# IMMUNE CELL INTERACTIONS WITH TARGET CELLS IN PHYSIOLOGICAL AND PATHOLOGICAL CONDITIONS OF THE NERVOUS SYSTEM

EDITED BY : Antje Kroner, Sven G. Meuth and Craig Stephen Moore PUBLISHED IN : Frontiers in Immunology, Frontiers in Neurology and Frontiers in Molecular Neuroscience

#### Frontiers eBook Copyright Statement

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

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

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

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

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

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-969-4 DOI 10.3389/978-2-88963-969-4

#### 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

# IMMUNE CELL INTERACTIONS WITH TARGET CELLS IN PHYSIOLOGICAL AND PATHOLOGICAL CONDITIONS OF THE NERVOUS SYSTEM

Topic Editors: Antje Kroner, Medical College of Wisconsin, United States Sven G. Meuth, University Hospital Münster, Germany Craig Stephen Moore, Memorial University of Newfoundland, Canada

Citation: Kroner, A., Meuth, S. G., Moore, C. S., eds. (2020). Immune Cell Interactions With Target Cells in Physiological and Pathological Conditions of the Nervous System. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-969-4

# Table of Contents


Ashley M. Cowie, Bonnie N. Dittel and Cheryl L. Stucky


Jacqueline A. Quandt, Pierre Becquart, Emily Kamma and John Hallenbeck

*66 Central Nervous System Remyelination: Roles of Glia and Innate Immune Cells*

Charbel S. Baaklini, Khalil S. Rawji, Greg J. Duncan, Madelene F. S. Ho and Jason R. Plemel


Nathaniel Enos, Hidehito Takenaka, Sarah Scott, Hai V. N. Salfity, Maia Kirk, Margaret W. Egar, Deborah A. Sarria, Denise Slayback-Barry, Teri Belecky-Adams and Ellen A. G. Chernoff


Nayana Gaur, Caroline Perner, Otto W. Witte and Julian Grosskreutz

# The iNOS Activity During an Immune Response Controls the CNS Pathology in Experimental Autoimmune Encephalomyelitis

Sandip Ashok Sonar and Girdhari Lal\*

*National Center for Cell Science, Pune, India*

#### Edited by:

*Antje Kroner, Medical College of Wisconsin, United States*

#### Reviewed by:

*Stefan Bittner, Johannes Gutenberg University Mainz, Germany Stella Tsirka, Stony Brook University, United States*

> \*Correspondence: *Girdhari Lal glal@nccs.res.in*

#### Specialty section:

*This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Immunology*

Received: *16 January 2019* Accepted: *15 March 2019* Published: *04 April 2019*

#### Citation:

*Sonar SA and Lal G (2019) The iNOS Activity During an Immune Response Controls the CNS Pathology in Experimental Autoimmune Encephalomyelitis. Front. Immunol. 10:710. doi: 10.3389/fimmu.2019.00710* Inducible nitric oxide synthase (iNOS) plays a critical role in the regulation of multiple sclerosis (MS) and experimental autoimmune encephalomyelitis (EAE). Previous studies have shown that iNOS plays pathogenic as well as regulatory roles in MS and EAE. However, how does iNOS alters the pathophysiology of the central nervous system (CNS) in neuronal autoimmunity is not clearly understood. In the present work, we show that treatment of mice with L-NAME, an iNOS inhibitor, during the antigen-priming phase primarily alters brain pathology, while in the subsequent effector phase of the immune response, the spinal cord is involved. Inhibition of iNOS during the priming phase of the immune response promotes the infiltration of pathogenic CD11b+F4/80−Gr-1<sup>+</sup> cells, but there is low recruitment of regulatory CD11b+F4/80<sup>+</sup> cells in the brain. Inhibition of iNOS during the effector phase shows similar pathogenic alterations in the spinal cord, instead of in the brain. Treatment of wild-type mice with L-NAME or mice having genetic deficiency of iNOS show lower MHC-II expression on the dendritic cells, but not on macrophages. Our data suggest that iNOS has a critical regulatory role during antigenpriming as well as in the effector phase of EAE, and inhibition iNOS at different stages of the immune response can differentially alter either the brain or spinal cord pathology. Understanding the cellular and molecular mechanisms through which iNOS functions could help to design a better strategies for the clinical management of neuroinflammation and neuronal autoimmunity.

Keywords: experimental autoimmune encephalomyelitis, inducible nitric oxide synthase, NOS2−/<sup>−</sup> neuroinflammation, central nervous system, autoimmunity

# INTRODUCTION

Nitric oxide (NO) is a small bioactive lipophilic molecule that diffuse across the cell membrane and controls many physiological functions of the body (1). NO production requires nitric oxide synthase (NOS), which catalyze the oxidation of L-arginine to L-citrulline (2, 3). In mammals, there are three different isoforms of NOS - endothelial NOS (eNOS), neuronal NOS (nNOS) and inducible NOS (iNOS) (4). Constitutive expression of eNOS and nNOS controls the vasodilation of vessels and neuronal functions, respectively (5). Several inflammatory stimuli can induce the expression of iNOS in various cell types such as macrophages, dendritic cells, neutrophils, epithelial cells in the gut and lung mucosa, smooth muscle cells, and stromal cells of secondary lymphoid organs (6–9). iNOS is also expressed in microglial cells, astrocytes, neurons in the central nervous system (CNS), and endothelial cells at the blood-brain barrier (BBB) (7, 10, 11). A clinical association between iNOS and pathogenesis has been reported in many organ-specific autoimmune inflammatory diseases, including multiple sclerosis (MS) and experimental autoimmune encephalomyelitis (EAE) (12–15).

Proper neuronal function requires the presence of the minimal physiological concentrations of NO in the CNS, with sustained high NO levels leading to detrimental effects (16). The active MS patients show high NO levels in the cerebrospinal fluid (CSF), and high concentrations of NO, peroxynitrite, and other reactive nitrogen species have been found to correlate with greater severity and chances of relapse of clinical symptoms (15, 17). The expression of iNOS in the CNS is very tightly regulated, and several intrinsic and extrinsic stimuli can induce its expression in immune cells (14, 18). The T cell-derived cytokine, IFN-γ induces the expression of iNOS in macrophages and microglial cells which leads to the generation of higher NO and peroxynitrite productions, and cause tissue destruction in the CNS (13, 19, 20). However, iNOS−/<sup>−</sup> mice are hypersusceptible to EAE, suggesting that iNOS may have a regulatory function during CNS inflammation and autoimmunity (21, 22). Several studies have shown that iNOS can regulate the function of regulatory dendritic cells (regulatory DCs) which in turn can induce apoptosis of inflammatory CD4<sup>+</sup> T cells and help in controlling the development of EAE (23–25). Furthermore, iNOS expression in macrophages is linked with the suppression of inflammasome activation-induced IL-1β production (26), as well as a reduction in the frequency of M1 macrophages (27). Myeloid cell-derived iNOS is also known to control the CD4<sup>+</sup> T cell response (28, 29). High levels of APCs-derived NO suppresses CD4<sup>+</sup> T cell response, while low levels favor the generation of a Th1 response (30). Th17-intrinsic iNOS has been shown to suppress Th17 response through nitration of the tyrosine residues of RORγt, and limiting its promoter binding capacity (31). While it is known that Th17 cells in the CNS express iNOS during EAE, the functional importance of iNOS production by Th17 cells in the inflamed CNS is not clearly understood. During chronic demyelination, a pathogenic phenotype of microglial cells has been found to be associated with iNOS expression (32, 33). Since both lesion-associated and non-associated astrocytes express iNOS, the contribution of astrocyte-derived iNOS is still unclear. Some in vitro experiments suggest that inflammatory cytokine-induced iNOS reduces the expression of myelin proteins and causes oligodendrocyte death in the mixed glial cultures (34). All these observations indicate that iNOS plays a dual role during neuronal autoimmunity. Anti-IFN-γ treatment and IFN-γR <sup>−</sup>/<sup>−</sup> mice show hypersusceptibility to the development of EAE, with preferential involvement of the brain stem and cerebellum, resulting in the atypical EAE symptoms with the critical participation of neutrophil effector function (35–37). Given that IFN-γ regulates the iNOS expression in several immune cells, how does iNOS controls the inflammation in the brain and the spinal cord, and whether iNOS performs different functions during the antigen-priming and effector phases of EAE is not known.

In the present study, we assessed the role of iNOS using L-NAME-mediated inhibition of its activity during various stages of the immune response in EAE, including the antigen-priming phase and the effector phases, accompanied by monitoring of cellular pathology in the CNS. Our results showed that inhibition of iNOS during the antigen-priming as well as effector phases of EAE worsened the disease, and histology indicated differential regulation of infiltration of CD11b+F4/80−GR-1 <sup>+</sup> and CD11b+F4/80<sup>+</sup> cells in the brain and spinal cord. iNOS inhibition during the antigen-priming phase selectively promoted the infiltration of inflammatory CD11b+F4/80−GR-1 <sup>+</sup> cells, while lowering the frequency of infiltration of CD11b+F4/80<sup>+</sup> cells into the brain. Conversely, inhibiting iNOS during the effector phase led to mostly CD11b+F4/80−GR-1 <sup>+</sup> cells migrating into the spinal cord. A similar phenotype with higher infiltration of CD11b+F4/80−GR-1<sup>+</sup> cells and reduced infiltration of CD11b+F4/80<sup>+</sup> cells in the CNS was observed in iNOS−/<sup>−</sup> mice or wild-type mice in which IFNγ, a known inducer of iNOS, was blocked. We show that iNOS plays a regulatory role in promoting the infiltration of CD11b+F4/80<sup>+</sup> suppressor cells, while at the same time inhibiting the mobilization of pathogenic CD11b+F4/80−GR-1<sup>+</sup> cells into the CNS.

#### RESULTS

#### Inhibition of NO Production in the Priming Phase Promotes Granulocytic Myeloid Cells Infiltration Specifically in the Brain

Active EAE was induced in C57BL/6 mice and given an intraperitoneal injection of NOS inhibitor L-NAME (100 mg/kg/ every day) in the antigen-priming phase (one injection/day for seven days). The inhibition of NO production by L-NAME at the antigen-priming phase significantly increased the severity of EAE mice, compared to control mice (**Figure 1A**). Interestingly, the severity of the clinical symptoms of EAE increased with time in mice that received L-NAME (**Figure 1A**). Analysis of the brain and spinal cord tissues showed enhanced infiltration of CD45<sup>+</sup> leukocytes, including CD4<sup>+</sup> T cells, in the brain but not in the spinal cord (**Figure 1B**). CD11b+F4/80<sup>+</sup> cells include mainly macrophages and monocytic myeloid-derived suppressor cells (Mo-MDSCs). The F4/80-expressing MDSCs are known to have a suppressive function in EAE (38), whereas CD11b+GR-1 <sup>+</sup> cells show a pathogenic phenotype (39). L-NAME-treated mice showed a significant reduction in the infiltration of CD11b+F4/80+GR-1<sup>−</sup> cells selectively in the brain, but not

**Abbreviations:** APCs, Antigen-presenting cells; BMDCs, Bone marrowderived DCs; CNPase, 2',3'-cyclic nucleotide 3'-phosphodiesterase; CNS, Central nervous system; CFA, Complete Freund's adjuvant; DCs, Dendritic cells; EAE, Experimental autoimmune encephalomyelitis; eNOS, Endothelial NOS; GFAP, Glial fibrillary acidic protein; iNOS, inducible NOS; L-NAME, N-ω-nitro-larginine methyl ester; MDSCs, Myeloid-derived suppressor cells; MOG, Myelin oligodendrocyte glycoprotein 35-55 peptide; Mo-MDSCs, Monocytic myeloidderived suppressor cells; MS, Multiple sclerosis; NO, Nitric oxide; NOS, Nitric oxide synthase; nNOS, neuronal NOS.

in the spinal cord (**Figure 1C**). However, L-NAME treatment significantly increased infiltration of CD11b+F4/80−GR-1<sup>+</sup> cells (mainly neutrophils) in the brain but not in the spinal cord (**Figure 1C**). Together, our results showed that inhibition of NOS in the priming phase of EAE mostly affected the infiltration of inflammatory CD4<sup>+</sup> T cells and GR-1<sup>+</sup> neutrophils in the brain, and reduced the frequency of CD11b+F4/80<sup>+</sup> cells, which might account for the severity of EAE in the later phase of the disease.

#### Inhibition or Deficiency of iNOS in the Antigen-Priming Phase Does Not Alter the Differentiation of Effector CD4<sup>+</sup> T Cells

Since L-NAME injections at day 0-7 coincided with the antigenspecific priming, activation, and differentiation of effector CD4<sup>+</sup> T cells, we measured the frequency of Th1 cytokine (IFN-γ) and Th17 cytokine (IL-17A) producing cells, and Foxp3<sup>+</sup> regulatory CD4<sup>+</sup> T cells in the spleen and lymph nodes. Our results showed no significant alteration in the intracellular expression of IL-17A and IFN-γ in CD4<sup>+</sup> T cells (**Figures 2A,B**) or γδ T cells (**Figure S1A**). To further confirm the role of iNOS in the priming-phase of EAE, we immunized wild-type and iNOS−/<sup>−</sup> mice with MOG in complete Freund's adjuvant (CFA). On day 8, we compared the differentiation of effector immune cells in their spleen and lymph nodes with those of L-NAME (d0-7) treated and untreated wild-type C57BL/6 mice. iNOS−/<sup>−</sup> mice also did not show any significant change in the expression of the Th lineage-specific transcription factors, T-bet, RORγt, Foxp3, and Eomes in CD4<sup>+</sup> T cells, as compared to the control group (**Figures 2C,D** and **Figures S1B,C**). However, L-NAME treated mice showed a lower frequency of T-bet expressing CD4<sup>+</sup> T cells in the draining lymph nodes, as compared to the control group (**Figures 2C,D**). We did not observe any significant changes in the expression of intracellular IL-17A, IFN-γ, and GM-CSF in the CD4<sup>+</sup> T cells or γδ T cells (data not shown). On day 28, L-NAME treated mice showed severe EAE symptoms but did not show a significant change in the percentages of Th1, Th17, and Tregs or IL-17A<sup>+</sup> and IFN-γ <sup>+</sup> or IL-17A+IFN-γ <sup>+</sup> γδ T cells in the secondary lymphoid organs (**Figures 2A,B** and **Figure S1A**). Together, these results suggest that inhibition of iNOS during the priming phase of EAE does not affect the differentiation of Th1, Th17, and Treg cells in the secondary lymphoid tissues.

#### Inhibition of iNOS in the Effector Phase Shows Increase Cellular Infiltration in the Spinal Cord

To understand how inhibition of NO production during the effector phase of EAE changes the pathophysiological phenotype and transmigration of effector immune cells in the CNS, L-NAME was administered (100 mg/kg/day; i.p.; day 8–15 of MOG injection) in C57BL/6 mice. Inhibition of NO production during the effector phase resulted in exacerbated EAE in these mice, as compared to control mice (**Figure 3A**). Interestingly, while increased infiltration of CD45<sup>+</sup> leukocytes and CD4<sup>+</sup> T cells was observed in the spinal cord in mice treated with L-NAME in the effector phase, infiltration of these cells in the brain was similar in L-NAME-treated and control mice (**Figure 3B**). This suggested that inhibition of NO production during the effector phase of EAE differentially diverts the inflammatory cells to the spinal cord and worsens the disease. Both meningeal and parenchymal regions showed higher infiltration in L-NAME-treated mice, as compared to control group (data not shown). Surprisingly, immunohistological analysis showed higher expression of iNOS in the brain-infiltrating CD45<sup>+</sup> leukocytes in the priming phase, as compared to that observed upon treatment with L-NAME in the effector phase (**Figure 3C**). However, inhibition of NO production during the effector phase showed a higher frequency of iNOS-expressing CD45<sup>+</sup> leukocytes in the spinal cord as compared to control group, or the mice that received iNOS inhibitor in the priming phase (**Figure 3D**). Analysis of iNOS expression in the brain and spinal cord astrocytes further revealed that the frequency of iNOS-expressing GFAP<sup>+</sup> astrocytes was significantly lower in the brains of mice treated with L-NAME during the antigen-priming phase as compared to control, whereas the spinal cord resident astrocytes did not show alteration in the number of iNOS-expressing cells (**Figures 3C,D**). These results indicate that L-NAME-mediated transient inhibition of NOS during the antigen-priming and effector phase of EAE leads to the reappearance of iNOS expression selectively in the infiltrated CD45<sup>+</sup> leukocytes and possibly in the CD45int CNS-resident microglial cells, in the brain and spinal cord, respectively. Consistent with the previous reports (21, 22), our results also showed that the a genetic deficiency of iNOS in mice leads to the development of more severe EAE than that observed in wild-type animals (**Figure 3E**). Immunohistological analysis of the brain and spinal cord in iNOS−/<sup>−</sup> mice showed significantly increased infiltration of CD45<sup>+</sup> leukocytes and CD4<sup>+</sup> T cells in the brain but not in the spinal cord (**Figure 3F**). These observations were consistent with those we obtained with the L-NAME injections in the early priming phase of the EAE, suggesting that lack of NO production in the initial stage of EAE directs pathological mechanisms specifically to the brain but not the spinal cord.

Furthermore, inhibition of NO production during the effector phase showed significantly reduced infiltration of CD11b+F4/80+GR1<sup>−</sup> cells in the brain and spinal cord as compared to control animals (**Figure S2A**). Interestingly, substantially higher infiltration of CD11b+F4/80−GR-1<sup>+</sup> cells was observed in the spinal cord but not in the brain tissue during effector phase iNOS inhibition as compared to control mice (**Figure S2A**).

To further confirm the role of iNOS in the infiltration of the CD11b+F4/80+GR-1<sup>−</sup> cells and CD11b+F4/80−GR-1<sup>+</sup> cells in the CNS, we induced active EAE in iNOS−/<sup>−</sup> mice and analyzed myeloid cell infiltration in the brain and spinal cord. Consistent with our observations with L-NAME treatment during the priming or effector phase, our results showed that deficiency of iNOS leads to reduced infiltration of CD11b+F4/80+GR-1<sup>−</sup> cells and increased infiltration of CD11b+F4/80−GR-1<sup>+</sup> cells in the brain and spinal cord, as compared to wild-type mice (**Figure S2B**). These results suggest that inhibition of NOS or a genetic deficiency of iNOS selectively promotes the neutrophilic infiltration and inhibits the infiltration of CD11b+F4/80+GR-1<sup>−</sup> cells in the CNS.

Frontiers in Immunology | www.frontiersin.org 4

two way ANOVA followed by Tukey's test (B), Student *t*-test (A,C). *n* = 5 mice/group.

Together, our results show that lack of iNOS during EAE development facilitates inflammation in both the brain and spinal cord, with enhanced inflammatory cell infiltrations specifically in the brain but not the spinal cord as compared to wild-type

images of the brain and spinal cord tissues stained with CD11b (gray), GR-1 (red). F4/80 (green), and nuclear stain DAPI (blue) are shown (left). The mean number of infiltrated CD11b+F4/80−GR-1−, CD11b+F4/80+GR-1<sup>−</sup> and CD11b+F4/80−GR-1<sup>+</sup> cells from at least 25–30 sections of the brain and spinal cord are plotted (right). Error bars represent ± standard error of mean (SEM) (A–C). Original magnification 400x (B,C). Scale bar 100µm (B,C). \**p* < 0.05, \*\**p* < 0.01, \*\*\*\**p* < 0.0001;

mice. However, differential preference observed with L-NAMEmediated inhibition of NOS during the priming versus effector phase of EAE. These results suggest that inhibition of NOS in the different stages of EAE differentially regulates the infiltration of

effector CD45<sup>+</sup> leukocytes, CD4<sup>+</sup> T cells, CD11b+F4/80−GR-1<sup>+</sup> cells, and CD11b+F4/80+GR1<sup>−</sup> cells in the CNS, and this may, in turn, account for the severity of the disease.

# Inhibition of iNOS During the Priming-Phase Reduces MHC Class II Expression on DCs and Reduces the Frequency of Monocytic-MDSCs in Secondary Lymphoid Tissues.

Myeloid cells like macrophages and dendritic cells serve as antigen presenting cells (APCs) and significant producers of iNOS, and play a crucial role in the pathogenesis of EAE (23, 24). Our result showed that EAE induced in iNOS−/<sup>−</sup> mice, as well as L-NAME-treated mice (day 0–7), had significantly reduced frequency of CD11c+I-Ab<sup>+</sup> DCs (**Figure 4A**), and significantly lower median fluorescence intensity (MFI) of the class II MHC molecule (I-A<sup>b</sup> ) on the DCs, as compared to control mice (**Figure 4B**)**.** Further analysis showed that among the various subsets of DCs, the CD11b+CD11c<sup>+</sup> and CD11b−CD11c<sup>+</sup> subsets showed reduced expression of class II MHC in L-NAMEtreated mice and iNOS−/<sup>−</sup> mice, as compared to control EAE mice (**Figure 4C**). L-NAME-treated and iNOS−/<sup>−</sup> mice also showed a reduced percentage of CD11b+F4/80<sup>+</sup> macrophages

FIGURE 3 | The Inhibition of iNOS during the effector phase exacerbates inflammation in the spinal cord, leading to the development of severe EAE. Active EAE was induced as in Figure 1. Mice were administered *i.p.* L-NAME from day 8 to 15 daily. Control groups were given *i.p.* PBS. (A) EAE clinical scores were monitored and plotted. (B) On day 28, brain and spinal cord cryo-sections were analyzed by immunofluorescence staining. Representative images of the brain and spinal cord tissue sections stained with CD45 (gray), CD4 (red), and nuclear stain DAPI (dark blue) are shown (left). Mean numbers of infiltrating CD45<sup>+</sup> and CD4<sup>+</sup> cells from 16 to 25 images of the brain and spinal cord were quantitated and plotted (right). (C,D) Representative images of the (C) brain and (D) spinal cord sections stained with GFAP (gray), iNOS (red), CD45 (green), and nuclear stain DAPI (dark blue) are shown (left). Magnified images of the regions marked with the dotted square and are shown at the bottom (D). Mean numbers of GFAP+iNOS<sup>+</sup> and CD45+iNOS<sup>+</sup> cells from 15 to 35 images of the brain and spinal cord were quantitated and plotted (right). (E) C57BL/6 and iNOS−/<sup>−</sup> mice were *s.c.* injected with 200 <sup>µ</sup>g MOG35−<sup>55</sup> (MOG) in CFA emulsion, and two doses of *i.v.* pertussis toxin (PTx, 200 ng/mouse) at day 0 and 2. Mice were monitored for the development of clinical symptoms of EAE and the mean clinical scores were plotted (5 mice/group). (F) Mice in (E) were sacrificed on day 20, and brain and spinal cord cryo-sections were analyzed by immunofluorescence staining. Representative images of the brain and spinal cord tissue sections stained with CD45 (gray), CD4 (red), and nuclear stain DAPI (dark blue) and analyzed are shown (left). Mean numbers of infiltrating CD45<sup>+</sup> and CD4<sup>+</sup> cells from 25 to 35 images of the brain and spinal cord were quantitated and plotted (right). Error bars represent ± SEM (A–F). Original magnification 400x (Brain; B,C,F); 200x (spinal cord; B,D,F). Scale bar, 100µm (Brain; B,C,F), 300µm (spinal cord; B,D,F). \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001; Student's *t-*test (A,B,F), two way ANOVA followed by Tukey's test (C,D). *n* = 5 mice/group.

in the spleen (**Figure 4D**), which may correspond to the lower infiltration of CD11b+F4/80<sup>+</sup> cells observed in the brain of L-NAME-treated mice, as compared to untreated mice (**Figure 1C** and **Figure S2B**). L-NAME treatment did not have any effect on the MFI of I-A<sup>b</sup> on non-DCs, whereas a lack of iNOS in iNOS−/<sup>−</sup> mice showed a significant decrease in the MFI of I-A<sup>b</sup> on non-DCs, as compared to control mice (**Figure 4E**). The F4/80+CD11b+Ly6ChiLy6G<sup>−</sup> cells mainly represent Mo-MDSCs, and are potent suppressors of effector CD4<sup>+</sup> and CD8<sup>+</sup> T cells, and known to express their suppressive effect via an iNOS-dependent mechanism (38). Our results showed a significantly reduced frequency of F4/80-expressing Mo-MDSCs in the iNOS−/<sup>−</sup> mice spleen, and to some extent in the L-NAME-treated mice, as compared to control mice (**Figure 4F**). These results suggest that inhibition or lack of iNOS can alter the generation of CD11b+F4/80<sup>+</sup> cells, including suppressive F4/80+Mo-MDSCs and mature antigen-presenting DCs in the spleen, leading to reduced infiltration of CD11b+F4/80+GR-1<sup>−</sup> cells and increased infiltration of CD11b+F4/80−GR-1<sup>+</sup> cells in the brain.

#### Neutralization of IFN-γ Shows iNOS Expression and Apoptosis of CNPase<sup>+</sup> Oligodendrocytes in the CNS

IFN-γ is a potent inducer of class II MHC molecules on APCs. It can render either inflammatory or tolerogenic function to DCs, depending on the presence or absence of a Toll-like receptor (TLR) and CD40L signaling, and is essential for the development of allograft tolerance. IFN-γ is also a known inducer of iNOS in a variety of cell types, including neutrophils, monocytes, macrophages, dendritic cells, microglial cells, and astrocytes (28). The IFN-γ <sup>−</sup>/<sup>−</sup> or IFN-γR <sup>−</sup>/<sup>−</sup> mice or antibody-mediated neutralization of IFN-γ signaling, confers hyper-susceptibility to the EAE. Similarly, our findings revealed that L-NAME-mediated inhibition of NOS in the early and effector phase of EAE causes severe EAE in wild-type mice.

Since our results showed that lack of iNOS reduced the expression of class II MHC molecule on DCs, we further investigated the link between IFN-γ and iNOS during EAE. We analyzed the influence of IFN-γ on iNOS-expression in astrocytes and CNS-infiltrated immune cells, and on apoptosis of the myelin-synthesizing cells, oligodendrocytes. For this purpose, we neutralized the IFN-γ with the anti-IFN-γ (100 µg/mouse; i.v.) mAb during MOG-induced active EAE. Consistent with several published reports (35, 36), our results showed exacerbation of the clinical symptoms of EAE with the neutralization of IFNγ (**Figure 5A**). We also observed an increased infiltration of CD45<sup>+</sup> leukocytes and CD4<sup>+</sup> T cells in the brain as well as spinal cord (**Figure 5B**). Both the meninges and parenchyma showed increased infiltration of immune cells in the anti-IFN-γ-treated group, as compared to the isotype control IgGtreated mice (data not shown). While anti-IFN-γ-treated mice showed significantly low infiltration of CD11b+F4/80+GR-1<sup>−</sup> cells, however, the numbers of CD11b+F4/80−GR-1<sup>+</sup> cells were dramatically increased in the brain and spinal cord, as compared to control group (**Figure 5C**). The neutralization of IFN-γ using mAb did not induce iNOS expression in the astrocytes (**Figure 5D**). However, with anti-IFN-γ-treatment, iNOS expression was significantly induced in CD45<sup>+</sup> leukocytes in both brain and spinal cord, as compared to isotype IgGtreated control mice (**Figure 5D**), suggesting that neutralization of IFN-γ leads to the reappearance of iNOS-expression on CD45<sup>+</sup> leukocytes and possibly in microglia, similar to the effects observed with L-NAME treatment.

We then asked whether increased infiltration of CD11b+F4/80−GR-1<sup>+</sup> cells could have an effect on the survival of oligodendrocytes during EAE. Our results showed that during EAE, the cleaved-caspase 3 signal co-localizes with CNPase<sup>+</sup> oligodendrocytes, suggesting that they are undergoing apoptosis (**Figure 5E**). The anti-IFN-γ-treated mice showed significantly higher co-localization of cleaved-caspase 3 with CNPase<sup>+</sup> oligodendrocytes in the brain, as compared to the isotype control group (**Figure 5E**), suggesting that the neutralization of IFN-γ signaling lead to increased apoptosis of oligodendrocytes during EAE. Furthermore, treatment with anti-IFN-γ mAb increased the infiltration of effector CD4<sup>+</sup> T cells, CD11b+F4/80−GR-1 <sup>+</sup> cells, and iNOS-expressing inflammatory immune cells, and reduced infiltration of suppressive CD11b+F4/80<sup>+</sup> cells (includes F4/80-expressing suppressive Mo-MDSCs) in the CNS. These could have contributed to increased oligodendrocyte and neuronal damage, consequently exacerbating the EAE pathology.

# DISCUSSION

In the present study, we show that iNOS critically regulates neuroinflammation at different phases of EAE by controlling the infiltration of immune cells in the brain and spinal cord. Inhibition of iNOS in the antigen-priming phase or effector phase of EAE exacerbates the symptoms of the disease via selectively increasing the infiltration of inflammatory CD11b+F4/80−GR-1<sup>+</sup> cells, while at the same time reducing the infiltration of CD11b+F4/80+GR-1<sup>−</sup> cells in the brain and spinal cord, respectively. Similarly, with anti-IFN-γ mAb treatment, we showed differential infiltration of CD11b+F4/80+GR-1<sup>−</sup> and CD11b+F4/80−GR-1<sup>+</sup> cells in the CNS during EAE. Furthermore, lack of iNOS or neutralization of IFN-γ promoted the infiltration of CD11b+F4/80−GR-1<sup>+</sup> cells in the brain or spinal cord and enhanced the inflammation and apoptosis of CNPase<sup>+</sup> oligodendrocytes.

Compelling evidence has suggested that various types of myeloid cells do infiltrate into inflamed CNS along with myelinreactive Th1 and Th17 cells (40). These cells contribute to the various inflammatory pathways leading to demyelination and axonal damage in the CNS. iNOS controls the function of a variety of myeloid and lymphoid cell populations in both a cell-intrinsic and extrinsic manner (41). iNOS is also critically involved in the pathogenesis of EAE and MS (17). However, how it affects neuroinflammation in the various phases of EAE was not clear. In the present work, we showed that inhibition of iNOS during the antigen-priming phase leads to a selectively increased infiltration of CD45<sup>+</sup> leukocytes and CD4<sup>+</sup> T cells selectively in the brain but not in the spinal cord. Together with

data from iNOS−/<sup>−</sup> mice, our results suggest that a lack of iNOS in the priming phase induces inflammation in the secondary lymphoid organs in a manner that causes pathogenic cells to be mobilized into the brain. The molecular details of trafficking of these cells across the blood-brain barrier to CNS and the effect of iNOS and possibly other isoform of NOS, such as eNOS in their transmigration need further investigation. Previous studies have shown that while iNOS does not affect the differentiation of Th1, Th2 and Tregs, it does influence Th17 differentiation (31). However, studies on the mechanism through which iNOS and NO influence Th17 differentiation have yielded mixed results. Studies in mice have revealed that this inhibitory function of endogenous iNOS is via the nitration of tyrosine residues in RORγt, and the inhibition of aryl hydrocarbon receptor (Ahr) signaling (31, 42). However, others have reported that iNOS and NO support human Th17 differentiation via the cyclic guanosine monophosphate (cGMP)-dependent protein kinase pathway, and endogenous iNOS play an essential role in the stability of human Th17 cells when differentiated in the presence of IL-1β, IL-6, and IL-23 (43). Our data show that inhibition of NOS or genetic deficiency of iNOS neither affect the generation of pathogenic Th1, Th17 cells, and Foxp3<sup>+</sup> regulatory CD4<sup>+</sup> T

FIGURE 5 | Neutralization of IFN-γ induces the expression of iNOS in the effector immune cells and results in apoptosis of CNPase<sup>+</sup> oligodendrocytes in EAE. Active EAE was induced in C57BL/6 mice. Anti-IFN-γ monoclonal antibody (XMG1.2, 100 µg/mouse) monoclonal antibody was *i.v.* injected at days 0, 4, 8, 12, 16 after MOG-immunization. (A) The clinical symptoms of EAE were recorded and mean clinical scores are plotted (5 mice/group). (B) On day 21, the brain and spinal cord cryo-sections were analyzed by immunofluorescence staining. Representative images of the brain and spinal cord tissue sections stained with CD45 (gray), CD4 (red), and nuclear stain DAPI (dark blue) are shown (left). Mean numbers of infiltrating CD45<sup>+</sup> and CD4<sup>+</sup> cells from 25 to 35 images of the brain and spinal cord were quantitated and plotted (right). (C) Representative images of the brain and spinal cord stained with CD11b (red), F4/80 (green), GR-1 (light blue), and nuclear stain DAPI (dark blue) are shown (left). Infiltration by indicated cell populations was quantified and plotted (right). (D) Representative images stained with GFAP (green), iNOS (red) and CD45 (blue) are shown (left). iNOS-expressing GFAF<sup>+</sup> cells and CD45<sup>+</sup> cells were quantified from 20 to 30 images of the brain and spinal cord and plotted as mean number of cells/field (right). (E) Representative images stained for cleaved caspase 3 (green), CNPase (red), CD45 (gray), and DAPI (blue) are shown (left). Cleaved-caspase 3 expressing CNPase<sup>+</sup> cells oligodendrocytes were quantified from 20 to 30 images of the brain and spinal cord and plotted as mean number of cells/field (right). Regions marked with white dotted square (D,E) are shown as magnified images at the right. Error bars represent ± SEM (A–E). Original magnification 200x (B, spinal cord), 400x (B, brain) and (C), 630x (D,E). Scale bar 50µm (D,E), 100 µm (B, brain) and (C) and 300µm (B, spinal cord). \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001, ANOVA followed by Tukey's test (A), ANOVA followed by Sidak's test (C), Student's *t*-test (B,D,E). *n* = 5 mice/group.

cells nor IFN-γ and IL-17A-expressing γδ T cells in the secondary lymphoid organs.

iNOS regulates the generation and function of regulatory DCs that control the effector function of CD4<sup>+</sup> and CD8<sup>+</sup> T cells, and helps in amelioration of EAE (23–25). CD11b+F4/80<sup>+</sup> cells and F4/80-expressing Mo-MDSCs are known to have a regulatory role in EAE (38). We show that a lack of iNOS function in the priming-phase results in a significant reduction in the frequency of CD11b+F4/80<sup>+</sup> cells as well as F4/80-expressing Mo-MDSCs in the spleen. L-NAME-treatment or lack of iNOS in mice yielded significantly lower infiltration of these cells in the inflamed CNS. In contrast, lack of iNOS significantly increased the previously reported pathogenic CD11b+F4/80−GR-1<sup>+</sup> cells in the brain (39). In contrast, inhibition of iNOS in the effector phase of EAE, when ongoing inflammation in the CNS causes demyelination and axonal damage, resulted in higher infiltration of CD4<sup>+</sup> T cells and CD11b+F4/80−GR-1<sup>+</sup> cells preferentially in the spinal cord, but not in the brain. These results suggest that temporal inhibition of iNOS during EAE modulates differential clinical pathology in the brain and spinal cord. iNOS inhibition during the antigen-priming phase of EAE showed a high frequency of iNOS-expressing CD45<sup>+</sup> leukocytes selectively in the brain, but not in the spinal cord. Since CNS-resident microglia express iNOS under inflammation and microglia are also characterized by CD45int expression, our study cannot exclude the potential role of iNOS-expressing microglia in exacerbation of EAE symptoms. Instead, this further opens the avenue to investigate the relative functions of iNOS-expressing microglia and circulation-derived macrophages in the early and late phases of neuroinflammation during EAE. The GFAP<sup>+</sup> astrocytes are also an important source of iNOS-expression during neuroinflammation (16). We also observed higher iNOSexpressing astrocytes in the brain and spinal cord in EAE mice. However, inhibition of iNOS in the priming phase selectively reduced the frequency of iNOS-expressing GFAP<sup>+</sup> astrocytes in the brain, whereas it was mostly unaffected upon inhibition of iNOS during the effector phase.

IFN-γ is the upstream regulator of iNOS expression in a variety of myeloid and lymphoid cells, and cells of nonhematopoietic lineages (9, 12). The neutralization of IFN-γ using anti-IFN-γ mAb during EAE in mice have been observed to result in more severe symptoms and pathology (44). IFN-γ is a known suppressor of Groα/KC (CXCL1) (45) and MIP-2 (CXCL2) (46), both being neutrophil chemoattractants that recruit neutrophils to the site of inflammation (47). In the absence of IFN-γ signaling, Th17 cells predominate infiltrate into the CNS, and predominantly recruit neutrophils in the CNS (48, 49), possibly via CXCL1- and CXCL2-mediated neutrophil chemoattraction. Neutralization of IFN-γ therefore promotes neutrophil trafficking to the inflamed CNS. Consequently, we reasoned that a lack of iNOS function during anti-IFNγ treatment (12) may contribute for bias differentiation of CD11b+F4/80<sup>+</sup> cells and F4/80-expressing Mo-MDSCs in the lymphoid organs, and infiltration of these cells in CNS may cause severe EAE. However, we also observed an increase in the expression of iNOS in the CNS-infiltrating immune cells and possibly CD45int microglia in anti-IFN-γ-treated mice, suggesting that in the absence of IFN-γ, some other factors can also induce the expression of iNOS in the inflamed CNS (50). Furthermore, the enhanced expression of iNOS in the CNS infiltrating immune cells was associated with increased apoptosis of myelin-synthesizing oligodendrocytes in anti-IFNγ-treated mice, suggesting that high level of iNOS produced by inflammatory CNS infiltrates can affect the remyelination process and contribute to severity of the disease. A study with cuprizoneinduced demyelination model reported that infiltration and the inflammatory function of CXCR2<sup>+</sup> neutrophils are required to induce oligodendrocyte damage and demyelination in addition to the toxic effect of cuprizone on mitochondrial function (51). The increased apoptosis of CNPase<sup>+</sup> oligodendrocytes in anti-IFNγ treated mice, might be in part due to increased neutrophilic infiltration in the CNS. A similar pathological mechanism may exist in iNOS-deficient, or L-NAME treated wild-type mice and warrants further investigation. We showed that MOG-induced inflammation caused the induction of iNOS-expression in the astrocytes, and that anti-IFN-γ treatment completely inhibits the iNOS expression in astrocytes. The numbers of iNOSexpressing astrocytes and the iNOS produced by them have different consequences. While low levels of iNOS or NO are beneficial, and their sustained high levels are detrimental to the CNS homeostasis (52). Therefore, the high levels of iNOS produced in the anti-IFN-γ recipients in our studies might have caused apoptosis of oligodendrocytes in the brain, and thus contributed to increased the clinical severity of the EAE. In conclusion, we have shown here that immunoregulatory role of iNOS safeguards the brain and spinal cord from inflammatory granulocytic infiltration during the antigen-priming and effector phase of EAE, respectively.

#### MATERIALS AND METHODS

#### Mice

Wild-type C57BL/6 and iNOS−/<sup>−</sup> (B6.129P2-Nos2tm1Lau/J) mice were obtained from The Jackson Laboratory (Bar Harbor, ME). Mice were maintained and bred in the Experimental Animal Facility of the National Centre for Cell Science (NCCS), Pune, India. All mice experiments were performed with the protocols approved by the Institutional Animal Ethics Committee. (Project Id: EAF/2016/B-257).

#### Antibodies and Reagents

Alexa Fluor 488-CD4 (GK1.5), APC-eFluor 780-CD4 (GK1.5), FITC-γδTCR (GL3), APC-γδTCR (GL3), APC-CD45 (30F-11), FITC-F4/80 (BM8), APC-Cy7-F4/80 (BM8), FITC-Ly6G (1A8), Alexa Fluor 647-Ly6C (HK1.4), biotin-CD11b (M1/70), APC-GR-1 (RB6-8C5), PE-GM-CSF (MP1-22E9), Brilliant violet 421-IL-17A (TC11-18H10.1), and purified anti-mouse GFAP (MCA-5C10) were purchased from BioLegend (San Diego, CA). Biotin-CD11c (N418), PE/Cy7-IFN-γ (XMG1.2), Pacific blue-Foxp3 (FJK-16s), PE-Foxp3 (MF-14), PE-Eomes (Dan11mag), eFluor 450-IA<sup>b</sup> (AF6-1201) were obtained from eBioscience (San Diego, CA). Purified anti-IFN-γ (XMG1.2), and rat IgG2b, k isotype control (LTF-2) were purchased from BioXcell (West Lebanon, NH). PE-Cy7-CD11b (M1/70) antibody was from BD Biosciences (San Diego, CA). Purified anti-iNOS antibody (EPR16635) was purchased from Abcam (Cambridge, MA). Purified cleaved caspase 3 (5A1E) and purified anti-CNPase (D83E10) was obtained from Cell Signaling Technology (Danvers, MA). N-ω-nitro-l-arginine methyl ester (L-NAME) was purchased from MP Biomedicals (Santa Ana, CA).

#### Induction of Active EAE

Wild-type C57BL/6 or iNOS−/<sup>−</sup> mice were given subcutaneous (s.c.) injections of an emulsion of MOG35−<sup>55</sup> (MOG) (200 µg/mouse) in complete Freund's adjuvant (CFA) containing Mycobacterium tuberculosis H37Ra (5 mg/ml), and also given intravenous injections (i.v.) of pertussis toxin (PTx, 200 ng/mouse) at day 0 and 2. L-NAME was administered intraperitoneally (i.p. 100 mg/kg) daily from day 0 to 7 (priming phase) or day 8 to 15 (effector phase). Control animals received i.p. injections of PBS. For the neutralization of IFNγ, animals were given anti-mouse IFN-γ mAb (clone XMG1.2; 100 µg/mouse) i.v. on days 0, 4, 8, 12, and 16. The animals that received isotype control IgG were used as a controls. Mice were followed for the development of clinical signs of EAE. Scoring of clinical symptoms was performed as follows; score 0, no symptoms; 1, limp tail or hind limb weakness but not both; 2, limp tail and hind limb weakness; 3, partial hind limb paralysis; 4, complete hind limb paralysis; and 5, death by EAE.

#### Immunofluorescence Staining of the Brain and Spinal Cord

Mice were sacrificed, ice-cold PBS was transcardially perfused, and the brain and spinal cord were harvested. The tissues were immediately snap-frozen in OCT compound (Sakura Finetek, Torrance, CA). Seven-micrometer-thick cryosections were prepared using a cryomicrotome. Sections were fixed with chilled acetone for 5 min, air dried, followed by washing with PBS. The tissue sections were blocked with 10% horse serum (Jackson ImmunoResearch, West Grove, PA) in PBS at room temperature (RT) for 30 min, followed by washing thrice with ice-cold PBS. The sections were incubated with primary antibodies overnight (12–14 h) at 4◦C, washed, further incubated with secondary antibodies at RT for 60 min, and then washed five times with PBS. The sections were stained with nuclear stain DAPI at RT for 5 min and washed twice with ice-cold PBS. Sections were mounted in mounting medium (Electron Microscopy Sciences, Hatfield, PA). Images were acquired on a Leica DMI6000 inverted fluorescent microscope (Leica Microsystems, Germany) at 100, 400, and 630x magnifications. Images were analyzed using Leica MMAF (Leica) or Image J software (National Institute of Health, Bethesda, MD).

The quantification of cells from microscopic images was performed using the MMAF software (Leica Microsystems, Germany). The single channel and multi-channel overlay images of sufficient magnification (original magnification, 400x) were used for quantification of the number of cells. At least 10 different microscopic fields per organ and more than three mice/group were analyzed. Each cell was analyzed for its cellular morphology, nuclear staining with DAPI, and specific markers. For evaluating cellular infiltration into the CNS, complete brain tissues sections were used.

# Flow Cytometry

Cells were harvested from spleen and draining lymph nodes of the mice. RBCs were lysed using ACK lysis buffer, and single cell suspensions were prepared. Cells were surface-stained with PE/Cy7-anti-mouse CD11b, PE/Cy5-anti-mouse CD11c, Alexa fluor 647-anti-mouse Ly6C, FITC-anti-mouse Ly6G, APC/Cy7 anti-mouse F4/80 and pacific blue-anti-mouse I-Ab (MHC class II) antibodies. Cells were incubated on ice in the dark for 30 min, washed with ice-cold PBS, and fixed with 1% paraformaldehyde. Cells were acquired on FACS Canto-II (BD Biosciences), and data were analyzed using the FlowJo software.

# Intracellular Cytokine Staining

Single cell suspensions were prepared from spleen and draining lymph nodes of mice. For transcription factor analysis, 1 × 10<sup>6</sup> cells were stained for the surface molecules CD4 and γδ TCR on ice for 30 min and washed with ice-cold PBS. Cells were subjected to fixation, and permeabilization with the Foxp3 fixation/permeabilization buffer kit (Biolegend), and intracellular staining for Foxp3, T-bet, RORγt, and Eomes was performed as per the manufacturer's instructions. For intracellular cytokine analysis, cells (6 × 10<sup>6</sup> cells/well) were stimulated with phorbol myristate acetate (PMA; 50 ng/ml) and ionomycin (850 ng/ml) in the presence of brefeldin-A (5µg/ml) and monensin (2µM) in 500 µl/well complete RPMI 1,640 medium in 24-well plates at 37◦C in a humidified 5% CO<sup>2</sup> incubator for 6 hours. Cells were collected, washed with PBS, and stained for the surface molecules CD4 and γδ TCR on ice for 30 min, and washed with icecold PBS. Cells were subjected to fixation and permeabilization using the Foxp3 fixation/permeabilization buffer kit (Biolegend) and intracellular staining for IL-17A, Foxp3, Eomes, and IFNγ was performed as per the manufacturer's instructions. Cells were acquired on FACS Canto II (BD Bioscience), and data were analyzed using the FlowJo software.

# Statistical Analysis

Statistical comparisons were performed using the GraphPad Prism 6 software (GraphPad, San Diego, CA). Unpaired twotailed Student's t-test was used to compare two variables. The ANOVA test was used to compare the means of more than two groups followed by appropriate multiple comparison tests. The other statistical methods used are described in the figure legends. A p < 0.05 was considered statistically significant.

# ETHICS STATEMENT

All mice experiments were performed with the approved protocols from the NCCS Institutional Animal Ethics Committee (Project Id: EAF/2016/B-257).

# AUTHOR CONTRIBUTIONS

SAS and GL designed the experiments, analyzed the data, and wrote the manuscript. SAS performed all the experiments.

# ACKNOWLEDGMENTS

We thank Dr. Ramanamurthy Boppana and Dr. Rahul Bankar for help with animal experiments. We also thank Dr. Jyoti Rao for critical suggestions and edits. SAS received senior research fellowship from the Council of Scientific and Industrial Research, Government of India. GL received NCCS intramural funding and grants from the Department of Biotechnology, (Grants numbers, BT/PR4610/MED/30/720/2012, BT/PR15533/MED/ 30/1616/2015; BT/PR14156/BRB/10/1515/2016), Science and Engineering Research Board (EMR/2016/007108), Ministry of Science and Technology, Government of India.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00710/full#supplementary-material

Figure S1 | Inhibition of iNOS or lack of iNOS does not affect the generation of Th1, Th17, and Tregs in the secondary lymphoid organs. (A) EAE was induced in C57/BL/6 mice as shown in Figure 1A. The intracellular expression of IL-17A and IFN-γ in the γδ T cells of the spleen and draining lymph nodes of the mice was monitored and plotted (left). The dot plot shown is gated on γδ T cells. Quantifications of the percentage of intracellular cytokines in the γδ T cells are shown (right). (B,C) C57BL/6 and iNOS−/−mice were *s.c.* injected with 200 µg

#### REFERENCES


MOG35−55 (MOG) in CFA emulsion, and two doses of *i.v*. pertussis toxin (PTx, 200 ng/mouse) at day 0 and 2. Mice were administered *i.p.* L-NAME from day 0 to 7 daily. Control groups were given PBS. On the day 8, single cell suspensions were prepared from spleen and draining lymph nodes, and expression of the transcription factors, Eomes, Foxp3, RORγt, and T-bet in CD4<sup>+</sup> T cells from the (B) spleen and (C) draining lymph nodes were analyzed using flow cytometry. Dot plots show the expression of the indicated transcription factors in CD4<sup>+</sup> T cells. Numbers in the dot plots show percentage of indicated molecules in γδ T cells (A) and CD4<sup>+</sup> T cells (B,C). The horizontal line denotes mean and error bars represents ± SEM (A). Student's *t*-test (A). *n* = 4 mice/group (A) and 5 mice/group (B).

