# T CELL DIFFERENTIATION AND FUNCTION IN TISSUE INFLAMMATION

EDITED BY : Amit Awasthi and Ritobrata Goswami PUBLISHED IN : Frontiers in Immunology

### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-614-3 DOI 10.3389/978-2-88963-614-3

### About Frontiers

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

### Frontiers Journal Series

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

### Dedication to Quality

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

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

### What are Frontiers Research Topics?

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

# T CELL DIFFERENTIATION AND FUNCTION IN TISSUE INFLAMMATION

Topic Editors:

Amit Awasthi, Translational Health Science and Technology Institute (THSTI), India Ritobrata Goswami, Indian Institute of Technology Kharagpur, India

Citation: Awasthi, A., Goswami, R., eds. (2020). T Cell Differentiation and Function in Tissue Inflammation. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-614-3

# Table of Contents


Sandip Ashok Sonar and Girdhari Lal

*39 Cyclic AMP Pathway Suppress Autoimmune Neuroinflammation by Inhibiting Functions of Encephalitogenic CD4 T Cells and Enhancing M2 Macrophage Polarization at the Site of Inflammation* Tatyana Veremeyko, Amanda W. Y. Yung, Marina Dukhinova,

Inna S. Kuznetsova, Igor Pomytkin, Alexey Lyundup, Tatyana Strekalova, Natasha S. Barteneva and Eugene D. Ponomarev


Eugene D. Ponomarev


Anke Fähnrich, Sebastian Klein, Arnauld Sergé, Christin Nyhoegen, Sabrina Kombrink, Steffen Möller, Karsten Keller, Jürgen Westermann and Kathrin Kalies


Suyasha Roy, Zaigham Abbas Rizvi and Amit Awasthi

# Editorial: T Cell Differentiation and Function in Tissue Inflammation

### Ritobrata Goswami <sup>1</sup> \* and Amit Awasthi <sup>2</sup> \*

*<sup>1</sup> School of Bioscience, Indian Institute of Technology Kharagpur, Kharagpur, India, <sup>2</sup> Translational Health Science and Technology Institute, Faridabad, India*

Keywords: CD4 T cell, CD8 T cell, cancer, autoimmune disease, inflammation

**Editorial on the Research Topic**

### **T Cell Differentiation and Function in Tissue Inflammation**

T cells constituting one of the arms of adaptive immune responses provide cell-mediated immunity against offending pathogens. Thymus is the maturation site for T cells that have been shown to be involved in cell-mediated immunity and humoral immune response in 1961–1962. It took another couple of decades to identify heterodimeric T cell receptor, which is crucial for the T cell activation, differentiation, and functions (1). In the next 25–30 years, several groundbreaking studies have contributed to the overall impact of T cells in modulating immune responses in health and diseases. T cell differentiation is one the key events that is absolutely essential for not only eliminating intra and extracellular pathogens but, upon dysregulation, could also lead to the onset of inflammation with exacerbate disease pathogenesis in autoimmune diseases. This Research Topic was developed to understand the complexity and molecular pathways that lead to the differentiation of Th cells that causes pathogenesis of disease. Under this Research Topic, a series of articles were published, which provided meaningful insights toward this emerging field. Briefly, this special issue is comprised of 8 original research papers, 5 full-length reviews, 3 mini-reviews, and 1 perspective to discuss the impact of T cell activation and differentiation in tissue inflammation. The original research articles included the role of CD4+ T cells in the pathophysiology of non-infectious uveitis and Graves' disease. The multi-faceted role of various subsets of CD4+ T cells have been reviewed extensively in tissue homeostasis, inflammatory bowel disease, osteoporosis, and neuroinflammation. These articles strongly support and provide new insight that harness the knowledge of Th cell differentiation may uncover novel therapeutic strategies to control inflammatory diseases.

Edited and reviewed by: *Loretta Tuosto, Sapienza University of Rome, Italy*

### \*Correspondence:

*Ritobrata Goswami ritobrata@gmail.com Amit Awasthi awasthi005@gmail.com*

### Specialty section:

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

Received: *31 January 2020* Accepted: *05 February 2020* Published: *21 February 2020*

### Citation:

*Goswami R and Awasthi A (2020) Editorial: T Cell Differentiation and Function in Tissue Inflammation. Front. Immunol. 11:289. doi: 10.3389/fimmu.2020.00289*

While CD4+ T cells work by releasing cytokines, CD8+ T cells are cytotoxic. A recent study has fueled the notion that CD8+ T cells might be important factor for longevity (2). Adoptive T cell treatment has shown immense potential to train the immune system in fighting against deadly diseases such as cancer. Tumor-specific CD8+ T cells are inserted into patients that target and attack cancer cells. There are clinical trials that have shown successful outcome in treating metastatic melanoma using adoptive T cell therapy. Patient T cells have been genetically modified with synthetic receptors generating chimeric antigen receptor T (CAR T) cells to specifically target surface antigen of cancer cells. Multiple targets are available for CAR T cell therapy including immunomodulatory antigens (PD-L1), overexpressed antigens (EGFR, HER2), aberrantly glycosylated proteins (MUC1). Suicide genes are being planned to be incorporated in CAR T cells to act as safety switch.

Differentiated CD4+ T cells play crucial role in providing beneficial immune responses against offending pathogens. Conversely, CD4+ T cells play various roles in the pathology of autoimmune inflammation. Effector CD4+ T cells, which were initially categorized as Th1 and Th2 cells by Mosmann et al. (3) have been expanded in the last 3 decades with the advent of Th17, Th9, Tfh,

**5**

and Th22 cells. Importantly, CD4+ T cells not only initiate specific immune responses; subsets of CD4+ T cells have also been identified that are able to inhibit the initiation of immune reactions and even downregulate established immune responses. These CD4+ T cells are termed regulatory T cells (Tregs) and have, because of their role in the immunopathogenesis of autoimmune diseases and their potential use in therapeutic applications, become the focus of intensive research. IL-10 secreting Tregs have been denoted as Tr1 cells that do not express Foxp3. Naïve CD4+ T cells can find their niche in inflamed tissues in some autoimmune disorders, which would otherwise be limited between circulation and secondary lymphoid organs. However, allergic inflammation from Th2 mediated responses to environmental allergens and Th1 mediated immunity is responsible for the generation of multiple organ-specific experimental autoimmune diseases in animals.

Differentiation and regulation of CD4+ T cells depend on a plethora of factors including strength of antigen-antibody interaction, amount of co-stimulation, cytokines present in the milieu, expression of transcription factors and their interaction with histone modifiers. During the development of thymocytes CD154 co-stimulation plays prominent role to the TCR repertoire diversity. CD154 deficiency attenuates the sharing of TCRβ clone compared to the wild-type in T-cell dependent immune responses, leading to incorrect editing of T-cell clonotypes during the negative thymic selection (Fähnrich et al.). As the appreciation of T helper subset plasticity increases, it becomes even more important to characterize them. Distinctly opposite T helper subsets can express the same receptor, secrete a common cytokine and be regulated by the same transcription factor. Study by Huang et al., urge caution in using LAG3/CD49b co-expression as standalone markers for Tr1 cell identification as they can also be expressed by Foxp3+ Tregs and CD8+ T cells. Further studies would dissect the physiological relevance of the expression of these markers in different T cell subsets. Several studies have indicated the participation of distinct T helper subsets in the pathophysiology of inflammatory disorders (4). In a model of neuroinflammation, Th17 cells have been demonstrated to receive help from Tfh cells for the inflammatory B-cell response (Quinn et al.). B cells regulated by Tfh cells could move to the CNS and undergo class switching that correlated with disease severity (Quinn et al.).

Significant amount of information has been generated by researchers on the outcome of CD4+ T cells in various inflammatory disorders in both mice and humans. Behavior of CD4+ T cells, are regulated by internal metabolic properties. Lipid metabolites can act as regulators of immune responses. Alteration of steroids pathways can affect inflammation and be responsible for the pathophysiology of various diseases. The enzyme cholesterol 25-hydroxylase, which synthesizes 25- OHC, can enhance IL-27-induced Tr1 cells (Vigne et al.). 25-OHC can negatively regulate Tr1 cells for the production of IL-10. T cell metabolism has been targeted for efficient cancer immunotherapy and altered effector T cell functions [(5), Roy et al.]. For the activation and proliferation of T cells, glucose provides the required energy (Roy et al.). Additional metabolites including lipids, ATP, nitric oxide, NAD also play crucial role in the differentiation of CD4+ T cells. Both mTOR and AMPK are sensors that regulate the metabolic checkpoints of T cell differentiation. mTOR inhibitors can attenuate glycolysis to induce memory T cell differentiation, while AMPK inhibitors decrease metabolism of fatty acids that in turn promote the differentiation of Th1 and Th17 cells. In contrast, activation of AMPK pathway can impart an analgesic effect in inflammatory pain by attenuating Il1β expression and blocking NF-κB activation (6). In naïve T cells, an increased AMP to ATP ratio is observed in the absence of TCR signaling leading diminished mTOR and sustained AMPK function (7). Interestingly during the lag phase of activated T cells, induced cytosolic calcium ions promote AMPK function in spite of decreased AMP level. However, in the growth phase of activated T cells increased ATP levels leads to sustained mTOR function. Depletion of the amino acids Trp and Arg can attenuate both the activation and function of effector T cells (5). Decreased level of oxygen and oxidative phosphorylation can increase PD-L1 expression on cancer cells. Activation of HIF-1 can induce effector properties of T cells by augmenting glycolysis and glutaminolysis. Other transcription factors that can act as metabolic checkpoints during T cell differentiation include BCL-6 (Tfh cell differentiation), IRF4 (Th2, Th17, Th9 cell differentiation), Foxo (Th9 cell differentiation), MYC (balance between Th17 and Treg cell differentiation). Drugs have been developed that target these various metabolic checkpoints to ameliorate various inflammatory diseases including Crohn's disease, ulcerative colitis, type 2 diabetes, rheumatoid arthritis, and chronic obstructive pulmonary disease. The activation of T cells is also mediated by essential trace metals including zinc. TCR signaling steps could potentially be altered by zinc (8). The zinc transporter Zip6, expressed on the surface of unstimulated T cells, is touted to be important bringing down the threshold of T cell activation. Vitamins also regulate the activation of T cells. Even though vitamin D blocks CD4+ T cell proliferation, it increases the number of CD4+CD25+Foxp3+ Treg cells. Therefore, T cell targeting via metabolic regulators represent exciting avenues for further investigations to regulate pathophysiology of inflammatory disorders and cancer.

Collectively, the articles published within the Research Topic highlight the emerging roles and underlying mechanisms of T cell differentiation and functions in tissue inflammation and their impact in the pathogenesis of inflammatory diseases. Based on the published work under this topic, it is further required to understand the functional dynamics of Th cell plasticity that lead to the ultimate outcome of immune-pathogenesis of diseases and lead to advancing our understanding for the immunological basis of diseases. The acquired immunological-based knowledge from the published articles will contribute further refined and novel immune strategies for inflammatory conditions.

### AUTHOR CONTRIBUTIONS

RG and AA conceived, designed, and wrote the manuscript. All the authors read and approved the final manuscript for publication.

### FUNDING

This work was supported by funds from the Department of Biotechnology and Department of Science and Technology, Government of India to AA. RG has been supported by a grant from INNO-INDIGO, Department of Science & Technology, Government of India.

### REFERENCES


### ACKNOWLEDGMENTS

We acknowledge and appreciate all contributing authors and their participation in this Research Topic. We express our gratitude to all reviewers for agreeing to participate in the peer review process and providing their comments and feedback on the manuscript.


**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 Goswami and Awasthi. 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.

*Solenne Vigne1†, Fanny Chalmin2†, Donovan Duc1 , Aurélie S. Clottu1 , Lionel Apetoh3 , Jean-Marc A. Lobaccaro4 , Isabelle Christen5 , Juan Zhang5 and Caroline Pot1,2\**

*1 Laboratories of Neuroimmunology, Division of Neurology and Neuroscience Research Center, Department of Clinical Neurosciences, Lausanne University Hospital, Lausanne, Switzerland, 2Department of Pathology and Immunology, University of Geneva, Geneva, Switzerland, 3 Faculté de Médecine, University of Bourgogne, INSERM U866, Centre Georges François Leclerc, Dijon, France, 4GReD, Université Clermont Auvergne, CNRS, INSERM, CRNH Auvergne, Clermont-Ferrand, France, 5Analytical Sciences and Imaging, Novartis Institutes for BioMedical Research, Basel, Switzerland*

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Hyun Park, National Cancer Institute, United States Ashutosh Chaudhry, Memorial Sloan Kettering Cancer Center, United States*

### *\*Correspondence:*

*Caroline Pot caroline.pot-kreis@chuv.ch*

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

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 07 July 2017 Accepted: 07 September 2017 Published: 25 September 2017*

### *Citation:*

*Vigne S, Chalmin F, Duc D, Clottu AS, Apetoh L, Lobaccaro J-MA, Christen I, Zhang J and Pot C (2017) IL-27-Induced Type 1 Regulatory T-Cells Produce Oxysterols that Constrain IL-10 Production. Front. Immunol. 8:1184. doi: 10.3389/fimmu.2017.01184*

The behaviors of lymphocytes, including CD4+ T helper cells, are controlled on many levels by internal metabolic properties. Lipid metabolites have recently been ascribed a novel function as immune response modulators and perturbation of steroids pathways modulates inflammation and potentially promotes a variety of diseases. However, the impact of lipid metabolism on autoimmune disease development and lymphocyte biology is still largely unraveled. In this line, oxysterols, oxidized forms of cholesterol, have pleiotropic roles on the immune response aside from their involvements in lipid metabolism. The oxysterols 25-hydroxycholesterol (25-OHC) and 7α,25-dihydroxycholesterol (7α,25-OHC) regulate antiviral immunity and immune cell chemotaxis. However, their physiological effects on adaptive immune response in particular on various subset CD4+ T lymphocytes are largely unknown. Here, we assessed oxysterol levels in subset of CD4+ T cells and demonstrated that 25-OHC and transcript levels of its synthesizing enzyme, cholesterol 25-hydroxylase, were specifically increased in IL-27-induced type 1 regulatory T (TR1) cells. We further showed that 25-OHC acts as a negative regulator of TR1 cells in particular of IL-10 secretion *via* liver X receptor signaling. Not only do these findings unravel molecular mechanisms accounting for IL-27 signaling but also they highlight oxysterols as pro-inflammatory mediators that dampens regulatory T cell responses and thus unleash a pro-inflammatory response.

Keywords: immunometabolism, CD4+ T cells, type 1 regulatory T cells, autoimmunity, oxysterols, cholesterol 25-hydroxylase, Epstein–Barr virus-induced G-protein coupled receptor 2 (EBI2), nuclear hormone liver X receptor

# INTRODUCTION

Oxysterols, oxidized forms of cholesterol, are essential precursors for bile acid and steroid biosynthesis. Apart from their basic metabolic properties, they have recently been ascribed with immunomodulatory functions. The enzyme cholesterol 25-hydroxylase (Ch25h) is the ratelimiting step to synthetize both 25-hydroxycholesterol (25-OHC) and 7α,25-dihydroxycholesterol (7α,25-OHC) from cholesterol. Both oxysterols modulate the immune response, 25-OHC controls viral infection in macrophages (1) and 7α,25-OH promotes macrophage and B cell trafficking within lymphoid structures (2). We showed that 7α,25-OHC promotes memory CD4<sup>+</sup> T cell migration to the target inflammatory organs during autoimmunity (3, 4). While authors have proposed oxysterols as pro-inflammatory mediators, others have submitted 25-OHC as an

**8**

anti-inflammatory intervener (5). Those contradictory results open the debate on the biological activities of oxysterols during the immune response. Furthermore, while the roles of oxysterols during innate immune response have been well studied in macrophages, their tasks during adaptive immune response remain largely unknown.

Adaptive immune homeostasis relies in part on orchestrated interactions among subsets of T cells with effector or regulatory functions. CD4<sup>+</sup> regulatory T cell subsets include naturally occurring CD4<sup>+</sup>CD25<sup>+</sup> Treg cells (nTregs), which can be defined by their expression of the forkhead-box transcription factor Foxp3, as well as peripherally induced type 1 regulatory T (TR1) cells that produce IL-10. The cytokine IL-27, mainly produced by antigen-presenting cells, promotes TR1 cell development. While initial animal studies suggested that IL-27 supported pro-inflammatory responses, the anti-inflammatory properties of IL-27 were exemplified in mouse models, where IL-27 injections reduced disease severity of experimental autoimmune encephalomyelitis (6–8). In addition, IL-27R-deficient mice show enhanced pro-inflammatory CD4<sup>+</sup> T cell response and enhanced autoimmunity susceptibility (9, 10) and die following exposure to parasitic and bacterial infections due to severe immunopathology (11). IL-27 downmodulates the immune responses through production of the immunosuppressive cytokines IL-10 (12) and IFN-γ (7) and by inhibiting pro-inflammatory cytokine, including IL-17, production (6). Interestingly, the oxysterol 7β,27-dihydroxycholesterol has been identified as an agonist for RORγt, a crucial transcription factor for IL-17-producing CD4<sup>+</sup> T cells (TH17 cells) and, thus, as a pro-inflammatory mediator (13). Those results suggest that oxysterols could act as fine tuners of the immune response.

Here, we show that the oxysterol 25-OHC is specifically induced by IL-27 *via* the signal transducer and activator of transcription factor 1 (Stat1) and interferon regulatory factor 1 (IRF1) signaling during CD4<sup>+</sup> T cell differentiation. 25-OHC further acts as a negative regulator on IL-10 production by lowering B-lymphocyte-induced maturation protein 1 (Blimp1) expression that contributes to IL-10 secretion by CD4<sup>+</sup> T cells (14). 25-OHC dampens anti-inflammatory cytokine production *via* the nuclear hormone liver X receptors (LXR) signaling and further promotes intracellular cholesterol accumulation, a process recognized to drive inflammation (15). Those results strengthen the pro-inflammatory role of 25-OHC during adaptive immune response by limiting the generation of IL-27-induced regulatory TR1 cells both *in vitro* and *in vivo*.

### MATERIALS AND METHODS

### Animals

*Ch25h<sup>−</sup>/*<sup>−</sup> mice were purchased from Jackson Laboratory. Stat1*<sup>−</sup>/<sup>−</sup>* mice on C57BL/6 background were a kind gift from Professors M. Mueller and D. Merkler (16), *Irf1<sup>−</sup>/<sup>−</sup>* by L. Apetoh (17), and *Lxr*αβ*−/−* by D. J. Mangelsdorf (18). Mice, on C57BL/6 background, were housed under specific pathogen-free conditions at Lausanne University Hospital. All experiments were undertaken in accordance with guidelines from the Cantonal Veterinary Services of states Vaud and Geneva.

### *In Vitro* T Cell Differentiation

Spleen and inguinal lymph nodes were obtained from 6- to 10-week-old mice and then mashed on a 70-µm mesh together with culture media to obtained single cell suspension. After erythrocyte lysis, naive CD4<sup>+</sup> T cells were purified by negative selection using immunomagnetic beads (Naive CD4<sup>+</sup>T cell Isolation Kit, Miltenyi Biotec) and stimulated for 3 days or for the indicated time on plate-bound antibodies against CD3 (145- 2C11, 1 µg/ml) and CD28 (PV-1, 1 µg/ml) without cytokines (TH0) or with mouse IL-27 [50 ng/ml (TR1); IL-12 (10 ng/ml) and anti-mouse IL-4 (11B11; 20 µg/ml) (TH1); IL-4 (20 ng/ml) and anti-mouse IFNy (XMG1.2; 20 µg/ml) (TH2); human TGFβ1 (2 ng/ml) (iTregs); TGF-β1 and IL-6 (20 ng/ml) (TH17)]. Cytokines were purchased from eBioscience, 25-OHC from Avanti Polar Lipis Inc., anti-CD3/CD28 monoclonal antibodies (mAbs) from BioXcell and LXR agonists T0901317 and GW3965 from Sigma.

### Oxysterol Extraction and Analysis using Ultra-High Performance Liquid Chromatography–Tandem Mass-Spectrometry (UHPLC–MS/MS)

Cell pellets (106 cells) re-suspended in water containing a mixture of deuterated internal standard compounds and 200 µM Butylated hydroxytoluene were lysed using a Precellys®24 (Bertin Technologies) before extraction. EtOH (ninefold in volume) was added in six steps to allow a slow protein precipitation under 4°C. One milliliter cell culture medium was used to extract oxysterols using the same slow protein precipitation method. Extracts were dried down and concentrated 10 times prior to injection into UHPLC-MS/MS. The oxysterols analysis was carried out on a Nexera UHPLC system (Shimadzu, Kyoto, Japan) coupled to a QTrap®6500 (ABSciex, Framingham, MA, USA) mass spectrometer. The UHPLC-MS/MS method was as previously described (19).

### Cytokine Measurements

Supernatants were collected after 48 h of culture and secreted cytokines measured by ELISA (eBioscience).

### Flow Cytometry

Cells, preincubated with mAb 2.4G2 (anti-CD16/32) to block Fc receptors were labeled with CD4 AlexaFluor 700 (GK 1.5, eBioscience). Cytokine-detection was performed by intracellular cytokine staining with anti-IFN-γ Alexa Fluor 488 (XMG1.2, eBiosciences) and anti-IL-10 BV421 (JE55-16E3, Biolegend). Cells (1.5 × 105 cells/well) were stimulated at 37°C with 10 ng/ml phorbol myristate acetate (PMA, Sigma), 1 µg/ml ionomycin (Sigma) for 4 h, and 5 µg/ml Brefeldin (Sigma) for 2 h and permeabilized using Foxp3/Transcription Factor Staining Buffer Set (eBiosciences). Data were acquired using a LSR II cytometer (BD Biosciences).

### CFSE Labeling

Naive CD4<sup>+</sup> T cells, suspended in 5-µM CFSE staining buffer (Molecular Probes) and incubated at 37°C for 10 min, were cultured as described above. After 5 days of culture, cell division was determined by measuring CFSE fluorescence in total cells. Cell viability was determined by measuring Fixable viability Stain 620 (FVS620, BD Bioscience). Data were analyzed using the FlowJo V10 software.

### Protein Isolation and Analysis

Total cell lysates were prepared in RIPA buffer (50 mM Tris–HCl pH 7.4, 2 mM EDTA pH8, 150 mM NaCl, 0.5% Na-deoxycholate, 0.1%SDS, 1% Non-idet P40) supplemented with protease inhibitors (Mini protease inhibitor, Roche). Samples were separated on a 8% SDS-polyacrylamide gel and transferred to nitrocellulose. Rat anti-mouse Blimp1 (5E7, Santa Cruz) at 1/200 dilution, rabbit anti-mouse β-Actin (N-21, Santa Cruz) at 1/600, Goat anti-rat or rabbit IgG-HRP 1/10000 were used and visualized by chemiluminescence (ECL, Amersham Pharmacia Biotech); relative density of Blimp1 expression analyzed using the ImageJ software.

# Quantitative Real-time PCR (RT-PCR)

RNA was extracted with Tryzol (Invitrogen Life Technologies), cDNA synthesized with random hexamers and Superscript II reverse transcriptase (Invitrogen Life Technologies) used as template for RT-PCR (Applied Biosystems® StepOne plus) with SYBR green Supermix (KAPA SYBR® FAST Universal, Labgene). Gene expressions were assessed with specific primers as follows: Ch25h (Fw CCAGCTCCTAAGTCACGTC Rev CACGTCGAAGAAGGTCAG), Cyp7B1 (Fw TTCCTCCACTCA TACACAATG Rev CGTGCTTTTCTTCTTACCATG), HSD3B7 (Fw AAGAGGCCAGCAATACCCAG Rev ACCATCCACAAAG TCAACG), Blimp1 (Fw GGAGGATCTGACCCGAAT Rev TC CTCAAGACGGTCTGCA), AhR (Fw CTCCTTCTTGCAAATC CTGC Rev GGCCAAGAGCTTCTTTGATG), c-maf (Fw GG CCATGGAATATGTTAATGACTTC Rev CCGCACTGGCTGA TGATG), IRF1 (Fw AGGCATCCTTGTTGATGTCC Rev AATT CCAACCAAATCCCAGG), LXR-β (Fw TTTGCTTTTCGCTCA GCAAGC Rev GGAGGCGAGAGTTGCCTCTG), SREBF1 (Fw GGGGAACTTTTCCTTAACGTGG Rev CGGGAAGTCACTG TCTTGGT), ABCA1 (Fw AGCACCGTGTCTTGTCTGAA Rev CATCGATGGTCAGCGTGTCA) and β-actin (Fw CCTGTATG CCTCTGGTCGTA Rev CCATCTCCTGCTCGAAGTCT). Values obtained with the SDS 2.2 software (Applied Biosystems) and gene expression calculated using the comparative method (2−ΔCt) for relative quantification by normalization to β*-actin* gene expression.

### *In Vivo* Treatment with Anti-CD3

*Ch25h<sup>−</sup>/<sup>−</sup>* and wild-type mice were treated with 20 µg of anti-CD3 (clone 2C11) or PBS i.p. every 3 days for a total of four times. Mice were sacrificed 4 h after the last treatment, single cell suspensions were prepared from mesenteric lymph nodes (MLNs).

### Statistical Analysis

Statistical analysis was performed using Prism software (Graph Pad software, La Jolla, CA, USA). Evaluations were performed with the unpaired Student's *t* test or with two-way ANOVA as appropriate. Two-tailed *p*-values < 0.05 were considered significant.

# RESULTS

# IL-27 Induces Ch25h Expression and Production of 25-OHC

While oxysterols are implicated in immune responses, their levels in T lymphocytes have not been assessed. We investigated the expression of oxysterol-converting enzymes on different subsets of T helper CD4<sup>+</sup> T cells differentiated *in vitro* into TH0, TH1, TH2, TH17, Foxp3iTregs, or TR1 cells. By quantitative RT-PCR analysis, we observed that Ch25h was highly expressed in IL-27-induced TR1 cells, compared to cells activated in the absence of differentiating cytokines (TH0) (**Figure 1A**, left panel). Cholesterol is converted by Ch25h into 25-OHC. We, therefore, applied UHPLC-MS/MS to analyze the extra- and intracellular oxysterol levels. Consistent with Ch25h-increased expression, 25-OHC production was specifically induced by IL-27 and detected at high levels in TR1 cells. Low level of Ch25h expression was observed in TH17 cells, but 25-OHC production was not increased in TH17 compared to TH0 subset (**Figure 1A**, right panels). 25-OHC can be further metabolized into 7α,25-OHC. However, 7α,25-OHC could not be detected in any subset of T cells. We further examined other oxysterol-converting enzyme expressions and observed marginal expressions of the oxysterol-converting enzymes Cyp27a1 (**Figure 1B**) and Cyp46a1 (**Figure 1C**) in all T cell subsets without any specific induction by differentiating cytokines. The oxysterols 27-OHC downstream Cyp27a1 (**Figure 1B**) and 24-OHC downstream Cyp46a1 (**Figure 1C**) were detected in all cell types at very low levels and their productions not affected by any cytokine combinations.

25-OHC can also be formed from cholesterol through autoxidation (20) or by alternate pathways (21). We, therefore, differentiated TR1 cells from both wild-type and Ch25h*<sup>−</sup>*/*<sup>−</sup>* CD4<sup>+</sup> T cells to assess Ch25h-independent production of 25-OHC. IL-27 induced a significantly increased expression of Ch25h in wild-type CD4<sup>+</sup> T cells starting after 16 h until 48 h of culture compared to Ch25h*<sup>−</sup>*/*<sup>−</sup>* cells, where no Ch25h expression was detected (**Figure 1D**). IL-27 did not induce 25-OHC production in the absence of Ch25h and no compensatory increase of 24-OHC or 27-OHC production was observed in Ch25h*<sup>−</sup>*/*<sup>−</sup>* TR1 cells (**Figure 1E**).

# IL-27 Induces Ch25h in a Stat1- and Irf-1-Dependent Manner

IL-27 signals through Stat1 and Stat3 (22, 23). Because Stat1 induces Ch25h expression in macrophages (1), we tested the ability of IL-27 to activate Ch25h expression in Stat1*<sup>−</sup>*/*<sup>−</sup>* cells. Genetic elimination of Stat1 resulted in the marked loss ability of IL-27 to induce Ch25h (**Figure 2A**, left panel). Moreover, the ability of IL-27 to induce 25-OHC in CD4<sup>+</sup> T cells was abrogated in the absence of Stat1 (**Figure 2A**, right panel).

Interferon regulatory factor 1 is a main transcription factor downstream Stat1 (17, 24) that has been proposed to drive Ch25h expression during viral infection (25). IRF1 is induced by IL-27 in a Stat1-dependent manner, with an early peak expression after 2 h of culture (**Figure 2B**, left panel). We further asked whether Ch25h expression was dependent on IRF1 and tested the ability

Figure 1 | IL-27 specifically induces cholesterol 25-hydroxylase (Ch25h) expression and 25-OHC production in CD4+ T lymphocytes. Naive CD4+CD62LhiCD25*<sup>−</sup>* T cells obtained from wild-type mice were differentiated into TH0, TH1, TH2, TH17, iTregs, and TR1 cells in the presence of anti-CD3 and anti-CD28 antibodies. RNA isolated from the CD4+ T cells after 24 h of culture was subjected to real-time PCR (RT-PCR) relative to the expression of mRNA encoding β-actin (2*<sup>−</sup>*ΔCT × 100,000) to examine oxysterol-converting enzyme expression, while oxysterol levels were assessed by LC–MS/MS in supernatants (extracellular) and in cell pellets (intracellular) after 3 days of culture. (A) Ch25h expression and 25-OHC production (B) Cyp27a1 expression and 27-OHC production. (C) Cyp46a1 expression and 24-OHC production. (D) RNA isolated at different time points of culture following activation with IL-27 from naive CD4+ T cells obtained from wild-type mice (closed squares) or Ch25h*−*/*−* mice (open squares), was subjected to RT-PCR to examine Ch25h expression. (E) Extracellular oxysterols levels measured by LC–MS/MS after 3 days of culture from wild-type mice or Ch25h*−*/*−* naïve CD4+ T cells differentiated in the presence (TR1) or absence (TH0) of IL-27. Data are shown from two or three independent experiments (\**p* < 0.05).

of IL27 to activate Ch25h mRNA levels in IRF1*<sup>−</sup>*/*<sup>−</sup>* CD4<sup>+</sup> T cells. Similarly to Stat1*<sup>−</sup>*/*<sup>−</sup>* T cells, IL-27 was not able to induce Ch25h in the absence of IRF1 (**Figure 2B**, right panel). T-bet, another transcription factor downstream of Stat1, is induced by IL-27 with high expression 20 h after culture initiation (**Figure 2C**, left panel). However, T-bet expression was independent of IRF1 as

Figure 2 | The transcription factors Stat1 and interferon regulatory factor 1 (IRF1) are mandatory for cholesterol 25-hydroxylase (Ch25h) induction by IL-27. Naive CD4+ T cells obtained from wild-type, Ch25h*−*/*−*, Stat1*−*/*−*, and IRF1*−*/*−* mice were differentiated without cytokines (TH0) or with IL-27 (TR1) as indicated. (A) Ch25h expression was assessed by real-time PCR (RT-PCR) relative to β-actin expression after 24 h in culture (left panel) and extracellular 25-OHC levels measured by LC-MS/MS (right panel). (B) IRF1 and Ch25h expression levels were assessed by RT-PCR at the indicated time points following activation (left panel) or after 24 h of culture (right panel). (C) T-bet and Ch25h expression levels were assessed by RT-PCR at indicated time points (left panel) or after 24 h of culture \**p* < 0.05. (D) IL-10 (left panel) and IFN-γ (right panel) secretions were measured by ELISA in the supernatants of T cells cultured for 48 h. Data are shown from one out of three independent experiments (\**p* < 0.05).

IRF1*<sup>−</sup>*/*<sup>−</sup>* T cells expressed T-bet at similar levels as wild-type T cells (**Figure 2C**, middle panel), suggesting that Ch25h expression is not downstream of T-bet. Indeed IL-27 could induce Ch25h expression in the absence of T-bet (**Figure 2C**, right panel). We, therefore, propose that the transcription factor IRF1, but not T-bet, is mandatory for IL-27-induced Ch25h expression.

TR1 cells are characterized by their secretion of IL-10 and IFN-γ (22). We observed that IRF1 and Stat1 are important to maintain both IL-10 and IFN-γ expression, as in the absence of each individual transcription factors, cytokine expressions were significantly reduced (**Figure 2D**).

### Ch25h-Deficient TR1 Cells Depict Higher IL-10 Production both *In Vitro* and *In Vivo*

The strong expression of Ch25h induced by IL-27 prompted us to investigate the role of 25-OHC during TR1 cell differentiation. Naive CD4<sup>+</sup> T cells from wild-type or Ch25h*<sup>−</sup>*/*<sup>−</sup>* mice were differentiated *in vitro* with IL-27 (TR1), without any cytokines (TH0) or with IL-12 and anti-IL-4 to generate TH1 as a control as they also express IFN-γ. The secretion of IL-10 (**Figure 3A**) and IFN-γ (**Figure 3B**) were notably enhanced in TR1 cells derived from Ch25h*<sup>−</sup>*/*<sup>−</sup>* cells (white bars) compared to wild-type cells (black bars). IFN-γ was not enhanced in TH0 nor in TH1 cells. Furthermore, the frequency of cell expressing IL-10 and IFN-γ was increased in Ch25h*<sup>−</sup>*/*<sup>−</sup>* compared to wild-type TR1 cells while they were not enhanced neither in TH0 nor in TH1 cells (**Figure 3C**). Oxysterols interfere with different cell type proliferation, including cancer cells (26). However, CFSE staining showed similar proliferation rates between Ch25h*−*/*−* and wildtype Tr1 cells (**Figure 3D**).

To further address the *in vivo* relevance of Ch25h in inducing IL-27-driven TR1 cells and the potential effect on regulating autoimmunity and tissue inflammation, we conducted repeated *in vivo* treatments with anti-CD3 to induce IL-10<sup>+</sup> regulatory T cells (27), that have been shown to be IL-27 dependent (28). We, thus, repeatedly administered anti-CD3 or PBS to C57Bl6 wild-type mice and assessed Ch25h expression in MLNs 4 h after the last injection. In line with our *in vitro* findings, Ch25h was significantly induced in wild-type but not Ch25h*<sup>−</sup>*/*<sup>−</sup>* CD4<sup>+</sup> T cells (**Figure 3E**). Since IL-10 is produced by TH17 cells (29), Foxp3<sup>+</sup> Tregs (30), and TR1 cells, we further analyzed the production of IL-10 by Foxp3*<sup>−</sup>* IL-17*<sup>−</sup>* CD4<sup>+</sup> CD3<sup>+</sup> TCRαβ+ T cells as previously published (28). Administration of anti-CD3 to wild-type mice resulted in a significant induction of IL-10<sup>+</sup> T cells in the MLNs that were significantly increased in Ch25h*<sup>−</sup>*/*<sup>−</sup>* mice (**Figure 3F**). Ch25h, thus, inhibits IL-10<sup>+</sup> T cell generation both *in vitro* and *in vivo*.

### 25-OHC Impairs IL-10 Expression from IL-27-Induced TR1 Cells

We further asked whether exogenous 25-OHC influences IL-10 production. Addition of 25-OHC during TR1 cell differentiation decreased IL-10 secretion in a dose-dependent manner (**Figure 4A**). 25-OHC did not inhibit TR1 cell proliferation assessed with CFSE, nor impacted Tr1 cell viability at concentration of 30 nM or lower (**Figure 4B**). At higher doses, in addition to the effects on IL-10 secretion, proliferation was inhibited and cell viability decreased (**Figure 4B**). We thus pursued our experiments with concentrations of 25-OHC that solely impacted cytokine production. We further questioned whether the unique addition of 25-OHC would compensate for the IL-10 phenotype noted in Ch25h*<sup>−</sup>*/*<sup>−</sup>* TR1 cell. The sole addition of 25-OHC (at 15 and 30 nM) dampened both IL-10 secretion (**Figure 4C**) and IL-10 frequency (**Figure 4D**) in Ch25h*<sup>−</sup>*/*<sup>−</sup>* T cells reversing to similar IL-10 level of wild-type TR1 cell.

