence system and brief exposure of the membrane to X-ray film (Kodak, Japan).

#### Flow Cytometry Analysis

KCs were collected, washed with fluorescence-activated cell sorting (FACS) buffer, and blocked with the anti-FcR antibody for 15 min. The cells were then stained by using rat antimouse TLR4/MD-2 complex PE or rat IgG2a isotype control PE in the dark at room temperature. Cells were incubated for 30 min, washed twice, and resuspended in 0.5 mL FACS buffer. Fluorescence was measured by flow cytometry and analyzed by using FlowJo software (Tree Star, Inc., Ashland, OR).

#### Transfection of Reporter Plasmids and Small Interference RNA (siRNA)

KCs were seeded at 1 × 10<sup>6</sup> cells per well in 24-well plates and allowed to rest for 1 day before transfection. Cells were transiently transfected with pAP1-Luc or pNF-κB-Luc and the pRL-TK Renilla plasmid by using Lipofectamine 2000. The activities of both firefly and Renilla luciferases were measured by using the Dual-Luciferase Reporter Assay System according to the manufacturer's instructions. To silence murine HO-1 in primary KCs, pre-confirmed HO-1 specific siRNA and non-targeting control siRNAs were transfected into freshly isolated primary KCs by using Lipofectamine 2000. At 24 h after transfection, the cells were used for subsequent experiments.

RT-qPCR and ELISA at 1.5 h after LPS/D-GalN challenge. Data are presented as mean ± SD. *n* = 10 (for survival rate analysis) or 6; \**P* < 0.05 and \*\**P* < 0.01.

# HO-1 Activity Assay

HO enzymatic activity was measured by bilirubin generation, as described previously (26). Briefly, frozen hepatic and cell samples were homogenized in lysis buffer (250 mM Tris·HCl, pH 7.4; 150 mM NaCl; 250 mM sucrose; 0.5 mM PMSF; 1 µg/µL leupeptin; 1 µg/µL aprotinin). The microsomal fraction was obtained by successive centrifugation and washing with 0.15 M KCl, followed by centrifugation (105,000 × g for 30 min). The pellet was solubilized in 0.1 M potassium phosphate by sonication and stored at −80◦C. The reaction was performed in a mixture containing 2–3 mg/mL protein microsomal fraction, 1 mM glucose-6-phosphate, 0.2 units/mL glucose-6-phosphate dehydrogenase, 0.8 mM NADPH, and 0.025 mg/mL hemin at 37◦C for 45 min. After chloroform extraction, the amount of extracted bilirubin was calculated by the difference in absorbance at 464 and 530 nm.

#### Statistical Analysis

The study data were expressed as mean ± standard deviation (S.D.). The differences between two groups were calculated by Student's t-test, and comparisons of multiple groups were evaluated by one-way analysis of variance (ANOVA), followed by the Tukey post hoc test. The survival rates were determined using Kaplan-Meier curves with log-rank tests. P < 0.05 were considered to be statistically significant.

# RESULTS

#### MF Improved Survival and Pathological Liver Injury Induced by LPS/D-GalN in Mice

First, we examined the survival rates induced by LPS/D-GalN in MF-treated and untreated mice. As expected, LPS/D-GalN resulted in high lethality, with 100% mortality occurring within 36 h. Pre-treatment of mice with MF improved survival rates in a dose-dependent manner; 70% of the mice survived to the end of the 48-h observation period in the 150 mg/kg MF-treated group (**Figure 1A**). With regard to the association of lethality with hepatic damage, we analyzed the liver enzymes and pathological changes in the serum aminotransferases and tissue sections. Serum concentrations of ALT/AST, which are released into the blood upon damage to liver cells by LPS/D-GalN, were markedly lower in MF-treated mice (**Figure 1B**). Similarly, there were clear

improvements in the pathology of the liver after MF treatment (**Figure 1C**). These results indicated that MF exerts protective activity against LPS/D-GalN-induced mortality and liver injury.

#### MF Alleviated Hepatic Expression of TNF-α mRNA and Protein Induced by LPS/D-GalN in Mice

After confirmation of the critical role of TNF-α in endotoxininduced acute liver injury, we attempted to determine whether MF affected LPS-induced TNF-α production. As shown, the expressions of hepatic TNF-α mRNA and protein was sharply increased in LPS/D-GalN-stimulated mice compared to in control mice; in contrast, the increased expression of TNF-α was dose-dependently decreased by MF treatment (**Figures 1D,E**, p < 0.05), suggesting that MF could inhibit the expression increases of TNF-α mRNA and protein induced by LPS/D-GalN.

#### MF Promoted Hepatic HO-1 Expression and Activity in LPS/D-GalN-Challenged Mice

HO-1 is an inducible heat shock protein that participates in cytoprotection. It has been shown that HO-1 exerts a protective effect against LPS/D-GalN-induced FHF. Therefore, we examined whether MF could modify the expression and activity of HO-1. Intriguingly, the expression and activity of HO-1 in hepatic tissues were significantly upregulated by MF treatment compared to those in LPS/D-GalN-treated mice (**Figures 2A–C**).

# HO-1 Is Responsible for the Effect of MF on LPS/D-GalN-Induced Liver Injury in Mice

ZnPP-IX, a widely used HO-1 inhibitor, could suppress HO-1 activity via the competitive conjugation with the porphyrin ring structure owing to its similarity to heme (27, 28). Hence, LPS/D-GalN-treated mice treated with or without MF were exposed to ZnPP-IX. As shown in **Figure 3A**, ZnPP-IX markedly blunted the improved survival of MF in LPS/D-GalN-treated mice. Likewise, ZnPP-IX blocked the effects of MF on the LPS/D-GalNinduced increases in ALT/AST activities and TNF-α production (**Figures 3B–D**).

#### KCs Are Involved in the Protective Effect of MF on LPS-Mediated Acute Liver Injury in D-GalN-Sensitized Mice

As hepatic TNF-α is produced mainly in inflammatory cells, especially in KCs, we then investigated whether KCs contributed to these effects. We injected mice with GdCl<sup>3</sup> to selectively deplete KCs in accordance with the previous study (29) and found that this almost completely abrogated LPS/D-GalNinduced mortality and hepatic damage, as indicated by improved survival and decreased serum liver enzymes (**Figures 4A,B**). Meanwhile, hepatic TNF-α mRNA and protein expression were also sharply diminished (**Figures 4C,D**). In contrast, adoptive transfer of KCs to GdCl3-treated mice restored these effects of the depletion of KCs, suggesting that KCs are required for LPS/D-GalN-induced acute liver injury and TNF-α production. In

keeping with those findings, KC depletion could also abolish the protection of MF on LPS/D-GalN-mediated acute liver injury; adoptive transfer of KCs restored the protective effect of MF on LPS/D-GalN-mediated acute liver injury and TNF-α production (**Figure 4**). Collectively, these data provided compelling evidence that KCs were involved not only in LPS/D-GalN-induced acute liver injury but also in the protective effects of MF on FHF induced by LPS/D-GalN.

### KCs Are Involved in MF-Mediated HO-1 Upregulation in LPS-D-GalN-Treated Mice

To assess whether HO-1 upregulation of MF was involved in KCs, we first used GdCl<sup>3</sup> to remove KCs in MF-treated LPS-D-GalN-induced mice. As expected, in **Figure 5**, GdCl<sup>3</sup> can be seen to have markedly abrogated MF-upregulated HO-1

FIGURE 3 | Inhibition of HO-1 removed the protective effect of MF against LPS/D-GalN-induced acute liver injury. Mice were pretreated orally with MF (150 mg/kg) at 1 and 7 h with or without ZnPP IX (40 mg/kg) i.p. 30 min prior to LPS/D-GalN challenge. (A) Survival rates of mice were monitored for 48 h after LPS/D-GalN challenge. (B) Serum ALT and AST activity were assessed at 6 h after LPS/D-GalN challenge. Hepatic TNF-α mRNA (C) and protein (D) expression were determined by RT-qPCR and western blotting, respectively, at 1.5 h after LPS/D-GalN. Data were presented as mean ± SD. *n* = 6, \**P* < 0.05, \*\**P* < 0.01.

FIGURE 4 | KCs are involved in the protective effect of MF on LPS-mediated acute liver injury in D-GalN-sensitized mice. Mice were injected i.p. with GdCl<sup>3</sup> 2 days before LPS/D-GalN challenge, KCs were i.v.-infused into mice at 24 h prior to LPS/D-GalN challenge, and MF pretreatment was given orally at 1 and 7 h prior to LPS/D-GalN challenge. (A) Survival rates of mice were monitored for 48 h after LPS/D-GalN challenge. (B) Serum ALT and AST activities were assessed at 6 h after LPS/D-GalN challenge. Hepatic TNF-α mRNA (C) and protein (D) expression were determined by RT-qPCR and ELISA, respectively, at 1.5 h after LPS/D-GalN challenge. Data are presented as mean ± SD. *n* = 6, \*\**P* < 0.01.

expression and activity in LPS-D-GalN-treated mice. Further, the adoptive transfer of KCs with control siRNA restored MFmediated upregulation of HO-1. However, adoptive transfer of KCs with HO-1 siRNA showed an inhibitory effect on HO-1 expression and activity (**Figure 5**, p < 0.01).

# KC-Derived HO-1 Is Key for the Effect of MF in Mice in Response to LPS/D-GalN

Finally, we investigated whether KC-derived HO-1 participated in the protective effects of MF on LPS/D-GalN-mediated acute liver injury in mice. To test this hypothesis, the HO-1 siRNA- or control siRNA-transfected KCs were transferred separately into mice with KC depletion. As shown in **Figure 6**, transferral of HO-1 siRNA- or control siRNA-transfected KCs into MF alonepretreated mice with KC depletion did not affect mouse survival rate, ALT/AST activities, or TNF-α production. However, in HO-1 siRNA-knockdown KCs, the protective effects of MF on LPS/D-GalN-induced FHF were reduced compared with those in HO-1 positive KCs, as indicated by the increased death rate, elevated ALT/AST activities, and enhanced TNF-α production (**Figures 6A–D**, p < 0.01). Overall, these results indicated that KC-derived HO-1 was crucial for the effects of MF on LPS/D-GalN-induced FHF.

# MF Inhibited LPS-Activated TLR4 Signaling and TNF-α Production in KCs

The observations above suggested that KCs might be a target of the effects of MF in response to LPS. To further verify this hypothesis, we isolated KCs from mice and examined whether MF also has an anti-inflammatory effect on the TNFα production from KCs stimulated by LPS. Triggering of TLR4 by LPS leads to the activation of transcript factors AP-1 and NF-κB, which promote the expression of TNF-α. The analysis of ELISA and RT-qPCR results revealed that MF remarkably inhibited LPS-induced expression of mRNA and protein TNFα (**Figures 7A,B**). Further, the assay of reporter genes showed that MF significantly decreased LPS-induced AP-1 and NFκB activities (**Figure 7C**). In parallel with these results, flow cytometry analysis indicated that MF also downregulated LPSenhanced TLR4 expression on the surface of KCs (**Figure 7D**). All these data supported the conclusion that MF could effectively suppress TLR4 signaling and TNF-α production triggered by LPS in KCs.

#### HO-1 Mediated the Suppression of MF on LPS-Triggered Inflammatory Responses in KCs

Given that the inhibitory role of MF in TLR signaling occurs through the HO-1 pathway in KCs, we first assessed the role of MF in the expression and activity of HO-1. Consistent with the above results observed in the animal experiment, the expression and activity of HO-1 were greater in MF-treated KCs than in untreated LPS-stimulated KCs (**Figures 7E–G**, p < 0.01). Further, knockdown of HO-1 with a specific HO-1 small interfering RNA markedly downregulated the HO-1 expression and activity induced by MF in LPS-stimulated KCs (**Figures 8A–C**). In line with these results, knockdown of HO-1 also abolished the beneficial role of MF on the TLR4 expression and TNF-α production induced by LPS (**Figures 8D–F**). These data indicated that HO-1 may mediate the anti-inflammatory effect of MF in KCs by downregulating the TLR4 signaling pathway.

# DISCUSSION

Fulminant liver failure is one of the most lethal clinical syndromes and is characterized by rapid deterioration of liver function, which leads to hepatic encephalopathy, systemic inflammatory response syndrome, and even multiorgan failure, which pose severe threats to a patient (4). Liver transplantation remains the only effective therapeutic strategy, but patient recovery is currently constrained by the limitations of organ donation (30). Accordingly, there is an urgent need to find effective drugs to ameliorate FHF and avoid severe secondary

complications. In this study, the protective role of MF and the underlying mechanism of action were explored in an LPS/D-GalN-induced mouse model of FHF. Pretreatment with MF significantly enhanced the survival of FHF mice, decreased the activities of ALT and AST in serum, and attenuated liver injury induced by LPS/D-GalN. MF administration not only reduced LPS-mediated TNF-α production but also inhibited KC activation by suppressing the TLR4/NF-κB signaling pathway. Furthermore, the inhibition of TNF-α by MF was indeed associated with the upregulation of HO-1 in KCs.

TNF-α is recognized as a critical inflammatory mediator in the pathogenesis of inflammation and infection-related diseases; moreover, its toxicity may directly damage hepatocytes (31). High amounts of circulating pro-inflammatory TNFα, the characteristic manifestation of FHF, ultimately induce multiple organ failure (32). TNF-α, via binding TNF receptor on the surface of hepatocytes as an exocellular death signal, can stimulate inner oxidative stress and initiate the apoptotic pathway, which triggers more extensive hepatocyte death and a serious inflammatory response (33). The administration of TNF-α was shown to accelerate liver injury in this model by neutralizing antibodies against TNF-α or inhibitors of TNF-α production; knock-out of TNF-α or TNF-α receptors completely abrogated LPS/D-GalN-induced FHF (34–36). In addition, the production and liberation of TNF-α are predominantly regulated by the TLR4/NF-κB signaling pathway during liver inflammation (37). Accordingly, the inhibition of the TLR4/NF-κB signaling pathway can decrease TNF-α production, which is an effective strategy for FHF alleviation (38, 39). In our study, MF suppressed TNF-α synthesis transcriptionally and translationally, as shown by the decrease in mRNA and protein of TNF-α after MF pretreatment in both mice and KCs. As shown by the reporter gene in KCs, the TLR4 signaling pathway was significantly inhibited by MF, suggesting that MF could reduce TNFα production via inhibition of TLR4/NF-κB signaling. The abrogation of TNF-α by MF pretreatment significantly retarded the progression of fulminant hepatic damage, which was a critical contribution to the mechanism of MF against acute liver injury.

Although various cell types in the liver are capable of producing TNF-α, KCs are acknowledged as the main source (40). KC-derived TNF-α is majorly involved in the development of FHF (41). Indeed, the liver injury in LPS/D-GalN-induced FHF is associated with activated KCs. The deletion of KCs by GdCl<sup>3</sup> not only completely blocked this disease but also inhibited TNF-α production (42). In contrast, the infusion of KCs restored the diseased state (43). Given these results, we hypothesized that the beneficial effect of MF on LPS/D-GalN-induced FHF was through the inhibition of KCs to produce TNF-α. To confirm

FIGURE 7 | MF inhibited LPS-activated TLR4 signaling and HO-1 upregulation in KCs. KCs were transfected with both pAP1-Luc or pNF-κB-Luc and the pRL-TK *Renilla* plasmid for 24 h, pretreated with PBS or MF (10−<sup>5</sup> M) 30 min prior to PBS or LPS challenge (100 ng/mL). (A) RNA was extracted from KCs at 3 h after LPS challenge and TNF-α mRNA was assayed by real-time RT-PCR. (B) Cell culture supernatant was obtained 3 h after LPS stimulation to determine TNF-α levels by using ELISA. (C) AP1 and NF-κB activities were measured at 30 min after LPS challenge. (D) The expression of TLR4 on the surface of KCs was analyzed at 3 h after LPS by using flow cytometry. HO-1 mRNA (E) and protein (F) expression and activity (G) were determined at 3 h after LPS challenge. Data are presented as mean ± SD. *n* = 6, \*\**P* < 0.01.

this, we utilized GdCl<sup>3</sup> to delete primary KCs and subsequently treated LPS/D-GalN-treated mice with MF. KC deletion indeed ameliorated the inflammatory response following LPS/D-GalN administration, although the activities of ALT and AST and the expression of TNF-α were not remarkably different in the mice with or without MF pretreatment. Moreover, after transferring KCs to the KC-deleted mice, the severe hepatic inflammation was restored, and the protective effect of MF was again observed, indicating the key role of KCs in the protection of MF against LPS/D-GalN-induced FHF.

HO-1, a member of the cellular defense system, fulfills a significant function in the elimination of oxidative stress products and the suppression of inflammatory response through multiple pathways. HO-1 exerts a protective effect by catalyzing the oxidative degradation of heme to carbon monoxide (CO), iron, and biliverdin-IXα, which scavenges endogenous injury factors to protect hepatocytes (11). Several studies have reported that HO-1 has a positive protective role in hepatitis and that the downregulation of HO-1 expression aggravated the severity of hepatic injury (44, 45). Consistent with this, it was observed that

LPS challenge and TNF-α mRNA expression was assayed by real-time RT-PCR (D). Cell culture supernatant was obtained at 3 h after LPS challenge and TNF-α protein expression was determined by ELISA (E). The expressions of TLR4 on the surface of KCs was analyzed at 3 h after LPS challenge by flow cytometry (F). Data are presented as mean ± SD. *n* = 6, \*\**P* < 0.01.

the expression and activity of HO-1 were elevated by MF in the LPS/D-GalN-treated mice, but the protective role of MF was not observed after the use of an HO-1 inhibitor (ZnPP-IX), which showed that the upregulation of HO-1 was a potential target for preventing the deterioration of hepatic injury. However, HO-1 exhibits distinctive regulatory states in different types of cells in the liver, and not every cell can be a therapeutic target for MF against FHF. Hepatic inflammation develops through cascade amplification (46). Thus, the restrictions of the target cell and the superior inflammatory response can fundamentally block the development of liver injury from the source. Activation of KCs has a clear link with the development of hepatic inflammation (47). To confirm this, KCs were examined, and the role of HO-1 in KCs was also investigated. MF pretreatment promoted the expression and activity of HO-1 in KCs, but the protection of MF was equally abated when KC-derived HO-1 expression was silenced by siRNA. Furthermore, when KCs with HO-1 or control siRNA were transferred to primary KC-deleted mice that were subjected to LPS/D-GalN challenge, the survival rate and the activities of transaminases were not improved, and meanwhile the high level of TNF-α. These data more comprehensively illuminated that the upregulation of KC-derived HO-1 was imperative for the attenuation of acute liver injury by MF.