Figure S2 | Inhibition of iNOS or its deficiency differentially regulates the infiltration of myeloid cells in the CNS. The brain and spinal cord tissue sections of mice from Figures 3A,E were stained with CD11b (red), F4/80 (green), GR-1 (light blue) and nuclear stain DAPI (dark blue). (A) Representative images of the brain and spinal cord of mice either untreated or treated with L-NAME in the effector phase of EAE are shown (upper). Magnified views of the areas marked with the dotted squares are shown next to the images. The mean number of infiltrated CD11b+F4/80−GR-1−, CD11b+F4/80+GR-1<sup>−</sup> and CD11b+F4/80−GR-1<sup>+</sup> cells from at least 11–14 fields of the brain and spinal cord sections were quantitated and shown (lower). (B) Representative images of the brain and spinal cord sections of wild-type and iNOS−/<sup>−</sup> mice with EAE at day 20 are shown (upper). Magnified views of the areas marked with the dotted squares are shown next to the images. The mean numbers of infiltrated CD11b+F4/80−GR-1−, CD11b+F4/80+GR-1<sup>−</sup> and CD11b+F4/80−GR-1<sup>+</sup> cells from at least 19–29 fields of the brain and spinal cord were quantitated and shown (lower). Original magnification, 400x (A,B). Scale bar, 100µm (A,B). <sup>∗</sup>*p* < 0.05, ∗∗*p* < 0.01, ∗∗∗*p* < 0.001, ∗∗∗∗*p* < 0.0001. Student's *t*-test (A,B). Error bars represents ± SEM (A,B). *n* = 5 mice/group.


autoimmune encephalomyelitis. J Immunol. (2007) 179:5228–37. doi: 10.4049/jimmunol.179.8.5228


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

# A T Cell Suppressive Circuitry Mediated by CD39 and Regulated by ShcC/Rai Is Induced in Astrocytes by Encephalitogenic T Cells

Cristina Ulivieri <sup>1</sup> \*, Domiziana De Tommaso<sup>1</sup> , Francesca Finetti <sup>1</sup> , Barbara Ortensi 2,3 , Giuliana Pelicci 2,3, Mario Milco D'Elios <sup>4</sup> , Clara Ballerini <sup>4</sup> and Cosima T. Baldari <sup>1</sup> \*

*<sup>1</sup> Department of Life Sciences, University of Siena, Siena, Italy, <sup>2</sup> Department of Experimental Oncology, European Institute of Oncology, Milan, Italy, <sup>3</sup> Department of Translational Medicine, Piemonte Orientale University "Amedeo Avogadro", Novara, Italy, <sup>4</sup> Department of Experimental and Clinical Medicine, University of Florence, Florence, Italy*

#### Edited by:

*Craig Stephen Moore, Memorial University of Newfoundland, Canada*

#### Reviewed by:

*Igal Ifergan, Northwestern University, United States Kingston H. Mills, Trinity College Dublin, Ireland*

#### \*Correspondence:

*Cristina Ulivieri cristina.ulivieri@unisi.it Cosima T. Baldari cosima.baldari@unisi.it*

#### Specialty section:

*This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Immunology*

Received: *21 December 2018* Accepted: *23 April 2019* Published: *10 May 2019*

#### Citation:

*Ulivieri C, De Tommaso D, Finetti F, Ortensi B, Pelicci G, D'Elios MM, Ballerini C and Baldari CT (2019) A T Cell Suppressive Circuitry Mediated by CD39 and Regulated by ShcC/Rai Is Induced in Astrocytes by Encephalitogenic T Cells. Front. Immunol. 10:1041. doi: 10.3389/fimmu.2019.01041* Multiple sclerosis is an autoimmune disease caused by autoreactive immune cell infiltration into the central nervous system leading to inflammation, demyelination, and neuronal loss. While myelin-reactive Th1 and Th17 are centrally implicated in multiple sclerosis pathogenesis, the local CNS microenvironment, which is shaped by both infiltrated immune cells and central nervous system resident cells, has emerged a key player in disease onset and progression. We have recently demonstrated that ShcC/Rai is as a novel astrocytic adaptor whose loss in mice protects from experimental autoimmune encephalomyelitis. Here, we have explored the mechanisms that underlie the ability of Rai−/<sup>−</sup> astrocytes to antagonize T cell-dependent neuroinflammation. We show that Rai deficiency enhances the ability of astrocytes to upregulate the expression and activity of the ectonucleotidase CD39, which catalyzes the conversion of extracellular ATP to the immunosuppressive metabolite adenosine, through both contact-dependent and–independent mechanisms. As a result, Rai-deficient astrocytes acquire an enhanced ability to suppress T-cell proliferation, which involves suppression of T cell receptor signaling and upregulation of the inhibitory receptor CTLA-4. Additionally, Rai-deficient astrocytes preferentially polarize to the neuroprotective A2 phenotype. These results identify a new mechanism, to which Rai contributes to a major extent, by which astrocytes modulate the pathogenic potential of autoreactive T cells.

Keywords: astrocyte, EAE, neuroinflammation, molecular adaptor, T cell signaling, ATP-degrading enzyme

# INTRODUCTION

Astrocytes are the first CNS resident cells encountered by infiltrating autoreactive T cells in multiple sclerosis (1, 2). Astrocytes contribute to neuroinflammation in multiple sclerosis and in the mouse experimental autoimmune encephalomyelitis (EAE) model by promoting encephalitogenic T-cell activation through their ability to act as antigen presenting cells (APC) and to upregulate the T-cell costimulatory molecules B7-1 and B7-2 (3–6). Intriguingly, T-cell suppression by astrocytes has also been documented, resulting from their ability to promote antigen-independent surface upregulation of inhibitory molecules on T cells, including the inhibitory receptor CTLA-4 and the ectonucleotidases CD39 and CD73 (7, 8). Additionally, astrocytes actively influence the generation and maintenance of effector T cells both by modulating CD4<sup>+</sup> T-cell polarization to Th1 cells and by supporting IL-2-dependent Treg cell survival (9, 10). The finding that different subsets of reactive astrocytes are induced following CNS injury, of which the A1 is neurotoxic and the A2 neuroprotective (11), adds further complexity to the role of astrocytes in CNS diseases.

Astrocytes are themselves targets of infiltrating autoreactive T cells. Antigen-independent, contact-dependent upregulation of the integrin ligands VCAM-1 and ICAM-1 on astrocytes cocultured with activated T cells has been reported (8). Additionally, infiltrating Th1 and Th17 cells modulate astrocyte function via contact-independent mechanisms involving the release of inflammatory mediators that promote astrocytic secretion of pro-inflammatory cytokines and chemokines while repressing expression of anti-inflammatory cytokines (12–15). Interestingly, while both microglia and astrocytes are targets of Th1-derived soluble factors, Th17 derived soluble factors preferentially act on astrocytes (13, 15), highlighting astrocytes as central mediators of T cell-mediated neuroinflammation.

The concentration of ATP and its metabolite, adenosine, in the CNS microenvironment has emerged as a central factor in the modulation of neuroinflammation in multiple sclerosis/EAE (16). Elevated extracellular ATP (eATP) is sensed as a danger signal, promoting inflammation, while adenosine exhibits strong anti-inflammatory and immunosuppressive activities (17). The ectonucleotidases CD39 and CD73 are responsible for the conversion of ATP to adenosine. Altered expression and/or function of these enzymes have been associated to multiple sclerosis (18). Additionally, activation of the adenosine receptor A2AR has been shown to attenuate CNS inflammation and EAE severity (19, 20), and conversely genetic ablation of A2AR to exacerbate the disease (21), underscoring a key role for adenosine in controlling disease development. In support of this notion, treatment of multiple sclerosis patients or EAE mice with inosine, which similar to adenosine binds to the A1A, A2A, and A3A receptors, ameliorates disease onset and severity by inhibiting inflammatory cell entry into the CNS, astroglial activation and demyelination (22).

We have recently reported that deficiency of ShcC/Rai, a member of the Shc family of protein adaptors, protects mice from demyelination and prevents reactive astrogliosis during EAE notwithstanding enhanced CNS infiltration by encephalitogenic Th17 cells, due to reduced astrocytic production of proinflammatory molecules in response to T cell-derived factors (23). Here we have addressed the outcome of Rai deficiency on the ability of astrocytes to generate a T cell suppressive microenvironment through eATP degradation. We show that Rai-deficient astrocytes have an enhanced ectonucleotidase activity and that they upregulate CD39 expression when exposed to conditioned media from encephalitogenic T cells, which results in their enhanced ability to suppress T cells through inhibition of TCR signaling and upregulation of CTLA-4.

# MATERIALS AND METHODS

#### Mice

Rai−/<sup>−</sup> mice in the C57BL/6J background (24, 25) and C57BL/6J controls were used. Animals were housed in a pathogen-free and climate-controlled (20 ± 2 ◦C, relative humidity 55 ± 10%) animal facility at the University of Siena. Mice were provided with water and pelleted diet ad libitum. All cages are provided with environmental enrichment in the form of nesting material and mouse houses. Procedures and experimentation were carried out in accordance with the 2010/63/EU Directive and approved by the Italian Ministry of Health.

#### Induction of EAE, Isolation of Glial Cells, and Generation of MOG-Specific T Cell Lines

EAE was induced in 8- to 10-week-old female mice (three Rai−/<sup>−</sup> and three wild-type C57BL/6J mice) by subcutaneous injection of 200 µg MOG35−<sup>55</sup> peptide emulsified in an equal volume of complete Freund's adjuvant (CFA) containing 6 mg/ml M. tuberculosis H37Ra (Difco Laboratories, Detroit, MI). Mice, selected by sex, age and strain, were randomly allocated to experimental groups and randomly treated. The experimental unit was single animal. We observed similar variance between the groups that were compared. On day 0 and 2 mice were injected i.p. with 300 ng B. pertussis toxin (Calbiochem, Darmstadt, Germany). Mice were monitored daily by two independent researchers and clinical scores were assigned according to the standard 0 to 5 scale (23, 26). Brain and spinal cords were isolated from EAE mice (15 days post-immunization) and total glial cells were obtained as described (23).

To generate MOG35−<sup>55</sup> specific T cells, splenocytes and lymph nodes were harvested at day 7 after immunization with MOG35−<sup>55</sup> peptide (three wild-type C57BL/6J mice) and expanded with 50µg/ml MOG35−<sup>55</sup> and 20 U/ml IL-2 in RPMI1640 with 10% BCS. After 7 days cells were re-stimulated with autologous bone marrow-derived dendritic cells, MOG35−<sup>55</sup> peptide and IL-2, for 7 days. Cells underwent 2 rounds of stimulation before being used. The frequency of GM-CSF-, TNFα-, IFNγ-, or IL-17a- producing cells among MOG-T cells have been assessed by flow cytometry (% GM-CSF<sup>+</sup> = 4.5 ± 1, % IL-17<sup>+</sup> = 19 ± 4, % IFNγ <sup>+</sup> = 55 ± 3, and % TNFα <sup>+</sup> = 27 ± 0.5).

#### Primary Astrocyte Culture and Treatments

Astrocyte cultures were prepared from newborn mice (15 Rai+/<sup>+</sup> and 15 Rai−/−) as described (27). Cerebral cortices were dissociated using the Neural Tissue Dissociation kit (T) (Miltenyi Biotec, Bergisch Gladbach, Germany) and the cells were cultured in flasks. For astrocytes monoculture, supernatants containing microglia were eliminated and adherent cells were trypsinized and replated. The purity of astrocytes was ≥95% as assessed by GFAP staining.

Treatment with IFNγ (10 ng/ml) or IL-17 (50 ng/ml) was performed in serum-free medium for ATP, adenosine and phosphate measurements or in complete medium for flow cytometric analysis and qRT-PCR analysis of CD39 and

*(Continued)*

FIGURE 1 | Rai−/−) astrocytes stimulated with IL-17 or IFNγ for 5 h or left untreated (-), then depleted of their culture supernatant and incubated with 1 mM ATP. Data are presented as mean fold change ± SD of specific enzymatic activities (nmol free phosphate/mg protein/min) in Rai+/<sup>+</sup> astrocytes and Rai−/<sup>−</sup> astrocytes, with unstimulated Rai+/<sup>+</sup> astrocytes taken as 1 (*n* = 3). (E) Quantification of adenosine in culture supernatants of astrocytes treated as in D. Data are presented as mean ± SD of adenosine concentration (µM) (*n* = 3). (F) Immunoblot analysis with anti-Rai or anti-CD39 antibodies of RanBPM-specific immunoprecipitates from total cell lysates of Rai+/<sup>+</sup> and Rai−/<sup>−</sup> astrocytes treated with IFNγ (10 ng/ml) for 15 min (*n* = 2). The quantification by laser densitometry of the levels of each of the proteins normalized to the level of RanBPM in each sample is shown (*n* = 2). 2-Way ANOVA and Mann–Whitney test \*\*\*\**p* < 0.0001, \*\**p* < 0.01, \**p* < 0.05.

CD73 expression and immunoprecipitation assays. Surface upregulation of CD39 and CD73 was analyzed in astrocytes stimulated for 120 h (peak of expression of CD39, as assessed in a preliminary time course analysis; **Supplementary Figure 2**) with pro-inflammatory cytokines. No surface upregulation of CD73 was found at any time point (data not shown). For the treatment with conditioned media from MOG-T cells, the culture medium was replaced with the culture supernatants from IL-2-stimulated MOG T cells in the presence or absence of a neutralizing anti-IFNγ mAb (e Bioscence). Alternatively, MOG T cells were added to astrocytes as such or previously pulsed with MOG35−<sup>55</sup> peptide.

# Splenocytes, CD4<sup>+</sup> T Cell Purification and Treatments

Mouse splenic mononuclear cells were separated by Mouse lympholyte gradient centrifugation (Cedarlane Laboratories, Netherlands) and resuspended in RPMI 10% BCS (two wild-type C57BL/6J mice).

Alternatively, CD4<sup>+</sup> T cells were enriched from spleen using DynabeadsTM UntouchedTM Mouse CD4 Cells Kit (Invitrogen).

Cells were treated with immobilized anti-CD3 (2 mg/ml; eBiosciences) and anti-CD28 (2 mg/ml; eBiosciences) mAb for 72 h, alone or in combination with either the nonhydrolyzable adenosine analog NECA (10µM) (Sigma-Aldrich) or supernatants from IFNγ-treated Rai−/<sup>−</sup> or Rai+/<sup>+</sup> astrocytes, in presence or absence of the ectonucleotidase inhibitor ARL67156 (100µM) (Sigma-Aldrich). Alternatively, cells were pre-treated with supernatants from IFNγ-treated Rai−/<sup>−</sup> or Rai+/<sup>+</sup> astrocytes (diluted 1:2 with culture medium) in the presence or absence of ARL67156 (100µM) for 1 h at 37◦C and activated with soluble anti-CD3 and anti-CD28 mAbs in presence or absence of 10 µM NECA.

#### eATP, Adenosine, and Ectonucleotidase Activity Measurements

ATP levels in the astrocyte supernatants and cells were measured using a luciferin/luciferase assay (ATP Determination Kit A22066; Invitrogen) and a luminometer (Berthold Lumat LB 9501) according to the manufacturer's instructions. Adenosine levels were measured on astrocytes supernatants using a fluorometric assay (Adenosine Assay Kit; Cell Biolabs, INC.) and a Fluorometer (TECAN) according to the manufacturer's instructions.

For determination of nucleotide hydrolysis free phosphate was measured using the Malachite Green Phosphate Assay Kit (POMG-25H) (BioAssay Systems) at 620 nm on a microplate reader, according to the manufacturer's protocol. Specific activity

was calculated using a calibration curve and expressed as nmol Pi released/mg protein/min.

Each sample was run in triplicate. Remaining cells were lysed in 0.02% SDS in phosphate-buffered saline (PBS) and protein content determined by the Pierce BCA protein assay kit (Thermo Fisher Scientific).

#### Cell Lysis, Immunoprecipitations, and Immunoblots

Two-Way ANOVA, \*\*\**p* < 0.001, \*\**p* < 0.01, \**p* < 0.05.

Cells were lysed in 1% (v/v) Triton X-100 in 20 mM Tris-HCl (pH 8), 150 mM NaCl in the presence of Protease Inhibitor Cocktail Set III (Calbiochem) and 0.2 mg Na orthovanadate/ml. Postnuclear supernatants were resolved by SDS-PAGE and transferred to nitrocellulose. Alternatively, postnuclear supernatants were immunoprecipitated using RanBPM polyclonal antibody (Proteintech) and protein A Sepharose (GE Healthcare). Immunoblots were carried out using peroxidase-labeled secondary antibodies (GE Healthcare) and a chemiluminescence detection kit (Bio-rad Laboratories Inc., Milan, Italy). Immunoblots were scanned and quantitated using ImageJ software.

#### Flow Cytometry and Proliferation Assays

Flow cytometric analysis of astrocytes, MOG-T cells and splenocytes was performed using AlexaFluor488-, PE-, PerCPconjugated anti-mouse antibodies to: GFAP (clone GA5; eBioscence), CD39 (clone 24DMS1; eBioscence), CD73 (clone TY11.8; Biolegend), CTLA-4 (clone UC10-4B9; Biolegend), IL-17A (clone TC11-18H10; Becton Dickinson), IFN-γ (clone XMG1.2; Becton Dickinson), GM-CSF (clone MPI-22E9; Biolegend), TNFα (Clone MP6-XT-22; Biolegend), and isotype control antibodies. Samples were acquired on Guava Easy Cyte cytometer (Millipore) and analyzed with FlowJo software (TreeStar Inc., Ashland, OR, USA).

Proliferation was measured on CFSE loaded cells (Molecular Probes, Thermo Fisher Scientific) by flow cytometry.

#### RNA Purification and RT-qPCR

Total RNA was isolated and purified from brain, astrocytes and splenocytes using the RNeasy Plus Mini Kit (Quiagen) according to the manufacturer's instructions. First-strand cDNAs were generated using the iScriptTM cDNA Synthesis Kit (Bio-Rad). RT-qPCR was performed using the SsoFastTM EvaGreen <sup>R</sup> supermix kit (BIO-RAD) and specific pairs of primers listed in **Supplementary Table 1**.

#### Statistical Analyses

One-way ANOVA with post-hoc Tukey or 2-way ANOVA with post-hoc Sidak test were used for experiments where multiple groups were compared. Mann–Whitney rank-sum tests were also performed to determine the significance of the differences between two groups. Statistical analyses were performed using GraphPad Prism Software (Version 8). A P < 0.05 was considered as statistically significant.

#### RESULTS

#### Rai Dampens CD39 Enzyme Activity in Astrocytes in Response to IFNγ Treatment

To address the impact of Rai deficiency on the eATP-degrading activity of astrocytes, ATP was quantified in culture supernatants from Rai+/<sup>+</sup> and Rai−/<sup>−</sup> astrocytes generated from newborn mice brain, stimulated or not with IL-17 or IFNγ. Lower levels of eATP were found in culture supernatants of Rai−/<sup>−</sup> astrocytes compared to Rai+/<sup>+</sup> astrocytes, despite the fact that the total levels of ATP were comparable (**Figures 1A,B**). No differences in surface CD39/CD73 expression were observed under these conditions (**Figure 1C**, **Supplementary Figure 1**), suggesting that Rai might modulate the eATP-degrading activity of astrocytes.

To test this possibility we compared the ATP-degrading activity of control and Rai−/<sup>−</sup> astrocytes stimulated or not with IL-17 or IFNγ by incubating cells depleted of their culture supernatant with 1 mM ATP and measuring free phosphate production. Both cytokines promoted ATP-degradation, with Rai−/<sup>−</sup> astrocytes hydrolyzing ATP more efficiently, both under basal conditions and following IFNγ or IL-17 treatment (**Figure 1D**). Quantification of adenosine in culture supernatants from astrocytes added with exogenous ATP showed that IFNγ, but not IL-17, enhanced adenosine production, which was further enhanced by Rai deficiency (**Figure 1E**). These data indicate that IFNγ modulates adenosine generation by astrocytes and that Rai dampens the activity of ATP-degrading enzymes in these cells.

To translate these results to the context of EAE we measured the ATP-degrading activity of astrocytes isolated from the spinal cord of Rai−/<sup>−</sup> and control EAE mice. Similar to the results obtained on astrocytes derived from the brain of newborn mice, Rai−/<sup>−</sup> astrocytes obtained from the CNS of EAE mice degraded ATP more efficiently compared to their wild-type counterparts (**Supplementary Figure 3**). These data suggest that the protective role of Rai deficiency in astrocytes toward neuroinflammation in EAE could be dependent at least in part on the ability of Rai to negatively control eATP degradation and adenosine generation.

#### Rai Couples CD39 to its Negative Regulator RanBPM

To address the mechanism responsible for the ability of Rai to negatively control eATP degradation we focused on the rate-limiting enzyme of the cascade which converts ATP/ADP to adenosine, namely CD39. Since the scaffolding protein RanBPM binds to the cytosolic tail of CD39 and downregulates its ectonucleotidase activity (28), we hypothesized that Rai may participate in this molecular complex to restrain CD39 function. Rai+/<sup>+</sup> and Rai−/<sup>−</sup> astrocytes were left untreated or were treated with IFNγ and post-nuclear supernatants were immunoprecipitated with anti-RanBPM antibodies. RanBPMspecific immunoprecipitates were analyzed by immunoblotting with anti-Rai and anti-CD39 Abs. Rai was found to associate with RanBPM in response to IFNγ (**Figure 1F**). Interestingly, the IFNγ-dependent association of RanBPM with CD39 was reduced in Rai−/<sup>−</sup> astrocytes compared to control astrocytes (**Figure 1F**), indicating that IFNγR signaling promotes CD39 activation and suggesting that Rai limits CD39 activity by promoting RanBPM recruitment to CD39.

#### Rai Negatively Controls the Contact-Dependent and -Independent Upregulation of CD39 and CD73 Elicited by Encephalitogenic T Cells

Astrocytes have been demonstrated to suppress recently activated CD4<sup>+</sup> T cells by inducing the upregulation of CD39/CD73 on their surface, which correlates with the acquisition of an immunosuppressive Th17 phenotype (8). Whether inflammatory

FIGURE 3 | Rai negatively controls the contact-dependent and -independent upregulation of CD39 and CD73 elicited by encephalitogenic T cells. (A,B) Flow cytometric analysis of surface CD73 and CD39 in Rai+/<sup>+</sup> (Astro WT) and Rai−/<sup>−</sup> (Astro Rai−/−) astrocytes untreated (-) or stimulated for 120 h with culture supernatants from MOG-specific T cells generated from WT mice (SN T-MOG). Data are presented as mean ± SD of mean fluorescence intensity (MFI) (A) or frequency of CD39highCD73high astrocytes (*n* > 3) (B). (C) Flow cytometric analysis of surface CD73 and CD39 expression in Rai+/<sup>+</sup> (Astro WT) and Rai−/<sup>−</sup> (Astro Rai−/−) astrocytes treated for 120 h with either culture supernatants from MOG-specific T cells generated from WT mice (SN T-MOG) or MOG-specific T cells depleted of their culture supernatants (T-MOG) in the presence (+Ag) or absence of MOG antigen. Representative dot plots are shown. The histograms show the frequency of the CD39highCD73high population. Data are presented as mean ± SD of the percentage of CD39highCD73high cells (*n* > 3). Representative dot plots are shown. Two-Way ANOVA, \*\*\*\**p* < 0.0001, \*\*\**p* <0.001, \*\**p* < 0.01, \**p* < 0.05.

T cells can in turn affect the expression of these ATP-degrading enzymes in astrocytes, and the role of Rai in this process, have as yet not been explored. At present no published data are available on surface expression of CD39 and CD73 in mouse primary astrocytes either in the basal state or in response to cytokines. Surface CD39 and CD73 was measured by flow cytometric analysis of control and Rai−/<sup>−</sup> astrocytes following prolonged (120 h) treatment with IL-17 or IFNγ. IL-17 had no effect on either CD39 or CD73 surface expression (**Figure 2A**). At variance, IFNγ was found to promote CD39 upregulation with, a slight, yet not significant, further increase in Rai−/<sup>−</sup> astrocytes compared to control astrocytes (**Figure 2A**). The increase in surface CD39 correlated with an increase in the levels of CD39 mRNA, as assessed by RT-qPCR. An increase in the levels of CD73 mRNA was also observed in Rai−/<sup>−</sup> astrocytes compared to control astrocytes, however this was not paralleled by an global concomitant increase in surface CD73 (**Supplementary Figure 4**). Interestingly, co-upregulation of surface CD39 and CD73 was observed in response to both IFNγ and IL-17 in a small subpopulation of astrocytes (**Figure 2B**). This CD39highCD73high subpopulation was larger in Rai−/<sup>−</sup> astrocytes compared to wild-type controls under steadystate conditions and was further expanded following IFNγ, but not IL-17, treatment (**Figure 2B**).

To mimic the CNS microenvironment shaped by infiltrating T cells during EAE, surface CD73 and CD39 were measured by flow cytometry on wild-type and Rai−/<sup>−</sup> astrocytes treated with conditioned media from MOG-specific T cells for 120 h. Under these conditions both CD39 and, to a lesser extent, CD73, were upregulated in both wild-type and Rai−/<sup>−</sup> astrocytes (**Figure 3A**). Additionally, a substantial increase in the abundance of the CD39highCD73high wild-type astrocyte subpopulation was observed, with a further significant increase in Rai−/<sup>−</sup> astrocytes (**Figure 3B**). No significant effect on the frequency of CD39highCD73high astrocytes was observed when an anti-IFNγ antibody was added to the conditioned media (data not shown), suggesting that other T cell-derived factors are responsible for the robust co-upregulation of CD39 and CD73 on this astrocyte subpopulation. These results indicate that encephalitogenic T cells promote co-upregulation of CD39 and CD73 on astrocytes in a contact-independent manner and that Rai deficiency results in an enhanced ability of astrocytes to respond to T cell-derived factors.

To understand whether surface ectonucleotidase expression on astrocytes can be further modulated by their physical contact with T cells, surface CD39 and CD73 were measured on wildtype and Rai−/<sup>−</sup> astrocytes co-cultured for 120 h in the presence or absence of MOG with encephalitogenic T cells previously depleted of their culture supernatant. Under these conditions an increase in the abundance of the CD39highCD73high astrocyte subpopulation was observed, independently of the presence of antigen (**Figure 3C**). However, Rai deficiency did not affect the contact-dependent upregulation of CD39 or CD73, as opposed to the enhancement observed in the presence of conditioned media from MOG-specific T cells (**Figure 3C**). Hence, encephalitogenic T cells elicit a co-upregulation of CD39 and CD73 on astrocytes in both a contact-independent and a contact-dependent but antigen-independent manner, and the contact-independent response is enhanced in astrocytes lacking Rai.

through cell-cell contact-independent mechanisms. (A) Flow cytometric analysis of CFSE-labeled splenic mouse cells from wild-type mice stimulated for 72 h with anti-CD3/CD28 antibodies (3+28) in combination with either NECA or supernatants from IFNγ-treated Rai−/<sup>−</sup> (SN AstroRai−/−) or Rai+/<sup>+</sup> (SN AstroWT) astrocytes in presence or absence of ARL67156 (100µM) (ARL). The graph shows the mean value ± SD of the percentage of CFSElow cells (proliferating cells) (*n* = 5). (B) Immunoblot analysis of ZAP-70 phosphorylation *(Continued)*

# Rai−/<sup>−</sup> Astrocytes Inhibit T Cell Proliferation by Suppressing TCR Signaling and Promoting Adenosine-Dependent CTLA-4 Upregulation

The enhanced ability of Rai−/<sup>−</sup> astrocytes to co-upregulate surface CD39/CD73 and hydrolyze eATP in the presence of the pro-inflammatory cytokines or factors released by encephalitogenic T cells suggests that Rai−/<sup>−</sup> astrocytes may suppress the activity of infiltrated T cells in an adenosinedependent manner. To test this hypothesis we measured the proliferation of splenic T cells activated by CD3/CD28 costimulation in the presence of conditioned media from IFNγ-treated Rai−/<sup>−</sup> or Rai+/<sup>+</sup> astrocytes, using a nonhydrolysable adenosine analog as control. Flow cytometric analysis of CFSE-labeled splenocytes showed that culture supernatants from both IFNγ-treated wild-type and Rai−/<sup>−</sup> astrocytes, but not from untreated astrocytes, inhibited T cell proliferation (**Figure 4A**, **Supplementary Figure 5**). These effects were neutralized by treatment with the CD39/CD73 inhibitor ARL67156 (**Figure 4A**), indicating that they were mediated by adenosine. ARL67156 alone had no effect on CD3/CD28-dependent proliferation (**Supplementary Figure 5**). Suppression of T cell proliferation was more profound in the presence of conditioned media from Rai−/<sup>−</sup> astrocytes (**Figure 4A**), consistent with their higher adenosine content (**Figure 1**).

Adenosine binding to A2AR results in an elevation in intracellular cAMP, which effectively inhibits TCR signaling through the PKA-dependent activation of the kinase Csk, a negative regulator of the initiating kinase Lck (29). Lck is required for the phosphorylation-dependent recruitment of the kinase ZAP-70 to the TCR, a key step for signal propagation (30). To explore the ability of astrocytes to modulate TCR signaling through their ATP-hydrolysing activity, splenic T cells were activated by CD3/CD28 costimulation in the presence of conditioned media from wild-type or Rai−/<sup>−</sup> astrocytes in the presence or absence of ARL67156, and the activation of ZAP-70 was measured by immunoblot using a phosphospecific antibody. Consistent with their inhibitory effect on T cell proliferation, supernatants from IFNγ-treated, but not from untreated, astrocytes suppressed ZAP-70 activation, with a higher efficiency for supernatants from Rai−/<sup>−</sup> astrocytes (**Figure 4B**, **Supplementary Figure 5**). Inhibition was fully relieved by the ectonucleotidase inhibitor (**Figure 4B**), supporting the notion that suppression of TCR signaling by astrocyte-derived factors is mediated by CD39/CD73-dependent adenosine production.

In addition to directly inhibiting TCR signaling, adenosine suppresses T cell responses by inducing the cAMP/PKAdependent expression of the inhibitory receptor CTLA-4, which blocks CD28-mediated costimulation (31, 32). Additionally, an upregulation of CTLA-4 expression has been reported in T cells exposed to astrocytes or to their conditioned media (7), suggesting a mechanistic link between these observations and our finding that astrocytes effectively degrade eATP. To address this issue, surface CTLA-4 was measured by flow cytometry on T cells exposed to conditioned media from IFNγ-treated wild-type or Rai−/<sup>−</sup> astrocytes. The levels of T cell surface CTLA-4 were higher in the presence of IFNγ-treated astrocyte culture supernatants, similar to adenosine-treated T cells. Significantly higher levels of CTLA-4 were observed in the presence of supernatants from Rai−/<sup>−</sup> astrocytes compared to wild-type astrocytes (**Figure 5A**). The enhanced ability of conditioned media from IFNγ-treated Rai−/<sup>−</sup> astrocytes to induce A2AR signaling compared with wild-type astrocytes was further supported by an enhancement of PKA activity and CREB phosphorylation (**Figure 5B**). Culture supernatants from untreated astrocytes had no effect on surface CTLA4 (**Supplementary Figure 5**). Supernatants from IFNγ-treated Rai−/<sup>−</sup> astrocytes, but not from IFNγ-treated WT astrocytes, induce significant CTLA4 upregulation also on CD3/CD28 activated T cells when comparing with activated T cells without addition of astrocytes supernatant (**Figure 5C**). This effect was completely neutralized by the ectonucleotidase inhibitor (**Figure 5C**), indicating that CTLA-4 upregulation by astrocytes is dependent on their ectonucleotidase activity. ARL67156 alone had no effect on CD3/CD28-dependent CTLA4 upregulation (**Supplementary Figure 5**).

While CTLA-4 can be rapidly expressed at the T-cell surface through the release of an intracellular pool stored in lysosomes (33), adenosine-dependent CTLA-4 upregulation involves de novo gene expression triggered by cAMP-dependent activation of the transcription factor CREB (34). To understand whether the increase in surface CTLA-4 observed in the presence of astrocyte culture supernatants was the result of transcriptional activation, we measured CTLA-4 mRNA levels on splenic T cells activated by CD3/CD28 costimulation in the presence of conditioned media from IFNγ-treated wild-type or Rai−/<sup>−</sup> astrocytes. Real-time RT-PCR analysis revealed an increase in the levels of the transcript for both the full length and the soluble form of CTLA-4, which results from alternative splicing and inhibits T-cell responses by binding B7 on APCs (35, 36) (**Figure 5D**). Collectively, these results indicate that astrocytes inhibit T cell activation and proliferation by suppressing TCR signaling and enhancing CTLA-4 expression through CD39/CD73-mediated adenosine production and cAMP/PKA signaling, which are enhanced in the absence of Rai.

#### A2 Reactive Astrocytes Are Induced by Encephalitogenic T Cells in the Absence of Rai

Astrocytes can polarize toward a neurotoxic A1 phenotype or a neuroprotective A2 anti-inflammatory phenotype depending on the disease. To date the signaling mechanisms responsible for the shift elicited by inflammatory cues remains largely unknown (11), and whether encephalitogenic T cells drive astrocyte polarization has as yet not been explored.

To investigate whether soluble factors released by encephalitogenic T cells induce astrocyte polarization, A1 or A2-specific transcripts were measured in astrocytes cultured for 24 h in conditioned media from encephalitogenic T cells. While no effect was detected in control astrocytes, under these

astrocytes either untreated or treated for 24 h with culture supernatants from MOG-specific T cells generated from WT mice. Data from 3 independent experiments, each carried out on the pooled astrocytes from at least 5 Rai+/<sup>+</sup> or 5 Rai−/<sup>−</sup> mice, are presented as mean value ± SD. The levels of the different transcripts were normalized to GAPDH, used as housekeeping gene. (B) Real-Time PCR analysis of the A1-specific transcripts H2-D1, Serping and C3 in the brain of Rai+/<sup>+</sup> EAE mice (EAE WT) and Rai−/<sup>−</sup> EAE mice (EAE Rai−/−) 15 days post immunization. Data are presented as mean value ± SD obtained on three Rai+/<sup>+</sup> and three Rai−/<sup>−</sup> EAE mice. 2-way ANOVA; \*\**p* < 0.01, \**p* < 0.05.

conditions a strong induction of the A2-specific transcripts Emp1 and S100a10 was detected in Rai−/<sup>−</sup> astrocytes. At variance, Rai deficiency did not affect the levels of the A1-specific transcript H2-D1 and Serping1 (**Figure 6A**). Consistent with the protective role played by Rai deficiency in the EAE mouse model, lower levels of the A1-specific transcript H2-D1, Serping1 and C3 were found in the brain of Rai−/<sup>−</sup> EAE mice compared with control EAE mice (**Figure 6B**). These data identify Rai as a signaling molecule that restrains the polarization of astrocytes to the neuroprotective A2 phenotype.

#### DISCUSSION

The cross-talk of astrocytes with encephalitogenic T cells is centrally implicated in multiple sclerosis pathogenesis (13, 37). Astrocytes respond to Th1 and Th17 cell-derived cytokines by producing factors that attract inflammatory cells. Additionally, they act as APC to promote effector T cell activation and expansion (38). However, activated astrocytes also deploy a variety of strategies to counteract inflammation and limit neuronal damage, including induction of Fas-mediated apoptosis of infiltrated T cells, skewing of T-cell polarization to a protective Th2 phenotype and Treg-dependent suppression of encephalitogenic T-cells (10, 39, 40). Here we document a new protective mechanism exploited by astrocytes to suppress T-cell activation and proliferation, which involves the upregulation of the astrocytic expression and activity of the ectonucleotidases CD39 and CD73 in response to pro-inflammatory factors released by encephalitogenic T cells. Additionally, we identify Rai as a negative regulator of this inhibitory circuitry.

The balance between eATP and adenosine has emerged as an important factor in the control of neuroinflammation to which both infiltrating T cells and astrocytes contribute. eATP boosts T-cell activation and promotes Th17 cell differentiation while inhibiting Treg cell differentiation and stability (41). Additionally eATP triggers microglia activation (42). On the other hand, ATP degradation to adenosine is a potent mechanism of Tcell suppression, and in fact CD39 has been established as a Treg cell marker that contributes to their inhibitory function (18, 43, 44). The function of CD39<sup>+</sup> Treg cells in MS is still unclear. Indeed in relapsing-remitting multiple sclerosis enhanced frequency of CD39<sup>+</sup> Treg cells has been reported both during relapse (45, 46) and during the remission phase (47). Adenosine suppresses TCR signaling by interacting with the adenosine receptor A2AR (48), which activates a cAMP/PKA axis that inhibits TCR signaling at multiple steps (29). We found that conditioned media from IFNγ-activated astrocytes were able to inhibit T-cell proliferation and that this effect was abrogated by an ectonucleotidase inhibitor, indicating a contact-independent, adenosine-mediated mechanism of Tcell suppression. Accordingly, proximal TCR signaling, which requires activation of the kinase Lck that is inhibited by cAMP (49), was impaired when T cells were activated in the presence of conditioned media from astrocytes. The ability of the ectonucleotidase inhibitor to reverse this effect highlights a major role for the ATP-degrading, adenosine-elevating activity of CD39 in T-cell suppression by astrocytes.

Interestingly, we found that the ATP-degrading activity of astrocytes contributes to the suppression of T-cell proliferation through an additional, cAMP-dependent mechanism involving upregulation of the inhibitory receptor CTLA-4. Astrocytes have been shown to induce the contact-independent CTLA-4 upregulation on activated T cells, suggesting the presence of soluble inhibitory factors (7). The fact that the enhancing effect of culture supernatants from IFNγ-activated astrocytes on T cell expression of CTLA-4 can be reversed by an ectonucleotidase inhibitor supports the notion that a major one among these factors is adenosine. Of note, we found that surface CTLA-4 upregulation was paralleled by an increase in the levels of specific transcripts, consistent with the fact that A2AR triggering on T cells promotes CTLA-4 transcription through its cAMP-elevating activity and the resulting activation of the transcription factor CREB (34). Indeed, we found that conditioned media from IFNγtreated astrocytes were able to trigger CREB activation in an ectonucleotidase-dependent manner.

Interestingly, the levels of surface CD39 expression were upregulated in response to long-term treatment with IFNγ but not IL-17, while surface CD73 expression was not affected, highlighting CD39 as a limiting factor in ATP degradation by astrocytes and indicating that this protective response may be elicited preferentially by Th1 cells. It is however noteworthy that both IFNγ and IL-17, and to an even greater extent conditioned media from encephalitogenic T cells, increase the abundance of a CD39<sup>+</sup> astrocyte subpopulation that co-expresses CD73, which may account for the increased ATP-degrading activity detected under these conditions. This finding supports the notion that astrocytes shift toward an immunosuppressive phenotype in a Th1/Th17-conditioned microenvironment. Of note, transcription of the gene encoding CD39 has been reported to be activated by cAMP (50). Taking into account the fact that astrocytes are able to promote CD39 upregulation on co-cultured activated T cells (8), a possible scenario is that the resulting adenosine-generating activity of T cells may trigger adenosine signaling on astrocytes, thereby promoting cAMP accumulation and transcriptional activation of CD39, which would in turn result in suppressive adenosine-mediated signaling in T cells. ROS-dependent upregulation of CD39 has been recently reported in CD8 T cells (51). While the impact of ShcC/Rai on ROS production has as yet not been investigated, our finding that surface CD39 expression was upregulated in response to longterm treatment with IFNγ opens the possibility that enhanced ROS generation may account for the higher CD39 expression also in astrocytes.

Our results identify Rai as a negative regulator of astrocytemediated, adenosine-dependent T-cell suppression. Indeed, the T cell suppressive effects of conditioned media from IFNγ-treated astrocytes were enhanced by Rai deficiency. This results both from the enhanced ability of Rai−/<sup>−</sup> astrocytes to degrade eATP to adenosine in response to short-term IFNγ treatment and from the greater increase in CD39 expression and frequency of the CD39+CD73<sup>+</sup> subpopulation after long-term IFNγ treatment compared to their wild-type counterparts. These results provide insights into the mechanisms responsible for the protective effect of Rai deficiency in astrocytes from encephalitogenic T celldependent neurodegeneration (23). Rai was initially identified as a molecular adaptor that couples the receptor tyrosine kinase Ret to Akt in neuronal cells (52). We showed that in T cells Rai limits antigen receptor signaling by impairing ZAP-70 recruitment to the activated TCR (53). The restraining effects of Rai on IFNγdependent CD39 expression and activity could be hypothesized to result from a similar mechanism involving the ability of Rai to exploit its adaptor function to interfere with IFNγR signaling.

That Rai is able to modulate the activity CD39 is intriguing. Studies on this ectoenzyme have been largely focused on the extracellular domain, which represents the most conspicuous part of the protein (54). Interestingly, recent evidence indicates that the short cytosolic tail is also implicated in the regulation of CD39 activity. Namely, RanBPM, an interactor of the small GTPase Ran, has been shown to associate with the cytosolic tail of CD39, which negatively regulates its activity, in B cells (28). RanBPM acts as a scaffolding protein, interacting with a variety of membrane proteins and receptors (55). Here we demonstrate that Rai forms a complex with RanBPM and promotes IFNγ-dependent recruitment of RanBMP to CD39, thereby restraining its function, which places Rai in the negative regulatory circuitry of CD39, accounting for the enhanced CD39 activity in Rai−/<sup>−</sup> astrocytes.

Reactive astrocytes may adopt two distinct phenotypes, A1 and A2, with A1 astrocytes being neurotoxic and A2 astrocytes neuroprotective (11). Although astrocyte conversion to the A1 phenotype has been shown to be modulated by activated microglia in human neurodegenerative diseases including multiple sclerosis (11), the underlying mechanism and the impact of encephalitogenic T cells on this process remain unknown. Our data provide evidence for a new role of Rai as a negative regulator of astrocyte polarization to the A2 phenotype, highlighting an additional mechanism involving astrocytes that contributes to attenuating EAE severity in Rai−/<sup>−</sup> mice (23).

In conclusion, the results presented in this report show a reciprocal interplay whereby pathogenic T cells trigger CD39 expression and activity on astrocytes, highlighting this ectonucleotidase as a hub where signals from T cells and astrocytes converge to modulate the pathogenic activity of T cells in the CNS. They moreover identify astrocytic Rai as a central player in this cross-talk which unleashes the pathogenic effects of infiltrated encephalitogenic T cells in the CNS by negatively regulating a protective CD39-based T cell suppression circuitry. Finally, they provide evidence that Rai negatively regulates the polarization of reactive astrocytes toward a neuroprotective A2 phenotype. Both the enhanced T cell suppressive activity of

REFERENCES


Rai-deficient astrocytes and their enhanced A2 polarization are likely to account for our finding that Rai deficiency in astrocytes prevents reactive astrogliosis and ameliorates EAE (23).

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations on the protection of animals used for scientific purposes, Directive 2010/63 EU of the European Parliament and of the Council. The protocol was approved by the Italian Ministry of Health.

# AUTHOR CONTRIBUTIONS

CTB, CU, CB, GP, and MMD contributed conception and design of the study. CU, DD, and FF performed the experiments. CTB, CU, and DD wrote the first draft of the manuscript. BO, GP, CB, and MMD provided key reagents and expertise. All authors contributed to manuscript revision, read, and approved the submitted version.

#### FUNDING

This work was supported by Fondazione Italiana Sclerosi Multipla Grant # 2014/R/3 to CTB.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01041/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 Ulivieri, De Tommaso, Finetti, Ortensi, Pelicci, D'Elios, Ballerini and Baldari. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Novel Sex-Dependent Target for the Treatment of Postoperative Pain: The NLRP3 Inflammasome

Ashley M. Cowie<sup>1</sup> , Bonnie N. Dittel 2,3 and Cheryl L. Stucky <sup>1</sup> \*

*<sup>1</sup> Department of Cell Biology, Neurobiology, and Anatomy, Medical College of Wisconsin, Milwaukee, WI, United States, <sup>2</sup> Blood Research Institute, Versiti, Milwaukee, WI, United States, <sup>3</sup> Department of Microbiology and Immunology, Medical College of Wisconsin, Milwaukee, WI, United States*

In recent years the innate immune system has been shown to be crucial for the pathogenesis of postoperative pain. The mediators released by innate immune cells drive the sensitization of sensory neurons following injury by directly acting on peripheral nerve terminals at the injury site. The predominate sensitization signaling pathway involves the proinflammatory cytokine interleukin-1β (IL-1β). IL-1β is known to cause pain by directly acting on sensory neurons. Evidence demonstrates that blockade of IL-1β signaling decreases postoperative pain, however complete blockade of IL-1β signaling increases the risk of infection and decreases effective wound healing. IL-1β requires activation by an inflammasome; inflammasomes are cytosolic receptors of the innate immune system. NOD-like receptor protein 3 (NLRP3) is the predominant inflammasome activated by endogenous molecules that are released by tissue injury such as that which occurs during neuropathic and inflammatory pain disorders. Given that selective inhibition of NLRP3 alleviates postoperative mechanical pain, its selective targeting may be a novel and effective strategy for the treatment of pain that would avoid complications of global IL-1β inhibition. Moreover, NLRP3 is activated in pain in a sex-dependent and cell type-dependent manner. Sex differences in the innate immune system have been shown to drive pain and sensitization through different mechanisms in inflammatory and neuropathic pain disorders, indicating that it is imperative that both sexes are studied when researchers investigate and identify new targets for pain therapeutics. This review will highlight the roles of the innate immune response, the NLRP3 inflammasome, and sex differences in neuropathic and inflammatory pain.

Keywords: NLRP3, interleukin-1β, sex differences, pain, tissue injury, innate immunity

#### INTRODUCTION

A unique combination of molecular and cellular factors can lead to acute and chronic pain conditions with varying pathologies. Despite this, pain is categorized into the following broad categories: inflammatory, neuropathic, and syndrome-based (e.g., fibromyalgia). There is overlap between these generalized categories. For example, inflammation can result in nerve damage, nerve injury involves inflammation, and syndrome-based pain can be neuropathic or inflammatory or both. Inflammatory pain occurs with

#### Edited by:

*Craig Stephen Moore, Memorial University of Newfoundland, Canada*

#### Reviewed by:

*Bradley Kerr, University of Alberta, Canada Nader Ghasemlou, Queen's University, Canada*

> \*Correspondence: *Cheryl L. Stucky cstucky@mcw.edu*

#### Specialty section:

*This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Neurology*

Received: *29 March 2019* Accepted: *28 May 2019* Published: *12 June 2019*

#### Citation:

*Cowie AM, Dittel BN and Stucky CL (2019) A Novel Sex-Dependent Target for the Treatment of Postoperative Pain: The NLRP3 Inflammasome. Front. Neurol. 10:622. doi: 10.3389/fneur.2019.00622*

**30**

peripheral tissue damage and the resulting tissue inflammation. Alternatively, neuropathic pain results from direct damage to nerves in the peripheral or central nervous systems. Postoperative pain has both inflammatory and neuropathic qualities (1). It is widely recognized that postoperative pain occurs as a result of the direct cutting of tissues and peripheral nerves at the surgical site.

Rodent models of postoperative pain have been consistently used to study the underlying causes of postoperative pain. Rodent models of surgical pain are strong preclinical models because the injury induced in the animal and human is similar, and therefore, these models likely recapitulate patient phenotypes and mechanisms (1–3). The most common postoperative pain model involves cutting through the skin and underlying muscle (flexor digitorum brevis), which reliably produces mechanical and heat hyperalgesia at the incision site (4–9). There is a robust immune response in this model that includes infiltration of neutrophils, macrophages, and lymphocytes. The immune response aids in wound healing, but also results in sensitization of sensory neurons to mechanical and heat stimuli (1, 10–13). The immune response begins at the incision site or site of tissue damage and moves proximally to the dorsal root ganglia and spinal cord.

There is a rapidly growing body of evidence demonstrating that the development and maintenance of postoperative pain are not solely dependent on the increased excitability of sensory neurons alone at the incision site, but they also depend on immune cell interactions with sensory neurons and activation of canonical immune receptors expressed by sensory neurons. Components of the innate immune system have emerged as crucial mediators in the development and maintenance of hypersensitivity following incision. Pattern-recognition receptors (PRRs) are part of the innate immune system and are among the first to be activated in response to tissue damage; their activation is important for the induction of immune responses leading to pathogen elimination and subsequent tissue repair (14). PRRs include cytosolic NOD-like receptors (NLRs) which, when activated, form inflammasomes. The NLR protein 3 (NLRP3) inflammasome is the best characterized NLR and has been shown to be critical in driving the immune response to sterile tissue damage (15), the type of inflammation that occurs with surgical incision. Additionally, NLRP3 is known to play a role in several painful conditions that arise from sterile tissue damage (16–30). Since the immune system is known to be sexually dimorphic, much recent attention has been given to understanding the sex differences and their causative factors that underlie painful conditions. However, little is known about the effects of sex on NLRP3 or the role of NLRP3 in postoperative pain. Therefore, this review provides a new insight into the relationship between NLRP3 and postoperative pain. Here we discuss the current understanding of sexual dimorphism in the innate immune system response to tissue injury and the role it plays in inflammatory and neuropathic pain conditions by focusing on the NLRP3 inflammasome.

# THE IMMUNE RESPONSE TO INCISIONAL INJURY

#### Immune Cell Involvement

Surgical incision results in local tissue injury, which destroys physical barriers between the body and environment, and increases the risk of exposure to environmental and commensal microbes. These consequences of surgery all lead to activation of the innate immune system and local inflammation. Inflammation occurs immediately following tissue injury as an attempt to clear debris and initiate healing. Initially immune cells such as mast cells, neutrophils, and monocytes/macrophages are recruited to the injury site by mediators that are released in tissues, by neurons and by tissue-resident immune cells (12, 31– 34). Recruitment and activation of different immune cells following injury occurs in the same sequence in both sexes. First, dermal mast cells regulate inflammation immediately following cutaneous wounding by releasing inflammatory mediators, thereby increasing vascular permeability and recruiting neutrophils (35, 36). The neutrophil recruitment is generally followed by monocyte/macrophage recruitment, which occurs 1–2 days following injury (12, 34, 37, 38). Macrophages play a dual role in wound healing, where initially they promote inflammation and then later, they switch to a reverse role where they promote the resolution of inflammation (34). Lastly, during the resolution of inflammation phase, T cells infiltrate the wound to aid in healing (34, 39).

# The NLRP3 Inflammasome and Interleukin-1β Production

Surgical trauma is aseptic and causes the release of damageassociated molecular patterns (DAMPs) (40). DAMPs are endogenous molecules that are released from damaged or dying cells and serve as a signal for tissue damage (41). Soluble DAMPs that are released as a result of incision include: heparan sulfate (42, 43), fibronectin (44, 45), hyaluronan (46–48), β-defensins (49–51), heat shock protein 70 (Hsp70) (52), and high mobility group box-1 (HMGB1) (53, 54). These DAMPs then bind to PRRs such as Toll-like Receptors (TLRs) on innate immune cells (mast cells, neutrophils, monocytes/macrophages) and sensory neurons, specifically Toll-like Receptor 4 (TLR4) (18, 49, 55–58). Stimulation of TLR4 leads to activation of the transcription factor NF-κB and upregulation of the synthesis of pro-inflammatory cytokines like interleukin-1β (IL-1β) (59). Stimulation of TLR4 also serves as the priming signal for NLRP3, the activator of IL-1β (59).