We then assessed if oxysterols impact the expression levels of transcription factors involved in IL-10 production. Ahr, c-maf, and Blimp1 control IL-10 expression during TR1 cell differentiation (14, 28, 31). While Ahr and c-maf expressions were not increased in Ch25h*<sup>−</sup>*/*<sup>−</sup>* TR1 cells (Figure S1A in Supplementary Material), Blimp1 expression level was increased in Ch25h*<sup>−</sup>*/*<sup>−</sup>* TR1 cells and downregulated by 25-OHC both at the mRNA (**Figure 4E**) and protein levels (**Figure 4F**). We further observed that Blimp1 expression was dependent on IRF1 and Stat1 signaling (Figure S1B in Supplementary Material), both of which showed to be critical for Ch25h expression. Those results suggest that 25-OHC negatively regulates IL-10 by dampening Blimp1 expression.

Altogether, these findings suggest that Ch25h-signaling pathway negatively regulates IL-10 expression in IL-27-induced TR1 cells.

### Oxysterols Inhibit IL-10 Secretion in an LXR-Dependent Manner in TR1 Cells

25-OHC are ligands for the extracellular receptor G-coupled protein receptor Epstein-Barr virus-induced G-protein coupled receptor 2 (EBI2) (32, 33) that is expressed on activated murine and human CD4<sup>+</sup> T cells (3, 4) and for intracellular receptors. We first assessed whether EBI2 receptor was involved in IL-10 inhibition. Neither IL-10 nor IFN-γ inhibitions by 25-OHC were mediated by EBI2 (data not shown). In addition to EBI2 binding, oxysterols activate transcription factors intracellularly. In this line, 7β,27-dihydroxycholesterol is a potent and selective activator for the transcription factor RORγt, a main transcription factor of TH17 cells (13). Furthermore, LXRs are established targets of 22-OHC and 25-OHC (34, 35). We, therefore, tested whether LXR activation would reproduce 25-OHC effects. We first observed that the LXR agonist T0901317 decreased IL-10 expression induced by IL-27 in a dose-dependent manner in wild-type and Ch25h*<sup>−</sup>*/*<sup>−</sup>* TR1 cells (**Figure 5A**). This LXR agonist was more potent in inhibiting IL-10 secretion in wild-type compared to Ch25h*<sup>−</sup>*/*<sup>−</sup>* TR1 cells (**Figure 5A**), suggesting a putative additive effect of LXR agonist in the presence of 25-OHC. In this line, sole addition of T0901317 decreased the secretion of IL-10 in the same range than 25-OHC alone; however, combined treatment with 25-OHC and T0901317 depicted additive effects on dampening IL-10 secretion in wild-type but not in Ch25h*<sup>−</sup>*/*<sup>−</sup>* TR1 cells (**Figure 5B**). No effects on proliferation were observed at the concentration used in this assay (**Figure 5C**).

To further investigate the specific role of LXR signaling in TR1 cells, we examined the expression pattern of LXR-α and LXR-β in naive CD4<sup>+</sup> T cells (TH0) and in IL-27-differentiated TR1 cells in the presence or absence of 25-OHC or LXR agonist

Figure 3 | Endogenous 25-OHC negatively regulates IL-10 production in TR1 cells *in vitro* and *in vivo*. Naive CD4+ T cells obtained from wild-type or Ch25h*−*/*−* mice as indicated were differentiated into TH0, TH1, and TR1 cells for 48 h. (A,B) IL-10 and IFN-γ cytokine productions in culture supernatants were assessed by ELISA. Data are shown from one of three independent experiments with similar results. Error bars represent the SD of triplicates in the same experiment (\**p* < 0.05). (C) Flow cytometric analysis of intracellular staining of IL-10 and IFN-γ of TH0, TR1 cells, and TH1 cells. (D) Proliferative responses were assessed by CFSE incorporation in TR1 cells, representative of one of three independent experiments with similar results. (E) Wild-type or Ch25h*−*/*−* mice were injected i.p. with 20 µg of antibodies to CD3 or PBS once every 3 days, for a total of three times. 4 h after the last injection, mice were sacrificed. CD4+ T cells from mesenteric lymph nodes (MLN) were then FACS sorted, RNA was then isolated and subjected to real-time PCR to examine cholesterol 25-hydroxylase (Ch25h) expression. (F) Frequency of IL-10+ CD4+ T cells from anti-CD3 or PBS treated mice was analyzed by flow cytometry in MLN (mean + SD of three mice). Data are shown from one out of two experiments (\**p* < 0.05).

GW 3965 used as positive control. LXR-α was not detected in any of the above conditions in accordance with previous reports (36). While LXR-β was upregulated in the presence of IL-27 (TR1 cells) compared to TH0, the addition of 25-OHC on TR1 cells significantly enhanced LXR-β mRNA expression to an extent comparable with that induced by GW3965 (**Figure 5D**).

Figure 4 | Exogenous 25-OHC inhibits TR1 cell differentiation. Wild-type or Ch25h*−*/*−* naive cells were differentiated with IL-27 in the presence of 25-OHC at the indicated concentration (or with 30 nM if not indicated) (A) IL-10 cytokine production in culture supernatants assessed by ELISA analysis. (B) Histogram profiles of CFSE (above) and fixable viability-labeled cells, with highly stained cells corresponding to dead cells (below). (C) IL-10 cytokine production assessed by ELISA analysis and (D) IL-10-expressing CD4+ T cells detected by intracellular staining and quantified by flow cytometry. Data shown are representative of one of three independent experiments with similar results (\**p* < 0.05). B-lymphocyte-induced maturation protein 1 (Blimp1) expression was evaluated by (E) quantitative real-time PCR relative to β-actin (*p* < 0.05). (F) Western blot on whole cell lysates. Relative density values of Blimp1 were calculated using ImageJ software. Data are shown from one of two independent experiments with similar results (\**p* < 0.05).

In contrast, the addition of 25-OHC or GW3965 in the absence of IL-27 had no significant effect on LXR-β mRNA expression (Figure S2A in Supplementary Material). To determine whether 25-OHC influences LXR transcriptional program in TR1 cells, we tested if 25-OHC could impact LXR-target gene expression particularly genes involved in *de novo* cholesterol biosynthesis as sterol regulatory element binding protein (SREBP1) and in cholesterol efflux as ATP-binding cassette transporter A1 (ABCA1). SREBP1 mRNA expression levels were significantly upregulated compared to TH0 when 25-OHC or GW3965 were added together with IL-27 (**Figure 5E**). Treatment of TR1 cells with 25-OHC resulted in a significant downregulation of ABCA1 mRNA expression whereas addition of GW3965 led to a robust induction of this gene (**Figure 5F**). In contrast, addition of 25-OHC without IL-27 had no significant effect on SREBP1 or ABCA1 mRNA expression (Figures S2B,C in Supplementary Material). These results suggest that LXR is more active in TR1 compared to TH0 cells.

We then investigated whether the inhibitory effect of 25-OHC on TR1 differentiation and IL-10 production was dependent on LXR signaling. We took advantage of cells deficient for LXRαβ. We observed that TR1 cells differentiated in the absence of LXRαβ displayed a significantly higher secretion of IL-10 (**Figure 5G**). 25-OHC was significantly more potent in inhibiting IL-10 production in the presence of LXR receptor similarly to the LXR agonist GW3965 (**Figure 5H**). To assess the implication of T cell proliferation, WT and LXRαβ*<sup>−</sup>*/*<sup>−</sup>* naive CD4<sup>+</sup> T cells were labeled with CFSE and activated with IL-27 in the presence or not of 25-OHC. The percentage of divided cells in response to TCR stimulus alone (anti-CD3, anti-CD28 activation) was significantly greater in LXRαβ*<sup>−</sup>*/*<sup>−</sup>* compared to WT TH0 cells (**Figure 5I**) in agreement with previous studies (36). However, WT and LXRαβ*<sup>−</sup>*/*<sup>−</sup>* TR1 cells underwent the same percentage of proliferation independently of 25-OHC addition. Altogether, these results indicate that 25-OHC regulates IL-10 production and that LXR signaling mediates, at least partially, the inhibitory effect of 25-OHC on TR1 cell polarization.

# DISCUSSION

Oxysterols have been ascribed functions in modulating the immune response. However, their pro-inflammatory and/or antiinflammatory contributions remain debated and scarcely studied during adaptive immune response. Here, we propose that 25-OHC dampens the secretion of the major anti-inflammatory cytokine IL-10 induced by IL-27 and thus assigns this oxysterol with a proinflammatory role during adaptive immune responses (**Figure 6**). Our findings are in line with publications assigning 25-OHC with both a pro-inflammatory function and an amplificatory inflammatory signal (1, 37–39).

We observed that Ch25h mRNA expression and 25-OHC levels are strongly induced by IL-27. Both Stat1 and Stat3 are phosphorylated upon IL-27 signaling, leading to transactivation of IL-10 (23, 40). While Stat3 is important for the expression of the transcription factors c-maf and Ahr, Stat1 induces the expression of the transcription factors T-bet, IRF1, and Blimp1 (17, 24). Interestingly, Stat1 (1) and IRF1 (25) can drive Ch25h induction in macrophages during viral infection. IRF1, initially identified as a TH1 cell-specific transcription factor, was further implicated in the biology of other T cell subsets such as TH9 cells (17, 41). In our study, we observed that Ch25h expression is dependent on the transcription factors Stat1 and IRF1 (**Figure 6**) but not T-bet. Those results suggest similarities in signaling pathways between innate (in particular macrophages) and adaptive immune responses in inducing cholesterol and oxysterol metabolism. Moreover, we showed that 25-OHC downregulated Blimp1 expression that is induced by IL-12 and IL-27 and promotes IL-10 production in T cells (42). Our observations suggest that 25-OHC suppress IL-10 secretion from TR1 cells by antagonizing Blimp1 expression.

Cholesterol is converted by the enzyme Ch25h to 25-OHC, which can be further metabolized into 7α,25-OHC in the presence of the cytochrome Cyp7b1. This latter cytochrome is abundant in the liver where it mediates bile acid synthesis. By contrast, Ch25h is poorly expressed in healthy liver, leading to an early suggestion that 25-OHC might generate other biological processes than bile acid production. We observed that IL-27 induced the expression of Ch25h but not Cyp7b1, leading to the production of 25-OHC but not of 7α,25-OHC. This is strengthened by the fact that the effect of 25-OHC on T cells is not dependent on EBI2 expression, a G-protein coupled receptor that binds 7α,25-OHC and 25-OHC with high and modest affinity, respectively.

25-OHC controls transcriptional activities intracellularly and binds to several transcription factors including RORγt (13, 43, 44) and LXRs (34, 35). LXRα and LXRβ have been implicated in cholesterol and fatty acid homeostasis *via* regulation of reverse cholesterol transport and *de novo* fatty acid synthesis. In addition, they regulate inflammatory gene expression and immune cell proliferation. Mechanistically, their effects are attributed to the inhibition of nuclear factor kappa-B and Stat1-mediated signaling pathways (45). LXRs biology has been studied in macrophages principally in atherosclerosis development (46) and more recently in T cells during autoimmune disorders, including experimental autoimmune encephalomyelitis and arthritis models (47–49). In agreement with previous studies, we here show that LXRβ is expressed in activated T cells (36). Interestingly, we observed that LXRβ mRNA is upregulated in IL-27-differentiated Tr1 cells and that addition of exogenous 25-OHC is able to significantly increase LXRβ expression in these cells but not in activated Th0 cells. Moreover, the measurement of 25-OHC within the cellular compartment of TR1 cells emphasizes our hypothesis that 25-OHC acts as an intrinsic transcriptional regulator of LXR. In TR1 cells 25-OHC induced the expression of SREBP1 and repressed the expression of ABCA1 (**Figure 6**) in contrary to what is known in macrophages in which oxysterols inhibit the maturation of SREBP *via* an LXR-independent pathway and induce the transcription of ABCA1 under conditions of cholesterol excess (35). Previous studies on lymphocytes have shown that genes encoding sterol transporters or fatty acids synthesis like ABCA1, ABCG1, or SREBP1 are strongly stimulated upon the addition of synthetic LXR agonist like GW3965 or T0901317 (36, 50). However, the influence of lipid metabolism on CD4<sup>+</sup> T lymphocyte function is still poorly understood.

Liver X receptors have been ascribed anti-inflammatory functions. They have been proposed to negatively regulate

Blimp1, known to induce IL-10. In addition, 25-OHC regulates cholesterol homeostasis by increasing sterol regulatory element binding protein (SREBP1) expression involved in its biosynthesis while simultaneously decreasing its efflux through the inhibition ATP-binding cassette transporter A1 (ABCA1) expression. In summary 25-OHC/LXR leads to both sustained inhibition of IL-10-immunosuppressive response and to the accumulation of cholesterol within TR1 cells.

macrophages inflammatory gene expression (51) and to inhibit TH17 cell generation and thus to mediate anti-inflammatory signals during adaptive immune responses (47). Moreover, LXRs agonist could affect different subsets of T cells including TH1, TH2, and iTreg by limiting T cell proliferation (50). We observed that low concentration of 25-OHC limits the anti-inflammatory response induced by IL-27 in TR1 cells *via* LXR signaling. The production of 25-OHC by TR1 cells is in agreement with the existence of an autocrine and paracrine 25-OHC/LXR amplification loop, inhibiting both TR1 polarization and cholesterol efflux while enhancing cholesterol production by TR1 cells (**Figure 6**). Finally, intracellular cholesterol accumulation has been shown to promote inflammation in innate immunity (15). Oxysterols can thus be considered as fine tuners of inflammation and cholesterol homeostasis during adaptive immune responses.

In humans, TR1 cells were first described in severe combined immunodeficient patients who had developed long-term tolerance to stem cell allografts, suggesting that these cells might naturally regulate immune responses in humans (52). However, human Treg play a deleterious role in cancer as they mediate suppression of antitumor responses and also interfere with immunotherapies. Tumor-associated human TR1 have been shown to be pro-tumorigenic, as they mediate immune suppression (53). Harnessing TR1 cells by modulating cholesterol pathways might open new tools in immunotherapy.

In conclusion, as one of the suppressive T cell subsets, TR1 cells have been described to regulate inflammation, graft-versus-host disease and autoimmunity by producing IL-10. However, excess of anti-inflammatory response may lead to uncontrolled infections or tumor development. The results presented in this study show that IL-27, a main inducer of TR1 cells, induces oxysterols to regulate the strength of the anti-inflammatory response. Taken together, our study identifies Ch25h and its biosynthetic product 25-OHC as negative regulators induced by IL-27 to maintain immune homeostasis *via* LXR signaling. Here, the induction of oxysterols would limit the induction of regulatory T cells to prevent excessive immune regulation that might favor the emergence of viral infections or cancers.

### ETHICS STATEMENT

All procedures and methods were approved by the Cantonal Veterinary Services (SCAV, autorisations VD 3025 and GE 1914).

### REFERENCES


### AUTHOR CONTRIBUTIONS

SV, FC, DD, and AC performed experiments and analyzed results; CP designed the research; LA provided IRF1*<sup>−</sup>*/*<sup>−</sup>* mice and scientific advises; J-ML provided LXRαβ KO mice and scientific advises; JZ and IC performed mass spectrometry analysis and provided scientific advises; and SV and CP elaborated the figures and wrote the paper.

### ACKNOWLEDGMENTS

The authors thank A. W. Sailer for scientific advises; L. Clivaz, Y. Yersin, V. Rochemont, and N. Budoo for technical assistance. This work was supported by the Swiss National Science Foundation (310030\_138430). FC was supported by the EMBO and the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS) fellowship exchange program, AC by the Hirsch foundation. CP holds stipendiary professorships of the Swiss National Science Foundation (#PP00P3\_157476).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://journal.frontiersin.org/article/10.3389/fimmu. 2017.01184/full#supplementary-material.


of 25-hydroxycholesterol. *Immunity* (2013) 38:92–105. doi:10.1016/j. immuni.2012.11.005


**Conflict of Interest Statement:** The authors have no conflict-of-interest to declare. IC and JZ are employees of Novartis Pharma AG and hold stock and stock options in their company. The authors have no additional financial interests to declare.

*Copyright © 2017 Vigne, Chalmin, Duc, Clottu, Apetoh, Lobaccaro, Christen, Zhang and Pot. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# MicroRNA-4443 Causes CD4**+** T Cells Dysfunction by Targeting TNFR-Associated Factor 4 in Graves' Disease

*Yicheng Qi1†, Yulin Zhou1†, Xinxin Chen1 , Lei Ye1 , Qianwei Zhang1 , Fengjiao Huang1 , Bin Cui1 , Dongping Lin2 , Guang Ning1 , Weiqing Wang1 and Shu Wang1 \**

*1Shanghai Clinical Center for Endocrine and Metabolic Diseases, Department of Endocrinology and Metabolism, Ruijin Hospital, Affiliated to Shanghai Jiao-Tong University School of Medicine, Shanghai, China, 2Department of Endocrinology and Metabolism, Shanghai Ninth People's Hospital, Affiliated Shanghai Jiao-Tong University School of Medicine, Shanghai, China*

Context: Aberrant CD4+ T cell function plays a critical role in the process of Graves' disease (GD). MicroRNAs (miRNAs) are important regulators of T cell activation, proliferation, and cytokine production. However, the contribution of miRNAs to CD4+ T cell dysfunction in GD remains unclear.

Objective: To investigate how certain miRNA causes aberrant CD4+ T cell function in GD patients.

Methods: We compared the expression pattern of miRNAs in CD4+ T cells from untreated GD (UGD) patients with those from healthy controls. The most significantly dysregulated miRNAs were selected and their correlations with clinical parameters were analyzed. The effect of miR-4443 on CD4+ T cells cytokines production and proliferation was assessed. The potential gene target was identified and validated.

Results: GD patients had unique pattern of miRNA expression profile in CD4+ T cells comparing to healthy subjects. miR-10a, miR-125b, and miR-4443 were the three most significantly dysregulated miRNAs. The elevated miR-4443 levels were strongly correlated with clinical parameters in an independent dataset of UGD patients (*N* = 40), while miR-4443 was normally expressed in GD patients with euthyroidism and negative TRAb level. We found that miR-4443 directly inhibited TNFR-associated factor (TRAF) 4 expression to increase CD4+ T cells cytokines secretion as well as proliferation through the NF-κB pathway. Furthermore, the TRAF4 levels in GD patients were inversely correlated with miR-4443, and knocking down TRAF4 had a similar effect with miR-4443 overexpression.

Conclusion: The increased expression of miR-4443 induced CD4+ T cells dysfunction by targeting TRAF4, which may cause GD.

Keywords: Graves' disease, CD4**+** T cells, microRNA-4443, TNFR-associated factor 4, NF-**κ**B

# INTRODUCTION

Graves' disease (GD) is a common immune-mediated disease. Lymphocytic infiltrations in thyroid lead to the production of autoantibody against thyrotropin receptor [thyroid stimulating hormone (TSH) receptor antibody (TRAb)], which then mimics the action of TSH, causing excessive thyroid hormone production and hyperthyroidism (1).

### *Edited by:*

*Ritobrata Goswami, Indian Institute of Technology, Kharagpur, India*

### *Reviewed by:*

*Ashutosh Chaudhry, Memorial Sloan Kettering Cancer Center, United States Ju Qiu, Shanghai Institutes for Biological Sciences (CAS), China*

### *\*Correspondence:*

*Shu Wang shuwang9999@163.com*

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

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 28 July 2017 Accepted: 17 October 2017 Published: 01 November 2017*

### *Citation:*

*Qi Y, Zhou Y, Chen X, Ye L, Zhang Q, Huang F, Cui B, Lin D, Ning G, Wang W and Wang S (2017) MicroRNA-4443 Causes CD4+ T Cells Dysfunction by Targeting TNFR-Associated Factor 4 in Graves' Disease. Front. Immunol. 8:1440. doi: 10.3389/fimmu.2017.01440*

**21**

The imbalance of Th1/Th2 cells and Treg/Th17 cells may alter the levels of pro- and anti-inflammatory cytokines, thus contributing to GD onset and development (2). Th1 cell could produce IFN-γ and enhance CXC chemokines production in GD (3). Th2 cell could produce IL-4, IL-5, IL-10, and costimulatory molecules, in turn increase B cell differentiation and antibodies release in GD (4). The proportion of Th17 cells increased in intractable GD patients (5). Our previous study found chemokine (C-C motif) ligand 20 (CCL20) was upregulated in CD4+ T cells from GD patients through NF-κΒ and MAPK pathways (6). CCL20 is a potent chemoattractant for Th17 cells and is closely related to IL17 signal activation. Additionally, we found CD40L overexpression on CD4+ T cells surface which serves as the downstream effector of osteopontin to produce immunoglobulin in GD (7). These findings suggested the dysfunction of CD4+ T cells played an important role in the pathogenesis of GD development. However, the potential mechanisms underlying CD4+ T cell dysfunction need to be clarified.

MicroRNAs (miRNAs) are important regulators of multiple immune pathways (8). Dysfunctions of miRNAs have been indicated in many autoimmune diseases. For example, decreased miRNA-142-3p/5p levels in CD4+ T cells caused T cell activation and B cell hyperstimulation in systemic lupus erythematosus (9). In rheumatoid arthritis, miR-146a was upregulated and T cell apoptosis was suppressed (10). In multiple sclerosis, miR-17 expression in CD4+ T cells was associated with natalizumab treatment and disease relapse (11). However, the miRNA expression profile in CD4+ T cells of GD patients and the pathogenesis underlying miRNA dysregulation needs further investigation.

In this work, we profiled the expression pattern of miRNAs of CD4+ T cells in GD patients and how they are related to GD development.

### SUBJECTS AND METHODS

### Subjects

Forty untreated GD (uGD) patients, 30 euthyroid GD (eGD) patients, 18 TRAb negative-conversion GD (nGD) patients, and 30 age- and sex-matched healthy control donors (hCD) were enrolled from Ruijin Hospital affiliated to Shang-hai Jiao Tong University School of Medicine. All of the uGD patients, without previous treatment, were newly diagnosed through patients' history, clinical manifestation, and laboratory examination. The presence of typical manifestation including heat intolerance, fatigue, increased appetite, increased sweating, weight loss, muscle weakness, tremors, and thyroid gland was diffusely enlarged. The abnormal result of laboratory examinations included high free T3 (FT3), FT4, and low sensitive TSH (sTSH) as well as high TRAb. eGD patients were treated with methimazole (MMI) for 2–4 months and reached normal FT3 and FT4 levels. nGD patients were treated with MMI until TSH, FT3, FT4, and TRAb levels returned to the normal range and remained stable for at least 3 months. Healthy subjects without any past or present history of thyroid disease were enrolled in this study. The subject characteristics and clinical information are shown in **Table 1**. The study was approved by the Research Ethics Board of Ruijin Hospital. Informed written consents were obtained from each participant.

### CD4**+** T Cell Isolation

Fresh peripheral blood mononuclear cells (PBMCs) from GD patients and healthy donors were isolated by Ficoll-Paque centrifugation (Sigma Aldrich, St. Louis, MO, USA). For the purification of CD4+ T cells from fresh PBMCs, positive selection by human CD4 Micro Beads (Miltenyi Biotec, Bergisch Gladbach, Germany) was used according to the manufacturer's instructions. The purity of CD4+ T cells was >95% as analyzed by flow cytometer (BD Biosciences, Bedford, MA, USA). The isolated CD4+ T cells were used for further research [microarray analysis, real-time reverse transcription–polymerase chain reaction (qRT-PCR), and culture].

### miRNA Microarrays

CD4+ T cell samples from three uGD patients and three hCD were used for microarray analysis, and the miRNA profiles were compared between the GD and control group. miRNA microarray profiling was performed using Agilent Human miRNA (8\*60K) V19.0 (Santa Clara, CA, USA) according to the manufacturer's recommendations. Briefly, miRNA molecular in total RNA was labeled using the miRNA Complete Labeling and Hyb Kit. After labeling, each slide was hybridized with 100 ng Cy3-labeled RNA using miRNA Complete Labeling and Hyb Kit. Next, the slides were washed in staining dishes with Gene Expression Wash Buffer Kit. The slides were then scanned using the Agilent Microarray


*a P* < *0.01, untreated GD (uGD) compared with healthy control donors (hCD).*

*bP* < *0.01, euthyroid GD (eGD) compared with uGD.*

*c P* < *0.01, TRAb negative-conversion GD (nGD) compared with uGD.*

*dP* < *0.05, nGD compared with uGD.*

Scanner and Feature Extraction software 10.7 with default settings. Raw data were normalized by the Quantile algorithm in the Gene Spring Software 11.0.

## T Cell Culture and Transfection

CD4+ T cells were rested in RPMI 1640 medium (Gibco, Carlsbad, CA, USA) plus 10% FBS and 1% penicillin/streptomycin overnight. For transfection, CD4+ T cells were transfected with 100 nM miRNA mimics or negative controls (Genepharma, China), 300 nM miRNA inhibitors or inhibitor negative controls (Genepharma), and 300 nM small-interfering RNAs targeting human TNFR-associated factor (TRAF) 4 or negative controls (Ribobio, China) using Lipofectamine 3000 (Invitrogen, Grand Island, NY, USA), according to the manufacturer's protocol. To evaluate the transfection efficiency, these oligonucleotides were labeled with Cy3. Six hours after transfection, we evaluated the transfection efficiency by fluorescence microscope. The percentages of Cy3 positive cells were about 60%. Twelve hours after transfection, the cells were stimulated with 5.0 µg/ml anti-CD3 and 5.0 µg/ml anti-CD28 mAbs (eBioscience, San Diego, CA, USA); the cells and supernatants were then collected for further analysis 48 h later.

## Real-time Reverse Transcription– Polymerase Chain Reaction

Total RNA was isolated using Trizol reagent (Invitrogen). cDNAs were synthesized from whole cellular RNA using a miScript Reverse Transcription Kit (Qiagen, Hilden, Germany). The expression levels of miRNAs were confirmed with a miScript SYBR Green PCR kit and miRNA-specific primers (Qiagen). RNU6-2 was used as the normalization control, and the Light Cycler 480 software (Roche Applied Science, Indianapolis, IN, USA) was used to analyze data. The quantity of mRNA was determined by reverse transcription using PrimeScript RT reagent Kit (Takara, Shiga, Japan) and real-time PCR with SYBR Premix Ex Taq (Takara) and normalized to β-actin. The gene specificity of all of the primers was confirmed by BLAST searches. The primers are presented in Table S1 in Supplementary Material. All of the reactions were performed in triplicate and the results were calculated by the ΔΔCt value and normalized against endogenous controls.

# ELISA

The concentrations of the indicated cytokines were measured quantitatively using ELISA kits according to the manufacturer's procedure. The ELISA kits were purchased from R&D Systems (Minneapolis, MN, USA). The optical density values were read at 450 nm.

# Western Blot Analysis

Cell lysates were subjected to western blot analysis according to standard protocols. After blocking, the membranes were incubated overnight at 4°C with primary antibodies to TRAF4 (Abcam, Cambridge, UK), p65, phosphorylated p65, IκBα, or phosphorylated IκBα (Cell Signaling Technology, Danvers, MA, USA). GAPDH (Cell Signaling Technology) was used as a normalized control. Next, the membranes were incubated with horseradish peroxidase-conjugated secondary antibody (Cell Signaling Technology). Protein bands were illuminated using ECL Prime Western Blotting Detection Reagent (GE Healthcare, Little Chalfont, Buckinghamshire, UK).

# Cell Proliferation Assays

After transfection with the mimics, inhibitors or controls for 12 h, human peripheral blood CD4+ T cells were plated in a 96-well plate and were activated with anti-CD3 and anti-CD28 mAbs. At the appropriate time, the cells were incubated with 10 µL of CCK-8 (Dojindo, Kumamoto, Japan) per well for 3 h at 37°C. The absorbance was determined at 450 nm using an enzyme-labeled instrument (Thermo Scientific, Rockford, IL, USA).

## Luciferase Activity Assay

The psiCHECK-2 vector (Promega, Madison, WI, USA) was used to clone the TRAF4 3′UTR sequence containing the putative miR-4443 binding sites, designated as wild type (WT). Because two putative miR-4443 binding sites were predicted in the TRAF4 gene, reporter plasmids of the corresponding mutation (Mut1 and Mut2) and both sites mutation (Mut3) were constructed. HEK293T cells were seeded in 24-well plates the day before transfection. For each well, 100 ng of wild-type or mutant TRAF4 3′-UTR psiCHECK-2 plasmid was transiently cotransfected with miRNA mimics or negative controls using Lipofectamine 2000 (Invitrogen). Cell lysates were harvested 48 h after transfection, and the cells were subjected to a Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's instructions. Renilla luciferase activities were normalized to firefly luciferase activities to control for the transfection efficiency.

# Statistical Analysis

SPSS version 17.0 and Graph Pad Prism 5.0 were applied in the statistical calculations. Student's *t*-test or the Mann–Whitney *U*-test was performed to compare the differences between the two groups. Correlations between the different variables were analyzed by simple correlation using Spearman's test. Multivariate analysis was performed by multiple linear regression analysis using miR-4443 as a dependent variable, and variables that were significant at *P* < 0.20 level in Spearman's correlations were used as covariates. Data are presented as mean ± SD. *P* < 0.05 was considered significant.

# RESULTS

# Upregulated miR-4443 Expression in CD4**+** T Cells of uGD Patients

We assessed 2006 human miRNAs in CD4+ T cells from three uGD patients and three healthy controls with Agilent Human miRNA array. As shown in **Figure 1A**, we found 11 miRNAs displayed significant differences (*P* < 0.05). Among them, eight miRNAs were suppressed, whereas three miRNA were enhanced in GD CD4+ T cells compared with normal CD4+ T cells. Based on fold change greater than 1.5 and the relative expression level, miR-10a, miR-125b, and miR-4443 were selected for real-time PCR validation. Consistently, we found that miR-10a (*P* < 0.001) and miR-125b level (*P* = 0.0014) was decreased and miR-4443 level (*P* < 0.001) was increased in an independent dataset (40 uGD and 30 controls, **Figure 1B**). Interestingly, the miR-4443 level was strongly correlated with FT3, FT4, and independently with TRAb (**Table 2**). In euthyroid group and TRAb negativeconversion group, the miR-4443 level was reduced to normal levels (eGD vs. hCD, *P* = 0.9961; nGD vs. hCD, *P* = 0.1337, **Figure 1C**).

We then evaluated whether miR-4443 upregulation in uGD patients is a consequence of hyperthyroidism or T cell hyperactivity. We isolated fresh CD4+ T cells from healthy individuals and cultured them with or without T3 treatment. After 24 h or 7 days of cell culture, no significant change was found. Activation of CD4+ T cells by anti-CD3/CD28 antibodies did not influence miR-4443 levels either (Figure S1 in Supplementary Material).

### miR-4443 As a Positive Regulator of CD4**+** T Cell Proliferation and Function

As shown in **Figures 2A,B**, overexpression of miR-4443 greatly increased proinflammatory cytokines, including IL-1β, IL-6 and IL-17, at both mRNA level and protein concentration. The expression of CCL21 was increased compared with that of the negative controls (**Figures 2C,D**). Both the mRNA and protein levels of IFNγ were upregulated in the miR-4443 mimic group (**Figures 2E,F**). Meanwhile, we found the plasma concentrations of IL-1β, IL-6, IL-17, CCL21, and IFNγ were higher in uGD than in eGD and nGD (Figure S2 in Supplementary Material). As illustrated in Table S2 in Supplementary Material, IL-6 and IFNγ concentrations correlated significantly with miR-4443 levels, IL-1β, IL-17, and CCL21 levels also have correlation trend with miR-4443 levels while *P* values > 0.05. Moreover, CCK8 assay showed miR-4443 mimics increased CD4+ T cells proliferation (**Figure 2G**). Consistent with these findings, miR-4443 inhibitors significantly decreased the levels of cytokines, chemokines, and the proliferation of CD4+ T cells obtained from uGD patients (**Figure 3**).

## Targeting TRAF4 by miR-4443 in CD4**+** T Cells

TNFR-associated factor 4, an inhibitor of inflammatory responses, was identified as one of the potential targets of miR-4443 according to TargetScan and miRDB. We constructed plasmids with the TRAF4 3′UTR including the wild-type sequence and mutated sequence (**Figure 4A**). Luciferase reporter assay showed that the WT luciferase activity, rather than Mut3 luciferase activity, was significantly reduced. miR-4443 binds more at the first putative binding sites because miR-4443 inhibited the luciferase activity of the Mut2 but had little inhibitory effect on the Mut1 luciferase activity (**Figure 4B**).

In CD4+ T cells transfected with miR-4443, TRAF4 mRNA and protein level was decreased (**Figure 4C**). In contrast, miR-4443

Qi et al. Function of miR-4443 in GD

inhibitors increased the TRAF4 mRNA and protein expression in CD4+ T cells from uGD patients (**Figure 4D**). Given that TRAF4 is a crucial inhibitor of the NF-κΒ pathway (12–14), we assessed the activity of key players of NF-κΒ pathway. As shown in **Figure 4E**, miR-4443 mimics increased phosphorylation of p65 (p-p65), total p65, and phosphorylation of IκBα (p-IκBα) levels, while reduced IκBα expression in CD4+ T cells from healthy controls. However, miR-4443 inhibitors downregulated p-p65, p65, and p-IκBα expression, while promoted IκBα expression in CD4+ T cells from uGD patients.

### TRAF4 Is Inversely Correlated with miR-4443 in GD

As shown in **Figure 5A**, TRAF4 mRNA levels were significantly downregulated in CD4+ T cells isolated from uGD patients

Table 2 | The association between miR-4443 and classic GD clinical parameters by Spearman correlation and multiple stepwise linear regression analysis.


*r, correlation coefficient;* β*, regression coefficient.*

(*P* = 0.037). Moreover, miR-4443 levels were inversely correlated with TRAF4 expression (*r* = −0.432, *P* = 0.035, **Figure 5B**). Furthermore, knock-down of TRAF4 expression had an effect similar to that of miR-4443 overexpression (**Figures 5C–E**), the expression levels of cytokines and chemokines were increased, the proliferation of CD4+ T cells was upregulated, and the NF-κΒ pathway was activated.

### DISCUSSION

In this study, we indicated miR-4443 was elevated in UGD patients and was significantly correlated with GD development. *In vitro* study, it demonstrated that miR-4443 induced the overexpression of cytokines, chemokines and the proliferation of CD4+ T cells, which could be explained by activated NF-κΒ pathway *via* targeting TRAF4.

MicroRNAs have been implicated in the pathogenesis of GD. Our previous study revealed that differentially expressed miRNAs in PBMCs were associated with GD and T3 exposure (15). A study subsequently indicated the expression profiles of mRNA/miRNA in Tregs might play important roles in GD development (16). Bernecker et al. found miR-200a decreased in CD4+ T cells from GD (17). However, the expression profile of miRNAs in CD4+ T cells from uGD patients and the effect of certain miRNA on GD CD4+ T cells have not been reported. In this study, we identified for the first time that miR-4443, miR-10, and miR-125b were differentially expressed between uGD patients and healthy controls. Further study confirmed miR-4443 was strongly correlated with

(E,F) Relative mRNA expression and protein levels of IFN-γ. (G) Relative proliferation activity of cultured CD4+ T cells. Bars show the mean ± SD of three independent experiments. \**P* < 0.05; \*\**P* < 0.01; \*\*\**P* < 0.001.

clinical GD parameters including FT3, FT4, especially the TRAb levels. After MMI treatment, miR-4443 levels decreased in EGD patients and TRAb-negative GD patients. Unlike previous study, we found there was no significant difference in the expression of miR-200a between uGD patients and controls by both microarray and RT-PCR analysis. This may be due to larger sample size and strict grouping in the current study.

Several miRNAs (e.g., miR-181c, miR-568) levels have already been affected during T cell activation (18, 19). However, the overexpression of miR-4443 that we observed in this study is not likely to be a downstream consequence of increased T lymphocyte activity because miR-4443 levels were not affected by anti-CD3/ CD28 antibodies. Besides, elevated concentrations of T3 and T4 are the major cause of GD signs and symptoms, and the effects of T3 on target tissues are roughly four times more potent than those of T4 (20). Interestingly, we previously found T3 treatment could directly affect the expression of miRNAs in cultured PBMCs from healthy subjects (15). However, T3 was not likely to change the miR-4443 levels in this study. Therefore, miR-4443 may be a potential cause of GD.