In conclusion, these data revealed that MF exhibited significant protection against LPS/D-GalN-induced acute liver injury and alleviated TNF-α production through the inhibition of the TLR4/NF-κB signaling pathway. The abrogation of TNFα was correlated with the upregulation of HO-1 expression. Furthermore, KCs played a key role in the protection of MF against acute liver injury and the promotion of KC-derived HO-1 was involved in the protective effects of MF. Collectively, MF attenuated acute liver injury induced by LPS/D-GalN through a reduction in TNF-α production by promoting KCs to upregulate HO-1 expression. This study has demonstrated a potential prophylactic agent with a novel mechanism of action against LPS/D-GalN-induced acute liver injury.

#### 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.

#### ETHICS STATEMENT

The animal study was reviewed and approved by the Animal Care and Use Committee of Chongqing Medical

#### REFERENCES


University. Written informed consent was obtained from the owners for the participation of their animals in this study.

# AUTHOR CONTRIBUTIONS

SY and GK performed experiments, analyzed data, and wrote the manuscript. SW, ZZ, and XY performed experiments and analyzed data. LZ and BW provided guidance and suggestions for the study. XG and JW conceived and supervised the study, designed experiments, analyzed and interpreted data, and wrote the manuscript.

#### FUNDING

This study was supported by the National Natural Science Foundation of China (81774124) and Chongqing Yuzhong District Science and Technology Project (20170104).

#### ACKNOWLEDGMENTS

The authors thank Ms. Rong Jiang for providing professional technical support for tissue section and analysis.


**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 Yang, Kuang, Zhang, Wu, Zhao, Wang, Yin, Gong and Wan. 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.

# Mesenchymal Stem Cell-Derived Extracellular Vesicles: A Novel Cell-Free Therapy for Sepsis

Yanwei Cheng1†, Xue Cao2† and Lijie Qin<sup>1</sup> \*

*<sup>1</sup> Department of Emergency, Henan Provincial People's Hospital, People's Hospital of Zhengzhou University, People's Hospital of Henan University, Zhengzhou, China, <sup>2</sup> Department of Rheumatology and Immunology, Henan Provincial People's Hospital, People's Hospital of Zhengzhou University, People's Hospital of Henan University, Zhengzhou, China*

Sepsis remains a serious and life-threatening disease with high morbidity and mortality. Due to the unique immunomodulatory, anti-inflammatory, anti-apoptotic, anti-microbial, anti-oxidative, and reparative properties, mesenchymal stem cells (MSCs) have been extensively used in preclinical and clinical trials for diverse diseases and have shown great therapeutic potential in sepsis. However, concerns remain regarding whether MSCs can become tumorigenic or have other side effects. Extracellular vesicles (EVs) are a heterogeneous group of membrane-enclosed particles released from almost any cell and perform an important role in intercellular communication. Recently, it has emerged that EVs derived from MSCs (MSC-EVs) appear to exert a therapeutic benefit similar to MSCs in protecting against sepsis-induced organ dysfunction by delivering a cargo that includes RNAs and proteins to target cells. More importantly, compared to their parent cells, MSC-EVs have a superior safety profile, can be safely stored without losing function, and possess other advantages. Hence, MSC-EVs may be used as a novel alternative to MSC-based therapy in sepsis. Here, we summarize the properties and applications of MSC-EVs in sepsis.

Keywords: mesenchymal stem cell-derived extracellular vesicles (MSC-EVs), sepsis-induced acute lung injury, sepsis-induced acute kidney injury, sepsis-associated cardiovascular disorder, sepsis-induced liver injury

#### INTRODUCTION

In the recent "sepsis-3.0" consensus (1), sepsis is defined as a life-threatening, multiorgan dysfunction caused by a dysregulated host response to infection. In view of the high morbidity and mortality, sepsis has been described as the quintessential medical disorder of the 21st century. Despite continuous progress in the development of therapeutic strategies, sepsis is still a major clinical problem and remains the leading cause of death (2, 3) in the critically ill patient population, due to uncontrolled inflammation together with immunosuppression. In such a context, an investigation into novel therapies to ameliorate sepsis would be urgently needed.

With the emergence of stem cells as potential therapeutic agents for diverse diseases, attempts to use stem cells, especially mesenchymal stem cells (MSCs), to ameliorate sepsis in animal models have increased exponentially. Growing preclinical data have suggested that MSCs can directly modify pathophysiology and underlying injury mechanisms in sepsis through the immunomodulatory, anti-bacterial, anti-inflammatory, anti-oxidative, anti-apoptotic, and reparative properties. It is worth noting that MSCs have also been confirmed as having a therapeutic effect on sepsis in two clinical trials (4, 5). Even then, the utilization of MSCs as therapeutic

#### Edited by:

*Lukas Martin, University Hospital RWTH Aachen, Germany*

#### Reviewed by:

*Sebastian Wendt, University Hospital RWTH Aachen, Germany Yasser Mohamed El-Sherbiny, Nottingham Trent University, United Kingdom*

> \*Correspondence: *Lijie Qin qinlijie1819@163.com*

*†These authors share first authorship*

#### Specialty section:

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

Received: *12 January 2020* Accepted: *23 March 2020* Published: *21 April 2020*

#### Citation:

*Cheng Y, Cao X and Qin L (2020) Mesenchymal Stem Cell-Derived Extracellular Vesicles: A Novel Cell-Free Therapy for Sepsis. Front. Immunol. 11:647. doi: 10.3389/fimmu.2020.00647*

Frontiers in Immunology | www.frontiersin.org 1 April 2020 | Volume 11 | Article 647

**270**

agents for sepsis has raised several concerns over the past decade, including senescence-induced genetic instability, the heterogeneity of MSCs populations, quality assurance challenges of MSCs, and finding optimal MSCs tissue sources. Most crucially, the significant hurdles that potential therapeutics face in the future are MSCs safety concerns. It has been reported that MSCs play a dual role in immune regulation; on the one hand, MSCs exhibit immunosuppressive function (6), while on the other hand, MSCs may act as antigen-presenting cells, inducing an immune response (7). In addition, MSCs can potentially

become tumorigenic either by direct malignant transformation of MSCs or indirectly by facilitating the growth of other tumor cells (8). So far MSCs have been found in several tumor types, including gastric adenocarcinoma (9), lipoma (10), and osteosarcoma (11), strongly indicating their involvement in tumor development. In light of this fact, the probability of the risks must be cautiously weighed against the potential benefits to every patient.

Recently, special attention has been paid to the extracellular vesicles (EVs) derived from MSCs (MSCs-EVs), which can function as shuttles for the delivery of a cargo that includes RNAs and proteins from parental to target cells and are as biologically active as the parent cells themselves (12–15). Hence, it is conceivable that the protective effect of MSCs on sepsis is at least partly due to the release of EVs.

# CHARACTERIZATION OF MSC-EVS (FIGURE 1)

EVs are a heterogeneous group of membrane-enclosed particles that can be released from almost any cell. Based on the current knowledge of their size and biogenesis, EVs can be classified into three broad groups: exosomes (Exos), microvesicles (MVs), and apoptotic bodies. Exos (40–150 nm) bud from the endosomal system, whereas MVs (100–1,000 nm) are shed directly from the plasma membrane, and apoptotic bodies (1,000–5,000 nm) originate from apoptotic cells (16–18). In general, apoptotic bodies can be removed by the commonly used isolation method of ultracentrifugation (19). It should be noted that each individual cell can release both Exos and MVs simultaneously. However, so far it is impossible to unanimously distinguish MVs and Exos. In addition, most studies have not clearly defined the origin of EVs under study. Therefore, we here use EVs as an umbrella term to include both MVs and Exos.

MSC-EVs are released from resting MSCs and can be markedly increased during cellular activation or cell stress (12, 20, 21). They contain a trove of bioactive substances such as proteins, lipids, and most importantly, nucleic acids, thereby playing an important role in immune modulation, proangiogenesis, anti-apoptosis, and tissue regeneration. Up to now, over 5,000 species of MSC-EVs proteins have been characterized, including mediators controlling self-renewal, differentiation, signal transduction, and additional MSCs antigens affecting the migration of MSCs (19, 22). In addition, MSC-EVs contain a number of adhesion molecules such as CD44, CD29, alpha 4-integrin, and alpha 5-integrin etc., which contribute to the identification of MSC-EVs (22). Accumulating evidence has demonstrated that MSC-EVs modulate immune response by producing cytokines, growth factors, and tolerogenic molecules, such as IL-10, IL-6, IL-37, and lipocalin-2 and transforming growth factor (TGF)-β, programmed death ligand-1(PD-L1), galectin-1, etc. (20, 23, 24). Other than proteins, MSC-EVs are enriched with nucleic acids, including mRNA, microRNA (miRNA), and DNA (25). It has been demonstrated that the transfer of miRNAs from MSC-EVs to target cells may be the underlying mechanism in alleviating injury of the kidney (26, 27), heart (28), liver (29), and brain (30). In recent years, more than 150 miRNAs have been identified in the cargo of MSC-EVs. These miRNAs are usually related to apoptosis, tumorigenesis, immune responses, angiogenesis, and organism development, such as miR-221, miR-23b, miR-125b, miR-451, miR-31, miR-24, miR-214, miR-122, miR-16, miR-150, and miR133b, etc. (19, 31–34). All the internal components in MSC-EVs are surrounded by a bilayer of lipids that play an important role in protecting them. Compared to their parent cells, EVs are enriched with some specific lipids (cholesterol, glycosphingolipids, and phosphatidylserine) (35–38), making them more functional. Our previous study (21) also showed that sphingosine-1-phosphate enriched within MSC-EVs could enhance the repair effect of MSCs for articular cartilage.

# THERAPEUTIC APPLICATIONS OF MSC-EVS IN SEPSIS

Through delivering proteins and nucleic acids, MSC-EVs seem to hold many functions of the MSCs themselves and are described as a new mechanism of intercellular communication (39). Nowadays, increasing studies have clearly confirmed that EVs alone are responsible for the therapeutic effects of MSCs in sepsisinduced organ dysfunction, including sepsis-induced acute lung injury, sepsis-induced acute kidney injury, sepsis-associated cardiovascular disorder, and sepsis-induced liver injury (40). Accordingly, MSCs-EVs may be an alternative therapy to MSCs and will be the next-generation therapeutic route for sepsis.

# MSC-EVs and Sepsis-Induced Actue Lung Injury

Acute lung injury (ALI) is the most common organ injury in septic patients and leads to a greater mortality. In the past few years, the beneficial effects of MSCs on sepsis-induced ALI have been shown to be attributed to the release of EVs. Using LPSinduced ALI in an ex vivo perfused human lung, Park et al. (41) reported that MSC-EVs significantly increased alveolar fluid clearance and reduced protein permeability and numerically lowered the bacterial load in the injured alveolus. In a mouse model of E. coli endotoxin-induced ALI, Zhu et al. (42) also confirmed the similar beneficial effects of human MSC-EVs and further demonstrated that the protection was in part due to the delivery of keratinocyte growth factor (KGF) mRNA from MSC-EVs to the injured alveolar epithelium and lung endothelium. This falls in good accordance with Monsel's study (43) that MSC-EVs improved survival in ALI from E. coli pneumonia via a

Keratinocyte growth factor; MVs, Microvesicles; MSCs, Mesenchymal Stem Cells; miR, microRNA; PDGF-D, Platelet-derived growth factor D; VEGF, Vascular

endothelial growth factor.

mechanism partially dependent on KGF secretion. The authors (43) further discovered that MSC-EVs could enhance monocyte phagocytosis of bacteria through transferring cyclooxygenase 2 (COX2) mRNA to monocytes, with a resultant increase in prostaglandin E2 (PGE2) secretion. PGE2 is essential for transforming the polarization of monocytes-macrophages M1 into M2 phenotype (44–46). In addition, Song et al. (47) showed that miR-146a promoted polarization of M2 macrophages and finally led to increased survival in septic mice. MSC-EVs also contain a substantial quantity of angiopoietin-1 (Ang-1) mRNA, which plays an essential role in vascular stabilization and resolving inflammation. In the current study, Tang et al. (48) demonstrated that the therapeutic benefit of EVs in ALI, and their immunomodulatory properties on macrophages, were partly mediated through their content of Ang-1 mRNA. Furthermore, the transfer of miR-451 and mitochondria from MSC-EVs to the alveolar epithelium and macrophages also contributed to preventing ALI (49). Altogether, these data indicate that

MSC-EVs may represent a novel therapeutic option for sepsisinduced ALI.

#### MSC-EVs and Sepsis-Induced Actue Kidney Injury

Mounting studies have provided convincing evidence that MSCs exert their renoprotective effects by releasing EVs in several acute kidney injury (AKI) models. AKI is commonly present in more than 20% of septic patients (50). However, data about the therapeutic effects of MSC-EVs in sepsis-induced AKI are scarce. Currently, the anti-apoptosis (51), pro-angiogenesis (52), and anti-inflammatory (53) properties of MSC-EVs have been considered as the important mechanistic approaches to ameliorate AKI induced by different causes. In AKI mice models induced by ischemia-reperfusion injury, Bruno et al. (27) found that MSC-EVs induced the expression of anti-apoptotic genes (Bcl-XL, Bcl2, and BIRC8) in renal tubular epithelial cells while simultaneously down-regulating pro-apoptotic genes (Casp1, Casp8, and LTA), thus reducing tubular cell apoptosis and conferring an anti-apoptotic phenotype necessary for tissue repair. Interestingly, the anti-apoptosis effect of MSC-EVs may be mediated in part by the transfer of miRNAs (miR-223 and miR-199a-3p) to tubular cells (54, 55). In addition, MSC-EVs can directly shuttle several pro-angiogenesis transcription factors including vascular endothelial growth factor (VEGF) (52), insulin-like growth factor 1 (IGF-1) (56), and basic fibroblast growth factor (bFGF) (57) to the damaged renal tubular epithelial cells and enhance angiogenesis, which is considered as an important step in tissue regeneration. In a mouse model of AKI induced by glycerol, Bruno (58) demonstrated that 15 µg of MSC-EVs injected intravenously activated the proliferative progress in viable tubular cells through their pro-angiogenesis effects. Importantly, the effects of MSC-EVs on the recovery of AKI were as effective as their parental MSCs. It is well known that excessive inflammation is the culprit leading to sepsis-induced AKI. MSC-EVs can modulate renal inflammation by reducing the release of several pro-inflammatory cytokines, including IL-1, IL-6, and TNF-α, while restoring the level of the anti-inflammatory cytokine IL-10 (26, 53, 59). This anti-inflammatory effect may be associated with the C-C motif chemokine receptor-2 (CCR2) contained within MSC-EVs, which can suppress macrophages function and attenuate renal injury (60). At present, MSC-EVs have been studied in patients with chronic kidney disease (61). The results showed that administration of MSC-EVs had no side effects and caused significant improvement of the kidney function. Collectively, there are good reasons to believe that MSC-EVs could be exploited as a new therapeutic approach for sepsis-induced AKI.

#### MSC-EVs and Sepsis-Associated Cardiovascular Disorder

Despite only a few studies having addressed potential cardioprotective effects of MSCs on sepsis-induced cardiovascular disease, the data still support that MSCs mediate their protection through the secretion of EVs. In septic mice induced by cecal ligation and puncture (CLP), MSC-EVs displayed cardioprotective benefits and enhanced survival of cardiomyocyte cells. In this research, Wang et al. (28) further proved that miR-223 enriched in MSC-EVs was critical for EVselicited action in sepsis. miR-223 can inhibit the expression of Sema3A and Stat3, negatively regulating the expression of many inflammatory genes, such as TNF-α, IL-6, and IL-1 (62). Another important piece of research by Tabet et al. (63) also mentioned that miR-223 was able to alleviate the inflammatory process in human coronary artery endothelial cells by downregulating intercellular adhesion molecule 1. In addition, some antiapoptotic miRNAs contained within MSCs-EVs showed a vital role in cardioprotection. For example, miR-221 can reduce the expression of p53 upregulated modulator of apoptosis and is more highly enriched in MSC-EVs than that in their parent MSCs. Numerous studies (34, 64) have confirmed that the delivery of miR-221 contained within EVs contributed to the cardioprotection by MSC-EVs. A study by Yu et al. (65) demonstrated that MSC-EVs enriched with anti-apoptotic miR-19a also restored cardiac function and reduced infarct size in a rat model of acute myocardial infarction (AMI). In addition to miRNAs, the protein of CD73 on the surface of MSC-EVs was suggested to overcome the pro-apoptotic milieu of the perfused myocardium (66). In AMI mice models, Bian et al. (67) reported that intramyocardial injection of MSC-EVs markedly promoted angiogenesis and reduced infarct size. They further identified that MSC-EVs might be a mixture of miR-150-enriched MVs and Exos (33). Moreover, the pro-angiogenesis effect of the other cargo of MSC-EVs had been investigated in vitro and in vivo MI models, including miR-210 (66), miR-21 (68), platelet-derived growth factor D (PDGF-D) (69), hypoxia-inducible factor 1 alpha (HIF1-α) (70), and extracellular matrix metalloproteinase inducer (EMMPRIN) (71) and angiopoietin (72). Recently, Xiao et al. (73) demonstrated that the beneficial effects offered by MSCs in MI model were at least partially because of improved autophagic flux through excreted EVs containing mainly miR-125b-5p. On the basis of these studies, MSC-EVs seem to be an ideal therapeutic agent for sepsis-induced cardiovascular disease in the near future.