NLRP3 is predominately expressed by cells in lymphoid organs and tissues that are highly populated by immune cells. These cells include but are not limited to mast cells, neutrophils, macrophages, monocytes, dendritic cells, and neurons in both the peripheral and central nervous systems (29, 60–62). The expression of NLRP3 in these cell types must be induced by inflammatory stimuli, which prevents uncontrolled release of IL-1β. NLRP3 requires two signals for canonical activation and for IL-1β secretion: the first signal primes the cell to express NLRP3 and pro-IL-1β, and the second signal induces inflammasome assembly and activation (41, 63, 64). NLRP3 forms a scaffold with apoptosis-associated speck-like protein containing a CARD (ASC) to provide a molecular platform for activation of pro-caspase-1, which collectively comprises the inflammasome (65). Activated caspase-1 cleaves pro-IL-1β into active IL-1β, which is secreted. Several DAMPs that are present after incision and that can serve as the activation signal for NLRP3 inflammasome assembly include: ATP (66, 67), reactive oxygen species (ROS) (68, 69), and low pH (70–72). The activation cascade for NLRP3 is summarized in **Figure 1**. Indeed, the presence of priming and activating DAMPs for NLRP3 activation after aseptic tissue injury implicate a role for NLRP3 in mediating the postoperative pain phenotype. We recently showed that NLRP3 is upregulated at the surgical site and drives postoperative mechanical pain-like behaviors in male mice, but not in female mice (13). This study provided the first evidence that NLRP3 drives postoperative pain and revealed that the immune-mediated mechanisms that underlie postoperative pain are sex-specific.

# SEXUAL DIMORPHISM IN THE IMMUNE RESPONSE

The importance of taking sex into consideration when studying painful injuries and their underlying mechanisms was recently highlighted when it was revealed by Sorge et al. that male mice require microglia and TLR4, whereas female mice require T cells to mediate chronic neuropathic pain (73, 74). In addition, hormones significantly contribute to sex-based differences in the immune response (75). Estradiol, progesterone, and testosterone are the primary hormones that affect the immune response. Female vertebrates have higher baseline estrogen and progesterone levels whereas male vertebrates have higher baseline testosterone levels. Estrogen, progesterone, and testosterone receptors are expressed on both adaptive (T cells and B cells) and innate (macrophages, dendritic cells, neutrophils, and natural killer cells) immune cells; the effects of hormones on these receptors are dose-dependent (76). Consequently, there are alterations in immune system function during pregnancy, menses, and menopause. Each of the three hormones mentioned above affects the immune system during injury or disease states in different ways, and therefore, the immune response to injury differs between males and females.

The level of immune cell infiltration and the extent of the innate immune response at an injury site are both affected by sex hormones. Estrogen suppresses mast cell release of histamine and as a result, fewer neutrophils are recruited to a wound site in females (77, 78). In regards to the effects of estrogen on the macrophage response to injury, Price et al. recently showed that a reduced number of macrophages is recruited to a postoperative tissue site in female mice compared to male mice (79). Additionally, high estrogen levels skew macrophages toward the M2 phenotype (anti-inflammatory) while high testosterone levels promote the M1 phenotype (proinflammatory). As a consequence of the M1 phenotype, males have higher expression of TLR4, NLRP3, and produce more IL-1β than females (75, 80–83). However, chronic estrogen exposure induces increased TLR4-mediated production of IL-1β in macrophages (84). Despite the lower levels of immune cell infiltrate within a wound in females as compared to males, cytokine levels in females are sustained longer than in males, and females have more tissue-resident immune cells than males (75, 78). Furthermore, data from our laboratory demonstrated that males have more IL-1β protein at the peri-incisional site than females (13). Whereas, we showed that NLRP3 mRNA was upregulated by incision to a similar extent in males and females, global deletion of NLRP3 decreased IL-1β levels and sensitization to mechanical stimuli only in males. This suggested that NLRP3 may be differentially regulated post-transcriptionally in males and females following tissue incision, where in females, the IL-1β production occurs independent of NLRP3. The activation pathway for NLRP3 has been suggested to differ in macrophages from male and female Systemic Lupus Erythematosus patients as well (85). In addition, males and females utilize TLR4 in a cell-specific manner. Stimulation of TLR4 on macrophages drives pain in male mice whereas stimulation of TLR4 on sensory neurons drives pain in female mice (86, 87). Furthermore, fibroblasts which play critical roles in the immune response and local environment during tissue injury, also produce IL-1β, and fibroblast IL-1β levels are differentially affected by testosterone and estrogen treatment (88). Considering all of the evidence above forsex-driven differences in the immune response to injury and the resulting differences in sensory neurons, it is imperative to take the sex of an individual into account when selecting and assessing the efficacy of pain interventions.

# PROINFLAMMATORY IL-1β AND POSTOPERATIVE PAIN

The primary function of IL-1β is to elicit a pro-inflammatory response to DAMPs (41). IL-1β is expressed by macrophages, monocytes, neutrophils, mast cells, glial cells, and sensory neurons (89–91). Secreted IL-1β exerts its proinflammatory effects through various mechanisms. These include increasing production of other inflammatory mediators via rapidly inducing their mRNA expression, increasing vascular permeability, recruiting immune cells, directly eliciting pain via binding of the IL-1β receptor on sensory neurons, and inducing neurogenic inflammation through sensory neuron sensitization and increased production of calcitonin gene-related peptide alpha (CGRPα) (91–94). IL-1β acts through its receptor type I IL-1 receptor (IL-1R1), which is ubiquitously expressed on neurons of the peripheral and central nervous systems (95, 96). When IL-1R1 binds IL-1β, the accessory protein IL-1R3 is recruited to induce intracellular signaling cascades via association of their intracellular Toll- and IL-1R-like (TIR) domains with signaling proteins (97). The cascade begins with the association of myeloid differentiation primary response gene 88 (MYD88) and interleukin-1 receptor–activated protein kinase (IRAK) 4 with the TIR domains. This leads to complex formation of IRAK1, IRAK2, and tumor necrosis factor–associated factor (TRAF) six and subsequent activation of transcription factors such as NF-κB to upregulate inflammatory genes.

Postoperative pain is characterized by persistent acute pain at the incisional site which is associated with release of proinflammatory cytokines, including IL-1β. Studies have found that IL-1β is significantly upregulated at the incision site (12, 13, 32, 98–101). Wolf et al. demonstrated that either systemic inhibition of IL-1β signaling by its receptor antagonist IL-1ra or deletion of IL-1R1 prevented the development and maintenance of postoperative mechanical hypersensitivity at the incision site (102). Other groups further demonstrated that inhibition of IL-1β signaling through antagonism of its receptor significantly decreased postoperative pain-like behavior in rodents (32, 98, 100). Furthermore, additional research has established that inhibition of the upstream mediators of IL-1β, such as TLR4 (98, 103), NF-κB (103), caspase-1 (104), or NLRP3 (13), decreases postoperative pain-like behaviors in rodents. General blockade of IL-1β signaling, like that obtained with FDA approved Anakinra (IL-1R1 antagonist), increases the rate of infections due to the necessity of IL-1β for bacterial infection clearance (105, 106). Whereas, inhibition of TLR4, NF-κB, and caspase-1 is more ubiquitous, inhibition of one inflammasome is more specific. Therefore, reduction of IL-1β but not complete depletion, through inhibition of only NLRP3 may avoid these complications while decreasing postoperative pain. However, inhibition of NLRP3 alone may only be effective in males but not females (13). Not only are sex differences prevalent in mice that received surgery, but they are present in human postoperative pain as well. For instance, a predicative factor of chronic postoperative pain is female sex (107–109). Thus, a therapeutic for the treatment of postoperative pain in females must target the unique factors that are required for the development of postoperative pain in females.

#### THE ROLE OF NLRP3 IN PAIN DISORDERS

Much is yet to be learned about NLRP3 and postoperative pain, however a role for NLRP3 in pain disorders is emerging (13). NLRP3 has been shown to be involved in the pathogenesis of both inflammatory and neuropathic pain conditions. Inflammatory pain depends on the sensitization of nociceptive neurons by proinflammatory mediators such as IL-1β (110, 111). In an acute model of dural inflammation, the injection of an "inflammatory soup" (comprised of histamine, serotonin, bradykinin, and prostaglandin E2 at pH 5.5) resulted in activated NLRP3 and caspase-1, and increased IL-1β expression in C fiber type neurons of the trigeminal ganglia (29). The inflammatory soup injection also resulted in pain-like behaviors which were alleviated by a caspase-1 inhibitor. In another inflammatory pain model, the complete Freund's adjuvant (CFA) model, NLRP3 was shown to be activated in the skin of rats (28). Electroacupuncture following CFA injection attenuated the expression of NLRP3 and ultimately eliminated the pain-like behavior (28). Additionally, NLRP3 has been demonstrated to be crucial for the pathogenesis of rheumatoid arthritis in both humans and rodents (27, 112, 113). Further, upregulation of NLRP3 has been shown to occur in rodent models of gout, and its inhibition or deletion ameliorated the pathology and pain (24, 25, 114– 116). Collectively, these data point to a key role for NLRP3 in inflammatory pain.

Neuropathic pain involves direct damage to nerves from injury or disease. IL-1β significantly contributes to traumatic neuropathic pain where its expression is upregulated in the dorsal root ganglia and spinal cord, as well as in damaged nerves in rodent models of neuropathic pain and in patients with neuropathic pain (18, 117–119). NLRP3 plays a role in various rodent models of neuropathic pain. Alleviation of sciatic nerve ligation neuropathic pain with miR-23a overexpression, or CXCR4 knockdown results in decreased NLRP3 expression (16). A study utilizing the chronic constriction sciatic nerve injury model of neuropathic pain demonstrated that NLRP3 is upregulated by nerve injury and that treatment with Peptide5, a Connexin 43 mimetic peptide that blocks hemichannels, decreased NLRP3 expression and mechanical pain-like behavior (17). In addition, chemotherapy-induced neuropathy models of neuropathic pain revealed that NLRP3 is upregulated in both oxaliplatin-induced nerve injury (19) and paclitaxelinduced nerve injury (20) models, and inhibition of NLRP3 decreased the mechanical pain-like behaviors in both models. In contrast to these findings, it was demonstrated that global knockout of NLRP3 had no effect on neuropathic pain in the spared nerve injury model of neuropathic pain (120). This is consistent with discrepant findings that challenge the view that microglia drive neuropathic pain exclusively in males (121–123). When compared, these studies demonstrate that different models of neuropathic pain (spared nerve injury, spinal nerve transection, spinal nerve ligation, and partial nerve ligation) do not produce the same findings. Together, these studies suggest that while NLRP3 contributes to a variety of etiologies of neuropathic pain it is dependent on the type of injury and the diverse factors that are likely involved in different injuries.

Although much remains to be discovered about the mechanistic causes of body-wide pain syndromes such as fibromyalgia, several studies have indicated a role for NLRP3 in fibromyalgia-associated pain, and NLRP3 was found to be upregulated in patients with fibromyalgia (21–23). Further research is needed in animal models of fibromyalgia and tissues from patients with fibromyalgia.

#### CONCLUSION

The discovery of inflammasomes has provided new insights into the molecular mechanisms underlying the innate immune system activation in inflammatory and neuropathic pain conditions. As discussed here, many inflammatory and neuropathic pain conditions, and specifically postoperative pain, involve the innate immune system and NLRP3. Therefore, modulators of NLRP3 may provide a novel, selective, and effective pain therapeutic target. Notwithstanding, our understanding of the functional roles and the mechanisms of activation of the NLRP3 inflammasome in pain conditions is in its infancy. Additionally, it is imperative that further research be conducted on the effect of the sex of an individual on NLRP3 function since all of the rodent studies on NLRP3 in the pain conditions discussed here, except for the report by our group (13), were performed in males only. In our study we revealed that there are significant sex differences when NLRP3 is deleted, suggesting that NLRP3 plays different roles in males and females following tissue injury (13). Additionally, the literature concerning the specific immune responses to perioperative incision in males and females is insufficient and far more studies that include females must be done. Therefore, we conclude that targeting NLRP3 may provide a novel approach to control pain, but that further research needs to uncover the mechanistic differences and roles of NLRP3 in wound healing following surgery in females and males.

#### DATA AVAILABILITY

No datasets were generated or analyzed for this study.

#### AUTHOR CONTRIBUTIONS

AMC wrote and edited the manuscript. CLS and BND edited the manuscript.

#### FUNDING

This work was supported by the National Institute of Neurological Disorders and Stroke grants NS040538 and NS070711 to CLS and F31GM123778 to AMC. The Research and Education Component of the Advancing a Healthier Wisconsin Endowment

#### REFERENCES


at the Medical College of Wisconsin provided partial support.

#### ACKNOWLEDGMENTS

The authors thank Neil Smith, freelance illustrator, for creation of **Figure 1**.


in monosodium urate crystal-induced inflammation. Arthritis Rheumatol. (2010) 62:2170–9. doi: 10.1002/art.27456


injury model. PLoS ONE. (2015) 10:e0133707. doi: 10.1371/journal.pone.01 33707


**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 Cowie, Dittel and Stucky. 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.

# Melatonin Suppresses Microglial Necroptosis by Regulating Deubiquitinating Enzyme A20 After Intracerebral Hemorrhage

Jianan Lu1†, Zeyu Sun1†, Yuanjian Fang1†, Jingwei Zheng<sup>1</sup> , Shenbin Xu<sup>1</sup> , Weilin Xu<sup>1</sup> , Ligen Shi <sup>1</sup> , Shuhao Mei <sup>1</sup> , Haijian Wu<sup>1</sup> , Feng Liang<sup>1</sup> \* and Jianmin Zhang1,2,3 \*

*<sup>1</sup> Department of Neurosurgery, The Second Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, China, <sup>2</sup> Brain Research Institute, Zhejiang University, Hangzhou, China, <sup>3</sup> Collaborative Innovation Center for Brain Science, Zhejiang University, Hangzhou, China*

#### Edited by:

*Antje Kroner, Medical College of Wisconsin, United States*

#### Reviewed by:

*Geert Van Loo, Flanders Institute for Biotechnology, Belgium Jorge Tolivia, Universidad de Oviedo, Spain*

#### \*Correspondence:

*Feng Liang frankfeng@zju.edu.cn Jianmin Zhang zjm135@zju.edu.cn*

*†These authors have contributed equally to this work*

#### Specialty section:

*This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Immunology*

Received: *01 March 2019* Accepted: *29 May 2019* Published: *14 June 2019*

#### Citation:

*Lu J, Sun Z, Fang Y, Zheng J, Xu S, Xu W, Shi L, Mei S, Wu H, Liang F and Zhang J (2019) Melatonin Suppresses Microglial Necroptosis by Regulating Deubiquitinating Enzyme A20 After Intracerebral Hemorrhage. Front. Immunol. 10:1360. doi: 10.3389/fimmu.2019.01360* Cell death is deeply involved in pathophysiology of brain injury after intracerebral hemorrhage (ICH). Necroptosis, one of the recently discovered forms of cell death, plays an important role in various diseases, including ICH. Previous studies have suggested that a considerable number of neurons undergoes necroptosis after ICH. However, necroptosis of microglia after ICH has not been reported to date. The present study demonstrated for the first time that necroptosis occurred in the microglia surrounding the hematoma after ICH in C57 mice, and melatonin, a hormone that is predominantly synthesized in and secreted from the pineal gland, exerted a neuroprotective effect by suppressing this process. When we further explored the potential underlying mechanism, we found that melatonin inhibits RIP3-mediated necroptosis by regulating the deubiquitinating enzyme A20 (also known as TNFAIP3) expression after ICH. In summary, we have demonstrated the role of microglial necroptosis in the pathogenesis of ICH. More importantly, A20 was identified as a novel target of melatonin, which opens perspectives for future research.

Keywords: intracerebral hemorrhage (ICH), necroptosis, microglia, melatonin, A20

#### INTRODUCTION

Intracerebral hemorrhage (ICH) is a significant cause of morbidity and mortality worldwide. As one of the most serious forms of stroke, ICH affects ∼2 million people worldwide each year (1– 3). Primary injuries after ICH are usually caused by the mechanical damage of the hematoma to the surrounding brain tissues. Following the primary injuries, secondary injuries, including inflammation and cell death, are extensively involved in the pathological processes following hemorrhagic events (4). Treatment of secondary injuries is one of the important interventions after ICH.

Cell death is a hallmark of secondary brain injury after ICH (5). Numerous experimental and clinical observations indicate a variety of cell death forms and mechanisms take place during hemorrhagic stroke (6). Among them, programmed cell death is highly correlated with the homeostatic mechanisms of the nervous system. Necroptosis, also known as programmed necrosis, shares upstream signaling elements with apoptosis, such as tumor necrosis factor (TNF), Fas-associated death domain (FADD), and Toll-like receptor (TLR) (7). However, apoptosis and necroptosis have distinct outcomes.

Apoptotic cells usually induce a non-inflammatory response, whereas necroptosis triggers an inflammatory response (8) due to the rapid loss of plasma membrane integrity prior to the exposure of phagocytic signal (9), which in turn causes the release of intracellular damage-associated molecular patterns (10), which induce inflammation. Microglia, which are primary immune cells of the central nervous system (CNS), undergo necroptosis in various pathological processes, such as ischemic stroke (11), retinal degeneration (12), and spinal cord injury (12). Therefore, it may be beneficial to inhibit necroptosis to reduce neuroinflammation and improve neuronal survival in the context of disease, particularly in microglia (13).

Melatonin (N-acetyl-5-methoxytryptamine) is a hormone that is predominantly synthesized in and secreted from the pineal gland. (14) In 1958, Lerner et al. were the first to isolate melatonin from bovine pineal gland extracts and named it according to its ability to aggregate melanin granules (15). Initial research on melatonin was focused on its regulation of circadian and circannual cycles (16). Later, the multifunctional roles of melatonin were further explored. Melatonin serves as an effective antioxidant in scavenging the highly toxic hydroxyl radical and other oxygen-centered radicals (17), and it promotes the immune response (18). No serious side effects or risks have been reported in association with the ingestion of melatonin (17). These findings suggest that melatonin might play a protective role in various pathological conditions. Therapeutic effects of melatonin in various diseases, including diabetes (19), cancers (20), cardiovascular diseases (21, 22), and CNS diseases (23) have been reported.

In CNS diseases, melatonin has been shown to contribute to the maintenance of cell homeostasis and survival by regulating inflammation, apoptosis, or autophagy after different types of brain injury (24). Melatonin reportedly has a neuroprotective effect after stroke, either ischemic (25, 26) or hemorrhagic (27). Melatonin has also been shown to inhibit necroptosis during myocardial ischemia and liver fibrosis (28–30), and this mechanism might explain its therapeutic effects after stroke. However, whether melatonin can inhibit necroptosis of microglia after ICH and how much protection can be achieved by inhibiting this process remain to be elucidated.

Serine/threonine kinase receptor interacting protein 1 (RIP1) and receptor-interacting protein 3 (RIP3) are important molecules in the process of necroptosis (7). A previous study revealed that RIP1 expression was significantly increased 1– 3 days after ICH (31), suggesting that necroptosis occurs in the acute phase after ICH. Upon inhibition of caspase activity, especially that of caspase 8 by genetic or chemical methods, RIP1 forms a necrosome with RIP3 by interacting via their homotypic interaction motif domains and activates their kinase activities (32–34). Subsequently, the pseudokinase mixed lineage kinase domain-like protein (MLKL), is activated as an executive molecule of necroptosis (35). In turn, MLKL translocates to the plasma membrane, where it forms pores and disrupts the plasma membrane integrity (36). Although the necrosome, the characteristic complex of necroptosis, is composed of both RIP1 and RIP3, RIP3 is considered to have a more critical role than RIP1 in the process of necroptosis (37–39). Previous studies have suggested that melatonin inhibits necroptosis mainly by inhibiting the kinase activity of RIP3 (28, 29). However, the specific mechanism of RIP3 regulation by melatonin has not been elucidated.

A20, also known as TNFAIP3, is a deubiquitinating enzyme (40). A20 is widely recognized as a potent anti-inflammatory protein linked to multiple human brain autoimmune diseases, neuro-degenerative diseases, and brain tumors, as well as stroke (41–44). Voet et al. reported that A20 controls microglial activation to regulate neuroinflammation, a function closely related to inhibition of NLR family pyrin domain-containing protein 3 (NLRP3) (45). Moreover, Onizawa et al. discovered that A20 restricts RIP3-dependent necroptosis (46). Thus, we hypothesized that melatonin can upregulate the expression of A20 in the brain, especially in microglia, after ICH, which in turn inhibits the expression of RIP3, thereby exerting a protective effect by inhibiting necroptosis.

In this study, we aimed to demonstrate the occurrence of microglial necroptosis after ICH and to evaluate the effect of melatonin on this process as well as the underlying mechanism, to improve the recovery of neurological function after ICH.

#### MATERIALS AND METHODS

#### Animals

C57 mice were purchased from SLAC Laboratory Animal Company Limited (Shanghai, China). In total, 214 male mice (8–10 weeks, 20–25 g) were used in this study. The mice were housed in a temperature- and humidity-controlled room under a standard 12-h light/dark cycle and had free access to food and water. The animal protocol was approved by the Institutional Ethics Committee of the Second Affiliated Hospital, Zhejiang University School of Medicine. The procedures were conducted according to the National Institutes of Health's Guide for the Care and the Use of Laboratory Animals and the ARRIVE (Animal Research: Reporting in vivo Experiments) guidelines. Experimental grouping was shown in **Supplementary Figure S1**.

#### ICH Model

The ICH model was established as previously described (47) (**Figure 1A**). Briefly, mice were anesthetized with 40 mg/kg 1% pentobarbital sodium via intraperitoneal injection. Under stereotactic guidance, a small cranial burr hole was made at a precise location (bregma coordinates: 0.5 mm anterior and 2.5 mm lateral to the midline). Autologous blood (30 µL) from the femoral artery was injected 3.5 mm deep into the right basal ganglia at a rate of 3 µL/min using a microinfusion pump, and the syringe was pulled out after 10 min.

#### Drug Administration

As described previously (48, 49), melatonin (Sigma, USA) was dissolved in dimethyl sulfoxide (DMSO) and diluted with 0.9% normal saline. A dose of melatonin (20 mg/kg) or vehicle (5% DMSO) was given to mice randomly via intraperitoneal injection 30 min before ICH induction.

ICH+melatonin groups before ICH and at 1, 3, 7 days after ICH. \**P* < 0.05.

# Neurobehavioral Function Assessment

Four evaluation methods were used to assess the voluntary activities and motor function of mice at 24 h, 72 h, and 7 days after ICH impairment.

Neurologic function was tested before and 1, 3, 7 days after ICH by assessing body symmetry, gait, climbing, circling behavior, front limb symmetry, and compulsory circling (50). Each test result was graded from 0 to 4, with a maximum deficit score of 24.

An adhesive removal test was conducted as previously described (51). Briefly, mice were accustomed to the experimental environment for 30 min. Then, an adhesive tape strip was placed on the left hairless part of the forepaws of the mice. Mice were then put into the testing cage and the time to feel and time to remove the strip by any behavior of the mice were recorded.

For the foot-fault test, mice were individually placed on a wired grid (50 × 55 × 52 cm length/width/height) with their paws. Behavior of the mice while they were moving was recorded for 1 min. Each successful foot placement onto the bar was recorded as a step. A foot fault was recorded when a paw slipped through the grid hole. The percentage of foot faults was calculated as: 100 × faults/(successful steps + faults) (52).

A rotarod test was performed as previously described (53). Mice were placed on a Rotamex 5 apparatus (Columbus Instruments). During the test, the speed was increased from 4 rpm to 40 rpm within 5 min. After adaptation for two consecutive days, mice were tested twice daily with at least 30 min between tests. The period to fall off the rotating rod was recorded, and the data were expressed as the mean from three trials.

#### Brain Water Content

We used a wet–dry method as previously described (54) to evaluate the brain water content at 72 h after ICH. Briefly, after euthanasia, mice (n = 6/group) were sacrificed and the brain hemispheres were collected and weighed immediately (wet weight). They were then dried at 100 ◦C for 48 h and weighed again to obtain the dry weight. The brain water content was calculated as follows: [(wet weight—dry weight)/(wet weight)] × 100% (55).

# Propidium Iodide (PI) Staining in vivo

PI staining in vivo to identify microglial necroptosis was conducted as previously described (31). For all experiments, PI (Beyotime, Shanghai, China) was administered intraperitoneally (10 mg/kg) 1 h before the mice were sacrificed. The brain tissue was cryoprotected by immersion in 15 and 30% sucrose solution. Then, brain sections (10µm) were cut along the anterior–posterior lesion and placed on poly-L-lysine-coated glass slides. Sections were visualized under a fluorescence microscope (OLYMPUS BX50/BX-FLA/DP70; Olympus Co.). PI+/Iba-1+ cells were counted by observers blinded to the experimental groups. Necroptotic microglia were counted in six microscopic fields per section, and the average number in each section was calculated.

#### Immunofluorescence

After the mice were anesthetized, transcardial perfusion was performed with 0.1 M PBS, followed by perfusion with 4% paraformaldehyde. The whole brain was immersed in 4% paraformaldehyde for 24 h, then cryoprotected in serial 15 and 30% sucrose solutions. The brain samples were then cut into coronal slices (9µm) and fixed on slides. After preprocessing with 5% BSA and 0.3% triton X-100, the sections were incubated at 4◦C overnight with primary antibodies, including anti-Iba-1 antibody (1:500, Abcam, Cambridge, UK, ab5079), anti-RIP1 antibody (1:200, Cell Signaling Technology, Danvers, MA, USA, CST#D94C12), anti-RIP3 antibody (1:200, Cell Signaling Technology, CST#D4G2A), anti-MLKL antibody (1:100, Santa Cruz Biotechnology, Santa Cruz, CA, USA sc-293201), anti-TNFAIP3 (A20) antibody (1:250, Abcam, ab13597). Then, the cryosections were incubated with secondary antibody [1:300, Thermo Fisher Scientific, Donkey anti-Goat IgG (H+L) Alexa Fluor 488, A32814; Donkey anti-Mouse IgG (H+L) Alexa Fluor 594, A32744; Goat anti-Rabbit IgG (H+L) Alexa Fluor 594, A-11037] at 37◦C for 1 h and washed three times with PBST. Finally, the sections were observed and analyzed using a fluorescence microscope (Olympus Co., Tokyo, Japan).

#### Western Blot Analysis

Brain samples around the hematoma were collected and lysed in RIPA lysis buffer (Beyotime, Shanghai, China). Western blotting was performed as previously described (56). Protein samples (40 µg/lane) were separated by 10% sodium dodecyl sulfate polyacrylamide gel and electrotransferred to polyvinylidene fluoride membranes (Millipore, Burlington, MA, USA). The membranes were blocked at room temperature for 1 h and then incubated overnight with primary antibodies, including: anti-RIP1 antibody (1:1000, Cell Signaling Technology, CST#D94C12), anti-RIP3 antibody (1:1000, Cell Signaling Technology, CST#D4G2A), anti-MLKL antibody (1:200, Santa Cruz Biotechnology, sc-293201), anti-TNFAIP3 (A20) antibody (1:1000, Abcam, ab13597), anti-RIP3 (phospho S232) antibody (1:1000, Abcam, ab195117), anti-MLKL (phosphor S345) antibody (1:1000, Abcam, ab196436), anti-NLRP3 (1:1000, Abcam, ab98151).

#### Small Interfering (si)RNA and Intracerebroventricular Injection

A20 siRNA or scramble siRNA (Genomeditech, Shanghai, China) mixed with transfection reagent (Engreen Biosystem, Auckland, New Zealand) was delivered via intracerebroventricular injection as previously described (57). Briefly, the injections into the right ventricle were performed using the following coordinates relative to bregma (0.2 mm posterior, 1.0 mm lateral, and 2.0 mm deep) at 48 h prior to surgery. After the injection was completed, the needle was left in the brain for 10 min, and the burr hole was blocked with bone wax.

# Single Cell Sorting

Single cell sorting was performed as described previously (58). In brief, brain samples were fixed by transcardial perfusion with PBS before extraction. After dissociation, cells were separated by 30% Percoll density gradient separation by centrifugation at 800 ×g for 30 min at 18◦C. Then, the cells were passed through a 70-µm nylon mesh. Cell populations were sorted on an Aria SORP instrument (BD Biosciences, San Jose, CA, USA). The gating strategy is shown in **Supplementary Figure S2**. CD45+ (PerCP CD45, 1:200, BD Biosciences, 557235) and CD11b+ (FITC CD11b, 1:200, BD Biosciences, 557396) cells were isolated as microglia. Immediately after sorting, the cells were stored at −80◦C until processed.

#### Cell Lines and Coculture

The mouse microglial cell line BV2 and the mouse hippocampal cell line HT22 were cultured (37◦C, 5% CO2) in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, 100 U/ml penicillin and 100µg/ml streptomycin. As secondary damage in ICH is mainly caused by oxidized hemoglobin, we used 100µM OxyHb to simulate the pathological process of ICH in vitro, as previously described (31). Melatonin was used at 1 mM to study the effect of melatonin on necroptosis in vitro. A Transwell coculture system was used to investigate the effect of BV2 cells on HT22 cells. BV2 cells were cultured at a density of 5 × 10<sup>4</sup> in Transwell inserts (pore size 0.4µm; Corning, Corning, NY, USA) placed above the HT22 neuronal layer.

FIGURE 2 | (*n* = 6/group). (D) MLKL expression. \**P* < 0.05 (*n* = 6 in each group), &*P* < 0.05 vs. ICH 1-day group (*n* = 6/group). (E,K) PI staining around ICH hematoma. A significant increase in PI+ microglia was observed in the ICH group. \**P* < 0.05 vs. sham group (*n* = 6/group). Melatonin treatment significantly decreased PI+ microglia compared with the levels in the ICH+vehicle group. &*P* < 0.05 vs. ICH+vehicle group (*n* = 6/group). (F) Expression of A20, RIP1, RIP3, and MLKL proteins in sorted microglia of each group. (G) A20 expression. (H) RIP1 expression. (I) RIP3 expression. (J) MLKL expression. \**P* < 0.05 vs. sham group (*n* = 6/group), &*P* < 0.05 vs. ICH 3-day group (*n* = 6/group).

# Cell Viability Assay and Cytotoxicity Assay

As necroptosis is a type of cell death, cell viability and cytotoxicity can reflect its occurrence in vitro. Cells were cultured in 96-well plates. The CCK-8 cell counting kit (Beyotime, Shanghai, China) was used to evaluate cell viability according to the manufacturer's instructions. Twenty microliters of CCK-8 solution was added to 200 µl of cell culture medium, after which the plates were incubated at 37◦C for 2 h. Then, the absorbance at 450 nm was measured. Cytotoxicity was evaluated using an LDH lactate dehydrogenase cytotoxicity test kit (Beyotime). Briefly, cells were cultured in serum-free medium for 24 h. Then, the plates were centrifuged at 400 ×g for 5 min. One hundred twenty microliters of supernatant of each well was transferred to a new 96-well plate, and the absorbance at 490 nm was measured immediately.

#### ELISA

To explore changes in inflammatory factors, a TNF ELISA KIT (Abcam, ab100747) was used according to the manufacturer's instructions to quantify the levels of TNF in BV2 cell supernatant.

#### Reactive Oxygen Species (ROS) Assay

ROS can be both an inducer and a product of necroptosis. ROS levels in cells were examined using a ROS Assay Kit (JianCheng, Nanjing, China) according to the manufacturer's instructions. In brief, cells were trypsinized, collected, and incubated with 10µM 2 ′ ,7′ -dichlorodihydrofluorescein diacetate at room temperature for 30 min. After two washes, intracellular ROS production was measured by fluorescence detection using a microplate reader at an excitation wavelength of 485 nm and an emission wavelength of 535 nm. Protein levels in the cells were measured using a detergent-compatible protein assay kit (Bio-Rad, Hercules, CA, USA). ROS levels are reported as fluorescence/mg protein.

#### Annexin V and PI Staining in vitro

Cells were cultured in six-well plates and subjected to melatonin treatment. Cells were trypsinized with 0.25% trypsin (without EDTA) and centrifuged at 1000 ×g for 5 min and resuspended in 300 µl of binding buffer. Subsequently, 1 µl of Annexin V and 1 µl of PI (Becton Dickinson, Franklin Lanes, NJ, USA) were added to the cell suspension. After 30 min of incubation at 37◦C in the dark, the cells were analyzed by flow cytometry (FACSCalibur; BD Biosciences, San Diego, CA, USA). Cells were first gated based on forward and side scatter, and necroptotic cells were determined as FITC+/PI+.

# Measurement of Mitochondrial Membrane Potential (1ψm)

BV2 cells were cultured in a six-well plate. The 1ψm was measured using a JC-1 kit (Beyotime, Shanghai, China) following the manufacturer's instructions. After rinsing with PBS, cells were incubated with JC-1 staining solution at 37◦C for 20 min, then washed with JC-1 staining buffer. Then, the cells were imaged using a fluorescence microscope (Olympus, Tokyo, Japan). The fluorescence intensity was measured using a fluorometric microplate reader (FilterMax F5, Molecular Devices, Sunnyvale, USA) at dual wavelengths: excitation and emission at 485 and 530 nm (to detect JC-1 monomer) and at 530 and 590 nm (to detect JC-1 polymer).

# Transmission Electron Microscopy

Mice were sacrificed and perfused with 0.9% saline and 4% PFA. Fragments of peri-hematoma tissues (∼1 mm<sup>3</sup> ) were collected and immersed in glutaraldehyde (2.5%) at 4◦C overnight. The tissues were preprocessed as previously described (59, 60). The samples were sliced into 100-nm sections and stained with 4% uranyl acetate and 0.5% lead citrate. The ultrastructure of the tissues was evaluated by transmission electron microscopy (Philips Tecnai 10, Netherlands).

# Statistical Analysis

All data are expressed as the mean ± standard deviation (SD) and were analyzed in GraphPad Prism v. 6.0. Data were tested for normality of distribution by the Kolmogorov-Smirnov test. Oneway ANOVA was applied for comparing multiple groups. The analyses were conducted in SPSS v. 22.0 (SPSS Inc.). Statistical significance was defined as P < 0.05.

# RESULTS

#### Melatonin Reduces Brain Edema, Improves Neurological and Motor Function After ICH

We evaluated the brain water content to explore the effects of melatonin treatment on ICH-induced brain edema. No significant differences were noted in the contralateral cortex, contralateral basal ganglia, or cerebellum between sham, ICH+vehicle, and ICH+melatonin groups. However, compared to the sham group, the brain water content was increased in the ipsilateral cortex and ipsilateral basal ganglia of mice in the ICH+vehicle group, whereas melatonin treatment significantly reduced the brain water content (**Figure 1B**). To explore the effect of melatonin on the neurological function of mice after ICH, we carried out various neurological tests before ICH and at 24 h, 72 h, and 7 days after ICH. Melatonin or vehicle (5% DMSO) was given to mice randomly via intraperitoneal injection 30 min before ICH induction. The neurological deficit scores were significantly increased after ICH and peaked at 72 h, whereas melatonin significantly improved neurological function (**Figure 1C**). The results of motor function assessment showed trends similar to those of neurological deficit scores (**Figures 1D–F**).

*(Continued)*

FIGURE 3 | corresponding sham group (*n* = 6/group), &*P* < 0.05 vs. corresponding ICH+vehicle group (*n* = 6/group). (D) RIP3 expression. \**P* < 0.05 vs. corresponding sham group (*n* = 6/group), &*P* < 0.05 vs. corresponding ICH+vehicle group. (E) MLKL expression. \**P* < 0.05 vs. corresponding sham group (*n* = 6/group), &*P* <0.05 vs. corresponding ICH+vehicle group. (F–I) Immunofluorescence staining showing the distribution of RIP1/RIP3/MLKL+ microglia surrounding hematoma at 3 days after ICH. Corresponding individually stained fluorescence images of higher magnification (40×) are shown in Supplementary Figure S3. \**P* < 0.05 vs. sham group (*n* = 6/group), &*P* < 0.05 vs. ICH+vehicle group (*n* = 6/group).

# Microglia Undergo Necroptosis After ICH

To explore the involvement of necroptosis of microglia after ICH, we performed a series of experiments. The expression of RIP1, RIP3, MLKL, which are executive molecules in necroptosis, was significantly increased at 24 h and 72 h after ICH (P<0.05 vs. sham group; **Figures 2A–D**). Correspondingly, cells that were costained for these molecules and Iba-1 were also remarkably increased at 72 h after ICH (P < 0.05 vs. sham group; **Figures 3F–I**). To further illustrate the occurrence of microglial necroptosis after ICH, we isolated microglia, and we found that at 3 days after ICH, the expression of necroptosis-related proteins (RIP1, RIP3, MLKL) in microglia was significantly higher than that in sham group (P < 0.05 vs. sham group; **Figures 2F–I**). We also counted PI+ microglia in frozen brain sections at 72 h after ICH. PI+/Iba-1+ cells were remarkably increased in brain tissues surrounding hematomas after ICH (P < 0.05 vs. sham group; **Figures 2J,K**). These results demonstrated that necroptosis occurred in microglia surrounding the hematoma after ICH.

#### Melatonin Plays a Protective Role by Suppressing Necroptosis in Microglia

To explore the effects of melatonin on necroptosis after ICH, the protein levels of RIP1, RIP3 and MLKL were detected by western blot analysis at 24 h and 72 h after ICH. As shown in **Figures 3A–E**, the expression of these proteins was significantly increased in the ICH+vehicle group compared to the sham group (P < 0.05), whereas melatonin treatment significantly suppressed the increases (P < 0.05 vs. ICH+vehicle group). Immunofluorescence staining revealed that cells costained for these molecules and Iba-1 were decreased in the ICH+melatonin compared to the ICH+vehicle group (P < 0.05; **Figures 3F–I**). Further, melatonin significantly reduced the expression of necroptosis-related proteins (RIP1, RIP3, MLKL) at 3 days after ICH in microglia (P < 0.05 vs. ICH 3-day group; **Figures 2F–I**). These results indicated that melatonin suppresses necroptosis of microglia in brain tissues after ICH.

#### Melatonin Reduces Mitochondrial Damage in Microglia After ICH

Mitochondrial damage has been shown to be involved in necroptosis (8, 61). Therefore, we used TEM to detect mitochondrial damage at 72 h after ICH. In the brain tissues of the ICH group, mitochondria with obvious vacuolization and swelling were observed (**Figure 4A**). Next, we determined the 1ψm of BV2 cells in vitro by JC-1 staining. As shown in **Figure 4B**, the green fluorescence intensity was increased in OxyHb+vehicle group compared to the sham group, indicating a decline of 1ψm; however, pre-incubation with melatonin attenuated the OxyHb-induced collapse of the 1ψm (P < 0.05). A microplate fluorescence read-out yielded similar results (**Figure 4C**). These results suggested that melatonin reduces mitochondrial damage after ICH by suppressing necroptosis.

#### Melatonin Prevents Cytotoxic and ROS Production-Enhancing Effects of OxyHb in vitro

**Figure 5A** shows representative photographs of OxyHb-treated and control BV2 cells. Swollen cells with ruptured plasma membrane can be observed in the OxyHb-treated group. We examined the effects of melatonin on necroptosis in vitro by cell viability and cytotoxicity assays. As shown in **Figures 5B,C**, cell viability was decreased in the OxyHb+vehicle group as compared to the control group (P < 0.05), whereas pretreatment with melatonin significantly improved cell viability (P < 0.05). The cytotoxicity assays suggested that OxyHb had a cytotoxic effect on BV2 cells (P < 0.05), which was prevented by melatonin pretreatment (P < 0.05). ROS were measured to investigate the effects of melatonin on ROS production in BV2 cells. As shown in **Figure 5D**, treatment with OxyHb remarkably promoted ROS production, which was suppressed by pretreatment with melatonin (P < 0.05). The above results indicated that melatonin improves BV2 cell viability and reduces cytotoxicity and ROS generation in vitro.

#### Melatonin Suppresses TNF Secretion by Microglia, Thus Reducing Neuronal Necroptosis

As the main immune cells in the nervous system, changes in microglia would affect the homeostasis of other cells, such as neurons. We firstly explored whether the inflammatory factor TNF, a common inducer of necroptosis, is involved in this process. The concentration of TNF was significantly higher in the culture medium of BV2 cells treated with OxyHb than that in that of the control group, whereas melatonin pretreatment significantly reduced TNF production (**Figure 5E**, P < 0.05). Next, we explored the effects of necroptotic microglia on neurons after ICH by using a Transwell system. HT-22 neurons were cocultured with BV2 cells exposed to various treatments. The neurons were then stained with Annexin V and PI and analyzed by flow cytometry. The necroptotic population (defined as Annexin V–/PI+) was increased in neurons cocultured with BV2 cells treated with OxyHb when compared with the control group, whereas pretreatment with melatonin suppressed this increase (**Figures 5F–H**; P < 0.05). These results indicated that melatonin suppresses TNF secretion by reducing microglial necroptosis, which in turn reduces neuronal necroptosis.

FIGURE 5 | BV2 cells in the OxyHb group showed cell swelling and plasma membrane rupture. (B) Results of cell viability assay. (C) Result of LDH assay. (D) Results of ROS production assay. (E) Results of TNF production assay in BV2 cell supernatant. OxyHb increased TNF production in BV2 cells, whereas melatonin significantly suppressed this. (F–H) Flow-cytometric analysis of HT22 cells cocultured with BV2 cells exposed to different treatments. HT22 cells cocultured with BV2 cells treated with OxyHb showed a higher necroptotic population (defined as PI+/FITC−), whereas melatonin significantly reduced the necroptotic population. \**P* < 0.05 vs. control group (*n* = 6/group), #*P* < 0.05 vs. OxyHb+vehicle group (*n* = 6/group), &*P* < 0.05 vs. OxyHb+melatonin+con-siRNA group.

#### A20 Is Expressed in the Acute Phase in Microglia After ICH, and Melatonin Promotes Its Expression and Thus Plays a Protective Role

We analyzed the expression of A20 at 72 h after ICH; A20 expression was significantly increased after ICH (**Figures 6A,B**; P < 0.05 vs. sham group). Treatment with melatonin significantly enhanced the expression of A20 after ICH (**Figures 6A,B**; P < 0.05 vs. ICH+vehicle group). Accordingly, immunofluorescence analysis revealed that A20 was expressed in the microglia surrounding the hematoma after ICH, and A20 expression was significantly increased after treatment with melatonin (**Figures 6C,D**, P < 0.05). To verify the above results, we measured A20 expression in microglia at 3 days after ICH. The results showed that A20 expression was significantly enhanced when compared with the sham group (**Figures 2E,F**; P < 0.05 vs. sham group). In line with the above results, melatonin treatment remarkably increased A20 expression after ICH (**Figures 2E,F**; P < 0.05 vs. ICH 3-day group). We investigated the relationship between A20 and RIP3-dependent necroptosis after ICH further by using A20 siRNA. As shown in **Figure 6E**, A20 siRNA transfection blocked the activating effect of melatonin on A20 expression (P < 0.05 vs. ICH+melatonin+con-siRNA group). Moreover, A20 siRNA blocked the effects of melatonin on RIP1, RIP3, MLKL expression (P < 0.05 vs. ICH+melatonin+consiRNA group). Melatonin reduced the phosphorylation of RIP3 and MLKL, which was prevented by A20 siRNA transfection (**Figures 6E–L**; P < 0.05). In vitro, A20 siRNA abolished the effects of melatonin on viability and ROS production in BV2 cells (**Figures 5B–D**). As A20 is generally considered to be a negative regulator of NLRP3, we also detected NLRP3 expression. The results showed that A20 siNRA lowered the promotive effect of melatonin on NLRP3 expression (**Figures 6E–L**; P < 0.05). These results indicated that melatonin inhibits RIP3 mediated necrotic apoptosis following ICH by regulating A20 expression.

# DISCUSSION

In this study, we demonstrated that melatonin inhibits necroptosis of microglia after ICH, thereby reducing brain edema and improving neurological function in the acute phase of ICH. In addition, we uncovered the potential underlying mechanism, in which melatonin regulates A20, thus affecting RIP3-dependent necroptosis (**Figure 7**).

Increasing evidence indicates that necroptosis plays a crucial role in the pathogenesis of several diseases with an inflammatory component (13). Shen et al. (31) found that necroptosis is an important mechanism of neuronal death after ICH, and treatment with necrostatin-1, a specific inhibitor of RIP1, could rescue the neurons from necroptosis, revealing a potential therapeutic target for ICH. Microglia are associated with necroptosis in different disease models (62–64); however, the exact underlying mechanism remained unclear. Moreover, there are no reports on microglial necroptosis after ICH. This study for the first time demonstrated that a significant amount of microglia around the hematoma did undergo necroptosis around 24 and 72 h after ICH. Although the relationship between necroptosis and inflammation has not been fully elucidated, necroptosis is generally thought to induce inflammation as, unlike apoptosis, it is essentially a type of necrotic cell death. In addition to inflammation, necroptosis causes some other cascade reactions, including mitochondrial damage, which exacerbates the release of ROS (32), which induce necroptosis. In line with previous experimental results (65), the present study suggested that mitochondrial damage and ROS accumulation after ICH might be caused by necroptosis. Therefore, necroptosis can play an intermediate role in the induction of inflammation by microglia after ICH. As resident macrophages in the CNS, microglia exert a variety of neuroprotective functions, such as phagocytosis and elimination, secretion of anti-inflammatory factors, growth factors, and promotion of the recovery of neurological function (66, 67). Thus, although microglia account for only approximately 10% of cells in the CNS, their death after ICH aggravates outcome. Therefore, we hypothesized that necroptosis plays an important role in the pathological processes after ICH.

A previous study revealed that A20 expression was increased after ICH and was associated with anti-inflammatory effects by inhibiting NF-κB activation (44). In addition to its NFκB-regulatory function, A20 is a ubiquitin-regulating enzyme, and increasing evidence shows that the central importance of its functions can be attributed to its ability to modulate ubiquitin-dependent signaling cascades (42). Ubiquitination is one of most prevalent post-translational protein modifications that regulate a variety of important physiological functions. During ubiquitination, ubiquitin, a highly conserved 76-aminoacid protein, is conjugated to lysine residues of diverse cellular proteins (66, 68). The E3 ubiquitin ligase activity of A20 has been shown to directly ubiquitinate RIP1, thus triggering proteasomal degradation of RIP1, which in turn inhibits TNF-induced NFκB activation (69). Therefore, in the present study, we mainly focused on the role of A20 as a deubiquitinating enzyme. As RIP1 is involved in apoptosis, it seems plausible that A20 could inhibit cellular apoptosis by promoting RIP1 degradation. However, a previous studies indicated that its anti-apoptotic effect is cell type- and stimulus-specific; thus, its protective effect is controversial (70). One study reported that A20 tends

to inhibit apoptosis via various mechanisms (41); however, other results indicated that in certain pathological conditions, such as IBD, A20 binds to RIP1 to enhance TNF-induced apoptosis (71). We did not further explore the effect of A20 on microglial apoptosis in this study. However, regardless of promotive or inhibitory effects, A20 is considered to directly

ICH+melatonin+con-siRNA group.

(I) p-RIP3 expression, (J) NLRP3 expression, (K) MLKL expression, and (L) p-MLKL expression. \**P* < 0.05 vs. sham group (*n* = 6/group), &*P* < 0.05 vs.

interact with RIP1 rather than via regulating NF-κB (71). As mentioned earlier, the necrosome is composed of both RIP1 and RIP3. Because RIP1 is involved in various biological processes, RIP3 is more valuable than RIP1 when studying necroptosis specifically. Moreover, RIP1 is not considered to be ubiquitinated in the necrosome, which suggests that A20 might regulate necroptosis by regulating RIP3. Furthermore, RIP3 has been regarded a specific activator of necroptosis because it regulates RIP1 phosphorylation (32) and switches TNF-induced cell death from apoptosis to necroptosis (33). The potential of A20 to suppress TNF-induced necroptosis by affecting RIP3 was observed in L929 cells and in primary mouse T cells (46, 72). In line herewith, in the present study, A20 regulated RIP3, thus inhibiting necroptosis.

Melatonin intervenes in different pathological processes through various mechanisms, e.g., it reduces inflammation (24, 73, 74), protects mitochondrial oxidoreductase and superoxide dismutase, thus improving ROS scavenging (75), and inhibits apoptosis (24, 76). In myocardial ischemia, melatonin inhibits necroptosis in cardiomyocytes by inhibiting RIP3 (29). In our study, melatonin inhibited necroptosis of microglia after ICH. We further explored the possible underlying mechanism. A20 is a regulatory protein of the NF-κB signaling pathway, with effects quite similar to those observed after treatment with melatonin. Accordingly, we found that melatonin inhibits RIP3-mediated necroptosis by promoting A20 expression. Interestingly, the NLRP3-regulatory effect of melatonin could be abolished by A20 siRNA transfection in vitro. This opens a new perspective for future exploration of NLRP3-mediated inflammation.

This study had some limitations that cannot be ignored. Firstly, melatonin might exert its neuroprotective effects after ICH by regulating various functions; the present study only explored its effect on necroptosis. Secondly, we only used cell lines for investigating the effect of melatonin on necroptosis in vitro; considering some subtle differences between cell lines and primary cells, out data would be more convincing had we used primary cells.