MicroRNAs are important regulators of T cell activation, proliferation and the cytokine production. For example, miR-125a, miR-21, miR-31, miR-23b, and miR-142-3p/5p can alter the expression of proinflammatory and anti-inflammatory cytokines *via* immune pathways (9, 21–23). It was previously found that miR-4443 expressed differently in EV71-infected human rhabdomyosarcoma (RD) cells and uninfected RD cells (24). Shefler et al. recently reported that miR-4443 presented in microvesicles derived from activated T cells could regulate mast cell activation by targeting PTPRJ gene (25). These suggested that elevated miR-4443 in GD CD4+ T cells plays a potential role in inflammation response. Importantly, we observed that overexpression of miR-4443 in normal CD4+ T cells increased GD-related cytokines including IL-1β, IL-6, IL-17, CCL21, and IFN-γ production. In contrast, inhibition of miR-4443 in uGD CD4+ T cells reduced these cytokines expression. IL-1β is a proinflammatory cytokine, which has variety of effects on GD, such as modification of thyroid epithelial tightness and induction of cytokine expression in thyroid cells (26). IFN-γ is important in initiating adaptive immune response, and can stimulate the production of CXC chemokines which are powerful chemotactic factor for recruiting Th1 cells to thyroid gland (3). IL-6, secreted by Th2 cell under humoral response, is involved in the activation of T cells, the production of antibodies by B cells and recruitment of dendritic cells in the thyroid (27, 28). Besides, the percentage of Th17 cell and IL-17 level are associated with the pathogenesis of GD (5). Increasing chemokines are confirmed having crucial roles in GD process. Our previous study showed that CCL21 was increased in UGD patients. It may consequently induce circulation of CC-chemokine receptor 7 (CCR7)-expressing cells to thyroid gland (29). It is likely that these increased expression cytokines are released into the circulatory system and up taken by thyroid gland to participate in the local immune response. Meanwhile, the percentage of CD4+ T cells and the ratio of CD4+/CD8+ cells were higher in both GD and Graves' ophthalmopathy than those of healthy controls (30). Our present study found overexpression of miR-4443 could promote CD4+ T cell proliferation and inhibition of miR-4443 level could reduce proliferation. Taken together,

miR-4443 participated in the GD pathogenesis through increas-

activities of NF-κB signaling. Bars show the mean ± SD of three independent experiments. \**P* < 0.05; \*\*\**P* < 0.001.

ing cytokines secretion and promoting CD4+ T cell proliferation. MicroRNAs bind to "seed" sequences in target mRNAs, leading to reduced expression of target mRNAs. According to bioinformatics tools, TRAF4 which could attenuate inflammatory responses was identified as a potential target of miR-4443. Moreover, there are two putative binding sites in the 3′UTR of TRAF4. With the dual-luciferase reporter assay, we indicated that miR-4443 mainly targeted the first binding site (positions 152–159 of TRAF4 3′UTR). Moreover, upregulation of miR-4443 reduced the expression of TRAF4 in healthy CD4+ T cells. The opposite effect was observed when downregulate miR-4443 level in uGD CD4+ T cells. We further found the expression of TRAF4 was decreased in uGD and was negatively correlated with miR-4443 level. All these results suggested that TRAF4 was a direct target gene of miR-4443 in GD CD4+ T cell. Interestingly, one recent study also found that miR-4443 acts in a tumor-suppressive manner by down-regulating TRAF4, which further confirms our findings (31).

TNFR-associated factor 4 is an atypical member of the TRAF superfamily (32). Unlike the other members, TRAF4 negatively regulates immune signaling. TRAF4 interacts with NOD2, TRAF6 and TRIF to inhibit NF-κΒ activation (12–14). NF-κΒ signal pathway is a major regulator of innate and adaptive immune responses (33). The classical NF-κΒ pathway is induced by signals including antigens, TLR ligands and cytokines, involved in controlling inflammation, survival and proliferation (34). Ligands for TNF receptor superfamily members engage the alternate NF-κΒ pathway which is crucial for lymphoid organogenesis. Marinis et al. found IKKα's phosphorylation of serine-426 on TRAF4 was required for the negative regulation of TRAF4 and also found macrophages stably expressing TRAF4 had dampened NF-κΒ responses, indicated by decreased IκBα phosphorylation (35). In experimental autoimmune encephalomyelitis model, the TRAF4-deficient mice had increased numbers of immune cell infiltration in the brain and a significantly higher expression of proinflammatory genes compared with that in the control mice (36). Similar result

was observed in the keratinocyte of TRAF4-deficient mice, showing basal-level and IL-17–induced activation of ERK1/2, JNK, P65, and p38 increased (37). Our results showed that the change of miR-4443 levels affected the NF-κΒ pathway activity. We subsequently observed TRAF4 function in CD4+ T cells, finding cytokines production and CD4+ T cell proliferation was increased obviously when silence the TRAF4 expression in CD4+ T cell by NF-κΒ pathway. Thus, we concluded that miR-4443 binds TRAF4 to activate NF-κΒ signal pathway, leading to aberrant T cell function.

One major limitation of this study is the relatively small sample size. Further long-term prospective investigations with larger sample size are needed to identify miR-4443 as biomarkers. Besides, we did not demonstrate that miR-4443 seems specific to GD because we did not investigate the miR4443 expression in CD4+ T cells from patients with other autoimmune disease.

In conclusion, we first revealed that miR-4443 levels were increased in CD4+ T cells from uGD patients and strongly associated with clinical parameters of GD. We also found *in vitro* assays that ectopic expression of miR-4443 targeted TRAF4 to induce aberrant CD4+ T cells cytokines secretion and proliferation through NF-κΒ pathway. These findings revealed that elevated miR-4443 contributes to the immune-pathogenesis of GD.

### REFERENCES

1. Weetman AP. Graves' disease. *N Engl J Med* (2000) 343:1236–48. doi:10.1056/ NEJM200010263431707

### AUTHOR CONTRIBUTIONS

Conceptualization: SW and WW. Data curation: YQ and YZ. Formal analysis: YQ and FH. Funding acquisition: SW. Investigation: YQ and XC. Methodology: XC and QZ. Project administration: SW, WW, and GN. Resources: LY and DL. Software: BC. Supervision: SW. Validation: YQ and YZ.

### FUNDING

This study was supported by the grants from the National Natural Science Foundation of China (nos. 81270872 and 81570707).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/article/10.3389/fimmu.2017.01440/ full#supplementary-material.

Figure S1 | (A) Expression of miR-4443 in CD4+ T cells from healthy human with or without T3 treatment. (B) Expression of miR-4443 in CD3/CD28 stimulated CD4+ T cells from healthy human. Bars show the mean ± SD of three independent experiments.

Figure S2 | Plasma concentrations of IL-1β, IL-6, IL-17, CCL21, and IFNγ. Bars show the mean ± SD. \*\*P < 0.01; \*\*\*P < 0.001.


beta chemokine CCL2 serum levels in chronic autoimmune thyroiditis. *Eur J Endocrinol* (2005) 152:171–7. doi:10.1530/eje.1.01847


**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 © 2017 Qi, Zhou, Chen, Ye, Zhang, Huang, Cui, Lin, Ning, Wang and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Differentiation and Transmigration of CD4 T Cells in Neuroinflammation and Autoimmunity

*Sandip Ashok Sonar and Girdhari Lal\**

*National Centre for Cell Science, Pune, India*

CD4+ T cells play a central role in orchestrating protective immunity and autoimmunity. The activation and differentiation of myelin-reactive CD4+ T cells into effector (Th1 and Th17) and regulatory (Tregs) subsets at the peripheral tissues, and their subsequent transmigration across the blood–brain barrier (BBB) into the central nervous system (CNS) parenchyma are decisive events in the pathogenesis of multiple sclerosis and experimental autoimmune encephalomyelitis. How the Th1, Th17, and regulatory Tregs transmigrate across the BBB into the CNS and cause CNS inflammation is not clearly understood. Studies with transgenic and gene knockout mice have unraveled that Th1, Th17, and Tregs play a critical role in the induction and resolution of neuroinflammation. However, the plasticity of these lineages and functional dichotomy of their cytokine products makes it difficult to understand what role CD4+ T cells in the peripheral lymphoid organs, endothelial BBB, and the CNS parenchyma play in the CNS autoimmune response. In this review, we describe some of the recent findings that shed light on the mechanisms behind the differentiation and transmigration of CD4+ T cells across the BBB into the CNS parenchyma and also highlight how these two processes are interconnected, which is crucial for the outcome of CNS inflammation and autoimmunity.

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Manu Rangachari, Laval University, Canada Markus Kleinewietfeld, VIB-UGent Center for Inflammation Research (IRC), Belgium*

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

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 31 August 2017 Accepted: 16 November 2017 Published: 29 November 2017*

### *Citation:*

*Sonar SA and Lal G (2017) Differentiation and Transmigration of CD4 T Cells in Neuroinflammation and Autoimmunity. Front. Immunol. 8:1695. doi: 10.3389/fimmu.2017.01695*

Keywords: blood–brain barrier, experimental autoimmune encephalomyelitis, CD4 T cells, neuroinflammation, transendothelial migration

# INTRODUCTION

Homeostasis of central nervous system (CNS) is maintained by various mechanisms operating in both the CNS and the peripheral immune system. Due to the presence of barriers, CNS antigens are not exposed to cells of the peripheral immune system, which ensures a lack of effector immune response to CNS antigens in the steady state (1). However, upon recognition of CNS-derived antigens or cross-reactive microbial antigens, the peripheral CD4<sup>+</sup> T cells have escaped from the central tolerance, mount a robust immune response, and infiltrate into the CNS. Such infiltration of CD4<sup>+</sup> T cells causes CNS autoimmune diseases such as multiple sclerosis (MS) and experimental

**Abbreviations:** APCs, antigen-presenting cells; BBB, blood–brain barrier; CFA, complete Freund's adjuvant; CSF, cerebrospinal fluid; CNS, central nervous system; EAE, experimental autoimmune encephalomyelitis; ECs, endothelial cells; ICAM-1, intercellular adhesion molecule 1; ILCs, innate lymphoid cells; LFA-1, lymphocyte function-associated antigen 1; MOG35–55, myelin oligodendrocyte glycoprotein peptide amino acid 35–55; MS, multiple sclerosis; TEM, transendothelial migration; Tregs, Foxp3+ regulatory T cells; Tr1 cells, T regulatory 1 cells; VCAM-1, vascular cell adhesion molecule 1; VLA-4, very late antigen 4.

autoimmune encephalomyelitis (EAE) (2). MS is a human autoimmune demyelinating disease of the CNS characterized by massive infiltration of inflammatory lymphocytes and myeloid cells into the brain and spinal cord, leading to demyelination, axonal damage, and loss of neuromuscular functions (3). Most of the clinical and pathological features of MS are recapitulated in the animal model, EAE, which is one of the important models used to study CNS inflammatory diseases. EAE is also used for evaluating the efficacy of several therapeutic strategies to control neuroinflammation and autoimmunity (3).

### ACTIVATION OF MYELIN-SPECIFIC CD4**<sup>+</sup>** T CELLS DURING CNS INFLAMMATION AND AUTOIMMUNITY

There is a long-standing hypothesis in the field that the activation of myelin-specific CD4<sup>+</sup> T cells requires a trigger from some environmental factors (4). EAE is induced by activating myelinreactive lymphocytes through peripheral immunization with myelin antigens. EAE can also be induced in susceptible animal hosts either by subcutaneous immunization (s.c.) with myelin antigens emulsified in complete Freund's adjuvant (CFA) or by the adoptive transfer of *in vitro*-activated myelin-specific CD4<sup>+</sup> T cell subsets such as Th1 and Th17 (5, 6). Among the various myelin proteins, proteolipid protein, myelin basic protein, and myelin oligodendrocyte glycoprotein (MOG), and their corresponding immunodominant peptides have been extensively used to induce EAE in different rodent hosts (2, 5). However, this also requires administration of pertussis toxin, highlighting the importance of environmental factors in the development of CNS pathology (7, 8). CD4<sup>+</sup> and CD8<sup>+</sup> T cells that have low affinity/avidity for myelin antigens, escape thymic selection, and are localized mainly to the secondary lymphoid organs, where they remain in the tolerant state under homeostatic conditions (9). The subcutaneous deposition of myelin peptide emulsion attracts and activates professional antigen-presenting cells (APCs), such as dendritic cells, macrophages, and B cells, at the site of injection. These APCs take up the antigens and migrate to the draining lymph nodes, where they process and present antigenic peptides to the T lymphocytes. Immunization (s.c.) of C57BL/6 mice with MOG35–55-CFA emulsion along with intravenous pertussis toxin were found to induce antigenspecific Th1 and Th17 cells in the draining lymph nodes, and at the same time limit the regulatory T (Treg) number and function (7, 10). Interestingly, T-cell receptor (TCR)-transgenic mice, such as 2D2 mice in which CD4<sup>+</sup> T cells are engineered to express MOG35–55-specific TCR, develop spontaneous CNS autoimmunity (11), suggesting the importance of CD4<sup>+</sup> T cells in EAE. By using several knockout and transgenic mice, molecules involved in the TCR and costimulatory and coinhibitory signaling in the activation, proliferation, and differentiation of myelin-specific CD4<sup>+</sup> T cells have been evaluated. Furthermore, several members of the TNF-receptor superfamily critically regulate the CD4+ T cell response both in the secondary lymphoid organs and inflamed CNS and perturb the pathology of EAE (12).

# DIFFERENTIATION OF MYELIN-SPECIFIC CD4**+** T CELLS

The naive CD4<sup>+</sup> T cells, when stimulated by myelin APCs and specific cytokines, differentiate into various effector and regulatory lineages (**Figure 1**). Th1 cells secrete IFN-γ and TNF-α and are critical for controlling intracellular pathogens and induction of delayed-type hypersensitivity response. Excess activation of Th1 is involved in many organ-specific inflammations, including MS and EAE (13). In the presence of a strong TCR signal, IL-12/ STAT4 and IFN-γ/STAT1 signaling induces the Th1-specific transcription factor T-bet, which amplifies IFN-γ/STAT1/T-bet signaling and drives Th1 differentiation (14, 15). Furthermore, T-bet cooperatively interacts with other transcription factors such to RUNX1, RUNX3, GATA3, IRF4, and BCL6 to inhibit the differentiation of alternative CD4<sup>+</sup> T cell subsets (16, 17). Mice deficient in Th1-associated factors such as T-bet and STAT4 are resistant to the development of EAE (18), whereas IFN-γ−/−, IFNγR<sup>−</sup>/<sup>−</sup>, and STAT1<sup>−</sup>/<sup>−</sup> mice develop more severe EAE (19). This suggests that Th1 cells play a critical role in the pathogenesis of EAE and MS through diverse mechanisms.

In humans and mice, various cytokines induces the differentiation of Th17 cells with diverse phenotypes and functions. The conventional Th17 cells generated in the presence of TGF-β1 and IL-6 are non-pathogenic during EAE, being involved in the maintenance of mucosal surface homeostasis and anti-bacterial defense (20, 21). However, several other factors such as IL-1β, IL-23, and TGF-β3 have been identified to favor the generation and maintenance of highly pathogenic Th17 cells during EAE (21–24). Mice deficient in IL-23 or IL-23R are completely resistant to the development of EAE (23, 25). Th17 cells that coexpress RORγt and T-bet and produce both IL-17A and IFN-γ, are highly pathogenic and preferentially recruited into the CNS, suggesting that T-bet enhances the pathogenicity of Th17 cells (13, 21, 26). The detailed development, phenotypic, and functional differences between pathogenic and non-pathogenic Th17 cells have been reviewed recently (21). Interestingly, a recent report shows that IL-23 induces a switch from CCR6 to CCR2 usage and controls the development and migration of highly encephalitogenic granulocyte macrophagecolony stimulating factor (GM-CSF)-expressing Th17 cells into the CNS, suggesting that homing receptors and pathogenic functions are imprinted during differentiation (27). It has been shown that IL-12 induces STAT4 signaling and also triggers GM-CSF expression in Th1 cells and promote EAE development (28). Interestingly, GM-CSF-producing Th1-like cells are also found in the cerebrospinal fluid (CSF) of MS patients (29), suggesting that GM-CSF may contribute to the pathogenic function of Th17, IFN-γ expressing ex-Th17 and Th1 cells. The chronic inflammatory signals can affect the transcriptional and/ or epigenetic signature and control the plasticity of Th1, Th17, and Tregs in the inflamed CNS and lymphoid organs during EAE (30). The IL-23-induced alternatively activated Th17 (T-bet<sup>+</sup>RORγt <sup>+</sup>) or ex-Th17 cells that acquire T-bet and IFN-γ and express negligible RORγt and IL-17 are more pathogenic (31, 32). Similarly, transdifferentiation of Th17 into Tregs, and Tregs to effector Th1, Th2, and Th17 cells are also known (33, 34).

FIGURE 1 | Generation of myelin-reactive effector Th1, Th17, and regulatory iTreg cells and their plasticity in the central nervous system (CNS) parenchyma during experimental autoimmune encephalomyelitis. Upon appropriate myelin antigen presentation, and in the presence of adequate costimulatory molecules and cytokine signaling, naive CD4+ T cells are activated and give rise to effector (Th1 and Th17) and regulatory (iTreg) T cells. Th17 cells are trafficked into the CNS mainly through the choroid plexus using CCR6–CCL20 interactions, whereas Th1 cells cross the blood–brain barrier mostly using CXCR3–CXCL9/10/11 interactions. The reactivation of the infiltrating Th1 and Th17 cells by the local antigen-presenting cells (APCs) in the CNS boosts the cytokine secretion and pathogenic potential of Th1 and Th17 cells. Under the influence of IL-12 and IL-23 produced by the APCs, Th17 cells acquire the Th1 (ex Th17) and highly pathogenic Th17 (RORγt <sup>+</sup>T-bet+) phenotype.

IL-23-induced phosphorylation and nuclear localization of STAT3/STAT4 heterodimer has been shown to control the generation of encephalitogenic Th1/Th17 cells (35). More studies are needed to understand the minimum essential cytokine stimuli is require to generate highly pathogenic Th17 cells and that may help in designing better therapeutic strategies to control the inflammation and autoimmunity.

Regulatory T cells are suppressive in nature and known to control the myelin-reactive CD4<sup>+</sup> effector T cell response and are therefore pivotal in regulating CNS inflammation during EAE (36, 37). Based on their developmental pathways, they are classified as natural Tregs (nTrgs; thymic-derived) or induced Tregs (iTregs; extrathymic-derived) (34, 38, 39). nTregs express Foxp3, a lineage-defining transcription factor, and CD25 (IL-2Rα) during their development in the thymus. iTregs are generated from naive CD4<sup>+</sup> T cells in the presence of TGF-β in the peripheral lymphoid tissue, which induces the expression of Foxp3 through STAT5 activation (40, 41). In both humans and mice Tregs, Foxp3 required for their suppressive capacity, and its deficiency is associated with the development of spontaneous autoimmunity (38, 39, 42). Tregs employ diverse contact-dependent (expression of CTLA4, FasL, and LAG3) and contact-independent (secretion of TGF-β and IL-10, deprivation of IL-2, and ectonucleotidases CD39/CD73-mediated conversion of extracellular inflammatory ATP/ADP into adenosine) mechanisms to inhibit the functions of myelin-reactive pathogenic T cells and other effector myeloid cells (43). The generation of myelin antigen-reactive CD4<sup>+</sup> T cell subsets and the plasticity of Th1 and Th17 cells in inflamed CNS during EAE are depicted in **Figure 1**.

Other subsets of CD4<sup>+</sup> T cells, such as Th9, T follicular helper (Tfh), T follicular regulatory, and T regulatory 1 (Tr1), are also reported to contribute to the development of neuroinflammation and autoimmunity. The adoptive transfer of MOG-specific Th9 cells is known to induce EAE in C57BL/6 recipients (44). Moreover, IL-9 is required for mast cell activation, which has previously been shown to degrade myelin during CNS inflammation (45). Tfh cells, which are mainly involved in the regulation of germinal center reaction, are also hypothesized to participate in the pathogenesis of MS and EAE by virtue of their ability to help in the formation of ectopic lymphoid follicles in the inflamed CNS (46). Tr1 cells which are differentiated *in vitro* by culturing in the presence of TGF-β plus IL-27 show the Foxp3<sup>−</sup>IL-10<sup>+</sup>IFN-γ+ phenotype. Tr1 cells are known to play a significant role in the development of transplantation tolerance (47), but their exact role in EAE is not known. However, IL-27Rα−/<sup>−</sup> mice are hypersusceptible to the development of EAE, possibly because of a lack of IL-27-mediated control of Th17, as well as the absence of Tr1-mediated suppression (48). There are several Th1- and Th17-associated molecules, which play an important role in the pathogenesis of EAE, and their deficiency affects the severity of the disease (**Table 1**).

### MIGRATION OF MYELIN-SPECIFIC CD4**<sup>+</sup>** T CELLS INTO THE CNS DURING INFLAMMATION AND AUTOIMMUNITY

Various cellular and molecular interactions help in maintaining the blood–brain barrier (BBB) integrity and immune quiescence into the CNS. It has been reported that astrocytes secrete Sonic hedgehog and endothelial cells (ECs) express Hedgehog receptors, and interaction of these receptor ligand promote the BBB formation and integrity (62). Migration of immune cells into the CNS parenchyma is a highly regulated process which occurs at various anatomical sites of the CNS, such as at the choroid plexus of the blood–CSF barrier, as well as across the BBB at postcapillary venules. The transmigration across the BBB is a very dynamic process and depends on a series of sequential and interdependent steps constituting tethering and rolling of immune cells, chemokine-induced activation, followed by polarization, crawling, the arrest of immune cells, and finally diapedesis of cells across the BBB ECs.

Intravital microscopic analysis of encephalitogenic T cell interactions with inflamed brain and spinal cord microvessels have revealed that the P-selectin glycoprotein ligand (PSGL-1)– P/E-selectin interaction mediates the initial rolling and tethering of CD4<sup>+</sup> and CD8<sup>+</sup> T cells (63). However, deficiencies of E- and P-selectin or PSGL-1 do not protect mice from EAE (64, 65), suggesting the redundant roles of these molecules during neuroinflammation. Followed by tethering, the α4β1-integrin on T cells interacts with endothelial vascular cell adhesion molecule 1 (VCAM-1) and is required for firm adhesion of T cells (66). Further studies are needed to clarify whether encephalitogenic T cells use alternative molecules for rolling and tethering onto the inflamed BBB.

The G protein-coupled receptors, such as chemokines and eicosanoids displayed on the luminal surface of the BBB ECs, trigger the integrin activation that leads to T cell firm arrest on the vascular endothelium. During EAE, ECs are shown to express CCL2, CCL19, and CCL21, which mediate firm arrest of CCR2<sup>+</sup> monocytes and DCs, and CCR7<sup>+</sup>CD4<sup>+</sup> T cells (67). However, the exact role of these interactions in the transendothelial migration (TEM) of encephalitogenic CD4<sup>+</sup> T cells remains to be determined. Mice that overexpress CCL19, CCL21, or CXCL10 molecules in the CNS do not show hypersusceptibility to the EAE development (68, 69), suggesting that the functions of these chemokine interactions are tightly regulated. Chemokine receptor-induced signaling leads to a conformational change in the integrin molecules on CD4<sup>+</sup> T cells, and which causes an increase in their affinity for their cognate ligands. Inflamed endothelial vessels in the CNS parenchyma upregulate the expression of the intercellular adhesion molecule 1 (ICAM-1) and VCAM-1, and their respective ligands, αLβ2 [lymphocyte function-associated antigen 1 (LFA-1)], and α4β1 [very late


antigen 4 (VLA-4)] integrins, are expressed on the encephalitogenic CD4<sup>+</sup> T cells (70). Multiple investigators have shown that LFA-1–ICAM-1 and VLA-4–VCAM-1 interactions are critically involved in the firm arrest of CD4<sup>+</sup> T cells onto the inflamed cerebral vessels or primary brain EC monolayers (71). Moreover, LFA-1–ICAM-1 interactions dictate the polarization and crawling of CD4<sup>+</sup> T cells onto the inflamed vessels ECs, but VLA-4–VCAM-1 interactions do not (66). The anti-α4-integrin antibody, natalizumab has been approved for the treatment of relapsing-remitting MS (72). Interestingly, α4β1-integrin– VCAM-1 interaction arrest encephalitogenic Th1 cells onto the spinal cord microvessels, whereas LFA-1–ICAM-1/2 regulates Th17 adhesion to the endothelial barrier in the brain. This suggests that Th1 cells preferentially use α4-integrin, whereas Th17 transmigration across the BBB is α4-integrin independent (73, 74). Other cell adhesion molecules such as activated leukocyte cell adhesion molecule and melanoma cell adhesion molecule (MCAM) are also known to regulate transmigration of the CD4<sup>+</sup> and CD8<sup>+</sup> T cells and CNS autoimmunity (75–77). The MCAM expression in lymphocytes are associated with GM-CSF, IL-22, and IL-17A/IFN-γ coproducing Th17 cells (75), and antibodymediated blocking of MCAM controls the CNS autoimmunity (75, 77). Recently, αvβ3-integrin has been shown to control Th17 migration into the CNS, and a lack of β3 subunits ameliorates EAE (78). A genetic deficiency or functional blocking of most of the integrins and their ligands have yielded varied results in controlling the disease. These discrepancies might be due to a difference in the induction of EAE models, wherein the contribution of Th1 and Th17 varied to a significant extent. Live cell imaging studies of TEM across ICAM-1 and ICAM-2-deficient brain endothelial monolayers have revealed the presence of an alternative pathway for CD4<sup>+</sup> T cell diapedesis (79).

Once arrested, CD4<sup>+</sup> T cells start crawling on inflamed CNS microvessels to search the sites for diapedesis. TEM of CD4<sup>+</sup> T cells across the BBB can occur through both, intercellular junctions and the cell body, known as paracellular and transcellular migration, respectively (80, 81). However, the physiological factors that dictate the choice of the transmigration route are not known, and thus forming an active area of investigation. Several studies have demonstrated that during TEM there is a fast and very dynamic remodeling of the endothelial junctional proteins that guide the migration of CD4<sup>+</sup> T cells through paracellular route (82). The involvement of some of the adhesion molecules and junctional proteins such as PECAM-1, CD99, Claudin-5, VE-cadherin, and JAMs in the regulation of paracellular TEM has been very well studied (83). Upon attachment of the CD4<sup>+</sup> T cells to the apical surface of inflamed brain ECs, a rapid clustering of ICAM-1 and VCAM-1 occurs around the transmigrating CD4<sup>+</sup> T cells, resulting in the formation of transmigratory cups enriched with actin filaments (84). These clustering events trigger the various signaling pathways, leading to generation of an intracellular calcium flux, phosphorylation of key molecules that regulate the actin cytoskeleton, and production of reactive oxygen and nitric oxide species, which ultimately result in junctional disassembly (85). Numerous factors such as shear force, cytokine-induced inflammatory changes in the brain ECs, type of lymphocytes, and levels of junctional tightness have been hypothesized as the potential factors that regulate transcellular TEM of cells at the BBB. A recent study showed that cytokine-induced increased levels of ICAM-1 on the apical surface of primary mouse brain microvascular cell monolayers promote the transcellular TEM of CD4<sup>+</sup> T cells possibly because of high occupancy of its receptor LFA-1 on CD4<sup>+</sup> T cells (80, 81). In addition, overexpression of the C-terminal deletion mutant form of ICAM-1 in primary brain endothelial monolayers inhibits the TEM of leukocytes by reducing transcellular migration (86). Recently, the critical role of the lateral border recycling compartment, a recently identified endothelial specific subcellular compartment (enriched with PECAM-1 and CD99), have been shown to support both paracellular and transcellular TEM (87). The caveolin-rich transmigratory cups that surround the migrating CD4+ T cells have also been associated with transcellular TEM (88). While transcellular migration is impaired in caveolin-1-deficient ECs, they show higher paracellular TEM, suggesting that in the absence of one pathway another route can compensate (88). Similarly, whether a lack of paracellular migration at the endothelial monolayers of high barrier tightness, such as the BBB favors the transcellular route remains to be determined.

After crossing the endothelial vessels of the BBB, CD4<sup>+</sup> T cells encounter the glial (glia limitans) basement membrane, and breaching this acellular structure represents the final step of trafficking into the CNS. The endothelial basement membrane at the BBB is characterized by the presence of laminin α4 and α5. It has been demonstrated that encephalitogenic CD4<sup>+</sup> T cells cross the endothelial basement membrane through α6β1-integrin– laminin α4 interactions (89). On the other hand, laminin α5 in the endothelial basement membrane inhibits migration (89). Under physiological conditions, CXCL12 is abundantly expressed on the abluminal surface of brain endothelial microvessels, which inhibits the migration of CXCR4+ leukocytes into the CNS parenchyma (90). The cytokine-induced CXCR7 expression on these endothelial microvessels changes the localization of CXCL12 to the luminal surface, resulting in TEM of CXCR4<sup>+</sup> leukocytes at the peak of the EAE (91). In contrast to the endothelial basement membrane, the glia limitans is enriched with laminin α1 and α2. Since encephalitogenic CD4<sup>+</sup> T cells do not interact with laminin α1 and α2, they depend on matrix metalloproteinases (MMPs) to cross the basement membranes. Various types of MMPs, such as MMP-2, MMP-7, MMP-8, and MMP-9, have been identified in the CSF and lesions of MS and EAE. During EAE, MMP2 and MMP9 expression are specifically increased, and their combined action is positively correlated with the migration of CD4<sup>+</sup> T cells across glia limitans (92). One of the targets of MMP2 and MMP9 is β-dystroglycan, a receptor which anchors astrocytic endfeet to the parenchymal basement membrane, leading to secretion of chemokines by the astrocytes at the glia limitans (93). Meningeal inflammation actively controls local CD4<sup>+</sup> T cell reactivation and transmigration into the CNS. It has been recently shown that Th17-derived IL-17 and IL-22 activate meningeal stromal cells, which support the *de novo* IL-17 responses in the meninges (94). Interestingly, a finding has extended our current view about the role of T-bet beyond the generation of pathogenic Th1/Th17 cells and showed that T-bet expressing NKp46<sup>+</sup> innate lymphoid cells (ILCs) promote meningeal inflammation and regulate EAE development by supporting Th17 migration into the CNS (95). Thus, the relay of coordinated signaling induced by cytokines, chemokines, and cell adhesion molecules in the ECs of the BBB and migrating CD4+ T cells orchestrate the multistep migration of encephalitogenic CD4<sup>+</sup> T cells into the CNS parenchyma.

### FUTURE PERSPECTIVE

Both genetics and environmental factors cooperate to program auto-reactive CD4<sup>+</sup> T cells to perform both pathogenic and regulatory functions during the course of autoimmune CNS pathologies. Recent evidence has suggested that the phenotype and functions of pathogenic Th1, Th17, and regulatory Tregs cells are regulated at various anatomic and physiological levels. The APCs in the periphery, tertiary lymphoid structures, stromal cells, and subsets of ILCs in the meninges, ECs at the BBB, and various CNS resident and infiltrating cells in the CNS parenchyma tightly control the activation, differentiation, and migration of CD4+ T cells that dictate the induction, maintenance, and resolution of autoimmune neuroinflammation. While considerable evidence already links Th1 and Th17 cells to the pathology of CNS autoimmunity, this list of cells continues to grow with recently identified subsets of CD4<sup>+</sup> T cells, the IL-9-secreting Th9 cells, and IL-10-secreting Foxp3<sup>−</sup> Tr1 cells (44, 96). However, the exact role of these cells and their associated molecules on the BBB, CNS resident cells, and other effector and regulatory leukocytes in the inflamed CNS parenchyma, and their overall impact in shaping neuroinflammation warrants further investigation. The ECs of the BBB have been recently shown to promote antigen-specific Th1 and Th17 migration through myelin-antigen presentation (97). However, the qualitative and quantitative differences in the

### REFERENCES


regulation of transmigration of Th1, Th17, Th9, Tregs, and Tr1 cells across the BBB is not known and needs further attention. Experiments with knockout mice have revealed a great deal of information about the role of CD4<sup>+</sup> T cell subsets and their lineage-associated transcription factors, cytokines, and homing receptors in the induction and propagation of CNS inflammation. The complete resistance of EAE in mice that lack T-bet, RORγt, IL-23R, and GM-CSF is attributed to the absence of pathogenic Th1 and Th17 functions (21, 26). However, these molecules are not exclusively expressed in CD4<sup>+</sup> T cells, and the contribution of other myeloid and lymphoid cells, including subsets of γδ T cells, ILC1 and ILC3 that express/respond to these molecules, needs to be further investigated. Therefore, to better understand the pathophysiology of CD4<sup>+</sup> T cells in autoimmune CNS diseases, we need to dissect out the contributions made by the other cell types that share the transcription factors, cytokines, and homing receptors of CD4<sup>+</sup> T cell lineages. A reductionist approach may help in probing the exact role of CD4<sup>+</sup> T cell subsets through the course of CNS inflammation and autoimmunity.

### AUTHOR CONTRIBUTIONS

SS and GL planned and wrote the manuscript.

### FUNDING

This work was supported by Department of Biotechnology (DBT), Government of India (Grants, BT/PR15533/ MED/30/1616/2015; BT/PR14156/BRB/10/1515/2016; and BT/ PR4610/MED/30/720/2012 to GL). SS received Senior Research Fellowship (SRF) from Council of Scientific and Industrial Research (CSIR), Government of India. Authors also thank Dr. Jyoti Rao for critical reading and edits.


the activation and differentiation of autoreactive T cells. *J Immunol* (1998) 161(12):6480–6.


and promotes their recruitment to the central nervous system. *Brain* (2012) 135(Pt 10):2906–24. doi:10.1093/brain/aws212


at sites of leukocyte extravasation in experimental autoimmune encephalomyelitis. *J Exp Med* (2006) 203(4):1007–19. doi:10.1084/jem.20051342


antigens by the brain endothelium guides antigen-specific T cell migration. *Elife* (2016) 5:e13149. doi:10.7554/eLife.13149

**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 © 2017 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) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*, Inna S. Kuznetsova1*

*,* 

### *Tatyana Veremeyko1 , Amanda W. Y. Yung1 Igor Pomytkin2 , Alexey Lyundup2 , Tatyana Strekalova3,4, Natasha S. Barteneva5,6 and Eugene D. Ponomarev1 \**

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Robert Axtell, Oklahoma Medical Research Foundation, United States Caroline Pot, Lausanne University Hospital, Switzerland*

### *\*Correspondence:*

*Eugene D. Ponomarev eponomarev@cuhk.edu.hk*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 17 November 2017 Accepted: 09 January 2018 Published: 25 January 2018*

### *Citation:*

*Veremeyko T, Yung AWY, Dukhinova M, Kuznetsova IS, Pomytkin I, Lyundup A, Strekalova T, Barteneva NS and Ponomarev ED (2018) Cyclic AMP Pathway Suppress Autoimmune Neuroinflammation by Inhibiting Functions of Encephalitogenic CD4 T Cells and Enhancing M2 Macrophage Polarization at the Site of Inflammation. Front. Immunol. 9:50. doi: 10.3389/fimmu.2018.00050*

*1School of Biomedical Sciences, The Chinese University of Hong Kong, Shatin, Hong Kong, 2Department of Advanced Cell Technologies, Institute for Regenerative Medicine, I.M. Sechenov First Moscow State Medical University, Moscow, Russia, 3Department of Neuroscience, Maastricht University, Maastricht, Netherlands, 4 Laboratory of Psychiatric Neurobiology, Institute of Molecular Medicine, I.M. Sechenov First Moscow State Medical University, Moscow, Russia, 5Department of Pediatrics, Program in Cellular and Molecular Medicine, Children's Hospital Boston, Harvard Medical School, Boston, MA, United States, 6School of Science and Technology, Nazarbayev University, Astana, Kazakhstan*

*, Marina Dukhinova1*

Although it has been demonstrated that cAMP pathway affect both adaptive and innate cell functions, the role of this pathway in the regulation of T-cell-mediated central nervous system (CNS) autoimmune inflammation, such as in experimental autoimmune encephalomyelitis (EAE), remains unclear. It is also unclear how cAMP pathway affects the function of CD4 T cells *in vivo* at the site of inflammation. We found that adenylyl cyclase activator Forskolin besides inhibition of functions autoimmune CD4 T cells also upregulated microRNA (miR)-124 in the CNS during EAE, which is associated with M2 phenotype of microglia/macrophages. Our study further established that in addition to direct influence of cAMP pathway on CD4 T cells, stimulation of this pathway promoted macrophage polarization toward M2 leading to indirect inhibition of function of T cells in the CNS. We demonstrated that Forskolin together with IL-4 or with Forskolin together with IL-4 and IFNγ effectively stimulated M2 phenotype of macrophages indicating high potency of this pathway in reprogramming of macrophage polarization in Th2- and even in Th1/Th2-mixed inflammatory conditions such as EAE. Mechanistically, Forskolin and/or IL-4 activated ERK pathway in macrophages resulting in the upregulation of M2-associated molecules miR-124, arginase (Arg)1, and Mannose receptor C-type 1 (Mrc1), which was reversed by ERK inhibitors. Administration of Forskolin after the onset of EAE substantially upregulated M2 markers Arg1, Mrc1, Fizz1, and Ym1 and inhibited M1 markers nitric oxide synthetase 2 and CD86 in the CNS during EAE resulting in decrease in macrophage/microglia activation, lymphocyte and CD4 T cell infiltration, and the recovery from the disease. Forskolin inhibited proliferation and IFNγ production by CD4 T cells in the CNS but had rather weak direct effect on proliferation of autoimmune T cells in the periphery and *in vitro*, suggesting prevalence of indirect effect of Forskolin

**39**

on differentiation and functions of autoimmune CD4 T cells *in vivo*. Thus, our data indicate that Forskolin has potency to skew balance toward M2 affecting ERK pathway in macrophages and indirectly inhibit pathogenic CD4 T cells in the CNS leading to the suppression of autoimmune inflammation. These data may have also implications for future therapeutic approaches to inhibit autoimmune Th1 cells at the site of tissue inflammation.