#### MSC-EVs and Sepsis-Induced Liver Injury

During sepsis, liver injury occurs frequently and contributes to the pathogenesis of multiple organ dysfunction, and has been considered as an early indicator of poor outcome in septic patients (74, 75). Studies in a range of liver disease models demonstrated that MSC-EVs could suppress inflammatory response, reduce hepatocyte apoptosis, and enhance liver regeneration. However, few applications of MSC-EVs into sepsisinduced liver injury are available. Nong et al. (76) studied the hepatoprotective effect of MSC-EVs in a mouse model of hepatic ischemia-reperfusion injury. They showed that human-induced pluripotent stem cell-derived MSC-EVs protected hepatocytes via reducing pro-inflammatory cytokine production, such as TNF-α, IL-6, and high mobility group box 1 (HMGB1). Likewise using a concanavalin-A-induced liver injury model, Tamura et al. (77) demonstrated that MSC-EVs decreased production of proinflammatory cytokines, while anti-inflammatory cytokine and regulatory T cell levels increased. Recently, the anti-apoptotic and anti-oxidant effects of MSC-EVs were also investigated. A study in a lethal mouse model of liver failure induced by Dgalactosamine/TNF-α showed that MSC-EVs containing Y-RNA-1 activated anti-apoptotic pathways, thus mitigating hepatic injury and enhancing survival (78). In carbon tetrachloride (CCl4)-induced liver injury, Yan et al. (79) demonstrated that MSC-EVs promoted the recovery of hepatic oxidant injury through the delivery of glutathione peroxidase 1 (GPX1). Additionally, Tan et al. (80) confirmed that MSC-EVs were capable of promoting hepatocyte regeneration by inducing the IL-6/STAT3 pathway and cell cycle progression. Proteins in MSC-EVs, such as interleukin 6 signal transducer (IL6ST), chemokine Ligand 2 (CXCL2), and hepatocyte growth factor (HGF) etc., were found to be involved in the liver-regeneration process (80). However, it remains to be determined whether miRNAs or other mRNAs within MSC-EVs may have any impact on the proliferation effect of the hepatocytes. According to these studies, MSC-EVs may be considered as a novel therapeutic approach for alleviating sepsis-induced liver injury, and correlative research should fill the gap in this field in the future.

#### ADVANTAGES AND CHALLENGES OF MSC-EVS THERAPY

Compared to MSC-based therapy, MSC-EVs therapy offers more advantages. First, MSC-EVs are highly stable and suitable for long-term storage without the addition of potentially toxic cryopreservatives (21, 81, 82). Second, MSC-EVs can induce stronger signaling in intercellular communication by directly transferring functional proteins and miRNAs to the recipient cells (83, 84). Third, MSC-EVs have no heterologous risk (85, 86) and no immune response after allogeneic application (87). Furthermore, a significant advantage of MSC-EVs is to circumvent the potential tumorigenicity of MSCs. So far, there is no evidence showing the oncogenic potential of MSC-EVs. A study (88) reported that MSC-EVs can inhibit tumor growth and exert an anti-tumor activity both in vitro and in vivo. Given this, MSC-EVs look to be a novel and promising therapeutic approach for sepsis.

However, there remain significant challenges that need to be addressed prior to the potential application of MSC-EVs in human sepsis. First, the protocol for isolation has to be standardized. Currently, no "gold standard protocol" for EVs isolation exists, instead different isolation methods are used to isolate different subsets of EVs even though the source is the same. Second, the suitable markers to analyze EVs at a single-vesicle level must be established. So far no consensus exists regarding the nomenclature of EVs or the markers that distinguish each type of EV once they have been secreted or shed from the cells. Third, there is a paucity of knowledge about the content within MSC-EVs. EVs contain a trove of cellular cargos, which are able to act as diverse functions and can vary widely between sources and conditions. It is unknown which effector molecules are important within EVs or which are not protective or harmful. Meanwhile, there is a critical need for exploiting possible ways to modify MSC-EVs composition

#### REFERENCES


and modulate their biological activities. Fourth, the therapeutic capacity of MSC-EVs derived from different sources in sepsis should be investigated. In addition, much work is needed on how to optimize the methods to produce MSC-EVs on a large scale, how to validate the dosage and half-life of MSC-EVs, and confirming the potential effects of MSC-EVs in the late stage of sepsis. Finally, the unknown negative effects of MSC-EVs have to be clarified.

#### CONCLUSIONS

Extracellular vesicles (EVs) are naturally released from almost any cell and participate in cell-to-cell communication in physiological as well as pathological processes by transferring their components, such as proteins, miRNA, mRNA, and even mitochondria. A study (89) considered EVs as possible culprits during the pathogenesis of sepsis, whereas EVs derived from MSCs showed striking therapeutic benefits in sepsis. In this review, we have summarized the recent knowledge related to the therapeutic applications of MSC-EVs in sepsis. These considerable preclinical data support the hypothesis that cellfree therapy with MSC-EVs could be a novel alternative MSC-based therapy in sepsis, especially in early stage sepsis. However, realizing this promising therapeutic approach of MSC-EVs would require extensive testing to validate their safety and efficacy.

#### AUTHOR CONTRIBUTIONS

YC and XC wrote the first draft of this article and designed the figure. LQ critically revised the manuscript for important intellectual content. All authors approved the final version.

# FUNDING

The present work was supported by the 23456 Talent Project of Henan Provincial People's Hospital to LQ, Research Startup fund of Henan Provincial People's Hospital to YC and XC.

of mesenchymal stromal cells (MSCs) in severe neutropenic patients (pts) with septic shock (SS) (RUMCESS trial). Blood. (2015) 126:2220. doi: 10.1182/blood.V126.23.2220


**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 Cheng, Cao and Qin. 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.

,

# Non-invasive Assessment of Mitochondrial Oxygen Metabolism in the Critically Ill Patient Using the Protoporphyrin IX-Triplet State Lifetime Technique—A Feasibility Study

Charles Neu1,2,3†, Philipp Baumbach1,2†, Alina K. Plooij 1,2, Kornel Skitek 1,2, Juliane Götze1,2

#### Edited by:

Pietro Ghezzi, Brighton and Sussex Medical School, United Kingdom

#### Reviewed by:

Robert Campbell, The University of Utah, United States Sergio Iván Valdés-Ferrer, Salvador Zubirán National Institute of Medical Sciences and Nutrition (INCMNSZ), Mexico

#### \*Correspondence:

Sina M. Coldewey sina.coldewey@med.uni-jena.de

†These authors have contributed equally to this work

#### Specialty section:

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

Received: 23 December 2019 Accepted: 03 April 2020 Published: 07 May 2020

#### Citation:

Neu C, Baumbach P, Plooij AK, Skitek K, Götze J, von Loeffelholz C, Schmidt-Winter C and Coldewey SM (2020) Non-invasive Assessment of Mitochondrial Oxygen Metabolism in the Critically Ill Patient Using the Protoporphyrin IX-Triplet State Lifetime Technique—A Feasibility Study. Front. Immunol. 11:757. doi: 10.3389/fimmu.2020.00757 Christian von Loeffelholz <sup>1</sup> , Christiane Schmidt-Winter 1,2 and Sina M. Coldewey 1,2,3 \* <sup>1</sup> Department of Anesthesiology and Intensive Care Medicine, Jena University Hospital, Jena, Germany, <sup>2</sup> Septomics

Research Center, Jena University Hospital, Jena, Germany, <sup>3</sup> Center for Sepsis Control and Care, Jena University Hospital, Jena, Germany

The imbalance of oxygen delivery and oxygen consumption resulting in insufficient tissue oxygenation is pathognomonic for all forms of shock. Mitochondrial function plays an important role in the cellular oxygen metabolism and has been shown to impact a variety of diseases in the intensive care setting, specifically sepsis. Clinical assessment of tissue oxygenation and mitochondrial function remains elusive. The in vivo protoporphyrin IX-triplet state lifetime technique (PpIX-TSLT) allows the direct, non-invasive measurement of mitochondrial oxygen tension (mitoPO2) in the human skin. Our recently established measurement protocol for the Cellular Oxygen Metabolism (COMET) Monitor, a novel device employing the PpIX-TSLT, additionally allows the evaluation of oxygen consumption (mitoVO2) and delivery (mitoDO2). In the intensive care setting, these variables might provide new insight into mitochondrial oxygen metabolism and especially mitoDO<sup>2</sup> might be a surrogate parameter of microcirculatory function. However, the feasibility of the PpIX-TSLT in critically ill patients has not been analyzed systematically. In this interim study analysis, we evaluated PpIX-TSLT measurements of 40 patients during the acute phase of sepsis. We assessed (a) potential adverse side effects of the method, (b) the rate of analyzable measurements, (c) the stability of mitoPO2, mitoVO2, and mitoDO2, and (d) potential covariates. Due to excessive edema in patients with sepsis, we specifically analyzed the association of patients' hydration status, assessed by bioimpedance analysis (BIA), with the aforementioned variables. We observed no side effects and acquired analyzable measurements sessions in 92.5% of patients (n = 37/40). Different measures of stability indicated moderate to good repeatability of the PpIX-TSLT variables within one session of multiple measurements. The determined limits of agreement and minimum detectable differences may be helpful in identifying outlier measurements. In conjunction with signal quality they mark a first step in developing a previously unavailable standardized measurement quality protocol. Notably, higher levels of hydration were associated with lower mitochondrial oxygen tension. We conclude that COMET measurements are viable in patients with sepsis. To validate the clinical and diagnostic relevance of the PpIX-TSLT using the COMET in the intensive care setting, future studies in critically ill patients and healthy controls are needed.

Keywords: sepsis, COMET, mitochondrial dysfunction, critically ill patients, protoporphyrin IX-triplet state lifetime technique, cellular oxygen metabolism, mitochondrial oxygen metabolism, mitochondrial oxygen tension

#### INTRODUCTION

Sepsis is defined as a life-threatening host response toward infection resulting in organ dysfunction (1). Despite advances in the pathophysiological understanding of this condition, research has not led to significant changes in sepsis therapy. It therefore remains one of the most prevalent critical conditions worldwide with only supportive therapy. A major hallmark of sepsis and septic shock in particular is disturbed tissue oxygenation. Hitherto, surrogate parameters of tissue oxygenation have been shown to be of limited reliability. Early goal-directed therapy, focusing on central venous oxygen saturation as a surrogate parameter of tissue oxygenation, for example, has shown no overall benefit in a recent meta-analysis (2) leaving clinicians without any evidence on which to base their decisions. Therefore, there is a need for new research into the direct measurement of tissue oxygenation.

In recent years, the mitochondrion has become a focus of medical research. The mitochondrion, as "powerhouse" of the cell, is responsible for the regulation of cellular oxygen metabolism and could therefore pose a potential target for the measurement of tissue oxygenation. Studies have demonstrated a pathophysiological involvement of mitochondrial function in, among others, cancer, heart, and age-related diseases (3– 5). In sepsis, studies have yielded first indications of a dysregulated mitochondrial function. Clinical trials showed increased mitochondrial protein synthesis in patients with sepsis (6, 7). Another study demonstrated a decreased mitochondrial function in biopsies from patients with sepsis (8). Thus far, clinical measurements were performed in muscle biopsies. As the analysis is ex vivo there may be pre-analytical confounders influencing the results. Also, the procedure is invasive and therefore unlikely to be used routinely for diagnostic purposes. Non-invasive direct measurements of mitochondrial function in patients could pose a feasible method to assess tissue oxygenation in patients with sepsis.

Mik et al. introduced the protoporphyrin IX-triplet state lifetime technique (PpIX-TSLT) for non-invasively measuring mitochondrial oxygen tension (mitoPO2) (9). In brief, the method is based on the delayed fluorescence of protoporphyrin IX (PpIX), the naturally occurring precursor of the heme molecule, which can be enriched in skin cells by external application of 5-aminolevulinic acid (5-ALA). The delayed fluorescence is induced by pulses of green light and is inversely correlated with mitochondrial oxygen tension. The recent development of the CE-certified Cellular Oxygen METabolism Monitor (COMET) now enables the application of the PpIX-TSLT in the clinical setting. As this device allows the direct measurement of oxygen metabolism on the cellular level, it could be employed as a diagnostic tool for patients with sepsis. Thus far, the COMET has been employed in a pharmacological study (10) and in healthy controls (11, 12).

The primary objective of this study was to test how feasible the PpIX-TSLT measurements are in the acute phase of sepsis during the treatment at the critical care unit. The secondary objectives were to assess the distribution and the stability of PpIX-TSLT variables for single measurements and to identify potential covariates, in particular the patients' fluid status.

#### METHODS

#### Patient Sample

This study is an intermediate analysis of patients from the study Identification of cardiovascular and molecular prognostic factors for the medium- and long-term outcomes of sepsis (ICROS, DRKS00013347, and NCT03620409). Patients with sepsis were recruited at the intensive care units of the Jena University Hospital. The inclusion and exclusion criteria of the study are presented in **Table 1**. For PpIX-TSLT measurements, contraindications were: allergies to contents of the Alacare <sup>R</sup> plaster (photonamic, Wedel, Germany), porphyria, skin conditions aggravated by sunlight, or increased sensitivity to light. For BIA measurements contraindications were: electronic implants (e.g., pacemaker) or active prostheses. Study physicians obtained written informed consent from either the patient or the patient's legal proxy if the patient was incapacitated. The study was approved by the ethics committee of the Friedrich Schiller University Jena (5276-09/17).

#### PpIX-TSLT Measurements

PpIX-TSLT measurements took place within 3 ± 1 days after the onset of sepsis. A 4 cm² patch containing 5-ALA (Alacare <sup>R</sup> , photonamic, Wedel, Germany) was applied to the clavipectoral triangle at least 5 h before the planned measurement to ensure sufficient accumulation of PpIX. Before application, the skin was cleaned and prepared with an abrasive paste (skinPure <sup>R</sup> , Nihon Kohden, Rosbach, Germany). For 48 h after application, the skin was protected from light with an additional patch. The measurements were performed with the COMET measurement

**Abbreviations:** 5-ALA, 5-aminolevulinic acid; BIA, bioimpedance analysis; BIVA, bioimpedance vector analysis; COMET, Cellular Oxygen Metabolism Monitor; mitoDO<sup>2</sup> : average, average mitochondrial oxygen delivery; mitoDO<sup>2</sup> : maximum, maximum mitochondrial oxygen delivery; mitoPO<sup>2</sup> : baseline, baseline mitochondrial oxygen tension; mitoPO<sup>2</sup> : post, post-mitochondrial oxygen tension; mitoVO<sup>2</sup> : average, average mitochondrial oxygen consumption; mitoVO<sup>2</sup> : maximum, maximum mitochondrial oxygen consumption; PpIX, protoporphyrin IX; PpIX-TSLT, protoporphyrin IX-triplet state lifetime; SEM, standard error of measurement.

TABLE 1 | Inclusion and exclusion criteria for the study.

#### Inclusion criteria

Sepsis or septic shock meeting Sepsis-3 criteria (1)

Onset of first infection-caused organ dysfunction no longer than 72 h before enrolment

At least 18 years of age

Written informed consent from the patient, legal representative or proxy, or preliminary consent after consultation of an independent medical doctor

#### Exclusion criteria

Cardiac surgery in the last 12 months

Significant heart disease:

Endocarditis

Higher degree valve disorders (severe valvular heart disease, symptomatic aortic valve stenosis, moderate mitral regurgitation with reduced ejection fraction)

Congenital heart defect (e.g., transposition of the great arteries, tetralogy of Fallot, atrioventricular septal defect)

Hemodynamically relevant shunting heart defect

Reduced cardiac output (EF < 45% or 10% below norm\*) prior to sepsis

Pulmonary arterial hypertension prior to onset of sepsis Myocardial infarction 12 months prior to onset of sepsis History of heart transplantation

Cardiopulmonary resuscitation 4 weeks prior to onset of sepsis

History of pneumonectomy

Child C liver cirrhosis

Contraindications for transesophageal echocardiography and insufficient quality of transthoracic echocardiography

Terminal chronic kidney disease with dialysis

Sepsis within 8 months prior to onset of sepsis

Pregnancy/breast-feeding

Therapy limitation/do-not-resuscitate order

Remaining life expectancy <6 months due to other causes than sepsis

Prior participation in this study Participation in another interventional study

\*Left ventricular ejection fraction >52% in men, >54% in women according to the American Society of Echocardiography and the European Association of Cardiovascular Imaging (13).

system (Photonics Healthcare, Utrecht, Netherlands). Before the measurement session, the sensor was shielded from light and applied to the prepared skin. During one session multiple measurements with the following parameters were performed (see also **Figure 1**): In the first 30 s, the mitoPO<sup>2</sup> : baseline was measured. Thereafter, we applied pressure to the sensor for 45 s to inhibit the microcirculation of that part of the skin and to measure oxygen consumption (mitoVO<sup>2</sup> : maximum and mitoVO<sup>2</sup> : average). Finally, the pressure was released to evaluate the re-oxygenation (mitoDO<sup>2</sup> : maximum and mitoDO<sup>2</sup> : average) and post-re-oxygenation mitoPO<sup>2</sup> (mitoPO<sup>2</sup> : post). A single measurement took 105 s and was performed at least three times per session. We used two complementary sigmoid functions to fit the raw PpIX-TSLT signals of each single measurement:

$$\frac{K\_1}{1 + e^{(-B\_1 \times (\chi - M\_1))}} + \frac{K\_2}{1 + e^{(-B\_2 \times (\chi - M\_2))}} + Z \tag{1}$$

The estimation of these function parameters allows a direct inference of the PpIX-TSLT variables. For details please [see (12)].

We used a self-developed program (Halley) under MATLAB (MATLAB and Statistics Toolbox Release 2017a, The MathWorks, Inc., Natick, Massachusetts, United States) for data management, data preparation and PpIX-TSLT variable estimation.

#### Bioimpedance Vector Analysis

The patient's fluid status was evaluated using the seca medical Body Composition Analyzer 525 (seca Germany, Hamburg, Germany). For bioimpedance vector analysis (BIVA), raw impedance variables, resistance (R) and reactance (Xc), were standardized to body height (R/height, Xc/height) in meters (14). Two characteristics of the resulting bivariate vector were analyzed: the phase angle [arc tangent of (Xc/R) × 180◦ /π] as vector orientation and the vector length (square root of Xc<sup>2</sup> height + R 2 height). Especially the latter one is an indicator of the hydration status (14, 15) and has already been proven to be applicable in the critical care setting (16). Short vectors, resulting from decreased resistance, indicate high levels of hydration and can indicate the presence of edema secondary to low oncotic pressure or endothelial barrier dysfunction. Long vectors, resulting from increased resistance, are an indicator of a low hydration status, i.e., dehydration. Where feasible, body height, and weight were obtained by means of a measuring tape and the Seca bed scale 985 (Seca Germany, Hamburg, Germany), respectively. Otherwise, information from the medical history was used.