In conclusion, this study revealed that melatonin exerts neuroprotective effects after ICH. We demonstrated that melatonin inhibits microglial necroptosis via the A20/RIP3 pathway, in turn reducing cell death, inflammation, and mitochondrial damage. These findings implicate A20 as a novel potential therapeutic target for ICH.

#### DATA AVAILABILITY

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

# ETHICS STATEMENT

The animal protocol was approved by the Institutional Ethics Committee of the Second Affiliated Hospital, Zhejiang University School of Medicine. The procedures were conducted according to the National Institutes of Health's Guide for the Care and the Use of Laboratory Animals and the ARRIVE (Animal Research: Reporting in vivo Experiments) guidelines.

# AUTHOR CONTRIBUTIONS

JiaZ is the principal investigator. JL, ZS, and FL contributed to the study design, performance, and manuscript draft. YF, JinZ, SX, and SM analyzed the experimental data. WX, LS, and HW revised the manuscript and polished the language.

# FUNDING

This work was supported by grants of the Major Research and Development Project of Zhejiang Province (2017C03021) to JiaZ and the Youth Fund of the National Natural Science Fund project (81701214) to HW.

# REFERENCES


# ACKNOWLEDGMENTS

We thank Susu Lou for support to JL.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01360/full#supplementary-material

Supplementary Figure S1 | Experimental grouping.jpg. *In-vivo* and *in*-*vitro* experimental groups.

Supplementary Figure S2 | FACS gating strategy.jpg. Gating strategy used to isolate microglia.

Supplementary Figure S3 | Single-stained fluorescence.jpg. Single-stained high-magnification (40×) fluorescence images.


pathway attenuates cardiac microvascular ischemia-reperfusion injury. J Pineal Res. (2018) 65:e12503. doi: 10.1111/jpi.12503


rats via suppressing apoptosis, inflammation, oxidative stress, DNA damage, and mitochondria injury. Transl Stroke Res. (2018) 9:74–91. doi: 10.1007/s12975-017-0559-x


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

Copyright © 2019 Lu, Sun, Fang, Zheng, Xu, Xu, Shi, Mei, Wu, Liang and Zhang. 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.

# Mucosal Administration of E-selectin Limits Disability in Models of Multiple Sclerosis

Jacqueline A. Quandt <sup>1</sup> \*, Pierre Becquart <sup>1</sup> , Emily Kamma<sup>1</sup> and John Hallenbeck <sup>2</sup>

<sup>1</sup>Department of Pathology and Laboratory Medicine, University of British Columbia, Vancouver, BC, Canada, <sup>2</sup>Stroke Branch, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, United States

E-selectin plays an important role in mediating the rolling of leukocytes along and thus, the subsequent extravasation across activated endothelial cells comprising the microvasculature of the blood brain barrier (BBB). In multiple sclerosis (MS) and other inflammatory disorders of the central nervous system (CNS), the microvasculature is altered and immune cells infiltrate the brain and spinal cord contributing to damage, demyelination and ultimately disability. While mucosal administration is typically used to affect lymphocyte hyporesponsiveness or tolerance to suspect autoantigens, intranasal administration to E-selectin has previously been shown to protect against CNS inflammatory insults. We characterized the potential for mucosal administration of E-selectin to modulate CNS autoimmunity in the experimental autoimmune encephalomyelitis (EAE) model of MS. Intranasally administered E-selectin reduced swelling by as much as 50% in delayed-type hypersensitivity reactions compared to ovalbumin-tolerized controls. Intranasal E-selectin delivery prior to disease induction with myelin oligodendrocyte glycoprotein (MOG)35–55 reduced disease severity and total disease burden by more than 50% compared to PBS-tolerized animals; this protection was not associated with differences in the magnitude of the autoimmune response. Examination after the onset of disease showed that protection was associated with significant reductions in inflammatory infiltrates throughout the spinal cord. Tolerization to E-selectin did not influence encephalitogenic characteristics of autoreactive T cells such as IFN-gamma or IL-17 production. Clinical disease was also significantly reduced when E-selectin was first delivered after the onset of clinical symptoms. Splenic and lymph node (LN) populations from E-selectin-tolerized animals showed E-selectin-specific T cell responses and production of the immunomodulatory cytokine IL-10. Transfer of enriched CD4+ T cells from E-selectin tolerized mice limited disability in the passive SJL model of relapsing remitting MS. These results suggest a role for influencing E-selectin specific responses to limit neuroinflammation that warrants further exploration and characterization to better understand its potential to mitigate neurodegeneration in disorders such as MS.

Keywords: multiple sclerosis, E-selectin, mucosal tolerance, experimental autoimmune encephalomyelitis, neuroinflammation

#### Edited by:

Craig Stephen Moore, Memorial University of Newfoundland, Canada

#### Reviewed by:

Denis Gris, Université de Sherbrooke, Canada Manu Rangachari, Laval University, Canada

> \*Correspondence: Jacqueline A. Quandt jquandt@pathology.ubc.ca

Received: 04 June 2019 Accepted: 22 July 2019 Published: 27 August 2019

#### Citation:

Quandt JA, Becquart P, Kamma E and Hallenbeck J (2019) Mucosal Administration of E-selectin Limits Disability in Models of Multiple Sclerosis. Front. Mol. Neurosci. 12:190. doi: 10.3389/fnmol.2019.00190

# INTRODUCTION

Multiple sclerosis (MS) is amongst the most common neurological diseases in young adults, with symptoms such as altered balance, vision or speech, weakness or paralysis, as well as extreme fatigue, depression and cognitive dysfunction. MS is a chronic inflammatory neurodegenerative disorder characterized histologically by vascular changes, immune infiltrates, loss of myelin and oligodendrocytes, and damage or loss of axons (Frohman et al., 2006). Ultimately, chronic disability is closely linked to neurodegeneration and irreversible axonal loss (Bjartmar and Trapp, 2003). Oligodendrocytes, myelin, neurons and axons are all damaged (Frischer et al., 2009) in MS and its primary animal model, experimental autoimmune encephalomyelitis (EAE). The demyelination and axonal injury observed in acute inflammatory lesions (Trapp et al., 1998; Kuhlmann et al., 2002) is thought to be driven by mechanisms distinct from those driving diffuse-axonal degeneration and atrophy associated with disease progression (Ingle et al., 2003; Kutzelnigg et al., 2005). Relapsing-remitting (RRMS) presents with alternating clinical attacks and periods of stability that transition into secondary progressive MS (SPMS) with progressive deterioration. Approved disease modifying therapies (DMT) for MS target inflammatory processes and have demonstrated significant benefit in studies of RRMS (Yong et al., 2018), yet only recently have DMT been approved to give modest but significant benefits in progressive or primary progressive MS (PPMS). Specifically, the B-cell depleting antibody, ocrelizumab, has been shown to significantly reduce the confirmed disability progression over 12 weeks in ocrelizumab patients compared to those taking placebo (Montalban et al., 2017). Siponimod, a compound which traps B and T cells in the body's immune organs has also shown evidence of slowing disability progression and has recently been approved for patients with active SPMS (Kappos et al., 2018).

Focal lesions are attributed to activated T cells that travel to the central nervous system (CNS) and trigger an inflammatory cascade as they encounter antigen (Bar-Or, 2008). CD4 and CD8 T cells may contribute directly and indirectly to disease but have been postulated to have both pathogenic and suppressive/regulatory roles over the disease course (Hauser et al., 2008; Racke, 2009; Johnson et al., 2010). Cells that induce EAE have a proinflammatory phenotype (Th1) producing interferon-γ (IFN-γ), tumor necrosis factor-α (TNFα), and granulocyte macrophage colony stimulating factor (GM-CSF), in contrast to any of the immunosuppressive cytokines interleukin-4 (IL-4), IL-5, IL-9 or IL-10 (Gor et al., 2003; Sospedra and Martin, 2005). Additionally, the role of IL-17 producing T helper cells (Th17), as key drivers of pathogenic autoimmunity has been highlighted in EAE (Aranami and Yamamura, 2008). Yet, a balance seems to exist over the course of disease between effector T cells and regulatory T cells (Korn et al., 2007). Regulatory T cells (Treg) are capable of limiting immunity either by interacting directly with pathogenic cells, or by synthesis of immunosuppressive factors including IL-10 or transforming growth factor-β (TGF-β; Banham et al., 2006). Specifically, Treg cells expressing forkhead transcription factor FoxP3 are key cells capable of preventing disease in animal models of autoimmunity (Sakaguchi et al., 2008). Several factors in the local tissue environment can influence the phenotype of a T cell; inflammatory mediators or other cytokines, as well as the nature of the antigen presenting cell (APC), can greatly influence the properties of effector cells. Autoimmunity is typically reduced or even prevented if the generation of these pathogenic cytokine-producing cells deviates to a more immunosuppressive phenotype (Zamvil and Steinman, 1990; Segal and Shevach, 1998). As such, therapies which are able to reduce the development of autoreactive T cells, alter the pathogenic profile of these cells, or encourage the generation of regulatory T cells hold promise.

Mucosal tolerance describes the phenomenon of limiting immune responsiveness to antigens delivered via oral or nasal routes that would typically be considered non-pathogenic (i.e., food). Tolerance induction has been demonstrated via the intranasal route and is thought to involve active suppression as it can be transferred with adoptive transfer of splenocytes (Unger et al., 2003). In this regard, antigens delivered to the nasal mucosa spurs activation after draining into the cervical lymph nodes (CLN) and spleen which leads to tolerance; the internal jugular superficial cervical lymph nodes (SCLN) are also key in mediating tolerance (Wolvers et al., 1999; Li et al., 2011). Intranasal and oral delivery of antigen for tolerization has been effective in animal models of autoimmunity, atherosclerosis, transplant rejection, and allergy (Maron et al., 2002; George et al., 2004; Mayer and Shao, 2004; van Puijvelde et al., 2006; Broere et al., 2008; Klingenberg et al., 2010).

E-selectin is a cell surface adhesion molecule rapidly and specifically induced on activated endothelium which along with P-selectin, mediates tethering and rolling of immune cells in vivo and in vitro (Lawrence and Springer, 1991, 1993; Frenette et al., 1996). E-selectin has been implicated in processes of rolling and recruitment in EAE (Engelhardt et al., 1997; Doring et al., 2007), and soluble plasma levels have been associated with inflammatory activity in MS (Kuenz et al., 2005). Nasal instillation of E-selectin was previously shown to limit damage secondary to both ischemic and hemorrhagic strokes and to be protective in a vascular cognitive impairment model (Takeda et al., 2002, 2004; Wakita et al., 2008). E-selectin administration gave rise to a cell-mediated protection in models of stroke (Chen et al., 2003) as well as limiting atherosclerosis in susceptible Apoe−/<sup>−</sup> animals (Li et al., 2011). This study was designed to test whether tolerization to human E-selectin via intranasal administration could alter the course and severity of disease in EAE as a model of MS.

#### MATERIALS AND METHODS

#### Intranasal Administration of E-selectin

Experiments were approved and conducted in accordance with the Guide for the Care and Use of Laboratory Animal Resources (1996) and the Canadian Council on Animal Care guidelines, with approval by the NINDS Animal Care and Use and UBC Animal Care Committees. Intranasal instillations were as follows: (1) PBS (Biowhittaker, Walkersville, MD, USA); and (2) recombinant human (hu) or mouse (m) E-selectin (**Supplementary Figure S1**) synthesized with a Baculovirus expression system with lectin and epidermal growth factor domains (prepared and provided by Novavax Pharmaceuticals, Rockville, MD, USA). Instillations were carried out with the animals under brief anesthesia with 4% isoflurane using a 10 µl pipette tip and micropipettor.

The tolerization regimens changed over the course of the study (**Figure 1**) depending on whether it was administration of E-selectin to limit delayed-type hypersensitivity (DTH; **Figure 1A**), prophylactic administration of E-selectin prior to active EAE induction to prevent EAE (**Figure 1B**), or therapeutic administration of E-selectin after the onset of clinical symptoms (**Figure 1C**) PBS (10 µl), E-selectin or ovalbumin (Ova; varied doses in 10 µl) were instilled into each nostril every other day for 10 days (total of five administrations), designated as a single-course regimen. For EAE studies, a repetitive (booster) regimen was also applied, where the booster would be repeated at 3-week intervals.

# Delayed-Type Hypersensitivity Reactions (DTH)

DTH was assessed in C57BL/6 animals receiving a single-course tolerization regimen (five administrations every other day over a total of 10 days per regimen) as previously described (Li et al., 2011). Fourteen days later, the animals were immunized in four spots on the flank with 55 µg E-selectin/200 µl of 0.5mg/ml complete Freund's adjuvant in PBS (Difco Laboratories, Detroit, MI, USA). Fourteen days later, ear thickness was measured with microcalipers and animals were subsequently re-challenged with 5 µg E-selectin/10 µl PBS injected subcutaneously into the ear. Ear thickness increase over baseline was measured 2 days later and presented as percentage increase and compared to increases following a PBS injection in the alternate ear.

#### EAE Induction

For active disease, female 8–10-week-old C57BL6/J mice (Jackson Labs) were immunized with 200 µg myelin oligodendrocyte glycoprotein (MOG)35–55 (MEVGWYRSPF SRVVHLYRNGK; Stanford Pan Facility, Stanford, CA, USA; Quandt et al., 2012). For passive disease, SJL/J mice (Harlan Laboratories) were immunized with 75 µg PLP139–151 (HCLG KWLGHPDKF, Stanford Pan Facility) and draining LN cells enriched with PLP for 3 days in culture before the transfer of 1.5 million blasted cells to healthy recipients as described (Anderson et al., 2004). Animals were monitored by a blinded assessor daily for clinical signs through the onset of disease and through recovery (typically day 30–35) or longer in chronic studies to day 45 or 60 according to the following: 0–0.5 indicated no disease/distal limp tail, 1.0 limp tail, 2.0 mild weakness in one or 2.25 in two hindlimbs/slipping on cage insert bars, 2.5 moderate-severe weakness in one or in 2.75 two hind limbs/getting stuck on cage insert bars, 3.0 paralysis in one or 3.5 both hind limbs, 4.0 hind limb paralysis plus weakness in one or in 4.5 both forelimbs, and five for moribund animals.

# Isolation of Immune Populations for Transfer or T Cell Proliferation and Cytokine Assays

The spleens or SCLN were aseptically removed from mice following CO<sup>2</sup> inhalation and euthanasia and transferred to cold sterile PBS and pressed through a sterile 40 µm filter to prepare a single-cell suspension. For transfer, cells were resuspended in a sterile PBS buffer and incubated with antibodies specific for non-T cell populations to prepare CD4+ T cell-enriched populations by negative selection (Miltenyi Biotec, Gladbach, Germany); a small sample was kept for characterization by flow cytometry which showed CD4+ T cell purities ranged from 93% to 96% in each experiment. Five million cells from tolerized animals were immediately transferred retroorbitally to recipient animals which had been induced for passive EAE as above. For analyses of immune responses or priming, 2–4 × 10<sup>5</sup> spleen or LN cells in serum-free X-vivo 15 media (BioWhittaker, Walkersville, MD, USA) were incubated with varying doses of E-selectin, control culture media, or concanavalin A (Con A; final concentration 2 µg/ml, Sigma, MO, USA) in 96-well round bottom plates for 48 h at 37◦C when tissue culture supernatants were harvested. Thymidine (1 mCi of [3H]-thymidine) was added for 16 h prior to harvesting the cells at 72 h and measurement of incorporated radioactivity. Cytokine analysis was performed using anti-mouse IFN-γ IL-10-, IL-17 or TGF-β-specific DuoSet kits from R&D Systems (Minneapolis, MN, USA) per protocol directions.

#### Serum Analyses

Blood was collected from the tail vein via sterile scalpel from mice anesthetized under 4% isoflurane. Samples sat at room temperature for 30 min in a Serum Gel Z/1.1 tube (Sarstedt, Numbrech, Germany) prior to processing per manufacturer's recommendations and frozen at −80◦C until assay. Serum E-selectin was measured using a mouse E-selectin-specific Duoset kit from R&D Systems as above and MOG35–55 specific IgG1 and IgG2a/c antibodies were detected as previously described (Neil et al., 2017).

#### Immunohistochemistry of EAE Tissues

Mice were deeply anesthetized with a ketamine/xylazine cocktail prior to transcardial perfusion with PBS followed by 10% buffered formalin and processed for immunohistochemistry (IHC; Quandt et al., 2012). Immunofluorescence with antibodies to SMI-32 (to hypophosphorylated neurofilament, Biolegend, San Diego, CA, USA) and myelin basic protein (MBP, Santa Cruz Biotechnology, Dallas, TX, USA) were used to assess infiltration, axonal damage as well as demyelination in the spinal cord. Tissue sections were examined with a Zeiss Axio Vert 200 Inverted Fluorescence Microscope (Oberkochen, DE, Germany), equipped with an Axiocam 506 monochrome camera at 20× with matched light intensity and exposure times across the entire experimental set. Analysis was performed with Zeiss software Zen (Version 2.3). Immunostained sections from each EAE mouse spinal cord were analyzed (n = 5 per treatment, three levels per mouse corresponding to the T8/T9, T12,

and L4/5 levels of the spinal cord for a total of 15 levels per treatment group examined). The sections were assessed for the degree of inflammatory infiltration by outlining and pooling the area of all infiltrates per section and calculating the percentage of white matter (WM) infiltrated by immune cells by taking this as a fraction of the WM area in that section. SMI32-positive regions of interest typically resembling spheroids with a staining intensity above the threshold set by averaging regions of WM in healthy animals were used to report the number of SMI32+ spheroids per square millimeter of WM.

#### Statistical Analysis

GraphPad Prism (version 7, La Jolla, CA, USA) was used for all analyses. A repeated-measures one-way analysis of variance (ANOVA) was used for normally distributed data or a repeated measures ANOVA on ranks (Friedman test) was performed when comparing multiple treatments. In EAE comparisons of multiple doses over time, a two-way ANOVA on ranks was performed followed by a Tukey multiple comparisons test vs. PBS-treated controls. Values represent mean ± standard deviation unless otherwise indicated. A Mann-Whitney comparisons test was performed in comparisons of EAE to healthy or sham/CFA immunized mice and significance reported at p < 0.05.

#### RESULTS

#### Intranasal Administration of E-selectin Specifically Limits Inflammation Secondary to Immunization With E-selectin or EAE Induction With MOG

In C57BL/6J mice, animals sensitized and later undergoing recall to either human or mouse E-selectin in vivo in a DTH assay (**Figure 1A**) showed similar responses: the average increase of ear thickness in mice sensitized with hE-selectin and recalled later with hE-selectin was 0.331 ± 0.02 mm and was similar to mice sensitized with msE-selectin and later challenged with mE-selectin at 0.324 ± 0.06 mm. In animals sensitized to huE-selectin, intranasal administration of huE-selectin limited swelling to 57.9 ± 4.2% compared to 128.5 ± 3.8% observed in animals receiving Ova intranasally (**Figure 2A**, 5 µg dose; p < 0.0001). Similarly, intranasal msE-selectin limited swelling in msE-selectin-sensitized mice although to a lesser degree (94.7 ± 5.4% vs. 126.2 ± 7.4% in Ova-tolerized animals **Figure 2B**; p = 0.02). A lower dose of 1 µg still significantly reduced hE-selectin responses whereas 1 µg of mouse E-selectin was not effective (data not shown). Notably, human and mouse E-selectin proteins share 81% amino acid similarity.

We next administered hE-selectin intranasally to test its potential to limit autoimmunity and clinical disease in the actively induced MOG35–55 chronic model of EAE. Animals received 0.1–10 µg of hE-selectin in three sessions which precede and span the induction of EAE as in **Figure 1B**. Compared to intranasal PBS, 5 µg of hE-selectin reduced clinical disability in animals to the greatest degree, although 0.1 and 10 µg doses were also effective (**Figure 2C**). Intranasal administration of 5 µg hE-selectin significantly reduced the cumulative disease burden over the study (37.1 ± 9.4 vs. 70.3 ± 13.7 in PBS, p = 0.05) and the average disease severity (1.16 ± 0.27 vs. 2.141 ± 0.42 in PBS, p = 0.05). Comparisons of the disease score over time showed that the means of each treatment were significantly different to PBS (PBS vs. 0.1 µg, p < 0.0001, PBS vs. 1 µg, p = 0.0002, PBS vs. 5 µg, p < 0.0001, PBS vs. 10 µg, p = 0.0007). We postulated that intranasal administration of E-selectin may limit the magnitude of a response to MOG in these animals, and examined spleens from animals culled at day 45 and tested recall ex vivo to both MOG and hE-selectin. Notably, stimulation indices to MOG were not significantly different between the two groups (**Figure 2D**). Furthermore, hE-selectin-specific responses could only be detected in populations derived from E-selectin tolerized animals. Notably, we have found that the isotype of antibody responses to MOG35–55 can be a sensitive means of detecting shifts in Th1 or Th2 profiles (Neil et al., 2017) and found that IgG1 and IgG2a antibody responses to MOG35–55 were comparable in PBS and hE-selectin serum from mice examined at day 45 tolerized with the regimen in **Figure 2C** (data not shown). To determine whether or not E-selectin tolerization interfered with the initial priming of encephalitogenic T cells, draining

FIGURE 2 | Intranasal E-selectin limits DTH responses and limits clinical disability in EAE. (A) Percentage change in ear thickness in animals sensitized to huE-selectin after intranasal administration of 5 µg E-selectin or Ova and challenged with PBS or huE-selectin in opposite ears. (B) Percentage change in ear thickness in animals sensitized to mE-selectin after intranasal delivery of mE-selectin or Ova. No differences in ear thickness were observed between ears receiving PBS challenge following either regimen. n = 5 mice per treatment group, bars represent the SEM. Mann-Whitney T-test comparing Ova to E-selectin tolerization. (C) huE-selectin was administered per the regimen in Figure 1B spanning the active induction of chronic EAE. 0.1 up to 10 µg of huE-selectin were administered and results represent the mean ± SEM of the animals over time. (D) Splenocytes isolated from animals at D.45 post-EAE induction were challenged ex vivo with different doses of MOG35–55 or huE-selectin for 72 h to examine the proliferation relative to untreated samples (stimulation index or SI). n = 3–4 animals per group, dots and error bars represent the average SI of responses from PBS (black) or huE-selectin tolerized (red) animals ± SD. Responses to huE-selectin in PBS tolerized animals were negligible. (E) Cells isolated from the draining axillary, brachial and inguinal lymph node (LN) 12 days after EAE induction were tested for proliferative responses to MOG35–55 (20 µg/ml), E-selectin (10 µg/ml) and anti-CD3 combined with anti-CD28 (n = 4–5 mice per group, stimulation index (SI) is proliferation relative to untreated samples from PBS—black or E-selectin-red-tolerized animals ± SEM).

inguinal, axillary and brachial LN populations were isolated from animals receiving one round of intranasal PBS or E-selectin and actively immunized to induce EAE. Animals were culled 12 days after immunization and tested for proliferative responses to MOG, hE-selectin, and anti-CD3 as well cytokine production over 48–72 h. Proliferation responses measured by stimulation indices to MOG35–55 and anti-CD3 were similar between PBS and E-selectin tolerized animals (**Figure 2E**); however, proliferation in response to E-selectin was only detected when E-selectin had previously been administered (p = 0.023). Both IL-17 and IFN-γ production, in response to 20 µg/ml of MOG35–55, was similar in cells derived from E-selectin tolerized LN as from those receiving PBS (596 ± 84.3 pg/ml vs. 654.8 ± 189 pg/ml and 426 ± 64.8 vs. 502 ± 72.1 pg/ml respectively). Polyclonal T cell activation with anti-CD3 and CD28 also yielded similar IL-17 and IFN-γ amounts from the two different treatment groups (n = 5 mice per treatment group). No IL-17 or IFNγ was detected in either group in response to E-selectin and production of TGF-β was negligible across all tolerization and stimulation groups (data not shown). IL-10 levels were near the lower limit of detection (0–16 pg/ml) and was not significantly different between PBS and E-selectin tolerized populations no matter the stimulus. At day 12, animals receiving PBS scored at 0, 2, 1.5, 1 and 2.5; animals receiving huE-selectin scored at 0, 0, 0, 0 and 2.75. Examination of three levels of spinal cord showed that animals which received intranasal E-selectin had inflammatory infiltrates in less than 1/3 of the tissue observed in PBS tolerized mice (**Figure 3A**; 1.12% ± 0.5 of the tissues vs. 4.27% ± 1.6 of the tissue, p = 0.012). No significant differences in the accumulation of SMI-32+ spheroids as a marker of axonal damage were observed at this time point (**Figures 3B–G**; data not shown).

#### Therapeutic Potential of E-selectin Tolerization and Immune Responsiveness

Administration of huE-selectin after the onset of disease symptoms was tested for the potential to treat disease in chronic EAE. Administration of huE-selectin after the onset of disease (starting on day 13 for 10 days as in **Figures 1C**, **4A**) reduced the average disease score over the study period (2.60 ± 0.13 vs. 2.12 ± 0.09, p = 0.0006). To characterize the specificity of the immune response to E-selectin, superficial cervical LN and splenic populations were examined from mice at the end of the study period (d.45 post-immunization) for recall to E-selectin and compared to the T cell mitogen Con A (**Figure 4B**). E-selectin-specific responses were only detected in animals that had been tolerized to E-selectin. Likely consistent with fewer T cells in the spleen than LN, the splenic responses to E-selectin were less than those observed in the LN (1.58 ± 0.18 vs. 5.72 ± 2.45, p = 0.03). Responses to Con A were similarly lower in spleens than LN in both groups of tolerized animals, and no differences were observed between the proliferative responses of cells derived from the PBS vs. E-selectin tolerized animals. Supernatants from spleen cells and LN cells isolated from PBS or E-selectin tolerized animals were tested for IL-10 production; IL-10 production in response to E-selectin was only detected in animals that had received E-selectin (**Figures 4C,D**) where it was significantly greater (p < 0.05) than that observed in PBS-tolerized animals. Notably, IL-10 production in response to Con A was similar from both PBS and E-selectin tolerized animals.

Our group had previously shown that transfer of splenic cells from E-selectin tolerized cells was able to confer protection in

experimental models of disease and we sought to test the same in EAE (**Figure 5**). To focus on CD4+ T effector populations, we purified CD4+ T cells from the superficial cervical LN or spleens of animals that had undergone two rounds of intranasal administration (per **Figure 1**) with PBS, OVA or E-selectin and transferred them into animals induced passively for PLP139−<sup>151</sup> EAE.

Animals receiving splenic cells from huE-selectin or Ova-tolerized animals or LN cells from Ova-tolerized animals all had similar average disease severities over the study period to PBS-tolerized animals (1.87 ± 0.30, 1.84 ± 0.31, 2.05 ± 0.31 and 1.95 ± 0.32); only animals receiving LN cells from E-selectin tolerized mice were different at 1.611 ± 0.25 (p < 0.05 over the disease course; **Figure 5A**). In a replicate of the study over

FIGURE 4 | Administration of E-selectin treats disability in chronic EAE and generates E-selectin-specific populations in immune organs. (A) Mice were immunized with myelin oligodendrocyte glycoprotein (MOG)35–55 to actively induce EAE. Once more than half the animals exhibited a limp tail (day 13) huE-selectin tolerization was initiated every other day for 10 days. Animals were followed to day 45. Average disease scores were significantly reduced over the treatment study, <sup>∗</sup>p < 0.05 t-tests vs. PBS-treated mice indicated those days which were significant. n = 13 and 14 for the PBS and E-selectin groups respectively. One of three representative experiments is shown. (B) Superficial LN and splenic populations were isolated from the mice in (A) at day 45 and challenged ex vivo with E-selectin or Con A to assess the response to antigen. Proliferative responses over 72 h to 20 µg/ml huE-selectin (left panel) were only observed in LN and spleen populations which had received E-selectin intranasally (red bars). Data represents spleen and cervical lymph nodes (CLN) from three mice per treatment group, average ± SD. (C,D) Supernatants from the splenic and LN populations isolated from animals undergoing E-selectin tolerization were examined for IL-10 production, PBS or huE-selectin. Spleens from n = 2 or from LN n = 3 mice were assayed per regimen treatment group, bars represent the average produced from duplicate or triplicate wells, error bars represent the SD. p < 0.05.

60 days, LN cells from huE-selectin tolerized animals had a significantly lower disease score over the study than cells from Ova tolerized animals (3.16 ± 0.15 vs. 2.51 ± 0.13, p < 0.0001; **Figure 5B**). LN populations were assessed by flow cytometry prior to transfer in each experiment, with 94.3%–96.5%

CD4 T cell purity and within that population 10.3%–12.6% FoxP3+/CD25+ cells in the Ova vs. huE-selectin transferred populations, showing no significant difference in T reg numbers between the two transferred populations.

# DISCUSSION

The microvasculature of the brain exhibits specialized tight junctions, low levels of vesicular trafficking and limited cell adhesion molecule (CAM) expression that otherwise enables immune cell movement out of the vessel (Reese and Karnovsky, 1967; Risau et al., 1998). Therapies which limit endothelial cell activation or immune cell recruitment across the blood brain barrier (BBB) offer promise in MS. Indeed, natalizumab, an antibody to α4 integrin the ligand for vascular CAM-1 (VCAM-1), was extremely effective in treating EAE (Zhang et al., 2003) and went on to show the ability to reduce relapses in MS patients by 68% compared to placebo, which far exceeded results for approved MS therapies at that time (Polman et al., 2006). However, this, and more recently approved efficacious therapies, have an increased risk profile, where depleting immune cell subsets or reducing immune surveillance with monoclonal antibodies were associated in rare instances with multifocal leukoencephalopathy (PML) as a rare yet serious adverse event (Piehl et al., 2011; Yong et al., 2018).

The current study uses a novel approach enabling expression of E-selectin under inflammatory conditions to localize and perpetuate an immunoregulatory response. Intranasal administration of E-selectin was a driver of specific immune responses in both the SCLN and spleen which were associated with the expression of the immunomodulatory cytokine IL-10 (Li et al., 2011), as previously described.

In our studies, we compared the efficacy of purified CD4+ T cells from SCLN to splenic populations and found E-selectin-specific cells derived from SCLN after two rounds of priming had efficacy. Many studies have reported intranasal antigen administration as a means to regulate CD4+ T cell function and have linked IL-10 to subsequent immune tolerance (Hoyne et al., 1996; Burkhart et al., 1999; Laliotou and Dick, 1999). In an experimental autoimmune uveitis model, authors showed a transfer of tolerance with a regulatory population of spleen cells (Burkhart et al., 1999; Laliotou and Dick, 1999). Studies have implicated APCs as key mediators of tolerance, particularly those in the regional drainage lymph nodes; a single intranasal administration was rapidly followed by T cell proliferation secondary to signaling networks in both SCLN and spleen (Wolvers et al., 1999). It may be that the multiplicity and boosting of these administrations in our model contributed to the increased efficacy observed upon transfer of SCLN-derived populations, which was also apparent in the increased proliferative responses observed when comparing E-selectin responses between the two immune organs.

The administration of E-selectin repeatedly, as with any antigen, has the capacity to generate both T and B cell responses specific to E-selectin. Although we previously showed that antibodies specific for E-selectin are generated using this approach, we showed that they are indeed of the IgG1 isotype, and are therefore a more immunsuppressive than inflammatory phenotype. This, combined with previous observations from other groups which show that neutralization of E-selectin with antibodies does not reduce the severity of EAE, suggests that the primary mechanisms of action here are not attributable to antibodies generated from T cell-aided B cell responses. It is well known that administration of antigens through mucosal routes rapidly generates T cell populations, immunomodulatory in nature, which, through their production of cytokines such as IL-10, can limit inflammation. Numerous reports have outlined the relative success of antigen-specific tolerance to limit EAE based on the administration of the candidate autoantigen driving the disease, although discussion surrounds the safety and also reliability of mucosal administration of antigen to suppress ongoing active disease (Burkhart et al., 1999). To test the efficacy of E-selectin administration in the most robust manner, we tested the approach prophylactically and therapeutically in active and passive models of chronic and relapsing-remitting MS, to overcome previous concerns that only in more of the milder relapsing disease settings would tolerance approaches have success (Liu and Wraith, 1995; Bai et al., 1997; Thurau et al., 1997).

The very nature of the tolerizing antigen in our approach may be the key to the success here; rather than having to energize or deviate the full weight of the immune response to the candidate autoantigen, E-selectin tolerization and the potential of E-selectin specific cells to act at the specific site of inflammatory responses and damage may significantly augment its potential to limit disease, without the risk of overt immunosuppression. Furthermore, in the case of MS where the definitive putative autoantigen remains elusive, the specific autoantigen need not be identified for this approach to yield benefit. Here, we target immunosuppressive cell populations towards the activated microvasculature and/or local immune organs; the specific sites remain to be identified. Future studies will be important to establish the site where these cells act to afford protection. There is indeed the possibility that the human E-selectin gives a more robust response as seen in the DTH because it differs from the mouse protein in sequence. It is also possible that any differences observed in the glycosylation patterns of otherwise homologous regions of the protein may be enhancing immunogenicity following intranasal administration. It is possible that E-selectin from another species from human may be more potent in humans, but may increase the potential for autoimmunity if high avidity cells specific for a non-human protein that have not been depleted through tolerance to self (human) antigen are indeed activated. Because self into self in our model was not a clear driver of autoimmunity, yet showed efficacy albeit lower than that observed with that from a different species, we believe a homologous system is a safe approach to consider in driving antigen-specific tolerance to E-selectin.

Despite E-selectin upregulation in association with activation of the vasculature in a variety of CNS insults or inflammatory settings and their models, a specific requirement for E-selectin in EAE development has been discounted. Intravital microscopy experiments showed that E-selectin is localized to regions of immune cell rolling and interactions with the superficial CNS microvasculature (Kerfoot et al., 2006); moreover, E- and P-selectin mediate early cerebrovascular T cell/endothelial interactions in EAE, and P-selectin glycoprotein ligand-1 (PSGL-1, one ligand for E- and P-selectin) is a primary mediator of E- and P-selectin-mediated T cell rolling in the spinal cord (Sathiyanadan et al., 2014). In contrast, studies have: (a) failed to localize E-selectin to parenchymal vessels over the course of EAE development; and (b) have shown that neutralization of E-selectin with antibodies or E-selectin knockout models show no differences in clinical disease or infiltrate composition compared to their wildtype counterparts (Engelhardt et al., 1997). In this regard, one concludes such rolling of T cells is not required for the initiation of EAE. However, elegant studies in relapsing and remitting remodels of disease have shown an important role for neo-epitope generation in potentiating disease through antigen presentation and the potentiation of immune responses to newly generated antigens within the CNS as proposed in antigenic ''spread'' (McMahon et al., 2005). Notably, relapse in chronic-relapsing EAE is less severe where CNS-draining lymph nodes have been removed (van Zwam et al., 2009). E-selectin draining to these lymph nodes during EAE or MS may be the driver for IL-10 responses that limit immune responses here, and foreseeably further potentiation. Future experiments will build on this initial study, and look to further characterize E-selectinspecific populations and their potential to limit inflammation both in vitro and in vivo.

In vitro models and the postulated pathogenesis of MS suggest that the BBB is not restrictive to trafficking of activated lymphocytes (Goverman, 2009). In our model, where T cells specific for E-selectin are generated and repeatedly challenged, this may lead to E-selectin specific cells that can gain entry to the CNS, even when the BBB has not been compromised. If these cells access the CNS and can go on through the cerebrospinal fluid to the recently characterized CNS meningeal lymphatics which then drain to cervical lymph nodes (Louveau et al., 2015), exposure here to E-selectin may drive their production of immunomodulatory factors such as IL-10 which may help to limit antigenic spread and the potentiation of immune responses that are more likely to start within the CNS. Additional studies are required to titer as well to localize E-selectin-specific T cells in our model over the course of EAE to better understand their full potential to actively and passively protect against and treat established disease in our models.

While rapid expression of E-selectin at the CNS microvasculature surface at a specified point in the disease process may be the most robust signal of inflammation in acute settings such as stroke, perhaps equally as important is the shedding of E-selectin. While not characterized over the course of disease, we did detect mouse E-selectin in the serum of animals undergoing acute attacks of EAE (ranging from 250 to 300 pg/ml); notably, these levels were highest in animals undergoing EAE that had been tolerized with Ova; levels were significantly lower in animals that had been tolerized with human E-selectin suggesting that the relative level of serum E-selectin may be linked to the severity of disease and/or inflammation. In our therapeutic studies of E-selectin administration after the onset of disease, we saw no difference in serum E-selectin levels between animals receiving PBS or E-selectin 1 month (588.0 ± 15.1 vs. 566.6 ± 11.9 ng/ml) or 2 months (310.5 ± 12.1 vs. 366.5 ± 13.3 ng/ml) after disease induction, although soluble E-selectin had clearly dropped by almost half over that time. Notably, sE-selectin is markedly increased in RRMS patients during an exacerbation or those with chronic progressive MS in comparison to controls (Tsukada et al., 1995). Notably, MS was distinct from HTLV-1 associated myelopathy (HAM) in this regard where levels of sE-selectin were similar between patients with HAM and controls (Tsujino et al., 1998). Authors also frequently found sE-selectin in the CSF during exacerbations in RRMS, supporting a role for E-selectin–mediated responses to endothelial cell activation or damage during an exacerbation (Droogan et al., 1996). In contrast, others found CSF and serum sE-selectin levels were similar between patients in relapse and those with other inflammatory (IND) and non-inflammatory neurological disease (NIND). Another report sought to determine whether concentrations of sE-selectin correlated with gadolinium-DPTA enhancement (a marker of enhanced BBB permeability) on MRI in patients with MS. Authors found levels of sE-selectin were significantly increased in PPMS patients compared to the near control levels of sE-selectin detected in RRMS and SPMS patients. Interestingly, 5 of the 10 primary progressive patients with high levels of sE-selectin showed very low activity on MRI scans. While there was no correlation between sE-selectin concentrations and contrast-enhancing lesions on MRI, a close correlation was observed between TNF-α and sE-selectin (Giovannoni et al., 1996). Other groups have documented the same high levels of serum sE-selectin in PPMS compared with RRMS and SPMS (McDonnell et al., 1999) and provide evidence for significant immunological heterogeneity in MS related to immune/endothelial cell interactions in different types of MS. It also goes against the dogma of lesser inflammatory involvement that has previously characterized PPMS. This unique association between serum E-selectin and PPMS may indeed highlight the importance of advancing the therapeutic consideration of E-selectin tolerization to limit neurodegeneration in MS, with relevance to particularly acute and active processes such as those implicated in PPMS.

#### DATA AVAILABILITY

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

#### ETHICS STATEMENT

#### Animal Subjects

All experiments were carried out in compliance with the Guide for the Care and Use of Laboratory Animal Resources (1996), the Canadian Council on Animal Care guidelines and approved by the NINDS Animal Care and Use and UBC Animal Care Committees.

#### AUTHOR CONTRIBUTIONS

JQ performed EAE induction, scoring, data collection, analyses, immunoassays, cell culture, tissue isolation and preparation, prepared the first and subsequent versions of the manuscript, formulated the research plan and series of experiments with input from JH; and oversaw the experimental approach and design, reviewed data and analyses and prepared the manuscript including revising the manuscript critically for technical and scientific accuracy with JH. PB and EK performed immunohistochemistry and analyses of that data. All authors have reviewed the manuscript in entirety and are in agreement with the content of this work.

#### FUNDING

This study was supported by funds from the Intramural Research Program of the Neuroimmunology Branch at the National Institutes of Health and post-doctoral support to Xinhui Li. JQ was supported by an MS Society of Canada Donald Paty Career Development Award and additional support was provided by philanthropic funds for our MS Neuroprotection Research Program through the UBC Hospital and VGH foundation.

#### ACKNOWLEDGMENTS

We would like to thank Jaebong Huh and Xinhui Li for technical assistance with aspects of EAE scoring, immunoassays and data analyses.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnmol.2019.001 90/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 Quandt, Becquart, Kamma and Hallenbeck. 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.

# Central Nervous System Remyelination: Roles of Glia and Innate Immune Cells

Charbel S. Baaklini <sup>1</sup> , Khalil S. Rawji <sup>2</sup> , Greg J. Duncan<sup>3</sup> , Madelene F. S. Ho<sup>1</sup> and Jason R. Plemel <sup>1</sup> \*

<sup>1</sup>Department of Medicine, Division of Neurology, Neuroscience and Mental Health Institute, Faculty of Medicine & Dentistry, University of Alberta, Edmonton, AB, Canada, <sup>2</sup>Wellcome Trust-Medical Research Council, Cambridge Stem Cell Institute, Cambridge Biomedical Campus, University of Cambridge, Cambridge, United Kingdom, <sup>3</sup>Department of Neurology, Jungers Center for Neurosciences Research, Oregon Health and Science University, Portland, OR, United States

In diseases such as multiple sclerosis (MS), inflammation can injure the myelin sheath that surrounds axons, a process known as demyelination. The spontaneous regeneration of myelin, called remyelination, is associated with restoration of function and prevention of axonal degeneration. Boosting remyelination with therapeutic intervention is a promising new approach that is currently being tested in several clinical trials. The endogenous regulation of remyelination is highly dependent on the immune response. In this review article, we highlight the cell biology of remyelination and its regulation by innate immune cells. For the purpose of this review, we discuss the roles of microglia, and also astrocytes and oligodendrocyte progenitor cells (OPCs) as they are being increasingly recognized to have immune cell functions.

#### Edited by:

Craig Stephen Moore, Memorial University of Newfoundland, Canada

#### Reviewed by:

Andrew David Greenhalgh, INRA UMR1286 Laboratoire NutriNeuro, France Samaneh Maysami, University of Manchester, United Kingdom

#### \*Correspondence:

Jason R. Plemel jrplemel@ualberta.ca

Received: 28 June 2019 Accepted: 04 September 2019 Published: 19 September 2019

#### Citation:

Baaklini CS, Rawji KS, Duncan GJ, Ho MFS and Plemel JR (2019) Central Nervous System Remyelination: Roles of Glia and Innate Immune Cells. Front. Mol. Neurosci. 12:225. doi: 10.3389/fnmol.2019.00225 Keywords: remyelination, microglia, oligodendrocyte, oligodendrocyte progenitor cells, astrocytes, white matter disease, aging

#### INTRODUCTION

Demyelination is a common feature of disease occurring after conditions such as spinal cord injury (SCI; Crowe et al., 1997; Powers et al., 2012; Plemel et al., 2014), multiple sclerosis (MS; Franklin and Ffrench-Constant, 2008; Plemel et al., 2017), stroke (Rosenzweig and Carmichael, 2015; Khodanovich et al., 2018), and traumatic brain injury (Mierzwa et al., 2015; Armstrong et al., 2016). Moreover, white matter loss occurs in Alzheimer's disease (Mitew et al., 2010; Carmeli et al., 2013; Zhan et al., 2014; Bejanin et al., 2017) and is a feature of aging that is correlated with cognitive decline (Bartzokis et al., 2012; Yeatman et al., 2014; Li et al., 2017). With age, myelin shows increasing signs of damage while at the same time myelin debris accumulates within microglia (Safaiyan et al., 2016; Hill et al., 2018), which can trigger cholesterol crystal formation and inflammasome activation (Cantuti-Castelvetri et al., 2018). In MS, more remyelination is associated with less disability (Bramow et al., 2010; Bodini et al., 2016), suggesting that enhancing remyelination is a viable therapeutic strategy that is only recently showing benefits in clinical trials (Plemel et al., 2017). However, for those with MS, remyelination is variable and prone to failure (Prineas et al., 1993; Patrikios et al., 2006; Patani et al., 2007; Bramow et al., 2010), especially in the context of aging which is known to slow the rate of remyelination (Shields et al., 1999; Goldschmidt et al., 2009; Ruckh et al., 2012; Brown et al., 2014; Frischer et al., 2015). In this review article, we focus on the innate immune cells of the central nervous system (CNS) and their role during remyelination. We include microglia as the primary CNS innate immune cells, but also recognize that other glia such as astrocytes and oligodendrocyte progenitor cells (OPCs) may sculpt the innate immune response.

# CELL BIOLOGY OF REMYELINATION

### Importance of Remyelination

Demyelination, or the loss of myelin, can be induced due to injury or disease. MS is characterized by inflammatory demyelination, possibly induced by monocytes, macrophage/microglia, and T-cells. For example, microglia/macrophage-derived oxidative species cause damage by inducing myelin and axonal injury (Haider et al., 2011; Nikic et al., 2011; Witte et al., 2019). The immune system produces demyelination by many other mechanisms, but this is not the focus of this review (Mayo et al., 2012; Lassmann and van Horssen, 2016; Mishra and Yong, 2016; Davies and Miron, 2018). Demyelination slows axonal conduction, and as suggested by computer simulations (Koles and Rasminsky, 1972; Waxman and Brill, 1978), may even result in a failure to propagate the signal past the demyelinated segment (Koles and Rasminsky, 1972; Waxman and Brill, 1978). The conduction block associated with demyelination is thought to relate to ionic disbalance once the myelin is removed. The loss of myelin unmasks potassium channels underneath the myelin sheath that can act as a current sink (Schauf and Davis, 1974; Bostock et al., 1978, 1981; Sherratt et al., 1980; Wang et al., 1993, 1995). To compensate, axons can upregulate voltage-gated sodium channels along their length (Foster et al., 1980; England et al., 1991; Black et al., 2007; Hamada and Kole, 2015). Potentially due to this upregulation or redistribution of voltage-gated sodium channels, it is possible to propagate the signal along a demyelinated segment, albeit more slowly (Felts et al., 1997). Associated with the axonal conduction block is a corresponding functional loss or alteration. For example, demyelination of the ventral medial geniculate nucleus (vMGN) or A1 region in the auditory cortex in mice results in both impaired sound frequency-specific responses and an increase in latency in auditory responses (Narayanan et al., 2018). Importantly, while remyelination may speed axonal propagation, it is not always linked to functional recovery (Duncan et al., 2009, 2018; Mozafari et al., 2010; Assinck et al., 2017a; Narayanan et al., 2018). Both the extent and severity of demyelination may dictate the functional consequences. For example, in an experimental murine model of SCI where demyelination is confined to the lesion, preventing remyelination did not change the spontaneous recovery, potentially because after SCI conduction is not sufficiently disrupted (Felts et al., 1997; Duncan et al., 2017, 2018).

Remyelination also maintains axonal integrity and attenuates axonal degeneration. If remyelination is slowed using irradiation to ablate OPCs in mice, there is more axonal degeneration (Irvine and Blakemore, 2008). However, this study is confounded by off-target effects of irradiation including blood-brain barrier (BBB) disruption (Diserbo et al., 2002), enhanced astrogliosis (Wilson et al., 2009), and activation of microglia (Hwang et al., 2006; Irvine and Blakemore, 2008). More specifically, if remyelination is accelerated by promoting oligodendrocyte differentiation in a cell-type specific manner, more axons are preserved (Mei et al., 2016). How might myelin protect axons? One likely mechanism is via the buffering of potassium through the potassium channel KIR4.1, expressed in oligodendrocytes (Schirmer et al., 2018). Potassium is released from axons as part of their action potential and conditional oligodendrocyte-specific removal of this KIR4.1 buffering results in late-onset axonal degeneration, suggesting that myelinic potassium buffering maintains axonal integrity. Oligodendrocytes also support axons by providing them with glycolytic metabolites via the myelin sheath (Fünfschilling et al., 2012; Saab et al., 2016; Micu et al., 2017). Even the myelin sheath may itself can act as a protective barrier by surrounding the axon from toxic reactive oxygen species (Nikic et al., 2011; Witte et al., 2019). The myelin sheath also shifts some of the metabolic demands from the axon to the oligodendrocyte. For example, when an axon is myelinated there is less sodium released for an axon potential, and therefore less energy is required to repolarize its membrane, yet the production of myelin is energetically expensive (Harris and Attwell, 2012). Other possible mechanisms of axonal support by oligodendrocytes include the release of oligodendrocyte-derived exosomes or ribosomes (Frühbeis et al., 2013; Shakhbazau et al., 2016). Oligodendrocytes can also secrete many other factors to boost neuronal health or survival in culture such as insulin-like growth factor 1 (IGF-1) and glial cell-derived neurotrophic factor (GDNF; Wilkins et al., 2001, 2003; Dai et al., 2003), which may support axons in vivo.

An analogous relationship between glial cells and axons exists in the peripheral nervous system (PNS). Schwann cells (SCs) are glial cells generally restricted to the PNS that arise from neural crest-derived SC precursors (Monk et al., 2015). Under normal conditions, SCs myelinate axons in the PNS. However, there is evidence that SCs can remyelinate in the CNS under inflammatory conditions. In fact, P<sup>0</sup> staining—a marker for SC-derived myelin sheaths—was observed in MS spinal cord lesions (Itoyama et al., 1983). Indeed, fate-mapping revealed that in LPC-demyelinated lesions, PDGFRa and NG2-expressing progenitor cells in the CNS produce the myelinating oligodendrocytes and SCs (Zawadzka et al., 2010).