Keywords: Th1 cells, neuroinflammation, Forskolin, cAMP, microRNA-124, macrophage, M2 polarization, ERK

### INTRODUCTION

Various pathways are known to affect differentiation and function of CD4 T cells *in vivo* during inflammation associated with autoimmunity or infection. One of most common and important pathways in the process is cAMP pathway that is known to be involved in negative regulation of T cell activation and proliferation (1). However, more detailed and recent studies demonstrated that cAMP-inducing agents *Cholera Toxin*, 8-bromo-cAMP, and FDA-approved drug Forskolin inhibited proliferation of Th1 but not Th2 effector cells *in vitro* (2). In addition, it was shown that *Cholera Toxin*-induced expansion rather than inhibition of Th17 cells *in vivo* (3). *Pertussis toxin*, which was specifically used for active induction of central nervous system (CNS) autoimmune inflammation, was also shown to activate adenylyl cyclase leading to increase in cAMP (4). However, *Pertussis toxin* stimulated rather than inhibited expansion of Th1 cells *in vivo* leading to development of CNS autoimmune inflammation (5). Moreover, selective inhibition of cAMP pathway *in vivo* in CD4 T cells demonstrated that cAMP was required for differentiation and proliferation of Th1 and Th17 cells but not Th2 and Tregs (6). Thus, exact role of cAMP pathway in the modulation of function of effector T cells during CNS autoimmune inflammation remains unclear. An important factor that could affect functions of T cells in the tissues during inflammation are tissue-resident and blood-derived macrophages that are recruited to the sited of inflammation and could be also affected by cAMP-inducing agents.

During inflammation, macrophages become activated under the influence of T-cell-derived cytokines or pathogens leading to two or more distinct (polarized) states. Polarization of macrophages toward the classic M1 phenotype is induced by Th1 cytokines such as IFNγ and the alternative M2 phenotype induced by Th2 cytokines such as IL-4 plays an important role in regulation of T cells functions during infection and autoimmune diseases (7). Recently, it was suggested that macrophages do not form stable populations, but rather have distinct phenotypes in response to various inflammatory stimuli (e.g., IFNγ vs. IL-4) and often form mixed phenotypes (7, 8), which has unpredictable impact on functions of T cells at the site of inflammation where macrophages serve as antigen-presenting cells. In normal conditions, the CNS has specific microenvironment where CNS-resident macrophages (also referred to as microglia) have intrinsic M2-like phenotype and express number of M2 markers (e.g., Ym1 and IL-4) and specific microRNAs (miRs) (e.g., miR-124) that promote M2 polarization (9–11). Moreover, CNS has internal source of IL-4, which plays critical role in suppression of neuroinflammation such as experimental autoimmune encephalitis (EAE) (9). M2 macrophages in normal CNS express low level of MHC class II and CD86 and do not effectively stimulate T cells.

Similar to human disease multiple sclerosis (MS), EAE is an inflammatory disease of the CNS that is initiated by autoimmune Th1 and Th17 cells that recognize self-antigen within myelin sheath [e.g., myelin oligodendrocyte glycoprotein (MOG)] (12). Th1 and Th17 cells produce number of cytokines such as IFNγ, TNF, and GM-CSF that mediate M1 polarization of macrophages that comprise ~70% of CNS inflammatory lesions and mediate most of the damage of neuronal tissue by producing TNF, nitric oxide (NO), etc. (7, 12–15). Since CNS has intrinsic M2-skewing microenvironment, macrophages in the CNS during EAE and other pathologies often exhibit mixed M1/M2 phenotype under the influence of CNS-derived IL-4 and Th1-derived IFNγ (16). We have previously found that during EAE macrophages exhibit dually activated phenotype expressing both M1 and M2 markers such as CD86 and Ym1 at the same time (9). These results were not surprising, since during EAE both M1 (IFNγ) and M2 (IL-4) cytokines were present in inflamed CNS. Therefore, it is important to find new pathways and co-factors that could skew macrophage polarization toward M2 in mixed inflammatory conditions during EAE.

One of the co-factors that help to skew macrophages toward M2 phenotype in the CNS is miR-124 (10). miRs are class of short (20–22 nucleotides) non-coding RNAs that regulate many physiological processes including macrophages polarization (11). We have previously shown that miR-124 is expressed in microglia in normal CNS contributing to M2 phenotype of microglial cells in resting state (10). During EAE, expression of miR-124 in activated microglia was decreased, while transfection of macrophages with miR-124 upregulated M2 markers such as arginase (Arg)1 and Mannose receptor C-type 1 (Mrc1) and downregulated M1 markers such as CD86, nitric oxide synthetase (NOS)2, and TNF leading to recovery from the disease (10). We also found that miR-124 is upregulated in bone marrow (BM)-derived and peritoneal macrophages in response to IL-4, but effect of IL-4 on upregulation of miR-124 was modest (17). Moreover, expression of miR-124 in microglia in normal CNS was IL-4/IL-13 independent, while IFNγ pretreated M1 macrophages failed to upregulate miR-124 in response to subsequent treatment with

**Abbreviations:** Arg, arginase; ATF, activating transcription factor; BM, bone marrow; CNS, central nervous system; CRE, cAMP response elements; EAE, experimental autoimmune encephalomyelitis; LFB, luxol fast blue; MOG, myelin oligodendrocyte glycoprotein; Mrc, mannose receptor; miR, microRNA; MS, multiple sclerosis; NOS, nitric oxide synthetase; REST, RE1-silencing transcription factor.

IL-4 (17). These data suggest the importance of other pathways that may stimulate miR-124 and change balance toward M2 in the CNS during inflammation.

Forskolin (coleonol) originate from Indian plant Coleus and currently used as food supplement or as a drug for traditional oriental medicine and anticancer therapy (18). We found that administration of Forskolin during EAE inhibited proliferation and IFN-γ production by CD4 T cells in the CNS and upregulated a number of M2-associated molecules miR-124, Arg1, Mrc1, Ym1, and Fizz1 and downmodulated M1-associated molecules MHC class II, CD86, and NOS2. This indicates very potent effect of this drug in the modulation of autoimmune CD4 T cells and macrophage activation/polarization at the site of tissue inflammation. To understand mechanisms, we were looking for pathways that inhibit pathogenic CD4 T cells, enhance action of IL-4 in M2 polarization and upregulation of miR-124 during CNS autoimmune inflammation. We found that Forskolin did not directly affect CD4 T cells, but promoted M2 polarization of macrophages. Further analysis demonstrated that Forskolin and/or IL-4 activated cAMP-responsive element-binding/activating transcription factor (CREB/ATF) family proteins that contributed to upregulation of miR-124 and M2 polarization.

### MATERIALS AND METHODS

### Mice

C57BL/6 mice were originally purchased from the Jackson Laboratory and bred locally at the Laboratory Animal Services Center at the Chinese University of Hong Kong. All animal procedures were conducted under individual licenses from Hong Kong government and approved by animal ethics committee form the Chinese University of Hong Kong.

### Cells

Bone marrow-derived macrophages were grown in DMEM media (Gibco) supplemented with 10% FBS (Gibco) and M-CSF (R&D; 10 ng/ml) for 5 days as described (10) and used for analysis. Peritoneal macrophages were isolated by peritoneal lavage and removal of non-adhered cells after 14–16 h of incubation at CO2 incubator (9). Flow cytometry analysis indicated 96–98% of CD11b<sup>+</sup>F4/80<sup>+</sup> cells in case of BM-derived and peritoneal macrophages. For macrophage polarization, the cells were treated with IL-4 (50 ng/ml), IFNγ (100 ng/ml), and/or Forskolin (30 µM; Sigma) for indicated periods ranging from 2 to 24 h. The optimal dosage of Forskolin was determined after performing dose–response curve (1–100 µM) for the upregulation of miR-124. ERK1/2 inhibitor U0126 was purchased from Cell Signaling, and another ERK1/2 PD184352 inhibitor was purchased from Abcam. Inhibitors U0126 and PD184352 were used at concentrations of 10 µM and 200 nM, respectively, according to the manufacturer's recommendations. Taurolidin (Sigma) was used at the dose of 100 µM. X5050 inhibitor was used at final concentration of 100 µM. NRSE inhibitory dsRNA was used at concentration of 20 nM as described (19). Primary cultures of mouse cortical neurons were performed as described earlier (10).

### EAE Induction

EAE was induced by subcutaneous immunization with 150 µg MOG35–55 (American peptide) in 4 mg/ml CFA (Diffco) as described earlier (10). Pertussis toxin (Sigma) was injected i.p. (150 ng/mouse) on day 0 and day 2 post-immunization. Individual animals were assessed daily for symptoms of EAE and scored using a scale from 0 to 5 as follows: 0, no disease; 0.5, weak tail or mild hind limb ataxia; 1, limp tail and/or hind limb ataxia; 2, hind limb paresis; 3, hind limb paralysis; 4, hind and fore limb paralysis; 5, death. Forskolin (3 mg/kg) or vehicle (PBS with 1% DMSO) was injected i.p. daily starting from disease onset (day 14) till the peak of disease (days 21–22). The optimal dosage of Forskolin for *in vivo* administration was selected by titrating down reported dosage of 4–5 mg/kg in previously published study for anticancer therapy in mouse model (20).

## Real-time PCR

For quantitation of various markers and molecules by realtime RT PCR, total RNA was isolated by Qiagene or MirVana (Applied Biosystem) kits from cultured BM-derived, or peritoneal macrophages, or form the spinal cords of PBS-perfused unmanipulated mice or mice on day 22 following EAE. For analysis of miRNA expression, real-time RT-PCR analyses were performed using TaqMan miRNA assays for miR-124 (Applied Biosystems) and relative expressions were calculated using the Δ*C*T method and normalized to uniformly expressed snoRNA55 (Applied Biosystems) as we described earlier (10, 21). For analysis of mRNA expression for NOS2 (forward primer, 5′-ACC CACATCTGGCAGAATGAG-3′; reverse primer, 5′-AGCCATG ACCTTTCGCATTAG-3′), Arg1 (forward primer, 5′-CTTGGCT TGCTTCGGAACTC-3′; reverse primer, 5′-GGAGAAGGCGTT TGCTTAGTTC-3′), and Fizz1 (forward primer, 5′-GCCAGGTC CTGGAACCTTTC-3′; reverse primer, 5′-GGAGCAGGGAGAT GCAGATGAG-3′), ATF3 forward 5′-AGCATTCACACTCTCC AGTTTCTCT-3′, reverse 5′-GGAGAAGACAGAGTGCCTGC-3′, IL-1β forward 5′-CTTCCAGGATGAGGACATGAGCAC-3′, reverse 5′-TCATCATCCCATGAGTCACAGAGG-3′, IL-6 forward 5′-CCTTCTTGGGACTGATGCTGGTG-3′, TNF forward 5′-AG CCGATGGGTTGTACCTTG-3′ reverse 5′-GTGGGTGAGGA GCACGTAGTC-3′, Mrc1 forward 5′-ACCACGGATGACCTGT GCTC-3′ reverse 5′-TGGTTCCACACCAGAGCCATC-3′, CCDN1 forward 5′-GCAGACCATCCGCAAGCATG-3′, reverse 5′-TGG AGGGTGGGTTGGAAATGAAC-3′, RE1-silencing transcription factor (REST) (for alternative splicing) forward 5′-CGACACATG CGGACTCATTC-3′, reverse 5′-GCATGTCGGGTCACTTCA TG-3′; Ym1 forward 5′-CCATTGGAGGATGGAAGTTTG-3′, reverse 5′-GACCCAGGGTACTGCCAGTC-3′ the relative expressions were calculated using the Δ*C*T method and normalized to the GADPH housekeeping gene (forward primer, 5′-ATGAC CACAGTCCATGCCATC-3′; reverse primer, 5′-GAGCTTCCC GTTCAGCTCTG-3′), and then relative level of expression was calculated in comparison to control conditions as we did earlier (10).

### Western Blotting

Western blotting analysis was performed according to standard protocol as we reported previously (10). Antibodies for ERK, phospho-ERK, CEBP, phospho-CEBP, CREB, phospho-CREB, REST, and β-actin were purchased from Cell Signaling.

### Flow Cytometry

Mononuclear cells were isolated from the CNS or spleens of mice with EAE on day 22 using Percoll gradients as described in our previous studies (13, 21). For analysis of microglia, macrophages, and lymphocyte populations and macrophage/microglia activation status, the cells were stained with anti-CD11b-AF480, anti-CD86-PE, anti-MHC class II-PE-Cy5 (all from BD Bioscience), and F4/80-APC (eBiosceince) or anti-CD45-APC-Cy7 (BioLegend). FcRs were blocked with a mAb specific for mouse FcR (2.4G2; BD Biosciences). CD4 T cells were stained for analysis of surface markers with anti-CD4-APC, or anti-CD4-APC and anti-TCRβ-PE, or anti-CD4-APC and anti-TCR-Vβ11-PE (all from BD Biosciences). These data were acquired on LSR Fortessa cytometer (BD Biosciences) and analyzed using FlowJo software (TreeStar Inc.). Absolute numbers of the cells were calculated by counting total number of mononuclear cells using hemocytometer and by multiplying by the percentages of particular populations obtained by flow cytometry.

# T-Cell and Macrophage Proliferation Assay

BrdU incorporation assay was used to assess proliferation of BM-derived macrophages or CD4 T cells *in vitro* or *in vivo* similar as was described in our earlier studies (12). To perform this analysis, we used BrdU kit from BD Biosciences following the manufacturer's instructions. BrdU was injected intraperitoneally (2 mg/ mouse) or added to cultures at final concentration 10 µM 14 h before analysis. Macrophages were stained for cell surface markers CD11b, F4/80, and for intracellular BrdU-FITC (BD Biosciences) and CD11b<sup>+</sup>F4/80<sup>+</sup> gated analyzed for BrdU incorporation by three-color flow cytometry. CD4 T cells were stained for cell surface markers CD4 alone, or CD4 and TCR-Vβ11, and intracellular BrdU and analyzed by two- or three-color flow cytometry.

# Cytokine Production by CD4 T Cells

For intracellular detection of IFNγ expression, *ex vivo* isolated or cultured CNS or splenic mononuclear cells were activated with phorbol myristate acetate (50 ng/ml) and ionomycin (1 µg/ ml; both from Sigma) in the presence of GolgiStop (1 µl/ml, BD Biosciences) for 4 h. Cells were washed, immediately stained for surface marker anti-CD4-APC, fixed/permeabilized using BD intracellular cytokine detection kit, and further stained for intracellular IFNγ using anti-IFNγ-PE (all reagents were purchased from BD Biosciences). CD4+ gated cells were analyzed by twocolor flow cytometry.

# T-Cell Recall Response

For CD4 T cell recall response, mice were immunized with 150 µg MOG35–55 (American peptide) in 4 mg/ml CFA (Diffco) as described earlier (10, 22). On day 7 post-immunization, CD4 T cells were isolated from draining lymph nodes and spleen using negative selection and magnetic beads and incubated in the presence of irradiated splenocytes, MOG35–55 peptide (5–10 µg/ ml) or stimulating anti-CD3 (BD Biosciences, clone 145-2C11) and anti-CD28 (BD Biosciences, clone 37.51) mAbs as described earlier (12). The experiments were performed in the presence of 30 µM of Forskolin or 0.1% DMSO in media (Control). After 72 h of incubation, the dead cells were removed using Ficoll™ gradient centrifugation, washed and stained for cells surface markers CD4 and TCRVβ11, and were analyzed for BrdU incorporation by three-color flow cytometry.

### Immunohistochemistry

Fixed frozen section of lumbar areas of spinal cord were prepared and stained with luxol fast blue (LFB, Sigma) as we described earlier (10).

# Statistical Analysis

The results are presented as mean ± SE. Unpaired Student's *t*-test was used to determine significance between two independent groups. *p* Values of less than 0.05 were considered to be significant. SigmaPlot program was used for the creation of the graphs and performing statistical analysis.

# RESULTS

### Forskolin Downregulate MHC Class II and CD86 on Microglia and Macrophages in the CNS during EAE Resulting in Decrease of Neuroinflammation and Recovery from the Disease

Based on literature, we hypothesized that administration of Forskolin would lead to activation of cAMP pathway in various cell types that might be beneficial for suppression of EAE by inhibiting autoimmune CD4 T cells and/or changing balance of macrophage polarization from M1 to M2. We started administration of Forskolin on the day of onset of the disease, when encephalitogenic IFNγ and IL-17 producing Th1 and Th17 cells were primed and already migrated into CNS leading to appearance of fist clinical symptoms (**Figure 1A**, day 14). Administration of Forskolin substantially ameliorated the disease leading to recovery, while control vehicle-treated group had peak of the disease on days 21–22 (**Figure 1A**). We found that mice treated with Forskolin had very low level of demyelination in the spinal cord, as determined by LFB staining (**Figure 1B**). Forskolintreated mice had substantially lower level of percentages and absolute numbers of CD11b<sup>+</sup>CD45hi macrophages/activated microglia and CD11b<sup>−</sup>CD45hi lymphocytes in the CNS on day 22 (**Figures 1C,D**). Forskolin-treated mice had also diminished level of expression of MHC class II and CD86 on CD11b<sup>+</sup>F4/80<sup>+</sup> microglia/macrophages in the CNS (**Figures 1E,G**) but not on CD11b<sup>+</sup>F4/80<sup>+</sup> macrophages in the spleen (**Figures 1F,G**). Thus, we found that Forskolin was effective in downmodulation of CNS autoimmune inflammation leading to decrease in expression of activation marker MHC class II and M1 marker CD86.

### Forskolin Upregulated Arg1 and Inhibited NOS2 in the CNS during EAE by Changing Balance toward M2

We hypothesized that downregulation of general activation marker MHC class II and M1 marker CD86 on microglia/macrophages in the CNS of Forskolin-treated mice with EAE was due to changing balance toward M2. To further verify this, we investigated the expression of additional numbers of M1 (NOS2, TNF), M2 markers (miR-124, Arg1, Mrc1, Ym-1, Fizz1), and general macrophage activation markers (IL-1β, IL-6) in the CNS of control group vs. Forskolin-treated group of mice with EAE on day 22. We found that miR-124, *Arg1*, *Mrc1*, *Ym1*, and *Fizz1* were upregulated in Forskolin-treated mice (**Figures 2A,B,D–F**), while *NOS2* was

Figure 1 | Analysis of the role of Forskolin in the modulation of neuroinflammation and expression of MHC class II and CD86 in the central nervous system (CNS) and periphery. EAE was induced and vehicle (PBS with 1% DMSO) or Forskolin were administrated i.p. on days 14, 15, 16, 17, 19, 20, and 21 postimmunization as described in Section "Materials and Methods." On day 22, mice were perfused with cold PBS and brains, spinal cords, and spleen were isolated for histology (B) and/or flow cytometry analysis (C–H). (A) EAE clinical course (mean ± SE) of total 11–12 individual mice for each group is shown. Injections of vehicle or Forskolin are indicated by arrows. (B) Histology analysis of extent of CNS myelination in coronal sections of lumbar area of the spinal cords of unmanipulated mice (top image) or mice with EAE on day 22 (bottom images) treated with vehicle (left image) or Forskolin (right image). Extent of myelination was assessed in lumbar area of the spinal cord using luxol fast blue (LFB) dye as described in Section "Materials and Methods" (bar: 50 μm; magnification: 400×). Quantitative analysis of percentage of myelin-free LFB-negative areas is shown on bar graph on the right. Mean ± SE of six separate images from three individual mice is shown (\*\*\*\**p* < 0.0001). (C,D) The flow cytometry analysis of the CNS mononuclear cells isolated from the mice with EAE on day 22 treated with vehicle or Forskolin. The mononuclear cells were isolated from the CNS of mice with EAE, stained for CD11b and CD45, and analyzed by flow cytometry as described in Section "Materials and Methods." The percentages of populations of CD11b+CD45low microglia (left gates), CD11b+CD45hi macrophages (upper right gates), and CD11b−CD45hi lymphocytes (lower right gates) are shown. The quantification of the absolute number of macrophages and lymphocytes in the CNS is shown in panel (D). (E–H) The flow cytometry analysis of expression of M1-associated markers in macrophages in the CNS (E,G) or spleens (F,H) of mice with EAE on day 22 treated with Vehicle or Forskolin. Mononuclear cells from CNS (E,G) or spleens (F,H) were stained for CD11b, F4/80, MHC class II, and CD86 and CD11b+F4/80+ cells were analyzed for the expression of MHC class II and CD86 by four-color cytometry as described in Section "Materials and Methods." Representative histograms for MHC class II (upper row) and CD86 (bottom row) are shown in panels (E,F). A solid line indicates staining for MHC class II or CD86 and a dotted line indicates staining for isotype-matched control. A bar indicates percentage of MHC class II or CD86-positive cells. Mean fluorescence intensity (MFI) of expression of MHC class II or CD86 is shown at the bottom of each histogram. Quantitative analyses of MHC class II and CD86 expression in macrophages in the CNS and spleens are shown in panels (G,H), respectively. In panels (D,G,H), mean ± SE of five individual animals is shown (\**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; NS, not significant).

Figure 2 | The effect of Forskolin on skewing balance toward M2 phenotype in the central nervous system during EAE. EAE was induced, and vehicle or Forskolin was administrated i.p. on days 14, 15, 16, 17, 19, 20, and 21 post-immunization similarly as for Figure 1. On day 22, mice were perfused with cold PBS, and the spinal cords were isolated for RNA isolation and analysis. Expression of M2 markers [microRNA (miR)-124 (A), *Arg1* (B), *Mrc1* (D), *Ym1* (E), and *Fizz1* (F)], M1 markers [*NOS2* (C), *TNF* (I)], and general pro-inflammatory markers [IL-1β (G), IL-6 (H)] was performed by real-time RT-PCR as described in Section "Materials and Methods." In panels (A–I), mean ± SE of triplicate is shown (\**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; NS, not significant).

downregulated (**Figure 2C**). We did not find statistically significant difference for *IL-1β*, *IL6*, and *TNF* (**Figures 2G–I**). However, *IL-1β* and *TNF* had trend to be elevated (**Figures 2G,I**) and *IL-6* had a trend to be decreased (**Figure 2H**). Notably, we observed the dramatic 19-fold upregulation for *Arg1* (**Figure 2B**) and almost undetectable level of *NOS2* (**Figure 2C**) in the spinal cord of Forskolin-treated mice with EAE. Dramatic upregulation of *Arg1* was in excellent agreement with our *in vitro* data (**Figure 2B**). Thus, we found that Forskolin skewed the balance toward M2 in the CNS during EAE leading to recovery form disease.

# Forskolin Decreased Proliferation and IFN**γ** Production by Autoimmune CD4 T Cells in the CNS, but Not in the Periphery

Since it was reported that Forskolin could decrease proliferation of T cells, we checked the level of proliferation of CD4 T cells in the CNS and periphery (spleen) during EAE on day 22. We found that that proliferation of CD4 T cells was decreased twofold from 20 ± 3 to 9 ± 2% in the CNS of Forskolin-treated mice (**Figures 3A,E**). The level of production of IFNγ by CD4 T cells was also decreased in the CNS 1.5-fold, indicating decreased level of activation and differentiation of pathogenic Th1 cells (**Figures 3B,F**). This was consisted with 6.3-fold decrease in absolute number of infiltrating CD4 T cells in CNS of mice with EAE on day 22 (**Figure 3G**). At the same time, Forskolin did not influence the level of proliferation of CD4 T cells in the spleen (**Figures 3C,H**). It was even trend for increase in IFNγ production in the spleen of Forskolin-treated mice with EAE on day 22; however, it was not statistically significant (**Figures 3D,I**). We also found that during peak of EAE (day 22) mRNA levels for both Th1 and Th17 cytokines IFNγ and IL-17A were decreased ~8-fold in the spinal cords of Forskolin-treated mice when compared with vehicle-treated group indicating decrease in activity of both Th1 and Th17 cells (data not shown). Thus, these data demonstrate that administration of Forskolin decreased proliferation and proinflammatory cytokine production by pathogenic CD4 T cells in the CNS at the site of inflammation, but not in the periphery.

## Forskolin Did Not Have Direct Effect on Proliferation of MOG-Specific Autoimmune CD4 T Cells

Although we found dramatic effect of Forskolin on EAE disease course, macrophage polarization, and proliferation of CD4 T cells in the CNS, effect of Forskolin on MOG-specific CD4 T cells could be secondary due to skewing macrophages toward M2, which poorly stimulate proliferation and differentiation of Th1 cells. It was reported that Forskolin directly inhibit proliferation of naïve T cells and T cell lines stimulated with anti-CD3, but the role of Forskolin on proliferation of effector T cells remained controversial (23). In our experiments, we started treatment after disease onset, when encephalitogenic T cells where already primed and become effector CD4 T cells. These effector CD4 T cells were not inhibited by Forskolin in the spleen and had level of proliferation and IFNγ production similar to control group. This may indicated that Forskolin had no direct effect on autoimmune T cells in the periphery. To address this, we further investigated whether Forskolin directly affect proliferation of primed MOGspecific CD4 T cells *in vitro* and found that Forskolin had very slight trend to decrease proliferation of CD4 T cells form 17 ± 1 to 15 ± 1% of BrdU-positive cells; however, this difference was statistically insignificant (**Figures 4A,B**; upper figures). When we compared extent of proliferation of MOG-specific enriched population of CD4<sup>+</sup>Vβ11<sup>+</sup> T cells, the level of proliferation was also very similar for control or Forskolin-treated cells comprising 70–80% of CD4<sup>+</sup>Vβ11<sup>+</sup>BrdU<sup>+</sup> cells (**Figures 4A,B**; bottom figures). Interestingly when we used anti-CD3/CD28 mAbs to elicit polyclonal stimulation in the same experiment instead of stimulation with MOG antigen, we found that Forskolin induced statistically significant 40% decrease in CD4 T cell proliferation (**Figures 4C,D**), as was reported earlier (2, 23). These data demonstrated that in contrast to polyclonally stimulated CD4 T cells, primed MOG-specific CD4 T cells were not susceptible to direct action of Forskolin. Thus, we demonstrate that Forskolin did not have significant direct effect on proliferation of autoimmune MOG-specific CD4 T cells in our model.

# Forskolin Synergize IL-4 in Upregulation of miR-124 and Arg1 Even in the Presence of IFN**γ**

Since it was recently demonstrated that cAMP contributed to M2 polarization (24, 25), we hypothesized that Forskolin could enhance IL-4 in M2 polarization. We analyzed the expression of M2-associated molecules miR-124, Arg1, Mrc1, Ym1, Fizz1, and M1-associated NOS2 in the presence of Forskolin in M0 (Control), M2 (IL-4), M1 (IFNγ), and mixed M1/M2 (IL-4 and IFNγ) conditions. Experiments demonstrated that combination of Forskolin with IL-4 resulted in 3- to 5-fold upregulation of miR-124 with average 3.3-fold (**Figure 5A**, IL4 FSK). However, more strikingly we found 3.4-fold upregulation of miR-124 when Forskolin was used together with IL-4 and IFNγ (**Figure 5A**, IL-4 IFN FSK). At the same time, IL-4 and IFNγ failed to upregulate miR-124. This indicates that Forskolin overcome inhibitory action of IFNγ in IL-4-induced miR-124 expression, when IL-4 was used together with IFNγ.

In addition to miR-124, Forskolin substantially enhanced IL-4 in induction of *Arg1* in M2 macrophages (**Figure 5B**). Forskolin synergized expression of *Arg1* by 18-fold when compared with the action of IL-4 alone (**Figure 5B**; IL4 and IL4 FSK). Forskolin upregulated *Arg1* by fivefold in mixed inflammatory conditions in the presence of IL-4 and IFNγ (**Figure 5B**; IFN IL4 and IFN IL4 FSK). This drug also enhanced expression of M2 marker *Mrc1*; however, it failed to induce it in mixed inflammatory conditions (**Figure 5C**). Finally, Forskolin did not have substantial effect on IL-4-induced expression of two other tested M2 markers *Ym1* and *Fizz1* (data not shown).

When we compared expression of M1 markers in IL-4- and IFNγ-activated macrophages, we found that Forskolin induced low level of *NOS2* expression by itself and enhanced induction of *NOS2* by IFNγ (**Figure 5D**). Although *Arg1* and *NOS2* were shown to be antagonistic markers, Forskolin induced both in mixed inflammatory conditions, but the level of *Arg1* was more than 10-fold higher than *NOS2* (**Figures 5B,D**, IFN IL4 FSK).

Figure 3 | The effect of Forskolin on proliferation and IFNγ production by autoimmune CD4 T cells in the central nervous system (CNS) and periphery during EAE. EAE was induced, and vehicle or Forskolin was administrated i.p. on days 14, 15, 16, 17, 19, 20, and 21 post-immunization similarly as for Figure 1. On day 21, BrdU was injected i.p. as described in Section "Materials and Methods." On day 22, mononuclear cells were isolated from CNS or periphery (spleen) as for Figure 1 and CD4 T cells were analyzed for the expression of surface markers, BrdU incorporation and intracellular IFNγ expression as described in Section "Materials and Methods." (A,C,E,H) To measure CD4 T cell proliferation, BrdU was injected i.p. 14 h before analysis and mononuclear cells were isolated from CNS (A,E) or spleen (C,H). The cells were stained for CD4 (*x*-axis) and BrdU (*y*-axis) and analyzed with two-color flow cytometry. BrdU incorporation by CD4+ gated cells is shown on representative contour-plots (A,C), and statistics is shown in panels (E,H). Percentages of BrdU-positive CD4+ gated cells are shown on the upper left quadrant of each contour-plot. Negative controls for BrdU staining are shown at left bottom quadrant of each contour-plot. (B,D,F,I) To measure production of IFNγ by CD4 T cells, mononuclear cells were isolated from CNS (B,F) or spleen (D,I) and stimulated with PMA/ionomycin in the presence of *GolgiStop* for 4 h as described in Section "Materials and Methods." Then, the cells were stained for CD4 (*x*-axis) and IFNγ (*y*-axis) and analyzed with two-color flow cytometry. IFNγ production by CD4+ gated cells is shown on representative contour-plots (B,D), and statistics is shown in panels (F,I). Percentages of IFNγ-positive CD4+ gated cells are shown on the upper left quadrant of each contour-plot. Negative controls for IFNγ staining are shown at left bottom quadrant of each contour-plot. (G) Mononuclear cells were stained with CD4 and TCRβ and absolute number of CD4+TCRβ+ cells was quantified as for Figure 8. In panels (E–I), mean ± SE of three to five individual mice is shown (\**p* < 0.05; \*\**p* < 0.01; NS, not significant).

cytometry. Total CD4+ gated cells are shown on upper contour-plots, while CD4+TCRVβ11+ gated cells are shown in bottom contour-plots in panel (A). Percentages of BrdU-positive CD4+ gated (top contour-plots) or CD4+TCRVβ11+ gated (bottom contour-plots) CD4 T cells are shown on the upper left quadrant of each contour-plot. Negative controls for BrdU staining are shown at left bottom quadrant of each contour-plot. Mean ± SE of four culture wells is shown in panel (B) (NS, not significant). (C,D) The influence of Forskolin on the proliferation of polyclonally stimulated CD4 T cells with anti-CD3/CD28. Percentages of BrdU-positive CD4<sup>+</sup> T cells are show on the upper left quadrant of each contour-plot (C). Negative controls for BrdU staining are shown in a left bottom quadrant of each contour-plot. Mean ± SE of four culture wells is shown in panel (B) (\*\*\**p* < 0.001).

Finally, Forskolin did not affect the expression of other M1 markers CD86 and *TNF* (data not shown).

Taken together, these data indicate that Forskolin synergize action of IL-4 in induction of miR-124 and *Arg1* in the M2-skewing condition in the presence of IL-4 and mixed inflammatory conditions (IL-4 with IFNγ) changing balance toward M2.

# Forskolin Activate ERK and Synergize Activation of ERK Pathway by IL-4

We further investigated mechanism by which Forskolin induce expression of M2 markers. It was reported that CEBPβ could become phosphorylated by ERK1/2 in response to IFNγ in mouse macrophage cell line that play an important role in IFNγinducible gene expression such as *NOS2* (26, 27). On the other hand, CREB–CEBPβ axis was shown to become activated in LPStreated macrophages resulting in expression of number M2 genes including *Arg1* and *Mrc1* while M1 genes remained unaffected (28). Thus, the roles of ERK–CEBPβ and CREB–CEBPβ axes in M1 vs. M2 remained unclear. We tested whether IL-4 and/or Forskolin affect activation CREB, ERK, and CEBPβ by inducing their upregulation and/or phosphorylation.

First, we investigated early kinetics of expression level and extent of phosphorylation of CREB, ERK, and CEBPβ after 10, 30, and 60 min of incubation with Forskolin. We found that CREB upregulation (**Figure 6A**; Figure S1 in Supplementary Material) and phosphorylation (**Figure 6B**; Figure S1 in Supplementary Material) was detected 10–60 min after incubation with Forskolin. We found very modest upregulation of CEBPβ (**Figure 6C**; Figure S1 in Supplementary Material) and only slight increase in CEBPβ phosphorylation (**Figure 6D**; Figure S1 in Supplementary Material) after 30–60 min of incubation with Forskolin. Forskolin induced subtle downregulation (**Figure 6E**; Figure S1 in Supplementary Material) and substantial increase

Forskolin/IL-4/IFNγ. After 24 h, the expressions of miR-124 (A), *Arg1* (B), *Mrc1* (C), and *NOS2* (D) were analyzed by real-time RT-PCR as described in Section "Materials and Methods." Mean ± SE of four to eight culture wells is shown (\**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001).

in the level of phosphorylation of ERK1/2, which was detectable after 10–60 min of incubation with Forskolin (**Figure 6F**; Figure S1 in Supplementary Material).