#### Descriptive Analysis

In descriptive analysis, means, standard deviations (SD), medians, as well as first and third quartiles (Q1/Q3) are reported. For categorical and dichotomous variables we report absolute and relative frequencies. The distribution of the PpIX-TSLT variables was assessed using histograms, Q-Q-plots, Shapiro-Wilk-Tests, and an estimation of kurtosis and skewness with corresponding standard errors (SE) and kurtosis excess (kurtosis – 3).

#### Stability of PpIX-TSLT Variables

To analyze the repeatability, i.e., stability, of PpIX-TSLT variables during one measurement session we took the following steps: First, descriptive mean differences between measurement pairs and corresponding p-values of the paired samples t-tests are reported. Second, Pearson correlation coefficients of all available measurement pairs are presented. Correlation coefficients can be interpreted in the following way: r < 0.10 negligible, r 0.1–0.39 weak, 0.40–0.69 moderate, 0.70–0.89 strong, 0.90– 1.00 very strong (17). Third, measurement pairs were analyzed using Bland-Altman plots to assess limits of agreement (LOA) (18, 19). In detail, the mean of a measurement pair is plotted against the difference between both single measurements (LOA = mean of the differences ± 1.96 × SD of the differences). The corresponding 95% confidence intervals (95%CI) for LOA were obtained using two-sided tolerance factors (20). At population level, 95% of PpIX-TSLT variable differences between single

PpIX-TSLT measurements within one session, according to our protocol, should lie within these LOA. Fourth, intra-class correlation coefficients (ICC) for all available measurement pairs using the two-way mixed effects analysis of variance (ANOVA) for single measures with absolute agreement were obtained (21). ICCs and corresponding confidence intervals can be interpreted as follows: ICC < 0.5 poor, 0.5 ≤ ICC ≤ 0.75 moderate, 0.75 ≤ ICC ≤ 0.90 good, and ICC > 0.90 excellent reliability (21). Finally, the standard error of measurement (SEM) as the square root of the mean square error term from repeatedmeasures ANOVA for all available measurement pairs (22) and the Minimum Detectable Difference (MDD, SEM × 1.96 × 2 1/2 ) are reported.

#### Exploratory Analysis of Potential Covariates

Potential covariates for PpIX-TSLT variables were analyzed parametrically and non-parametrically with Pearson correlation coefficients and Spearman's rank correlation coefficients, respectively. PpIX-TSLT variables from multiple single measurements were averaged before correlative analysis. Results were additionally visualized with scatterplots and regression lines from the simple linear regression models (PpIX-TSLT variables served as dependent variables). We considered the following variables: sex, age, BIVA variables (see above), PpIX-TSLT-associated variables (duration of 5-ALA application, average signal quality, room, sensor, skin, and body temperature, goodness of fit of the fitting procedure), physiological data (heart rate, systolic and diastolic blood pressure, SpO2, hemoglobin, fluid balance), and treatment-associated data (initial SOFA score, catecholamine status, ventilation status). Treatment-related variables were obtained from electronic patient records on ICU (Copra System, Berlin, Germany).

For statistical analysis we used SPSS Statistics 24 (IBM Corporation, Armonk, NY, USA) and R [Version 3.5.1, Vienna, Austria (23)]. We applied a significance level of 5% and report two-sided p-values.

#### RESULTS

#### Sample Description

**Figure 2** summarizes the study inclusion. Of 332 screened patients, 45 patients with sepsis were enrolled in the study. After excluding patients who withdrew consent, died before the measurement or had contraindications, PpIX-TSLT measurements were performed on 40 patients. 37 patients with reliable PpIX-TSLT measurements were included in the primary analysis. Demographic and clinical characteristics of the study sample are displayed in **Table 2**.

#### Primary Analysis

In 37 (92.5%) of the 40 included patients with sepsis, reliable PpIX-TSLT measurement sessions could be recorded. In the 3 patients with unreliable measurement sessions either the signal quality was too low and/or mitoVO<sup>2</sup> could not be induced consistently during multiple measurements. We did not observe TABLE 2 | Demographics and clinical characteristics (n = 37 patients).


For continuous measures median, first and third quartile (Q1/Q3) are displayed. For categorical variables number (n) and percent (%) are shown.

<sup>a</sup>Any duration of invasive ventilation. Non-invasive ventilation > 6 h.

<sup>b</sup>Dopamine > 5 µg/kg/min or epinephrine ≤ 0.1 µg/kg/min or norepinephrine ≤ 0.1 µg/kg/min.

<sup>c</sup>Dopamine > 15 µg/kg/min or epinephrine > 0.1 µg/kg/min or norepinephrine > 0.1 µg/kg/min.

any side effects of the 5-ALA application or the PpIX-TSLT measurement itself in any patient.

#### Secondary Objectives

#### Descriptive Statistics and Variable Distribution

The descriptive statistics of the PpIX-TSLT variables are displayed in **Table 3**. Additional information on variable distribution is provided in **Supplement S-1**. Shapiro-Wilk tests indicated that mitoPO<sup>2</sup> variables were normally distributed (p > 0.05). The p-values of the Shapiro-Wilk tests for the other PpIX-TSLT variables were significant (p < 0.01). All PpIX-TSLT parameters were positively skewed and kurtosis excess values ranged between −2.4 (mitoPO<sup>2</sup> : baseline, platykurtic) to 3.0 (mitoVO<sup>2</sup> : average, leptokurtic). After removing outlier values (values >Q<sup>3</sup> + 1.5 times the interquartile range, see also **Table 3**) none of the p-values of the Shapiro-Wilk tests reached significance (p > 0.05) indicating normally distributed values.

#### Stability of PpIX-TSLT Measurements

MitoPO<sup>2</sup> : baseline values showed a mean difference of 4.42 mmHg between subsequent measurements, resulting in a significant p-value of the corresponding paired t-test (**Table 4**). In addition, both mitoDO<sup>2</sup> variables differed significantly between


#### TABLE 3 | Descriptive statistics for PpIX-TSLT variables.

All PpIX-TSLT variables from one measurement session were averaged before descriptive analysis.

mitoPO2, mitochondrial oxygen tension; mitoVO2, mitochondrial oxygen consumption; mitoDO2, mitochondrial oxygen delivery.

Thresholds are defined as Q<sup>1</sup> – 1.5 × interquartile range and Q<sup>3</sup> + 1.5 × interquartile range.

SD, standard deviation; 95%CI, 95 percent confidence interval, Q1/Q3, first and third quartile.

TABLE 4 | Intra-session stability of the PpIX-TSLT variables (n = 36 patients with at least 2 reliable PpIX-TSLT measurements).


Mean difference (1mean), corresponding standard deviation (SD), and 95% confidence intervals (95%CI) for PpIX-TSLT measurement pairs are shown. Additionally, Pearson correlation coefficients (r), corresponding p-values (pr), and p-values of the paired t-tests (pttest) for PpIX-TSLT measurement pairs are displayed. Further, limits of agreement (LOA) with corresponding 95%CI and p-values are shown. P < 0.05 are printed in bold.

subsequent measurements (p < 0.01) with a mean difference of 1.07 (mitoDO<sup>2</sup> : maximum) and 0.76 (mitoDO<sup>2</sup> : average), respectively. MitoPO<sup>2</sup> : Post and both mitoVO<sup>2</sup> variables did not differ significantly between subsequent measurements. All variables were moderately (mitoDO2) to strongly (all others) correlated between subsequent measurements (**Table 4**). The limits of agreement and corresponding Bland-Altman-Plots for subsequent measurements are displayed in **Table 4** and **Supplement S-2**. The intraclass correlation coefficients for all PpIX-TSLT variables showed significant p-values (p < 0.001) and ranged between 0.652 (moderate) for mitoDO<sup>2</sup> variables and 0.805 (good) for mitoVO<sup>2</sup> variables (**Table 5**). Finally, the standard error of measurement (SEM) and Minimum Detectable Difference (MDD) of PpIX-TSLT variables are displayed in **Table 5**.

#### Correlative Analysis

Statistically significant associations between PpIX-TSLT variables and tested covariates are shown in **Table 6**. The descriptive statistics for BIVA-variables and potential covariates are shown in **Supplement S-3**. In summary, mitoPO<sup>2</sup> : baseline correlated positively with height-standardized vector length of BIVA. Maximum and average mitoVO2-variables showed correlation coefficients of 1. The same applies for mitoDO<sup>2</sup> variables. For this reason, only results for mitoVO<sup>2</sup> : average and mitoDO<sup>2</sup> : average are reported. MitoVO<sup>2</sup> variables were positively correlated with the goodness of fit (R 2 ) of the fitting function. MitoDO<sup>2</sup> variables were negatively correlated with the duration of 5-ALA application, the average signal quality during PpIX-TSLT measurement and body height. In addition, mitoDO<sup>2</sup> variables were positively correlated with the room temperature. Finally, the mean signal quality of one measurement was positively correlated with the duration of the 5-ALA application and negatively correlated with the room temperature. All other tested covariates showed no statistically significant correlation coefficients with PpIX-TSLT variables (data not shown). MitoPO<sup>2</sup> : baseline tended to differ between female (median: 69.95, Q1|3: 54.03 | 74.08) and male patients (median: 57.19, Q1|3: 44.36 | 69.25). In addition, we found significant sex differences between mitoDO<sup>2</sup> : average with higher values for females (median: 5.25, Q1|3: 3.45 | 9.98) compared to males (median: 3.73, Q1|3: 2.55 | 5.27, U = 84, p = 0.022). Neither the ventilation status nor the catecholamine dosage was significantly associated with any of the PpIX-TSLT variables. After adjusting for multiple testing using the Bonferroni-Holm method, only the p-value for the association between mitoVO<sup>2</sup> variables and the goodness of fit (R 2 ) reached significance (adjusted p-values not shown).

#### DISCUSSION

#### Feasibility of PpIX-TSLT Measurements of Mitochondrial Function in Patients With Sepsis

In this study, we report for the first time direct in vivo assessment of mitochondrial oxygen metabolism by PpIX-TSLT measurements in a cohort of patients with sepsis. Thus


TABLE 5 | Intraclass correlation coefficients (ICC) with corresponding 95% confidence intervals, standard error of measurement (SEM), and minimum detectable difference (MDD) for PpIX-TSLT variables (n = 36 patients with at least 2 reliable PpIX-TSLT measurements).

P < 0.05 are printed in bold.

ICC < 0.5 poor, 0.5 ≤ ICC ≤ 0.75 moderate, 0.75 ≤ ICC ≤ 0.90 good, ICC > 0.90 excellent (21).

TABLE 6 | Main findings of the correlation analyses for the PpIX-TSLT variables and potential covariates.


Spearman's rank correlation coefficient (ρ) with the corresponding p-value (p<sup>ρ</sup> ) and the Pearson correlation coefficient (r) with the corresponding p-value (pr) are shown. P < 0.05 are printed in bold.

mitoPO2, mitochondrial oxygen tension; mitoVO2, mitochondrial oxygen consumption; mitoDO2, mitochondrial oxygen delivery.

† Obtained from height-standardized resistance and reactance values.

far, only reports in healthy subjects or surgical patients have been published. In our cohort of 40 patients, PpIX-TSLT measurements yielded analyzable datasets in 92.5% of patients. Also, no side effects of the measurements were observed. We therefore conclude that PpIX-TSLT measurements with the COMET are feasible in patients with sepsis in the ICU setting. In our previous study, analyzable results were obtained from 75% of healthy subjects (12). We believed at the time that compliance problems concerning 5-ALA application may have contributed to this relatively low success rate. As the ICU offers a standardized environment, protocol adherence was very high. Hence, our results may confirm our assumption and stress the importance of controlling the duration of 5-ALA application. Of the three measurements that failed in this study, one was due to low signal quality. This may have been due to inadequate skin absorption of 5-ALA despite using a standardized protocol. In the other two cases, mitoVO<sup>2</sup> could not be induced. This phenomenon is not entirely understood and deserves future attention.

#### Distribution and Stability of PpIX-TSLT Variables

After removing outliers, all variables were normally or near-tonormally distributed. In our previous trial, PpIX-TSLT variables were also distributed near to normal (12). Therefore, we conclude both parametric and non-parametric analyses may be applicable for PpIX-TSLT variables.

All variables were moderately to highly correlated between replicate measurements. All ICCs of PpIX-TSLT variables showed significant p-values and ranged from moderate to good. Although t-tests showed significant differences in mitoPO<sup>2</sup> : baseline and both mitoDO<sup>2</sup> variables between iterative measurements, the corresponding effect sizes were low (d = 0.32, d = 0.27, and d = 0.27, respectively). Taken together, the stability of replicate measurements in one session can be seen as moderate to good. We nonetheless recommend multiple measurements in one session. Judging from the LOA and MDD, we can determine that an increase in mitoPO<sup>2</sup> : baseline of 35 mmHg or a decrease of 25 mmHg between measurements is probably due to incorrect measurement and the measurement should be repeated. Similarly, absolute changes of 2 mmHg/s for mitoVO<sup>2</sup> : average and 5 mmHg/s for mitoDO<sup>2</sup> : average could indicate incorrect measurements. All these results are very similar to values determined in our previous study of healthy controls (12).

#### Potential Covariates of PpIX-TSLT Measurements

We identified potential covariates of the PpIX-TSLT measurements. We demonstrated a positive association between mitoPO<sup>2</sup> : baseline and BIVA vector length, which in turn correlates negatively with hydration status (14, 16). Therefore, more hydrated patients (i.e., with edema), may show lower mitoPO2: baseline values. Thus, we recommend considering the presence of edema or conditions with low intravascular oncotic pressure as influence factors when employing the PpIX-TSLT in patients with sepsis. The duration of the 5-ALA plaster application was positively associated with signal quality. One should therefore perform the measurement using a standardized minimum duration of 5-ALA application. Furthermore, room temperature was associated with mitoDO<sup>2</sup> : average and signal quality. Signal quality could be improved by performing measurements in a cool environment if possible. The patient's sex posed another covariate. MitoPO<sup>2</sup> : average tended to differ between female and male patients. In addition, we found significant sex differences between mitoDO<sup>2</sup> : average. Both variables showed higher values for females compared to males. This result could indicate the need for sex-specific normal values of PpIX-TSLT variables. Interestingly, neither the status of ventilation nor the dosage of catecholamines showed significant associations with PpIX-TSLT variables.

#### Limitations

The generalizability of the results for our secondary objectives are limited by the relatively small number of patients (n = 37) and selective exclusion criteria (i.e., exclusion of patients with preexisting cardiac conditions). Furthermore, in correlative analysis for the identification of potential covariates we did not adjust for multiple testing. Thus, our results need to be confirmed in larger trials. Especially the potential influence of the patient's hydration status on PpIX-TSLT variables should be examined. A direct estimation of total body water using BIA may be useful but is strongly dependent on body weight and height. As measuring body weight accurately in intensive care patients is difficult, we restricted our analyses to the weight-independent BIA variables (BIVA). Due to the fluctuation of physiological variables in critically ill patients, we only analyzed the short-term stability of PpIX-TSLT variables, with iterative measurements within minutes. The long-term stability of PpIX-TSLT variables was not analyzed.

# CONCLUSION

We conclude that PpIX-TSLT measurements with the COMET are feasible in the critical care setting. Despite the moderate to good stability of the PpIX-TSLT variables using our protocol, we recommend the recording of multiple measurements during one session to increase the reliability of results. The determined limits of agreement and the minimum detectable differences may help to identify potential outlier measurements and additionally improve data quality. Future studies in larger cohorts of critically ill patients are needed to determine the clinical and diagnostic

# REFERENCES


relevance of the PpIX-TSLT using the COMET. Furthermore, healthy controls should be measured to generate normal values for the PpIX-TSLT variables. This study poses a first step toward an evidence-based approach in the assessment of tissue oxygenation in the intensive care unit.

# DATA AVAILABILITY STATEMENT

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

# ETHICS STATEMENT

The studies involving human participants were reviewed and approved by Ethics Committee of the Friedrich Schiller University Jena. The patients/participants provided their written informed consent to participate in this study.

# AUTHOR CONTRIBUTIONS

SC, CN, and PB: conception and design of the study. CN, PB, AP, KS, JG, and CS-W: performance of measurements. PB, CN, AP, KS, and JG: clinical data collection. PB, KS, and AP: data analysis and statistical analysis. CN, PB, AP, KS, JG, CS-W, and SC: drafting the manuscript for important intellectual content. CN, CL, PB, and SC: revising the manuscript prior to submission. All authors carefully reviewed and approved the manuscript.

# FUNDING

This study was funded by the Federal Ministry of Education and Research within the Centre for Innovation Competence Septomics (Research Group Translational Septomics, Grant 03Z22JN12 to SC). The funding source had no involvement in the study design, the collection, analysis, and interpretation of data, in the writing of the report, or in the decision to submit the article for publication.

# ACKNOWLEDGMENTS

We thank the study nurse team of the department of Anesthesiology and Intensive Care Medicine of the Jena University Hospital for their support.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.00757/full#supplementary-material


statement from the american heart association. Circ Res. (2016) 118:1960–91. doi: 10.1161/RES.0000000000000104


**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 Neu, Baumbach, Plooij, Skitek, Götze, von Loeffelholz, Schmidt-Winter and Coldewey. 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 Effect of β2-Adrenoceptor Agonists on Leucocyte-Endothelial Adhesion in a Rodent Model of Laparotomy and Endotoxemia

Mansoor Nawaz Bangash<sup>1</sup> \*, Tom E. F. Abbott <sup>2</sup> , Nimesh S. A. Patel <sup>2</sup> , Charles Johnston Hinds <sup>2</sup> , Christoph Thiemermann<sup>2</sup> and Rupert Mark Pearse<sup>2</sup>

*<sup>1</sup> Department of Critical Care & Anaesthesia, University Hospitals Birmingham NHS Trust, Birmingham, United Kingdom, <sup>2</sup> Centre for Translational Medicine & Therapeutics, William Harvey Research Institute, Queen Mary University of London, London, United Kingdom*

#### Edited by:

*Rudolf Lucas, Medical College of Georgia - Augusta University, United States*

#### Reviewed by:

*Massimo Collino, University of Turin, Italy Christian Lehmann, Dalhousie University, Canada*

#### \*Correspondence:

*Mansoor Nawaz Bangash mansoor.bangash@uhb.nhs.uk*

#### Specialty section:

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

Received: *13 January 2020* Accepted: *27 April 2020* Published: *21 May 2020*

#### Citation:

*Bangash MN, Abbott TEF, Patel NSA, Hinds CJ, Thiemermann C and Pearse RM (2020) The Effect of* β2*-Adrenoceptor Agonists on Leucocyte-Endothelial Adhesion in a Rodent Model of Laparotomy and Endotoxemia. Front. Immunol. 11:1001. doi: 10.3389/fimmu.2020.01001* Background: The β2-adrenoceptor agonist dopexamine may possess anti-inflammatory actions which could reduce organ injury during endotoxemia and laparotomy. Related effects on leucocyte-endothelial adhesion remain unclear.