# Process of Remyelination

Remyelination is a regenerative process that requires the sequential activation and recruitment of OPCs to areas of demyelination, followed by their differentiation into new oligodendrocytes—but also less frequently into SCs—and subsequent myelin deposition around demyelinated axons (Franklin and Ffrench-Constant, 2008, 2017; Zawadzka et al., 2010; Plemel et al., 2017). OPCs are found throughout the adult mouse brain white matter and gray matter (Dimou et al., 2008; Rivers et al., 2008; Kang et al., 2010). After a demyelinating insult, OPCs transition to a reactive state with the upregulation of key transcription factors such as Nkx2.2, Olig2 and Sox2 (Levine and Reynolds, 1999; Fancy et al., 2004; Zhao et al., 2015). In the lesion, there is a release of many growth factors such as platelet-derived growth factor (PDGF) and fibroblast growth factor (FGF; Hinks and Franklin, 1999; Messersmith et al., 2000). These factors and others are thought to regulate OPC lineage progression by increasing proliferation and migration towards the lesion (Messersmith et al., 2000; Murtie et al., 2005; Dehghan et al., 2012). OPCs proliferate and migrate in the days following demyelination or injury, as part of their recruitment (Redwine and Armstrong, 1998; Hughes et al., 2013). Following OPC recruitment, they differentiate and mature into myelinating oligodendrocytes, which involves the ensheathment and enwrapping of the denuded axon with a new myelin sheath (Tripathi et al., 2010; Zawadzka et al., 2010; Crawford et al., 2016b). Differentiated oligodendrocytes start expressing myelin proteins such as myelin basic protein (MBP), which allows for the compaction of myelin membranes (Kimura et al., 1989; Watanabe et al., 2002; Polito and Reynolds, 2005; Snaidero et al., 2014). Much of the process of oligodendrocyte differentiation is due to key transcription factors. For example, the transcription factor myelin regulatory factor (MyRF) is required for oligodendrocyte maturation and myelination (Emery et al., 2009) and effective remyelination in adulthood (Duncan et al., 2017). MyRF binds to enhancer sequences and directly drives the expression of myelin genes (Bujalka et al., 2013); it is necessary for remyelination after chemical and traumatic demyelinating injuries (Duncan et al., 2017, 2018). Remyelinating oligodendrocytes can adopt different phenotypes, as shown by single-nucleus RNA sequencing (Jäkel et al., 2019). Interestingly, oligodendrocytes in the white matter of MS patients compared to non-diseased controls have different subclustering based on their transcriptomes. This shift in oligodendrocyte phenotypes may reflect an impaired capacity to remyelinate but the functional consequences of these different phenotypes have not yet been directly determined.

Much of the research on remyelination is conducted on demyelinating rodent models. Using these models, it has been shown that OPCs, and not mature oligodendrocytes, are the cells most responsible for remyelination (Crawford et al., 2016a). Transplantation of mature oligodendrocytes into demyelinated lesions do not remyelinate denuded axons (Targett et al., 1996). Oligodendrocyte fate-mapping in mice also demonstrated that mature oligodendrocytes do not engage in remyelination (Crawford et al., 2016a). Surprisingly, new data provides evidence that mature oligodendrocytes can remyelinate in larger mammals. Duncan et al. (2018) find the presence of oligodendrocytes with both thick and thin myelin sheaths using a feline irradiated food-induced demyelination (FIDID) model and vitamin B12 deficiency and in a non-human primate model. Given that remyelination is typically, but not always associated with thin myelin sheaths (Blakemore, 1974; Duncan et al., 2017), one explanation for these results is that spared oligodendrocytes with a thick, unaffected, myelin sheath can re-extend their processes and remyelinate adjacent denuded axons (Duncan et al., 2018). An alternative possibility with this study is that myelin sheath thickness may be differentially regulated by the axon. In a recent study, Yeung et al. (2019) investigated oligodendrogenesis in humans with MS using a specific carbon dating strategy. After cell division, the daughter cell's DNA will have a radioactive carbon concentration that is proportional to concentration of radioactive carbon in the atmosphere. As a result of atomic bomb testing starting in the mid-late 1940s, the atmospheric radioactive carbon increased making it possible to date daughter cells after the 1940s based on the nuclear radioactive carbon levels. Using this strategy, Yeung et al. (2019) measured new oligodendrocyte production in shadow plaques, which are regions of intermediate lipid staining that are thought to be remyelinated. Surprisingly, they found that new oligodendrocyte production was similar in shadow plaques to those found in the non-diseased brain, suggesting that in humans the mature oligodendrocytes can remyelinate denuded axons. However, in this one study, these conclusions are based off of 11 samples. Moreover, this study found that OPCs did turn over in shadow plaques, but due to the levels of atmospheric radioactive carbon, this was only apparent for those born in the mid-1960s onwards. Only one shadow plaque was measured for oligodendrocyte carbon after this critical mid-1960s period. The question of whether adult oligodendrocytes can remyelinate will, therefore, require further validation in future studies.

#### ROLES OF MICROGLIA AND INFILTRATING MACROPHAGES IN REMYELINATION

Microglia are immunocompetent glial cells of the CNS (Streit, 2002). They are the CNS parenchymal macrophages that arise developmentally from erythromyeloid precursors in the yolk sac (Ginhoux et al., 2010; Kierdorf et al., 2013). They are unevenly distributed according to the CNS region and range from 5 to 12% of glia in the mouse brain and 0.5%–16.6% of human brain parenchymal cells (Lawson et al., 1990; Mittelbronn et al., 2001). They are important for controlling many neurodevelopmental processes. For example, neural precursor cells proliferate in the presence of microglia in vitro (Antony et al., 2011). Also, microglia conditioned media promotes the differentiation of neural precursor cells into neurons as well as astrocytes (Nakanishi et al., 2007; Antony et al., 2011). Microglial ablation results in neuronal apoptosis and a decrease in spine density in young mice indicating microglia promote synaptogenesis and the survival of neurons (Ueno et al., 2013; Miyamoto et al., 2016). Microglia also regulate myelinogenesis through the secretion of growth factors like IGF-1, which is critical for Mbp expression in young mice (Wlodarczyk et al., 2017).

# Microglia Response to Injury

Microglia regulate homeostasis by surveying their microenvironment but are highly responsive to injury or disease as laser-induced injury in the mouse neocortex results in microglial extensions surrounding the site of injury (Davalos et al., 2005; Nimmerjahn et al., 2005). When there is more damage over a longer period of time, for example following focal demyelination with LPC, microglia can retract their processes and become more spheroidal (Plemel et al., 2018). These morphological attributes of activated microglia, as well as similar expression patterns, have made it difficult to differentiate microglia from other macrophages such as border-associated macrophages in the CNS that include meningeal, choroid plexus and perivascular macrophages (Goldmann et al., 2016; Mrdjen et al., 2018), as well as monocyte-derived macrophages (Butovsky et al., 2014). Most studies do not differentiate between these cell types. As such, in this review article, these cells will be referred to as microglia/macrophages.

Microglia are surveillant cells that are highly responsive to environmental cues. In adults, microglia self-renew with modest proliferation (Nimmerjahn et al., 2005; Elmore et al., 2014; Kawabori and Yenari, 2015). In the uninjured state of the CNS, in vivo imaging revealed that ramified microglia continuously scan their microenvironment by undergoing structural changes including filopodia extension and retraction (Nimmerjahn et al., 2005; Bernier et al., 2019). By this surveillance mechanism, using two-photon microscopy of living murine microglia, Davalos et al. (2005) demonstrate that they detect and act accordingly to damage-associated molecular patterns (DAMPs). Microglia respond to disease conditions through a combination of receptors such as pattern recognition receptors, purinergic and fractalkine receptors, and cytokine receptors (Hickman et al., 2013). Microglia likely responds to hundreds, if not thousands of molecules, many in undefined ways. Certain molecules elicit specific responses, for example, the subsequent activation of purinergic receptors leads to the activation of the phagocytic pathway in rat microglia, that involves the clearance of apoptotic cells, both in vitro and in vivo (Davalos et al., 2005; Haynes et al., 2006; Koizumi et al., 2007; Bernier et al., 2019).

#### Microglia During Remyelination

In the context of remyelination, microglia/macrophages can have many beneficial roles. However, these cells also contribute to autoimmune-related toxicity (Heppner et al., 2005; Ajami et al., 2011; Goldmann et al., 2013; Rothhammer et al., 2018). The dichotomous nature of microglia and macrophages is incompletely understood. One important function of microglia/macrophages following demyelination is the removal of inhibitory myelin debris present in the lesion. This myelin debris has been shown to negatively regulate OPC lineage progression (Syed et al., 2008; Plemel et al., 2013) and consequently, remyelination (Kotter et al., 2006). As microglia are phagocytotic cells, they contribute to myelin debris clearance (Sosa et al., 2013; Rawji et al., 2018). This phagocytosis is reflected by the presence of myelin proteins such as MBP in microglia/macrophages in the white matter following demyelination (Sosa et al., 2013).

Microglia also secrete an array of signaling molecules, including cytokines, chemokines and growth factors; many of these factors signal via receptors on oligodendrocyte lineage cells. For instance, the proinflammatory cytokine tumor necrosis factor-α (TNF-α), secreted by microglia/macrophages as well as astrocytes in mouse cuprizone-induced lesions, promotes OPC expansion and remyelination through TNF receptor 2 (TNFR2) on NG2<sup>+</sup> cells (Arnett et al., 2001; Voss et al., 2012). Interleukin-1β (IL-1β), also secreted by microglia/macrophages and astrocytes, is important for remyelination potentially by controlling oligodendrocyte lineage cell's survival by stimulating IGF-1 secretion from microglia/macrophages in cuprizone-treated mice (Mason et al., 2001). Recently, it was shown microglia can modulate remyelination through the secreted enzyme transglutaminase 2 (TG2). TG2 interacts with OPC-specific Adhesion G Protein-Coupled Receptor G1 (ADGRG1) to promote OPC proliferation. Either, microgliaspecific loss of TG2 or OPC-specific loss of ADGRG1 impairs remyelination in two murine demyelination models (Giera et al., 2018). As microglia/macrophages transition from a more pro-inflammatory state in the days after demyelination to a more immunoregulatory phenotype at a later time point, they secrete Activin-A in mouse lesions (Miron et al., 2013). Activin-A release, in turn, interacts with activin receptors on OPCs, which are required for developmental myelination and remyelination (Dillenburg et al., 2018).

It is important to note that some of the cytokines that have pro-remyelinating roles can be implicated in mediating pathogenesis, including, for example, TNF-α that is required for autoimmune demyelination (Ruddle et al., 1990; Selmaj et al., 1991; Baker et al., 1994; Steeland et al., 2017). In addition, IL-1β from neutrophils and monocyte-derived macrophages drives neuroinflammation in EAE mice (Lévesque et al., 2016; Paré et al., 2018). Indeed, in autoimmune demyelination the removal of a key NF-kB regulator, TAK1, from microglia and CNS macrophages prevents EAE, suggesting an important role of these cells in the execution of autoimmune demyelination (Goldmann et al., 2013).

#### Microglia Response Is Heterogeneous

Recent transcriptomic analyses in both mice and humans revealed that the microglia response to acute demyelination is more heterogeneous than previously thought (Hammond et al., 2019; Masuda et al., 2019). In the resting state, microglia typically adopt a homeostatic phenotype characterized by the expression of such markers as CX3CR1, the purinergic receptor P2RY12, and TMEM119 (Krasemann et al., 2017; Masuda et al., 2019). As microglia become activated, they shift their phenotype. One recently characterized microglia phenotype is the damageassociated microglia (DAM), which are a subset of microglia revealed by single-cell RNA sequencing in Alzheimer's Disease transgenic mice (Keren-Shaul et al., 2017). DAM are linked to neurodegeneration and are characterized by the upregulation TREM2, a phagocytic marker (Deczkowska et al., 2018). Many of these markers associated with DAM are also upregulated in microglia following demyelination such as Lpl, Cst7 and Apoe, suggesting that there is an overlapping response between neurodegeneration and demyelination (Keren-Shaul et al., 2017; Hammond et al., 2019). The transition to a less pro-inflammatory microglia/macrophage phenotype is promoted by the secreted enzyme, interleukin 4 induced 1, which is induced in microglia in response to interleukin 4 (Psachoulia et al., 2016). Recently, it was shown that necroptosis of microglia/macrophages is important in the phenotypic shift during remyelination, suggesting that microglia/macrophages may not transition between phenotypes but instead die, only to be replaced by alternative phenotypes (Lloyd et al., 2019).

# Infiltrating Macrophages During Remyelination

The functional overlap of monocytes-derived macrophages and microglia during remyelination is still unclear. Monocytederived macrophages are mononuclear phagocytic cells of the hematopoeitic stem cell lineage (Jakubzick et al., 2017). Under inflammatory conditions in the CNS, circulating monocytes cross the BBB and mature into macrophages (King et al., 2009; Caravagna et al., 2018). These infiltrating macrophages are evident in both MS lesions and in the different models of demyelination/remyelination such as EAE, LPC and cuprizonetreated animal models (Hiremath et al., 1998; Trebst et al., 2001; King et al., 2009; Miron et al., 2013; Vogel et al., 2013; Kuhlmann et al., 2017). Monocytes that express C-C Motif Chemokine Receptor (CCR2), known as inflammatory monocytes, migrate into the CNS in response to C-C Motif Chemokine Ligand 2 (CCL2). Following cuprizone toxicity or in aged mice following focal demyelination, the vast majority of monocytes require CCR2 for their entry (Ruckh et al., 2012; Lampron et al., 2015). CCR2-knockout mice are resistant to EAE (Fife et al., 2000) and blocking monocyte infiltration prevents EAE in mice (Ajami et al., 2011) suggesting that monocytes are required for autoimmune-mediated demyelination in part by the production of reactive oxygen species (Nikic et al., 2011; Locatelli et al., 2018).

The roles of monocyte-derived macrophages are just beginning to be characterized during remyelination and can be difficult to delineate from microglia as experimentally manipulating these cells while leaving the microglial response intact is challenging. During remyelination, the early use of clodronate liposomes reduces remyelination. Clodronate liposomes are artificial lipid vesicles that are taken up preferentially by phagocytic cells and kill circulating monocytes and macrophages (van Rooijen, 1992; Van Rooijen and Sanders, 1994; Popovich et al., 1999; Kotter et al., 2006; Döring et al., 2015), but they can also ablate microglia (Kumamaru et al., 2012; Han et al., 2019). It is, therefore, still unclear whether clodronate liposomes impair remyelination by killing microglia or macrophages or both. One important correlate of remyelination is the removal of inhibitory myelin debris for which young monocyte-derived macrophages participate (Ruckh et al., 2012). However, it is still unclear whether microglia or macrophages predominate during myelin debris phagocytosis. Taken together, CNS-infiltrating macrophages promote demyelination, but the overlap in functions between microglia and macrophages during remyelination is less clear.

Taken together, microglia and macrophage likely possess overlapping functions that are not completely defined, which as a whole are required for remyelination (**Figure 1**). One critical function is the phagocytosis of inhibitory myelin debris, but other key roles of these cells include secretion of growth factors and the regulation of the immune response through cytokines and other immunoregulatory molecules. The response of these cells is heterogeneous and exciting new technologies such as single-cell RNA sequencing will allow for the characterization of their response during the continuum of remyelination.

#### ROLES OF OTHER INNATE IMMUNE CELLS DURING REMYELINATION

Following demyelination microglia/macrophages predominate, but other innate immune cells could potentially participate in remyelination. The other cells of the innate immune system include natural killer cells, mast cells, eosinophils, basophils, neutrophils, and dendritic cells. Many of these innate immune cells can inflict tissue damage (Mayo et al., 2012; Boutajangout and Wisniewski, 2013; Courties et al., 2014), but little is known about their role in remyelination. In the context of demyelinating diseases, several of these innate immune cells have a described role during demyelination. For instance, neutrophils are important players in disease pathogenesis as mice lacking the key neutrophils chemokine receptor, C-X-C motif chemokine receptor 2 (CXCR2), are resistant to cuprizoneinduced oligodendrocyte cell death and demyelination (Liu et al., 2010). Eosinophils also contribute to demyelination as preventing their infiltration into the CNS in EAE mice decreases disease severity (Gladue et al., 1996). Mast cells, which are found in MS lesions, are also important in disease pathogenesis as blocking their ability to degranulate in EAE rats reduces disease severity (Olsson, 1974; Dimitriadou et al., 2000). To date, less is known about how eosinophils, basophils and neutrophils regulate remyelination. Dendritic cells, which are professional antigen producing cells, may have pro-remyelinating roles (Pusic et al., 2014). Dendritic cells challenged with interferon γ (IFNγ) release exosomes containing microRNA such as miR-219. These dendritic cell-derived exosomes improve remyelination in cultured rat hippocampal slices ex vivo, however, it is still unclear how dendritic cells regulate remyelination in vivo. Given the complex interplay between various innate immune cells and their known involvement in repair of other systems, such as during wound healing (MacLeod and Mansbridge, 2016), there are likely other unknown roles for these cells during remyelination.

# ROLES OF ASTROCYTES IN REMYELINATION

Astrocytes have important roles in both CNS development and homeostasis such as during synaptogenesis, neurotransmission, BBB formation and maintenance, among other roles (Sofroniew and Vinters, 2010; Molofsky and Deneen, 2015). In response to neuroinflammation, astrocytes respond in a process called reactive astrogliosis (Zamanian et al., 2012). Reactive astrocytes have both potentially beneficial and detrimental roles during remyelination; indeed, these functions may relate to a particular phenotype that the astrocytes adopt (Liddelow and Barres, 2017; Liddelow et al., 2017). Astrocytes may promote/inhibit remyelination directly, but could also signal through microglia to promote remyelination. For example, the ablation of reactive astrocytes in cuprizone-treated mice impairs recruitment of microglia to the demyelinating lesion (Skripuletz et al., 2013). Lowered microglia recruitment is associated with reduced clearance of inhibitory myelin debris, which impairs oligodendrocyte maturation (Syed et al., 2008; Plemel et al., 2013) and remyelination (Kotter et al., 2006). The inverse is likely also true, whereby microglia can regulate astrocyte function. Microglia activated by a TLR4 agonist, LPS, induces a newly identified neurotoxic phenotype of astrocytes. The neurotoxic astrocyte may be detrimental for OPC maturation as it has been shown to induce oligodendrocyte and neuronal apoptosis (Liddelow et al., 2017). As microglia/macrophages display a temporal shift from a predominantly pro-inflammatory phenotype to an

anti-inflammatory phenotype (Miron et al., 2013), it is unknown whether the astrocytes within a demyelinating lesion adopt a similar temporal shift in phenotype and whether this may in turn yield a spectrum of beneficial or detrimental functions for remyelination.

One role of astrocytes is mediated by endothelin-1, a peptide that has been characterized to potently induce vasoconstriction. Following demyelination in animal models and in MS patients' lesions, astrocytes begin expressing endothelin-1 (ET-1; Hammond et al., 2014). ET-1 is secreted by astrocytes and endothelial cells and acts in paracrine and autocrine manners in the CNS (Hostenbach et al., 2016). ET-1 can activate receptors on vascular smooth muscle cells, inducing vasoconstriction and leading to cerebral hypoperfusion, which is commonly observed in MS patients (Law et al., 2004; Varga et al., 2009). Low oxygen tension, at least during development in mice, also impairs OPC maturation (Yuen et al., 2014). Following acute demyelination, astrocytes respond to ET-1 via their endothelin receptor B by upregulating the Notch1 receptor ligand, jagged1 (Hammond et al., 2014, 2015). Astrocyte-induced Notch1 activation is known to inhibit OPC differentiation and remyelination. Interestingly, this astrocyte jagged1-Notch1 interaction resolves to allow OPC differentiation after a period of OPC expansion. Astrocytes could, therefore, be delaying OPC differentiation to allow sufficient proliferation of OPCs (Zhang et al., 2009; Hammond et al., 2014). This notch-mediated interaction may be relevant to MS considering a histological analysis of lesions shows that astrocytes express jagged1 and OPCs express Notch1 (John et al., 2002). This is important as it is thought that a differentiation and maturation failure of OPCs is a significant contributor to remyelination failure in MS lesions (Kuhlmann et al., 2008; Duncan et al., 2017).

# Astrocytes as Regulators of the Extracellular Matrix (ECM)

Astrocytes are also major modulators of the extracellular matrix (ECM). The ECM is composed of several molecules which can influence OPC function (Pu et al., 2018). Astrocytes are major producers of the ECM molecules high molecular weight hyaluronan (Back et al., 2005), Tenascin-C (Gutowski et al., 1999), fibronectin (Stoffels et al., 2013), and various members of the chondroitin sulfate proteoglycan family (Jones et al., 2003; Siebert et al., 2014). In addition to providing chemical signals, the mechanical stiffness of the ECM significantly influences the proliferation and differentiation of OPCs in vivo. It was recently shown that aging rodents have a stiffer ECM, resulting in decreased OPC proliferation and differentiation (Segel et al., 2019). When the mechanosensitive ion channel PIEZO1 is inhibited, OPCs are able to restore their loss of function in proliferation and differentiation. In addition to inhibiting OPC maturation and axon regeneration, these molecules exert immunomodulatory effects on several members of the immune system (Stephenson et al., 2018). For example, when chondroitin sulfate proteoglycans are added to murine macrophages in vitro, these cells significantly upregulate their production of pro-inflammatory cytokines, matrix metalloproteinases, as well as displaying an increase in migration (Stephenson et al., 2018). In this study, these ECM molecules were observed around perivascular cuffs in the EAE model. Pharmacologically inhibiting the synthesis of these molecules in this model resulted in reduced immune cell infiltration and decreased clinical severity (Stephenson et al., 2019). Although not yet addressed, the production of such inhibitory molecules may be influenced by the phenotypic state adopted by astrocytes.

#### Astrocyte Role in Lipid Metabolism

Astrocytes play a central role in lipid metabolism. As the BBB limits the entry of several lipoproteins from the peripheral circulation, the CNS has developed a tightly regulated system in which to regulate the production of lipid species such as arachidonic acid, docosahexanoic acid, and cholesterol (Moore, 2001; Dietschy, 2009). Astrocytes are the major producers of cholesterol and the cholesterol carrier, apolipoprotein E (APOE) under homeostatic conditions (Boyles et al., 1985). Cholesterol is a major constituent of cell membranes and myelin and is therefore critical in normal CNS functioning (Linton et al., 1991). Rodent neurons do not readily synthesize cholesterol and depend on astrocytes for their source of cholesterol (Nieweg et al., 2009). As astrocyte-derived cholesterol is important for normal neuronal functioning, it is likely that transport of cholesterol to oligodendrocytes from astrocytes or other glial cells such as microglia is important in myelin synthesis. Indeed, it was recently shown that demyelinated white matter lesions from aging mice have deficient reverse cholesterol transport that was associated with reduced remyelination (Cantuti-Castelvetri et al., 2018). Cantuti-Castelvetri et al. (2018) found that stimulating reverse cholesterol transport enhanced remyelination in the aging CNS, highlighting the role of lipid metabolism in myelin synthesis. Whether cholesterol exported by astrocytes and macrophages/microglia can be directly taken up by OPCs responding to a demyelinated lesion was not addressed in this study and is yet to be determined. This area is of direct importance as several small molecules capable of stimulating oligodendrogenesis act on the cholesterol biosynthesis pathways (Hubler et al., 2018). Importantly GFAP-expressing reactive astrocytes in EAE mice have decreased cholesterol synthesis, which may limit remyelination efficiency (Itoh et al., 2018), suggesting that autoimmunity may impair cholesterol synthesis.

Medications used to treat hypercholesterolemia (statins) seem to show some benefit in secondary progressive MS (Chataway et al., 2014). High-dose statin treatment resulted in a 43% reduction in brain atrophy compared to placebo. As statins have effects on the immune system, it is unclear whether the therapeutic efficacy of statins in secondary progressive MS are mediated through cholesterol modulation or other indirect mechanisms (Eshaghi et al., 2019). It is important to note that statins did not influence relapse risk, disease progression or disability scores in people with MS in combination with interferon therapy (Bhardwaj et al., 2012).

#### Astrocytes Regulate CNS Schwann Cell Myelin and Secrete Growth Factors

Astrocytes are thought to regulate the degree to which remyelination is favored by OPC-derived oligodendrocytes or SCs (Zawadzka et al., 2010; Monteiro de Castro et al., 2015). In the absence of astrocytes, more SC remyelination is observed. The mechanism by which astrocytes regulate the fate choice of OPCs is unclear but may be due to interactions in BMP/Wnt signaling within the lesion microenvironment (Ulanska-Poutanen et al., 2018). Although SCs remyelination appears to re-establish conduction capacity within CNS axons (Smith et al., 1979; Blight and Young, 1989), it is not yet clear whether SCs differ from oligodendrocytes in the provision of metabolic support to axons. Similarly, it is unclear how different myelin sheath thicknesses between oligodendrocytes and SCs affect the timing of axonal conduction.

Astrocyte secretion of neurotrophins, a class of proteins that induce the growth and survival of neuronal cells, are involved in remyelination. For example, astrocytes can regulate myelin protein synthesis by the release of brain-derived neurotrophic factor (BDNF; Fulmer et al., 2014). In the cuprizone model of demyelination, the activation of OPC-expressed BDNF receptor TrkB promotes remyelination as reflected by an increase in OPC differentiation, the number of remyelinated axons and myelin sheath thickness (Fletcher et al., 2018). Furthermore, astrocytes secrete the mitogen platelet-derived growth factor-A (PDGF-A) to act on OPCs, which can promote proliferation (Wolswijk and Noble, 1992; Redwine and Armstrong, 1998; Frost et al., 2003). Lesions from MS also display PDGFRα expression on proliferating cells (Maeda et al., 2001). Astrocytes are also thought to produce leukemia inhibitory factor, ciliary neurotrophic factor, and insulin-like growth factor-1, all of which have been implicated in supporting OPC maturation (Moore et al., 2011).

Taken together, the role of astrocytes during remyelination is complex and incompletely understood (**Figure 2**). Release of growth factors and lipid metabolism is required for remyelination, yet astrocytes also produce several inhibitory ECM molecules that impair remyelination. The expression of notch ligands by astrocytes also serves as a break for remyelination. How astrocytes are regulated through the continuum of remyelination is yet to be defined, but may provide an explanation for their diverse roles during remyelination.

#### NON-MYELINATING ROLES OF OLIGODENDROCYTE LINEAGE CELLS DURING REMYELINATION

The non-myelinating roles of oligodendrocyte lineage cells are being increasingly realized and may provide new avenues for therapeutic intervention. During development and into adulthood, OPC differentiation is restricted to forming mature oligodendrocytes (Kang et al., 2010; Young et al., 2013; Huang et al., 2019) but during traumatic injury or demyelination OPCs act as multipotent progenitor cells (Richardson et al., 2011; Crawford et al., 2014) capable of differentiating into other neural lineage cells such as astrocytes and SCs (Zawadzka et al., 2010; Assinck et al., 2017b; Hackett et al., 2018; Huang et al., 2018). Beyond differentiating into other lineages, OPCs respond to disruptions in tissue homeostasis in a manner reminiscent of microglia. For example, OPCs proliferate and extend their processes to surround local injury sites (Hughes et al., 2013), albeit the kinetics of this process extension are slower than microglia (Davalos et al., 2005; Nimmerjahn et al., 2005). OPCs also migrate to areas of injury and proliferate to compensate for the loss of adjacent OPCs, again reminiscent of microglia. The high proliferative capacity and robust response to tissue injury suggest OPCs themselves may be major regulators of repair. Indeed a number of studies now indicate that the response of OPCs is critical for modulating inflammation, altering glial scarring and potentially regulating angiogenesis. The subsequent sections will review the evidence that OPCs are subverted to directly perform additional functions in CNS repair with an emphasis on how modulating these functions may offer new therapeutic strategies.

#### Inflammatory Nature of Oligodendrocyte Progenitor Cells

The rapid proliferation and migration to areas of tissue damage leave OPCs ideally situated to regulate subsequent immune responses. OPCs express many microglia-enriched genes but often at lower levels. For example, the homeostatic microglia marker CX3CR1 is expressed by OPCs (Voronova et al., 2017). Accordingly, OPCs express a number of critical modulators of the inflammatory response including IL-33 and the low-affinity Fc receptor (Fcgr2b) during EAE (Falcão et al., 2018), the latter of which is normally only expressed by microglia (Zhang et al., 2014). Transcriptomic analyses of OPCs reveal an increase in expression of the inflammasome-associated cytokines IL-1β and the chemokine CCL2 during demyelination (Moyon et al., 2015). It should be noted, however, that OPCs were isolated by Moyon et al. (2015) based on PDGFRα expression, which is also enriched in a population of pericytes (Assinck et al., 2017b), so these results should be interpreted with caution. However, recent single-cell RNA sequencing of OPCs during EAE (Falcão et al., 2018) and in MS (Jakubzick et al., 2017) confirms that these cells express a number of proinflammatory genes typically restricted to microglia/macrophages (Butovsky et al., 2014; Zhang et al., 2014). For example in response to IFNγ, OPCs go on to express the antigen presentation molecules MHC-I/II (Falcão et al., 2018). OPCs use MHCI/II to activate T-cells both in vivo and in vitro (Falcão et al., 2018; Kirby et al., 2018), and the capacity to cross-present antigens to CD8 T-cells was confirmed in vivo (Kirby et al., 2018). Taken together, inflammatory demyelination triggers OPCs to express pro-inflammatory cytokines and subsequently present antigens to T-cells.

Given that OPCs can perpetuate an inflammatory response raises the possibility that they might directly induce damage in autoimmune disease. Indeed, the induction of pro-inflammatory genes in OPCs during autoimmune demyelination may be necessary for subsequent demyelination. Blocking the activation of NF-κB by deleting Act1 in NG2+ glia blocks the development of EAE-induced demyelination and diminishes inflammatory gene expression following the adoptive transfer of MOG35–55 Th-17 cells (Kang et al., 2013). Thus, the co-option of OPCs into a pro-inflammatory phenotype during Th-17-mediated autoimmune disease likely drives subsequent tissue damage. Additionally, the co-opting of OPCs to perpetuate an inflammatory response may leave them unable to adequately differentiate into new remyelinating oligodendrocytes, a possibility that remains to be proven. Indeed, the adoptive transfer of autoimmune T-cells (Baxi et al., 2015) or engraftment of lymphocytes from people with MS into mice with LPC-induced lesions (El Behi et al., 2017) slows remyelination, which is consistent with OPC immune activities impairing their differentiation. Likewise, following cuprizone demyelination, the transfer of effector T cells or IFNγ impairs oligodendrocyte differentiation and is associated with the increased presentation of antigens by OPCs to CD8+ T cells (Kirby et al., 2018).

This co-option of OPCs into an inflammatory phenotype during autoimmune demyelination raises the intriguing possibility that the pro-inflammatory OPC phenotype likely drives tissue damage and may impair effective remyelination. Current FDA-approved therapeutics for MS target autoimmune damage in the CNS, and as such, they may have an unexpected benefit to diminish the co-option of OPCs, therefore, ''freeing'' them so that they can effectively remyelinate. In accordance, both the switch of microglia to a less inflammatory phenotype (Miron et al., 2013) and the transfer of regulatory T-cells promote remyelination following chemical demyelination (Dombrowski et al., 2017).

# Oligodendrocyte Lineage Cells During Glial Scarring

When the extent of trauma or tissue injury necessitates glial scar formation, OPCs' role is to modify other cells that contribute to scar formation such as astrocytes or pericytes

(Hackett et al., 2016; Huang et al., 2018) and in some cases directly such as astrocytes (Rodriguez et al., 2014; Hesp et al., 2018). With a milder injury, OPCs have a much-reduced capacity to differentiate into astrocytes (Zawadzka et al., 2010). If proliferating NG2+ cells—which are composed of OPCs, pericytes, SCs and activated microglia/macrophage following traumatic SCI (McTigue et al., 2006)—are depleted, glial scar formation is dramatically impaired (Hesp et al., 2018). The result of NG2+ cell depletion is prolonged hemorrhage, larger lesions, and more pronounced edema (Hesp et al., 2018). Taken together, the proliferation and activation of NG2+ glia, which prominently includes OPCs, is necessary for proper formation of the glial scar that restricts ongoing secondary damage to tissue following traumatic injury. Given that scar formation is associated with ECM molecules inhibitory to remyelination, this raises the intriguing possibility that OPCs indirectly regulate ECM molecules known to inhibit remyelination.

#### Oligodendrocyte Lineage Cells Interact With and Regulate the Vasculature

During development, OPCs migrate along blood vessels in order to successfully distribute throughout the parenchyma, a step that is necessary for their subsequent differentiation into myelinating oligodendrocytes (Tsai et al., 2016). The secretion of stromal cell-derived factor 1 (SDF1) by endothelial cells attracts OPCs by binding to CXCR4, which is abundantly expressed on OPCs when Wnt-tone is high (Tsai et al., 2016). However, signaling between OPCs and endothelial cells is bidirectional, and given OPCs' location along blood vessels in development (Tsai et al., 2016), they are ideally positioned to sense oxygen levels and regulate angiogenesis. OPCs sense oxygen tension within the parenchyma via HIF1/2α (Yuen et al., 2014), a protein sensor that is stabilized under hypoxic conditions (Wang et al., 1995; Majmundar et al., 2010). Experimentally stabilizing HIF in OPCs—and therefore mimicking hypoxia for these cells—promotes angiogenesis in both the cortex and corpus callosum likely through the secretion of soluble factors (Yuen et al., 2014). A mechanism emerges in which endothelial cells are necessary for the initial migration of OPCs throughout the parenchyma, which in turn, drive angiogenesis until oxygen tension is sufficient to support proper oligodendrocyte differentiation and myelination (Yuen et al., 2014; Tsai et al., 2016). Paracrine signaling between OPCs and endothelial cells is therefore critical for proper developmental myelination.

Interestingly, recent studies suggest the angiogenic activities of OPCs can be subverted under pathological conditions. In development, heightened OPC clustering along blood vessels disrupts BBB function resulting in lymphocyte trafficking in the CNS parenchyma. Like development (Tsai et al., 2016), OPCs migrate along blood vessels during remyelination (Niu et al., 2019). In active MS lesions, OPCs become trapped and cluster along blood vessels. It is unclear if this heightened clustering is due to local hypoxia within MS lesions or inflammation. Cytokines like IL-1β stimulates OPCs to secrete factors that promote angiogenesis in vitro, suggesting inflammation may also regulate OPC-dependent angiogenesis. Excessive Wnt signaling in OPCs not only induced clustering of OPCs but also increased the expression of Wif1, which suppresses tight junction formation in endothelial cells (Niu et al., 2019). Indeed hypoperfusion are present in both the NAWM and NAGM in MS potentially inducing a hypoxic environment (Law et al., 2004; Varga et al., 2009). Inflammation may also subvert OPCs to associate with blood vessels and block their ability to fully differentiate in MS. Given that excessive interactions between OPCs and blood vessels during development are sufficient to trigger pathological immunity, it will be interesting to examine if this interaction promotes pathology in models of inflammatory demyelination like EAE.

Taken together, OPCs are increasingly being recognized for their roles other than differentiating into oligodendrocytes during remyelination (**Figure 3**). OPCs morphologically respond to injury in a way comparable to microglia, they secrete cytokines, and they regulate scar formation. In these capacities OPCs are an alternative CNS-innate immune cell. OPCs are also responsive to oxygen levels and are important regulators of angiogenesis.

# REMYELINATION DECLINES WITH AGE

Remyelination, as with any regenerative process, declines in efficiency with age (Shields et al., 1999; Sim et al., 2002; Ruckh et al., 2012). OPC proliferation and differentiation, which are necessary processes for remyelination, are reduced with age (Sim et al., 2002). In fact, dorsally-derived OPCs in the spinal cord are the predominant source of remyelination in young mice, but contribute less in aged animals (Crawford et al., 2016b). Age can intrinsically alter OPCs directly as old OPCs display a different global DNA methylation profile that alters and/or diminishes their response to pro-remyelinating factors (Zhou et al., 2019). Aging also decreases the recruitment of histone deacetylases (HDACs) in OPCs—epigenetic regulators of gene expression—that leads to a decrease in remyelination efficiency. In OPCs, HDACs downregulate the expression of many transcription factors that inhibit OPC differentiation, including Hes5 and Sox2. They, therefore, regulate remyelination by controlling OPC differentiation (Liu et al., 2006; Shen et al., 2008).

Age-impaired remyelination is also due to impaired innate immune cell function. The removal of inhibitory myelin debris by microglia/macrophages is impaired by aging (Natrajan et al., 2015; Safaiyan et al., 2016; Cantuti-Castelvetri et al., 2018; Rawji et al., 2018). In LPC-induced lesions, aging microglia/macrophages display a decrease in microenvironment surveillance activity and a decrease in phagocytic activity (Rawji et al., 2018). Furthermore, aging microglia/macrophage lysosomal degradation of myelin is diminished as reflected by the accumulation of myelin proteins in microglial lysosomes of aging mice (Safaiyan et al., 2016). This accumulation of myelin debris inside of microglia could be due to a deficiency in reverse cholesterol transport as remyelination efficiency increased when mice with LPC-induced focal demyelination were treated with HβCD, a drug that stimulates cholesterol efflux from cells (Cantuti-Castelvetri et al., 2018). These findings raise the possibility that targeting the phagocytic pathway of microglia/macrophages could improve the efficiency of the myelin debris clearance.

Why myelin debris clearance declines are likely multifaceted? Recently, a CRISPR-Cas9 knockout screen identified CD22 as a negative regulator of microglia phagocytosis that increases with age (Pluvinage et al., 2019). Many factors, such as Retinoid-X-Receptor-α (RXR-α), fractalkine receptor CX3CR1 and Galectin-3 (Gal-3) can also influence the ability of microglia to phagocytose (Lampron et al., 2015; Natrajan et al., 2015; Reichert and Rotshenker, 2019). RXR-α is a ligand-activated transcription factor that controls a variety of genetic programs, including immune cell-related functions (Dawson and Xia, 2012). Natrajan et al. (2015) demonstrate that RXR-α knockout from young macrophages in mice LPC-induced lesion delays myelin debris clearance and remyelination, and also that stimulating RXR-α in monocytes from MS patients in vitro improves this normally inefficient myelin debris phagocytosis. There is now an ongoing clinical trial using an RXR-α agonist (EudraCT number: 2014-003145-99). Similarly, the chemokine receptor CX3CR1—enriched in microglia—is critical for myelin debris clearance following cuprizone-induced demyelination (Lampron et al., 2015). Ruckh et al. (2012) found that in the LPC-induced demyelination model, the systemic circulatory system of young mice can restore the efficiency of OPC proliferation, differentiation, and remyelination in aged mice to levels near those observed in young mice. They were able to show these differences by joining the circulatory systems of young and old mice in pairs through heterochronic parabiosis. This rejuvenation of remyelination suggests that reversing the peripheral immune system can partially restore

age-dependent decline in remyelination potential. In fact, several human trials such as the Stanford Parkinson's Disease Plasma Study (NCT02968433) infuse plasma from young people to diseased patients. However, a proper evaluation of the blood-borne rejuvenating factors is needed before any similar trial can be conducted for demyelinating diseases such as MS.

Peripheral changes during aging can also impart effects onto innate immune cells such as microglia. For example, peripheral blood from aged mice is sufficient to lower the levels of type II interferon in the choroid plexus, which is linked to cognitive decline, a common symptom observed in MS patients (Franklin et al., 1989; Baruch et al., 2014). Microglia also have interferon phenotypic changes during aging, in this case, driven in part by higher levels of IFN-I cytokines (Deczkowska et al., 2017) that likely come from peripheral sources. Furthermore, retro-orbital injection of plasma from aging mice to young mice is sufficient to activate microglia (Yousef et al., 2019). Peripheral immune changes are therefore important in regulating the innate immune system in the CNS.

# CONCLUSION

Remyelination involves a complex interplay between the oligodendrocyte lineage cells that produce the myelin, the astrocytes and the other innate immune cells that regulate the microenvironment of remyelination. Microglia and CNS-infiltrating macrophages regulate the lesion environment and make it suitable for remyelination by activities such as the removal of inhibitory myelin debris. Astrocytes secrete growth factors and provide key lipids species, but also produce inhibitory ECM molecules and the Notch ligand Jagged. OPCs are the key source of remyelinating oligodendrocytes but are now recognized to also regulate angiogenesis, respond to injury and secrete cytokines suggesting they may also be a brain resident immune cell. Given that boosting remyelination spares venerable axons (Mei et al., 2016), more work is required to examine how these cells interact with one another during remyelination.

#### AUTHOR CONTRIBUTIONS

CB, KR, GD and JP drafted and reviewed this manuscript. MH and CB constructed the figures. JP supervised the drafting of this manuscript.

#### REFERENCES


#### FUNDING

KR is supported by the MS Society of Canada. GD is supported by a National Multiple Sclerosis Society fellowship. This work was funded by the University of Alberta, the University Hospital Foundation Medical Research Competition, the Azrieli Foundation and Brain Canada, through the Canada Brain Research fund, with support of Health Canada and the Funding Partner(s).


levels, induces astrocyte activation and damages blood vessel integrity. Mol. Neurobiol. 56, 6184–6196. doi: 10.1007/s12035-019-1502-9


glia to axons: evidence from animal models and MS patient brains. Mol. Biol. Rep. 43, 495–507. doi: 10.1007/s11033-016-3990-2


experimental autoimmune encephalomyelitis model of demyelinating disease. J Neurosci 30, 16383–16390. doi: 10.1523/jneurosci.3411-10.2010


nanoruptures drives axon degeneration in a model of multiple sclerosis. Neuron 101, 615.e5–624.e5. doi: 10.1016/j.neuron.2018.12.023


**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 Baaklini, Rawji, Duncan, Ho and Plemel. 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.

# Mimicry of Central-Peripheral Immunity in Alzheimer's Disease and Discovery of Neurodegenerative Roles in Neutrophil

Joseph Park <sup>1</sup> , Sung Hoon Baik <sup>2</sup> , Inhee Mook-Jung<sup>2</sup> , Daniel Irimia<sup>3</sup> \* and Hansang Cho1,3,4 \*

*<sup>1</sup> The Nanoscale Science Program, Department of Mechanical Engineering and Engineering Science, Department of Biological Sciences, Center for Biomedical Engineering and Science, University of North Carolina at Charlotte, Charlotte, NC, United States, <sup>2</sup> Department of Biochemistry and Biomedical Sciences, Seoul National University, Seoul, South Korea, <sup>3</sup> Department of Surgery, BioMEMS Resource Center, Harvard Medical School, Massachusetts General Hospital, Charlestown, MA, United States, <sup>4</sup> Department of Biophysics, Institute of Quantum Biology, Sungkyunkwan University, Suwon, South Korea*

#### Edited by:

*Antje Kroner, Medical College of Wisconsin, United States*

#### Reviewed by:

*Andre Ortlieb Guerreiro Cacais, Karolinska Institute (KI), Sweden Yoshiro Ohara, Kanazawa Medical University, Japan*

#### \*Correspondence:

*Daniel Irimia daniel\_irimia@hms.harvard.edu Hansang Cho h.cho@uncc.edu*

#### Specialty section:

*This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Immunology*

Received: *24 June 2019* Accepted: *03 September 2019* Published: *25 September 2019*

#### Citation:

*Park J, Baik SH, Mook-Jung I, Irimia D and Cho H (2019) Mimicry of Central-Peripheral Immunity in Alzheimer's Disease and Discovery of Neurodegenerative Roles in Neutrophil. Front. Immunol. 10:2231. doi: 10.3389/fimmu.2019.02231* Neuroinflammatory roles of central innate immunity in brain parenchyma are well-regarded in the progression of neurodegenerative disorders including Alzheimer's disease (AD), however, the roles of peripheral immunity in central nervous system (CNS) diseases are less clear. Here, we created a microfluidic environment of human AD brains: microglial neuroinflammation induced by soluble amyloid-beta (Abeta), a signature molecule in AD and employed the environment to investigate the roles of neutrophils through the central-peripheral innate immunity crosstalk. We observed that soluble Abeta-activated human microglial cells produced chemoattractants for neutrophils including IL6, IL8, CCL2, CCL3/4, CCL5 and consequently induced reliable recruitment of human neutrophils. Particularly, we validated the discernable chemo-attractive roles of IL6, IL8, and CCL2 for neutrophils by interrupting the recruitment with neutralizing antibodies. Upon recruitment, microglia-neutrophils interaction results in the production of inflammatory mediators such as MIF and IL2, which are known to up-regulate neuroinflammation in AD. We envision that targeting the crosstalk between central-peripheral immune community is a potential strategy to reduce immunological burdens in other neuroinflammatory CNS diseases.

Keywords: neuroinflammation, neurodegeneration, Alzheimer's disease, microglia, neutrophil, chemotaxis, cellular interaction, microfluidics

# INTRODUCTION

Alzheimer's disease (AD) is the most common form of dementia. It affects more than 35 million people worldwide (1, 2). It is characterized by neuropathological features including amyloid-beta (Abeta) plaques, neurofibrillary tangles, and neuroinflammation that lead into synaptic dysfunction and a progressive deterioration of cognitive functions (2–5). A growing body of evidence suggests that neuroinflammation is the critical hallmark of AD and central innate immune cells (microglia and astrocytes) have the major roles of neuroinflammation in a central nervous system (CNS) (6–9). Particularly, microglia, a macrophage resident in brains, produce both beneficial neuroinflammatory responses and detrimental neurotoxic effects in the progression of AD (10–12). Therefore, the regulation of microglial activation has been proposed as a potential therapeutic approach in AD treatment (13).

In contrast to the field's increasing understanding of the neuroinflammatory roles of central innate immunity in AD, comparatively little is known about their interactions with peripheral immunity. Peripheral immune cells including Tcells, B-cells, monocytes, and neutrophils, have been found in the brains of AD human patients and corresponding AD animal models (14–17). Among them, neutrophils are the body's most abundant and frontline immune cells, respond with both central and peripheral immune effectors during the progression of AD (17). The migration of neutrophils is initiated by chemokine-triggered activation. Neutrophils actively crawl through the endothelium and accumulate in tissues where they released neutrophil extracellular traps (NETs) and IL-17 which are markers of sterile inflammation and toxic cytokines, respectively. Neutrophil depletion or inhibition of trafficking via LFA-1 blockade reduced Alzheimer's disease– like neuropathology and improved memory in mice showing cognitive dysfunction in transgenic Alzheimer's disease mouse models (18). However, the mechanism study of neutrophil recruitment and induction of neuropathological changes of AD still remains unclear. The challenge of the study is due to, in part, the limited readouts in vivo, the similar profiles of peripheral inflammatory roles compared to the central immunity by microglia and astrocytes, and inconsistent observation and/or variation in AD animal models. Recently, microfluidic organ-onchips demonstrated successfully to provide feasible, regulated, accessible microenvironments to study multicellular interaction and/or multiorgan behaviors in quantitative, reproducible, and large-scaled manners (19).

Here, we present a microfluidic system mimicking centralperipheral innate immunity in AD for the first time. We identified IL6, IL8, CCL2, CCL3, and CCL5 as mediators released by soluble Abeta-stimulated microglia. We validated their contribution to neutrophil recruitment by inhibiting with neutralizing antibodies. Moreover, the recruited neutrophils induce the secretion of inflammatory factors such as MIF and IL2, which may further stimulate the glia cells. Taken together, these studies suggest that central innate immune cells have a critical role in recruiting peripheral immune cells and mitigating AD pathogenesis.

#### RESULTS

#### Replicated AD Microenvironment by Using a Chemotactic Microfluidic Platform

We established an in vitro system to study the crosstalk between microglia and neutrophils in the context of AD (**Figure 1**). The system consists of two compartments: central circular and surrounding annular compartments representing regions of cerebral tissues and blood streams, respectively. The two compartments are connected with microchannels (10 × 5 × 500 µm<sup>3</sup> in width, height, and length) representing a mechanical barrier between two regions, which only chemoattracted neutrophils can penetrate through. The long and thin migration channels serve as mechanical constraints to avoid spontaneous entrance of inactivated neutrophils and activated microglia. Firstly, we cultured human adult microglia cells in the central compartment in addition of soluble Abeta for 3 days to induce the release of microglial inflammatory mediators in the context of AD. Later, we plated freshly isolated human neutrophils on the annular compartment under effective gradients of microglial soluble factors from the central compartment. We tested 200 independent conditions on eight arrays of twenty-five devices in single well plates and measured neutrophil mobility at a single cell resolution in a real-time.

#### Assessment of Microglial Neuroinflammation Activated by Soluble Abeta

We assessed the status of microglial inflammation by measuring inflammatory cytokines. We found that proinflammatory mediators, CCL2, IL-6, and IL-8 increase significantly by 1.4-, 1.9-, and 2.3-folds, respectively and chemokines, e.g., CCL3/4, CCL5 released from activated microglia only under stimulation of Abeta (**Figures 2A,B**). Secretion of anti-inflammatory markers, such as IL-1RA, IL-4, IL-10, and TGF-ß were below the limit of detection. These data collectively suggest that soluble Abeta induces microglial proinflammation quantified by using our microfluidic platform.

To identify the microglial activation, Abeta-treated microglial cells were monitored for three days using a time-lapse imaging microscopy. At the early 24 h, microglia remained as branched filopodia in all directions, morphological features typically associated with "resting" microglia (**Figure 2C**, upper, −Abeta). At 48 h after cell seeding and Abeta incubation, microglia cells became larger (**Figure 2C**, lower, +Abeta) and transformed into an amoeboid shape while the length and area of microglial somata increased in all directions (**Figures 2C,D**). In addition, microglia treated with soluble Abeta showed up-regulation of microglial activation markers such as CD11b and CD68 (**Figures 2C,E**, **Supplementary Figure 1**).