Second, we investigated long-term effect of Forskolin and/or M1 (IFNγ) and M2 (IL-4) activating stimuli on expression and extent of phosphorylation of CREB, ERK, and CEBPβ after 24 h of incubation with Forskolin. We found that CREB was upregulated by IFNγ and IL-4 (**Figure 7A**; Figure S1 in Supplementary Material), while the highest level of CREB phosphorylation was observed in macrophages treated with IL-4 together with Forskolin (**Figure 7B**, *IL4/F*; Figure S1 in Supplementary Material), which was blocked by ERK inhibitor U0126 (**Figure 7B**, *IL4/F/U*; Figure S1 in Supplementary Material). Interestingly, we found that IL-4 caused the highest level of CREB expression at 24 h time-point indicating important role of cAMP in conventional IL-4-polarized macrophages (**Figure 7A**). Most of activating stimuli (IFNγ alone, IL-4 alone, or IL-4 with Forskolin, or IFNγ with Forskolin) upregulated CEBPβ with little difference in extent of phosphorylation of CEBPβ (**Figures 7C,D**; Figure S1 in Supplementary Material). ERK inhibitor did not downregulate expression of CEBPβ (**Figure 7C**). ERK expression was not influenced by most of activating stimuli while IFN/Forskolin and IL-4/Forskolin even decreased it (**Figure 7E**; Figure S1 in Supplementary Material). However, phosphorylation of ERK1/2 was slightly enhanced by IL-4 or Forskolin, but the highest level of ERK phosphorylation was observed when IL-4 was added together with Forskolin (**Figure 7E**; Figure S1 in Supplementary Material). In contrast, IFNγ alone did cause substantial phosphorylation of ERK, but Forskolin with IFNγ substantially enhanced ERK1/2 phosphorylation. The ERK inhibitor U0126 blocked phosphorylation of ERK1/2 caused by IL-4 together with Forskolin, or IFNγ with Forskolin (**Figure 7E**). Thus, we concluded that Forskolin enhanced action of IL-4 in phosphorylation of ERK1/2 and CREB, but ERK pathway did not play a substantial role in CEBPβ phosphorylation.

## ERK Inhibitors Block Upregulation of Arg1, Mrc1, and miR-124 Induced by Forskolin with IL-4

We found that Forskolin activated ERK1/2 pathway and stimulated expression of M2 markers. We hypothesized that Forskolin induced expression of M2 markers and miR-124 *via* activation ERK1/2. However, currently, it is not clear whether ERK1/2 promote M2 polarization or not. It was reported that ERK1/2 activation contributed to both M1 and M2 phenotypes (26, 29–32). The role of ERK was shown to be important for Th2-mediated allergic lung inflammation (33), while the role of miR-124 was also shown in our studies to be important for EAE and allergic lung inflammation (10, 17). Therefore, we further investigated the role of ERK pathway in upregulation of miR-124, *Arg1*, *Mrc1*, and *NOS2* induced by IL-4 together with Forskolin. We found that in the presence of ERK inhibitor U0126 the expression of miR-124 fell below control level (**Table 1**, miR-124). In addition to U0126, we used another ERK inhibitor P184352 to confirm inhibitory effect of U0126. Both ERK inhibitors U0126 and P184352 inhibited Arg1 expression by 58 and 98%, respectively (**Table 1**, Arg1). Both inhibitors U0126 and P184352 increased expression of NOS2 by fourfold and sevenfold, respectively (**Table 1**, NOS2). At the same time, U0126 decreased expression of *Mrc1* by 50%, upregulated expression of *Ym1* by 25%, and had no effect on *Fizz1* expression (**Table 1**). Thus, we found that ERK pathway substantially contributed to upregulation of miR-124 and *Arg1* and downregulation of *NOS2* caused by the simultaneous action of IL-4 and

β-Actin was used as a loading control. (A,C,E) Quantitative analysis of relative expression levels of CREB, CEBPβ, and ERK normalized to β-actin is shown. Representative image is shown in Figure S1 in Supplementary Material. (B,D,F) Quantitative analysis of relative expression levels of p-CREB, p-CEBPβ, and p-ERK normalized to total CREB, CEBPβ, and ERK1/2, respectively, is shown. Representative image is shown in Figure S1 in Supplementary Material. In panels (A–D), mean ± SE of three separate experiments is shown.

Forskolin. At the same time, we found that Forskolin enhanced IFNγ to induce expression of *NOS2* (**Figure 5D**, NOS2). We investigated whether ERK pathway contributed to this process as it was described earlier for RAW 264.7 cell line (26). We did find that ERK pathway partially contributed to IFNγ-inducible *NOS2* expression since U0126 decreased expression of *NOS2* by 21% (**Table 2**, NOS2). However, inhibition of ERK pathway resulted in twofold increase in *TNF* expression (**Table 2**), indicating that ERK partially contributed to IFNγ-inducible NOS but not other M1 marker TNF. Thus, ERK pathway contributed to upregulation of M2 markers miR-124 and *Arg1* (**Table 1**) and partially *NOS2* and downmodulation of *TNF* (**Table 2**). These data demonstrated

Figure 7 | Effect of Forskolin on activation of downstream signaling pathways CREB, CEBPβ, and ERK1/2 in M0 macrophages (Control), or during M2 (IL-4) vs. M1 (IFNγ) polarizing conditions. Bone marrow-derived macrophages were grown for 5 days in the presence of M-CSF and were treated with Forskolin, or IL4, or IFNγ, or Forskolin/IL4, or Forskolin/IFNγ, or Forskolin/IL4/IFNγ for 24 h as in Figure 1. ERK1/2 inhibitor U0126 was used to block ERK1 and ERK2 phosphorylation (see Materials and Methods). Expressions of CREB, CEBPβ, and ERK1/2 and their phosphorylated forms (p-CREB, p-CEBPβ, and p-ERK1/2) were analyzed by Western blot as described in Section "Materials and Methods." β-Actin was used as a loading control. (A,C,E) Quantitative analysis of relative expression levels of CREB, CEBPβ, and ERK normalized to β-actin is shown. Representative image is shown in Figure S1 in Supplementary Material. (B,D,F) Quantitative analysis of relative expression levels of p-CREB, p-CEBPβ, and p-ERK normalized to total CREB, CEBPβ, and ERK1/2, respectively, is shown. Representative image is shown in Figure S1 in Supplementary Material. In panels (A–D), mean ± SE of three separate experiments is shown. Abbreviations: IFN, IFNγ; IL4, IL-4; F, Forskolin; U, U0126.

importance of ERK pathway in the expression of M2 markers miR-124 and Arg1.

## ERK Inhibitor Block Upregulation of Arg1, Mrc1, and Cyclin D2 and Upregulate NOS2 in IL-4 Activated BM-Derived Macrophages

Since ERK inhibitors inhibited IL-4/Forskolin-induced expression of miR-124 and *Arg1* (**Table 1**), we hypothesized that ERK pathway is important for upregulation of M2 markers in conventional M2 macrophages polarized with IL-4 without Forskolin. Our hypothesis was also based on the finding that IL-4 alone but not IFNγ alone caused ERK1/2 phosphorylation (**Figure 7**). It was suggested that ERK contribute to M2 phenotype for tumorinfiltrating macrophages (TIMs) (29, 30), but the role of ERK has not be proven to be essential for IL-4-activated macrophages as a general concept. We further compared expression of several M2 markers in IL-4-polarized BM-derived macrophages with and without U0126 inhibitor. We found that ERK pathway was critical

Table 1 | Effect of ERK inhibitors U0126 and PD184352 on IL-4/Forskolininduced expression of microRNA (miR)-124, *Arg1*, *NOS2*, *Mrc1*, *Ym1*, and *Fizz1*. a


*a Bone marrow (BM)-derived macrophages were grown for 5 days in the presence of M-CSF and were analyzed as untreated (control) or treated with IL-4 and Forskolin without or with ERK inhibitors U0126 or PD184352 for 24 h as for Figure 5. Relative levels compared to control are shown for all samples. Mean* ± *SE of three to six separate wells are shown. These data are representative of three separate experiments.*

*bp* < *0.001 when compared to IL-4/Forskolin (IL-4 FSK) treated cells. c p* < *0.0001 when compared to IL-4/Forskolin (IL-4 FSK) treated cells.*

*dp* < *0.01 when compared to IL-4/Forskolin (IL-4 FSK) treated cells.*

*ND, not determined.*

Table 2 | Effect of ERK inhibitor U0126 on IFNγ/Forskolin-induced expression of *NOS2* and *TNF*. a


*a Bone marrow (BM)-derived macrophages were grown for 5 days in the presence of M-CSF and were analyzed as untreated (control) or treated with IFN*γ *and Forskolin without or with ERK inhibitor U0126 for 24 h as for Figure 5. Relative levels compared to control are shown for all samples. Mean* ± *SE of three to six separate wells are shown. These data are representative of three separate experiments. bp* < *0.01 when compared to IFN*γ*/Forskolin (IFN*γ *FSK) treated cells.*

*c p* < *0.05 when compared to IFN*γ*/Forskolin (IFN*γ *FSK) treated cells.*

for IL-4-induced upregulation of *Arg1*, *Mrc1* (**Figures 8A,B**), but not *Ym1* and *Fizz1* (**Figures 8C,D**). Moreover, *Ym1* was further upregulated by ERK inhibitor U0126 (**Figure 8D**). We also investigated the level of expression of member of CREB/ ATF family ATF3 as a transcriptional factor that is regulated by cAMP and found that ATF3 was upregulated by IL-4, suggesting possible involvement of this transcription factor in M2 polarization (**Figure 8E**). ERK inhibitor U0126 even further upregulated ATF3 in IL-4-treated macrophages (**Figure 8E**) demonstrating that other pathways and/or co-factors were responsible for the upregulation of ATF3. Thus, we confirmed ERK pathway was important for the upregulation of *Arg1* and *Mrc1* by IL-4. We also found that in addition to CREB, ATF3 might be also involved in M2 polarization process but an expression of this factor was ERK independent.

It was previously reported that M2 macrophages have proliferative capacity *in vivo* upon systemic administration of IL-4 (34). Since reported proliferation of tissue-resident macrophages in peritoneal cavity (35) and microglia in the CNS during EAE (14) could be connected with activation of ERK pathway, we investigated the level for the expression of Cyclin D (CCDN1). CCDN1 plays a central role in the regulation of cell division and it is required (but not sufficient without other co-factors such as CDK4 and CDK6) for the progression from G1 to S phase of cell cycle (36). We found that in BM-derived macrophages *CCDN1* was expressed at the baseline level, which was not surprising since these cells were expanded in culture in the presence of M-CSF. However, IL-4 further upregulated mRNA for Cyclin D1 (**Figure 8F**, IL4) and expression of *CCDN1* was completely blocked by ERK inhibitor (**Figure 8F**, IL4 U0126). This indicates that both M-CSF and IL-4 contributed to ERK-dependent expression of *CCDN1* in BM-derived macrophages. We confirmed reported study that M-CSF induced ERK phosphorylation (37) in BM-derived macrophages (data not shown) suggesting that both IL-4 and M-CSF could contribute to the ERK-dependent expression of *CCDN1*. Despite notable upregulation of *CCDN1* expression by IL-4, we did not find very dramatic difference in the proliferation of IL-4-treated macrophages *in vitro* probably due to baseline level of M-CSF-driven proliferation of these cells (**Figures 8G,H**). However, ERK inhibitor U0126 did decrease proliferation of IL-4-treated BM-derived macrophages by 60%, which was consistent with reduced *CCDN1* expression in the presence of ERK inhibitor (**Figures 8G,H**). Thus, we found that ERK pathway is critical for the expression of *Arg1*, *Mrc1*, and *CCDN1* and downmodulation of *NOS2*.

### ERK Inhibitor Upregulate Expression of M1 Marker CD86 in IL-4-Activated Macrophages and NOS2 in IFN**γ**-Activated Macrophages

We further investigated and found that ERK inhibitor U0126 upregulated expression of general activation marker MHC class II and M1 marker CD86 in IL-4-treated BM-derived macrophages, but not in IFNγ-treated macrophages (**Figure 9A**; **Table 3**). In addition, ERK inhibitor further upregulated expression of *NOS2* (**Figure 9B**), but not *TNF* (**Figure 9C**). We observed a trend of ERK inhibitor upregulating MHC class II in IFNγ-treated macrophages (**Figure 9A**), but this trend not statistically significant (**Table 3**). Thus, we found that ERK pathway was antagonistic for the expression of activation marker MHC class II and M1 markers CD86 and *NOS2*.

### ERK Inhibitor Block Upregulation of Arg1, Mrc1, and Cyclin D1 in IL-4-Activated Resident Peritoneal Macrophages

We confirmed our results obtained on BM-derived proliferating macrophages in non-proliferating resident peritoneal macrophages. This confirmation was also important to prove essential role of ERK in M2 polarization of various types of IL-4-stimulated macrophages. Similar to BM-derived macrophages, ERK pathway was critical for IL-4-induced upregulation of *Arg1* and *Mrc1*, but not *Ym1* or *Fizz1* (**Figures 10A–D**). *ATF3* was even further upregulated by ERK inhibitor U0126 when compared to BM-derived macrophages (**Figure 10E**). As expected, *CCDN1* was not expressed in nonproliferating peritoneal macrophages, but it was induced by IL-4 and completely blocked by ERK inhibitor (**Figure 10F**).

Figure 8 | Analysis of the role of ERK pathway in upregulation of M2 markers and proliferative capacity of bone marrow (BM)-derived macrophages polarized with IL-4. BM-derived macrophages were grown for 5 days in the presence of M-CSF and were analyzed as untreated (control) or treated with IL-4, or IL-4 with ERK inhibitor U0126 for 24 h as for Figure 2. (A–F) Analysis of M2 marker expression. To assess expression of M2-associated and proliferation markers, RNA was isolated and the expression of *Arg1*, *Mrc1*, *Ym1*, *Fizz1*, *ATF3*, and *CCDN1* (Cyclin D1) was analyzed by real-time RT-PCR as described in Section "Materials and Methods." (G,H) Analysis of macrophage expansion. To assess macrophages proliferation, BrdU was added 14–16 h before analysis to cell cultures, after which the cells were stained for surface markers CD11b and F4/80 and intracellular BrdU and analyzed by three-color flow cytometry as described in Section "Materials and Methods." Representative contour-plots for CD11b+F4/80+ gated cells are shown in panel (G). Expression of F4/80 is shown on *x*-axis and expression of BrdU is shown on *y*-axis. Percentages of BrdU-positive cells are shown on upper right quadrants of each contour-plot. Quantitative analysis for mean ± SE of percentages of CD11b+F4/80+BrdU+ cells is shown in panel (H). In panels (A–F,H), mean ± SE of three to five culture wells is shown (\*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001; \*\*\*\*\**p* < 0.00001).

Figure 9 | Analysis of the role of ERK pathway in upregulation of M1 markers in bone marrow (BM)-derived macrophages polarized with IFNγ. BM-derived macrophages were grown for 5 days in the presence of M-CSF and were analyzed as untreated (control) or treated with IFNγ, or IFNγ with ERK inhibitor U0126 as described in Section "Materials and Methods." (A) Analysis of expression of surface markers MHC class II and CD86 on BM-derived macrophages. To assess expression of macrophage activation marker MHC class II and M1 marker CD86, macrophages were stained for CD11b and F4/80, MHC class II and CD86. CD11b+F4/80+ gated cells were analyzed for the expression of MHC class II and CD86 by four-color flow cytometry as described in Section "Materials and Methods." A representative histogram for the expression of MHC class II (upper row) and CD86 (bottom row) of CD11b+F4/80+ gated cells is shown. The solid line indicated staining for MHC class II or CD86, and the dotted line indicated staining for isotype-matched control. The bar shows percentage of MHC class II or CD86-positive cells. The mean fluorescence intensity (MFI) of expression of MHC class II or CD86 is shown at the bottom of each histogram. (B,C) Analysis of expression of nitric oxide synthetase (NOS)2 and TNF M1 markers. To assess expression of two other M1-associated markers *NOS2* and *TNF*, RNA was isolated and expressions of *NOS2* and *TNF* were analyzed by real-time RT-PCR as described in Section "Materials and Methods." In panels (B,C), mean ± SE of three to five culture wells is shown (\*\*\**p* < 0.001; \*\*\*\**p* < 0.0001).

Table 3 | Effect of ERK inhibitor U0126 on expression of cell surface markers MHC class II and CD86 in M2 (IL-4)- and M1 (IFNγ)-polarized bone marrow (BM) derived macrophages.a


*a BM-derived macrophages were grown for 5 days in the presence of M-CSF and were analyzed as untreated (control) or treated with IL-4 or IFN*γ *without or with ERK inhibitor U0126 for 24 h as for Figure 5. After, which macrophages were stained for CD11b, F4/80, MHC class II, and CD86. CD11b*+*F4/80*+ *gated cells were analyzed for the expression of MHC class II and CD86 by four-color flow cytometry as described in Section "Materials and Methods." Mean* ± *SE of mean fluorescence intensity of MHC class II or CD86 on macrophages from four separate wells are shown. These data are representative of two separate experiments.*

*bp* < *0.0001 when compared to IL-4 only treated cells (IL-4).*

*c Not significant difference when compared to IFN*γ *only treated cells (IFN*γ*).*

*dp* < *0.01 when compared to IL-4 only treated cells (IL-4).*

We confirmed that IL-4-treated peritoneal macrophages did not proliferate *in vitro* (data not shown), indicating that other co-factors (e.g., M-CSF) and other pathways (e.g., Akt) (35) are likely to be required for their reported IL-4-driven proliferation *in vivo*. Thus, these data demonstrate the importance of ERK pathway in IL-4-induced expression of Arg1, Mrc1, and Cyclin D2 in peritoneal macrophages. This also supports our general concept that ERK is essential pathway for M2 activation.

### Forskolin Upregulate miR-124 in BM-Derived Macrophages in CREB/ATF3-Dependent Manner

We have previously found that miR-124 plays an important role in M2 macrophage polarization in the CNS and the periphery

U0126 for 24 h as for Figure 8. To assess expression of M2-associated and proliferation markers, RNA was isolated and the expression of *Arg1*, *Mrc1*, *Ym1*, *Fizz1*, *ATF3*, and *CCDN1* (Cyclin D1) was analyzed by real-time RT PCR as for Figure 3. In panels (A–F), mean ± SE of three to five culture wells is shown (\**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; NS, not significant).

leading to suppression of EAE (10, 17). Here, we found that Forskolin affected miR-124 expression in macrophages. We hypothesized that Forskolin induced transcription of precursors of miR-124 affecting cAMP-responsive transcription factors. By performing *in silico* analysis of promoter areas of miR-124 precursor RNAs pre-miR124-1, pre-miR-124-2, and pre-miR-124-3, we found the presence of binding sites for several transcription factors from cAMP-responsive CREB/ATF family of transcription factors in pre-miR124-2 and pre-miR-124-3 promoter areas (Figure S2 in Supplementary Material). The most important factors from the list of potential regulators of miR-124 (Figure S1 in Supplementary Material) were ATF3 and CREB. CREB was upregulated by Forskolin (**Figure 6A**), while ATF3 was upregulated in M2 macrophages by IL-4 (**Figures 8E** and **10E**). It was Veremeyko et al. Forskolin Suppress Neuroinflammation

shown to be induced by IFNβ, a known FDA-approved drug (38). In addition, ATF3 was recently found to be induced in myeloid cells by dimethyl fumarate, another drug used for MS treatment (39, 40). ATF3 was also shown to promote anti-inflammatory TGFβ signaling and inhibited TNF, which is associated with the phenotype of M2 macrophages (41–43). However, when we used taurolidin, a known activator of ATF3 (44), we found only modest 1.5-fold upregulation of miR-124 (Figure S3 in Supplementary Material). It was recently shown that CREB-CREBβ axis play a major part in induction of expression of M2-associated genes *Arg1* and *Mrc1* (28). Forskolin induced early phosphorylation of CEBPβ (**Figure 6B**) showing activation of CREB–CEBPβ axis in macrophages by Forskolin. Thus, we found that Forskolin substantially upregulated miR-124 in macrophages acting through activation of CERB.

## REST Does Not Play a Role in Upregulation of miR-124 in Unmanipulated or Forskolin-Treated Macrophages

It was shown that miR-124 is highly expressed in neurons due to inactivation of transcriptional repressor REST (45). We also found that in contrast to neuronal cells, REST (but not the other alternative splicing isoform REST4) was highly expressed in macrophages on mRNA (Figure S4A in Supplementary Material) and protein (Figure S4B in Supplementary Material) levels. Importantly, the REST expression was not downregulated by Forskolin (Figures S4A,B in Supplementary Material). Thus, REST was not downregulated or functionally inhibited by Forskolin. Moreover, REST inhibitor X5050 failed to upregulate miR-124 in unstimulated macrophages (Figure S4C in Supplementary Material), while application of REST inhibitory RNA (NRSE) resulted in only modest 1.5-fold upregulation of miR-124 (Figure S4D in Supplementary Material). Thus, we demonstrated that Forskolin upregulated miR-124 in macrophages in RESTindependent manner.

### DISCUSSION

This study showed that cAMP pathway is more important to modulate function of macrophages rather than T cells during EAE. We demonstrated the importance of ERK pathway for M2 polarization induced by IL-4 alone or by the synergetic action of Forskolin and IL-4 leading to the upregulation of miR-124, *Arg1*, and *Mrc1*. Forskolin also synergized action of IFNγ leading ERK phosphorylation and upregulation of NOS2; however, this effect was about 10-fold weaker when compared with stimulation of *Arg1*. In addition, we found that in contrast to Forskolin and IL-4, IFNγ alone did not induce ERK phosphorylation. We have previously found that miR-124 promote expression of *Arg1* and *Mrc1* and inhibit *NOS2* (10, 17) further contributing to M2 phenotype. Summarizing all our data, we proposed the model in which IL-4 and Forskolin activated ERK pathway that promote expression of miR-124, Arg1, and Mrc1 (**Figure 11**). Forskolin alone or in combination with IFNγ also promoted NOS2, but this pathway is likely antagonized by miR-124, which we found to be critical for M2 phenotype of

Figure 11 | Model of changing balance toward M2 by Forskolin in mixed inflammatory conditions in the presence of IFNγ and IL4 in the central nervous system (CNS) during EAE. Our study demonstrated that IFNγ results in the upregulation of M1 marker nitric oxide synthetase (NOS)2 in an ERK-independent manner, since IFNγ alone did not result in phosphorylation of ERK. However, IFNγ and Forskolin resulted in ERK1/2 phosphorylation, contributing to NOS2 expression. IL-4 alone resulted in upregulation of M2 markers microRNA (miR)-124, arginase (Arg)1, and Mcr1 in an ERKdependent manner. Forskolin synergized IL-4 to cause ERK phosphorylation and further upregulation of miR-124, Arg1, and Mrc1 in ERK-dependent manner. Upregulation by IL-4 and/or Forskolin miR-124 further upregulates Arg1 and downregulates NOS2, contributing to skewing macrophages toward M2 by Forskolin in the presence of both IL-4 and IFNγ. Thus, Forskolin skews balance toward M2 in mixed inflammatory conditions such as the CNS autoimmune inflammation during EAE.

macrophages in the CNS (10) (**Figure 11**). Our *in vitro* studies were further tested in *in vivo* in mouse model of autoimmune neuroinflammation demonstrating strong therapeutic effect of Forskolin and changing balance toward M2 in the CNS in mouse model of MS.

The exact role of ERK pathway in macrophage polarization remains enigmatic. It was shown that ERK pathway is required for macrophage development and M2 skewing of monocytes under the action of M-CSF during their growth and differentiation vs. M1 phenotype when monocytes are expanded in differentiated under the action of GM-CSF (30, 37). However, the role of ERK pathway in conventional M2 macrophages that were polarized with IL-4 was not investigated in details. Much less is known about the activation of ERK pathway in macrophages during EAE in mixed inflammatory conditions (IL-4 and IFNγ) and possible outcome of activation of this pathway in macrophages for the development and resolution of neuroinflammation. Our data confirmed that M-CSF-induced ERK phosphorylation, but we also demonstrated that IL-4 induced ERK1/2 phosphorylation, which was dramatically increased when IL-4 was used together with Forskolin. Moreover, we demonstrated that ERK1/2 phosphorylation was critical for the expression of miR-124, *Arg1*, *Mrc1*, and Cyclin D1 in M2 macrophages that were polarized by IL-4 with or without Forskolin. In addition to M2 activation by IL-4, it was shown in diabetes model that TGFβ1 in combination with high level of glucose resulted in ERK-dependent upregulation of Arg1, suggesting that ERK pathway is involved regulation of Arg1 induced by other M2 stimuli such as TGFβ1 (46). These data support our findings that ERK inhibitor blocked upregulation of Arg1 caused by IL-4. In further support of our data with ERK inhibitors, it was recently shown that anticancer drug puerarin inhibited ERK pathway in TIMs or in IL-4-treated macrophages by skewing balance toward M1 (29). Thus, stimulation of ERK pathway in macrophages appeared to be beneficial for M2-skewing and downmodulation of Th1/17-driven autoimmune inflammation, and *vice versa*, inhibition of ERK pathway is advantageous for M1 skewing in tumor microenvironment and stimulation of anticancer immunity. Taken together, modulation of ERK pathway in macrophages open new possibilities to change M1/M2 balance in various spectrum of disorders ranging from diabetes and cancer to autoimmune disorders and allergy.

ERK pathway is traditionally connected with proliferative capacity of the cells; however, its role in proliferation of macrophages is still not clear (47). It was discovered that tissueresident M2 macrophages (including CNS-resident microglia during EAE) have high-proliferating capacity *in vivo* when compared with M1 macrophages, but molecular basis for this phenomenon remained not known (14, 34). It was shown that Forskolin-induced proliferation of thioglycollate-elicited macrophages demonstrating important role of CREB in proliferation of macrophages during inflammation (48). Our study linked Forskolin with ERK activation and highlighted importance of ERK pathway in terms of upregulation of expression of Cyclin D1 and enhancement of macrophage proliferation by IL-4. However, we also found that activation of ERK pathway is required but not sufficient to induce and mediate proliferation of M2 macrophages. Further studies will elucidate the role of other pathways besides ERK in this process. One possible candidate is Akt pathway, which is also induced by M-CSF and probably IL-4 (35, 48).

ERK pathway was shown to be important *in vivo* for allergic lung inflammation (33). In EAE model, it was recently shown that inhibition of ERK resulted in downregulation of GM-CSF, a cytokine that promote M1 polarization (49, 50). However, the role of ERK is difficult to estimate *in vivo* during autoimmune neuroinflammation since besides macrophages, ERK inhibitors also affect CD4 T cells and dendritic cells (51, 52). Our study suggests that stimulation of ERK pathway is beneficial during neuroinflammation by changing balance toward M2 in inducing miR-124. We previously demonstrated that miR-124 was very important *in vivo* for EAE and allergic lung inflammation models promoting M2 polarization while not affecting CD4 T cells and dendritic cells (10, 17). Therefore, Forskolin is promising drug that stimulate miR-124 in macrophages and suppress autoimmune neuroinflammation.

The role of cAMP in M2 polarization was recently highlighted in several studies (24, 25, 28). It was also demonstrated that treatment of cultured macrophages with 8-bromo-cAMP together with IL-4 synergistically activated Arg1 promoter (53). We confirmed that cAMP-elevating agent Forskolin also synergized IL-4 to induce 18-fold upregulation of Arg1. Our study also demonstrated for the first time important role of ERK pathway that most likely modulate or serve as a co-factor for other known pathways such as CREB–CEBPβ axis and STAT6 (28, 53). Currently, the role of cAMP in the modulation of ERK activity remains controversial. It was shown that during macrophage development cAMP inhibit M-CSF-induced ERK expression demonstrating negative role in macrophages activation/polarization (54). Our study clearly demonstrate that cAMP activate not only CREB but also ERK, which play an important role in M2 polarization. In addition to CREB, we also found that other members of this family of cAMP-induced transcriptions factors such as ATF3 are upregulated in IL-4-treated M2 macrophages. Quite interesting that ATF3 is downregulated sevenfold by Forskolin as shown for human macrophages (55). Our data also demonstrated that ATF3 was downregulated fivefold by Forskolin in mouse BM-derived macrophages (data not shown). Thus, CREB and ATF3 are regulated by Forskolin in opposite ways. This implies that upregulation of ATF3 in M2 macrophages treated with ERK inhibitors could happen due to compensatory mechanism. In the CNS, ATF3 was upregulated early in EAE during preclinical stage, when few infiltrating leukocytes were migrating into CNS (56), suggesting that ATF3 was most likely induced in microglial cells, which become activated before the onset of EAE (14). Thus, cAMP pathway has multiple actions in M2 skewing of macrophages *in vitro* acting through multiple transcriptional factors, which require further investigation.

The mechanisms how Forskolin skew macrophages toward M2 *in vivo* during EAE is even complex when compared with *in vitro* models since Forskolin affect many different cell types. Macrophages are known to be critical for onset of EAE (57), but this disease is also initiated by autoimmune Th1 and Th17 effector cells (58). Forskolin was shown to inhibit proliferation of T cells *in vitro* during their priming by inhibiting IL-2 signaling (23). In our model, we started administration of Forskolin on the day of disease onset (day 14), when Forskolin affected already primed effector cells. The role of cAMP pathway is more complex in case of effector CD4 T cells, since there are contradictory studies demonstrating either inhibition or stimulation of Th1 and Th17 cells by this pathway (6, 23, 59). Thus, there is a possibility that Forskolin also promote expansion of Th1 and Th17 cells *in vivo* during EAE. If this is the case, we demonstrated that direct action of Forskolin on macrophages could overcome direct effect on T cells. However, what Forskolin was doing with encephalitogenic effector CD4 T cells *in vivo* remained enigmatic, since Forskolin could affect T cells indirectly by modulating function of antigen-presenting cells in the CNS. To address this conundrum, we performed *in vitro* experiment investigating direct effect of Forskolin on the proliferation of primed MOG-specific enriched population of CD4<sup>+</sup>Vβ11<sup>+</sup> T cells. Our study demonstrated that Forskolin did not have significant effect on proliferation of CD4 T cells that were restimulated with MOG peptide. Same results we received when we analyzed MOG-specific enriched population of CD4<sup>+</sup>Vβ11<sup>+</sup> cells: comparable level of proliferation was observed. On the other hand, we confirmed previously published studies that during polyclonal stimulation of T cells with anti-CD3, Forskolin decreased CD4 T cell proliferation. Our *in vitro* data indicate that MOG-specific autoimmune CD4 T cells were not substantially affected by Forskolin. Thus, we strongly believe that it is unlikely that Forskolin directly inhibits effector autoimmune CD4 T cells during EAE; rather, Forskolin affect macrophages, which result to indirect inhibition of encephalitogenic T cells in the CNS. Since Forskolin did not skew balance toward to M2 in the periphery (spleen) and did not have direct effect of autoimmune CD4 T cells, this likely explains why proliferation and IFNγ production by CD4 T cells in the spleen was not decreased by Forskolin during EAE. We believe that Forskolin affected macrophages in the CNS but not in the spleen is due to the presence of specific CNS microenvironment such as presence of internal source of IL-4 in the CNS (9). Indeed, our data indicate that Forskolin skew macrophages toward M2 in the presence of IL-4 or IL-4 and IFNγ.

In addition to macrophages, Forskolin is known to affect neuronal cells activating BDNF–TrkB–CEBPβ signaling pathway. In neurons, CREB–CEBPβ axis increase survival, repair, and plasticity (59, 60). Finally, Forskolin may affect oligodendrocytes and astrocytes by elevating cAMP and/or ERK1/2 levels and promoting their maturation and differentiation (61–63). We found increased LFB staining in the spinal cords of Forskolin-treated mice in comparison to within unmanipulated (control) mice; this increase could not be attributed to staining artifact. This suggests that Forskolin might improve myelination during recovery from EAE. Further experiments will establish other cell targets for this drug during EAE/MS.

Taken together, our data demonstrated new ERK-dependent mechanism of action of Forskolin in skewing of macrophages toward M2 *in vitro* and *in vivo* during EAE. Our study demonstrate an important role of M1/M2 balance in the CNS for the regulation of autoimmune inflammation and stimulation of pathogenic T cells, which has potential applications for the usage of Forskolin for the therapy of MS and other Th1-mediated autoimmune disorders.

### ETHICS STATEMENT

The study was performed in accordance with the recommendations of the ARRIVE guidelines (http://www.nc3rs.org.uk/ arrive-guidelines). All animal procedures were conducted under individual licenses from the Department of Health of Hong Kong government and approved by animal ethics committee form the Chinese University of Hong Kong.

### AUTHOR CONTRIBUTIONS

EP and TV conceived the study. TV and EP designed experiments. TV, AY, MD, IK, and EP conducted experiments. NB and EP performed flow cytometry data analysis. IP, AL, and TS contributed reagents and materials. TV, AY, IP, AL, TS, NB, and EP analyzed the data. TV, IP, AL, TS, NB, and EP prepared the manuscript.

### ACKNOWLEDGMENTS

We thank Brenda Chiang (Harvard University) for help in preparation of the manuscript.

# FUNDING

The work was supported by Research Grant Council-Early Career Scheme grant, reference no. 24100314 (Hong Kong) and by "5-100" Russian Academic Excellence Project.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00050/ full#supplementary-material.

Figure S1 | Effect of Forskolin on the activation of downstream signaling pathways CREB, CEBPβ, and ERK1/2 at a baseline level of BM-derived macrophages (Control), or M2 (IL-4) vs. M1 (IFNγ) polarizing conditions. BM-derived macrophages were grown for 5 days in the presence of M-CSF and were treated with Forskolin, or IL4, or IFNγ, or Forskolin/IL4, or Forskolin/ IFNγ, or Forskolin/IL4/IFNγ for durations of 10, 20, and 30 min or for 24 h. ERK inhibitor U0126 was used to block ERK1 and ERK2 phosphorylation (see Materials and Methods). Expressions of CREB, CEBPβ, and ERK1/2 and their phosphorylated forms (p-CREB, p-CEBPβ, and p-ERK1/2) were analyzed by Western blot as described in Section "Materials and Methods." β-Actin was used as a loading control. The same stripped membrane was used to evaluate the expression of all proteins including β-actin. These data are representative of three separate experiments (abbreviations: I, IFNγ; 4, IL-4; F, Forskolin; U, U0126).

Figure S2 | *In silico* analysis of the presence of cAMP-response elements (CRE) in promoter areas of microRNA (miR)-124 precursor molecule premiR-124-3 and potential transcription factors, which upregulate miR-124 expression in macrophages. (A) Mapping of cAMP-response elements within 2,000 bp promoter region upstream of miR-124 precursor molecule premiR-124-3. A list of transcription factors that could potentially bind CRE sites are shown on the right. Two selected transcription factors CREB and activating transcription factor (ATF)3 are marked in red. Similar CRE sites for binding of CREB/ATF family transcriptions factors were also found for pre-miR-124-2 but not pre-miR-124-1 precursor molecule (data not shown). (B) Promoter area for pre-miR-124-3 with CRE sites are shown for mouse chromosome 2 and human chromosome 20.

Figure S3 | Influence of activating transcription factor (ATF)3 activator taurolidin on microRNA (miR)-124 expression in bone marrow (BM)-derived macrophages. BM-macrophages were treated with taurolidin for 24 h and miR-124 expression was analyzed as described in Section "Materials and Methods." Mean ± SE of triplicate is shown (\*\**p* < 0.01).

Figure S4 | The role of RE1-silencing transcription factor (REST) in the expression of microRNA (miR)-124 in macrophages. Bone marrow (BM)-derived macrophages were cultured for 5 days as described in Section "Materials and Methods." Then, samples that were untreated (Control), treated with Forskolin (A,B), or treated with REST inhibitors (C,D) for the indicated periods were analyzed for the expression of REST (A,B) or miR-124 (C,D). (A) Analysis of the expression of REST and REST4 in BM-derived macrophages on the mRNA level by RT-PCR. BM-derived macrophages were analyzed as untreated (Control) or treated with Forskolin (FSK) for 2–4 h. Mouse primary cortical neurons (Neu) were used as a control with a low level of REST and REST4 expressions. Negative control (Neg) was a medium. MW indicate molecular weight marker (in bp). GADPH loading control is shown below. (B) Analysis of the modulation of the expression of REST in BM-derived macrophages that were untreated for 4 or 24 h or treated with Forskolin for 4 or 24 h. Expressions of REST were analyzed with Western blot as described in Section "Materials and Methods." Actin was used as a loading control. Representative image is shown on the left, quantitative analysis of expression of REST normalized do β-actin is shown on the right. Mean ± SE of three separate experiments is shown. (C) BM-derived macrophages were incubated with REST inhibitor X5050 for 6 or 16 h (overnight), after which the expression of miR-124 was assessed by real-time RT-PCR as described in Section "Materials and Methods." (D) BM-derived macrophages were incubated with REST inhibitory dsRNA NRSE for 16 h, after which expression of miR-124 was assessed by real-time RT-PCR as described in Section "Materials and Methods." REST inhibitory dsRNA NRSE resulted in 1.5-fold upregulation of miR-124. Mean ± SE of triplicate is shown (\*p < 0.05).