Methods: Thirty anesthetized Wistar rats underwent laparotomy followed by induction of endotoxemia with lipopolysaccharide and peptidoglycan (*n* = 24) or sham (*n* = 6). Animals received dopexamine at 0.5 or 1 µg kg−<sup>1</sup> min−<sup>1</sup> (D0.5 and D1), salbutamol at 0.1 µg kg−<sup>1</sup> min−<sup>1</sup> , or saline vehicle (Sham and Control) for 5 h. Intravital microscopy was performed in the ileum of the small intestine to assess leucocyteendothelial adhesion, arteriolar diameter, and functional capillary density. Global hemodynamics and biochemical indices of renal and hepatic function were also measured.

Results: Endotoxemia was associated with an increase in adherent leucocytes in post-capillary venules, intestinal arteriolar vasoconstriction as well-reduced arterial pressure and relative cardiac index, but functional capillary density in the muscularis was not significantly altered. Dopexamine and salbutamol administration were associated with reduced leucocyte-endothelial adhesion in post-capillary venules compared to control animals. Arteriolar diameter, arterial pressure and relative cardiac index all remained similar between treated animals and controls. Functional capillary density was similar for all groups. Control group creatinine was significantly increased compared to sham and higher dose dopexamine.

Conclusions: In a rodent model of laparotomy and endotoxemia, β2-agonists were associated with reduced leucocyte-endothelial adhesion in post-capillary venules. This effect may explain some of the anti-inflammatory actions of these agents.

Keywords: adrenergic β<sup>2</sup> receptor agonists, dopexamine, inflammation, microcirculation, surgery

# INTRODUCTION

Complications following major gastrointestinal surgery have a significant impact on both short and long-term survival (1–3). Inotropic agents may have important effects on outcomes for this patient group (4). Dopexamine is a dopamine analog with agonist activity at β2-adrenoceptor and dopaminergic receptors. This spectrum of activity confers vasodilator actions in addition to chronotropic and mild inotropic effects (5). Dopexamine has been used to increase cardiac output and hence tissue oxygen delivery in several trials of peri-operative haemodynamic therapy (6, 7). Other cardiovascular effects of dopexamine may include improved tissue microvascular flow and oxygenation (8). Various groups have studied the effects of dopexamine in patients following major gastrointestinal surgery (6, 9), with promising results, although the findings of a recent large trial were inconclusive (7).

Investigators have previously demonstrated potent antiinflammatory effects of β2-adrenoceptor agonists (10–18), in particular dopexamine (19–22). However, it is unclear that β2-adrenoceptor agonism is responsible for the dopexaminedependent reduction of leucocyte-endothelial adhesion seen in several endotoxemia studies (19, 20). The findings of previous laboratory and clinical investigations suggest dopexamine may improve tissue microvascular flow and oxygenation (8, 19, 20, 23–26), and it is thought that these effects may account for much of the potential benefit of inotropic agents in the critically ill (27). However, in a previous laboratory study from our group, the haemodynamic actions of dopexamine infusion appeared to be less important than anti-inflammatory effects, including decreased plasma cytokine levels, modulation of neutrophil CD11b surface expression, and decreased pulmonary neutrophil infiltration (28). We sought to further clarify how leucocyte-endothelial adhesion under endotoxemia might relate to the β2-adrenoceptor agonist effects of dopexamine, and its effects on arterial pressure, cardiac output and the microcirculation.

We therefore investigated the effects of dopexamine on leucocyte-endothelial adhesion within the microcirculation. Our hypothesis was that, in a rodent model of laparotomy and endotoxemia, dopexamine would decrease leucocyte-endothelial adhesion in intestinal post-capillary venules, through β2 adrenoceptor mediated actions. The relative contribution of β2 adrenoceptor agonism to these effects was assessed by using the β<sup>2</sup> adrenergic agonist salbutamol as a comparator.

# MATERIALS AND METHODS

Thirty male Wistar rats (240–340 g) received a standard diet and water ad libitum before the experiments. All procedures were performed with institutional approval and in accordance with the United Kingdom Home Office Guidance on the Operation of the Animals (Scientific Procedures) Act 1986 under the project license PPL 70/6526. Anesthesia was induced by intraperitoneal injection of thiopental (120 mg kg−<sup>1</sup> ) and maintained with supplementary injections administered according to regular testing for limb withdrawal to a standard stimulus. Animals were placed on a warming mat to maintain a core temperature of 37 ± 0.5◦C. A tracheostomy was performed, following which a short section of polyethylene tubing (internal diameter, 1.67 mm) was inserted to maintain airway patency and to facilitate spontaneous respiration. The right carotid artery was cannulated to allow blood sampling and continuous monitoring of heart rate (HR) and mean arterial pressure (MAP). The left jugular vein was cannulated for drug and fluid administration.

A 2 cm midline incision was then made through the abdominal wall to expose the peritoneum. Following laparotomy, bowel was evacuated into a moist cotton receptacle. Blunt dissection was then performed to access the abdominal vasculature. After isolation from the vena cava, a 1.5 mm ultrasonic aortic transit time flow probe (MA1.5PRB; Transonic Systems Inc., Ithaca, USA) was placed on the infra-renal aorta to measure aortic blood flow allowing calculation of relative stroke volume and relative cardiac index. The bowel was then replaced in the abdominal cavity, except for a loop of ileum just proximal to the caecum. The exposed bowel was kept moist by the application of 0.9% saline drops through a pipette. The laparotomy incision above and below the exit of the terminal ileal loop from the abdomen was then closed with 5.0 vicryl to prevent excessive insensible fluid losses. The animal was maintained on a warming mat on an intravital microscopy platform and placed in the right lateral position so the ileal loop fell on to a raised section of the platform at the level of the laparotomy incision. The temperature of the raised section was thermostatically controlled at 37.5◦C to ensure the temperature of the exposed bowel was similar to the core temperature. This position did not interfere with the ability of the ultrasonic probes to measure aortic blood flow. Subsequently the bowel was covered with Saran wrap to prevent evaporative losses from its surface and maintain bowel microvascular integrity (29). This was followed by a 5 ml kg−<sup>1</sup> bolus of normal saline to replace insensible fluid losses and a 15 min stabilization period to allow microvascular flow to stabilize. A first set of arterial blood samples was then taken (see below), the volume being replaced with an equal volume of normal saline. Animals were allowed to stabilize for 15 min before being allocated randomly to one of five groups (sham, control, D0.5, D1, S).

Endotoxemia was induced over a 10 min period in four of five groups by administering 1 ml kg−<sup>1</sup> of a solution containing Escherichia coli lipopolysaccharide 0111:B4 (LPS, 1 mg ml−<sup>1</sup> ) and peptidoglycan (PepG, 0.3 mg ml−<sup>1</sup> ) intravenously, the sham group received 0.9% saline vehicle alone. In all groups this was followed by an infusion of 0.9% saline at 4.3 ml kg−<sup>1</sup> h −1 though different doses of dopexamine or salbutamol were added to the D0.5, D1, and S groups' infusion fluid. This resulted in dopexamine infusion rates of 0.5 and 1 µg kg−<sup>1</sup> min−<sup>1</sup> for groups D0.5 and D1, respectively, and a salbutamol infusion rate of 0.1 µg kg−<sup>1</sup> min−<sup>1</sup> in group S. This dose of salbutamol was selected as previous studies conducted in isolated guinea–pig tracheal preparations showed a 10-fold greater potency of salbutamol at the β2-adrenoceptor when compared with dopexamine (30). Intravital microscopy in the intestinal ileum was performed after 150 min, midway through resuscitation. It was not possible to measure global hemodynamics during this procedure. The experiment ended after 5 h of resuscitation when the heart and lungs were excised.

# Analysis of Plasma Lactate, Base Deficit, and Renal and Hepatic Function

Two hundred microliter of blood was taken at baseline and at the end of the experiment for measurement of plasma lactate concentration (Accutrend Lactate; Roche Diagnostics, Basel, Switzerland) and base deficit (Radiometer ABL77, Copenhagen, Denmark). A 1ml blood sample was also taken at the end of the experiment for measurements of urea, creatinine, alanine aminotransferase, and aspartate aminotransferase by a commercial veterinary laboratory (IDEXX Laboratories Ltd, Sussex, UK).

# MEASUREMENT OF AORTIC BLOOD FLOW

A 1.5 mm perivascular probe was applied with water-soluble sonicating gel and sited as described earlier. The probe was connected to a TS420 monitor (Transonic Systems Inc., Ithaca, USA), which was connected to a Powerlab/8SP monitoring system (AD Instruments). This allowed continuous recording of aortic blood flow and HR, and calculation of relative stroke volume and relative cardiac output (relative as infra-renal aortic blood flow is not equivalent to cardiac output). Aortic blood flow was indexed to body weight to provide a measure of changes in relative stroke volume index (SVI) and relative cardiac index (CI). Probe calibration was performed daily according to the manufacturer's instructions before experiments.

#### Intravital Microscopy (IVM)

Fifteen minutes before the midpoint of fluid resuscitation, 0.2 ml of 0.17 g L−<sup>1</sup> rhodamine 6G (Sigma-Aldrich, Gillingham, UK) was administered intravenously to enhance the visibility of leucocytes. The animal platform was then transferred to the stage of an intravital microscope. Fluorescence microscopy was carried out using an Olympus BX61W1 microscope (Carl Zeiss Ltd.) connected to an Olympus BXUCB lamp, Uniblitz VCMD1 shutter driver and DG4-700 shutter instrument. Recordings were captured using Slidebook 5.0 software (Intelligent Imaging TTL) and saved for later offline analysis. All images were taken at x40 magnification. Leucocyte rolling and adhesion (>30 s stationary) was quantified in ileal post-capillary venules: the course of microvessels of the ileal submucosal layer was followed from collecting venules (V1) to postcapillary venules (V3), the latter being selected for analysis. Vessel length and diameter was measured and recorded. Images were recorded for a minimum of 40 s. A further 0.2 ml of 250 mg kg−<sup>1</sup> ml−<sup>1</sup> of FITC labeled bovine albumin (Sigma-Aldrich) was then administered intravenously in order to measure functional capillary density (FCD) and arteriolar diameters: the course of microvessels of the ileal submucosal layer was followed from supply arterioles (A1) to pre-capillary arterioles (A3), the latter being selected for analysis. Vessel diameter was measured and recorded. Capillaries were identified in the circular and longitudinal layers of the ileum and images were recorded for a minimum of 40 s. These images were later analyzed offline. The platform was then removed from the microscope stage and observations continued as before.

Recordings of intravital videos were stored electronically. These files were later analyzed offline using Slidebook 5.0 Reader [Intelligent Imaging Innovations (3i)] by an observer blinded to the experiment groups. Leucocyte adhesion was quantified and indexed to endothelial surface area (mm<sup>2</sup> ), calculated from the diameters and lengths of the vessel segments studied and assuming cylindrical vessel geometry. Firmly adherent leucocytes were defined as those that did not move or detach from the endothelial lining within an observation period of 30 s. FCD was calculated as the total length of perfused capillaries indexed to the visualized area (mm−<sup>1</sup> ).

#### Statistical Analysis

Data were presented as Mean (SEM) unless expressed otherwise and specifically. Kolmogorov-Smirnov normality testing was performed for all groups. Normally distributed data were tested using one-way analysis of variance (ANOVA) for comparison across all groups at a given time point. Post-testing was performed with Bonferroni's tests. Occasionally when ANOVA revealed significant results but post-tests did not indicate which group was responsible, t-tests (with or without Welch's correction depending on the variance of data) were performed to gain additional insight to the data. When data were not normally distributed in at least one group for any measurement, data were expressed as median (IQR) and the Kruskal-Wallis test was used in place of one-way ANOVA with Mann Whitney post-tests (and a Bonferroni correction). Two-tailed paired t-tests were used to compare hemodynamics at baseline with those at other time points for animals within the same group. Data were analyzed with PrismGraph 4.0 (GraphPad Software, San Diego, USA). Significance was set at p < 0.05.

# RESULTS

Baseline characteristics and fluid management are described in (**Table 1** and **Supplementary Table 1**). There were no significant differences between groups regarding weight or volume of fluid received. Animals in the sham group required a slightly greater dose of thiopental to maintain anesthesia (**Supplementary Table 1**). There were no baseline differences in hemodynamics, base deficit, lactate or hematocrit. In the sham group, MAP and HR did not change significantly but CI and SVI increased progressively (**Table 1**, **Figure 1**, **Supplementary Table 1**, **Supplementary Figures 1, 2**). Compared with the sham group and baseline, controls had a significantly higher HR (p < 0.05) and a significantly lower SVI and CI at 5 h (**Table 1**, **Figure 1**, **Supplementary Figures 1, 2**). At this point control group plasma base deficit and lactate were increased compared with sham animals, the latter significantly (p < 0.05) (**Table 1**, **Supplementary Figure 3**). Compared to shams, there were more firmly adherent leucocytes (control: 703 ± 86 mm−<sup>2</sup> vs. sham: 186 ± 68 mm−<sup>2</sup> ; p < 0.001), and fewer


*There were no significant differences between groups in baseline hemodynamics or volumes of fluid administered. Over the course of the experiment in shams SVI significantly increased and CI also tended to increase, though HR and MAP did not change. In contrast HR significantly increased, while SVI and CI significantly decreased in control and dopexamine groups. MAP was significantly decreased, though this was not a consistent finding. HR also significantly increased in group S and although CI and SVI remained relatively stable MAP decreased significantly. A significantly lower lactate was seen in shams compared to controls. Data presented as mean (SEM) when all groups normally distributed, otherwise median (IQR) if* ≥ *1 group not normally distributed. One-way ANOVA (post hoc Bonferroni's test,* \**p* < *0.05,* \*\**p* < *0.01,* \*\*\**p* < *0.001 vs. controls). Paired t-tests of baseline vs. end experiment for mean changes (*\**p* < *0.05,* \*\**p* < *0.01,* \*\*\**p* < *0.001).*

FIGURE 1 | Relative Stroke Volume Index during 5 h of laparotomy and endotoxemia recorded every 30 min. Values not plotted for t180–t210 [animals were undergoing IVM at this time and aortic flow and HR could not be measured (therefore relative SVI could not be calculated)]. Relative stroke volume index in controls differed significantly for most of the experiment and until its end when compared to shams. However, no significant differences in relative stroke volume index were observed between controls and groups treated with dopexamine or salbutamol at any time. The mean change in relative stroke volume index from baseline to end experiment was also significant in all groups except salbutamol treated animals (also see Table 1). Data presented as mean (SEM). One-way ANOVA at each time point (Bonferroni's post-tests, \*\**p* < 0.01, \*\*\**p* < 0.001 vs. controls).

rolling leucocytes in the post-capillary venules of control animals (**Figure 2**, **Supplementary Figure 4**). Intestinal arteriolar diameters were reduced in control animals [control: 21 ± 2µm vs. sham: 39 ± 3µm; p < 0.01; **Figure 2** (mid)] although FCD in the muscularis and its component circular and longitudinal layers did not differ significantly from shams [**Figure 2** (right), **Supplementary Figure 4**]. Endotoxemia was associated with acute kidney injury but not liver dysfunction (**Figure 3**).

Dopexamine infusion had no significant effect on any haemodynamic parameters when compared to

FIGURE 2 | (Left) Numbers of adherent leucocytes per square mm of endothelium in ileal post-capillary venules at 2.5 h of laparotomy and endotoxemia (*n* = 6 all groups). Numbers of vessels observed per group ranged from 8 to 18. Sham, dopexamine, and salbutamol treated groups demonstrated significantly less adhesion than controls. Data presented as mean (SEM). One-way ANOVA (Bonferroni's post-tests, \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001 vs. controls). (Mid) Intestinal arteriolar diameters of the ileum at 2.5 h of laparotomy and endotoxemia (*n* = 6 all groups). Numbers of vessels measured per group ranged from 8 to 19. When compared to shams and unlike controls, ileal arteriolar diameters were not significantly reduced in dopexamine and salbutamol treated groups. Data presented as mean (SEM). One-way ANOVA (Bonferroni's post-tests, \*\**p* < 0.01 vs. shams). (Right) Intestinal functional capillary density in longitudinal layers of the ileal muscularis at 2.5 h of laparotomy and endotoxemia (*n* = 6 all groups). Number of images per group ranged from 16 to 25. Groups were significantly different with respect to longitudinal FCD at 2.5 h. Data presented as mean (SEM). One-way ANOVA *p* = 0.024 (no groups positive in post-tests, although *p* < 0.01 when comparing control and D1 group with unpaired *t*-test with Welch's correction).

significant acute kidney injury in controls. The degree of injury was significantly less in the D1 group. All data presented as mean (SEM). Plasma urea and creatinine: One-way ANOVA (Bonferroni's post-tests, \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001 vs. controls). ALT: One-way ANOVA, *p* = 0.0246 (no groups positive in post-tests).

controls, except for an increase in heart rate (**Table 1**, **Figure 1**, **Supplementary Table 1**, **Supplementary Figures 1, 2**). Similarly, dopexamine was not associated with any improvement in plasma lactate or base deficit in endotoxaemic animals (**Table 1**, **Supplementary Figure 3**). The major finding of this study was that at the mid-point of resuscitation dopexamine significantly reduced leucocyte adhesion in post-capillary venules when compared to controls (D0.5: 409 ± 65 mm−<sup>2</sup> , p < 0.05 vs. control; D1: 361 ± 66 mm−<sup>2</sup> , p < 0.01 vs. control) (**Figure 2**). Furthermore, dopexamine prevented the reduction in arteriolar diameter observed in control animals [**Figure 2** (mid)]. The effects of dopexamine on FCD were complex. There were significant differences between groups in longitudinal muscle FCD (one-way ANOVA p = 0.024) [**Figure 2** (right)], but post-tests did not show which group was responsible for this difference. However, isolated t-tests comparing each group against controls reveal that only the D1 group had a significantly reduced longitudinal FCD compared to controls (unpaired t-test with Welch's correction, p = 0.0034). Dopexamine had no effect on FCD in the circular layer of the muscularis at any dose. When comparing total muscularis FCD for all groups, differences fell outside of the limits of statistical significance (one way ANOVA p = 0.058) (**Supplementary Figure 4**). Regarding organ function, renal function was improved in the D1 group compared to controls, but not the D0.5 animals (**Figure 3**).