#### Validation of Microfluidic Chemotaxis for Neutrophils

We induced gradients of chemoattractants for neutrophils from the central toward the annular compartments and validated the design of microchannels effective for measuring chemotactic activities of neutrophils. We defined a neutrophil recruitment index, R.I. (**Figure 3A**, **Supplementary Figure 2**), representing the fraction of neutrophils recruited to the central compartment by chemoattractants. We assessed the R.I. values under various conditions (fMLP: N-formyl-methionyl-leucylphenalanine, Abeta, and culturing medium) and observed discernable neutrophil recruitment only in addition of fMLP to the central compartment (**Figures 3A,B**). In this study, we used fMLP for a positive control as fMLP triggers p38 pathway overriding PI(3)K and is certain for effectiveness of neutrophil chemoattraction (20). However, neutrophil recruitment was not discernable with Abeta only and the medium during the observation for 6 h. The results successfully demonstrate that our platform can engage neutrophils with chemical cues actively while avoiding spontaneous entrance of neutrophils with the mechanical barriers. Neutrophil recruitment reached

the peak activity at 18 ± 5 µm/min within the first 6 h (**Supplementary Figure 3**) and the activity was proportional to the concentration of fMLP, such as 1.6 times faster at 10 nM fMLP compared to 1 nM fMLP (**Figure 3B**).

We further analyzed the migratory activities of human neutrophils in a single cellular level. We found that 88.5 ± 3.8% of active neutrophils migrated forwards and 11.5 ± 1.7% migrated backward 10 nM fMLP, whereas 82.2 ± 3.4% migrated forward and 17.8 ± 2.8% backward 1 nM fMLP showing the persistent migration (**Figures 3C,D**). However, too much fMLP of 1µM resulted in a 63 % reduction in the total fraction of migrating cells compared to 10 nM (**Supplementary Figure 4**).

#### Inducement of Neutrophil Recruitment by A-Beta Activated Microglia

To probe the crosstalk between microglia and neutrophils in the context of AD, we cultured microglia cells in the central compartment at various cell numbers (5,000, 10,000, 20,000 cells per device) with the treatment of soluble Abeta42 (22 nM) for 48 h, and then introduced neutrophils to the surrounding compartment. Microglia cells were pre-labeled with a red membrane dye to monitor their morphogenesis. Microglia were incubated alone for 24 h with reduced 1% FBS PrigrowIII medium without Abeta. Followed by microglial incubation with soluble Abeta, human neutrophil labeled with a green membrane dye were plated on the annular compartment. Neutrophil immediately started to migrate toward the central compartment of microglia of 5,000 with Abeta (**Figure 4A**). With the increase of microglia in the central compartment, neutrophil migration became more discernable and the number of recruited neutrophils increases proportional to the microglia, presuming that A-beta activated microglia release neutrophil chemoattractants. The responding neutrophils showed the enhanced polarity and distinct trailing edge toward microglial compartment with Abeta (**Figure 4A**, **Supplementary Figure 5**, **Supplementary Videos 1, 2**). However, neutrophil migration and the basal cell polarity were not noticeable in the conditions of microglia only, Abeta only, nor culture medium only (**Figure 4B**, **Supplementary Video 3**).

To confirm neutrophil activation, neutrophils co-cultured with microglial cells with or without Abeta for 5 min, were thoroughly washed away and the numbers of adherent neutrophils were counted. It showed that neutrophils cocultured with Abeta-activated microglia enhanced cellular adhesion to the surface, confirming the neutrophil activation (**Supplementary Figure 6**).

amoeboid shape (activated). A microglia activation marker of CD11b (green) was upregulated with Abeta stimulation. Both the membrane area (D) and the amounts of expressed CD11b (E) increase in the activated microglia with Abeta. ndevice = 3, ncell = 150. Scale bars: 30µm. All parameters are presented as mean ± SEM.

#### Identification of Microglial Chemoattractants That Induce Neutrophil Recruitment

We investigated which of microglial soluble factors are responsible for neutrophil recruitment. Firstly, we measured multiple cytokines in conditional media of Abeta-activated microglia and tested neutrophil chemotaxis for selected cytokines: IL-6, IL-8, CCL2, CCL3/4, and CCL5 (**Figure 4C**). We compared the recruitment activity by counting the neutrophil R.I. and found the significant recruitment by IL8 at 10 nM (R.I. = 12 ± 2.2), IL6 at 12 nM (R.I. = 8.7 ± 1.1) and discernable recruitment by CCL2 (R.I. = 5.1 ± 0.72), CCL3 (R.I. = 4.9 ± 0.51), CCL5 (R.I. = 4.2 ± 0.31) at 10 nM, respectively.

We also assessed the persistent movement of neutrophils by utilizing a single-cellular, high-resolution, and spatiotemporal imaging microscope. Neutrophils showed the highest persistent migration (∼92%) toward Abeta-stimulated microglial cells, compared to single soluble factors of IL8, IL6, CCL2, CCL3, and CCL5 (79.2, 75.2, 73.1, 64.2, and 51.2%, respectively) (**Figure 4D**). Such the high persistency was consistent with the variation in microglial cell numbers.

#### Validation of Microglial Chemoattractants That Induce Neutrophil Recruitment

To confirm the contribution of microglia-derived cytokines to neutrophil recruitment, we inhibited neutrophil migration toward Abeta-activated microglia by adding individual neutralizers for CCL2, CCL3/4, CCL5, IL6, and IL8 to the central compartment. We observed the discernable reduction of neutrophil recruitment by neutralizing antibodies against IL8, IL6, and CCL2 (**Figure 4E**). As reported in literature, IL8 appears to be the most dominant chemoattractant among others. IL8 neutralization reduced neutrophil recruitment most significantly by 50.2%, while IL6 (34.1%), CCL2 (22.1%), CCL3 (9.1%), and CCL5 (7.9%) neutralizers showed moderate reduction of neutrophil recruitment, respectively (**Figure 4E**).

behavior of neutrophils responding to fMLP (10 nM, 1 nM), Abeta (22 nM), and a negative control (medium). (D) Measuring the migration persistence for the first 1 h shows the dominancy of the forward migration of neutrophils. Combined data from ndevice = 3, nneutrophil = 250 independent experiments. Scale bars: (A) 500µm, (B) 50µm. Data represented as mean ± SEM.

#### Assessment of Neutrophil Inflammation Activated by Microglial Soluble Factors

To assess the roles of recruited neutrophils in the progression of AD pathology, we investigated neutrophil inflammation. We compared the morphology of neutrophils resting in the annular compartment (**Figure 5A** left) and activated in the central compartment (**Figure 5A** right). Microglia-activated neutrophils extended their bodies by 1.8 folds (**Figure 5B**). In a 37-human cytokine assay, we found two additional cytokines, MIF and IL2 (1.7-fold and 1.67-fold, respectively) from the recruited neutrophils compared to A-beta activated microglia only (**Figures 5D,E**). The other 5-soluble factors from Abeta-activated microglia remained unchanged after neutrophil recruitment. Other key proinflammatory and neurotoxic cytokines including interleukin 1β (IL-1β), interleukin 17 and tumor necrosis factor alpha (TNF-alpha) were below the detection limit of the kit used (**Figure 5C**). This measurement suggests that the recruitment of neutrophils may exacerbate microglial neuroinflammation activated by Abeta in AD.

# DISCUSSION

Microglial cells are cerebral innate immune cells having neuroinflammatory roles for AD pathology. Based on our previous study of neutrophil accumulation on Abeta plaques in AD animal models (17), we hypothesized that neutrophils, the frontline of peripheral immune cells are engaged with neuroinflammation not directly with Abeta but mediated by microglia. Our microscale platforms enabled comprehensive studies of the crosstalk between central and peripheral immunity and overcame significant limitations of previous studies (21–23). For example, the study of neutrophil involvement in previous models lack control over the complex cellular interactions among multiple types of brain cells and leukocytes (23). Another

FIGURE 4 | Reconstruction of human neutrophil recruitment mediated by human microglia in AD. (A) Fluorescent images show discernable neutrophil recruitment (green) modulated by Abeta-stimulated microglia (red). (B) Measurement of single cellular migration shows instant activation and saturation of neutrophil migration responding to co-culturing conditions: microglia (µG) of 20,000 cells per device, 10,000 cells per device, 5,000 cells per device with Abeta at 22 nM, and negative controls (Abeta only, 20,000 microglia only). Neutrophil chemotaxis was assessed with various soluble factors: Abeta at 22 nM, CCL2 at 10 nM, CCL3 at 10 nM, CCL5 at 10 nM, IL6 at 12 nM, and IL8 at 10 nM by counting the accumulated neutrophils (C) and measuring the persistence (D). (E) The effectiveness of microglia-derived chemokines was confirmed by showing the reduced neutrophil accumulation to Abeta-stimulated microglia in addition of neutralizing antibodies for individual chemokines. Statistical significance is denoted by \*\**P* < 0.20, \*\*\**P* < 0.10, \*\*\*\**P* < 0.01 ndevice = 4 (B–E), ndevice = 3, nneutrophil = 250. Scale bars: 500µm (upper right), 50µm (lower pannel). All parameters are presented as mean ± SEM. \*not detected.

key feature of our platform is the capability of isolating activated neutrophils from resting neutrophils, which allows selective monitoring the activities of interest in the AD progression (21) and on-chip investigation of drug efficacy, such as, inhibitor treatment against neutrophil recruitment and microglial inflammation (23).

One of the intriguing findings is that several mediators contribute to the robust recruitment of human neutrophils,

FIGURE 5 | Inducement of additional proinflammatory mediators from recruited human neutrophils. (A) Fluorescent images show the discernable morphology change of activated neutrophils (central compartment) compared to non-activated neutrophils (annular compartment). (B) Intensity profile shows the significantly enlarged neutrophils after recruitment (black dotted line) compared to non-activated ones (light gray dotted line). (C) 21 human cytokines/chemokines for inflammation-related soluble factors were measured. Among them, MIF and IL2 are the additional soluble factors secreted from recruited neutrophils in comparison with other conditions. (D) A radar plot highlights the increase of IL6, IL8, CCL2, CCL3/4, CCL5 from Abeta-stimulated microglia (dotted black line) in comparison with resting microglia (dotted gray line). (E) A radar plot highlights the increase of MIF and IL2 from Abeta-stimulated microglia and recruited neutrophil (dotted black line) in comparison with Abeta-stimulated microglia only (dotted gray line). ndevice = 5 (A–E). Scale bars: 25µm. All parameters are presented as mean ± SEM. \*: not detected.

which is similar to the findings in AD mouse models (24, 25). CCL3 has been found to be associated with AD (26, 27), reported as expressed by neurons and microglia in post-mortem brains from AD patients (28) and upregulated in experimental models mimicking both amyloid and Tau deposits (29). In addition, a CCL3 polymophism has been associated with AD (30) and CCL3 secretion was found dependent on apoplipoprotein ε4, the greatest genetic risk factor for sporadic AD (31). CCL5 is most commonly involved in AD neurodegenerative processes, which regulates the activities of normal T cells. During AD pathogenesis, an elevated level of the glial CCL5 chemokine is observed in the microcirculatory system of the brain (32). Its elevated levels contribute to the recruitment of immunecompetent cells, which occurs concurrently with increased rates of neuronal deaths (33). In addition to these factors, we identified two proinflammatory-specific proteins released during the crosstalk in between activated microglia and neutrophils. IL2 and MIF are known to enhance human-neutrophil inflammatory activities (25). Especially, MIF (macrophage inhibitor factor), was observed in serial brain sections of transgenic APP AD mice and stained MIF immunolabeling in neutrophils in association with Abeta plaques in the transgenic mouse brain sections (34). In addition, functional studies in murine and human neuronal cell lines revealed that Abeta-induced toxicity could be reversed significantly by a small amount of MIF, showing the beneficial effect in AD mediated by microglia (25). Finally, MIF levels in the brain cytosol and cerebrospinal fluid (CSF) of AD patients showed marked increase compare to and age-matched controls. MIF are regarded as an important proinflammatory mediator by observing in the brain cytosol and cerebrospinal fluid (CSF) of AD patients and affecting tau hyper-phosphorylation, which is found in the late stage of AD (35). IL-2 was found to attenuate Abeta pathology and synaptic impairment in AD mice by engaging astrocytes for clearing Abeta and consequently improving synaptic plasticity (36).

# CONCLUSION

We developed a microfluidic platform to explore the crosstalk between microglia and neutrophils in the context of AD and suggested how peripheral immune cells, neutrophils are involved in AD pathology mediated by Abeta-activated microglia to enhance the beneficial neuroinflammation. Our platform will serve as a physiologically relevant in vitro AD brain model for the study of neuroinflammation and a reliable tool for the validation of therapeutic strategy for neuroinflammatory immunotherapy.

#### METHODS

#### Cell Lines, Media, and Reagents

Immortalized human adult microglia cells (SV40-microglia) were created by a company (ABM Inc., Montreal, Canada). The cells were plated onto T25 cell culture flasks (BD Biosciences, San Jose, CA, USA) and maintained in Prigrow III Medium (ABM Inc) supplemented with 10 % FBS (Life Technologies) and 1 % Pen/Strep (Invitrogen) in a CO<sup>2</sup> cell culture incubator. The cell culture medium was changed every 3 days until cells were confluent.

For neutrophil isolation, human peripheral blood samples from healthy volunteers, aged 18 years and older, were collected by a company (Zen-Bio Inc., Research Triangle Park, NC, USA). Peripheral blood was drawn in a 10-mL tube containing a final concentration of 5 mM EDTA (Vacutainer; Becton Dickinson). Nucleated cells were isolated using a HetaSep gradient, followed by the EasySep Human Neutrophil Enrichment Kit (STEMCELL Technologies, Vancouver, Canada) according to the manufacturer's protocol.

#### Microfluidic Device Fabrication

Negative photoresists, SU-8 5 and SU-8 100 (MicroChem, Newton, MA, USA), were sequentially patterned using standard lithography on a 4" silicon wafer to create a mold for cell migration channels of 5µm in height and chemokine compartments of 100µm in height. A mixture of a base and a curing agent with a 10:1 weight ratio (SYLGARD 184 A/B, Dow corning, Midland, MI, USA) was poured onto the SU-8 mold and cured for 1 h at room temperature under vacuum and, subsequently, cured for more than 3 h in an oven at 80◦C. The cured poly dimethyl-siloxane (PDMS) replica was peeled off from the mold and holes were punched for fluid reservoirs. Arrayed holes were also laser-cut (Zing 24, Epilog Laser, Golden, CO, USA) into an acrylic plate of 6 mm in thickness. The machined membrane and the plate were glued together using uncured PDMS and incubated at 80◦C overnight. This assembly was irreversibly bonded first to the PDMS replica using oxygen plasma at 50 mW, 5 cm, for 30 s (PX-250, March Plasma Systems, Petersburg, FL, USA), and later to a glass-bottomed single well plate (P384G-1.5-10872-C, MatTek Corp., Ashland, MA, USA). Immediately after the bonding, 10 µL of poly (l-lysine) solution (PLL, M.W. 70,000–150,000, 1.0 mg/mL, Sigma-Aldrich Co. LLC) was injected into each platform and incubated for 2 h at a room temperature to promote cellular adhesion. PLL-treated surface was rinsed with autoclaved and 0.2µm filtered water (AM9920, Life Technologies).

# Membrane Staining of Microglia

Before the experiment, cells were washed using a serum-free medium and the cell membrane was labeled with red fluorescent dyes (PKH26PCL Red Fluorescent Cell Linker, Sigma-Aldrich). Briefly, after centrifugation (1,200 rpm.g for 5 min), the cells were re-suspended in 1 mL of Diluent C (G8278, Sigma-Aldrich) and immediately mixed with 4 µL of dye solution. The cell/dye mixture was incubated at room temperature for 4 min and periodically mixed by pipetting to achieve a bright, uniform, and reproducible labeling. After the incubation, the staining was stopped by adding an equal volume (1 mL) of 1 % BSA in PBS and incubating for 1 min to remove excess dye. Unbound dye was washed by centrifugation and suspending cells in culture medium (10<sup>6</sup> cells/mL). Ten µL of the cell suspension was injected into each platform and 100 µL of a culturing medium was added into side and central extra wells. The loaded micro-devices were then incubated at 37◦C supplied with 5% CO2.

# Isolation and Membrane Staining of Neutrophils

Human neutrophils were isolated from fresh human blood samples within 2 h of blood collection by using EasySepHuman Neutrophil Enrichment Kits (STEMCELL Technologies) and following the manufacturer's instructions. After isolation, the neutrophils were washed using 1x PBS, and immediately mixed with 4 µL of dye solution (PKH67 Green Fluorescent Cell Linker, Sigma-Aldrich). The cell/dye mixture was incubated at room temperature for 4 min and periodically mixed by pipetting to achieve a bright, uniform, and reproducible labeling. After the incubation, the staining was stopped by adding an equal volume (1 mL) of 1% BSA in PBS and incubating for 1 min to remove excess dyes. Unbound dyes were washed by centrifugation and suspending cells in culture medium (10<sup>6</sup> cells/mL). Ten µL of the cell suspension was injected into each annular compartment. The loaded microfluidic-device were then incubated at 37◦C supplied with 5% CO<sup>2</sup> in 10 min for the cell attachment and used later in microscope.

#### Time-Lapse Imaging

To track individual cells at a high throughput, we prepared 24 devices on a single-well plate and time-lapse imaged the entire devices in a large-area mode of 8 × 8 mm<sup>2</sup> with a 15 % stitching by using a fully automated microscope (Ti-E/NIS Elements, Nikon Inc.) integrated with a heated incubating stage (INU-TIZB-F1, Tokai Hit Co., Ltd., Shizuoka, Japan), set at 37.5◦C and 5% CO<sup>2</sup> with moderate humidity. We imaged cells at every 15 min for 6 h in three channels—FITC, Cy5, and a bright field with a 10x objective lens and a perfect focusing system in a phase contrast mode.

#### Immunostaining

Before immunofluorescent staining, we rinsed the cells twice with DPBS. To fix, cells were incubated in fresh 4% paraformaldehyde aqueous solution (157-4, ElectronMicroscopy Sciences) for more than 15 min at RT followed by rinsing twice with DPBS. To permeabilize, cells were incubated in 0.1% Triton X-100 in PBST (phosphate buffered saline with 0.1% tween <sup>R</sup> 20) for 15 min at RT. To block, cells were incubated in 3% human serum albumin for overnight in PBST at 4◦C. After incubating with the primary antibody solutions for 24 h at 4◦C, the cells were washed five times. The following antibody (and dilutions) were used: anticd11b (1:100, Life Technologies).

# Multiple Human Cytokine Array Kit

A human cytokine array kit was purchased from R&D Systems (Catalog # ARY005, Human Cytokine 37-membrane kit array) and utilized following the protocol provided by the manufacturer. Capture antibodies have been spotted in duplicate on nitrocellulose membranes. Cell culture supernatants are mixed with a cocktail of biotinylated detection antibodies. The sample/antibody mixture is then incubated with the array. Any cytokine/detection antibody complex present is bound by its cognate immobilized capture antibody on the membrane. Streptavidin-Horseradish Peroxidase and chemiluminescent detection reagents are added. A signal is produced, proportional with the amount of cytokine bound. Chemiluminescence is detected in the same manner as a Western blot and the resulting cytokine release profiles were quantified with ImageJ.

#### Microglial Stimulation by Soluble Abeta

After stabilizing microglia in the central compartment, the culturing medium was replaced by a medium containing Abeta 42 (PAM-4349-v, Amyloid Beta-Protein, Peptides International, Inc.) at 22 nM of 2% FBS and microglia were incubated for 72 h.

#### Statistical Analysis

Data, expressed as mean ± SEM, were compared using either a two-tailed Student's t-test when comparing two groups/conditions or one-way ANOVA followed by a post hoc test when comparing 3 or more groups/conditions. P < 0.05 was considered significant.

# DATA AVAILABILITY

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

# AUTHOR CONTRIBUTIONS

JP and HC designed, fabricated, and tested devices, designed experiments, performed immunostaining, statistical quantification, generated figures, wrote and edited the manuscript. JP, SB, IM-J, DI, and HC conceived the ideas and edited the manuscript. All authors read and edited the manuscript extensively.

# FUNDING

This work was supported by the Pioneering Funding Award funded by Cure Alzheimer's Fund (CAF, HC), Duke Energy Special Initiatives funded by Charlotte Research Institute (CRI, HC) and Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2015R1A6A3A03019848, JP), the National Institute of Health (AG059236-01A1, HC).

# ACKNOWLEDGMENTS

We appreciate Dr. You Jung Kang (UNCC) for critical reviewing our manuscript.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02231/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 Park, Baik, Mook-Jung, Irimia and Cho. 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.

# Meningeal Foam Cells and Ependymal Cells in Axolotl Spinal Cord Regeneration

Nathaniel Enos, Hidehito Takenaka, Sarah Scott, Hai V. N. Salfity † , Maia Kirk, Margaret W. Egar ‡ , Deborah A. Sarria, Denise Slayback-Barry, Teri Belecky-Adams and Ellen A. G. Chernoff\*

*Department of Biology, Indiana University-Purdue University Indianapolis, Indianapolis, IN, United States*

#### Edited by:

*Craig Stephen Moore, Memorial University of Newfoundland, Canada*

#### Reviewed by:

*Akira Satoh, Okayama University, Japan Emanuele D'amico, University of Catania, Italy*

> \*Correspondence: *Ellen A. G. Chernoff*

> *echernof@iupui.edu*

#### †Present address:

*Hai V. N. Salfity, Division of Thoracic Surgery, Cardiothoracic Department, Duke University Health Center, Durham, NC, United States ‡Retired, Cleveland Heights, OH,*

#### Specialty section:

*United States*

*This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Immunology*

Received: *30 July 2019* Accepted: *15 October 2019* Published: *01 November 2019*

#### Citation:

*Enos N, Takenaka H, Scott S, Salfity HVN, Kirk M, Egar MW, Sarria DA, Slayback-Barry D, Belecky-Adams T and Chernoff EAG (2019) Meningeal Foam Cells and Ependymal Cells in Axolotl Spinal Cord Regeneration. Front. Immunol. 10:2558. doi: 10.3389/fimmu.2019.02558* A previously unreported population of foam cells (foamy macrophages) accumulates in the invasive fibrotic meninges during gap regeneration of transected adult Axolotl spinal cord (salamander *Ambystoma mexicanum*) and may act beneficially. Multinucleated giant cells (MNGCs) also occurred in the fibrotic meninges. Actin-label localization and transmission electron microscopy showed characteristic foam cell and MNGC podosome and ruffled border-containing sealing ring structures involved in substratum attachment, with characteristic intermediate filament accumulations surrounding nuclei. These cells co-localized with regenerating cord ependymal cell (ependymoglial) outgrowth. Phase contrast-bright droplets labeled with Oil Red O, DiI, and DyRect polar lipid live cell label showed accumulated foamy macrophages to be heavily lipid-laden, while reactive ependymoglia contained smaller lipid droplets. Both cell types contained both neutral and polar lipids in lipid droplets. Foamy macrophages and ependymoglia expressed the lipid scavenger receptor CD36 (fatty acid translocase) and the co-transporter toll-like receptor-4 (TLR4). Competitive inhibitor treatment using the modified fatty acid Sulfo-N-succinimidyl Oleate verified the role of the lipid scavenger receptor CD36 in lipid uptake studies *in vitro*. Fluoromyelin staining showed both cell types took up myelin fragments *in situ* during the regeneration process. Foam cells took up DiI-Ox-LDL and DiI-myelin fragments *in vitro* while ependymoglia took up only DiI-myelin *in vitro*. Both cell types expressed the cysteine proteinase cathepsin K, with foam cells sequestering cathepsin K within the sealing ring adjacent to the culture substratum. The two cell types act as sinks for Ox-LDL and myelin fragments within the lesion site, with foamy macrophages showing more Ox-LDL uptake activity. Cathepsin K activity and cellular localization suggested that foamy macrophages digest ECM within reactive meninges, while ependymal cells act from within the spinal cord tissue during outgrowth into the lesion site, acting in complementary fashion. Small MNGCs also expressed lipid transporters and showed cathepsin K activity. Comparison of <sup>3</sup>Hglucosamine uptake in ependymal cells and foam cells showed that only ependymal cells produce glycosaminoglycan and proteoglycan-containing ECM, while the cathepsin studies showed both cell types remove ECM. Interaction of foam cells and ependymoglia *in vitro* supported the dispersion of ependymal outgrowth associated with tissue reconstruction in Axolotl spinal cord regeneration.

Keywords: foam cells, foamy macrophages, ependymal cells, meningeal fibrosis, spinal cord regeneration, axolotl regeneration, lipid uptake, myelin uptake

# INTRODUCTION

This research examines a previously unidentified phenomenon in spinal cord regeneration involving the accumulation of foamy macrophages in fibrotic meninges during amphibian spinal cord regeneration. In the transected spinal cord, during gap regeneration, of the adult Axolotl (an aquatic salamander, Ambystoma mexicanum) foamy macrophages took up lesion site lipids and myelin while degrading ECM. These innate immune system cells were also found in close association with Axolotl ependymal cell (ependymoglial) outgrowth that remodels the cord following injury. The interaction of foamy macrophage with ependymal cells in vitro modified ependymal cell behavior related to mesenchymal outgrowth.

The role of ependymal cells has been investigated extensively in urodele spinal cord regeneration. Studies include ependymal growth factor and retinoid responses, ECM formation and removal, cytoskeletal changes, remodeling of radial processes and epithelial to mesenchymal transition, association with axonal outgrowth, stem cell properties and neurogenesis, and dorsalventral patterning of the regenerating cord (1–17). The role of a meningeal reaction in urodele spinal cord regeneration has a far less extensive body of work (5, 12, 13). The present research explores aspects of the urodele spinal meninges response complementary to the earlier studies.

Meningeal fibrosis occurs after penetrating spinal cord injury (SCI) in urodele amphibians (newts and salamanders), as it does in mammals [rev. (10, 15)]. Penetrating mammalian SCI induces a meningeal (fibrotic) scar that inhibits axonal regrowth directly and reinforces the astrocytic (gliotic) scar (18, 19). This dual scarring process forms a permanent barrier to axonal regrowth. In urodeles, fibrotic meninges is remodeled and excluded to the periphery of regenerating cord, a process that involves ependymal outgrowth and digestion of extracellular matrix (10, 12, 15, 20). Stensaas (5) and Zukor et al. (12) showed an intimate association of reactive meninges with multiple cell types in transected newt spinal cord. Reactive newt meninges and cord outgrowth were shown to contain macrophages that contact regenerating neurons and ependymoglia during the regenerative process (12).

Foamy macrophages, also known as foam cells, foamy phagocytes or foamy histiocytes, are of monocyte origin and distinguished by the "foamy" appearance of their extensive lipid inclusions in histological preparations (21, 22). They can fuse into "osteoclast-like" MNGCs (21, 23, 24). Foamy macrophages can serve as sinks for lipoproteins and myelin fragments in pathological neural conditions, such as multiple sclerosis (21, 25–27). They can be, at least transiently, beneficial in this pathology (22, 27).

Foamy macrophages form from monocyte-derived M2 macrophage (anti-inflammatory macrophage) precursors (26, 28, 29). Features of foam cells in vivo and in vitro include: clusters of lipid inclusions that are phase contrast bright, stain with Oil Red O or the indocarbocyanine dye DiI, production of the cysteine proteinase cathepsin K, activity of the lipid scavenger receptor CD36, uptake of oxidized low density lipoprotein (Ox-LDL), and uptake of myelin fragments. These features are characteristic of live cell lipid droplets, foam cells and osteoclast-like MNGCs derived from foam cells (21, 25–27, 30–33).

In mammalian SCI, foamy macrophages form only within injured spinal cord tissue, where they take up myelin and contribute to a pro-inflammatory environment (34). Accumulation of foamy macrophages has not been shown within injured mammalian spinal meninges (34, 35). Macrophages have been described within injured salamander spinal cord, as well, and many immune responsive genes are upregulated shortly after Axolotl SCI (12, 36, 37). However, foamy macrophages have not previously been reported in salamander cord or meninges.

Uptake of the toxic lipid metabolites after neural injury can be approximated in vitro by uptake of Ox-LDL (38). A common lipid transport mechanism involved in the uptake of Ox-LDL uses CD36, a class B scavenger receptor/fatty acid translocase (25, 39). In atherosclerosis and other pathological conditions, CD36 and Toll-like Receptor-4 (TLR4), along with TLR6, act together in lipid uptake and inflammatory behavior (40). CD 36 is also involved in fusion of macrophages to form MNGCs (23, 24). These studies suggest the use of an Ox-LDL uptake model and examination of the role of CD36 in Axolotl meningeal foam cell lipid transport.

In many neural pathologies, foamy macrophages and MNGCs also take up myelin sheath products by phagocytosis. Myelin debris persists for extended periods in mammalian spinal cord lesion sites and is sequestered in macrophages (41, 42). Extensive myelin fragment uptake by foamy macrophages occurs within active and chronic-active plaques in the CNS in multiple sclerosis (25–27, 43). In animal models of amyotrophic lateral sclerosis, foamy macrophages are involved in myelin uptake during Wallerian degeneration in the peripheral nerves, associated with loss of axons and neuromuscular synapses (44–46). In Charcot-Marie-Tooth disease, a group of peripheral nervous system (PNS) demyelinating disorders, foamy macrophages with myelin inclusions are found next to poorly myelinated or demyelinated axons (47). Foamy, myelin-containing macrophages are also found in association with peripheral nerve degeneration in aging mice (48). In some of these pathologies the literature is contradictory on the pro-inflammatory or anti-inflammatory nature of the foam cells involved in this processes, depending on the type of experimental system, stage of disease or the markers examined (27, 46, 47, 49–51). The question whether these foam cell effects are beneficial or harmful is even more complex. The work presented here showed uptake of myelin in situ and in vitro similar to that in seen in mammalian neural pathologies.

A third critical property of foamy macrophages in the nervous system, in addition to oxidized lipoprotein and myelin uptake, is their ability to degrade ECM (32). This process appears to be universal in foamy macrophages and occurs in osteoclastlike MNGCs associated with atherosclerosis and MNGC tumors (52, 53). Foamy macrophages and MNGCs digest ECM using cathepsin K and MMP9 as major secreted proteolytic factors (32, 52–55). Secreted proteases can be concentrated on the matrix within the sealing ring, a set of cytoskeletal and membrane specializations also seen in osteoclasts (24, 32). The examination of foamy macrophage and ependymal cathepsin K expression extends our prior studies of matrix proteinase activity from

Parasag, parasaggital; D, day; DIC, differential interference contrast. Magnification bar is shown in the lower portion of each image.

ependymal cells in Axolotl spinal cord regeneration by Chernoff et al. (20).

The studies reported here characterize distribution and functionality of foamy macrophages and some MNGCs in the injury-reactive Axolotl spinal cord meninges, starting at the histological level. In transected adult Axolotl bodyregion (non-tail) cord, transmission electron microscopic (TEM) morphological studies, plus collagen and proteoglycan staining, show that interstitial meningeal ECM fills the lesion site and wraps the regenerating transected stumps. DiI labeling identify lipid-laden mononucleated cells attached to the fibrotic meningeal ECM. Within the lesion site, lipid-laden and multinucleated cells share ECM-filled space with the reactive ependymal cells that grow out from the spinal cord. Transmission electron microscope (TEM) studies show co-localization of foamy macrophages and reactive ependymal cells within lesion site fibrillar collagen. A primary tissue culture system uses lipid stains to identify neutral and polar lipids. Functional studies indicate that the foamy macrophages and some associated MNGCs take up lipid via the scavenger receptor CD36, co-expressed with toll-like receptor 4 (TLR4), and produce cathepsin K.

Our studies are the first to indicate that foamy macrophages are present in reactive Axolotl spinal cord meninges and participate in spinal cord regeneration.

#### METHODS AND MATERIALS

#### Surgery and Tissue Culture Surgery

Axolotls were obtained from the Ambystoma Genetic Stock Center, University of Kentucky and maintained at 20–22◦C in 20% Holtfreter's Salts Solution. Transdermal anesthetic tricaine methane sulfonate (Finquel; Syndel, Formerly Western Chemical) was used in 20% Holtfreter's Salts Solution, with thimerosal for disinfection of the surgical field (5 ml/L of a 20 g/L 88% ethanol thimerosal stock), adjusted to pH 7 with sodium bicarbonate. Finquel concentration was adjusted for the size of the animal. Animals were anesthetized in 0.5 g/l Finquel for adult; >20 cm, 2–3 years old; 0.3 g/L for juveniles; 10–15 cm, 6 months old. Before surgery, animals were injected with the antibiotic Amikacin (2.5 mg/ml, 0.75 ml for an adult; 0.5 ml for a juvenile). The lesioning procedure is described in detail in Chernoff et al. (56) and will not be repeated here. Postsurgically, lesioned-cord animals were treated in 20% Holtfreter's solution at 12◦C in a BOD incubator for 3 days, with daily water changes and amikacin injections, then kept at 20–22◦C in the vivarium through the regeneration process. All husbandry, surgery, analgesia, and euthanasia was performed following the IUPUI School of Science IACUC approved protocol SC 280R.

#### Tissue Culture

For these experiments sets of four to six animals, either sex, age matched, but any color morph, were lesioned. Explants were isolated following the procedure described in Chernoff et al. (56) and O'Hara and Chernoff (6). Two weeks outgrowth (see dashed linesin **Figure 1**) was isolated, cut to size and cultured on poly-Dlysine/fibronectin-coated dishes. Explants were divided among 2–3 dishes. Each experiment was repeated at least three times. Culture contained Leibovitz L-15 medium, 5 mM Hepes Buffer, and progesterone 20 nM. Finally a stock solution containing 5µg/ml insulin, 100µg/ml transferrin, 100µM putrescine, and 30 nM selenium was added (Sigma Chemical). EGF 20 ng/ml, 1% axolotl serum, and 1% Pen-Strep/Fungizone (Gibco) was added to L-15 medium, which was used for the cultures. pH was adjusted to 7.6.

#### Culture Dish Preparation

Thirty-five millimeter polystyrene tissue culture dishes were coated with 100µg/ml poly- D -lysine in HEPES-buffered saline solution, pH 7.4. The dishes were incubated at 37◦C for at least 30 min, then rinsed twice with HEPES buffered saline pH 7.4. Fibronectin (75µg/ml) was added to each dish and incubated at 37◦C for 1 h and rinsed with HEPES buffered saline. HEPES buffered saline contains 0.01 M HEPES, 0.01 M KCl, and 0.013 M NaCl in water, adjusted to pH 7.4. The dishes were rinsed with medium prior to addition of explants.

#### Histological Stains Wholemount

Animals were anesthetized and 3 cm spinal column segments were excised fixed in 4% paraformaldehyde, pH 7.4 on a rocker platform at room temperature for 20 min, then overnight at 4◦C. After rinsing in Hanks' BSS, the tissue was incubated for 1 day in 0.5% neutral calcium chelator EGTA to soften bone, rinsed in HBSS and the spinal cord removed.

To identify the ECM collagen and proteoglycan wholemount segments of axolotl spinal cord were stained. Proteoglycan was stained with alcian blue alone. Collagen and proteoglycan were co-stained with metanil yellow and alcian blue. Samples were stained in alcian blue for 5 min, and in metanil yellow for 1 min. After staining, samples were rinsed briefly in three changes of 10% ethanol for 5 min to remove excess dye.

All samples were viewed and photographed in their wholemount state using a Nikon stereomicrocope and photographed using a Nikon DXM-F digital camera system.

#### Trichrome Stain

Slides with paraffin sections were deparaffinized in xylene, twice for 5 min each. The slides were then placed in decreasing concentrations of ethanol, then distilled water for 30 s at each step. Deparaffinized slides were stained using Weigert's iron hematoxylin, followed by Biebrich scarlet-acid fuchin and phosphotungstic/phosphomolybdic acid following kit instructions (Sigma Chemical Co.). Slides were then soaked for 5 min in aniline blue. After the finally staining step, the slides are placed in 1% acetic acid for 2 min, twice, then rinsed in distilled water for 5 min, twice. Finally, slides are dehydrated and cleared in increasing concentrations of ethanol and finally xylene for 10 min in each solution. Histomount (Pella) permanent mounting medium is then used to mount the slides.

#### Transmission Electron Microscopy

Animals were anesthetized and either perfused, or spinal cord segments were fixed in situ for several minutes in cacodylatebuffered tri-aldehyde fixative (57). Tissue was removed and fixation continued in Kalt's fixative for 24 h. Tissue was post-fixed in buffered 2% osmium tetroxide, dehydrated, and embedded in araldite epoxy resin. Plastic-embedded tissue was sectioned on a Porter-Blum MT-2 ultra-microtome with glass knives for 1-micron sections, or with a diamond knife for thin sections. Thick sections were collected on glass slides and stained with 1% toluidine blue. Thin sections (50–80 nm) were made at selected intervals, collected on bare grids or on Padget-film-supported-1 hole grids, stained with lead-citrate, and examined with a Tecnai G2 12 Bio Twin (FEI, Hillsboro, OR) equipped with AMT CCD Camera (Advanced Microscopy Techniques, Danvers, MA). Two older images were obtained using a Philips 400 transmission electron microscopy.

#### Actin Staining

Samples were fixed in 4% paraformaldehyde in HBSS for 20 min. Fixative was rinsed out with multiple changes of Hanks' Balanced Salts Solution (HBSS) pH 7.6. Cultures were permeabilized for 5 min with 0.1% Triton-X-100 in HBSS. Wholemount samples were permeabilized for 10 min. Triton-X-100 was removed with three 5 min distilled water rinses. Rhodamine-Phalloidin (Invitrogen) stock was prepared in methanol as directed, diluted 1:200 into the sample buffer and used at a concentration of 1.5 units/ml. Cultures were incubated for 10 min and wholemount samples were incubated for 20 min, then rinsed in HBSS twice, 10 min each. HBSS was removed and DAPI/antifade (Invitrogen) added to cover the samples in small culture dishes, which were coverslipped and viewed.

# Lipid Labels and Inhibitor

#### DiI Staining

Cell cultures were fixed with 4% paraformaldehyde at room temperature for 20 min, rinsed in HBSS, then incubated with DiI (1 mg/ml in 100% ethanol) for 15 min. Dishes were rinsed briefly in ethanol to remove unbound DiI. Cultures were incubated in HBSS for 1 h to allow DiI to partition from membranes to lipid droplets. Antifade with DAPI (Invitrogen) was added, cultures coverslipped and viewed with a Nikon Eclipse TE 2000- U inverted phase contrast and fluorescence microscope.

#### Oil Red O Staining

For neutral lipid staining, a stock solution of Oil Red O (Sigma Aldrich) was prepared by dissolving 0.25 g/40 ml of 2-propanol. Solution was warmed at 37◦C to dissolve the dye. Before use, a 3:2 dilution of the dye/propanol was made with water and 0.22 micron filtered. Working solution was prepared immediately before use. Cultures were fixed in paraformaldehyde, rinsed in HBSS, then water. After a brief rinse with 60% 2-propanol the working dye solution was added for 1–2 min. Staining was observed on an inverted microscope. Dye was removed, dishes briefly rinsed once with 60% 2-propanol then distilled water (10 min) and digital images captured.

#### Polar Lipid Live Stain

To visualize neutral and polar lipids together, the DyRect Live-Cell Neutral Lipid Imaging Kit (Marker Gene Technologies, Inc.) was used. In this live foam cell stain, neutral lipids fluoresce green and polar lipids fluoresce red. The green fluorescence overlaps with green autofluorescence in our cells, so only the polar lipid staining properties were useful. DyRect reagent was reconstituted in ethanol at 1 mg/ml and diluted to 1µg/ml in culture medium immediately before use. The dye-medium containing cultures were incubated at room temperature for 1 h, rinsed in HBSS and imaged live.

#### Ox-LDL Uptake

DiI-labeled Ox-LDLs (Invitrogen L3482) were diluted to 25 ug/ml in E3 medium (see Axolotl Tissue Culture section for composition) and applied to cultures for 24 h. The cultures were rinsed with HBSS and fresh medium was applied before observation via fluorescence microscopy.

#### CD36 Inhibition

CD36 inhibition medium was created by adding sulfo-Nsuccinimidyl oleate to E3 medium at a final concentration of 100 uM. Culture dishes were incubated with the CD36 inhibition media for 24 h, then DiI labeled Ox-LDLs (25 ug/ml) were added to the dish for 24 h incubation. Culture dishes were rinsed with HBSS, and fresh E3 media was added before observation under fluorescence microscopy.

#### Glucosamine Uptake, Autoradiography

For glucosamine uptake experiments, explant cultures were established and grown for 6 days. Five µCi (185 µBq) of D- [6-3H(N)]-glucosamine (Perkin-Elmer) was added per ml of E3 medium and the cultures were incubated for 24 h. Labeled medium was removed and cultures rinsed three times in HEPES buffered saline, fixed in ice-cold 5% acetic acid in ethanol for 10 min at 4◦C, rinsed three times with 100% ethanol, and air dried. The dry plates were coated with 0.5 ml Kodak-NB2 nuclear tracking emulsion (Eastman Kodak Co) and allowed to expose for 2 weeks at 4◦C in a light-tight box containing desiccant. After 2 weeks the plates were brought to room temperature, developed with Kodak D-19 developer for 10 min at room temperature, rinsed, fixed with Kodafix (diluted 1: 3) for 10 min, rinsed again, and allowed to air dry. Coverslips were applied to the dishes with Histomount permanent mounting medium (Ted Pella).

#### Immunohistochemistry Paraffin Sections

Dissected tissues were fixed in 4% paraformaldehyde in HBSS, pH 7.6, at 4◦C for at least 1 h, rinsed in HBSS and dehydrated in a graded ethanol series, followed by two xylene rinses, paraffin penetration and paraffin embedding. The sections were cut to 10µm thickness by microtome and "baked" at 60◦C overnight onto Superfrost/Plus Microscope glass slides (Fisher). After deparaffinizing and rinsing with PBT (phosphate-buffered saline, PBS, plus 0.1% Tween-20), the sections were placed in 90◦C 0.01 M citrate buffer (pH 6.0) for 10 min, for post-fixation antigen recovery, unless otherwise noted. Sections were then treated with blocking buffer as follows: PBT with 10% normal goat serum (NGS), diluted 1:1 with Superblock (Pierce Chemical). Primary antibody was added in HBSS and incubated overnight at 4◦C. After rinsing three times with HBSS, sections were incubated with Alexa Fluor 594 secondary antibody (1:2,000, Invitrogen) for 2 h at room temperature. CD36 antibody (R&D Systems MAB25191) was diluted to 2.5 ug/ml. After washing with HBSS twice, sections were mounted in SlowFade Gold antifade reagent with DAPI as the nuclear counterstain (Invitrogen) and coverslipped. Specimens were observed with a Nikon Eclipse E800 fluorescence/DIC microscope.

#### Cell Cultures

Culture dishes were fixed with 4% paraformaldehyde for 30 min at room temperature and rinsed three times with HBSS before incubation in SuperBlock blocking buffer in PBS (Thermo). Dishes were rinsed three times with HBSS, and the appropriately diluted primary antibody applied and incubated overnight. Then dishes were rinsed three times with HBSS, and Alexa Flour 595 secondary antibody was applied and incubated for 2.5 h, rinsed three times with HBSS, mounted in SlowFade Gold antifade reagent with DAPI and coverslipped. Specimens were observed with a Nikon Eclipse TE 2000-U inverted phase contrast and fluorescence microscope or a Keyence BZ-X Fluorescence, phase, DIC Microscope. TLR4 polyclonal Ab (Novus Bio NB100- 56580SS) and TLR4 polyclonal Ab (Novus Bio NB100-56581SS) were diluted to 5 ug/mL in HBSS. CD36 Ab (R&D Systems MAB25191), and CD36 Ab (Novus Bio NB400-145SS) were diluted to 2.5 ug/mL in HBSS. Cathepsin K Ab (Abcam ab19027) was applied at 1:75 dilution in HBSS (antibody information shown in table form in **Supplemental Figure 1**).

# Myelin Experiments

#### Extraction and Labeling

Twelve axolotl brains were isolated and homogenized in a 0.32 M sucrose solution with a sterile plastic pestle in 1.5 ml microcentrifuge tubes. The homogenate was layered over a 0.8 M sucrose solution and centrifuged at a G-force of 16,000 (Sorvall Biofuge Pico, radius = 8.5 cm, 13,000 rpm) for 15 min. The myelin was collected at the interface of the two sucrose solutions. 1,1′ -dioctadecyl-3,3,3′ 3 ′ -tetramethylindocarbocyanine perchlorate (DiI) in ethanol (1 mg/mL) was applied to the myelin extract and incubated for 30 min at RT. The labeled myelin extract was rinsed three times with HBSS to remove any unbound DiI.

#### Endogeneous Myelin Staining

Fluoromyelin Red (Fisher F34652) was added at 1/300 dilution in HBSS to fixed cultures and incubated for 40 min then rinsed in HBSS. DAPI and antifade (Invitrogen) was added to the dishes and coverslipped. Cultures were imaged with fluorescence microscopy.

#### Myelin Uptake

DiI-labeled myelin extract (100 µL/2 ml culture medium) was added to the 35 mm dish cultures and incubated for 48 h. Dishes were rinsed with HBSS and new E3 media added before cultures were observed by fluorescence microscopy.

#### Statistical Analysis

Ninety-two unique photographic fields from >16 cultured explants in untreated controls from seven separate untreated regeneration control experiments were compiled and sorted into categories of outgrowth: condensed ependymal outgrowth with no foam cells, condensed ependymal outgrowth with foam cells, dispersed ependymal cells with foam cells, dispersed ependymal cells without foam cells and mixed forms. Each form was tabulated for each of the seven experiments. Using Graphpad Prism, A one-way analysis of variance (ANOVA) with Tukey's post-hoc Multiple Comparisons Test was performed. In this type of analysis significance is first determined by ANOVA, then an stricter adjusted p-value is determined for each comparison (58).

# RESULTS

#### Overview of Meninges, Meningeal Matrix, and Foamy Macrophages

The results provide a structural overview of the organization of reactive Axolotl meninges, fibrotic ECM and associated foamy macrophages and MNGCs. Control cords were compared with the regeneration stage when invasive meninges have produced interstitial ECM and reactive ependymal cells grow out from the central.

**Figure 1A** shows a trichrome stained histological crosssection of intact adult lumbar region Axolotl spinal cord. The normal meningeal layers were very thin and stain with aniline blue, indicating the presence of collagen. A stereoscopic image of a wholemount preparation of control cord was stained with metanil yellow for collagen and alcian blue for acidic proteoglycans (**Figure 1B**). The entire cord and the nerve roots are ensheathed with the collagen-containing meninges (**Figure 1B**). Small areas of alcian blue represent mainly basal lamina associated with capillaries (**Figure 1B**). Following transection, meningeal fibrosis produced a mass of interstitial ECM seen here in an unstained stereoscopic wholemount preparation (**Figure 1C**). The tapered regenerating outgrowth of the cord was embedded in the meningeal matrix. In a parasagittal paraffin section, the trichrome stain showed a large amount of aniline blue-stained fibrillar collagen with cells interspersed (purple) across the lesion site (**Figure 1D**). The violet material on the right was associated with a fibrin clot.

One bulbous end of a 2-weeks regenerating cord showed several features (**Figure 1E**). An extensive amount of alcian blue-stained sulfated proteoglycan-containing ECM overlayed metanil yellow-stained fibrillar collagen-containing ECM. This meningeal investment of the regenerating cord stump showed numerous dark orange cells on the ECM (arrows). This appearance reflected the large amount of yellowish lipid seen against the stained ECM. In regenerating spinal cord labeled solely with alcian blue and dissected in a parasagittal plane, the lipid-laden cells were found throughout the fibrotic ECM at the tip of the regenerative outgrowth (**Figure 1F**).

Organic solvents used to embed material for sectioning extracted most of lipid present in the lipid-laden cells like those shown in **Figures 1E,F**, precluding lipid staining in paraffin sections. However, DAPI nuclear label in a cross-section of stump near the lesion site showed the presence of numerous multinucleated cells in fibrotic meninges surrounding the spinal cord in the regenerating tissue (**Figures 1G,H**).

To characterize the lipid-laden cells described in **Figure 1**, regenerating cord was labeled with the lipophilic membrane stain DiI and the fluorescent F-actin probe rhodamine-phalloidin. **Figure 2A** shows a wholemount view of unstained spinal cord with outgrowth from one side of a transected, regenerating cord. Dark melanocytes were present on stump meninges, as they are on control cords. The dashed line shows the margin of injuryreactive material accumulated near the cut end (**Figure 2A**). A region of this regenerating outgrowth is shown in the DiI wholemount stain that follows (**Figure 2B**). DiI/DAPI labeling showed a region of lipid-containing cells concentrated in the meninges, on the regenerative outgrowth (**Figure 2B**). These lipid-laden cells were not present on the surface of the stump distant from the transection (∼2 cm from the regenerating end; **Figure 2C**). DiI staining of the distal stump showed only the white matter axonal myelin just beneath the cord surface (**Figure 2C**).

Rhodamine-phalloidin labeling for F-actin showed a small group of MNGCs on the reactive meninges of another lesioned cord (**Figure 2D**). The distinctive podosome actin dots of the sealing ring were seen around the margin of the MNGCs, along with regions of the more interior ruffled border (**Figure 2D**). These structures were localized to the margins of spread MNGCs, like that shown in lesion site outgrowth in vitro (**Figure 2E**). The outgrowth shown in **Figure 2E** was from a distal end region of the regenerating cord like that shown in **Figure 2A**.

Fine-structural detail of ECM, ependymal cells, foam cells and MNGCs in regenerating Axolotl cord was obtained by transmission electron microscopy (TEM). A toluidine bluestained plastic thick section of a 2-weeks lesion site showed an overview of a lesion site examined by TEM (**Figure 3A**). A MNGC, in which nuclei were visualized through several sections, is circled with a dashed yellow line. TEM examination of this area showed juxtaposition of the MNGC, ependymal cells and foam cells (**Figure 3B**, enlarged area **Supplemental Figure 2**). At increased magnification, masses of vimentin intermediate filaments were seen around one of the MNGC nuclei [**Figure 3C**; (59)]. An enlargement of the zone of intermediate filaments is shown in **Supplemental Figure 2B**. The dark structure next to the MNGC appeared to be a telopode [**Figure 3C**; (60)].