# REFERENCES


via dimethylfumarate or cyclosporine A. *J Dermatol Sci* (2017) 87:246–51. doi:10.1016/j.jdermsci.2017.06.005


phenotype: IL-4 and cyclic AMP synergistically activate the arginase I promoter. *J Immunol* (2013) 191:2290–8. doi:10.4049/jimmunol.1202102


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

*Copyright © 2018 Veremeyko, Yung, Dukhinova, Kuznetsova, Pomytkin, Lyundup, Strekalova, Barteneva and Ponomarev. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Variations in cellular responses of Mouse T cells to adenosine-5**′**- Triphosphate stimulation Do not Depend on P2X7 receptor expression levels but on Their activation and Differentiation stage

*Hanaa Safya1 , Amine Mellouk1 , Julie Legrand2 , Sylvain M. Le Gall2,3, Mohcine Benbijja2,4, Colette Kanellopoulos-Langevin5 , Jean M. Kanellopoulos6 and Pierre Bobé1,2\**

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Ashutosh Chaudhry, Memorial Sloan Kettering Cancer Center, United States Girdhari Lal, National Centre for Cell Science, India*

> *\*Correspondence: Pierre Bobé pierre.bobe@u-psud.fr*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 11 August 2017 Accepted: 08 February 2018 Published: 27 February 2018*

### *Citation:*

*Safya H, Mellouk A, Legrand J, Le Gall SM, Benbijja M, Kanellopoulos-Langevin C, Kanellopoulos JM and Bobé P (2018) Variations in Cellular Responses of Mouse T Cells to Adenosine-5*′*- Triphosphate Stimulation Do Not Depend on P2X7 Receptor Expression Levels but on Their Activation and Differentiation Stage. Front. Immunol. 9:360. doi: 10.3389/fimmu.2018.00360*

*1UMR1174, INSERM, Université Paris-Sud, Orsay, France, 2 Institut André Lwoff, CNRS, Université Paris-Sud, Villejuif, France, 3UMR 970, INSERM, Université Paris Descartes, Paris, France, 4UMR 1012, INSERM, Université Paris-Sud, Le Kremlin Bicêtre, France, 5UMR 1149, INSERM, Université Paris Diderot, Paris, France, 6UMR 9198, I2BC, CNRS, Université Paris-Sud, Orsay, France*

A previous report has shown that regulatory T cells (Treg) were markedly more sensitive to adenosine-5′-triphosphate (ATP) than conventional T cells (Tconv). Another one has shown that Tregs and CD45RBlow Tconvs, but not CD45RBhigh Tconvs, displayed similar high sensitivity to ATP. We have previously reported that CD45RBlow Tconvs expressing B220/CD45RABC molecules in a pre-apoptotic stage are resistant to ATP stimulation due to the loss of P2X7 receptor (P2X7R) membrane expression. To gain a clearer picture on T-cell sensitivity to ATP, we have quantified four different cellular activities triggered by ATP in mouse T cells at different stages of activation/differentiation, in correlation with levels of P2X7R membrane expression. P2X7R expression significantly increases on Tconvs during differentiation from naive CD45RBhighCD44low to effector/memory CD45RBlowCD44high stage. Maximum levels of upregulation are reached on recently activated CD69+ naive and memory Tconvs. Ectonucleotidases CD39 and CD73 expression levels increase in parallel with those of P2X7R. Recently activated CD69+ CD45RBhighCD44low Tconvs, although expressing high levels of P2X7R, fail to cleave homing receptor CD62L after ATP treatment, but efficiently form pores and externalize phosphatidylserine (PS). In contrast, naive CD45RBhighCD44low Tconvs cleave CD62L with high efficiency although they express a lower level of P2X7, thus suggesting that P2X7R levels are not a limiting factor for signaling ATP-induced cellular responses. Contrary to common assumption, P2X7R-mediated cellular activities in mouse Tconvs are not triggered in an all-or-none manner, but depend on their stage of activation/differentiation. Compared to CD45RBlow Tconvs, CD45RBlowFoxp3+ Tregs show significantly higher levels of P2X7R membrane expression and of sensitivity to ATP as evidenced by

**Abbreviations:** ConA, concanavalin A; mAb, monoclonal antibody; P2X7R, P2X7 receptor; PS, phosphatidylserine; Treg, regulatory T lymphocyte; Tconv, conventional T lymphocyte.

their high levels of CD62L shedding, pore formation and PS externalization observed after ATP treatment. In summary, the different abilities of ATP-treated Tconvs to form pore or cleave CD62L depending on their activation and differentiation state suggests that P2X7R signaling varies according to the physiological role of T convs during antigen activation in secondary lymphoid organs or trafficking to inflammatory sites.

Keywords: P2X7, CD39, CD73, regulatory T lymphocyte, CD62L shedding, pore formation, phosphatidyslerine exposure, cell death

### INTRODUCTION

The P2X7 receptor (P2X7R) belongs to the P2X receptor family of adenosine-5′-triphosphate (ATP)-gated cation channels. The type of molecular and cellular responses induced by P2X7R depends on the length of stimulation by 0.5–1 mM concentrations of ATP in its tetra-anionic form (ATP4<sup>−</sup>) (1). Since P2X7R plays a major role in both innate and adaptive immunity, its involvement in the development of inflammatory and autoimmune diseases is extensively studied. Moreover, there is an increasing interest in the potential of P2X7R antagonists to treat a variety of inflammatory conditions (2). However, protective or detrimental effects of P2X7R on disease onset and/ or development have been observed (3, 4). These seemingly contradictory reports emphasize the need for an in-depth investigation on how the various cellular functions triggered by the ATP/P2X7R signaling pathway are regulated in immune cells under normal and pathological situations. Brief stimulation of P2X7R induces cation-specific ion channels formation (5) and phosphatidylserine (PS) exposure in the plasma membrane (6). Prolonged activation of P2X7R results in the formation of nonselective membrane pores, permeable to molecules with a molecular mass up to 900 Da. Continuous activation of P2X7R can lead to cell death by apoptosis (7–9) and/or necrosis (10, 11), depending on the cell type. In contrast, P2X7R may trigger an anti-apoptotic or growth promoting activity (12, 13). Numerous physiological functions have been attributed to P2X7R; notably, activation of caspase-1 (14, 15), maturation and secretion of cytokines such as IL-1β, IL-18, IL-6, and TNF-α (3, 14, 16–18), migration of leukocytes (19), and killing of intracellular pathogens in macrophages (20, 21). Moreover, P2X7R activation triggers proteolytic cleavage of membrane proteins such as the homing receptor L-selectin (CD62L), the low affinity receptor for IgE (CD23) (22, 23), TNF-α (24), IL-6 receptor (25), and the amyloid precursor protein (26). P2X7R also regulates the early signaling events involved in T-cell activation. Upon antigen stimulation, T lymphocytes release ATP, which induces Ca2<sup>+</sup> influx, NF-AT activation, and IL-2 production through P2X7R activation (27–30). Moreover, ATP plays a crucial role in regulating the differentiation of CD4<sup>+</sup> T cells into Th17 cells (31, 32). During chronic inflammation, ATP could also facilitate the conversion of regulatory CD4<sup>+</sup> T cells (Treg) into Th17 cells (33). Sensitivity to ATP varies among different T-cell subpopulations. Thus, CD8<sup>+</sup> T cells from the spleen, lymph nodes, or liver exhibit low levels of P2X7R membrane expression and ATP sensitivity, whereas both are displayed at high levels in intestinal CD8<sup>+</sup> T cells (34). Concerning CD4 Tregs, although significantly higher levels of *p*2X7 mRNA were found in CD4<sup>+</sup> Tregs compared to CD4<sup>+</sup> conventional T lymphocyte (Tconvs) (33), contradictory reports have been published about their sensitivity to P2X7R-induced cell death. In one report, CD25<sup>+</sup>CD4<sup>+</sup> Tregs are markedly more sensitive to P2X7R stimulation than CD25<sup>−</sup>CD4<sup>+</sup> Tconv (35). In another, the sensitivity of CD4<sup>+</sup> Tregs to P2X7R stimulation is normal and similar to that of CD4<sup>+</sup> Tconvs, provided that the latter express, like Tregs, low levels of the RB isoform of the transmembrane tyrosine phosphatase CD45 (CD45RB) (36). In Tconvs, the sensitivity to ATP-induced PS externalization and cell death appears to inversely correlate with the levels of CD45RB membrane expression. Thus, CD4<sup>+</sup> T cells expressing low levels of CD45RB (CD45RBlow) are significantly more sensitive to ATP stimulation than their counterpart expressing high levels of CD45RB (CD45RBhigh) (36, 37). However, the levels of P2X7R membrane expression were not determined in these studies. Moreover, we have shown that effector T lymphocytes become totally resistant to P2X7R stimulation following the plasma membrane expression of the B220 isoform of CD45 (or CD45RABC) (38) during the process of activation-induced cell death (39–41). The resistance of B220<sup>+</sup> T lymphocytes to ATP stimulation is due to the loss of P2X7R expression at the plasma membrane, as it is retained in the cytosol (38).

High and low levels of CD45RB expression on mouse T cells are a feature of naive and antigen-activated cells, respectively (42). Therefore, one could conclude from previous reports (36, 37) that activated T cells (CD45RBlow) are more sensitive to ATP-induced PS externalization and cell death than naive T cells (CD45RBhigh). However, we found that recently activated naive CD69<sup>+</sup>CD45RBhighCD44low Tconvs show a significantly reduced ability to proteolytically cleave CD62L compared to naive CD69<sup>−</sup>CD45RBhighCD44low T cells although they express higher levels of P2X7R. The reverse situation was observed for ATP-induced pore formation, and to a lesser extent for PS externalization, which were significantly upregulated in recently activated naive compared to naive Tconvs. To our knowledge, this is the first report describing a complete dissociation of ATP-induced cellular activities during the activation and/ or differentiation of Tconvs, regardless of the levels of P2X7R and ectonucleotidases CD39 and CD73 membrane expression. Compared to CD45RBlow Tconvs, Foxp3<sup>+</sup> Tregs that have an activated phenotype (CD45RBlow or CD25<sup>+</sup>) show higher levels of P2X7R membrane expression and of sensitivity to ATP. Thus, our present data show that the regulation of T cell sensitivity to ATP is far more complex than previously considered, as we found that P2X7R-mediated cellular activities in T-cell subsets are not dependent on the levels of P2X7R membrane expression and not triggered in an all-or-none manner. They depend on their stage of activation/differentiation.

## MATERIALS AND METHODS

### Reagents

Adenosine-5′-triphosphate, phorbol myristate acetate (PMA), concanavalin A (ConA), EGTA, and KN-62 (1-[N,O-bis(5- Isoquinolinesulfonyl)-N-methyl-L-tyrosyl]-4-phenylpiperazine) were purchased from Sigma-Aldrich (St. Louis, MO, USA). YO-PRO-1 and YO-PRO-3 dyes and BAPTA-AM were from Life Technologies (Carlsbad, CA, USA). Metalloprotease inhibitor GM6001 was from Chemicon International (Temecula, CA, USA). ATP solutions were prepared extemporaneously from 100 mM stock solution (pH 7.4) stored at −20°C. Because divalent ions affect the potency of ATP4<sup>−</sup> to bind P2X7R, the cell medium used to activate PX7R contains low concentrations of Mg2<sup>+</sup>.

## Mice

Wild-type C57BL/6J (B6) and *P2rx7* knockout B6.129P2- *P2rx7tm1Gab*/J (P2X7R KO) (16) mice originally from The Jackson Laboratory (Bar Harbor, ME, USA) were maintained in our animal facilities (CNRS SEAT UPS44, Villejuif, France and animalerie NeuroPSI, Orsay, France). B6.Cg-*Foxp3tm1Mal*/J (Foxp3GFP) (43) mice were kindly provided by Dr Géraldine Schlecht-Louf (INSERM UMR 996, France). All the experiments were conducted in accordance with French (décret n° 2013-118) and EU (directive 86/609/EEC) guidelines for the care of laboratory animals and approved by our local research ethics committee (CEEA 59).

## Flow Cytometry Immunophenotyping Assays

Spleen cell suspensions were phenotyped by flow cytometry using fluorescent-conjugated monoclonal antibody (mAb): anti-CD90.2/Thy1.2 (clone 30-H12), anti-B220 (clone RA3-6B2), anti-CD45RA (Clone 14.8), anti-CD45RB (clone C363.16A), anti-CD45RC (C363-16A), anti-CD4 (clone GK1.5), anti-CD69 (clone H1.2F3), anti-CD44 (clone IM7), anti-CD62L (clone MEL-14), anti-CD197/CC-chemokine receptor 7 (CCR7) (clone 4B12), CD39 (clone 24DMS1), and CD73 (clone TY/11.8) (all from eBioscience). P2X7R was detected using a rabbit polyclonal anti-P2X7R serum described in Le Gall et al. (38) and fluorescentconjugated goat anti-rabbit IgG F(ab)′2 secondary antibodies (eBioscience). Fluorescent-conjugated rat IgG2a, IgG2b or Armenian hamster IgG mAbs were used as the isotype control (eBioscience). Use of mAb to mouse Fcγ receptor (eBioscience) avoided non-specific antibody binding. Data acquisition was performed at the Flow cytometry core facility at I2BC, CNRS UMR 9198.

## CD62L Shedding, PS Exposure, Pore Formation, and Cell Death Assays

Spleen cells suspended in RPMI 1640 medium (Invitrogen, France) were treated with ATP or PMA in a humidified 5% CO2

atmosphere at 37°C for 30 min or 2 h, depending on the assay. After washing with RPMI 1640 medium, cells were resuspended in FACS buffer (eBioscience) and stained for 30 min on ice with phenotype-specific fluorescent mAbs and fluorescentconjugated anti-CD62L mAb to assess CD62L shedding. PS cell surface exposure was detected on mAb-labeled cells using FITC- or PE-Annexin V apoptosis detection kit according to the manufacturer's specifications (eBioscience, France). To quantify P2X7R-mediated pore formation, ATP treatment was performed in the presence of either the green-fluorescent YO-PRO-1 (molecular weight 629 Da) or the orange-fluorescent YO-PRO-3 (molecular weight 655 Da) nucleic acid dyes, depending on the fluorochromes used in the phenotyping step. Cell morphology (FSC/SSC) and Annexin V staining were used to quantify dead/ dying cells (Annexin V<sup>+</sup> FSClow SSChigh) by flow cytometry. In some experiments, cells were pretreated with metalloprotease inhibitor GM6001, P2X7R antagonist KN-62, intracellular calcium chelator BAPTA-AM (10 µM) or extracellular calcium chelator EGTA (5 mM) for 30 min at 37°C with 5% CO2 prior treatment with ATP or PMA.

### Transfection and Flow Cytometry Assays

The COS7 epithelial cell line was transfected transiently with a pCDEF3 expression vector containing CD45RABC cDNA (kindly provided by Dr A. Weiss, UCSF, San Francisco, CA, USA). At 48 h after transfection, the cells were stained with FITC-conjugated anti-CD45RA (clone 14.8), PE-conjugated anti-CD45RB (clone 16A), APC-conjugated anti-CD45RC (clone GL24), and PE Cy5.5-conjugated anti-CD45RABC (clone RA3-6B2) mAbs, and analyzed by flow cytometry.

### Statistical Analysis

Data are reported as mean ± SEM. Comparisons between untreated and treated groups were made by Student's *t*-test. Degrees of significance are indicated as follows: \**p* ≤ 0.05, \*\**p* ≤ 0.01, \*\*\**p* ≤ 0.001.

# RESULTS

### ATP-Mediated Cellular Activities and P2X7R Membrane Expression in T Cells with either High or Low Expression of CD45RB

Effector T cells express low levels of the CD45RB (42). Previously, we have shown that effector CD45RBlow T cells become resistant to ATP stimulation when they reach a preapoptotic stage characterized by the plasma membrane expression of B220 (or CD45RABC) (38). Therefore, reports (36, 37) showing that CD45RBlow effector T cells are notably more sensitive to BzATPmediated PS exposure and cell death than CD45RBhigh naive T cells appear contradictory to our previous findings (38). The different ligands (ATP vs. BzATP) used to activate P2X7R could explain the discrepancy between our data and those of Taylor et al. However, we favor the hypothesis that the anti-CD45RB mAb used to gate CD45RBhigh T cells in these reports (36, 37) also detect B220<sup>+</sup> (or CD45RABC<sup>+</sup>) T cells because the anti-CD45RB mAb (clone 16A) recognizes an exon B-dependent epitope (44). To test our hypothesis, CD45-negative COS7 epithelial cells were transfected with a plasmid vector encoding mouse CD45RABC (or B220) and stained with fluorescent anti-B220, anti-CD45RA, anti-CD45RB and anti-CD45RC mAbs. We observed that anti-CD45RB mAb recognized CD45RABC<sup>+</sup> COS7 cells, but not untransfected COS7 cells (CD45RABC<sup>−</sup>), confirming that anti-CD45RB mAb cannot distinguish between CD45RB and CD45RABC isoforms. Likewise, anti-CD45RA and anti-CD45RC mAbs recognized CD45RABC<sup>+</sup> COS7 cells (Figure S1 in Supplementary Material). Therefore, we have quantified ATP-mediated CD62L shedding, pore formation, PS exposure and cell death in B220-negative T cells with either high or low CD45RB cell surface expression (**Figure 1**). In agreement with previous studies (37), we found CD45RBlow T cells displayed higher sensitivity to ATP-mediated PS externalization than CD45RBhigh (**Figure 1C**). Moreover, CD45RBhigh T cells displayed markedly lower sensitivity to ATP-mediated CD62L

P2X7R KO T-cell subsets (gray histograms). The histograms are representative of six individual mice.

shedding (**Figure 1A**), pore formation (**Figure 1B**) and cell death (**Figure 1D**) than CD45RBlow T cells. Altogether, our data suggest that P2X7R-mediated cellular activities are poorly triggered in naive CD45RBhigh T cells compared to CD45RBlow activated T cells.

Membrane P2X7R levels on T cells could be a limiting factor for triggering the various cellular responses induced by ATP, and reduced levels of P2X7R membrane expression could account for the lower sensitivity of CD45RBhigh naive T cells to ATP. Therefore, we quantified P2X7R on CD45RBhigh and CD45RBlow T cells using a rabbit polyclonal anti-P2X7R serum and flow cytometry. We found that P2X7R is notably less expressed on naive CD45RBhigh T cells than on activated CD45RBlow T cells (**Figure 1E**), suggesting that (1) ATP-mediated CD62L shedding, pore formation, PS externalization and cell death are triggered in the presence of high levels of P2X7R membrane expression, which are weakly express on the surface of naive T cells; (2) T-cell activation upregulates the expression of P2X7R making T cells able to perform all ATP-induced cellular activities.

### ATP-Mediated Cellular Activities and P2X7R Membrane Expression in Conventional and Tregs with CD45RBlow or CD25**+** Phenotype

The CD45RBlowCD4<sup>+</sup> T-cell subset encompasses conventional and Tregs (45). Different behaviors of Tconvs and Tregs upon ATP stimulation have been reported previously (35, 36). While one report concluded that Foxp3<sup>+</sup>CD25<sup>+</sup>CD4<sup>+</sup> Tregs were markedly more sensitive to ATP-induced PS exposure and cell death than CD25−CD4+ Tconvs (35), another reported a similar susceptibility to ATP-induced PS exposure and cell death between CD4<sup>+</sup> Tconvs and Tregs provided, however, that both T-cell subsets displayed the CD45RBlow phenotype (36). In an attempt to clarify these contradictory results, spleen cells from B6.Cg-*Foxp3tm1Mal*/J (Foxp3GFP) and P2X7R KO mice have been stimulated with ATP for 30 min or 2 h, and the levels of CD62L shedding, PS exposure, pore formation and cell death have been measured in CD4<sup>+</sup> Tconvs (either CD25<sup>+</sup> or CD45RBlow) and Foxp3<sup>+</sup>CD4<sup>+</sup> Tregs, which are mostly CD45RBlow (**Figures 2** and **3**).

conventional T lymphocytes (Tconvs) with CD25+ and/or CD45RBlo phenotype. (A–D) Spleen cells from Foxp3GFP B6 mice were either left unstimulated or stimulated with 500 µM ATP for 30 min (A–C) or 2 h (D) in the presence or absence of YO-PRO-3 fluorescent probe. Cells were subsequently stained with fluorescent monoclonal antibodies against phenotypic markers CD90, B220, CD4, CD45RB, and CD62L as well as Annexin V fluorescent probe. CD62L shedding, pore formation, phosphatidylserine (PS) exposure, or cell death were assessed by flow cytometry on gated GFP+ (Foxp3+) or GFP− (Foxp3−) CD90+CD4+B220<sup>−</sup> T cells with CD45RBhi, CD45RBlo, CD25+, or CD25− phenotype. Cell morphology (FSC/SSC) and Annexin V staining were used to quantify dead/dying cells (Annexin V+ FSCloSSChi). Results on CD62L shedding (A), pore formation (B), PS exposure (C), or cell death (D) are expressed as the ratio ± SEM (six mice per group) between the percentage of cells expressing or non expressing cellular activities (A–D) in the presence or absence of ATP, respectively. Data are representative of six independent experiments. Asterisks denote statistically significant differences between the indicated groups (\*\*\**p* ≤ 0.001). (E,F) P2X7R membrane expression on CD25+ or CD45RBlow Foxp3+ Tregs, and CD45RBhi or CD45RBlo CD90+CD4+B220− Tconvs was measured using rabbit polyclonal anti-P2X7R antiserum (1:100) and fluorescent-conjugated goat anti-rabbit IgG F(ab)′2 secondary antibodies. P2X7R-staining histograms of wild-type T cells (black histograms) are overlaid on P2X7R-staining histograms of P2X7R KO T cells (gray histograms). The histograms are representative of at least 6 individual mice.

We found that CD25<sup>+</sup>Foxp3<sup>+</sup> Tregs were three to four times more sensitive to ATP-mediated CD62L shedding (**Figure 2A**) and pore formation (**Figure 2B**) than activated CD25<sup>+</sup>Foxp3<sup>−</sup> Tconvs. However, CD25<sup>+</sup>Foxp3<sup>−</sup> Tconvs and CD25<sup>+</sup>Foxp3<sup>+</sup> Tregs displayed similar sensitivity to ATP-mediated PS externalization (**Figure 2C**) and cell death (**Figure 2D**). When ATP sensitivity was analyzed in Foxp3<sup>+</sup> Tregs and Foxp3<sup>−</sup> Tconvs expressing a CD45RBlow phenotype, we found again a greater sensitivity of CD45RBlowFoxp3<sup>+</sup> Tregs to ATP-mediated CD62L shedding (**Figures 2A** and **3A**) and pore formation (**Figures 2B** and **3B**) compared to ATP-treated CD45RBlowFoxp3<sup>−</sup> Tconvs, but not to ATP-mediated PS externalization (**Figures 2C** and **3C**) and cell death (**Figures 2D** and **3D,E**) that was similarly high in CD45RBlow Tconvs and CD45RBlow Tregs. As expected, CD4<sup>+</sup> T cells from P2X7R KO mice were totally resistant to ATP-induced cellular activities (data not shown). Altogether our results demonstrate a higher sensitivity of Tregs to P2X7R-mediated cellular activities compared to Tconvs.

The levels of P2X7R membrane expression are significantly higher on Foxp3<sup>+</sup> Tregs than on Foxp3<sup>−</sup> Tconvs (**Figure 2E**), and especially on CD45RBhigh naive Tconvs (**Figure 2F**). Interestingly, and in contrast with Foxp3<sup>+</sup> Tregs, P2X7R expression on CD45RBlow Tconvs is heterogeneous (**Figure 2F**). This would be in keeping with the phenotypic and functional heterogeneity of Tconvs and subsequently lead us to analyze the levels of P2X7R membrane expression and functions during their differentiation from naive to memory state.

### ATP-Mediated Cellular Activities and P2X7R Membrane Expression in Recently Activated Naive and Memory CD69**+** T Cells

**Figures 1**–**3** suggest that the cellular activities triggered by ATP correlate positively with the levels of P2X7R membrane expression. Moreover, these expression levels vary among Tconvs according to their state of activation and/or differentiation. Therefore, we have analyzed ATP-induced CD62L shedding, pore formation and PS exposure in B220<sup>−</sup>CD90<sup>+</sup> Tconvs expressing CD69 (**Figure 4**), which is the earliest surface marker expressed during T-cell activation. The expression of CD69 marks the activation of naive and memory T cells. CD69 is expressed within < 4 h of activation and peak expression is between 18 and 48 h after stimulation (46). The percentage of T lymphocytes that express CD69 is usually low in the spleen of B6 mice (10.2 ± 2.8%, *n* = 10 mice). Therefore, B6 mice were injected intravenously with 7 µg/g of T-cell mitogen ConA for

fluorescent probe. Cells were subsequently stained with fluorescent monoclonal antibodies against phenotypic markers CD90, CD4, B220, CD45RB, and CD62L as well as Annexin V fluorescent probe. CD62L shedding (A), pore formation (B), PS exposure (C), or cell death (D,E) were assessed by flow cytometry on gated GFP<sup>+</sup> (Foxp3+) or GFP− (Foxp3−) CD90+CD4+B220− T cells with CD45RBhi or CD45RBlo phenotype. Numbers reported in the dot plots indicate the percentages of CD62L<sup>+</sup> cells (A), YO-PRO-3+ cells (B), Annexin V+ FSChiSSClo living cells (C), or Annexin V+ FSCloSSChi dead/dying cells (D,E) (either CD45RBhi or CD45RBlo) in the gated CD90+CD4+B220− T-cell population. Data are representative of six independent experiments with six mice per group per experiment.

histograms) are overlaid on P2X7R-staining histograms of P2X7R KO T-cell subsets (gray histograms). Bar graph shows mean fluorescence intensity (MFI) ± SEM of P2X7R of 6 individual mice. Results are expressed as delta MFI (ΔMFI = MFIwild type − MFI KO), i.e., change in MFI relative to P2X7R KO Tconvs. Asterisks denote statistically significant differences between the indicated groups (\*\**p* ≤ 0.01, \*\*\**p* ≤ 0.001).

18 h to massively increase the numbers of splenic CD69<sup>+</sup> T cells (61 ± 7.5%, *n* = 14 mice) and their mean fluorescence intensity (MFI) (53 vs. 378). In contrast, the ConA treatment had no effect in both spleen cell number and the percentage of splenic T cells (Figure S2 in Supplementary Material). Unexpectedly, we found that CD69<sup>+</sup> T cells, but not CD69<sup>−</sup> T cells, failed to cleave their CD62L molecules after ATP treatment, both in terms of the percentages of CD62L+ T cells and levels of CD62L membrane expression (**Figure 4A**, and data not shown). In contrast, CD69<sup>+</sup> T cells efficiently form pore (**Figure 4B**) and externalize PS (**Figure 4C**) after ATP treatment. The inability of recently activated CD69<sup>+</sup> T cells to shed CD62L molecules was not related to defective proteolytic activity of ADAM-17 since both CD69<sup>−</sup> and CD69<sup>+</sup> T cells can shed CD62L after PMA treatment (**Figure 4A**). The expression of CD69 coincides with a strong upregulation of P2X7R membrane expression (**Figure 4D**) that is mainly observed on effector/memory CD45RBlow T cells (**Figure 5**).

In summary, recently activated CD69<sup>+</sup>CD45RBlow T cells, although expressing high levels of P2X7R, appeared mostly resistant to ATP-induced CD62L cleavage, indicating that P2X7R-mediated cellular activities are not triggered in an allor-none manner in Tconvs and vary according to their stage of activation.

# ATP-Mediated Cellular Activities and P2X7R Membrane Expression in Effector/ Memory CD45RBlowCD44high T Cells

The CD69+CD45RBlow T-cell subset is heterogeneous and encompasses recently activated naive and memory T cells. Therefore, the membrane expression levels of adhesion molecule CD44, CCR7, CD45RB, CD69, and P2X7R has been used to further explore the sensitivity of naive, recently activated, effector and central memory T cells to ATP in conjunction with the levels of P2X7R membrane expression. CD4<sup>+</sup> T cells (either CD69<sup>−</sup> or CD69<sup>+</sup>) with high or low expression levels of CD45RB and CD44 naive and effector/memory markers have been identified and gated using a sequential gating strategy (Figure S3 in Supplementary Material). The differentiation of naive CD45RBhighCD44low T cells into effector/memory CD45RBlowCD44high T cells is accompanied by a significant increase in P2X7R membrane expression (**Figure 6A**). A further significant upregulation of P2X7R membrane expression was observed on both naive CD45RBhighCD44low and memory CD45RBlowCD44high T-cell subsets following the expression of the early activation marker CD69 (**Figure 6A**). Interestingly, the sensitivity to ATP of naive CD45RBhighCD44low, effector memory CD45RBlowCD44high (either CCR7− or CCR7+) and recently activated CD69<sup>+</sup> (either naive or memory) T cells was not strictly

fluorescent-conjugated goat anti-rabbit IgG F(ab)′2 secondary antibodies. In contrast with Figure 4, mice had not been injected with concanavalin A to maintain a high frequency of naive and memory resting T cells. P2X7R-staining histograms of wild-type T-cell subsets (black histograms) are overlaid on P2X7R-staining histograms of P2X7R KO T-cell subsets (gray histograms). The histograms are representative of at least 6 individual mice. Asterisks denote statistically significant differences between the indicated groups (\**p* ≤ 0.05; \*\**p* ≤ 0.01; \*\*\**p* ≤ 0.001).

correlated with the levels of P2X7R membrane expression especially for ATP-induced CD62L shedding (**Figures 6B–D** and data not shown). ATP-treated naive CD62LhighCD45RBhighCD44low and effector/memory CD62LhighCD45RBlowCD44high T cells shed efficiently CD62L (**Figure 6B**). The ability to shed CD62L significantly decreased after the upregulation of activation marker CD69 particularly in naive CD45RBhighCD44low T cells (**Figure 6B**). In contrast, naive CD45RBhighCD44low T cells had a poor ability to form pore and externalize PS after ATP treatment (**Figures 6C,D**). A marked upregulation of the ability to form pore and externalize PS was observed on naive CD45RBhighCD44low T cells following antigenic-activation (CD69<sup>+</sup>), which is retained at the effector/ memory stage (**Figures 6C,D**). Because P2X7 displays a low affinity for ATP, we have evaluated whether the resistance to CD62L shedding observed in recently activated naive T cells upon stimulation with 0.5 mM of ATP could be overcome with higher doses of ATP. However, recently activated CD69<sup>+</sup>CD45RBhighCD44low T cells did not recover their ability to cleave CD62L after treatment with 1 or 2 mM ATP (data not shown). Finally, we found that calcium from the extracellular space and/or the intracellular stores was involved in pore formation and PS exposure in Tconvs, but not CD62L shedding, since they were significantly reduced in the presence of extracellular (EGTA) and/or intracellular (BAPTA-AM) calcium chelator (**Figure 7A**). As expected, pretreatment of spleen cells with P2X7R antagonist KN-62 inhibited ATP-induced cellular responses (**Figure 7A**). Likewise, all T-cell subsets from P2X7R KO mice were resistant to 0.5–2 mM ATP stimulation, as shown by a complete lack of CD62L shedding, pore formation and PS exposure (Figure S4B in Supplementary Material). As observed in CD69+ T cells (**Figure 4A**), ADAM-17 metalloproteases were not defective in ATP-treated central memory CD62LhighCD45RBlowCD44high T cells since PMA treatment was able to induce the cleavage of CD62L. Moreover, the shedding of CD62L induced by ATP in naive CD45RBhighCD44low T cells or by PMA in effector/memory CD45RBlowCD44high T cells could be prevented by the metalloprotease inhibitor GM6001, in a similar dose dependent-manner (Figure S4A in Supplementary Material), confirming the specificity of CD62L cleavage.

were assessed by flow cytometry on gated naive CD45RBhiCD44lo (either CD69+ or CD69−) and effector/memory CD45RBloCD44hi (either CD69+ or CD69−) CD90+B220− T cells. The gating strategy presented in Figure S3 in Supplementary Material has been followed. Results on CD62L shedding (B), pore formation (C), or PS exposure (D) are expressed as the mean percentage ± SEM of CD62L+, YO-PRO-3+, or Annexin V+ cells after ATP stimulation. Asterisks denote statistically significant (\**p* ≤ 0.05, \*\**p* ≤ 0.01, \*\*\**p* ≤ 0.001) differences between ATP-stimulated groups. Data are representative of at least six independent experiments with 7 mice per group per experiment.

To summarize, despite high levels of P2X7R membrane expression, CD45RBlowCD44high T cells, especially upon activation (CD69<sup>+</sup>), are strongly resistant to ATP-induced CD62L cleavage, but not pore formation and PS exposure, demonstrating that P2X7R-mediated cellular activities vary during their activation and differentiation into effector or central memory Tconvs, independently of the levels of P2X7R membrane expression.

### CD39 and CD73 Membrane Expression on Naive, Recently Activated and Effector/ Memory T Cells

Ectoenzymes CD39 and CD73 sequentially degrade extracellular ATP to adenosine. To examine whether the variation of P2X7R-mediated cellular responses of Tconvs might be due to enhanced hydrolysis of ATP following CD39 and/or CD73 overexpression, we have analyzed the levels of CD39 and CD73 membrane expression in Tconvs according to their state of activation and differentiation. Flow cytometry analyses show that the levels of CD39 and CD73 membrane expression increase with T-cell differentiation from naive CD45RBhigh CD44low to effector/ memory CD45RBlowCD44high stage (**Figure 7B**). Interestingly, the expression of the early activation marker CD69 by memory CD45RBlowCD44high Tconvs is accompanied by a strong upregulation of CD73 (**Figure 7B**). However, Tconvs strongly expressing ectonucleotidases also present higher levels of ATP-induced pore formation and PS externalization. Thus, it is likely that these ectonucleotidases do not control P2X7-mediated cellular responses in Tconvs.