The infusion of salbutamol was associated with a similar pattern of hemodynamics to those observed in dopexamine treated animals (**Table 1**, **Figure 1**, **Supplementary Table 1**, **Supplementary Figure 2**), and there was also no improvement in indices of tissue perfusion (**Supplementary Figure 3**). Compared to controls, salbutamol significantly reduced leucocyte-endothelial adhesion (S: 365 ± 49 mm−<sup>2</sup> , p < 0.01 vs. control) (**Figure 2**). Salbutamol also appeared to prevent arteriolar vasoconstriction, but unlike dopexamine was not associated with any change in FCD in any layer of the muscularis [**Figure 2** (mid, right), **Supplementary Figure 4**] and had no effect on organ injury (**Figure 3**).

#### DISCUSSION

The principal findings of this experiment were that in a rodent model of laparotomy and endotoxemia, clinically relevant doses of dopexamine were associated with decreased leucocyteendothelial adhesion and reduced arteriolar constriction in the intestinal microvasculature. However, with the exception of an increase in heart rate, dopexamine infusion was not associated with any systemic cardiovascular effects and in particular, relative cardiac index was not improved. In the higher dose dopexamine group an amelioration of renal dysfunction as assessed by plasma creatinine was observed. Almost all these findings were replicated by the β2-adrenoceptor agonist salbutamol though salbutamol failed to improve renal function (28).

This study provides evidence that dopexamine modulates the inflammatory response by reducing leucocyte-endothelial adhesion. We have previously demonstrated that in addition to reducing the pro-inflammatory cytokine response, dopexamine may also reduce the expression of leucocyte surface integrins following endotoxemia. We have also previously demonstrated a reduction in neutrophil infiltration in the lung of dopexamine treated endotoxaemic rats (28). It is possible that these observations are linked, such that during endotoxemia, a dopexamine mediated reduction in surface integrin expression results in reduced leucocyte-endothelial adhesion and consequential neutrophil transmigration into the tissues. Although this experiment does not elucidate the cellular events that result in reduced leucocyte-endothelial adhesion, the similar effects of salbutamol suggest a β2-adrenoceptor mediated mechanism. In this regard it has been shown that β2- and nonβ2-adrenoceptor mediated elevations of cAMP reduce leucocyte adhesion (31, 32), whilst tonic activity of Protein Kinase A prevents β2-integrin activation (33). Other factors including, but not limited to, an amelioration of pro-inflammatory cytokines or drug effects on vascular endothelium may be of equal or greater relevance to these observed changes. Furthermore, the relevance of reduced leucocyte-endothelial adhesion to the effect of dopexamine on organ function is also unclear because although salbutamol reduced leucocyte adhesion in post-capillary venules it did not ameliorate renal injury.

The reduction in arteriolar vasoconstriction by dopexamine is consistent with the preservation of arteriolar diameters in villus arterioles and hepatic sinusoids observed in previous endotoxemia studies with dopexamine. In those studies, dopexamine treated endotoxaemic animals had an associated preservation of total organ and microvascular blood flow compared to untreated groups (23, 26). However, in our previous study we observed a reduction in ileal red cell flux during endotoxemia (28). Assuming the same effect occurred in this study, then the similar muscularis FCD in control and sham animals suggests that any perfusion defect (assessed by FCD) must occur in the mucosa, as has been shown in other studies using intravital microscopy (34). The non-significant trend to reduction in longitudinal FCD in dopexamine treated animals may then indicate a dopexamine mediated re-distribution of blood from the outer layer of the ileum to the hypoxia prone mucosa. In this regard it is interesting to note that although salbutamol and dopexamine had similar effects on arteriolar diameters, salbutamol did not show any tendency to produce this effect on longitudinal layer FCD, did not improve renal function and showed no tendency to a reduced plasma lactate either. These differences, despite the similarity of effects otherwise, are notable and warrant further study.

Previous clinical studies in major surgery including those by our own group have emphasized the role of cardiovascular optimisation in enhancing tissue oxygen delivery to reduce perioperative morbidity (7, 8). Early studies drove clinicians to suggest this approach was beneficial as it reduced potentially harmful tissue ischaemia (35). The importance of maintaining tissue perfusion is still supported by modern studies from perioperative medicine where MAP is clamped at higher levels and shown to reduce the incidence of post-operative morbidity (36). This may also relate to improvements in microvascular perfusion (37). However, we previously showed in surgical patients kept to a narrow MAP range that peri-operative strokevolume guided fluid management protocols with continuous 0.5 µg kg−<sup>1</sup> min−<sup>1</sup> dopexamine infusion produced improvements in tissue oxygenation but without beneficial effects on markers of systemic inflammation or organ dysfunction (8). We also showed in a rodent model that dopexamine 1–2 µg kg−<sup>1</sup> min−<sup>1</sup> brings about potent immunomodulatory effects that are associated with improved organ function despite MAP and microvascular flow being similar to controls (28). This suggests that under surgical conditions therapeutic benefit is achievable through modulation of the host response to tissue injury and that further increases in tissue oxygenation or blood flow when perfusion is already guaranteed are redundant. In support of this, an analysis of surgical trials shows that patients with higher levels of baseline systemic inflammation are more likely to develop surgical complications (38). Similarly patients with impaired preoperative microvascular function (who are known to have higher levels of inflammatory markers) are also more likely to suffer later complications (37, 39). Considering that previous studies have shown that surgical stress is associated with an upregulation of chemokines in the peritoneum and lungs, modulation of the host response to surgery becomes a potential explanation for the beneficial effects of dopexamine (40). However, the failure of salbutamol to improve renal function while producing a very similar spectrum of immune effects to dopexamine suggests that the mechanisms of renal protection with dopexamine are not necessarily only related to β2-adrenoceptor mediated effects on leucocyte-endothelial adhesion.

Differential abilities of dopexamine and salbutamol to increase cellular cAMP may explain divergent effects on renal function (41). This might be the case if dopaminergic receptor activation by dopexamine further increased cAMP levels above that provided by β2-adrenoceptor activation (5). The importance of increasing cAMP is that regulated cell death that causes tissue injury in acute kidney injury is inhibited by cAMP mediatedpathways, as is mitochondrial biogenesis which is required for enhanced recovery from cell stress (42, 43). On the other hand β2-adrenoceptor activation has also been shown to alter systemic metabolism to increase tissue tolerance of injury (44). Therefore, unexpected differential effects of the drugs at β2 adrenoceptors or in cAMP generation provide two mechanisms through which tissue damage can be minimized during the hours during and after emergency laparotomy and elective major abdominal surgery. This might also provide some explanation for the opposite findings in trials of β2-adrenoceptor agonism in acute respiratory distress syndrome (45). In those trials a week long infusion of higher doses of salbutamol were used to try and improve outcomes through a reduction in extravascular lung water but resulted in increased mortality. On the other hand in peri-operative medicine shorter-term infusions of similar agents are used to try and minimize tissue damage and organ dysfunction that can result from major surgery—trials in this setting with dopexamine have not shown any signal to increased mortality (7).

Several findings of this study are consistent with previous investigations. The haemodynamic effects of endotoxemia with or without dopexamine were replicated here and are in keeping with other studies (20, 28). Findings of intestinal arteriolar constriction are in keeping with the intense splanchnic vasoconstriction and rapid reduction of blood flow seen following endotoxemia and shock in rodents (20, 28, 46). This study, in keeping with other studies, found arteriolar constriction could be ameliorated by dopexamine (23). Similar studies have found an increase in adherent leucocyte numbers in intestinal or mesenteric post-capillary venules that could be ameliorated by dopexamine (19, 20). A significant increase in adherent leucocytes (reduced by dopexamine) and a decrease in rolling leucocyte numbers in post-capillary venules at two and a half hours would likely have resulted in leukopenia, as found in other studies (19–23). However, some findings of the experiment reported here, such as the failure of endotoxemia to decrease longitudinal and circular muscularis functional capillary density are not consistent with previous studies (20). These inconsistencies are likely to be the result of differences in the endotoxin serotype, dose, method of administration and fluid loading conditions of each experiment. Our study also appears to contradict the findings of Schmidt et al. regarding the role of β2-adrenoceptor agonism in leucocyte-endothelial adhesion (19). Importantly, our study design avoided the ablation of both endogenous and exogenous β2–adrenoceptor agonism that may account for differences in findings between the two studies.

Our study has several strengths. The use of IVM gave qualitative and quantitative data that is unobtainable from laser Doppler flowmetry studies (20). The nature of endotoxemia was modified, using peptidoglycan, which increases the generalizability of these findings outside of Gram negative septicaemia alone. Although the duration and nature of endotoxemia differed from our previous study, the model behaved in a similar fashion to our previous study with respect to hemodynamics, markers of perfusion and resultant organ dysfunction. With respect to biochemical markers of tissue perfusion, it is possible that the lack of statistical significance in the D1 group where lactic acidosis and base deficit were less severe (as in our previous study) is the result of smaller sample sizes. If correct, it is notable that salbutamol neither showed any signal to an amelioration of plasma lactate nor resulted in any amelioration of renal dysfunction. This could suggest there are additional important mechanisms of action of reducing organ injury that dopexamine possesses (as discussed above).

There are also limitations to the study performed. Although there were many similarities with our previous experiments, fundamental differences in design mean that it is not possible to be certain that the models behaved in an identical fashion. Secondly, although the use of IVM permitted direct visualization of the intestinal microcirculation, expected differences in FCD were not seen between shams and controls. This may have related to the mild severity of the model (note no significant hepatic dysfunction was observed in any group), to visualizing the intestinal microcirculation too early in the course of the experiment or even to the volume of fluid administered. In this regard the inability to observe changes in the microvascular bed over the entire course of the experiment was a weakness of this study. Furthermore, although arteriolar diameters and muscularis FCD were observed, the inability to measure centerline red cell velocity, mucosal FCD and mucosal inflow arteriolar diameters prevents a complete picture of the distribution of intestinal blood flow being made. Although reductions in leucocyte-endothelial adhesion were observed, it is not possible to ascertain the relative importance of this phenomenon to the reduction in organ injury seen in this or previous experiments. Derivation of the surrogates relative stroke volume and relative cardiac index from infra-renal blood flow is recognized in the literature. Nevertheless, it should be noted that cardiac index may have differed between groups and the effects of different doses of vasoactive drugs may have led to differing organ blood flows above the level of measurement, none of which could have been detected by an infra-renal flow probe. Regarding the use of dopexamine and salbutamol, dose equivalence was based on previous studies in isolated tracheal preparations (30). Although hemodynamics were similar between D1 and S groups suggesting the dose selection was probably correct, it is still not possible to be certain that the effect at the β2-adrenoceptor was identical for both drugs. This may be compounded by the fact that salbutamol is a mix of two enantiomers, adding further complexity to the comparison. In this regard the use of a selective β2-adrenoceptor antagonist to further disentangle the role of this drug effects will be useful in future studies. Finally, although the use of peptidoglycan increases the generalizability of these findings outside of Gram negative septicaemia, the choice of an endotoxin based model may still be criticized for lack of a true clinical correlate.

In summary, we present experimental evidence confirming that clinically relevant doses of dopexamine reduce leucocyteendothelial adhesion in the intestinal microvasculature and are associated with improved renal function at clinically relevant doses. As a consequence of our experiments some avenues warrant further research. The relationship between β2 adrenoceptor signaling and downstream effects on leucocyte CD11b expression, tissue tolerance mechanisms and inhibition of regulated cell death deserve further attention. The effect of dopexamine on microvascular recruitment and its relationship to cardiac index under differing fluid regimes and also the effect of dopexamine on the distribution of microvascular blood flows are two areas that also warrant further study given the disparity seen in results of our studies and others (8, 20, 23). Although peri-operative dopexamine use has been shown to be safe in randomized controlled trials and a Bayesian analysis of the OPTIMIZE trial suggested a high probability of superiority of treatment efficacy, a new randomized controlled trial of peri-operative optimisation using β2-adrenoceptor agonists including dopexamine is underway and will inform clinicians definitively regarding the role of peri-operative dopexamine and haemodynamic optimisation (7, 47, 48).

#### DATA AVAILABILITY STATEMENT

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

#### REFERENCES


#### ETHICS STATEMENT

The protocol was approved on 17/10/2011 under the project license number PPL 70/6526, by the AWERB (Animal Welfare and Ethical Review Body of Queen Mary University of London). All procedures were performed in accordance with the United Kingdom Home Office Guidance on the Operation of the Animals (Scientific Procedures) Act 1986.

#### AUTHOR CONTRIBUTIONS

MB carried out the in vivo studies and statistical analysis of all data. NP provided advice during in vivo studies and on study design. TA took offline measurements from intravital microscopy videos. CT participated in the design and coordination of the study and provided guidance throughout. RP conceived the study, participated in its design and co-ordination, and helped with statistical analysis. RP, CH, CT, and MB together drafted the manuscript. All authors read and approved the final manuscript.

#### FUNDING

This work was supported by research grants from a British Journal of Anaesthesia/Royal College of Anaesthetists Project Grant and an Intensive Care Society Young Investigator Award. NP was supported by a Kidney Research UK Post-Doctoral Fellowship (PDF4/2009). TA was supported by a jointly funded Medical Research Council and British Journal of Anaesthesia Clinical Research Training Fellowship. This work forms part of the research themes contributing to the translational portfolio of Barts and The London Cardiovascular Biomedical Research Unit, which was supported and funded by the National Institute of Health Research. RP is a National Institute for Health Research (UK) Research Professor.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.01001/full#supplementary-material

outcomes after surgery: a cochrane systematic review. Br J Anaesth. (2013) 111:535–48. doi: 10.1002/14651858.CD004082.pub5


an intravital microscopic study in the rat. Crit Care Med. (2001) 29:597–600. doi: 10.1097/00003246-200103000-00023


on clinical outcomes in acute respiratory distress syndrome (BALTI-2): a multicentre, randomised controlled trial. Lancet. (2012) 379:229–35. doi: 10.1016/S0140-6736(11)61623-1


**Conflict of Interest:** RP has received equipment loans from LiDCO Ltd and has performed consultancy work for Edwards Lifesciences and Massimo Inc. RP is a member of the associate editorial board of the British Journal of Anaesthesia. CT is CEO of William Harvey Research Limited, which is a CRO and has conducted contracted research in the area of critical care. CT is also Senior Associate Editor for the journal Shock.

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

Copyright © 2020 Bangash, Abbott, Patel, Hinds, Thiemermann and Pearse. 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.

# Csf2 Attenuated Sepsis-Induced Acute Kidney Injury by Promoting Alternative Macrophage Transition

Yiming Li <sup>1</sup> , Pan Zhai <sup>2</sup> , Yawen Zheng<sup>3</sup> , Jing Zhang<sup>1</sup> , John A. Kellum<sup>4</sup> and Zhiyong Peng1,4 \*

*<sup>1</sup> Department of Critical Care Medicine, Zhongnan Hospital of Wuhan University, Wuhan, China, <sup>2</sup> Department of Neurology, Hubei Province Hospital of Tradition Chinese Medicine, Wuhan, China, <sup>3</sup> Department of Urological Organ Transplantation, The Second Xiangya Hospital of Central South University, Changsha, China, <sup>4</sup> Center of Critical Care Nephrology, Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States*

#### Edited by:

*Lukas Martin, University Hospital RWTH Aachen, Germany*

#### Reviewed by:

*Basilia Zingarelli, Cincinnati Children's Hospital Medical Center, United States Zoltan Nemeth, Morristown Medical Center, United States*

> \*Correspondence: *Zhiyong Peng zn001590@whu.edu.cn*

#### Specialty section:

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

Received: *22 November 2019* Accepted: *02 June 2020* Published: *07 July 2020*

#### Citation:

*Li Y, Zhai P, Zheng Y, Zhang J, Kellum JA and Peng Z (2020) Csf2 Attenuated Sepsis-Induced Acute Kidney Injury by Promoting Alternative Macrophage Transition. Front. Immunol. 11:1415. doi: 10.3389/fimmu.2020.01415* Sepsis is a systemic inflammatory state that occurs in response to infection and significantly increases mortality in combination with acute kidney injury (AKI). Macrophages accumulate in the kidney after injury and undergo a transition from a proinflammatory (M1) phenotype to an alternatively activated (M2) phenotype that is required for normal repair. However, the specific signals that regulate the transition from the M1 to M2 phenotype *in vivo* are unknown. Here, we found an unexpected role of Colony stimulating factor 2 (Csf2) in controlling macrophage transition *in vitro* and in a mouse model of sepsis induced by cecal ligation and puncture (CLP). We first co-cultured human M1 macrophages with HK-2 cells and characterized cytokine/chemokine profiles via Luminex. Of the cytokines and chemokines that were overexpressed in medium from M1 macrophages cocultured with human kidney-2 (HK-2) cells compared with that from M1 macrophages cultured alone, Csf2 and IL6 showed the greatest increases. Csf2 was exclusively secreted by HK-2 cells but not by M1 macrophages. Furthermore, recombinant human Csf2 protein promoted transition of M1 macrophages to the M2 phenotype in a dose and time-dependent manner. The apoptosis and reactive oxygen species (ROS) release induced by M1 macrophages in HK-2 cells was attenuated after exposure to exogenous Csf2. In addition, the switch from the proinflammatory M1 phenotype to the M2 phenotype occurred via the p-Stat5 pathway, which was activated by Csf2. Importantly, we found that intraperitoneal injection of a Csf2-neutralizing antibody after CLP aggravated kidney injury and suppressed tubular proliferation, subsequently decreasing survival. However, administration of recombinant mouse Csf2 protein could rescue mice with sepsis. Together, our results indicate that Csf2 plays critical roles in regulating macrophage transition via activation of p-STAT5. These data form a foundation upon which new therapeutic strategies can be designed to improve the therapeutic efficacy of cytokine-based treatments for sepsis-induced AKI.