**Figure 3D** shows a higher magnification view of an area between foam cells and ependymal cells. A large amount of fibrillar collagen was present and appeared to be in the process of engulfment (**Figure 3D**). **Figures 3E,F** shows a foam cell among meningeal cells, deeper in the lesion site. Fibrillar collagen was absent from this zone. Other white blood cells are found within the 2 weeks regenerate lesion site: **Supplemental Figures 3A,B** show a macrophage engulfing red blood cells and a lymphocyte among ependymal and meningeal cells (61, 62).

Later in the regeneration process ependymal outgrowth re-epithelializes to reform the central canal, re-extend radial processes, re-form endfeet and form channels through which regenerating axons extend (2, 5, 10). **Figure 4** shows the 4–5 weeks period of regeneration at which the regenerating ends of the cord had met and axonal regrowth is underway to restore the full thickness spinal cord. In **Figure 4A**, a trichromestained paraffin cross-section of a 5 weeks regenerate lesion site shows an asymmetrical regenerate with enlarged central canal. The meninges were excluded to the periphery and were still highly reactive, filling the entire space between the regenerating cord and neural arch and vertebral body. **Figure 4B** shows a higher magnification view of a section like **Figure 4A**. The enlarged central canal contained extruded cellular material. The collagenous denticulate ligament (arrow) was adjacent to the regenerating cord. This is significant because the denticulate ligament is an extension of the pia mater, and indicated pial restoration around the regenerating cord. A toluidine-stained plastic section showing a cross-section of a proximal stump from a 4 weeks regenerating cord showed the reactive meninges adjacent to the white matter (**Figure 4C**). This region contained a denticulate ligament profile, comparable to **Figures 4A,B**. The reactive meninges contained foamy cells and a variety of white blood cells and meningeal cells (**Figure 4C**).

#### Ependymal and Foamy Macrophage Lipid Content and Uptake

The presence of neutral and polar lipid in both ependymal cells and foamy macrophages was shown by using Oil Red O stain for neutral lipid, and DiI staining and DyRect, a proprietary stain, for polar lipids. CD36 was identified as a significant lipid transporter in these cells, in association with TLR4, through antibody localization in vivo and in vitro and treatment with a CD36 inhibitor in vitro. Exposure to DiI-labeled Ox-LDL showed the capacity for additional uptake of Ox-LDL in vitro in foamy macrophages.

#### Neutral and Polar Lipids

When 14D regenerating spinal cord outgrowth was placed in culture, accumulation of small lipid droplets in phase contrast images of live ependymal outgrowth, and a larger mass of lipid in foamy macrophages can be clearly seen (**Figure 5A**). The fluorescent lipophilic cationic indocarbocyanine dye, DiI, is used here as a direct lipid probe. The amount of DiI stained foamy macrophage lipid was greater, and lipid droplets generally larger, than that in ependymal cells, but both cell types were labeled (**Figure 5B**). To examine the presence of neutral lipids, Oil Red O staining was performed. Oil Red O stains both ependymal cells and foamy macrophages showing that some of the lipid in both cell types was neutral lipid (**Figures 5C,D**). At higher magnification both labeled and unlabeled lipid droplets can be seen in foamy macrophages and ependymal cells (**Figures 5C,D** insets).

Identification of the types of lipid labeled by DiI has never been clear, beyond its known intercalation behavior into cell membrane phospholipid layers (63). While this implies an affinity for polar lipids, the targets are not certain. To further explore polar lipid content, a proprietary polar lipid label was applied to live cultures. Ependymal and foam cells both showed polar lipid content (DyRect Kit; **Figures 5E,F**). The polar lipid label showed strong overlap with the DiI staining, suggesting polar lipid components were detected by DiI in our cells. In addition to mononucleated foamy macrophages, small MNGCs with lipid also stained strongly for polar lipids (**Figure 5F**). The **Figure 5F** inset indicates the location of 4 nuclei. Nuclear counterstaining was not possible because this was a live cell labeling process and cells could not be permeabilized. The neutral lipid staining properties of the DyRect reagent were not useable in the Axolotl cells. The neutral lipid probe fluorescence overlapped too

strongly with green autofluorescence and was of no utility (see **Supplemental Figure 4**).

#### Lipid Transporters

The fatty acid translocase/lipid scavenger receptor CD36 is the best candidate for lipid transport in the injured Axolotl spinal cord based on foamy macrophage behavior in other tissues (25). An antibody was identified that reacted with CD36 in paraffin-embedded Axolotl tissue and in vitro. In other sources of foam cells, TLR4 co-localizes with CD36 and assists in the lipid transport process, so TLR4 localization was also examined (40). A combination of two TLR4 antibodies to different TLR4 sites was identified that reacts with strongly Axolotl cells in vitro, and was effective in paraffin sections.

In intact adult Axolotl spinal cord sections, the lipid scavenger receptor CD36 was seen in the ependymal endfeet and in small zones of the ependymal cell bodies (**Figures 6A,B**). In the proximal stump region of the lesioned spinal cord, CD36 was expressed in the reactive meninges, as well as the ependymal cells (**Figures 6C,D**). In vitro, CD36 was strongly expressed by foamy macrophages and small MNGCs (**Figures 6E** inset). Reactive ependymal cells retained CD36 expression in vitro, as well (**Figure 6F**).

FIGURE 3 | TEM showing close relationship between MNGC, ECM, ependymal cells, and foam cells. (A) Thick section of plastic embedded 2-weeks spinal cord regenerate outgrowth stained with toluidine blue. Yellow dashed circle surrounds MNGC. (B) Thin section TEM of lesion site shown in (A). A portion of the MNGC is circled in yellow (dashed line). Arrow indicates portion of the ruffled border. Foamy macrophages and ependymal cells are present. (C) A portion of the ruffled border, and perinuclear intermediate filaments are shown. Engulfed myelin fragments are present within the cytoplasm. Engulfed myelin enlarged in inset (arrows) (D) Foamy macrophage engulfing fibrillar collagen within the lesion site. Arrows show intracytoplasmic collagen. (E) Deeper within the lesion site a foamy macrophage is present among ependymal and meningeal cells. (F) Deeper within the lesion site an ependymal cells show a network of processes (arrows). MNGC, multinucleated giant cells; RB, ruffled border; RBC, red blood cell; Epen, ependymal cells; Ifs, intermediate filaments; Col, collagen; Mening, meningeal cell; macr, macrophage; Nucl, nucleus. Magnification bar is shown in the lower portion of each image.

Localization of CD36 in foamy macrophages was, generally, central/perinuclear and overlapped the zone of lipid droplets (**Figure 6E**). CD36 was also present in ependymal cells (**Figure 6F**).

TLR4 was expressed in both the foamy macrophages and ependymal cells (**Figures 6G,H**). The region of TLR4 localization in ependymal cells was comparable that seen for CD36 expression, but smaller that that seen in the foamy macrophages (**Figures 6F,H**). TLR4 was not detected in large, lipid droplet-free MNGCs (**Figure 6G**).

# Ox-LDL Uptake in vitro

At the time of isolation, all lesion site foamy macrophages and ependymal cells contained lipid acquired in situ and both cell types expressed lipid transporters (**Figures 5**, **6**). Results shown in **Figure 7** address the ability of foamy macrophages

and reactive ependymal cells to take up additional lipid in vitro. Explant cultures were established and exposed to DiI-Ox-LDL. A subset of foamy macrophages took up significant amounts of DiI-Ox-LDL (**Figures 7A–E**). High uptake foam cells were found migrating on the plastic dish, as well as on and under the explants (**Figures 7A,B,D,E**). The limited number of MNGCs seen in these cultures did not take up DiI-Ox-LDL (e.g., **Figure 7A**). Ependymal cells also did not take up DiI-Ox-LDL in vitro (**Figure 7B**). All of the foamy macrophages and ependymal cells contained extensive lipid stores (**Figures 7A–C**), so it is not clear whether those not actively transporting did so because they had reached maximum LDL content, or whether the culture conditions were not fully optimal for lipid uptake.

Functional involvement of CD36 in the in vitro Ox-LDL uptake was shown by combining DiI-Ox-LDL uptake with treatment using the CD36 inhibitor sulfo-N-succinimidyl oleate, a modified fatty acid (64). Typical dense DiI-Ox-LDL uptake label was seen in actively transporting control cultures (**Figure 8A**). Co-treatment with the Ox-LDL label and the oleate inhibitor reduced the area of uptake to small patches and streaks (**Figure 9B**). This inhibition of Ox-LDL uptake shows a functional role for CD36 in lipid uptake in the Axolotl spinal meninges foam cells.

# Myelin Uptake

Myelin fragments produced from damaged white matter in urodele SCI are seen within lesion site cells in TEM studies [**Figure 2C**; (5, 12)]. To determine whether the foamy macrophages and/or ependymal cells were serving this function in our experimental system, the presence of myelin engulfed in situ (lesion site myelin uptake) was examined along with the ability of lesion site cells to take up additional myelin in vitro.

To avoid the enormous amount of autofluorescence found in Axolotl lesion site white matter, endogenous myelin content was assayed by staining primary cultures with fluoromyelin red. Fluoromyelin-labeling of myelin fragments taken up while in situ, were seen in cells within the lesion site explants (**Figure 9A**), in ependymal cells (**Figures 9A,B**) and foamy macrophages migrating on the culture dishes (**Figures 9B–D**). In areas of outgrowth showing large numbers of foam cells mixed with ependymal cells, more myelin label appeared in the foam cells (**Figure 9B**). Within a given culture there were areas of foam cells with different amounts of myelin stain (**Figures 9C,D**).

The continuing ability of lesion site cells to take up myelin fragments was also assayed in vitro. The probe was DiI-labeled Axolotl brain myelin. New myelin uptake was seen by cells within the explants (**Figures 10A,E,F**). In the cellular outgrowth, some zones of ependymal cells showed extensive myelin uptake

Ependymal (white arrows) and foamy macrophages (yellow arrows) in regenerating explant cultures are labeled with polar lipid marker DiI. Fluorescence/Phase contrast image. (C) Ependymal cell lipid droplets label with neutral lipid marker Oil Red O. Phase contrast image. (D) Foamy macrophages lipid droplets also label with neutral lipid marker Oil Red O. Phase contrast image Insets in (C,D) show higher magnification images of ependymal cells and foamy cells, respectively. (E,F) Ependymal cells (white arrows in E), foamy macrophages (yellow arrows in E,F), and MNGCs (white arrow in F) all labeled with a commercial polar lipid marker. Fluorescence/Phase image, Inset in (F) shows a higher magnification of an MNGC with 4 nuclei, fluorescence only. Regen, regenerating; DIV, days *in vitro*. Magnification bar is shown in the lower portion of each image.

(**Figure 10A**), while in others there was little (**Figures 10C,D**). The regions with extensive ependymal myelin uptake appeared to be free of foam cells, as in **Figure 10A**. Foamy macrophages showed robust DiI-myelin uptake in vitro (**Figures 10B–D**). A 2 days period of exposure was used for the labeled myelin uptake. Cells were cultured without DiI-myelin for an additional 12 days, during which the ependymal cells appeared to turn over most of the labeled myelin (**Figure 10A** compared with **Figure 10E**). During the 12 days chase period foamy macrophages seemed to retain more of the labeled myelin, or turnover had released free DiI in these cells which partitioned into lipids stored within the cells (**Figure 10F**).

FIGURE 6 | CD36 and TLR4 in regenerating cord *in vivo* and *in vitro*. (A) The scavenger receptor CD36 was detected on the endfeet of ependymal cells in control cord paraffin cross-sections. Fluorescence image. (B) CD 36 is also present on ependymal cell bodies in intact adult Axolotl spinal cord. Fluorescence image. (C) In paraffin sections from regenerating cord stump CD36 was detected in ependymal endfeet in the reactive meninges. Fluorescence image. (D) CD36 also present on ependymal cell bodies in regenerating cord, proximal stump. Fluorescence image. (E) *In vitro*, CD36 is found on foamy macrophages from 10D regenerate spinal cords 6 days *in vitro* (yellow arrows). Orange arrows show three nuclei in CD36<sup>+</sup> MNGC. Inset in (E) shows a CD36<sup>+</sup> MNGC with six nuclei. (F) CD36<sup>+</sup> Ependymal cells (white arrows) in explant cultures from a 10D regenerate spinal cord 6 days *in vitro*. (G) TLR4 was detected in foamy macrophages (yellow arrows) in culture on cells from 14 days regenenerates, 17 days *in vitro*. (H) Ependymal cells from 14 days regenenerates, 17 days *in vitro* are also TLR4<sup>+</sup> (white arrows). D, day; Regen, regenerating; MNGC, multinucleated giant cells; TLR4, toll like receptor 4. Magnification bar is shown in the lower portion of each image.

# Cathepsin K

The cysteine protease cathepsin K catabolizes several ECM molecules, including collagen, elastin, and gelatin. We hypothesized that cathepsin K present in Axolotl cells within the regenerating cord might be involved in catabolizing the fibrous collagen in the lesion site, allowing the regenerating cells to form

contrast in (C) and overlaid with fluorescence to highlight the foamy macrophages that have taken up DiI-Ox-LDL (yellow arrows). (E) Shows foamy macrophages that have migrated out of the explants and taken up DiI-Ox-LDL in culture. D, day; Regen, regeneration. Magnification bar is shown in the upper portion of images A and E, and the lower portion of images (B–D).

a more regeneration-friendly region in which new ECM might be secreted. To determine if cathepsin K was expressed in the lesion site, cells in vivo and in vitro were labeled with cathepsin K antibody. **Figure 11A** shows a reactive meningeal flatmount from the lesion site where cells were labeled with cathepsin K antibody. In culture, there was expression of the enzyme in cells within the explants and in foamy macrophages and ependymal cells growing out of the explants (**Figures 11B–E**). In favorably oriented foamy macrophages, the cathepsin K was seen concentrated between the foam cell and the plastic substratum within the sealing ring (**Figures 11B,C**). In migrating ependymal cells, substantial portions of the cytoplasm were cathepsin K-positive (**Figures 11D,E**). Small MNGCs also expressed cathepsin K (**Figure 11E**).

#### Glycosaminoglycan Synthesis

Both foam cells and ependymal cell produce ECM-degrading enzymes [**Figure 11**; (20)], but ependymal cells also rebuild the regenerating spinal cord, including reforming the basal lamina of the glia limitans (10). Comparison of the ECM synthetic capacity of foamy macrophages and ependymal cells was performed using <sup>3</sup>H-glucosamine incorporation and autoradiography as an assessment of glycosaminoglycan and proteoglycan synthetic capacity (65).

Extensive <sup>3</sup>H-glucosamine uptake by mesenchymal ependymal cells occurred over a 24-h incubation in established cultures (**Figures 12A,C**). Label occurs in the cytoplasm and nucleus, and there is deposition of material onto the substrate near the explants where ependymal cells were most densely distributed. The nuclear label is associated with nuclear pore complex incorporation of O-linked N-acetyl glucosamine synthesized from the <sup>3</sup>H-glucosamine (66). No incorporation of <sup>3</sup>H-glucosamine was seen in or around any of the foamy macrophages, identified on the basis of their lipid droplet content comparable to those cells labeled with Oil Red O and DiI (**Figures 5**, **12B,C**). This indicates a major difference in lesion-site remodeling roles between ependymal cells and foamy macrophages.

#### Foam Cell/Ependymal Interaction

To observe foam cell interaction with the ependymal cells, regenerating tissue was placed in culture conditions that maintain mesenchymal ependymal outgrowth and proliferation of reactive Axolotl spinal cord ependymal cells (56). Pieces of tissue with reactive meninges and ependymal cells grown in culture typically showed three stages of growth: (1) at 3 h after the start of culture, the attachment period, the explant was firmly adherent and cells at the margins started to extend processes (**Figure 13A**), (2) after 1 day of culture (**Figure 6B**), injury-reactive ependymal cells were migrating out of the explants and meningeal tissue was spreading, and (3) at 4 days in culture, the explants had extensive, dispersed ependymal outgrowth with foamy macrophages on and among the ependymal cells (**Figure 13C**). The extensive outgrowth period persists though at least 22 days in vitro.

Labeling of cultured 2-weeks regenerating spinal cord with fluorescent-phalloidin showed localization of F-actin in sealing rings of foamy macrophages on the ependymal outgrowth (**Figure 13D**). At 4 weeks of regeneration, the ependymal outgrowth was reconnected between the cranial and caudal stumps (10). When 4-weeks lesion site regenerate tissue was excised and cultured, the foam cells were no longer localized with the ependymal cells (**Figure 13E**).

In analysis of cellular arrangement in culture, outgrowth could be divided into 5 forms of foamy macrophage/ependymal interaction: (1) condensed outgrowth with no foamy macrophages (**Figure 14A**), (2) condensed outgrowth with foamy macrophages (**Figure 14B**), (3) dispersed outgrowth with foamy macrophages (**Figure 14C**), (4) dispersed outgrowth without foamy macrophages (**Figure 14D**), and (5) mixed (**Figure 14E**). Ninety-two photographic fields from >16 explants in 7 experiments were sorted into the categories listed above. A one-way analysis of variance (ANOVA) with a post-hoc Tukey analysis showed a highly statistically significant difference

FIGURE 9 | Endogenous myelin uptake is robust in ependymal cells and foamy macrophages. Fourteen days regenerate spinal cord explants were cultured for 8 days *in vitro* and stained with fluoromyelin label to reveal endogenous myelin uptake from the lesion site. Fluorescence/phase images. (A) The explant periphery is heavily labeled as is a zone of ependymal outgrowth (white arrows). There are a few, scattered myelin-containing foam cells (yellow arrows). (B) In a mixed zone of outgrowth foam cells are labeled (yellow arrows), but ependymal cells show little label (white arrows). (C) A zone with heavily myelin-laden foam cells (yellow arrows). White arrow shows one myelin-laden ependymal cell. (D) A zone of foam cells with little endogenous myelin uptake (yellow arrows). Endog, endogenous; Epend, ependymal. Magnification bar is shown in the lower portion of each image. Magnification bars are in the lower portion of images.

between the presence of foamy cells with dispersed vs. condensed ependymal cells (**Figure 14F**). The occurrence of dispersed ependymal growth with foamy macrophages was far more prevalent when compared to the dispersed ependymal cells without foamy macrophages. This difference is highly significant with a p-value < 0.0001. Conversely, a condensed form of reactive ependymal cell outgrowth occurs in the absence of foam cells, compared to the occurrence of condensed outgrowth in the presence of foam cells (p-value < 0.0001). Further test details are included in **Supplemental Figure 5**.

# DISCUSSION

#### Diagrammatic Summary

The cell and ECM associations shown in **Figures 1**–**3** are interpreted diagrammatically in **Figure 15A**. Fibrillar collagen

a region with ependymal cell outgrowth only, the explants and ependymal outgrowth took up significal levels of labeled myelin fragments (white arrows). (B) In an area of mixed, dispersed foam cell and ependymal cell outgrowth, only the foam cells are heavily labeled. (C) An denser area of dispersing ependymal cells (white arrows) plus foam cells (yellow arrows) shows heavy myelin uptake *in vitro* only in the foam cells. (D) A sparsely labeled explants has strongly labeled foam cell outgrowth (yellow arrows) and unlabeled ependymal cells (white arrows). (E,F) Show a 12 days myelin-free chase period following 2 days of DiI-Axolotl-myelin uptake. (E) Shows part of the same region shown in (A). Labeled myelin in the ependymal outgrowth is greatly reduced (white arrows). The explants is still labeled. (F) Even after the 12 days chase period, foam cells. D, day; Epend, ependymal; FC, foam cell. Magnification bar is shown in the lower portion of each image.

from the meninges was found throughout the lesion site between the retracted spinal cord stumps after transection. Sulfated proteoglycan was found throughout the outgrowth, but concentrated closer to periphery. Meningeal cells invaded from the periphery while reactive ependymal cells withdrew their radial processes, became mesenchymal and migrated out into the lesion site (16). Macrophages and foamy macrophages were present.

**Figure 15B** shows a composite model of markers to be used throughout the experimental results, illustrating foam cell specializations that were characteristic to this cell type and are indicators of function within the Axolotl cord lesion site. On

tissue from 16D regenerating cord were co-labeled for cathepsin K (red) and DAPI (blue). Cells on the surface of the meninges are positive for cathepsin K (dashed white line). Explants from 16D regenerate cord, cultured for 10DIV, showed labeled ependymal cells (B,D,E), foamy macrophages (B, dashed square), and an MNGC (E). Yellow arrows indicate foam cells, white arrows indicate ependymal cells. Ruffled borders sequestering cathepsin K between foam cells and the culture substratum were also detected (yellow arrow, C). Yellow arrows indicate nuclei in MNGCs positive for cathepsin K in (E). Regen, regenerating; D, day; DIV, days *in vitro*; MNGC, multinucleated giant cells. Magnification bar is shown in the lower portion of the images.

the unattached cell surface, is the lipid transporter CD36 which would mediate uptake of lipids, to be sequestered within the cytoplasm. TLR4/TLR6 acts in concert with CD36 in transport and are also known to be involved in fusion into MNGCs (23). The sealing ring structures and cathepsin K are show on and associated with the ECM-attached surface.

#### Organization of the Lesion Site

There are species differences in urodele SCI responses, just as there are in rodents (10, 67). The ependymal outgrowth process during gap regeneration in newts is more in the form of an epithelioid bulb than mesenchymal outgrowth, but newt transection lesion sites share features with those seen here in the Axolotl (5, 12, 16, 68). In early stages of spinal cord regeneration in both Notophthalmus viridescens (Eastern Red-Spotted Newt) and Ambystoma mexicanum (Axolotl) a fibrillar collagen-rich ECM wraps the regenerating tissue (Zukor et al., Figure 7; our **Figure 1E**). In the Axolotl, fibrillar collagencontaining matrix encases and bridges the cut ends of the cord, and cells grow out into that material (**Figures 1C,D**). Zukor et al.

(12) showed the accumulation of CSPG in the meninges after transection in the newt. The Axolotl ECM that supported foam cell, MNGC and ependymal cell invasion also contained sulfated proteoglycans (**Figures 1E,F**). In both the newt and Axolotl there is little sulfated proteoglycan present before injury (**Figure 1B**). Compared to mammalian SCI ECM, it still not entirely clear whether the interstitial matrix that forms with meningeal fibrosis in urodeles is intrinsically non-inhibitory in composition or organization, or whether regeneration proceeds because of successful removal of this material (69–71). Production of cathepsin K by the foamy macrophages, MMP and cathepsin production by ependymal cells, plus known MMP production by foamy macrophages in other tissues suggests that removal of the ECM material is likely a strong component of the process [**Figure 11**; (20, 72)].

Axolotl lesion site ECM appeared to have two zones: one closely wrapped around the regenerating cranial and caudal ends of the cord plus the material between the cut ends. Following fixation, they were separable, the wrapped material remained firmly attached to the regenerating cord (**Figures 1C,E,F**). Inflammatory response cells accumulate in this material in both the newt and Axolotl [(12); **Figures 1**–**4**; **Supplemental Figure 2**]. Though not explicitly identified, the newt TEM images shows lipid-laden

cells among the immune response cells accumulated in the reactive meninges (12). In the Axolotl lesion site, the foamy macrophages were concentrated in and on the meninges of the regenerating cranial and caudal stumps (**Figures 1**, **2**).

In situ, MNGCs are found in clusters on and in the fibrotic meninges (**Figures 1G,H**, **2D**, **3A–C**). It is not known if representative numbers of these cells are growing out of the lesion site explants in vitro, or whether they are so strongly attached to lesion site ECM that they are under-represented on the

culture dish. Cluster of MNGCs are not seen in the culture dish outgrowth (**Supplemental Figure 6**). Further studies of axolotl meningeal MNGC formation and behavior are required.

# Identification of Ependymal Cells

In these culture conditions (+EGF, fibronectin coating), the cells that grow out of the lesion site explants are ependymal cells, foamy macrophages, MNGCs and, rarely, melanocytes [**Figures 2E**, **5**–**11**, **13**; **Supplemental Figure 7**; (6, 16, 56, 73)]. Other cord or meningeal cells fail to exit the explants. The Axolotl ependymal cells have been thoroughly characterized in prior marker studies. Intact Axolotl cord ependymal cells are cytokeratin and glial fibrillary acidic protein (GFAP)-positive and the cytokeratins and GFAP are lost during epithelial-mesenchymal transition (6, 16, 73). All of the injury-reactive Axolotl ependymal cells express the stem/progenitor cell marker mRNA-binding protein Musashi1 (16). A Musashi1 expression image is shown, again, at reviewer request along with GFAP intermediate filaments in the process of turnover. The appearance of the large ependymal nuclei is also quite distinctive (56). Examples of control culture Musashi1 localization and perinuclear GFAP in vitro are shown in **Supplemental Figure 8**.

# Identification of Axolotl Foamy Macrophages

The identification of the lipid-laden mononucleated cells in the Axolotl lesion site in situ and in vitro as foamy macrophages was based on systematic examination of markers used in mammalian foamy macrophages. The labeling of lipid droplets with Oil Red O was primarily perinuclear (**Figure 5**), as it was in mammalian foamy macrophages (29). The cells also took up DiI-Ox-LDL like mammalian foam cells [**Figure 7**; (28, 74)]. The lipid transporters localized in the Axolotl foamy macrophages were CD36 and TLR4, the primary transporter and co-transporter used by mammalian foamy macrophages [**Figure 6**; (75)]. An active role for CD36 in lipid transport was supported by the specific inhibition of Ox-LDL uptake by the modified fatty acid Sulfo-N-succinimidyl Oleate [**Figure 8**; (64)]. Axolotl spinal cord foamy macrophages took up myelin fragments in vivo and in vitro, like foam cells in diseases including multiple sclerosis, and within mammalian spinal cord after SCI [**Figures 2C**, **9**, **10**; (27, 34)]. Myelin uptake by macrophages in mammalian SCI and multiple sclerosis models is a phagocytic process, and membrane-bound myelin inclusions are seen in our foamy macrophages and smaller MNGCs by TEM [(41, 76); **Figure 3C** inset]. Axolotl foamy macrophages produced cysteine proteinase cathepsin K characteristic of mammalian foam cells, and cathepsin K was seen localized within sealing rings [**Figure 11**; (54)]. Finally, TEM examination shows ultrastructural features like a ruffled border and perinuclear intermediate filaments seen in mammalian foamy macrophages [**Figures 3B,C**; **Supplemental Figure 2**; (59)].

### Misidentification of Axolotl Foamy Macrophages

Macrophages have been described in injured salamander spinal cord, but not foamy macrophages (36). Results presented here (**Figures 1B**, **2C**) showed that foamy macrophages were absent from distal stump and control Axolotl cord but accumulated in the fibrotic meninges within 1–1.5 mm on each side of the transection site. There is a history of noting, but not accurately identifying, these cells in the urodele regeneration literature. Our early studies show the presence of lipid-laden mono- and multinucleated cells in reactive axolotl cord ependymal cultures that were considered to be osteoclasts and their precursors that had migrated from neural arch following laminectomy performed to expose the cord for transection (56). Maier and Miller (77) noted the presence of what appear to be foamy macrophages among newt limb blastema cells in culture, but called them signet cells because they had eccentrically positioned nuclei. Washabaugh and Tsonis (78) identified these cells as signet cells following Maier and Miller [(78), **Figure 1B**] in regenerating newt limb blastema cultures, noting the presence of granules that appear to be lipid droplets. In a newt spinal cord regeneration study, Zukor et al. (12) showed macrophages and other white blood cells in and around the newt spinal cord lesion site by TEM, and some of these appear to have foamy cytoplasm [(12), **Figures 7C,D**]. The present studies are the first to show that these lipid-laden cells concentrate in and on the meninges at the lesion site in urodele spinal cord regeneration and to characterize them as foamy macrophages based on functional markers, lipid labeling and uptake studies.

#### Possible Source of Foamy Macrophages

Two waves of macrophage recruitment have been described in mammalian SCI: first M1 (pro-inflammatory) macrophages of splenic origin, then M2 (anti-inflammatory) macrophages from either bone marrow or resident tissue immune cells (79). In a mouse spinal cord contusion system, M1 macrophages precursor cells were carried in the vasculature of the leptomeninges (mainly the arachnoid layer), which exit through the subarachnoid space to become M1 macrophages within the cord (80). Precursors to the M2 macrophages enter the brain ventricular-choroid plexus, travel through the cerebrospinal fluid, then invade the injured spinal cord and became M2 macrophages (80). It appears that conversion to mammalian cord foamy macrophages occurs within the spinal cord tissue (34, 35). Control Axolotl cord and stump tissue meninges have no foam cells (**Figures 1B**, **2C**), but they are abundant in the lesion site. The source of these foamy macrophages and MNGCs is not yet known. Future studies will be required to determine whether the entire population is recruited from circulating monocytes and macrophages or if resident macrophages proliferate locally and invade the lesion site.

#### Multinucleated Giant Cells

Activity differs between large and smaller MNGCs in our tissue culture system: smaller MNGCs and mononucleated foamy macrophages were more like each other than were large and small MNGCs. Cathepsin K activity, expression of the lipid transporter CD36 and co-transporter TLR4 were found in mononucleated foam cells and small MNGCs (≤6 nuclei, **Figures 6E,G**, **11E**). TLR4 was absent from very large MNGCs (**Figure 6G**), consistent with the absence of lipid droplets. Very large MNGCs were generally lipid droplet free: one with 36 nuclei is shown in **Figure 6G** and **Supplemental Figure 6D** along with other non-lipid-containing MNGCs (**Supplemental Figure 6**), so their origin or metabolic state may be very different from the smaller lipid-laden MNGCs. Small MNGCs displayed cathepsin K activity (**Figure 11E**), but large MNGCs were not seen in any of the cathepsin K antibody labeled cultures, so the relationship of large numbers of nuclei and ECM proteolysis is not yet known in our system.

#### Cytoskeleton Identity and Function

F-actin localization showed that the foamy macrophages (**Figure 13D**) and MNGCs (**Figure 2D**) attach to the reactive meninges with characteristic podosome-studded sealing rings. These integrin and actin-containing structures are present in other types of macrophages and osteoclasts, where they are associated with a polarized (substratum-attached vs. free surface) morphology needed for normal secretory and transport functions (24, 32).

TEM studies show Axolotl meningeal foamy macrophages and MNGCs with masses of perinuclear intermediate filaments (**Figure 3C**; **Supplemental Figure 2B**). Vimentin intermediate filament accumulation is characteristic of mononucleated osteoclast precursors, osteoclasts and osteoclast-like MNGCs in mammals and clustering of nuclei in MNGCs occurs within a netlike "nest" of vimentin intermediate filaments (59).

#### Cathepsin K in vivo and in vitro

The cysteine protease cathepsin K is probably best known for its role in osteoclast pit production on bone (24, 32). Cathepsin K also performs a role in foamy macrophage and macrophage-derived MNGC ECM degradation (32, 52–55). Foamy macrophages in the Axolotl spinal cord lesion site showed cathepsin K localization (**Figure 11A**) and, in vitro, foamy macrophages and small MNGCs were all cathepsin K positive (**Figures 11B,C,E**). In vitro, cathepsin K was localized within the sealing ring between the cell and culture substrate and also within the cytoplasm of foamy macrophages (**Figures 11B,C,E**).

In the normal CNS, cathepsin K is expressed in mouse choroid plexus ependymal cells, but it is not a universal feature of intact ependymal cells throughout the mouse CNS (81). It is not known how widely distributed cathepsin K might be in the urodele CNS.

The activation of cathepsins is a complex process. Association of cathepsins with negatively charged glycosaminoglycan chains of proteoglycans allows autocatalytic activation (82). In regenerating urodele spinal cord there is abundant sulfated proteoglycan, including CSPG [**Figure 1**; (12)]. CSPG, specifically, has been shown to be involved in extracellular autoprocessing of pro-cathepsin K in osteoclasts (83), and is a good candidate for the same role in reactive Axolotl ependymal cells, foamy macrophages and MNGCs. This response to the fibrotic meningeal sulfated proteoglycan would mediate matrix degrading activity locally, contributing to prevention of permanent scar formation.

#### Intracellular Distribution of Cathepsin K

Ependymal cell cathepsin K localization was markedly asymmetrical (**Figures 11B,D,E**) and could be related to leading edge/trailing edge polarity. In studies of tumor cells, related to understanding metastatic migration, attempts have been made to assign cathepsin localization to the leading edge or trailing edge of migrating cells. In stationary breast carcinoma cells, the localization of cathepsin B is perinuclear, but it is concentrated on one side of the nucleus in a moving cell (84). It is reported that this is the trailing edge of the cell (84). In a study of cathepsin H localization in a prostate cancer cell line, the cathepsin co-localizes with talin, a leading-edge protein associated with focal adhesions in cell migration (85). The cathepsin H was also abundant around the nuclei. The Jevnikar et al. (85) paper is the only one showing a leading edge or trailing edge marker (talin) co-localized with a cathepsin.

In our cultured foamy macrophages and MNGCs, the cathepsin K was localized either within the sealing ring (**Figures 11B,C**) or in a polarized fashion in elongated cells (**Figures 11B,E**). In primary T-lymphocytes and T-lymphocyte lines, cathepsin X is found strongly localized in the both the leading edge and in the uropod, the trailing, deadhesive structure of lymphocytes (86).

While associated with leading edge/trailing edge polarity morphologically, correlation of cathepsin localization with a mechanism of directional cell migration is not possible at this time in either the ependymal cells or foamy macrophages.

#### Lipid in Ependymal Cells and Foamy Macrophages

TEM examination has shown that normal urodele ependymal cells contain some small lipid droplets. During regeneration, newt tail cord TEM examination shows the accumulation of lipid droplets in ependymal cells (2). The cytoplasm of intact mammalian ependymal cells label with DiI, but the magnification and resolution of the published images in that study do not permit identification of lipid droplets (87). In other mammalian studies, normal, young, non-pathological CNS ependymal cells do not contain stores of lipid. In the ventral portion of the brain, lateral ventricles of young mice stained with Oil Red O shows little lipid in the ependymal cells (88). There is more lipid in middle-aged mice and a high level of lipid in aged mouse ependymal cells (88). Similarly, there is a large increase of lipid in Alzheimer's Disease choroid plexus ependymal cells connected to expression of receptors for the transcytosis of LDL, lipoprotein receptor-related proteins-1 and 2 (megalin) (89, 90). So lipid accumulation in ependymal cells is associated with a disturbed or pathological state in ependymal cells.

The combined Oil Red O, DiI and DyRect labeling showed both neutral lipid and polar lipid content in both ependymal cells and foamy macrophages (**Figure 5**). The Axolotl ependymal cells and foamy macrophages also expressed the lipid scavenger receptor CD36 in vivo and in vitro indicating a common mechanism of lipid uptake (**Figure 6**). Therefore, both ependymal cells and foamy macrophages could be participating in removal of toxic lipids released or formed after neural injury. The source of neutral and polar lipids in both populations of cells could involve uptake of native and oxidized lipoproteins as well as myelin breakdown products from the Axolotl spinal cord lesion site.

#### Autofluorescence

Autofluorescence in Axolotl foamy macrophages confounds the use of green fluorochromes, including the Dyrect neutral lipid fluorochrome (**Supplemental Figure 4**). The source is not necessarily unique, but foamy macrophages can take up lipofuscin (ceroids) in pathological conditions (Dvorak and Monahan-Early, 1992). Foam cells exposed to Ox-LDL accumulate lipofuscin, and lipofuscin autofluorescence has even been used as a marker for phagocytic CNS macrophages (74, 91). Lipofuscin can also be liberated by neuronal death. Autofluorescence of lipofuscin has an emission spectrum in the region of 430–490 nm, with a maximum in the yellow range (92). This would bleed through typical green fluorescence filters. The degree of autofluorescence of foamy macrophages and ependymal cells varies, suggesting differential uptake of autofluorescent compounds or differential exposure to them within the lesion site. Maier and Miller (77) noted the autofluorescence of what they termed "signet cells" from regenerating newt limb blastemas.

# Myelin Uptake

The myelin uptake behavior seen in the present study could be beneficial in regeneration. The axolotl form of the myelin-associated inhibitory molecules Nogo-A (axNogo), the Nogo receptor and myelin associated glycoprotein (MAG) are expressed in the Axolotl CNS, including during regeneration, suggesting that they are not inhibitory. Localization of axNogo and MAG in urodeles is primarily in gray matter neurons and ependymal cells, unlike the myelin/oligodendrocyte localization in mammals (93, 94). Despite these differences in the effects of myelin-associated molecules between mammals and urodeles, lesion site myelin was removed by ependymal cells and foamy macrophages during activity in vivo and in vitro (**Figures 9**, **10**). One additional consideration regarding removal of myelin is the role this process plays in modulating M1 pro-inflammatory to M2 anti-infammatory phenotype of foamy macrophages in multiple sclerosis models (22, 27, 34, 95–97). Pro- or anti-inflammatory properties of Axolotl spinal cord foamy macrophages remain to be studied.

There were distinct differences in initial myelin content in ependymal cells placed in culture: some regions of explants outgrowth labeled heavily and some not at all. This could be dependent on their initial location within the lesion site. The foamy macrophages are concentrated around the regenerating ends of the cord, while the ependymal cells are growing out into this zone from the spinal cord stumps. In regions close to the transection site, the foamy macrophages may be the initial cell population to engulf myelin fragments, outcompeting the laterarriving ependymal cells. Deeper within the lesion site, where there are fewer foamy macrophages (**Figures 2A,B**), ependymal cells may be the primary cells to take up myelin.

# Are Some Cells "Full" of Ox-LDLIn Culture?

Lipid droplets are the most visible form of cellular lipid storage. A mature lipid droplet is described as a mass of neutral lipid within a single phospholipid leaflet membrane (98). Lipid droplets have a complex genesis and breakdown process within cells. Neutral lipids are generated enzymatically in the ER and can be trafficked to many parts of the cell including and the nucleus, lysosomes and vacuoles. Intracellular storage of lipids reflects a balance between uptake, consumption, interconversion of lipid forms and release (98, 99). Cellular stress and immune factors play significant roles in lipid balance (99).

The mechanisms underlying lipid balance in macrophages have been studied most extensively in non-neural diseases, such as atherosclerosis, non-alcoholic fatty liver disease (NAFLD) and pulmonary alveolar proteinosis (PAP), including the conversion of high levels of LDL to Ox-LDL related to foamy macrophage formation and function (29, 100). Ox-LDLs are not recognized by a macrophage's native LDL receptor, but they are recognized by the CD36 scavenger receptor (29, 75). Lipid uptake via scavenger receptors is not subject to feedback regulation and can lead to excessive accumulation of lipids in affected cells (75, 90, 101). Cholesterol esters are hydrolyzed enzymatically into free cholesterol, which can be utilized by cellular processes, exported from the cell, or converted back to cholesterol esters to prevent toxic effects from excess free cholesterol. Export of high-density lipoprotein is the primary means of lipid efflux from cells under normal conditions, though there is some passive free cholesterol efflux (29, 100).

Cholesterol ester hydrolysis may be actively inhibited by Ox-LDLs after prolonged exposure (102, 103). Exposure to Ox-LDLs can lead to lysosomal accumulation of lipids, rather than cytosolic accumulation of lipid droplets (102–104). However, there are species-specific differences in Ox-LDL response: free cholesterol and cholesterol esters were present in pigeon macrophage lysosomes after Ox-LDL exposure, while lipid accumulation was cytosolic in mouse macrophages (103). Lysosomal dysfunction caused by Ox-LDLs may not be easily reversed, which could explain why some of our foam cells don't take up much additional Ox-LDL in vitro (**Figure 7**). Other modified LDLs can be metabolized without adversely affecting essential lysosomal acidity and our foam cells are likely taking up many types of modified LDLs, but those with a high existing Ox-LDL burden would be inactive in the DiI-Ox-LDL uptake experiments (105, 106). Prolonged lysosomal dysfunction could be why our foam cells remain so full of lipid long after removal from the injury site. The absence of DiI-Ox-LDL uptake in ependymal cells in vitro (**Figures 7A,B**, **8A**) could reflect natural uptake of less oxidized forms of LDL in vivo, or a greater tendency toward lysosomal dysfunction.

#### ECM Degradation vs. Synthesis ECM Degradation

Cells migrate into the Axolotl cord lesion site through a collagenand proteoglycan-rich ECM (**Figures 1D–F**). The regenerating cord shows a massive increase in the amount of ECM investing the cord, compared with the normal cord meningeal ECM, with a disproportional increase in sulfated proteoglycan content (**Figures 1B,E**).

ECM is known to be phagocytosed and degraded intracellularly by a variety of amphibian cell types, including anuran (frog and toad) macrophages during the ECM turnover associated with metamorphosis (107). TEM studies, here, show foam cells are attached to, and embedded in, fibrillar collagen (**Figures 3B,D**). TEM examination shows phagocytosed fibrillar collagen in a foamy MNGC (**Figure 3D**; **Supplemental Figure 2A**). In an enlargement of **Figure 3B** (**Supplemental Figure 2A**) interstitial fibrillar collagen can be seen among a group of ependymal cells, but none of them showed phagocytosed collagen. Either phagocytosis of the fibrillar collagen is accomplished only by the foamy macrophages, or reactive ependymal cell collagen uptake is in the form of smaller peptides resulting from extracellular degradation. In **Figure 3F** an image from a site in a zone farther from the transection site where the interstitial collagen has already been removed: no intracellular collagen was seen.

The ependymal cathepsin K production described here, and ependymal MMP production characterized previously in Axolotl cord regeneration, show that ependymal cells can participate actively in ECM degradation mediated by secreted proteases [**Figure 11**; (20)]. Foamy macrophages are known to produce MMP9 and cathepsin K in diseases including atherosclerosis, and MMP9 in multiple sclerosis (53, 108, 109).

Axolotl foamy macrophages and smaller MNGCs participate in ECM removal by secretion of cathepsin K as well as phagocytosis (**Figures 3D**, **11A–C,E**; **Supplemental Figure 2A**). MNGCs are seen in or on the reactive meninges (**Figures 1G,H**, **2D**, **3A,B**) and are known to participate in ECM degradation in other system including MNGC tumors (32, 52). In our primary tissue culture system, smaller MNGC actively produce cathepsin K, while all of the foam cells do so (**Figure 11**). A full understanding of the ECM proteolytic repertoire of cord lesion site MNGCs will require much larger numbers in future studies involving stimulated fusion in vitro.

#### ECM Synthesis

The participation of ependymal cell endfeet in reforming the glia limitans has long suggested that they are producing ECM in later stages of the regeneration process (2). The expectation regarding ECM production by meningeal foam cells is not at all clear. In a few experimental disease-related models, a fibrotic role has been indicated for macrophages. This includes exposure of human macrophages to native LDL which stimulates proteoglycan secretion, a phenomenon with implications for vascular wall trapping of apolipoprotein Bcontaining lipoproteins in atherosclerotic plaques (110). This phenomenon may be related to the production of a proteoglycan form of macrophage colony stimulating factor in atherosclerotic plaque macrophages (111). In mice with surgical sponge-induced granulomas, and in atherosclerotic plaques, a variety of genes associated with fibrosis are expressed in transcriptome analysis: several collagen peptides and the small proteoglycans decorin and biglycan (112). This expression is higher in foamy macrophages vs. non-foamy macrophages (112).

The in vitro <sup>3</sup>H-glucosamine uptake results presented here suggest that ependymal cells are producing glycosaminoglycans and proteoglycans during the ependymal outgrowth process, depositing new ECM even as they are engaging in lesion site matrix turnover (**Figures 14A,C**). The foamy macrophages, however, show no evidence of ECM production from the <sup>3</sup>Hglucosamine uptake studies (**Figures 14B,C**).

#### Foamy Macrophage/Ependymal Interactions: Beneficial to Regeneration or Not?

In mammalian SCI, macrophages support scar formation and inhibit axonal regrowth (55, 113, 114). Macrophages are required in Axolotl limb and heart regeneration, so there is evidence that innate immune system cells can have a positive role in urodele regeneration (115, 116). In mouse heart, this positive role for macrophages does not extend beyond the neonatal stage (117).

Less is known about macrophages and macrophage/target cell interactions in urodele spinal cord regeneration. A study using lectin probes shows macrophages within the injured Axolotl spinal cord (36). In the newt studies by Zukor et al. (12), TEM examination shows macrophages in contact with neurons in an early stage of axonal regrowth called "wisping," as well as being present in the reactive meninges (12). In the Zukor et al., **Figure 7D**, a lipid-laden macrophage that could be a foam cells is shown in contact with an axon, and foam cell-like white blood cells are shown in the reactive meninges (12), **Figure 7C**. Though not identified as such, foamy macrophages may be present in the regenerating newt cord.

The present study examined the interaction between foamy macrophages and ependymal cells (**Figures 13**, **14**). Mononucleated foamy macrophages were abundant in Axolotl cord lesion site tissue in situ and in vitro (**Figures 2B,E**, **6C**, **13C**). Statistical analysis showed a highly significant association of the presence of foamy macrophages with an increased degree of dispersal of ependymal cells in vitro (**Figures 2E**, **5A,C**, **13**, **14**). It is not known whether the interaction involves cell-cell contact or secreted factors.

The co-migration of foamy macrophages and ependymal cells was strong during the period of adult Axolotl cord gap regeneration corresponding to mesenchymal ependymal outgrowth (2 weeks, **Figures 13A–D**), while at 4 weeks of regeneration, when the ependymal cells from cranial and caudal stumps have reconnected, the foam cells were no longer present (**Figure 13E**). It is not yet known whether the foamy macrophages and MNGCs are excluded to the periphery with the meninges, undergo cell death or both.

The position of these cells on and within the reactive meninges investing the regenerating stumps suggests a model in which the foam cells and smaller MNGCs act on the invasive meninges from the "outside-in," while the ependymal cells act from within the cord into the lesion site to remove fibrotic meningeal ECM [**Figures 1E–H**, **2A,B**, **15A**; (16, 20)]. Because they are attached to and within the invasive meninges, the foamy macrophages are concentrated in the lesion site distal to the zone of ependymal outgrowth. The ependymal cells grow out to meet and mix with the foamy macrophages.

Foamy macrophages can release growth factors, such as FGF2 and TGFβ, that are bound by ECM (118). Decellularized human meninges retains FGF2 and VEGF, suggesting that the release of ECM-bound growth factors by the foamy macrophages proteases could be of importance in the Axolotl cord lesion site (119). Urodele ependymal cells respond to TGFβ and FGF2 (6, 120).

The relationship of foamy macrophages to ependymal outgrowth, and their known activity, places the macrophages in a position to open the way for ependymal cells through ECM in vitro during ependymal outgrowth. In combination with the ECM degrading and synthetic capacity of the reactive ependymal cells, this juxtaposition of the foamy macrophages and ependymal cells could help maintain directional outgrowth across the lesion site.

#### DATA AVAILABILITY STATEMENT

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

#### ETHICS STATEMENT

The animal study was reviewed and approved by IUPUI School of Science Institutional Animal Care and Use Committee (SARC).

# AUTHOR CONTRIBUTIONS

NE, HT, HS, ME, TB-A, and EC contributed the conception and design of the study. SS developed the wholemount staining procedures and their application to this study. TB-A performed the statistical analysis. ME and DS performed the TEM analysis and interpretation. DS planned and performed the examination of the time course of meningeal regeneration. DS-B provided the

#### REFERENCES


immunological interpretation essential for experimental design. MK developed and applied the wholemount actin organization studies. TB-A, EC, and DS-B developed the diagrammatic interpretation of the lesion site, foamy macrophages, and MNGCs. EC, TB-A, and NE wrote, and EC and TB-A coordinated the revisions to the manuscript. HT and HS initiated this project as students in the laboratory of EC.

#### FUNDING

This work was funded by the Indiana University Foundation Regenerative Biology Fund I380010385.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02558/full#supplementary-material


Red O staining and DiI-oxLDL uptake. Cytotechnology. (2010) 62:473–81. doi: 10.1007/s10616-010-9290-0


regeneration. J Clin Invest. (2014) 124:1382–92. doi: 10.1172/JCI 72181


**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 Enos, Takenaka, Scott, Salfity, Kirk, Egar, Sarria, Slayback-Barry, Belecky-Adams and Chernoff. 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 CD40-ATP-P2X<sup>7</sup> Receptor Pathway: Cell to Cell Cross-Talk to Promote Inflammation and Programmed Cell Death of Endothelial Cells

#### Carlos S. Subauste1,2 \*

*<sup>1</sup> Division of Infectious Diseases and HIV Medicine, Department of Medicine, Case Western Reserve University, Cleveland, OH, United States, <sup>2</sup> Department of Pathology, Case Western Reserve University, Cleveland, OH, United States*

#### Edited by:

*Antje Kroner, Medical College of Wisconsin, United States*

#### Reviewed by:

*Rishein Gupta, University of Texas at San Antonio, United States Bruce Hostager, The University of Iowa, United States Steven F. Abcouwer, Michigan Medicine, United States*

> \*Correspondence: *Carlos S. Subauste carlos.subauste@case.edu*

#### Specialty section:

*This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Immunology*

Received: *28 May 2019* Accepted: *02 December 2019* Published: *17 December 2019*

#### Citation:

*Subauste CS (2019) The CD40-ATP-P2X*<sup>7</sup> *Receptor Pathway: Cell to Cell Cross-Talk to Promote Inflammation and Programmed Cell Death of Endothelial Cells. Front. Immunol. 10:2958. doi: 10.3389/fimmu.2019.02958* Extracellular adenosine 5′ -triphosphate (ATP) functions not only as a neurotransmitter but is also released by non-excitable cells and mediates cell–cell communication involving glia. In pathological conditions, extracellular ATP released by astrocytes may act as a "danger" signal that activates microglia and promotes neuroinflammation. This review summarizes *in vitro* and *in vivo* studies that identified CD40 as a novel trigger of ATP release and purinergic-induced inflammation. The use of transgenic mice with expression of CD40 restricted to retinal Müller glia and a model of diabetic retinopathy (a disease where the CD40 pathway is activated) established that CD40 induces release of ATP in Müller glia and triggers in microglia/macrophages purinergic receptor-dependent inflammatory responses that drive the development of retinopathy. The CD40-ATP-P2X<sup>7</sup> pathway not only amplifies inflammation but also induces death of retinal endothelial cells, an event key to the development of capillary degeneration and retinal ischemia. Taken together, CD40 expressed in non-hematopoietic cells is sufficient to mediate inflammation and tissue pathology as well as cause death of retinal endothelial cells. This process likely contributes to development of degenerate capillaries, a hallmark of diabetic and ischemic retinopathies. Blockade of signaling pathways downstream of CD40 operative in non-hematopoietic cells may offer a novel means of treating diabetic and ischemic retinopathies.