### DISCUSSION

Although the ATP/P2X7R pathway is recognized as an important regulator of T cell functions (27–30), the respective sensitivities

and CD62L. After 30 min, cells were diluted in annexin-binding buffer and stained with Annexin V fluorescent probe. CD62L shedding, pore formation, or phosphatidylserine exposure were assessed within the gated CD69− CD44lo B220−CD90+ T-cell subpopulation by flow cytometry. Bars show the mean percentages ± SEM (*n* = 4–9 mice) of CD62L+ cells, YO-PRO-1+ cells and Annexin V+ cells after ATP stimulation (+) in the presence (+) or the absence (–) of KN-62, EGTA or BAPTA-AM. (B) CD39 and CD73 membrane expression was measured by flow cytometry on naive CD45RBhiCD44lo and effector/memory CD45RBloCD44hi T cells (either CD69− or CD69+). Mice had not been injected with concanavalin A, and the gating strategy presented in Figure S3 in Supplementary Material has been followed. Asterisks show statistically significant (\**p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\**p* ≤ 0.001) differences between ATP-stimulated groups. Data are representative of three independent experiments.

of Tregs and Tconvs to extracellular ATP are discussed and this point needs further clarification. In one report Tregs appeared to be markedly more sensitive to ATP than Tconvs, whereas in another, Tregs and Tconvs displayed similar high sensitivity to ATP provided that both T-cell populations expressed low levels of membrane phosphatase CD45RB (35, 36). However, given the large cellular and functional heterogeneity of the CD45RBlow Tconv population, no definitive conclusion can be drawn from these previous studies regarding the sensitivity of Tconvs to ATP. Thus, we have previously reported that effector CD45RBlow T cells become totally resistant to ATP at the preapoptotic stage (38). The sensitivity of T cells to ATP has also been correlated with P2X7R mRNA or protein expression levels (29, 33, 47). However, these studies were conducted in whole T-cell populations and not in T-cell subsets, which could express different levels of P2X7R depending on the stage of activation. Several reports reviewed in Ref. (48) do not show a strong correlation between mRNA expression levels and protein abundance, weakening previous conclusions on the relationship between P2X7R mRNA levels and ATP-induced cellular functions. Previous studies on the quantification of P2X7R protein expression have used Western blotting in whole-cell lysates from splenic T cells (29), which does not allow to discriminate between cell surface and intracellular P2X7R. Indeed, we have reported that P2X7R can accumulate in the cytosol of a T-cell subset without cell surface detection of P2X7R (38). Therefore, our present study aims to gain a clearer Safya et al. P2X7R Activities in T-Cell Subsets

picture on the sensitivity of the major subsets of T cells to extracellular ATP in relation with the levels of P2X7R membrane expression on each subset. Through four different cellular activities (CD62L shedding, pore formation, PS externalization, cell death) triggered following the stimulation of P2X7R, we have (1) compared the sensitivity of Tconvs and Foxp3<sup>+</sup> Tregs to ATP; (2) evaluated the ATP sensitivity of Tconvs at different stages of activation and differentiation; (3) quantified P2X7R membrane expression on Tconvs at different stage of activation and differentiation. In agreement with previous data (36, 37), we found that activated CD45RBlow Tconvs displayed significant higher sensitivity to ATP-induced PS externalization and cell death than naive CD45RBhigh Tconvs. Moreover, we found that activated CD45RBlow Tconvs also displayed significant higher sensitivity to ATP-induced CD62L shedding and pore formation than naive CD45RBhigh Tconvs (**Table 1**). In contrast with previous report (36), we do not find that CD45RBlowFoxp3<sup>+</sup> Tregs display higher sensitivity to ATP-induced cell death and PS externalization than activated CD45RBlowFoxp3<sup>−</sup> Tconvs (**Table 1**), but rather show significant higher sensitivity to ATP-induced CD62L shedding and pore formation. We suggest that the anti-CD45RB mAb used to control CD45RB expression levels in Tregs and Tconvs (36) could account for the discrepancy between this report and our present data. Indeed, we show that this mAb recognizes an epitope shared by both CD45RB and B220 molecules, suggesting that Taylor et al. (36) assessed P2X7R activity in a pool of B220<sup>+</sup> and CD45RBhigh Tconvs. Moreover, we had reported that B220-expressing effector Tconvs were totally refractory to ATP stimulation (38). Thus, in our present experiments, P2X7Rinduced cellular activities were analyzed in B220-negative gated T cells. The difference in ligands used to activate P2X7R might also explain the discrepancy observed between previous studies (36) and our present data. Indeed, it has been reported that BzATP triggers pore formation in lymphocytes with an EC50 value of about 15 µM compared to 85 µM for ATP. Thus, BzATP stimulates P2X7-induced pore formation up to 30% more than ATP (49). In all our experiments, splenocytes have been activated with the physiological agonist ATP instead of BzATP, to avoid an overstimulation of P2X7R-mediated cellular responses that could have masked potential differences between T-cell subsets.

Interestingly, we show that P2X7R-mediated cellular activities in Tconvs are not triggered in an all-or-none manner and their expressions depend on the stage of activation and differentiation. Thus, among the activated T-cell subsets, we found that recently activated CD69<sup>+</sup> T cells showed a significant reduction in their ability to shed CD62L in the presence of ATP despite significant higher levels of P2X7R expression than CD69<sup>−</sup> T cells. In contrast, CD69<sup>+</sup> T cells display significantly higher efficiency to form pore and externalize PS than CD69-negative T cells. CD69 expression marks the activation of both naive (CD44low) and memory (CD44high) T cells. Therefore, using a more detailed phenotypic characterization of Tconvs, we have compared the levels of P2X7R membrane expression and sensitivity to ATP of naive (CCR7<sup>+</sup>CD45RBhighCD 44low) and effector/memory (CD45RBlowCD44high either CCR7<sup>+</sup> or CCR7<sup>−</sup>), a recently activated (CD69<sup>+</sup>) (naive or memory) Tconvs. Naive CD62LhighCD45RBhighCD44low Tconvs had a notably higher ability to shed CD62L in the presence of ATP than lymphoidhoming central memory CD62LhighCCR7+CD44high T cells, thus suggesting that P2X7R does not play a central role in the shedding of homing-receptor CD62L at least during secondary immune responses. Antigenic stimulation of naive CD45RBhighCD44low Tconvs, as evidenced by the upregulation of CD69, leads to both decreased ability to shed CD62L and increased ability to form pore (**Table 1**). CD62L regulates the homing of naive and central memory T cells to secondary lymphoid organs whereas sphingosine 1-phosphate receptor 1 (S1PR1) controls T cell egress from these organs. CD62L is shed from the plasma membrane of T cells following activation. The S1RP1 ligand, S1P, is expressed at low concentration in lymphoid organs and at high concentration in circulatory fluids. The S1P gradient promotes S1PR1-dependent migration of T cells from secondary lymphoid organs into the blood and lymphatic circulation. CD69 expressed on recently activated T cells causes internalization and degradation of S1PR1, delaying T cell egress (50). The shedding of CD62L on activated T cells prevents their reentry into lymph nodes, and favors the acquisition of T-cell effector functions. It is a slow process that reaches its maximum 4–6 h postactivation (51). The reduced capacity of recently activated CD69<sup>+</sup> naive T cells to shed CD62L following autocrine activation of P2X7R could amplify the transient retention of recently activated T cells in the lymph nodes caused by the CD69-mediated inhibition of S1RP1, favoring a full differentiation of activated T cells into effector cells. Moreover, the higher capacity of recently activated CD69<sup>+</sup> naive T cells to form pore compared to CD69<sup>−</sup> naive T cells could participate to their full activation by increasing calcium entry through P2X7R acting as a costimulatory factor involved in the strength of T cell


*a Treated with 0.5 mM ATP.*

+*: low level;* ++*: moderate level;* +++*: high level;* ++++*: very high level.*

*bCell death was determined in* Figure 1 *(panel D) on whole CD45RBlow and CD45RBhigh Tconvs in the absence of CD69 and CD44 phenotyping.*

activation. Although it has been suggested that extracellular calcium is not a critical factor in pore formation (52), we found that calcium from the extracellular space and/or the intracellular stores was involved in pore formation and PS exposure in Tconvs, but not CD62L shedding. This finding agrees with our previous report showing that P2X7R-induced amyloid precursor protein shedding is independent of extracellular calcium (26). Because pore formation induces robust extracellular calcium influx (53), our data emphasize the role of P2X7R in the regulation of early signaling events involved in T-cell activation, as previously reported (27, 28, 30).

The expression of P2X7R splice variants (12, 13, 54) could account for the dissociation of ATP-induced cell activities observed during activation and differentiation of Tconvs. However, the analysis of the P2X7a and P2X7k splice variants by real time RT-PCR have shown that the Tconvs express the P2X7k variant whatever their stage of activation/differentiation (unpublished data). Membrane-bound ATPases play an important role in the regulation of cell sensitivity to ATP by regulating the extracellular concentration of ATP. Thus, ectoenzymes CD39 and CD73 sequentially hydrolyze ATP into ADP, AMP, and adenosine. However, the dissociation in ATP-induced cell activities that we observed in Tconvs is probably not linked to the levels of ATPase membrane expression. Indeed B220<sup>+</sup> Tconvs, which express low levels of CD39, are totally resistant to ATP stimulation (23). Moreover, we show herein that the levels of CD39 and CD73 membrane expression increase in parallel with those of P2X7R and maximum levels of ectonucleotidases upregulation are reached on Tconvs displaying the higher levels of ATP-induced pore formation and PS externalization. Likewise, although Tregs express high levels of membrane ectonucleotidases CD39 and CD73 (55), they are highly sensitive to ATP-induced cellular functions (**Table 1**). The ability of Tregs to form high levels of P2X7 membrane pore might favor their regulatory functions by increasing the amount of ATP molecules released in the pericellular space, and their subsequent hydrolysis to the adenosine immunosuppressive molecule by CD39 and CD73 (55).

To summarize, CD45RBlow or CD25<sup>+</sup> Foxp3<sup>+</sup> Tregs show high levels of P2X7R membrane expression and sensitivity to ATP. In contrast, P2X7R-mediated cellular activities in CD4<sup>+</sup> Tconvs are not dependent on the levels of P2X7R membrane expression and not triggered in an all-or-none manner, and depend on their stage of activation/differentiation. Thus, the P2X7R surface expression on CD4<sup>+</sup> Tconvs is in the following order: CD45RBhiCD44lo cells < CD45RBloCD44lo < CD45RBloCD44hi cells. In each of the three subsets, CD69<sup>+</sup> recently activated T cells express very significantly more P2X7R than their CD69<sup>−</sup> counterpart. However, the naive CD69<sup>−</sup>CD45RBhiCD44lo T cells bearing the lowest surface P2X7R yield the highest CD62L shedding response but give a weak PS exposure response and do not have a significant pore formation. Recently activated CD69<sup>+</sup> naive CD45RBhiCD44lo Tconvs show a significant reduction in their ability to proteolytically cleave CD62L compared to inexperienced naive CD69<sup>−</sup> T cells. The reverse situation is observed for ATP-induced pore formation, and to a lesser extent for PS externalization, which are significantly upregulated in recently activated naive CD69<sup>+</sup> Tconvs compared to naive CD69<sup>−</sup>CD45RBhiCD44lo Tconvs. Effector/memory CD69<sup>−</sup>CD45RBloCD44hi T cells show high levels of P2X7R membrane expression and sensitivity to ATP. Recently activated CD69<sup>+</sup> effector/memory CD45RBloCD44hi T lymphocytes with the highest amounts of surface P2X7R respond strongly to ATP by PS exposure and pore formation but yield weaker CD62L shedding compared to their CD69<sup>−</sup> counterpart.

### ETHICS STATEMENT

All the experiments were conducted in accordance with French (décret no. 2013-118) and EU (directive 86/609/EEC) guidelines for the care of laboratory animals and approved by our local research ethics committee (CEEA 59).

## AUTHOR CONTRIBUTIONS

PB, HS, AM, and JL conceived and designed the experiments. HS, AM, JL, SL, and MB performed the experiments. PB, HS, AM, JL, SL, and JK analyzed and interpreted the data. PB and HS wrote the manuscript with the assistance of all authors. JK and CK-L critically reviewed the manuscript.

## FUNDING

This work was funded by grants from ANR—Agence Nationale de la Recherche (ANR-07-BLAN-0089-02 and ANR-13-ISV6-0003). Hanaa Safya was the recipient of a National Council for Scientific Research (CNRS) Lebanon fellowship.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00360/ full#supplementary-material.

Figure S1 | Anti-CD45RB antibodies do not distinguish CD45RB from CD45RABC (B220) isoforms. COS-7 fibroblasts were transfected with a pCDEF3 expression vector containing CD45RABC (B220) cDNA. 48 h later, transfected cells were stained with FITC-conjugated anti-CD45RA, PE-conjugated anti-CD45RB, APC-conjugated anti-CD45RC, and PE Cy5.5-conjugated anti-CD45RABC or isotype controls, and analyzed by flow cytometry. (A) R1 and R2 gates delineate B220-negative and B220+ COS-7 cells, respectively. (B) Expression of CD45RA, CD45RB or CD45RC in COS-7 cells gated on R2. Flow cytometry histograms obtained with anti-CD45RA, anti-CD45RB or anti-CD45RC monoclonal antibody (—) are overlaid on histograms obtained with isotype controls (- - -). At least 20,000 events were analyzed from each sample. Data are representative of three independent experiments.

Figure S2 | Massive increase of CD69 expression on splenic T cells from concanavalin A (ConA)-treated mice. B6 mice were injected i.v. with 7 µg/g of T-cell mitogen ConA or phosphate-buffered saline. They were euthanized 18 h after injection, spleen cells were counted and stained with fluorescent monoclonal antibodies against phenotypic markers CD90, B220, CD4, and CD69 or isotype controls, and analyzed by flow cytometry. At least 20,000 events were analyzed from each sample. Asterisks indicate statistically significant differences between groups (\*\*\**p* ≤ 0.001).

Figure S3 | Representative dot plots and histograms showing sequential gating for analyses of CD62L expression levels on CD45RBhiCD44lo naive and CD45RBloCD44hi effector/memory CD4+ T cells (either CD69− or CD69+). Spleen cells from B6 mice were stained with either fluorescent monoclonal antibodies against phenotypic markers CD90, B220, CD4, CD45RB, CD44, CD62L, and CD69 or isotype controls, and analyzed by flow cytometry. Mice had not been

injected with concanavalin A to maintain a high frequency of naive and memory resting T cells. FSC vs. SSC dot plots were used to select single, viable cells and exclude debris, dead cells and doublets. CD90 vs. B220 dot plot was used to select B220−CD90+ T cells in the FSChi SSClo-gated cell population (Gate 1) and exclude B220+CD90+ T cells. The CD90+B220− T cells were then further gated by the expression of CD4 (Gate 2). The relative expression of CD45RB and CD44 was analyzed on the CD4+ T-cell subset to identify CD45RBhiCD44lo naive cells (Gate 3) and CD45RBloCD44hi effector/memory cells (Gate 4). Naive and memory/effector CD4+ T cells were then identified by the expression of CD69. Finally, the percentages of T cells expressing CD62L were gated on CD69− (Gate 5) and CD69+ (Gate 6) cells. At least 20,000 events were analyzed from each sample.

Figure S4 | Inhibition of CD62L shedding by metalloprotease inhibitor and absence of P2X7R signaling in P2X7KO mice. (A) Spleen cells from B6 mice (*n* = 6 mice/group) were preincubated with 0, 10, 25, 50, 100 µM of metalloprotease inhibitor GM6001 for 15 min at 37°C, and then either left

### REFERENCES


unstimulated or stimulated with 2 mM adenosine-5′-triphosphate (ATP) for 30 min at 37°C. Cells were subsequently stained with monoclonal antibody (mAb) against CD62L and phenotypic markers. Cell surface expression of CD62L was assessed by flow cytometry in the gated CD90+, CD45RBhiCD44lo naive, and CD45RBloCD44hi effector/memory CD4+ T cells. Bars represent the percentages of initial CD62L surface expression after ATP stimulation in the presence or absence of metalloprotease inhibitor. (B) Spleen cells from B6 and P2X7KO mice were either left unstimulated or stimulated with 2 mM ATP for 30 min in the presence or absence of YO-PRO-3 fluorescent probe. Cells were subsequently stained with fluorescent mAbs against phenotypic markers CD90, B220, CD4, CD45RB, CD44, and CD62L as well as Annexin V fluorescent probe. CD62L shedding, pore formation or phosphatidylserine exposure were assessed within the gated CD45RBhiCD44lo T-cell subset by flow cytometry. Flow cytometry histograms display the staining of spleen cells from B6 or P2X7RKO mice with anti-CD62L antibody, YO-PRO-3, or Annexin V probe, left unstimulated (black line) or following ATP stimulation (gray line). Numbers inside each panel correspond to relevant cell percentages in unstimulated/stimulated conditions.


independently of TCR-MHC engagement and is insensitive to the action of Foxp3+ regulatory T cells. *J Immunol* (2008) 180(3):1565–75. doi:10.4049/ jimmunol.180.3.1565


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

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

*James L. Quinn1,2, Gaurav Kumar <sup>2</sup> , Agnieshka Agasing1,2, Rose M. Ko2 and Robert C. Axtell2 \**

*1Department of Microbiology and Immunology, University of Oklahoma Health Sciences Center, Oklahoma City, OK, United States, 2Arthritis and Clinical Immunology Program, Oklahoma Medical Research Foundation, Oklahoma City, OK, United States*

Both T cells and B cells are implicated in the pathology of multiple sclerosis (MS), but how these cells cooperate to drive disease remains unclear. Recent studies using experimental autoimmune encephalomyelitis (EAE) demonstrated that the TH17 pathway is correlated with increased numbers of ectopic B-cell follicles in the central nervous system (CNS). As follicular T helper (TFH) cells are regulators of B cell responses, we sought to examine the role of TFH cells in EAE induced by the transfer of myelin-specific TH17 cells (TH17-EAE). In this study, we first confirmed previous reports that B-cells are a major cell type infiltrating the CNS during TH17-EAE. In addition, we found that B cells contribute to the severity of TH17-EAE. Class-switched B-cells in the CNS were positively correlated with disease and, strikingly, the severity TH17-EAE was diminished in B cell deficient mice. We next focused on the role TFH cells play in TH17-EAE. We found substantial numbers of CXCR5+PD1+CD4+ TFH cells in the CNS tissue of TH17- EAE mice and that at the peak of disease, the number of infiltrating TFHs was correlated with the number of infiltrating B-cells. Using congenic CD45.1+ donor mice and CD45.2<sup>+</sup> recipient mice, we determined that the TFH cells were recipient-derived, whereas IL-17<sup>+</sup> cells were donor-derived. We assessed whether myelin-specific TFH cells are capable of inducing EAE in recipient mice and found that transferring TFH cells failed to induce EAE. Finally, we tested the effects of blocking TFH trafficking in TH17-EAE using an antagonistic antibody against CXCL13, the chemokine ligand for CXCR5 on TFH cells. We found anti-CXCL13 treatment significantly reduced TH17-EAE disease. This treatment blocked CD4+ T cells from entering the CNS, but had no effect on infiltration of B cells. Strikingly, this antibody treatment had no measurable effect on TH17 disease in B cell-deficient mice. These data demonstrate that infiltrating TFH cells are a key cell type that contributes to an inflammatory B cell response in TH17-EAE and provide evidence for targeting TFH cells as a treatment for neuro-autoimmune diseases like MS.

Keywords: experimental autoimmune encephalomyelitis, Th17, TFH, B cells, CXCL13, multiple sclerosis

### INTRODUCTION

Multiple Sclerosis (MS) is a neuro-inflammatory disorder, which results in the infiltration of immune cells into the central nervous system (CNS) and demyelination of neurons (1). CD4<sup>+</sup> T helper (TH) cells play a critical role driving disease in both MS and the mouse model of the disease, experimental autoimmune encephalomyelitis (EAE) (2, 3). In fact, the transfer of myelin-specific T helper cells

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Shiv Pillai, Harvard Medical School, United States Ju Qiu, Shanghai Institutes for Biological Sciences (CAS), China*

> *\*Correspondence: Robert C. Axtell bob-axtell@omrf.org*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 14 November 2017 Accepted: 12 February 2018 Published: 27 February 2018*

### *Citation:*

*Quinn JL, Kumar G, Agasing A, Ko RM and Axtell RC (2018) Role of TFH Cells in Promoting T Helper 17-Induced Neuroinflammation. Front. Immunol. 9:382. doi: 10.3389/fimmu.2018.00382*

**74**

alone is capable of inducing EAE disease in healthy recipient mice. While both TH1 and TH17 cells are capable of inducing disease, each cell type results in a unique disease pathology. EAE induced by the transfer of myelin-specific TH17 cells (TH17-EAE) is characterized by elevated numbers of B cells and neutrophils in the CNS compared to TH1-induced EAE (3, 4). Additionally, TH17-EAE mice develop aggregates of B cells and T cells in the meninges of CNS tissue, also known as ectopic follicles (4).

The recent successes of several B cell-depleting therapies demonstrate B cells also play a critical role in MS and EAE disease progression (5–8). The mechanisms though which B cells lead to disease activity are threefold. First, plasma cells generate autoantibodies (9). Second, B cells produce inflammatory cytokines such as IL-6 (10). Finally, B cells can act as antigen presenting cells to T cells and help maintain an autoreactive T cell response (11). Although previous work has clearly demonstrated that both B cells and helper T cells contribute to disease progression, it is not currently clear if and how these two populations cooperate in MS.

CXCL13 is a chemokine that plays a critical role in the homing and co-localization of B cells and T cells to the follicles of lymphoid organs as well as the formation of ectopic follicles outside of lymphoid organs (12, 13). It has been shown in several disease models that a TH17 signature is associated with elevated levels of CXCL13 in the blood and tissue (14, 15). In EAE mice specifically, CXCL13 is elevated in TH17-EAE mice compared to actively immunized mice (4). Additionally, in MS patients, CXCL13 is significantly elevated in the peripheral blood and found in active lesions of the brain (16–18).

The primary receptor of CXCL13, CXCR5, is found on subsets of both B cells and T cells. The primary T cell subset that expresses CXCR5 is follicular T helper (TFH) cells (19). TFH cells have been well studied in the follicles of secondary lymphoid tissue in the context of infection, where they directly interact with proliferating B cells to aid in germinal center (GC) formation, affinity maturation, and maintenance of memory B cells (20). In MS patients, a recent study has shown a positive correlation between CXCR5<sup>+</sup> TFH cell numbers in blood and increases in disability (21). Additionally, effective Laquinomod treatment of EAE mice was associated with decreased numbers of TFH cells and B cell responses (22). Though these reports show strong correlative data indicating that TFH cells may contribute to disease activity in MS and EAE, there is currently no study that definitively shows the causal effects of TFH cells in this disease.

The purpose of our study is to better understand the role of CXCR5<sup>+</sup> TFH cells in autoimmune CNS inflammation. Our results demonstrate elevated levels of TFH cells in the CNS tissue of TH17-EAE mice. Additionally, we found that donor-derived TH17 cells first infiltrate the CNS and are followed by a second wave of infiltration, which include both TFH cells and B cells. These two populations are positively correlated and this relationship indicates that these TFH cells are promoting a pro-B cell environment within the CNS tissue. This role of TFH cells is further demonstrated by the reduction of TH17-EAE disease following anti-CXCL13 treatment. Additionally, we observed no significant effect on TH17-EAE in B cell-deficient mice treated with CXCL13 antibody. This shows an important role for CXCR5<sup>+</sup> T cells in line with their known ability to support B cell function and begins to answer important questions regarding their contribution to disease progression.

# MATERIALS AND METHODS

### Mice

C57BL/6 (WT, CD45.2<sup>+</sup>), B6.SJL-Ptprca Pepcb/BoyJ (CD45.1<sup>+</sup>), and B6.129S2-*Ighmtm1Cgn*/J (μMT) mice were purchased from Jackson Laboratory and bred in the Oklahoma Medical Research Foundation mouse facility. All animals were treated in compliance with the institutional guidelines. All experiments were performed using 8- to 10-week-old female mice.

# EAE Induction

For adoptive transfer EAE, age- and sex-matched donor mice were subcutaneously immunized with 150 µg MOG35–55 peptide (Genemed Synthesis, Inc.) emulsified in complete Freund's adjuvant (5 mg/ml heat-killed *M. tuberculosis*), followed by an intraperitoneal (IP) injection of 250 ng of *Bordetella pertussis* toxin (List Biological Laboratories, Inc.) in 200 µl of PBS at 0 and 2 days postimmunization. Ten days postimmunization, spleens and lymph nodes were collected and mechanically disrupted to generate a single-cell suspension. For TH17-EAE, the cells were cultured at 2.5 × 106 cells/ml for 72 h and stimulated with 10 µg/ml MOG35–55, 10 ng/ml IL-23, and 10 µg/ml IFN-γ antibody in complete RPMI media (23). For TFH-EAE, cells were cultured with 10 µg/ml MOG35–55, 20 ng/ml IL-6, 20 ng/ml IL-21, 10 µg/ml IFN-γ antibody, 10 µg/ml IL-4 antibody, and 20 µg/ml TGF-β antibody in complete RPMI media as previously described (24). On Day 3, cells were collected and 5 × 106 cultured cells were transferred into healthy recipient mice by IP injection.

Mice were monitored daily for clinical signs. Paralysis was assessed using a standard clinical score ranging from 0 to 5 with scores corresponding to the following phenotypes: 0, no disease; 1, loss of tail tone; 2, partial hind-limb paralysis; 3, complete hindlimb paralysis; 4, forelimb paralysis; and 5, moribund/dead.

# Isolation of CNS-Infiltrating Cells

Cells were isolated from the brainstem, cerebellum, and spinal cords of PBS-perfused mice. CNS homogenates were incubated with 5 µl/mL DNAse (Sigma) and 4 mg/ml collagenase (Roche) at 37°C for 40 min. and purified using a Percoll (GE Healthcare) gradient.

# CXCL13 Antibody Treatment

Anti-mouse CXCL13 and isotype antibodies were provided by Dr. Maurice Zauderer (Vaccinex). Beginning on the day of transfer, mice were treated with 30 mg/kg of the antibodies in phosphate buffer saline, intraperitoneally, twice a week until sacrifice.

# Quantitative Real-time PCR

Following culture, CD4<sup>+</sup> T cells were isolated using a magnetic CD4 negative enrichment kit (Miltenyi Biotec). Total RNA was extracted using the RNeasy Mini Kit (Qiagen) and reversetranscribed into cDNA by iScript cDNA Synthesis Kit (Bio-Rad). Q-PCR was performed using iQ SYBR Green Supermix (Bio-Rad) and expression levels of genes were normalized to a reference gene β-actin. The primer pair for CXCR5 is forward, 5′-ACTCCTTACCACAGTGCACCTT-3′; and reverse, 5′-GGAAACGGGAGGTGAACCA-3′. Primers for BCL6 are forward, 5′-CACACCCGTCCATCATTGAA-3′; and reverse, 5′-TGTCCTCACGGTGCCTTTTT-3′. Primers for IL-17A are forward, 5′-GGCCCTCAGACTACCTCAAC-3′; and reverse, 5′-AGCTTCCCAGATCACAGAGG-3′. Primers for β-actin are forward, 5′-GACGGCCAGGTCATCACTATTG-3′; and reverse, 5′-AGGAAGGCTGGAAAAGAGCC-3′. Naïve control CD4<sup>+</sup> cells were obtained from unimmunized wild-type splenocytes.

### Histology

Spinal cords and brains were fixed in 4% paraformaldehyde in PBS, paraffin embedded, cut, and stained with H&E and Luxol Fast Blue, according to standard protocols. For fluorescent microscopy, mice were perfused with PBS followed by 4% paraformaldehyde. Spinal cords were fixed in 4% paraformaldehyde for 4 h then placed in 20% sucrose for 48 h. Samples were embedded in OCT Compound (Sakura Finetek) and cryosectioned (7 µm) on the coronal plane. Slides were blocked with 5% normal donkey serum for 1 h and stained overnight with anti-B220 (1:400) (BioLegend, clone RA3-6B2) and anti-CD3 (1:250) (Abcam, ab5690). Slides were stained with secondary antibodies AlexaFluor 488 donkey anti-rat (Life Technologies) and AlexaFluor 546 donkey anti-rabbit (Invitrogen) for 1 h and counterstained with DAPI (Life Technologies, ProLong Diamond Antifade Mountant with DAPI). Images were collected using a Ziess LSM-710 Confocal.

### MOG Recall Assay

At the peak of TH17-EAE, spleens were collected and mechanically disrupted to generate a single-cell suspension. The cells were cultured at 2.5 × 106 cells/ml for 72 h and stimulated with 0 or 10 µg/ml MOG35–55 in complete RPMI media. On Day 3, supernatants were collected, and IL-17 was measured by ELISA (eBiosciences).

# Flow Cytometry

The following surface antibodies were used for flow cytometry; αCD4-PECy7 or ef450 (GK1.5), αPD1-ef450 (J43), αGL7-eF450 (GL-7),αIgM-eF450 (II/41), αB220-PE (RA3-6B2), and Streptavidin-PE or PECy7 were purchased from eBiosciences. αPD1- APC (EH12.2H7), αIgD-PE (11-26c.2a), αCD19-PerCPCy5.5 or FITC (6D5), αCD19-PerCPCy5.5 or FITC (6D5), αB220-AF488 (RA3-6B2), αCD45.1-BV711 (A20), αCD45.2-BV605 (104), and αICOS-AF488 (C398.4A) were purchased from BioLegend. αCXCR5-Biotin was purchased from BD Biosciences.

To detect intracellular cytokines, cells were stimulated with PMA/ionomycin along with GolgiStop for 3 h at 37°C in RPMI. They were then permeabilized with Cytofix/Cytoperm (BD Biosciences) and stained with αIL-17A-FITC (TC11-18H10.1), or IL-21R/Fc chimera (R&D Systems) followed by PE-conjugated affinity-purified F(ab′)2 fragment of goat anti-human Fcγ Ab (Jackson ImmunoResearch Laboratories).

All flow cytometry data were acquired on a BD LSRII or BD FACSAriaIIIu and analyzed with FlowJo software.

# Statistics

Data are presented as means ± SEM and statistical significance was determined using a two-tailed Mann–Whitney test, Student's *t*-test, or Kruskal–Wallis test when more than two groups were analyzed. Correlation was assessed by linear regression with calculated *R*<sup>2</sup> and significance. For all data sets, differences were considered statistically significant for *p* < 0.05. All statistical analyses were made using Prism 7 (GraphPad).

# RESULTS

## B Cells Accumulate into the CNS of TH17-EAE Mice and Contribute to Disease Severity

Recent studies have linked the TH17 pathway with the formation of GCs in spleens and ectopic B cell follicles at sites of inflammation (25, 26). Specifically in EAE, it was reported that TH17-EAE mice develop ectopic follicles and GC-like structures in their CNS tissues (4). For our study, we first sought to verify B cell infiltration is a feature of our TH17-EAE model; a model where we expand myelin-specific TH17 cells from MOG35–55-immunized mice with IL-23 and anti-IFNγ, and transfer these donor cells into healthy recipient mice (23). We assessed CNS tissue for B cells at 5, 9, and 15 days post-transfer, which correspond to before disease onset, early disease, and peak disease, respectively. We found an increasing number of B cells infiltrating the spinal cords and brains as EAE disease progressed with the highest numbers of B cells found at the peak of disease (Day 15 post-transfer of TH17 cells) (**Figures 1A,B**). We performed immunohistochemistry on frozen spinal cord sections and observed that B cells cluster in the meningeal areas of the spinal cord in close proximity to CD3<sup>+</sup> T cells (**Figure 1C**). We next characterized the phenotype of the B cells in the CNS of TH17-EAE mice at the peak of disease. CXCR5 expression was significantly decreased on B cells in the CNS tissue compared to B cells in the spleen (**Figure 1D**). In addition, we observed that a greater percentage of the total B cell population (CD19<sup>+</sup>B220<sup>+</sup>) in the spinal cord was class-switched (IgM<sup>−</sup>IgD<sup>−</sup>) and had a GC phenotype (GL7<sup>+</sup>PNA<sup>+</sup>) compared to the B cell population in the spleens (**Figures 1E,F**). Strikingly, we found at the peak of disease that the number of class-switched B cells correlated with increased weight-loss in mice, which is indicative of greater disease severity (**Figure 1G**). These data confirm the previous reports showing that B cell responses occur in the CNS during TH17-EAE and we now show that they are positively correlated with increased disease severity.

Because we observed B cells infiltrating into the CNS in our TH17-EAE model, we next examined whether B cells contribute to disease progression in our disease model. To answer this question, we compared the development of TH17-EAE in wild-type mice (WT) with its development in B cell-deficient mice (μMT). We found that both WT and μMT mice developed disease, however, the μMT mice had significantly lower scores compared to WT mice (**Figures 2A,B**). As expected, μMT mice were almost completely devoid of spinal cord-infiltrating B cells (**Figure 2C**), but we also found that the μMT mice had fewer CD4<sup>+</sup> T cells infiltrating the CNS compared to WT mice (**Figure 2D**).

Figure 1 | B cell responses occur in the central nervous system of TH17-EAE. (A) Representative flow cytometry plots of the percentage of viable B220+CD19<sup>+</sup> B cells in the spinal cords of TH17-experimental autoimmune encephalomyelitis (EAE) at 5, 9, and 15 days post-transfer of TH17 cells. Gated on total viable cells. (B) Percentage of total viable cells and absolute number of B cells infiltrating the spinal cord and brain of TH17-EAE. Statistical significance from Day 5 to Day 15 was determined using Kruskal–Wallis test (\*\**p* < 0.01, \**p* < 0.05). Data are from one experiment representative of two experiments (*N* = 3 mice per group/ experiment). (C) Representative image of a spinal cord section from TH17-EAE stained with anti-CD3 Ab (red), anti-B220 Ab (green), and DAPI (blue). (D) Expression levels of CXCR5 (measured by mean fluorescent intensity) on viable B220+CD19+ B cells in spleen and spinal cord tissue was measured by flow cytometry. Tissues were collected at the peak of disease. Statistical significance was determined using Student's *t*-test (\*\*\*\**p* < 0.0001) and *N* = 5. (E) Representative flow cytometry plots of CD19+IgM−IgD− class-switched B cells in the spleen and spinal cord of mouse with TH17-EAE. Gated on viable B220+CD19+ B cell population. Percentage of class-switched B cells of the total CD19+ B cell population in the spleens and spinal cords of mice with TH17-EAE. Statistical significance from was determined using Student's *t*-test (\*\**p* < 0.01) and *N* = 4. (F) Representative flow cytometry plots of CD19+GL7+PNA+ germinal center B cells in the spleen and spinal cord of mouse with TH17-EAE. Gated on viable CD19+ B cell population in the spleens and spinal cords of mice with TH17-EAE. Statistical significance was determined using Student's *t*-test (\*\**p* < 0.01) and *N* = 4. (G) Correlation between CD19+IgM−IgD− class-switched B cells and decrease in body weight at the peak of disease. Statistically significant correlations were determined using linear regression and *N* = 8.

Taken together, these data demonstrate that B cells are promoting TH17-EAE disease severity and suggest that there is cooperation between B cells and T cells driving this pathology.

## TFH Cells Infiltrate the CNS of TH17-EAE Mice

TFH cells are primarily found in follicles of the spleens and lymph nodes during an immune response and facilitate B cell activity in the GC (20). Because we found that T cells are in close proximity to B cells in the CNS of TH17-EAE (**Figure 1C**), we speculated that these T cells are functioning as TFH cells and are providing help to B cells. Therefore, we assessed the frequency and numbers of CNS-infiltrating T helper cells that co-express the TFH markers CXCR5 and PD1 in mice with TH17-EAE. At the peak of disease, we found that TFH cells comprise 16.2 ± 7.6% of the total CD4<sup>+</sup> T cell population within the spinal cords of mice with TH17-EAE (**Figures 3A,B**). In contrast, we found very few TFH cells in the spleens of these TH17-EAE mice (**Figure 3A**). We next examined the kinetics of TFH cell infiltration into the CNS tissue. Similar to what we observed with B cells (**Figure 1B**), we found that the brain and spinal cord tissue had increases in the TFH cell population as disease developed in these mice (**Figure 3B**). We also characterized other molecules on the infiltrating T helper cells and found that the CXCR5<sup>+</sup>PD1<sup>+</sup>CD4<sup>+</sup> T cells express higher levels of ICOS and the cytokine IL-21, molecules associated with TFH function, compared to CXCR5<sup>−</sup>PD1<sup>−</sup>CD4<sup>+</sup> T cells (**Figures 3C,D**).

We next assessed whether CXCR5+PD1+ TFH cells were correlated with B cell infiltration and activity in the inflamed CNS of mice with TH17-EAE. At the peak of disease, we found that the numbers of B cells within the CNS are positively correlated with the numbers of TFH cells (**Figure 3E**). Furthermore, we found there is a significant positive correlation between the percent of CXCR5<sup>+</sup>PD1<sup>+</sup> TFH cells (in the CD4<sup>+</sup> population) with class-switched B cells (in the total B cell population) in the CNS tissue (**Figure 3F**). Interestingly, this correlation with class-switched B cells was not observed with other CD4<sup>+</sup> T cell subsets (CXCR5<sup>−</sup>PD1<sup>−</sup>CD4<sup>+</sup>) which indicates the correlation between TFH cells and class-switched B cells is not due to a general increase of infiltrating CD4<sup>+</sup> T cells in the CNS. Instead, these correlative data suggest that there is a functional interaction between B cells and TFH cells specifically in the CNS that may promote disease progression in this model of EAE.

# TFH Cells Constitute a Second Wave of CNS-Infiltrating T-Cells

To better understand the development of these infiltrating TFH cells in the CNS tissue, we wanted to first determine their origin. We generated myelin-specific TH17 cells from congenic CD45.1<sup>+</sup> mice and transferred these cells into CD45.2<sup>+</sup> recipient mice. We found that early in disease development (Day 9) the majority of CD4<sup>+</sup> T cells in the CNS tissue are donor-derived CD45.1<sup>+</sup>, but by the peak of disease (Day 15) we found the recipient-derived CD45.2<sup>+</sup> cells comprise the majority of the CD4<sup>+</sup> T cells in the CNS (**Figure 4A**). This shift indicates the transferred cell population traffics into the CNS first and is followed by a second wave of immune cells derived from the recipient mouse. We next determined the phenotype of the donor and recipient cells infiltrating the CNS. At Day 15, we observed that IL-17-producing cells in the CNS were almost entirely donor-derived (**Figure 4B**). Conversely, we found that the majority of the TFH cells in the CNS were derived from the recipient animals (**Figure 4C**). These data demonstrate that TFH cells are not derived from the donor TH17 cell population, but constitute a second wave of infiltrating cells that are derived from the recipient mouse.