Keywords: sepsis, acute kidney injury, macrophage transition, Csf2, cytokine-based therapy

# INTRODUCTION

Sepsis is a complex clinical syndrome characterized by a systemic inflammatory response to infection. Acute kidney injury (AKI) is one of the most frequent and serious complications and contributes to high mortality. Up to 60% of cases with sepsis are complicated with AKI, and approximately half of AKI cases are related to sepsis (1–3). The disease burden from AKI results in an estimated \$10 billion of additional costs for the health care system in the United States (4), and severe AKI is associated with a mortality of 45–70% (2, 5–7).

Pathogen-associated molecular patterns (PAMPs) activate resident macrophages and kidney parenchymal cells, leading to the secretion of proinflammatory cytokines and chemokines that cause nonspecific tissue damage (8). An excessive inflammatory response and oxidative stress are considered to be the main mechanisms of septic AKI (9). Kidney monocytes have been implicated in pathogenesis and healing in mouse models of AKI; these cells infiltrate the injured kidney shortly after neutrophils in the early stage of injury, differentiate into macrophages, and contribute to tissue injury (10). Various experimental models suggest that extended inflammatory macrophage activation augments inflammation. Increased inflammatory cell infiltration and cytokine/chemokine release contribute to AKI (11), and many of these proinflammatory mediators can trigger cell death pathways. Furthermore, the removal of these activated cells and cytokines results in improved survival (12).

Indeed, prolonged inflammatory activity results in further tissue damage, ultimately inhibiting the reparative phase of injury resolution (13). The inflammatory milieu is reversed by the actions of reparative cells via secretion of anti-inflammatory cytokines and pro-proliferative signals. In disease states and injury, macrophages expand and play distinct but important roles in the immune response. They acquire a spectrum of phenotypes that range from highly inflammatory at the beginning of disease to highly reparative toward the resolution of injury. M1 macrophages exacerbate injury by producing proinflammatory cytokines (14), resulting in the recruitment of other inflammatory cells (15). M1 macrophages also secrete antipathogenic molecules, such as NO generated by inducible nitric oxide synthase (iNOS) and ROS, both of which can also induce mitochondrial damage and apoptosis (16). Activated M2 macrophages recognize and downregulate high levels of inflammatory proteins. M2 macrophages also promote the deposition of extracellular matrix by producing arginase, and they inhibit inflammatory immune cell activity via the secretion of resolvents, lipoxins, matrix metalloproteinases, and TGFβ, which target and cleave chemokines and chemoattractants (17). Moreover, depletion of macrophages in the reparative stages results in a significant increase in kidney injury biomarkers (18). However, these two disparate roles of macrophages, from inflammation and injury in the early stage to tissue repair and remodeling in the recovery stage, remain to be further studied. Therefore, identification of the trigger that promotes the transformation from the M1 phenotype to the M2 phenotype may provide therapeutic targets to relieve inflammation and promote tissue repair.

In this study, we showed that macrophage polarization from M1 to M2 in the kidney promotes repair and attenuates septic AKI. M1 macrophages were cocultured with human kidney cells to assess the interaction between these two cell types. Csf2 secreted by kidney tubular cells promoted the repair of tubular cell injury by inducing M0/M1-to-M2 transformation in vitro. To further determine how tubular cells promote M2 transformation, we examined the JAK-p-STAT5 pathway in cultured macrophages. Finally, a neutralizing Csf2 antibody and recombinant Csf2 protein were injected intraperitoneally to evaluate alternative macrophage activation and tubular cell injury during the repair stage after AKI. These findings not only address our knowledge gaps regarding the detrimental roles of M1 in AKI but also identify an unexpected role of Csf2 in regulating M1 to-M2 transformation. Modulating Csf2 signaling could improve the therapeutic efficacy of cytokine-based therapies in septic AKI.

#### MATERIALS AND METHODS

#### Cell Culture

The human kidney-2 (HK-2) and THP-1 cell lines were purchased from the Cell Bank of the Chinese Academy of Sciences. We followed the cell culture methods of Li et al. (19). Then, 10µg/mL LPS (Sigma, L3129) was used to treat HK-2 cells for 18 h. Next, 100 nM phorbol-12-myristate-13-acetate (PMA) was applied to induce THP-1 cell differentiation into M0 macrophages. As shown in **Figure 1A**, different methods were used to promote M1 or M2 differentiation. For the coculture experiments, the LPS-treated HK-2 cells were washed with PBS and then cocultured with M1 or M0 macrophages in 5% fetal bovine serum (FBS)-containing medium for several additional days. The Transwell chambers (0.4µm pore size) used for coculture were purchased from Thermo Fisher. HLA-DR is a marker of M1 macrophages. CD206, CD163 and IL-10 are M2 markers. CD68 is usually used to label M0 macrophages.

#### Cytokine and Chemokine Assays

Multiplex kits for measuring cytokines and chemokines were purchased from Bio-Rad (Austin, TX, USA). Plates were measured and analyzed with Bio-Plex Manager version 6.1 (Luminex, Austin, TX, USA), sold by Wayen Biotechnologies, Inc. (Shanghai, China). The 27-plex cytokine panel and 40-plex chemokine panel kits were used to measure the concentrations of cytokines and chemokines in the supernatant. Twelve cytokines overlapped between these two detection kits.

#### Cell Viability Assays

M1 macrophages were treated with recombinant human Csf2 (215-GM-050, Novus, USA). Cell viability was then tested with a Cell Counting Kit 8 (Dojindo Molecular Technologies, Japan) at 48 h after coculture. The absorbance at 450 nm was measured using a Thermo Scientific Microplate Reader.

#### Flow Cytometry Assay

At 48 h after coculture with M1 macrophages for 48 h, HK-2 cells were collected. The cells were fixed with 70% ethanol. The

*(Continued)*

FIGURE 1 | Slides were directly visualized using an Olympus fluorescence microscope at a 20X magnification, scale bar = 20µm. Representative images were from two independent experiments. (C) M0/M1/M2 macrophages were plated on 0.4 mm Transwell inserts and grown over a 2-day period. HK-2/LPS-HK2 cells were plated in 24-wells. The HK-2 cells were subjected to a viability assay. ROS release was detected in HK-2 cells by flow cytometry. LPS-HK2 cells were pretreated with LPS. Data are represented as the mean ± SD from *n* = 3 experiments. The significance of differences was tested using Student's *t*-test. \**p* < 0.05 and \*\**p* < 0.01. (D) Cytokines and chemokines in the supernatant after HK-2/LPS-HK-2/M1 culture or M1+ LPS-HK-2 coculture were assessed with a 27-plex cytokine panel and a 40-plex chemokine panel. Twelve cytokines overlapped between these two detection kits. Fifty-five cytokines and chemokines are shown in the heatmap. The Z-score was used to depict the variation between different samples. The culture media were pooled from three experiments and then subjected to Luminex.

fixed cells were washed with PBS and then treated with DCFH-DA according to the Reactive Oxygen Species Assay Kit (Yeasen, Shanghai, China). In the apoptosis experiment, Annexin V-FITC/PI (Annexin V-FITC/PI Apoptosis Detection Kit, Yeasen, Shanghai, China) was used to label apoptotic cells. The stained cells were analyzed using a Becton Dickinson flow cytometer (Franklin Lakes, NJ).

#### Cell Migration Assays

Transwell chambers (8-µm pore size, Corning, USA) were used to perform the cell migration assays. M1 macrophages (2 × 10<sup>5</sup> ) were seeded in the insert of the chamber. After coculture with LPS-treated HK-2 cells for 48 h, cells that migrated to the lower surface were stained with crystal violet (Sigma-Aldrich, St. Louis, MO) and photographed.

# ELISA

TIMP2 and Csf2 levels in the cell culture supernatant were measured with an ELISA kit (Bio-Swamp Life Science, Shanghai, China) according to the manufacturer's instructions. To detect IL-10 and TNF-α in kidney tissues, the kidney cortex was homogenized, and the protein concentration was determined using the Coomassie blue method. IL-10 and TNF-α levels in tissue and cell culture supernatant were measured with ELISA kits. Each sample was measured in triplicate.

# RNA Extraction and qRT-PCR

Total RNA was extracted from HK-2 cells, macrophages and kidney tissues using the YPH EASY spin tissue/cell RNA quick extraction kit (YPH, Beijing, China). Furthermore, cDNA was synthesized with the ReverTra Ace Kit (Toyobo, Osaka, Japan). The SYBR Green real-time polymerase chain reaction (PCR) kit was used to detect the expression of target mRNA. The human primer sequences were as follows: Csf2 sense, 5′ -TAC CTTTGTTGCAGCTGCTG−3 ′ , and anti-sense, 5′ -CACCACCC ACTGTTCGCTG−3 ′ ; CD163 sense, 5′ -TTTGTCAACTTGAGT CCCTTCAC−3 ′ , and anti-sense, 5′ -TCCCGCTACACTTGTTT TCAC−3 ′ ; and IL10 sense, 5′ -GACTTTAAGGGTTACCTGGG TTG−3 ′ , and anti-sense, 5′ -TCACATGCGCCTTGATGTCTG-3 ′ . The mouse primer sequences were as follows: Csf2 sense, 5 ′ -GGCCTTGGAAGCATGTAGAGG-3′ , and anti-sense, 5′ -GG AGAACTCGTTAGAGACGACTT−3 ′ ; CD163 sense, 5′ -ATGG GTGGACACAGAATGGTT-3', and anti-sense, 5′ -CAGGAGCG TTAGTGACAGCAG-3'; and IL10 sense, 5′ -GCTCTTACTGA CTGGCATGAG-3′ , and anti-sense, 5′ -CGCAGCTCTAGGAG CATGTG-3′ .

# Western Blot Analysis

The cells or kidney tissues were lysed, and lysates were prepared. Membranes were incubated with primary Abs specific for the following proteins: Csf2 (Abcam, ab54429), CD206 (Proteintech, 18704-1-AP), ARG1 (Proteintech, 16001-1-AP), t-Stat5 (Abcam, ab227687), p-STAT5 (Cell Signaling Technology, mAb #4322), and Jak2 (Proteintech, 17670-1-AP). Anti-glyceraldehyde 3 phosphate dehydrogenase (GAPDH, Proteintech, 60004-1-Ig) was used as a loading control.

#### Immunofluorescence

Immunofluorescence staining of the kidney was performed using paraffin or frozen sections as previously described (19). The sections were incubated with HLA-DR and CD206 antibodies (1:100). The slides were then exposed to a Cy3-conjugated secondary antibody for HLA-DR (red) and a fluorescein isothiocyanate (FITC)-conjugated secondary antibody for CD206 (green). The sections were visualized using fluorescence microscopy (Olympus, Tokyo, Japan).

# Sepsis-Induced CLP Model

C57BL/6 mice (male, 10–12 weeks old) were subjected to CLP to induce sepsis as previously described (20). Briefly, the animals were anesthetized with isoflurane. Under aseptic conditions, a 2-cm midline laparotomy was created below the diaphragm to expose the caecum. The cecum was ligated at the middle with a 5-0 silk suture and punctured twice with a 22-gauge needle. The caecum was then squeezed gently to extrude a small amount of feces through the perforation site. Animals were resuscitated with 1 ml of saline subcutaneously (s.c.) after CLP. We followed the methods of Li et al. (19). To minimize variability across experiments, the CLP procedure was always performed by the same investigator. All experiments were performed in accordance with the Animal Care and Use Committee of Wuhan University.

# Csf2 Antibody and Csf2 Protein Administration to Mice

For Csf2 neutralization experiments, five doses of 300 mg of anti-mouse Csf2-neutralizing antibody (BioXcell, BE0259-5MG) (13) or rat IgG isotype control (BioXcell, BE0089-5MG) in 100 µl of PBS were injected intraperitoneally daily starting 24 h after CLP. For rescue experiments, recombinant mouse Csf2 protein (2.5 µg/mouse/d, NOVUS) (21) in 100 µl of PBS was intraperitoneally administered five times after CLP. We used the 7-day survival rate as the endpoint for our survival study.

#### Evaluation of Renal Function

Mice were sacrificed at 48 h following Csf2 antibody or Csf2 protein administration. Heparinized blood was centrifuged to separate the plasma. Serum creatinine (Scr) and blood urea nitrogen (BUN) were detected using commercial kit reagents (Institute of Jiancheng Bioengineering, Nanjing, China).

#### H&E Staining

Kidneys were fixed in a 10% buffered formaldehyde solution. The tissue was processed routinely and paraffin embedded. Paraffin blocks were cut using a microtome at a thickness of 4µm, and the sections were deparaffinized, hydrated, and stained with hematoxylin and eosin (H&E). Photomicrographs were obtained randomly from the kidney renal cortex.

#### TUNEL Assay and Immunohistochemistry

Forty-eight hours following Csf2 antibody or Csf2 protein administration, kidney tissues were harvested and fixed with 4% paraformaldehyde for 30 min at room temperature. The TUNEL assay was performed using a TUNEL Apoptosis Detection Kit (Yeason, Shanghai, China) according to the manufacturer's instructions. Briefly, 4-µm kidney sections were deparaffinized and rehydrated. Then, the sections were incubated with TUNEL reagent mixture for 30 min at room temperature and washed with PBS three times for 5 min each time. 4 ′ ,6-Diamidino-2-phenylindole was used to stain nuclei. For immunohistochemistry, a Ki67 antibody (Proteintech, 27309-1- AP) was used to stain the tissue sections. The numbers of total cells and Ki67+ cells were determined.

#### Statistical Analysis

All tests were analyzed with SPSS version 20.0 or GraphPad Prism 8.0. Student's t-test was used for comparisons between the two groups. Survival data were analyzed using the log-rank test. In all comparisons, P < 0.05 indicated statistical significance. The data are presented as the mean and standard deviation.

# RESULTS

# Changes in Cytokines/Chemokines Produced by Macrophages and HK-2 in Response to LPS

An in vitro model based on differentiation of the THP-1 human monocyte cell line into M1 or M2 macrophages was first established. THP-1 cells grew in suspension with a rounded morphology, and they became attached to the plate in the presence of PMA. The schematic diagram shows that THP-1 cells differentiated into M1 and M2 macrophages (**Figure 1A**). Immunofluorescence staining for the macrophage markers HLA-DR (M1 marker) and CD206 (M2 marker) was used to confirm the phenotype of these macrophages. After LPS and IFN-γ treatment, the macrophages expressed high levels of HLA-DR (M1, red), while exposure to IL4 and IL13 resulted in the expression of CD206 (M2) (**Figure 1B**). Following activation, M0, M1 or M2 macrophages were cocultured with HK-2 cells pretreated with LPS for 18 h. Cell viability of HK-2 cells was significantly higher after coculture with M2 macrophages and lower after coculture with M1 macrophages. Similarly, the release of ROS was increased in the M1+ LPS-HK2 group compared to the M0+ LPS-HK2 group (**Figure 1C**).

Macrophages at various tissue sites have different cytokine profiles (22). Thus, we first measured a panel of 55 cytokines and chemokines in conditioned medium (**Figure 1D**). The absolute value of cytokines and chemokines were shown in **Supplementary Table 1**. Most proinflammatory cytokines, such as IL8 and MIF, were increased in HK-2 cells after exposure to LPS. Moreover, CXCL1, IL-1β and tumor necrosis factor (TNF) α were secreted at higher levels by M1 macrophages cocultured with HK-2 cells than by M1 macrophages cultured alone. Of the cytokines and chemokines that were overexpressed in medium from M1 + LPS-HK-2 cultures compared with medium from M1 macrophages cultured alone, IL-6 and Csf2 showed the greatest increases (147.5- and 114.3-fold, respectively, for M1 + LPS-HK-2 vs. M1). The changes in cytokine profiles between M1 macrophages cultured alone and macrophages cocultured with LPS-HK-2 cells confirmed that crosstalk existed between macrophages and HK-2 cells. These cytokine and chemokine array results also indicated that several cytokines might regulate the differentiation of macrophages. Csf2 was reported to facilitate the development of the immune system and promote defense against infections (23). These results indicate that macrophages interacted with kidney cells and that Csf2 was increased in the coculture medium.

#### Csf2 Was Derived Exclusively From HK-2 Cells and Decreased HK-2 Cell Apoptosis

Csf2 is considered a mediator by which T cells communicate with myeloid populations during tissue inflammation (24). To verify the specific role of Csf2 in sepsis, exogenous Csf2 was added to the medium in the M1 macrophage and HK-2 cell coculture system. The viability of HK-2 cells was decreased after exposure to M1 macrophages. However, additional Csf2 treatment rescued the viability of HK-2 cells (**Figure 2A**). The apoptosis rate was assayed to determine the antiapoptotic effect of Csf2. The total apoptosis percentage, including early apoptosis (Annexin Vpositive and PI-negative) and late apoptosis (Annexin V and PI double-positive), of HK-2 cells was increased in the M1 + LPS-HK-2 group (10.49% vs. 18.18%, LPS-HK-2 vs. M1 + LPS-HK-2), but the apoptosis percentage decreased to 7.76% after exposure to additional Csf2 (**Figure 2B**). Next, the Transwell assay was performed to investigate the ability of Csf2 to attract M1 macrophages. Fewer M1 macrophages migrated toward the HK-2 cells when Csf2 was added to the medium (**Figure 2C**). Moreover, the concentration of TIMP-2, a biomarker of kidney cell stress, was also reduced (**Figure 2D**). We next measured the concentration of Csf2 in this coculture system. We found that Csf2 secretion increased significantly in a time-dependent manner (**Figure 2E**). This increased Csf2 could be derived from either HK-2 cells, M1 macrophages, or both. To determine the source of Csf2, qRT-PCR was performed to detect the relative expression of Csf2 in HK-2 cells and M1 macrophages. Interestingly, Csf2 was mainly secreted by HK-2 cells but not M1 macrophages (**Figures 2F,G**). This increase in Csf2

(Annexin V-FITC+) was shown in the right panel. Three independent experiments were performed. \**p* < 0.05 and \*\**p* < 0.01; Student's *t*-test. (C) M1 macrophages were seeded in the upper chamber. After coculture with LPS-treated HK-2 cells for 48 h, M1 macrophages that migrated to the lower chamber were stained with crystal violet. Representative photos were obtained from three independent experiments at a 200× magnification. (D) The concentration of TIMP2 in M1and LPS-HK-2 culture media with or without Csf2 was detected by ELISA. (E) The concentration of Csf2 in the medium of LPS-stimulated HK-2 cells cocultured with M1 macrophages was detected by ELISA. Three independent experiments were performed. \**p* < 0.05 and \*\**p* < 0.01; Student's *t*-test. (F,G) The mRNA of Csf2 was measured independently in HK-2 cells and M1 macrophages after coculture for the indicated times. Data are the mean ± SD of 3 separate experiments. The significance of differences was tested using Student's *t*-test. \**p*<0.05. (H) Western blot analysis was used to detect the content of Csf2 in HK-2 cells after coculture with M1 macrophages. Three independent experiments were performed.

expression was confirmed by Western blotting in HK-2 cells after coculture with M1 macrophages (**Figure 2H**). Taken together, these findings clearly demonstrated that Csf2 could decrease HK-2 cell apoptosis and improve cell viability and that Csf2 was derived exclusively from HK-2 cells.