Keywords: CD40, glia, retina, endothelial cell, ATP, cytokine, diabetes

# INTRODUCTION

Glia orchestrate homeostasis in neural tissue through cell-to-cell interactions. Communication among glial subsets and communication between glia and other cells of the nervous system are also important during the development of disorders with an inflammatory component. ATP released by astrocytes appears to cause neuroinflammation by activating pro-inflammatory responses in microglia (1). Retinopathies caused by diabetes and ischemia are driven to a significant extent by chronic inflammation (2–4). The CD40 pathway is activated in these retinopathies and CD40 has emerged as a central mediator of inflammatory responses and pathology in these disorders (5–7). Herein I will review our work that identified CD40 expressed in retinal Müller glia as a trigger for secretion of ATP that in turn engages the P2X<sup>7</sup> receptor leading to pro-inflammatory cytokine production by monocyte/macrophages/microglia and programmed cell death of retinal endothelial cells (7, 8). Through this process, CD40 present in a non-hematopoietic cell amplifies inflammation and causes tissue pathology.

#### CD40

CD40 is a member of the TNF receptor superfamily that is expressed in various hematopoietic and non-hematopoietic cells including antigen-presenting cells (B cells, dendritic cells, and monocyte/macrophages), endothelial cells, epithelial cells, vascular smooth muscle cells, retinal Müller glia, fibroblasts, and neurons (6, 9–13). CD154 (CD40 ligand) is expressed primarily on activated CD4<sup>+</sup> T cells, platelets, and is also present as a biologically active soluble protein present in plasma (14, 15).

Studies in patients with congenital absence of functional CD154 (Hyper IgM syndrome, X-HIM) provide clinical evidence for the central role of this pathway in adaptive immunity (16). CD40-CD154 interaction promotes dendritic cell maturation inducing licensing of these cells for efficient T cell priming (14, 17, 18). This pathway stimulates IL-12 secretion by dendritic cells that in turn promotes CD4<sup>+</sup> T cell differentiation into Th1 cells (14, 17, 18). It also supports CD8<sup>+</sup> cytotoxic T lymphocytes (CTL) development and prevents CTL exhaustion (19). CD40-CD154 interaction promotes pro-inflammatory cytokine production by macrophages and activates effector functions that are central to control of intracellular pathogens (14, 18). Indeed, the most important clinical feature of patients with X-HIM is the increased susceptibility to opportunistic infections normally controlled by cell-mediated immunity (16). The CD40-CD154 pathway is also central for humoral immune responses including B cell proliferation, germinal center formation, antibody production, immunoglobulin class switch, and the generation of B cell memory (14, 17, 18).

In contrast to hematopoietic cells, little is known about the physiologic role of CD40 in non-hematopoietic cells. It has been proposed that CD40 promotes survival of neurons in the brain since old CD40−/<sup>−</sup> mice (16 months of age) have reduced expression of neurofilament isoforms and exhibit evidence compatible with increased neuronal programmed cell death (TUNEL<sup>+</sup> neurons) (12). In addition, in developing neural tissue, CD40 promotes axon growth in sympathetic neurons and has effects on dendrite growth that vary depending on the class of neurons: CD40 promotes dendrite growth in hippocampal excitatory neurons while it suppresses dendrite growth in striatal inhibitory neurons (20, 21). It is not known whether CD40 regulates the development and survival of retinal neurons. Moreover, the physiologic function of CD40 expressed in non-hematopoietic compartments in other organs is unclear. This may be explained by the low levels of CD40 expression in these compartments under basal conditions. In contrast, CD40 is upregulated in various inflammatory disorders and, through ligand engagement, CD40 triggers pro-inflammatory responses in endothelial cells, vascular smooth muscle cells and epithelial cells that play a key role in the pathogenesis of various disorders such as inflammatory bowel disease, systemic lupus erythematosus, rheumatoid arthritis, multiple sclerosis, graft rejection, and atherosclerosis (22, 23). These responses include increased protein expression of adhesion molecules, chemokines, metalloproteinases, and tissue factor (22, 23). The effects of CD40 ligation on retinal non-hematopoietic cells are discussed below.

# DIABETIC AND OTHER ISCHEMIC RETINOPATHIES

Diabetes mellitus has become one of the most important health problems in the world. It is estimated that there are 422 million patients with diabetes worldwide (World Health Organization; www.who.int/diabetes/global-report). Diabetic retinopathy (DR) is a major complication of diabetes and eventually occurs in ∼35% of patients with diabetes (24). In addition, DR is the most common cause of vision loss among working-age adults in developed countries (25). The development of DR appears to be multifactorial and mechanisms such as oxidative stress, increased polyol and hexosamine pathway flux, protein kinase C activation, increased formation of advanced glycation-end products and alterations in systemic and local lipid metabolism have been linked to the development of the disease (26, 27). Ample experimental data indicate that low-grade chronic inflammation also plays an important role in the development of DR (2–4).

The vitreous of patients with DR (28) and retinal endothelial cells from diabetic humans and rodents exhibit increased expression of ICAM-1, an event that promotes adherence of leukocytes to the retinal vasculature (leukostasis) (29, 30). This phenomenon is important since blockade of ICAM-1– CD18 interaction diminishes the development of degenerate capillaries in diabetic mice (31). These structures are a hallmark of early diabetic retinopathy and are formed as a consequence of the death of endothelial cells and pericytes, leading to the transformation of capillaries into collapsed sheaths of collagen/extracellular matrix structures that lack blood flow (32). The ensuing ischemia can promote transition to proliferative DR (PDR) that is characterized by retinal neovascularization. DR is also accompanied by increased expression of TNF-α and IL-1β (33–36). Microglia/macrophages express TNF-α in the diabetic retina (34). TNF-α and IL-1β play a pathogenic role in DR since they contribute to diabetes-induced degeneration of retinal capillaries (37, 38). Inducible nitric oxide synthase (NOS2) is expressed in the retinas of patients with DR and of diabetic rodents (39, 40). Furthermore, diabetic NOS2−/<sup>−</sup> mice have reduced retinal leukostasis and capillary degeneration (41, 42). CCL2 levels are increased in the vitreous fluid in patients with PDR (43) and in retinas of diabetic rodents (44). This chemokine appears to play a pathogenic role in DR since there is a correlation between CCL2 protein levels in the vitreous with the severity of DR (43).

Ischemic retinopathies including those caused by central retinal artery occlusion, retinal vein occlusion, and retarded retinal vascular development in premature infants are important causes of permanent visual impairment and blindness in adults and children (45–47). Like DR, inflammatory responses including TNF-α, IL-1β, nitro-oxidative stress, and chemokines likely play an important role in the pathogenesis of these diseases (48). Retinal injury induced by ischemia/reperfusion (I/R) is a commonly used animal model of ischemic retinopathy (49). I/Rinduced retinopathy is characterized by retinal inflammation, loss of ganglion cells, and development of capillary degeneration (49). I/R of the retina causes upregulation of ICAM-1, TNF-α, IL-1, NOS2, and COX-2 (5, 49–53). These responses are pathogenic since approaches to inhibit them are protective against retinal pathology (50–54).

#### CD40 IN THE DEVELOPMENT OF DIABETIC AND I/R-INDUCED RETINOPATHIES

CD40 is expressed in the retina at the level of endothelial cells, Müller glia (important macroglia in the retina), microglia, ganglion cells, and retinal pigment epithelial cells (5, 6, 55, 56). The levels of CD40 expression are low under basal conditions. However, induction or upregulation of CD40 expression is a feature of inflammatory disorders driven by CD40 (57). Indeed, CD40 mRNA is upregulated in the retina of mice with diabetes and mice subjected to retinal I/R (5, 6). Immunohistochemistry and flow cytometry studies to assess protein expression revealed that CD40 is upregulated in retinal endothelial cells, Müller glia and microglia of diabetic mice (6). Importantly, CD40−/<sup>−</sup> mice are protected from I/R-induced retinopathy and early diabetic retinopathy (5, 6).

Ligation of CD40 in retinal Müller glia upregulates ICAM-1, CCL2, NOS2 at the protein level and stimulates PGE<sup>2</sup> production (5, 6, 58). CD40 ligation in retinal endothelial cells upregulates ICAM-1 and CCL2 protein levels (5, 6, 58). Retinal endothelial cells also produce CXCL1 following CD40 stimulation, a response that is markedly potentiated by a low concentration of IL-1β (5).

CD40 is central for the development of retinal inflammation and retinopathy induced by I/R. In contrast to wild-type mice, CD40−/<sup>−</sup> mice subjected to I/R are protected from upregulation of ICAM-1, CXCL1, NOS2, and COX-2 mRNA levels (5). The reduced expression of NOS2 and COX-2 is explained at least in part by diminished recruitment of NOS2<sup>+</sup> COX-2<sup>+</sup> leukocytes into the retina of CD40−/<sup>−</sup> mice (5). Importantly, the loss of ganglion cells and the development of capillary degeneration are markedly attenuated in ischemic retinas of CD40−/<sup>−</sup> mice (5). The protection from development of ischemic retinopathy observed in CD40−/<sup>−</sup> mice is likely explained by diminished leukocyte infiltration and reduced expression of pro-inflammatory molecules since blockade of ICAM-1, NOS2, or COX-2 protect from retinal pathology after ischemia (51, 52, 54). Altogether, CD40 is a central mediator of inflammation and neuro-vascular degeneration after I/Rinduced injury of the retina. The model that likely explains these findings is as follows: ischemia-induced activation of CD40 in retinal endothelial cells triggers ICAM-1 and KC/CXCL1 upregulation leading to recruitment of NOS2 and COX-2 expressing leukocytes that would in turn promote neurovascular degeneration in the retina (5). However, it is also possible that Müller glia from ischemic retinas could be a source of increased NOS2 and/or COX-2 expression after activation via CD40.

The upregulation of CD40 and CD154 indicate that this pathway is activated in diabetes. CD40 protein expression in increased in the retina of diabetic mice and in the kidneys of patients with diabetic nephropathy (6, 59). CD40 mRNA levels are upregulated in the retinas of diabetic mice (6). Peripheral blood mononuclear cells from poorly controlled patients with type I diabetes exhibit increased mRNA levels of the functional type I isoform of CD40 (60). It is not known whether changes in micro RNA that control CD40 transcription [i.e., miR-155, miR-424, miR-503 (61, 62)] explain the upregulation of CD40 mRNA. In addition, CD154 protein levels are elevated in the blood from patients with diabetic microangiopathy and mice with diabetes (7, 63, 64). CD154 upregulation is biologically relevant since serum CD154 from diabetics triggers pro-inflammatory responses in endothelial cells and monocytes (63). It is likely that CD154 levels are also increased in the retina because microthrombosis occurs in diabetic retinopathy and activated platelets express CD154 (65).

CD40 is relevant to DR since diabetic CD40−/<sup>−</sup> mice are protected from upregulation of ICAM-1 in retinal endothelial cells, leukostasis, upregulation of TNF-α, IL-1β, and NOS2 mRNA levels, retinal protein nitration and elevated CCL2 mRNA levels in the retina (6, 7, 58). Importantly, diabetic CD40−/<sup>−</sup> mice do not develop capillary degeneration (6, 7). Taken together, CD40 is critical for development of various inflammatory responses in the diabetic retina and the development early DR (6, 7).

#### CD40 IN MÜLLER GLIA RECRUITS INFLAMMATORY RESPONSES IN BYSTANDER MICROGLIA/MACROPHAGES

Leukocytes are recognized key players in the development of inflammatory disorders. Indeed, expression of NOS2 or poly(ADP-ribosyl) polymerase 1 (PARP1) in bone marrow cells is necessary for the development of early DR (66). Similarly, CD40 in hematopoietic cells has been deemed a central driver of inflammation. However, studies in mice using bone marrow chimeras revealed that CD40 expressed in non-hematopoietic cells is also required for inflammation (5). Absence of CD40 in the retina inhibits ICAM-1 mRNA upregulation, leukocyte recruitment to the retina and neurovascular degeneration after I/R of the retina (5). Importantly, studies using transgenic mice have established that CD40 expression in a nonhematopoietic cell—Müller glia—is sufficient for development of an inflammatory disorder (7).

Müller glia link with neurons and capillaries, and are central to retina homeostasis (67, 68). Müller glia become dysfunctional and acquire expression of proinflammatory genes in diabetic and other ischemic retinopathies (69–71). The fact that Müller glia express CD40 raised the possibility that CD40 present in these cells may be an important activator of inflammation and retinal injury. Studies in transgenic mice that expressed CD40 restricted to Müller glia demonstrated that, after induction of diabetes, the presence of CD40 in these cells was sufficient for upregulation of ICAM-1, NOS2, TNF-α, IL-1β, CCL2 mRNA levels as well as for development of leukostasis and capillary degeneration (7). This work identified CD40 in Müller glia as a central regulator of inflammation and development of early diabetic retinopathy.

Despite the fact that CD40 in Müller glia from diabetic mice drives TNF-α and IL-1β in vivo, work done in vitro revealed that human and rodent Müller glia are unable to secrete these pro-inflammatory cytokines in response to CD40 ligation even though these cells react to CD40 stimulation (CCL2 secretion and ICAM-1 protein upregulation) (7). This apparent discrepancy raised the possibility that CD40 in Müller glia acts on bystander microglia/macrophages to promote expression of TNF-α and IL-1β.

Testing whether Müller glia activated by CD40 induce IL-1β and TNF-α production in bystander monocytes/macrophages was done by adding human CD154 to human CD40<sup>+</sup> Müller glia incubated with CD40<sup>−</sup> human monocytic cells (to avoid the effects of direct CD40 ligation on these cells), or by adding human CD154 to human CD40-expressing mouse Müller glia incubated with mouse macrophages (human CD154 does not stimulate mouse CD40 expressed in macrophages) (7). While Müller glia and monocyte/macrophages failed to secrete TNF-α and IL-1β in response to CD154, addition of CD154 to the co-culture of these cells triggered TNF-α and IL-1β production (7). The in vitro studies have an in vivo correlate since diabetic mice that express CD40 restricted to Müller glia upregulate TNF-α protein levels in microglia/macrophages but not in Müller glia while the latter cells upregulate CCL2 protein levels (7). Taken together, these studies revealed that Müller glia activated by CD40 induce proinflammatory responses in bystander microglia/macrophages.

#### THE CD40-ATP-P2X7 PATHWAY AND INFLAMMATORY RESPONSES IN BYSTANDER MICROGLIA/MACROPHAGES

ATP functions not only as a neurotransmitter for neurons but can also be secreted by non-excitable cells (72, 73). Moreover, various cell types express P2 purinergic receptors. These receptors are divided into ATP-gated ionotropic P2X receptors and metabotropic, G protein-coupled P2Y receptors (72, 73). The seven subtypes of P2X receptors are ligand-gated channels permeable to Ca2+, Na+, and K+. P2X<sup>7</sup> receptor is characterized by the ability to form large trans-membrane pores in response to repetitive or prolonged exposure to ATP (72, 73). P2X<sup>7</sup> receptor is key for IL-1β and TNF-α secretion by microglia/macrophages stimulated with ATP (74, 75). Indeed, secretion of ATP by astrocytes may cause P2X7-dependent microglial activation that would drive neuroinflammatory and degenerative disorders (76).

In vitro and in vivo studies were conducted to determine whether CD40 acts through ATP-P2X<sup>7</sup> signaling to induce cytokine production in bystander myeloid cells. These studies showed that CD40 is an inducer of ATP release in Müller glia (7). Moreover, purinergic signaling explains TNF-α and IL-1β secretion in bystander monocytes/macrophages incubated with Müller glia activated by CD40. Blockade of the P2X<sup>7</sup> receptor either by pharmacologic approaches, knockdown of P2X<sup>7</sup> or the use of macrophages from P2X−/<sup>−</sup> <sup>7</sup> mice results in marked inhibition of TNF-α and IL-1β secretion (7). In addition, a purinergic receptor ligand (Bz-ATP) enhances cytokine production by monocytic cells (7).

As described above, studies in diabetic transgenic mice that express CD40 only in Müller glia revealed that TNF-α is expressed in a distinct compartment—microglia/macrophages (7). Moreover, P2X<sup>7</sup> receptor mRNA levels are enhanced in the retinas of diabetic mice and P2X<sup>7</sup> receptor protein expression is increased in microglia/macrophages from these animals (7). This is relevant since increased levels of P2X<sup>7</sup> receptor facilitate the effects of the receptor (77). Mice treated with the P2X<sup>7</sup> receptor inhibitor BBG as well as P2X−/<sup>−</sup> <sup>7</sup> mice are protected from diabetes-induced upregulation of IL-1β and TNFα mRNA levels (7). The mice are also protected from increased expression of ICAM-1 and NOS2, molecules that are upregulated by IL-1β and TNF-α (78, 79). Taken together, Müller glia activated by CD40 secrete extracellular ATP and drive P2X<sup>7</sup> receptor-dependent pro-inflammatory cytokine expression in bystander microglia/macrophages in vitro and in vivo (**Figure 1** and **Table 1**). These findings support a model whereby CD40 engagement in non-hematopoietic cells triggers inflammatory

in Müller glia are increased in diabetes. CD40 ligation in Müller glia causes PLCγ1 (PLC) activation and secretion of extracellular ATP. P2X<sup>7</sup> receptor is upregulated in microglia/macrophages in the diabetic retina. ATP binds P2X<sup>7</sup> receptor leading to secretion of TNF-α and IL-1β. Created in part with BioRender.com.


responses not only in these cells (i.e., chemokine and adhesion molecule upregulation) but also amplifies inflammation by enabling bystander myeloid cells to secrete pro-inflammatory cytokines in a manner dependent on ATP-P2X<sup>7</sup> receptor.

Increased intracytoplasmic Ca2<sup>+</sup> triggers ATP release (80) and CD40 elevates intracytoplasmic Ca2<sup>+</sup> levels (81, 82). Indeed, BAPTA-AM, a chelator of intracellular Ca2+, impairs CD40 mediated ATP release in Müller cells (7). In addition, CD40 ligation in Müller glia causes rapid Tyr783 phosphorylation of phospholipase Cγ1 (PLCγ1) (7), a signaling molecule that increases intracytoplasmic Ca2<sup>+</sup> (83), and pharmacologic inhibition of PLC impairs ATP release by Müller glia activated by CD40 (7). Thus, CD40 ligation phosphorylates PLCγ1 and CD40 likely functions via PLCγ1 to trigger ATP release in Müller glia.

# CD40 IN MÜLLER GLIA AND PROGRAMMED CELL DEATH OF BYSTANDER RETINAL ENDOTHELIAL CELLS

Retinal endothelial cells undergo programmed cell death (PCD) in the diabetic retina (32, 84–86). This process would contribute to the development of capillary degeneration, a central feature of early diabetic retinopathy (32). CD40 is necessary for the development of capillary degeneration (6, 7) and yet, ligation of CD40 in endothelial cells does not induce PCD likely because CD40 typically triggers pro-survival signals (87). This raised the possibility of CD40 promoting death of retinal endothelial cells by acting through other cells of the retina. Müller cells were a likely culprit since they encircle retinal endothelial cells.

Whereas direct CD40 ligation in retinal endothelial cells does not cause PCD, CD40 stimulation enhances PCD of endothelial cells when they are incubated with CD40<sup>+</sup> Müller cells (8). This effect is not driven by NOS2, oxidative stress, TNF-α, IL-1β, or Fas ligand (8). As described above, CD40 ligation in Müller glia increases release of ATP. CD40 ligation in retinal endothelial cells upregulates P2X<sup>7</sup> receptor expression making these cells susceptible to ATP-induced PCD (8). Indeed, pharmacologic inhibition of P2X<sup>7</sup> receptor prevents PCD of the endothelial cells (8). These results are consistent with the ability of the P2X<sup>7</sup> to form trans-membrane pores that are permeable to hydrophilic molecules of up to 900 Da (88) and mediate cell death (89, 90). The in vitro studies described above have an in vivo correlate since retinal P2X<sup>7</sup> mRNA levels and P2X<sup>7</sup> receptor expression in retinal endothelial cells are increased in diabetic mice in a CD40-dependent manner (8), CD40 appears to be necessary for PCD of retinal endothelial cells from diabetic mice (8), and CD40 is known to be required for retinal capillary degeneration (6, 7). Taken together, CD40 has a dual role in promoting PCD of retinal endothelial cells: it causes release of extracellular ATP by Müller glia and makes retinal endothelial cells susceptible to P2X7-driven PCD (**Figure 2** and **Table 1**). The latter effect may be explained by CD40-driven upregulation of the P2X<sup>7</sup> receptor in endothelial cells that would overcome the prosurvival signals activated by CD40 ligation. This mechanism may contribute to increased susceptibility to ATP-mediated PCD that appears to occur in diabetes (91). Other potential mechanisms by which CD40 increases susceptibility to P2X<sup>7</sup> receptor-mediated PCD may include modulation of ATP-gated channel expression, ectoATPase activity, and/or coupling to downstream cell signaling pathways that promote cell death. Finally, while CD40 induced activation of ATP-P2X<sup>7</sup> receptor signaling mediates PCD of retinal endothelial cells, CD40 may also promote death of these cells and capillary degeneration through mechanisms that include: enhancement of retinal leukostasis, upregulation of NOS2, TNF-α, and IL-1β in the retinas of diabetic mice (6, 7), events linked to PCD of retinal endothelial cells and capillary degeneration (31, 37, 38, 40, 42, 86).

In summary, the studies discussed here discovered the CD40- ATP-P2X<sup>7</sup> receptor pathway and revealed that this pathway links

a macroglia to microglia/macrophages and endothelial cells for the induction of inflammatory responses and endothelial cell death, respectively. By enabling myeloid cells to secrete TNFα and IL-1β, this process would circumvent the poor capacity of CD40 to directly trigger secretion of these cytokines in nonhematopoietic cells, thus causing amplification of inflammation. These findings may be operative in I/R-induced retinopathy given its similarity to DR. The CD40-ATP-P2X<sup>7</sup> receptor pathway may also be relevant to neuro-inflammatory and neuro-degenerative brain disorders. For example, astrocytes acquire CD40 expression after incubation with IFN-γ (92), neural tissue injury or in a transgenic mouse model of amyotrophic lateral sclerosis (93), a disease driven by CD40. Thus, the CD40-ATP-P2X<sup>7</sup> receptor pathway may potentiate pro-inflammatory cytokine production by microglia further driving neuro-inflammation.

#### REFERENCES


Finally, this pathway may be functional in other diseases driven by CD40 such as inflammatory bowel disease, atherosclerosis and lupus nephritis. CD40 present in non-hematopoietic cells of the intestine, blood vessels and kidney may induce release of ATP that would bind purinergic receptors present in infiltrating myeloid cells.

The existence of the CD40-ATP-P2X<sup>7</sup> receptor pathway may have therapeutic implications. Pre-clinical data revealed that administration of anti-CD154 mAb to inhibit CD40- CD154 signaling effectively controlled various inflammatory and neurodegenerative disorders (22, 23). Unfortunately, anti-CD154 mAbs caused thromboembolic complications in humans that are unrelated to inhibition of CD40 (94). Targeting signaling pathways downstream of CD40 may represent an alternative approach to treat CD40-driven diseases. CD40 functions by recruiting TNF Receptor Associated Factors (TRAF) to its TRAF2,3 or TRAF6 binding sites (95). Blockade of CD40-TRAF2,3 signaling markedly impairs pro-inflammatory responses in non-hematopoietic cells (58, 96). Blocking this signaling pathway may also inhibit pro-inflammatory responses in neighboring myeloid cells. Pharmacologic approaches to inhibit CD40-TRAF2,3 signaling (cell penetrating CD40- TRAF2,3 blocking peptide or small molecule CD40-TRAF2,3 inhibitor) may prove an effective approach to treat diabetic and ischemic retinopathies, and potentially other CD40-driven inflammatory disorders. Given that TRAF6 is critical for dendritic cell maturation and development (96, 97), CD40 mediated IL-12 production by dendritic cells (98) and induction of antimicrobial effector mechanisms in macrophages (99, 100), pharmacologic inhibition of CD40-TRAF2,3 signaling would minimize the risk of opportunistic infections by leaving CD40-TRAF6 signaling intact.

#### AUTHOR CONTRIBUTIONS

The author confirms being the sole contributor of this work and has approved it for publication.

#### FUNDING

CS was funded by NIH-R01 EY018341 and NIH-R01 EY019250.

#### ACKNOWLEDGMENTS

The author thanks all the members of the Subauste lab for their feedback on this manuscript.


proliferative lupus glomerulonephritis. Arthritis Rheum. (2003) 48:719–27. doi: 10.1002/art.10856


**Conflict of Interest:** The author declares 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 Subauste. 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 Chitinases as Biomarkers for Amyotrophic Lateral Sclerosis: Signals From the CNS and Beyond

Nayana Gaur <sup>1</sup> \*, Caroline Perner 1,2, Otto W. Witte1,3 and Julian Grosskreutz 1,3

<sup>1</sup> Hans Berger Department of Neurology, Jena University Hospital, Jena, Germany, <sup>2</sup> Center for Immunology and Inflammatory Diseases, Massachusetts General Hospital, Charlestown, MA, United States, <sup>3</sup> Jena Center for Healthy Ageing, Jena University Hospital, Jena, Germany

Amyotrophic lateral sclerosis (ALS) is a late-onset neurodegenerative condition, most widely characterized by the selective vulnerability of motor neurons and the poor life expectancy of afflicted patients. Limited disease-modifying therapies currently exist, which only further attests to the substantial heterogeneity associated with this disease. In addition to established prognostic factors like genetic background, site of onset, and age at onset, wide consensus on the role of neuroinflammation as a disease exacerbator and driver has been established. In lieu of this, the emerging literature on chitinases in ALS is particularly intriguing. Individual groups have reported substantially elevated chitotriosidase (CHIT1), chitinase-3-like-1 (CHI3L1), and chitinase-3-like-2 (CHI3L2) levels in the cerebrospinal, motor cortex, and spinal cord of ALS patients with multiple—and often conflicting—lines of evidence hinting at possible links to disease severity and progression. This mini-review, while not exhaustive, will aim to discuss current evidence on the involvement of key chitinases in ALS within the wider framework of other neurodegenerative conditions. Implications for understanding disease etiology, developing immunomodulatory therapies and biomarkers, and other translational opportunities will be considered.

#### Edited by:

Sven G. Meuth, University Hospital Münster, Germany

#### Reviewed by:

Stefanie Schreiber, University Hospital Magdeburg, Germany Marc Pawlitzki, University Hospital Münster, Germany

#### \*Correspondence:

Nayana Gaur nayana.gaur@med.uni-jena.de

#### Specialty section:

This article was submitted to Multiple Sclerosis and Neuroimmunology, a section of the journal Frontiers in Neurology

Received: 28 November 2019 Accepted: 14 April 2020 Published: 27 May 2020

#### Citation:

Gaur N, Perner C, Witte OW and Grosskreutz J (2020) The Chitinases as Biomarkers for Amyotrophic Lateral Sclerosis: Signals From the CNS and Beyond. Front. Neurol. 11:377. doi: 10.3389/fneur.2020.00377 Keywords: neurodegeneration, biomarker (BM), neuroinflammation, chitinases, amyotrophic lateral sclerosis (ALS)

#### INTRODUCTION

#### ALS and Neuroinflammation

Amyotrophic lateral sclerosis (ALS) is the most prevalent form of adult-onset motor neuron disease and clinically presents with the relentless destruction of primarily (but not exclusively) upper and lower motor neurons (UMN, LMN). Riluzole, the sole treatment available, confers only modest effects via a median increase of 2–3 months in survival; most patients eventually succumb to respiratory failure. Although there is a pressing need for treatment modalities that tackle disease aggressiveness, therapeutic development has been severely constrained by the disease's characteristic heterogeneity; this stems from age-at-onset and site-of-onset, presence of disease-associated mutations, and comorbidities, including frontotemporal dementia (FTD) (1). Progression and survival rates are also highly variable; while the median survival is 2–3 years from symptom onset, some patients present with a disease duration of over 10 years (2). Cellular and animal studies have provided elegant evidence that neuroinflammation contributes to ALS pathology and that concomitant glial dysregulation is necessary for motor neuronal

**132**

degeneration (3–5). Numerous immunological changes, including the functional alteration and pro-inflammatory phenotype of circulating myeloid cells (6), dysregulated leukocytic chemokine receptor expression (7), the reduction of regulatory T cells (8), and cytotoxic T cell infiltration, have also been reported in patients (9).

Despite this, there remains a paucity of biological tools that adequately capture the neuroinflammatory response across the disease; this may partially explain the failure of immunomodulatory therapies to date. Biomarkers that reflect target engagement and assess the efficacy of novel treatments are therefore crucial. Although molecular imaging studies of microglial activation are underway, fluid-based biomarkers are more accessible and can provide important insights into disease pathomechanisms. For instance, cerebrospinal fluid (CSF) and humoral levels of the neurofilament proteins have been validated as robust diagnostic and prognostic markers for ALS. Several inflammatory cytokines have also been reported as dysregulated in ALS, including TNF-α, MCP-1, and IL-6 (10–12). In lieu of this, recent reports of elevated chitinase levels in ALS are particularly interesting, as these have already been reported as surrogate markers of a chronic inflammatory response in non-neuronal conditions.

#### Mammalian Chitinases: Novel Players in Neurodegeneration?

The chitinases belong to the family 18 glycosyl hydrolases (GH18) and are characterized by their ability to cleave chitin, a natural polysaccharide found in the coating of various pathogens. The GH18 family is ubiquitously expressed across a wide range of organisms, from bacteria to humans; evolutionary conservation in the latter is particularly interesting, given the lack of endogenous chitin synthesis. This has led to the view that chitin is a defense target for the mammalian immune system or an "immune stimulator." Indeed, it is recognized by several pattern recognition receptors and can trigger associated immune responses in a fragment-size and tissue-dependent manner (13). Mammalian chitinases include the enzymatically active chitinases chitotriosidase (CHIT1) and acidic mammalian chitinase (AMCase) that can degrade chitin, and the chilectins (CLs) chitinase 3-like 1 and -like 2 (CHI3L1, CHI3L2). Despite being able to bind chitin with high affinity, the CLs possess no chitinolytic activity, owing to the absence of the catalytic motif. CHIT1 is primarily expressed by cells of myeloid lineage, particularly mature macrophages (14, 15). Like CHIT1, CHI3L1 is absent in monocytes and strongly upregulated during later stages of macrophage differentiation (16). CHI3L1 is also produced by reactive astrocytes and associated with chronic neuroinflammation, as will be further discussed in the Section Chitinases Across the ALS-FTD Spectrum (17–19). While CHI3L2 hasn't been as extensively studied, expression has been noted in chondrocytes, synoviocytes, and alternatively activated "M2" macrophages (20).

Although the exact roles of these moieties remain to be fully elucidated, it is clear that they extend beyond innate immunity against chitin-containing pathogens. Chitinases have been reported in the context of adaptive Th2 response mediation (21, 22), tissue remodeling and repair, and, most recently, oligodendrogenesis (23). Dysregulated chitinase levels have been reported in several chronic neurodegenerative conditions, including Alzheimer's disease (AD) and FTD. In vitro evidence suggests that, at least in ALS, they may act in a "feed-forward" loop that sustains neuroinflammation and exacerbates disease, as illustrated in **Figure 1**. For instance, in a transgenic rat model, TDP-43 induced astrocytic CHI3L1 up-regulation; in turn, synthetic CHI3L1 caused neuronal death in a dosedependent manner (19). Similarly, Raju et al. reported that CSF from ALS patients impacted cell viability and upregulated CHIT1 expression in murine microglial cultures (24). Subsequent exposure to CHIT1 itself caused microglial activation, indicating again a "self-propagating" inflammatory mechanism (25).

This review, while not exhaustive, will summarize current evidence for chitinase dysregulation in ALS and its implications for understanding disease etiology and progression, and therapeutic and biomarker development. CHIT1, CHI3L1, and CHI3L2 will be focused on, since these have been most extensively studied in a neurodegenerative context.

# THE CHITINASES IN ALS

# Evidence Concerning Diagnostic Potential

Varghese et al. (26) were the first to report chitinases in the context of ALS; using quantitative mass spectrometry (MS) and ELISA-based validation in an independent cohort, they showed that CSF levels of CHIT1, CHI3L1, and CHI3L2 were significantly elevated in ALS patients relative to healthy controls (HCs). This elevation has since been confirmed by several studies using a range of proteomic and transcriptomic methods (27–33). Recent studies have predominantly focused on assessing discriminatory power with regard to mimic conditions, and other neurodegenerative diseases. However, studies have differed with respect to (a) the chitinases and secondary targets investigated; (b) cohort demographics; (c) bio-fluids assessed, and (d) experimental and analytical methods used (**Table 1**). Thompson et al. subsequently investigated all three chitinases and reported that they were significantly higher in the CSF of ALS patients relative to HCs, mimics, and asymptomatic mutation carriers (MCs) and that increased CHIT1 levels corresponded to active forms of the enzyme. While all three could reliably distinguish ALS from HCs and mimics, they were outperformed by pNfH. Furthermore, all three chitinases performed poorly in distinguishing ALS from primary lateral sclerosis (PLS) (28). A

**Abbreviations:** AD, Alzheimer's Disease; ALS, Amyotrophic Lateral Sclerosis; ALSFRS-R, Amyotrophic Lateral Sclerosis Functional Rating Scale Revised; CHIT1, Chitotriosidase; CHI3L1, Chitinase 3-Like 1; CHI3L2, Chitinase 3-Like 2; CLs, Chi-lectins; fALS, Familial ALS; FTD, Frontotemporal Dementia; gALS, genetic ALS; GH18, family 18 Glycosyl Hydrolases; HCs, Healthy Controls; IFNγ, Interferon gamma; IL-6, Interleukin 6; LMN, Lower Motor Neuron; LPS, Lipopolysaccharide; MCs, Mutation Carriers; MCP-1, Monocyte Chemoattractant Protein 1; MS, Mass Spectrometry; NDCs, Neurological Disease Controls; NfL, Neurofilament Light Chain; PLS, Primary Lateral Sclerosis; pNfH, Phosphorylated Neurofilament Heavy Chain; PR, Progression Rate; sALS, sporadic ALS; sAPPβ, soluble Amyloid Precursor Protein Beta; Th2, T Helper Type 2; TNF-α, Tumor Necrosis Factor alpha; TREM2, Triggering Receptor expressed on Myeloid Cells 2; UMN, Upper Motor Neuron.

prior MS-based study by the same group also noted a modest fold change for only CSF CHIT1 and CHI3L2 between ALS and PLS (31). Similarly, while Steinacker et al. (33) recommended CHIT1 as a potential differential diagnostic marker for ALS, they also noted that levels were increased in other neurodegenerative conditions and that pNfH and NfL had superior discriminatory power. In the same vein, Gille et al. (10) reported that elevated CSF CHIT1 and CHI3L1 levels were only weakly specific to ALS patients relative to neurological disease controls (NDCs). Observations of significant ALS-associated chitinase elevations in blood have been limited, barring one study that reported significantly elevated CHIT1 activity in dried blood spots (30) and another that noted higher CHIT1 levels in a genetic ALS (gALS) cohort (27) (both relative to HCs). This, coupled with reports of poor correlations between peripheral and CSF chitinase levels, makes a blood-based marker unlikely.

Applicability as stand-alone diagnostic markers is also likely to be constrained by the effect of functional variants. For instance, polymorphisms in the CHI3L1 locus contribute to almost 15% of the variance in CSF CHI3L1 levels (34). Likewise, duplication in exon 10 of the CHIT1 gene reduces both expression and activity; although this polymorphism is highly prevalent in European populations, no significant differences in genotype frequency have been observed between ALS patients and healthy individuals (27, 30). Additionally, presence of the CHIT1 polymorphism has no influence on neurofilament levels or age of onset in patients, making a causative role in ALS pathogenesis unlikely. Importantly however, both CHIT1 expression and activity are significantly elevated in ALS patients (relative to HCs) independent of genotype and other factors like gender and age, indicating that disease status—rather than the presence of the polymorphism—determines the extent of dysregulation (12, 27, 30).

#### Evidence Concerning Prognostic Potential

The prognostic potential of the chitinases has been examined in relation to several clinical outcomes, including disease severity (overall ALSFRS-R score), the ALSFRS-R-derived progression rate (PR), survival, and disease duration, with several conflicting results as discussed below. The majority of the results discussed here focus on CSF, as almost no robust and consistent links between blood chitinase levels and prognostic factors have been

#### TABLE 1 | An overview of recent studies investigating chitinases in the context of ALS.


(Continued)

Chitinases in

Amyotrophic Lateral Sclerosis


Amyotrophic Lateral Sclerosis; GFAP, glial fibrillary acidic protein; gFTD, genetic Frontotemporal Dementia; HCs, Healthy Controls; IB, Immunoblotting; IHC, Immunohistochemistry; IL-6, interleukin 6; LC-MS, Liquid ChromatographyMass Spectrometry; MCs, Mutation Carriers; MCP-1, Monocyte chemoattractant protein 1; MS, Mass Spectrometry; NDCs, Neurological Disease Controls; NDegs, Neurodegenerative Disease Controls; NfL, neurofilament light chain;PLS, Primary Lateral Sclerosis; PMT, Post-mortem Tissue; pNfH, phosphorylated neurofilament heavy chain; RT-qPCR, Real-time Quantitative Polymerase Chain Reaction; sALS, sporadic ALS; sAPPβ, soluble amyloid precursor proteinbeta; sFTD, sporadic Frontotemporal Dementia; TNF-α, tumor necrosis factor alpha.

reported. It is worth noting, however, that studies have only now begun to examine CHIT1 enzymatic activity in addition to protein levels and that links between the periphery and prognostic factors, as reported by Pagliardini et al. (30), may yet emerge.

#### Links With Disease Severity and Progression

Evidence for a link with disease severity and progression has been tenuous at best. Martinez-Merino et al. (12) controlled for CHIT1 genotype and reported that while ALS patients had significantly elevated CHIT1 activity, it correlated with neither disease severity nor progression. Thompson et al. (28) reported a significant albeit modest correlation between CHIT and CHI3L2 levels—but not CHI3L1—and PR after controlling for gender, age at onset, and site of onset; however, a stronger correlation was noted for pNfH. Conversely, Illán-Gala et al. (32) and Andres-Benito et al. (35) reported that CSF CHI3L1 levels correlated with PR to almost the same degree as CSF NfL levels.

Gille et al. (10) noted that both CSF CHIT1 and CHI3L1 only weakly correlated with PR at time of sampling; however, "fast" progressors had significantly higher levels of CHIT1 and CHI3L1 than "slow" progressors. One study reported that CSF CHIT1 also significantly correlated with both disease severity and PR (inversely) and to almost the same magnitude as NfL and pNfH. However, these correlations did not persist when patients were stratified based on PR, despite "fast" progressors having significantly higher levels of CHIT1 (33). Chen et al. (29) too reported no significant differences in CHIT1 levels between PR-stratified patients.

It is worth noting that establishing any association between the chitinases and PR is likely confounded by the lack of any external consensus on the thresholds for "high" or "low" PR. These are often arbitrarily set based on individual cohorts, thus constraining inter-study comparability and potentially occluding genuine biological signals.

#### Links With Disease Duration

Evidence for an association with disease duration has also been inconsistent, even by the few studies that have included longitudinal sampling. CSF CHIT1 activity did not significantly differ between patients stratified based on time since onset to sampling (12). A MS-based study reported a small increase in CSF CHI3L1 levels over time in patients who had low levels at onset (31). However, a subsequent ELISA-based verification noted that CSF chitinase levels in ALS and PLS patients did not significantly increase over a follow-up period of ∼2 years, even when patients were stratified by PR (28). Similarly, no significant associations between CSF CHIT1 and CHI3L1 and disease duration were observed in a cohort of 105 ALS patients (10). Indeed, evidence from asymptomatic ALS and FTD MCs suggests that chitinase elevation is a feature of the early symptomatic phase of the disease and is unlikely by itself to trigger disease onset, given that no significant differences were observed between patients with either genetic or sporadic disease (27).

#### Links With Survival and Mortality

Studies examining survival have also reported discrepant results. Di Rosa et al. analyzed microarray datasets and reported that patients with a shorter survival had significantly higher CHI3L1 and CHI3L2 in their motor cortex than those that survived longer; levels also inversely correlated with survival in the entire patient cohort (36). Cox proportional hazards analyses have also revealed a significant association between CSF CHIT1 levels and mortality, while one study reported that the association was independent of pNfH levels, another by the same group reported the opposite (28, 31). However, neither study had data on other prognostic factors, e.g., respiratory and C9orf72 status, thus precluding a definitive conclusion on the influence of CHIT1. In contradiction, Gille et al. (10) reported that CSF CHI3L1, but not CHIT1, significantly affected mortality; this is compelling because they included data for eight established prognostic markers. The authors did not however compare how the chitinases performed relative to neurofilaments. Building on this, Illán-Gala et al. (32) also reported that increased CSF CHI3L1 levels were associated with shortened survival, even after adjustment for sex, age at onset and site of onset, NfL levels, and ALSFRS-R score at time of sampling. Taken together however, the currently available evidence doesn't unequivocally establish the degree to which the chitinases influence survival and whether they outperform established prognostic factors.

#### Links With Additional Indices

Although data are limited, some studies have also begun to examine a wider range of clinical outcomes; for instance, peripheral CHIT1 activity was significantly inversely correlated with forced vital capacity (30). Additionally, CSF CHIT1 and CHI3L1 levels correlated with the number of regions clinically affected by both UMN and LMN and only UMN degeneration, respectively (10, 28). Frontotemporal cortical thickness, as assessed by structural MRI, directly correlated with the CSF sAPPβ:CHI3L1 ratio in both ALS and FTD patients (32). Finally, whether chitinase levels also reflect the poorer outcomes associated with factors like bulbar onset or genetic status (e.g., C9orf72) needs further investigation.

#### CHITINASES ACROSS THE ALS-FTD SPECTRUM

Studies focusing on the broader ALS-FTD spectrum have noted that the two conditions present with specific chitinase dysregulation patterns. When examined alongside glial activation markers, these suggest different underlying inflammatory processes: increased microglial (as evidenced by CHIT1) and astroglial (as evidenced by CHI3L1) activation in ALS and FTD, respectively.

For instance, although CSF CHIT1 is elevated in FTD patients relative to both HCs and asymptomatic MCs, it is significantly higher in ALS patients (27, 33). Furthermore, CHIT1 immunostaining in post-mortem spinal cord tissue was observed only in ALS cases, where it co-localized with IBA1-positive microglia and CD68-positive macrophages, and not in other neurodegenerative disorders, including FTD and AD (27, 33). Conversely, despite considerable overlap, CSF CHI3L1 levels were higher in patients with sporadic FTD relative to those with sporadic ALS, albeit only slightly. CHI3L1 elevation also correlated with cognitive dysfunction, as assessed by the Edinburgh Cognitive and Behavioral ALS Screen (ECAS), suggesting that it skews more closely to the FTD phenotype (28). Illán-Gala et al. (32) reported that although neither absolute CSF CHI3L1 levels nor the sAPPβ:CHI3L1 ratio significantly differed between FTD and ALS patients, CHI3L1 and global cognitive performance only correlated in the FTD subgroup. Furthermore, a robust inverse correlation was noted between the sAPPβ:CHI3L1 ratio and the FTD-Clinical Dementia Rating score in FTD patients. CHI3L1 immunoreactivity has been observed in astrocytes, but not microglia and neurons; its expression correlates with GFAP, particularly in acute inflammatory conditions like multiple sclerosis, suggesting that CHI3L1 is indicative of reactive astrocytosis (18, 19, 37). Crucially, negligible CHI3L1-positive astrocytes were observed in post-mortem ALS cortical tissue and no significant differences in GFAP mRNA in the spinal cord were noted between ALS patients and HCs (18, 35). CSF GFAP levels were also significantly increased in FTD patients while they were unaffected in ALS patients (27).

In summary, while the chitinases may not be specific markers for either condition, they allude to distinct neuroinflammatory profiles. If corroborated by other modalities, e.g., PET imaging (38), these profiles could help delineate the underlying pathology and provide specific targets for immunomodulatory therapy.

#### CHITINASES IN THE BROADER NEUROINFLAMMATORY AND NEURODEGENERATIVE MILIEU

While much remains unknown about their cellular origin, it is evident that chitinase expression is not exclusive to ALS. It has been noted in multiple neurodegenerative conditions, where it predicts both clinical severity and longterm risk (39–41). The chitinases also robustly correlate with established neurodegenerative markers, including, e.g., the neurofilaments (10, 27, 31) and both total and phosphorylated tau (40, 42). Studies investigating multivariate panels have additionally reported close links to other inflammatory mediators. For instance, CSF chitinase levels correlated with MCP-1, and C-reactive protein in ALS patients and soluble TREM2 in cognitively unimpaired individuals (10, 28, 43). Transcriptomic studies have shown that CHIT1 correlates with IL-16, IL-18, and CHI3L1 and CHI3L2 with complement C1s subcomponent (36, 41). Therefore, it is probable that the chitinases reflect the inflammation that is characteristic of the wider neurodegenerative process. Given the evidence from post-mortem co-localization studies and that significant dysregulations have been primarily observed in CSF rather than blood, we further speculate that the chitinases are proxies for reactive gliosis. It is worth noting, however, that systemic conditions may also influence chitinase levels, potentially "masking" alterations in blood.

While there is considerable overlap between neurodegenerative conditions, expression patterns differ, underscoring the different pathomechanisms at play; for instance, while CSF CHI3L1 increases as cognitive deficits worsen along the AD continuum, no similar associations have been noted with the ALSFRS-R, the primary indicator of disease severity in ALS (40). However, limitations with using the ALSFRS-R and derived parameters have been previously described (44). Instead, disease progression models could be particularly informative, as they allow interpretation of biomarker profiles within the disease course.

It is also imperative to expand beyond studying the chitinases as just fold changes within a case–control paradigm, given the evidence that they act as active immune modulators rather than just passive indicators of pathology. For instance, TNF-α, LPS, and IFN-γ stimulation increased both CHIT1 expression and activity in human macrophages (45). Conversely, CHIT1, CHI3L1, and AMCase stimulation increased the transmigratory capacity of leukocytes from patients with multiple sclerosis (46).

In conclusion, studies should address how immune activation—vis-à-vis chitinase elevation—presents across the ALS disease course, whether it differs between glial cell types and what the functional consequences are. Studies also need to account for physiological aging, given multiple reports that it influences chitinase levels (27, 31, 47).

#### CONCLUSIONS AND FUTURE DIRECTIONS

What can be concluded of the chitinases holds true for all biomarkers; no single molecule can capture all the pathogenic processes at play in a disease as heterogeneous as ALS. This is particularly relevant in the case of inflammatory markers: these cannot be viewed in isolation because of their functional abundance and intricate signaling networks. It is the interaction with the disease microenvironment and the interplay between different cell types that drives pathology, rather than the singular action of a specific target. Multivariate biomarker panels are more likely to capture the dynamic immune

#### REFERENCES


signatures associated with different functional disease phases and identify optimal treatment windows and patients who would most benefit from immunomodulatory therapies. Therefore, the chitinases represent valuable additions to the current immunobiomarker repertoire; while their diagnostic and prognostic efficacy is unlikely to supersede that of the neurofilaments, they can assist with subtle distinctions between different neurodegenerative conditions and delineate the mechanisms underlying glial dysregulation. Additional mechanistic studies could focus on how the chitinases reflect the dynamicity of glial cell responses across the disease. For instance, current evidence already indicates that the chitinases reflect a neuroinflammatory component that is common to both genetic and sporadic forms of ALS (27). Future prospective studies could focus on recruiting MCs and following them as they transition to clinical disease to better understand how chitinase elevation manifests, what triggers it, and how it relates to other modalities.

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

NG was supported by a doctoral scholarship (Landesgraduiertenstipendien) from the Graduate Academy of Friedrich Schiller University, Jena, Germany. This work was also supported by the German Bundesministerium für Bildung und Forschung (BMBF) grant ONWebDUALS to JG under the aegis of the EU Joint Programme-Neurodegenerative Disease Research (JPND) and a BMBF grant PYRAMID to JG in the framework of the ERANET E-RARE program.

#### ACKNOWLEDGMENTS

We extend our sincere thanks to Klara Metzner for bibliography curation and management. **Figure 1** was created with www.BioRender.com.

responses in ALS mice. Brain Behav Immun. (2011) 25:1025–35. doi: 10.1016/j.bbi.2010.12.008


selectively trigger death of spinal motoneurons. Proc Natl Acad Sci USA. (2019) 116:2312–7. doi: 10.1073/pnas.1815961116


**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 Gaur, Perner, Witte and Grosskreutz. 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.