# Csf2 Mediated Macrophage Transition Toward an M2 Phenotype via the p-STAT5 Pathway

To study the effect of Csf2 on macrophages, M1 macrophages were treated with Csf2 for 72 h. We detected changes in the expression of HLA-DR (red, M1 macrophage) and CD206 (green, M2 macrophage). CD206 fluorescence was increased at the cell membrane of M1 macrophages (**Figure 3A**). Among the different doses of Csf2, the mRNA expression of the M2 phenotype markers CD163 and IL-10 was increased, with the highest increase at a dose of 25 ng/ml (**Figure 3B**). Additionally, immunofluorescence staining showed that the expression of CD206 on the M1 macrophage surface increased over time following coculture with HK-2 cells after Csf2 exposure (**Figure 3C**). We then wondered whether Csf2 had a similar effect on M0 macrophages. The number of CD68 (red, M0 marker) and CD206 (green, M2 macrophage) double-positive cells increased in a dose-dependent manner in response to Csf2 (from 0.5 to 25 ng/ml, **Figure 3D**). This finding indicates that M1 macrophages differentiated into M2 macrophages in the presence of Csf2.

Macrophage reprogramming is mediated by the JNK, PI3K/Akt, Notch, JAK/STAT, TGF-β, and TLR/NF-κB pathways (25). The JAK-STAT pathway plays an important role in regulating macrophage transformation (26, 27). To determine whether the JAK-STAT signaling pathway was involved in tubular cell-mediated macrophage activation, we analyzed STAT5 signaling pathways in M1 macrophages at multiple time points after Csf2 treatment. Phosphorylated STAT5 levels were much higher in macrophages after 48-72 h of Csf2 exposure (**Figure 3E**). Moreover, the M2 markers CD206 and ARG1 were also increased. These findings support the hypothesis that M1 macrophages shift toward an M2 phenotype through the STAT5 pathway when exposed to mediators such as Csf2.

#### Effect of Csf2 on Kidney Function and Survival Rates in a Sepsis Model

A previous study showed that CLP minimally affected survival in Csf2−/<sup>−</sup> mice while markedly reducing survival in wildtype mice (28). We hypothesized that the Csf2 antibody may aggravate AKI and decrease survival in a sepsis model. To test our hypothesis, a Csf2-neutralizing antibody or recombinant mouse Csf2 protein was intraperitoneally injected five times starting at 24 h after CLP. As shown in **Figure 4A**, five doses of antimouse Csf2-neutralizing antibody or recombinant mouse Csf2 protein were injected intraperitoneally. Serum was harvested 48 h following Csf2 antibody or Csf2 protein administration. Scr and BUN levels in the CLP group were increased compared with those in the SHAM group. Notably, intraperitoneal injection of the Csf2 antibody significantly exacerbated renal dysfunction, as evaluated by Scr and BUN levels. However, Scr and BUN levels were decreased after Csf2 injection compared with IgG isotype injection (**Figures 4B,C**). The survival rate decreased to 15.8% at 7 days after Csf2 antibody injection, compared with 36.8% after Csf2 isotype injection. Surprisingly, injection of Csf2 after CLP significantly reduced mortality (**Figure 4D**). These data revealed that blockade of Csf2 after CLP aggravated kidney injury and decreased survival and that Csf2 could rescue sepsis in mice.

# Blockade of Csf2 After CLP Inhibited Kidney Macrophage Transition and Reduced Tubular Cell Proliferation

Our data indicate that Csf2 could promote the transformation from M1 to M2 macrophages in vitro. Therefore, we hypothesized that blocking Csf2 signaling may inhibit macrophage transition in the kidneys of septic mice. To test whether Csf2 signaling also regulates macrophage transition in vivo, we first tested the expression of markers of M1 and M2 macrophages in kidney tissue. Double HLA-DR/CD206 immunofluorescence staining indicated that CD206 (red, M2 marker) expression was decreased after Csf2 antibody injection (**Figure 5A**), which indicated a decline in kidney macrophage transition toward the M2 phenotype after neutralizing antibody administration. In addition, the M2 markers CD163 and IL-10 were significantly reduced in the kidney cortex following Csf2 antibody treatment compared with isotype antibody treatment (**Figures 5B,C**), whereas the mRNA expression of Csf2 remained unchanged in these two groups (**Figure 5D**). Western blotting was used to detect the expression of p-STAT5 and Jak2. p-STAT5 and Jak2 activation decreased in the kidney cortex after anti-Csf2 antibody administration (**Figure 5E**).

To better and more objectively characterize the effect of Csf2 on M2, we detected the concentrations of IL-10 and TNFα in kidney tissue. As shown in **Figures 5F,G**, we found that the concentration of TNFα was much higher in the CLP + Ab group than in the CLP + ISO group, while the concentration of IL-10 was decreased following Csf2 antibody administration. The TUNEL assay was performed to detect the apoptosis rate. Kidneys from mice treated with the Csf2 neutralizing antibody showed a marked increase in the number of TUNEL-positive tubular epithelial cells (**Figure 5H**). We then attempted to confirm the histopathological damage among the different groups. H&E staining was performed on kidney tissue slices. H&E staining showed that tubular epithelial swelling and brush border injury were even worse in the CLP + Csf2 Ab group than in the CLP + ISO group (**Figure 5I**). AKI with proximal tubule death is usually followed by a wave of tubular proliferation, peaking at 48 h after injury, to restore tubular cell mass. The surviving epithelial cells have an equivalent capacity for repair (29, 30). Ki67, a marker of proliferation, was detected to characterize the repair process. The percentage of Ki67-positive cells was significantly decreased by Csf2–neutralizing antibody injection compared with isotype antibody injection after CLP (**Figure 5J**). These findings support our hypothesis that the Csf2 antibody significantly inhibited macrophage transition toward

FIGURE 3 | macrophages exposed to different doses of Csf2. Data are the mean ± SD of three separate experiments. The significance of differences was tested using Student's *t*-test. \**p* < 0.05. (C) Representative immunofluorescence images of M1 macrophages showing the expression of HLA-DR (red, M1 marker) and CD206 (green, M2 marker) after coculture with LPS-HK-2 for 4 days. Slides were directly visualized using an Olympus fluorescence microscope at a 20X magnification, scale bar = 20µm. (D) Representative immunofluorescence images of M0 macrophages showing the expression of CD68 (red, M0 marker) and CD206 (green, M2 marker) following Csf2 stimulation at different doses for 48 h. Slides were directly visualized using an Olympus fluorescence microscope at a 20X magnification, scale bar = 20µm. (E) Western blot analysis was used to detect CD206 (M2 marker), ARG1 (M2 marker), total STAT5 and phosphorylated STAT5 in M1 macrophages at the indicated times. Three independent experiments were performed.

the M2 phenotype via the JAK2-STAT5 pathway and reduced tubular cell repair.

#### DISCUSSION

AKI is a global health problem with high morbidity and mortality (31). Macrophages are highly heterogeneous in terms of origin and function. Different subpopulations play various roles at different stages in the course of renal disease. Accumulating evidence has suggested that macrophages shift from an M1 like proinflammatory state during the early phase after injury to an M2-like reparative state during the tubular recovery phase. The signals that instruct macrophages to alter their gene expression profiles and promote tubular repair after septic AKI are unknown. Here, we revealed the molecular mechanism by which kidney tubular cells induce macrophage transformation. We found that Csf2 derived from tubular cells induced M0/M1 to-M2 differentiation and subsequently played a protective role in sepsis-induced AKI. Our findings suggest a novel therapeutic target for septic AKI.

Csf2 was one of the top 2 differentially expressed cytokines in the M1 + LPS-HK-2 group compared with the M1 group. Csf2 has been reported to regulate complement- and antibody-mediated phagocytosis, antigen presentation, leukocyte chemotaxis, microbicidal capacity, dendritic cell homeostasis, and adhesion (32, 33). Csf2 is produced by a variety of cells, such as fibroblasts, macrophages, endothelial cells, DCs, neutrophils, T cells, and tissue-resident cells, during inflammatory/autoimmune reactions. Additionally, Csf2 has various effects on the regulation of myeloid populations, including survival, activation, differentiation, and mobilization (34, 35). Csf2-dependent inflammatory pathways in monocytes (and macrophages) are likely to be critical for the purported

FIGURE 5 | and CD206 (M2 marker, green) after neutralizing Csf2 antibody or isotype antibody injection. Data are shown as the mean ± SD of 6 mice per group per experiment and were pooled from two independent experiments. (B–D) The indicated M2 markers (CD163 and IL10) and Csf2 were assessed using qRT-PCR. Data are shown as the mean ± SD of six mice per group per experiment, and two independent experiments were performed. \**p* < 0.05; two-tailed unpaired Student's *t*-test. (E) Western blotting was utilized to detect total STAT5 and phosphorylated STAT5 in the kidney cortex. Six mice were used per group per experiment, and two independent experiments were performed. (F,G) The concentrations of TNFα and IL10 in the kidney cortex were assayed by ELISA. Data are shown as the mean ± SD of 6 mice per group per experiment. (H) The apoptosis rate was analyzed in the kidneys of mice by TUNEL. Data are shown as the mean ± SD of 6 mice per group per experiment. (I) The collected kidneys from the four groups were stained with H&E. Scale bars: 20µm. Kidney histology from CLP mice showing significant tubular epithelial swelling and brush border injury. Data are shown as the mean ± SD of six mice per group per experiment and were pooled from two independent experiments. (J) Photomicrograph of kidney tissue sections immunohistochemically stained for Ki67. Dense brown nuclear immunohistochemical staining indicates Ki67-positive status. Quantification of Ki67-positive tubular cells in the outer medulla. Data are shown as the mean ± SD of six mice per group per experiment and were pooled from two independent experiments, \*\**p* < 0.01; two-tailed unpaired Student's *t*-test.

role of Csf2 in inflammation, autoimmunity and host defense (36). However, the role of Csf2 in sepsis-induced AKI remains largely unknown. We found that administration of Csf2 in the coculture system decreased the apoptosis rate of HK-2 cells. Furthermore, fewer M1 macrophages migrated to the lower chamber, and the AKI biomarker TIMP-2 was also decreased in the culture medium. These results demonstrate a protective role of Csf2 in LPS-induced kidney injury. Basal levels of Csf2 are low under homeostatic conditions but can be quickly elevated during infection or inflammation. Modulating the functional homeostasis of macrophages can be of great benefit for sepsisinduced AKI (37). Therefore, it is reasonable to speculate that Csf2 could play a central role in mitigating sepsis-induced AKI.

Surprisingly, the source of Csf2 in our model system was exclusively secretion by HK-2 cells, not macrophages. Furthermore, Csf2 induced the M2 transition in a dosedependent manner at doses up to 25 ng/ml, and 3 days were required for M1-M2 transition after exposure to Csf2 in vitro. In Fleetwood et al.'s study, Csf2-dependent murine bone marrowderived macrophages (BMM) produced significantly more TNFα and IL-6 compared with Csf1-dependent murine BMM. Based only on increased expression of pro-inflammatory cytokines, Csf2-treated monocytes/macrophages after exposure with LPS produce more cytokines have been termed "M1-like" (38). However, Csf2 is often used to generate DCs populations in BMM (39–41), and DCs express TLR2 and TLR4 (42, 43). No wonder that Csf2-dependent BMM secreted more proinflammatory cytokines after exposure to LPS. In our study, M1 macrophages polarization was induced with THP-1 cells by incubation with LPS and IFN-γ according to routine protocol (44, 45). After stimuli macrophage expressed the typical M1 makers, and Csf2 induced the transformation from M1 to M2. These results are consistent with the pathological process, in which proinflammatory macrophages (M1) predominate during the early injury phase (first 24–48 h), when tubular apoptosis is prominent, and local macrophages begin to express M2 markers during the tubular repair phase (3–5 days), when tubular epithelial cells are growing and remodeling the damaged nephron (46, 47).

Disturbances in macrophage plasticity might compromise immune responses and lead to the development of disease conditions (25). The phenotypic changes and functions of macrophages are regulated by various signaling pathways. Transcription factors, including MAPK, JNK, and NF-κB, are involved in M1 programming (48). PPARγ-deficient macrophages are resistant to M2 polarization and promote insulin resistance (49). IRF, C/EBPβ and STAT regulate the M2 phenotype (26). The core factors that control macrophage phenotype and function in sepsis-induced AKI are still unclear. Our results showed that the Jak2-STAT5 pathway was involved in tubular cell-mediated macrophage transition. The Csf2 receptor, a dodecamer structure, was activated after Csf2 treatment, recruited Jak2 and led to downstream activation of STAT5, thus promoting the M1 to-M2 transition. The Jak2-STAT5 pathway was activated at 24 h and reached a peak at 48 h after Csf2 treatment. We identified tubule-derived Csf2 as a STAT5 activator mediating the macrophage transition from a proinflammatory to a proreparative phenotype.

Because Csf2 promotes the transition from the M1 phenotype to the M2 phenotype in vitro, we also examined whether a Csf2 antibody or recombinant Csf2 could contribute to macrophage differentiation in a murine sepsis model. Csf2 blockade through antibodies strongly inhibited M1 macrophage differentiation and increased mortality in CLP-induced septic kidney injury. Consistent with the data obtained in mice subjected to injection of a Csf2-neutralizing antibody, Csf2 protein administration improved the survival rate, suggesting that Csf2 signaling may promote the transition to M2 macrophages. M2 macrophages attenuate sepsis-induced AKI by upregulating IL-10 expression and suppressing TNF-α secretion. Moreover, the reparative process in kidney injury was inhibited following administration of the Csf2 antibody. Our result is also consistent with a randomized controlled trial of 39 patients with sepsis who had immune dysfunctions: Csf2 therapy in these patients was safe and effectively restored the immunocompetence of monocytes (50).

We have sufficient supporting evidence to propose that tubule cell-secreted Csf2 induces macrophage transformation and protects against AKI, although there are some limitations to our study. Previous research has reported that the production of Csf2 is induced by proinflammatory cytokines such as IL-1α, IL-1β, IL-12, and TNF and inhibited by IL-4, IFN-γ, and IL-10 (51, 52). In our study, Csf2 was increased at an early time point (18 h) in the kidney, but the stimulus resulting in Csf2 overexpression remains unknown. As Csf2 may serve as a potential therapeutic target for sepsis-induced AKI, host defense parameters, including circulating cytokines and the bacterial load, should be evaluated in a CLP model following Csf2 treatment. Moreover, whether

other organs are affected by Csf2 treatment is unknown and must be further investigated.

#### CONCLUSIONS

In summary, we investigated the regulatory function of Csf2 in sepsis-induced AKI and found that Csf2 secreted by HK-2 cells could promote macrophage transition toward the M2 phenotype via the p-STAT5 signaling pathway. Activated proinflammatory macrophages became polarized to the M2 phenotype, which dampens the proinflammatory response. Csf2 treatment reduced tubular cell apoptosis and subsequently improved AKI and survival (**Figure 6**). Csf2-mediated macrophage transition is an important mechanism of renal epithelial cell repair after kidney injury and could provide an alternative method for the treatment of septic AKI in the future.

#### 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 Animal Care and Use Committee of Wuhan University.

#### REFERENCES

1. Yegenaga I, Hoste E, Van Biesen W, Vanholder R, Benoit D, Kantarci G, et al. Clinical characteristics of patients developing ARF due to sepsis/systemic

#### AUTHOR CONTRIBUTIONS

YL: conceived the project, designed the project, extract and analyzed data, drafted the manuscript, and approved the final manuscript. YZ: drafted the part of the discussion and background of the manuscript. YL, PZ, and JZ: conducted the experiments. JK and ZP: designed the project, edited the manuscript, and approved the final version. All authors contributed to the article and approved the submitted version.

#### FUNDING

This work was supported by the National Natural Science Foundation of China (Nos. 81772046, 81560131), and the Health Commission of Hubei Province (No. WJ2017Z008).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.01415/full#supplementary-material

Supplementary Table 1 | The absolute value of cytokines and chemokines in the supernatant.

inflammatory response syndrome: results of a prospective study. Am J Kidney Dis. (2004) 43:817–24. doi: 10.1053/j.ajkd.2003.12.045

2. Uchino S, Kellum JA, Bellomo R, Doig GS, Morimatsu H, Morgera S, et al. Acute renal failure in critically ill patients: a multinational, multicenter study. JAMA. (2005) 294:813–8. doi: 10.1001/jama.294. 7.813


in dendritic cells but not in macrophages. Nat Commun. (2014) 5:3039. doi: 10.1038/ncomms4039


placebo-controlled multicenter trial. Am J Respir Crit Care Med. (2009) 180:640–8. doi: 10.1164/rccm.200903-0363OC


#### **Conflict of Interest:** JK discloses grant support and consulting fees from Astute Medical.

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

Copyright © 2020 Li, Zhai, Zheng, Zhang, Kellum and Peng. 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.