# TFH Cells Are Not Effective Inducers of EAE

Our data demonstrate that TFH cells comprise a distinct subset of immune cells in the CNS of mice with TH17-EAE, which are correlated with increased B cell infiltration and disease activity. We next questioned whether TFH cells themselves are capable of migrating into the CNS and inducing EAE without the transferred

Figure 3 | CXCR5+ TFH cells are present in the central nervous system (CNS) of TH17-experimental autoimmune encephalomyelitis (EAE) mice. (A) Representative flow cytometry plots of CXCR5+PD1+CD4+ T cells in the spleens and spinal cords of mice with TH17-EAE mice at Day 15 post-transfer. Gated on the total viable CD4+ cell population. (B) Percentage of CD4+ cells and absolute number of CXCR5+PD1+CD4+ T cells infiltrating the spinal cord and brain of TH17-EAE. Statistical significance from Day 5 to Day 15 was determined using Kruskal–Wallis test (\**p* < 0.05). Data are from one experiment representative of two experiments (*N* = 3 mice per group/experiment). (C) Expression levels of ICOS (measured by mean fluorescent intensity) in CXCR5+PD1+CD4+ T cells and CXCR5−PD1−CD4+ T cells was determined by flow cytometry. Statistical significance was determined using Student's *t*-test (\*\*\**p* < 0.0001) and *N* = 8. (D) Expression levels of intracellular IL-21 (measured by mean fluorescent intensity) in CXCR5+PD1+CD4+ T cells and CXCR5−PD1−CD4+ T cells was determined by flow cytometry. Statistical significance was determined using Student's *t*-test (\*\*\**p* < 0.0001) and *N* = 8. (E) Correlation between TFH and B cell numbers was assessed at the peak of disease. Statistically significant correlations were determined using linear regression. Data are compiled from two experiments (*N* = 10 mice). (F) Correlation between the percent of CD19+B220+IgM−IgD− class-switched B cells (of total CD19+B220+ population) and the percent of CXCR5+PD1+CD4+ T cells and CXCR5−PD1−CD4+ T cells (of total CD4+ population) in the CNS tissues was assessed at the peak of disease. Statistically significant correlations were determined using linear regression. Data are compiled from two experiments (*N* = 10 mice).

TH17 cell activity. We differentiated myelin-specific TFH cells (IL-6, IL-21, anti-IFNγ, anti-IL-4, anti-TGFβ) (24) and myelinspecific TH17 cells (IL-23 and anti-IFNγ) (23) and compared the ability of these cells to infiltrate into the CNS and induce EAE in healthy recipient mice. After *in vitro* culturing, we verified that the TFH-polarizing and TH17-polarizing conditions yielded the expected T cell phenotypes. We found that the TFH cells expressed high levels of BCL6 and CXCR5 compared to naïve T helper cells and TH17 cells. Conversely, the TH17 cells expressed high IL-17A compared naïve T helper cells and TFH cells (**Figure 5A**). Additionally, we measured PD1 expression on the cultured cells and found *in vitro*-derived TFH cells had PD1 levels comparable

autoimmune encephalomyelitis was induced in wild type CD45.2+ mice with TH17 cells derived from CD45.1+ donor mice. (A) At day 9 and day 15 post transfer of TH17 cells, spinal cord infiltrating cells were isolated and assessed for the percentage of CD4+ gated cells that were derived from donor (CD45.1) or recipient (CD45.2) mice. In the spinal cord at day 15 post transfer, the percentage of (B) IL-17+CD4+ T cells and (C) CXCR5+PD1+CD4+ TFH cells that were donor-derived (CD45.1+ gated) or recipient-derived (CD45.2+ gated) was assessed. Data are from one experiment representative of two experiments (*N* = 3 mice per group/experiment).

to *in vivo*-derived TFH cells in the CNS. This was significantly elevated compared to naïve CD4<sup>+</sup> T cells (**Figure 5B**). We then transferred the TFH or TH17 cells into healthy recipient mice. In contrast to the myelin-specific TH17 cells that induced severe EAE by Day 13, we found that the myelin-specific TFH cells did not induce severe disease even by 50 days post-transfer of cells (**Figure 5C**). In addition, we found that TFH cells did not traffic to the CNS of mice receiving TFH cells; however, we did find substantial numbers of TFH cells homing to the spleen of these mice (**Figure 5D**). These data demonstrate that TFH cell alone do not initiate EAE disease but require TH17 cells to gain entry to the CNS to and promote EAE disease severity.

### CXCL13 Antibody Treatment Inhibits EAE

The data described above led us to hypothesize that TFH cells require a TH17 response to gain entry in the CNS, and, once in the CNS, the TFH cells facilitate the inflammatory response and contribute to disease severity. To address this hypothesis, we assessed the clinical effects of blocking TFH cell trafficking in TH17-EAE. CXCL13 is a chemokine which binds to CXCR5 and acts on both

CXCR5+CD4+ T cells from the spinal cord, CD4+ T cells from TFH culture, and naïve CD4+ T cells was determined by flow cytometry. Statistical significance was determined using one-way ANOVA (\*\*\*\**p* < 0.0001) and *N* = 3 samples per group. (C) The cultured TH17 and TFH cells were then transferred into healthy C57BL/6 mice and EAE disease was monitored daily up to 50 days (Note: TH17-EAE mice were sacrificed at Day 13 due to animal welfare concerns). (D) Numbers of TFH cells in the spinal cords and spleens from mice with TFH-EAE and TH17-EAE were assessed by flow cytometry. TFH numbers in the spleens of naïve mice were also measured. Statistical significance was determined using Student's *t*-test (\*\**p* < 0.01) and *N* = 5 mice per group.

TFH cells and B cells to form GCs in the spleen during an immune response (27). Recent studies have shown that CXCL13 contributes to diseases driven by TH17 activity (28, 29). We treated TH17-EAE induced in C57BL/6 mice with anti-CXCL13 or an isotype control twice a week for the duration of the experiment. We observed that the mice treated with anti-CXCL13 had significantly attenuated disease compared to isotype control-treated mice (**Figure 6A**). Histology was performed on spinal cords to confirm the EAE scores was measured (**Figures 6B,C**). Mice treated with an isotype control had large inflammatory demyelinated lesions in the spinal cords, whereas the anti-CXCL13 treated mice had smaller inflammatory lesions with less demyelination.

We next assessed whether the anti-CXCL13 treatment altered the composition of T cell and B cells infiltrating the CNS. We found that the percentage of B cells, out of the total cell population infiltrating the spinal cord, was similar in the anti-CXCL13 treated mice (9.34 ± 1.55%) compared to isotype control-treated mice (8.95 ± 2.65%) (**Figure 6D**). We found decreased percentages of CNS-infiltrating TFH cells within the CD4+ cell population in the anti-CXCL13-treated mice (5.54 ± 3.7%) compared to control mice (15.5 ± 2.6%), with *p*-values nearing statistical significance (*p* = 0.0732) (**Figure 6E**). In addition, we found that the TH17 cells were not affected by anti-CXCL13 treatment. Percentages of IL-17<sup>+</sup> cells of the infiltrating CD44<sup>+</sup>CD4<sup>+</sup> T cell were similar in the anti-CXCL13 treated mice (9.1 ± 4.6%) and control mice (8.6 ± 1.9%)(**Figure 6F**). We also found no change in the MOG-specific IL-17 response in the spleen (**Figure 6G**).

Finally, we determined whether B cells are required for the efficacy of anti-CXCL13 treatment. We induced TH17-EAE in μMT mice and compared the effect of anti-CXCL13 treatment on the severity of EAE. As we observed in untreated mice (**Figure 2**), isotype control-treated μMT mice had less severe clinical scores compared to isotype control-treated WT mice (mean maximum scores for μMT = 1.6 ± 0.40 and

WT = 2.6 ± 0.37; *p* = 0.047). This was further confirmed by histology (**Figures 6B** and **7B**). In contrast to WT mice, however, we observed that anti-CXCL13 treatment had no effect on paralysis, infiltration, or demyelination of μMT mice (**Figures 7A–C**). Similar to wild-type mice, we observed a trend toward a decreased TFH population in anti-CXCL13-treated mice compared to isotype control-treated mice (**Figure 7D**). We also found that anti-CXCL13 treatment had no effects on the TH17 population infiltrating the CNS nor in the MOGspecific IL-17 response in the spleen (**Figures 7E,F**).

The precise mechanism behind the efficacy of the anti-CXCL13 treatment still needs to be fully assessed. As we found that B cells have reduced CXCR5 expression in the inflamed CNS (**Figure 1D**) and that anti-CXCL13 treatment reduced CNSinfiltrating TFH cells but not B cells (**Figures 6D,E**), we speculate that blocking CXCL13 directly disrupts a critical function of TFH cells in promoting an inflammatory B cell response within the CNS of mice with TH17-EAE.

### DISCUSSION

Previous studies have demonstrated a fundamental link between TH17 cells and B cell responses in autoimmune disease (30). It has been shown that a deficiency in IL-17 results in reduced GC formation, lowered antibody secretion, and ineffective B cell chemotaxis in the spleens of autoimmune mice (29). In EAE, the transfer of myelin-specific TH17 cells in mice induces high numbers of CNS-infiltrating B cells and ectopic follicles in the CNS tissue (4). Prior to our current study, it was unclear how TH17 cells drive disease or B cell responses in the CNS of these mice. Our current study suggests that TFH cells cooperate with TH17 cells to induce inflammatory B cell responses in the CNS and increase disease severity. Our data demonstrate that the CXCR5<sup>+</sup> TFH cells in abundance within the CNS of TH17-EAE are highly correlated with B cell activity and severity of disease. Furthermore, we found that inhibition of TFH cells in the CNS with an anti-CXCL13 antibody treatment effectively reduces the severity of TH17-mediated EAE.

One question we addressed was: from what lineage do the TFH cells originate? A recent report revealed that TH17 cells differentiate into TFH cells in the Peyer's patches and provide help for a B cell response in the gut (31). Interestingly, they showed that IL-23 was indispensable for the maintenance of the TH17 phenotype in the gut and that IL-23 blocked their differentiation into TFH cells. The results in our EAE model reinforce this relationship. When we transferred IL-23-stimulated TH17 cells

Figure 7 | CXCL13 antibody treatment is not effective protection against TH17-experimental autoimmune encephalomyelitis (EAE) in B cell-deficient mice. TH17-EAE was induced in μMT mice and treated with either anti-CXCL13 or an isotype control twice a week beginning at Day 0 post-transfer of TH17 cells. (A) Disease score was monitored. Data were compiled from three experiments and statistical significance at each day was determined using a Mann–Whitney test (*N* = 10 total mice per treatment group). Histological analysis of representative spinal cord sections stained with H&E and Luxol Fast Blue in mice treated with (B) control antibody or (C) anti-CXCL13. Representative FACS plots of the percentage of spinal cord-infiltrating (D) CXCR5+PD1+ TFH cell (in the CD4+ cell gate), and (E) IL-17+ cells (in the CD4+CD44+ cell gate) of mice treated with anti-CXCL13 or isotype control. (F) MOG-specific IL-17 responses in the spleen were assessed through ELISA (*N* = 3 mice per group).

to induce TH17-EAE, we found that these donor cells remain IL-17-producing cells in the CNS and do not take on a TFH phenotype. Instead, we found that the CNS-infiltrating TFH cells are derived from the recipient mice and require an inflammatory TH17 response to traffic into the CNS. It remains unclear what factors are driving the differentiation of the TFH cells *in vivo* and we have ongoing efforts to answer this question. However, as we only observe TFH cells within the CNS and not in the blood, spleen, or lymph nodes, we speculate that the environment of the inflamed CNS tissue harbors the factors that drive TFH differentiation.

Our study reinforces the hypothesis that B cell responses play an important role in driving severe TH17-driven autoimmunity and we now show that blocking a B cell response in the TH17-EAE model is beneficial in reducing disease. Yet, studies using EAE induced with active immunization with rodent MOG35–55 have shown that B cells possess both pro- and antiinflammatory effects on disease (10, 32, 33). The seminal report, by Matsushita et al. (33), demonstrated that administration of CD20 antibody treatment before the induction of EAE exacerbated disease, whereas treatment of mice with established paralysis reduced disease. Subsequent studies have identified that IL-10 and IL-35 production are key mechanisms for the regulatory function of B cells and that IL-6 production and antigen presentation are key mechanisms for the inflammatory effects of B-cells in EAE (10, 34–36). Other studies have shown regulatory effects of B cells in EAE models induce by the transfer of myelin-specific TH1-cells (37). As CNSinfiltration of B cells is a feature of TH17-EAE and not TH1- EAE (4), B cells are likely to have apposing effects on disease activity in these two models. Considering our current data in the context of these previous reports, we speculate that the adoptive transfer TH17-EAE model bypasses the initial priming of myelin reactive T cells in the secondary lymphoid tissues. This avoids the regulatory effects B cells have on the early events of EAE induction seen by Matsushita et al. (33) and reveals the inflammatory effects that CNS-infiltrating B cells have on TH17-induced disease.

Recent reports have shown that ectopic follicles in the CNS and TFH cells in the blood are more prominent in secondary progressive MS patients compared to the relapsing-remitting version of this disease (21). Furthermore, CXCR5<sup>+</sup> TFH cell numbers in blood are correlated with increased disability (21). These reports are highly suggestive that TFH cells play a key role in MS, especially during the progressive-debilitating phase of this disease. Our study in mice provides an important first step in understanding the relationship between TH17 cells, TFH cells, and B cells during neuro-inflammatory diseases like MS. These data provide strong evidence that blocking TFH function may be an effective strategy for treating MS, including patients with progressive versions of disease.

### ETHICS STATEMENT

All procedures and methods with animal experimentation were approved by the OMRF IACUC committee.

### AUTHOR CONTRIBUTIONS

JQ, GK, AA, and RK performed experiments and analyzed results; RA and JQ designed the research and wrote the manuscript.

# ACKNOWLEDGMENTS

The authors would like to thank Dr. Maurice Zauderer for providing them with anti-mouse CXCL13 and isotype antibodies.

### REFERENCES


This manuscript was funded by grants awarded to RA from the National Multiple Sclerosis Society (RG-1602-07722) and the NIH (1R01EY027346) and a training grant awarded to JQ from the NIH (T32 AI 7633-15).

onset of disease and severe cortical pathology. *Brain* (2007) 130:1089–104. doi:10.1093/brain/awm038


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

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

# Fresh evidence for Platelets as neuronal and innate immune Cells: Their Role in the Activation, Differentiation, and Deactivation of Th1, Th17, and Tregs during Tissue inflammation

### *Eugene D. Ponomarev\**

*Faculty of Medicine, School of Biomedical Sciences, The Chinese University of Hong Kong, Shatin, Hong Kong*

Recent studies suggest that in addition to their common function in the regulation of thrombosis and hemostasis, platelets also contribute to tissue inflammation affecting adaptive immunity. Platelets have a number of pro-inflammatory and regulatory mediators stored in their α-granules and dense granules, which are promptly released at sites of inflammation or tissue injury. Platelet-derived mediators include cytokines (IL-1α, IL-1β, and TGFβ1), chemokines (CXCL4 and CCL3), immunomodulatory neurotransmitters (serotonin, dopamine, epinephrine, histamine, and GABA), and other low-molecular-weight mediators. In addition, activated platelets synthesize a number of lipid pro-inflammatory mediators such as platelet-activating factor and prostaglandins/ thromboxanes. Notably, platelets express multiple toll-like receptors and MHC class I on their surface and store IgG in their α-granules. Platelet-derived factors are highly effective in directly or indirectly modulating the priming and effector function of various subsets of T cells. Besides secreting soluble factors, activated platelets upregulate a number of integrins, adhesion molecules, and lectins, leading to the formation of platelet–T cells aggregates. Activated platelets are able to instantly release neurotransmitters acting similar to neuronal presynaptic terminals, affecting CD4 T cells and other cells in close contact with them. The formation of platelet–T cell aggregates modulates the functions of T cells *via* direct cell–cell contact interactions and the local release of soluble factors including neurotransmitters. New data suggest an important role for platelets as neuronal and innate-like cells that directly recognize damage- or pathogen- associated molecular patterns and instantly communicate with T cells.

Keywords: platelets, inflammation, CD4 T cells, glycolipids, damage-associated molecular pattern, neurotransmitter, autoimmunity

### INTRODUCTION

Platelets are small non-nucleated cells around 2–3 μm in diameter, produced by megakaryocytes in the bone marrow by a budding process (1). Platelets have developed a secretory machinery with multiple storage vesicles, mitochondria, mRNA, ribosomes, signal transduction pathways, and multiple receptors acting in many instances as large nucleated cells (2). Most importantly, in the peripheral blood, platelets outnumber mononuclear cells by almost 100-fold and constitute the most abundant

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Manu Rangachari, Laval University, Canada Silvia Deaglio, Università degli Studi di Torino, Italy*

> *\*Correspondence: Eugene D. Ponomarev eponomarev@cuhk.edu.hk*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 04 December 2017 Accepted: 14 February 2018 Published: 02 March 2018*

### *Citation:*

*Ponomarev ED (2018) Fresh Evidence for Platelets as Neuronal and Innate Immune Cells: Their Role in the Activation, Differentiation, and Deactivation of Th1, Th17, and Tregs during Tissue Inflammation. Front. Immunol. 9:406. doi: 10.3389/fimmu.2018.00406*

population of circulating cells after erythrocytes (1, 2). Platelets are known as a cell type that immediately reacts to damage to the blood vessels and participates in blood clot formation, but their role in the regulation of immune cells such as CD4 T cells is still largely underestimated (1–4).

Recent research suggests that platelets participate in inflammation by producing a number of pro-inflammatory mediators (1, 5). Platelets store many mediators in their vesicles (granules), which are swiftly released in a manner similar to the release of neurotransmitters from neuronal presynaptic synapses (2, 6, 7). The molecular mechanisms of the calcium-dependent activation of the secretory machinery and the fusion of vesicles with plasma membrane *via* specific docking molecules (e.g., SNAREs, VAMPs, Syntaxins) are very similar for platelets and neuronal cells and aim to release a number of neurotransmitters from platelets with the most abundant monoamine serotonin, followed by the other biogenic amines epinephrine, dopamine, and histamine (6, 8–11). Platelets also have inhibitory neurotransmitter GABA, but at lower concentrations than in biogenic amines (12). Similar to postsynaptic neurons, immune cells, including CD4 T cells, have multiple receptors for neurotransmitters (e.g., serotonin, dopamine receptors), which provide a direct path by which platelets can instantly communicate with CD4 T cells (13, 14). Similar to neuronal synapses, platelets and T cells are capable of making direct contact with other cells such as antigen-presenting cells (immunological synapses) *via* a number of specific adhesion molecules and integrins (1, 15–17). Certain adhesion molecules (e.g., NCAM or CD56) are expressed in both neurons and subsets of activated T cells, while other adhesion molecules (e.g., ALCAM or CD166) are expressed in neuronal cells, T cells, and platelets, and have a high level of structural homology with NCAM (17–21) (**Figure 1**).

Besides platelet-derived neurotransmitters (serotonin, dopamine, epinephrine, histamine, and GABA), there are other mediators that are either released as soluble factors or appear on the plasma membrane of activated platelets as receptors that directly affect CD4 T cells. These factors include cytokines, chemokines, and potent lipid mediators such as platelet-activating factor (PAF) and thromboxane A2 (2, 22). Activated platelets also release IgGs, which are stored in their α-granules (23). Finally, platelets have a large number of integrins, adhesion molecules, and lectins, which are located inside the granules and are recruited to the platelet plasma membrane when the granules fuse with the plasma membrane (e.g., CD62P) (**Table 1**) (6, 15, 16). Adhesion molecules play an important role in the formation of platelet–T cell contacts, in a manner similar to the formation of neuronal synapses (17) (**Figure 1**). Although it is known that platelets release multiple soluble factors and upregulate multiple integrins and adhesion molecules during their activation, it is still not clear which activating stimuli are responsible for the release of proper factor and/ or proper surface receptor. It is also not clear how specific is the action of single platelet-derived factor on the proliferation and differentiation of various subsets of CD4 T cells. In an attempt to resolve these questions, we take the opportunity in this review to draw attention to some recently discovered pathways of platelet activation in response to tissue damage and discuss the outcomes of each particular pathway for the most common types of CD4 T cells: Th1, Th2, Th17, and Tregs.

Figure 1 | Communication of platelets with CD4 T cells has many similarities with the interaction of presynaptic and postsynaptic neurons. The process of platelet degranulation is very similar to the process of the release of neurotransmitters by presynaptic neurons. In both presynaptic neurons and platelets, neurotransmitters (e.g., serotonin, dopamine), and other mediators are stored in specific vesicles inside the cells. During the process of neuronal or platelet activation, specific vesicles are fused with the surface membrane (using the same docking molecules for platelets and neurons such as VAMP and SNARE), and the vesicle content is released. Both CD4 T cells and postsynaptic neurons have detergent-resistant membrane domains (lipid rafts) with neurotransmitter receptors (e.g., serotonin, dopamine receptors) that promote the further activation of postsynaptic neuron or T cells when stimulated. Both neuronal and platelet–T cell synapses are stabilized with adhesion molecules such as ALCAM, NCAM, and various integrins. ACLAM adhesion molecules and integrins are expressed by neurons, platelets, and activated T cells, and NCAM is expressed by neurons and subsets of activated T cells. During inflammation, platelets are able to directly interact with postsynaptic neurons or activate T cells recognizing specific glycolipids (sialylated gangliosides) and glycoproteins (ALCAM, NCAM) within lipid rafts *via* specific receptors (CD62P, Siglecs, CLRs). AChRs, acetylcholine receptors; CLRs, C-type lectin receptors; DA, dopamine; DARs, dopamine receptors; GluRs, glutamate receptors; HRs, histamine receptors; β2ARs, β2-adrenoreceptors; 5HT, serotonin; 5HTRs, serotonin receptors.

## PLATELET RECOGNITION OF DAMAGE-ASSOCIATED MOLECULAR PATTERN (DAMP) SIGNALS AND DIRECT COMMUNICATION WITH CD4 T CELLS

The classic view of the initiation of inflammation and adaptive immune response holds that the sensing by innate immune cells (macrophages, dendritic cells) of pathogen or initial tissue


Table 1 | Platelet-derived soluble factors and surface molecules that affect proliferation, differentiation, and migration of Th1, Th17, Th2, and Tregs.

damage is followed by the stimulation of adaptive immunity by the CD4 T cells, which in turn help to activate CD8 T cells and B cells. It is commonly agreed that innate immune cells become activated by pathogens through specific receptors such as toll-like receptors (TLRs), C-type lectin receptors (CLRs), and NOD-like receptors, but it is much less clear how these innate cells sense initial damage during sterile inflammation in the absence of infection (37). Several candidate molecules such as HMGB1 and HSPs [both of which are released by apoptotic and necrotic cells and activated platelets at the site of injury (2, 38)] have been proposed recently as activating ligands for TLRs on classic innate cells such as dendritic cells or macrophages or CD8 T cells, but many other pathways remain to be discovered (37). In this respect, platelets are well known as cells that sense initial damage to blood vessels within seconds, in the absence of any infection. Indeed, they recognize highly glycosylated structures of extracellular matrix (ECM) that become exposed to platelets when blood vessel endothelial cells become damaged. These structures include collagen, von Willebrand factor, laminin, and fibronectin, among others (7). However, the exposure of glycosylated structures of ECM to blood-derived cells is an integral part of any type tissue inflammation, when the blood vessel walls become permeable for platelets along with other blood-derived cells (e.g., leukocytes) and serum proteins (e.g., immunoglobulins) (39).

In addition to glycosylated components of ECM that are recognized by platelets in the tissues, we recently found that platelets recognize specific glycolipids (sialylated gangliosides) that are present in the detergent-resistant rigid membrane domains [also referred to as lipid rafts (40)] of postsynaptic neurons (**Figure 1**) and on astroglial cells that comprise the blood–brain barrier (21). Similar glycolipid-rich lipid rafts are also present on the surface of activated T cells in the area of immunological synapse, when T cells interact with antigen-presenting cells or platelets (41) (**Figure 1**).

Gangliosides are also present in the blood vessels (GD3), and certain organs including the brain (GM1, GD3, GT1b, and GQ), pancreas, testis, lungs, and heart are enriched with sialylated gangliosides (21). CD4 T and CD8 T cells are known to require certain gangliosides for their activation and for the assembly of lipid rafts in T-cell receptor (TCR) area during their activation (41). Particularly, it has been shown that CD4 T cells require GM1a and CD8 T cells require GM1b for their activation (42). We found that platelets recognize ganglioside GM1, GT1b, and GQ, which occur abundantly in neuronal cells, particularly in lipid rafts of postsynaptic neurons (21) (**Figure 1**). Most strikingly, the intravenous injection of biochemically isolated neuronal lipid rafts caused massive platelet activation and degranulation in mice, leading to anaphylactic shock (21). The exact mechanisms of this recognition of ganglioside-rich lipid rafts by platelets remains unclear, but the process evidently involves several receptors on platelets, including CD62P and possibly a number of lectins (e.g., Siglec-H, Siglec-15, CLEC-2) (21). An attractive hypothesis is that platelets can also recognize specific gangliosides on the surface of lipid rafts of activated T cells similar to those of neurons *via* specific lectins and highly glycosylated adhesion molecules such as ALCAM and/or NCAM (17) (**Table 1**). In support of this hypothesis, our group demonstrated that *st3gal5*-deficient mice that lacked several sialylated gangliosides such as GM1, GT1b, and GQ do not develop Th1- and Th17-mediated experimental autoimmune encephalomyelitis (EAE) (21). Two studies (one of which was by ourselves) have also recently demonstrated that mice with depleted platelets do not develop severe EAE (21, 43). Therefore, recent findings suggest that platelets are able to sense initial tissue damage-associated signals such as sialylated gangliosides in the lipid rafts within damaged tissues. As mentioned earlier, ganglioside-rich lipid rafts are present in tissue stromal and infiltrating activated immune cells such as T cells. Therefore, the evidence suggests the presence of an important link between the sensing of initial tissue damage by platelets in the absence of infection and the proliferation and differentiation of CD4 T cells at the site of injury.

# PLATELET RECOGNITION OF PATHOGEN-ASSOCIATED MOLECULAR PATTERN (PAMP) SIGNALS

Besides being able to sense tissue damage, platelets also have a number of the receptors that directly interact with pathogens. These receptors include TLRs (TLR2, TLR3, TLR4, TLR7, TLR9), Siglecs (Siglec-H, Siglec-7, Siglec-15), and CLRs (CLEC-2). The role of TLRs, CLEC-2, and Siglecs on platelets has recently been reviewed (4). Platelets do not have MHC class II, but they express MHC class I along with the co-stimulatory molecules CD40 and CD86, and are capable of presenting antigen to CD8 T cells (44). In this review, we focus on the role of these interactions for the stimulation of CD4 T cells. Our data strongly suggest that in myelin oligodendrocyte glycoprotein (MOG)-TCR transgenic mice immunized with MOG35-55 peptide with complete Freund's adjuvant (CFA), the proliferation of MOG-specific CD4 T cells was greatly decreased when platelets were depleted (17). Moreover, the production of IFNγ and IL-17 by MOG-specific CD4 T cells was also significantly decreased in immunized mice with depleted platelets (17). This indicates an important role for platelets during T cell priming *in vivo* when mice are immunized with antigen (MOG) with CFA. We believe that a similar result will be observed not only for MOG but also for most common antigens. Indeed, mice with depleted platelets displayed an elevated level of bacterial load during infection, suggesting a poor immune response against pathogens (3). Our interpretation of these results is that platelets become activated by CFA or pathogens (bacteria) and produce a number of pro-inflammatory factors that support the proliferation and differentiation of CD4 T cells toward Th1 and Th17 at the site of immunization or infection. Similar results were found for induction of EAE in mice with depleted platelets, when disease was significantly diminished (21). To induce EAE, C57BL/6 mice were immunized with MOG35-55/ CFA, with subsequent administration of *pertussis toxin*, at day

0 and day 2. Without *pertussis toxin*, EAE could not be induced in MOG35-55/CFA-immunized mice. However, we managed to induce EAE in the absence of *pertussis toxin* when we administered intravenously biochemically isolated neuronal lipid rafts, which caused massive platelet activation and degranulation (21). Thus, platelet activation became an effective substitute for the action of *pertussis toxin,* which is known to activate T cells and increase the permeability of blood vessels (45). We concluded that platelet activation and degranulation were important events for the stimulation of autoimmune Th1 and Th17 cells in an EAE model.

The precise nature of the reaction by platelets to CFA and/ or *pertussis toxin* remains unclear, but we believe that TLRs (e.g., TLR4) and Siglecs (e.g., Siglec-H) are involved in this process. CELC-2 could be also involved, since it has been reported that this activating receptor directly binds to HIV (4). All these receptors on platelets interact with mycobacteria in CFA. Interestingly, we found that the administration of biochemically isolated neuronal lipid rafts caused massive platelet degranulation and the complete fragmentation of the platelets into platelet-derived microparticles (PMPs), leading to thrombocytopenia (21). The formation of PMPs and the presence of thrombocytopenia is a hallmark of many infections and autoimmune diseases such as hemorrhagic fever (46) or multiple sclerosis (17). Moreover, PMPs have PAF on their surfaces (22), which have the ability to stimulate Th17 cells (17) (**Table 1**).

Another possibility is that massive platelet activation induced by lipid rafts acts as a systemic pro-inflammatory signal inducing the activation of innate cells through the activation of the inflammasomes and other pro-inflammatory pathways. In this hypothesis, platelets would be able to recognize DAMP and PAMP signals along with glycolipids in the lipid rafts and to directly stimulate CD4 T cells. This stimulation is important for the development of the normal immune repose to pathogens and for the development of autoimmune diseases such as multiple sclerosis. We believe that similar mechanisms of stimulation of Th1 and Th17 cells occur in several other types of Th1/17 mediated autoimmune diseases.

# MODULATION OF FUNCTION OF Th1, Th17, AND Th2 CELLS BY PLATELETS

To better understand the role of platelet-derived factors in the modulation of the functions of CD4 Cells, several research groups have performed co-culture (co-incubation) experiments on platelets with T cells. Several studies have indicated an important role for platelet-derived serotonin/5HT in the functions of T cells *in vitro* and *in vivo* (13, 47). Our own studies with polyclonally stimulated human CD4 T cells with anti-CD3/CD28 mAbs indicate that 5HT stimulates proliferation and IFNγ production by CD4 T cells and has no effect on Th17 cells (17). At the same time, platelet-derived PAF and CXCL4 stimulate the differentiation of Th17 cells (17) (**Table 1**). We have demonstrated that the co-incubation of platelets with CD4 T cells did not increase the number of Th2 cells. We even found a tendency for a reduction in the percentages of Th2, demonstrating that platelets skew the balance toward Th1/Th17 (17). A study with a higher number of added platelets demonstrated that platelet-derived factors CXCL4 and CCL5 also enhanced Th1 and Th17, while producing little or no effect on Th2 cells (28). Taken together, these studies suggest that 5HT, CXCL4, CCL5, and PAF play an important role in the stimulation of Th1 and T17 cells, as summarized in **Table 1**. Furthermore, platelets also produce CXCL1, which has been reported to stimulate Th17 cells, and may enhance or substitute for the effect of other platelet-derived factors in the stimulation of Th17 cells (**Table 1**). Among other platelet-derived factors, epinephrine has been shown to stimulate Th17 and Th2 cells. However, the stimulation of Th2 cells was indirect and was derived through the effect on antigen-presenting cells (**Table 1**). In other words, most platelet-derived factors stimulated Th1 and Th17 cells but not Th2 cells, thereby skewing the balance toward Th1 and Th17.

# MODULATION OF THE FUNCTION OF Tregs BY PLATELETS

In our studies utilizing a physiological platelet to CD4 T cell ratio of 15 to 1 or less, we did not find stimulation of Tregs in our co-culture experiments (17). Indeed, we identified a trend for a reduction in the percentage of Tregs, much as we found for Th2 cells. However, when a high number of platelets (~100 to 1) were involved, we found that the percentage of Tregs actually increased (unpublished). Our findings were confirmed by the results of an earlier study, involving the stimulation of Tregs at a high platelet to CD4 T cell ratio (250 to 1), where the effect of the stimulation of IL-10-producing Tregs was blocked by antipan-TGFβ polyclonal antibodies (28). These studies suggested that platelet-derived TGFβ1 (**Table 1**) is capable, under certain conditions, of stimulating Tregs. A recent study has confirmed that platelets stimulate the formation of Tregs *in vivo* in the case of tumor-infiltrating CD4 T cells (48). Thus, platelet-derived TGFβ1 is able to stimulate Tregs, but only when platelets are present in high numbers. At first sight, this finding seems at odds with the fact that TGFβ1 is found in platelet α-granules in relatively high concentrations (49). The most likely explanation is that TGFβ1 is secreted by platelets into blood serum (and probably transported from serum into platelets) where this cytokine is mostly present in its biologically inactive form (50). Therefore, platelet-derived TGFβ1 is mostly inactive and should be activated by additional pathways. At the same time, platelets have the ability to secrete pro- and anti-inflammatory mediators that produce an opposite outcome for CD4 T cells (**Table 1**). The most critical question here is how selectively the process of secretion of anti-inflammatory (e.g., TGFβ1) vs. pro-inflammatory (e.g. CXCL4) agents is regulated *in vivo* during various pathologies. An attractive hypothesis is that various α-granules could contain different substances, which are selectively released upon specific types of stimulation. In support of this hypothesis, it has been shown that one α-granule contains fibrinogen, and another—vWF (51). Thus, there is a possibility that TGFβ1 is located in other granules than CXCL4. Future research should allow this uncertainty to be resolved.

# FORMATION OF PLATELET–T CELLS AGGREGATES AND MODULATION OF FUNCTION OF T CELLS

Besides stimulating CD4 T cells *via* soluble factors, platelets also form direct contacts with T cells, leading to their stimulation or inhibition. The role of CD40 on platelets that interact with CD40L on T cells has been shown to be important for stimulation for CD4 T cells (**Table 1**). However, several recent studies have demonstrated that platelets bind to CD4 T cells *via* multiple receptors, including integrins (αIIbβ3), P-selectin (CD62P), and possibly other lectins such as Siglec-H, Siglec-7, and CLEC-2 (**Table 1**). All of these receptors are able to bind to glycolipids and glycoproteins within lipid rafts, perturbing the interaction of CD4 T cells with antigen-presenting cells (17, 21). Several studies (including one by ourselves) have demonstrated that the formation of platelet–-CD4 T cell aggregates results in the downmodulation of proliferation and differentiation toward Th1 and Th17 (16, 17). We found that during the advanced stages of human autoimmune disease multiple sclerosis, platelets become exhausted in the content of soluble factors in their granules, but upregulated integrins, lectins, and adhesion molecules effectively bind CD4 T cells, perturbing their interaction with antigenpresenting cells or endothelial cells. This eventually inhibited T cell activation and migration to the site of tissue inflammation in a mouse EAE model (17). Similar results have also been recently found for rheumatoid arthritis patients (16). Thus, the formation of platelet-CD4 T cell aggregates mostly inhibited Th1 and Th17 cells during the resolution of tissue inflammation.

# CONCLUDING REMARKS

Recent studies have shown that platelets have two unique properties (1). They directly sense damage- and pathogen-associated molecular patterns (2). They swiftly react and directly communicate with CD4 T cells. Importantly, platelets also play a dual role in the regulation of tissue inflammation (1). During the initiation of inflammation, they produce a number of soluble substances that stimulate proliferation, differentiation, and migration Th1 and Th17 cells (2). During the resolution of inflammation, platelets inhibit proliferation and Th1 and Th17 differentiation by forming platelet–T cell aggregates. Interestingly, platelets act in many instances much like neuronal cells, and platelet-–T cell interactions represent a good model for "neuroimmunological" synapse and direct interaction between the nervous and immune systems. These findings should lead to the opening of new pathways for targeting platelets and either stimulating the immune response during infection, vaccination, or antitumor immunotherapy or inhibiting it during chronic autoimmune diseases such as rheumatoid arthritis, type I diabetes, and multiple sclerosis.

# AUTHOR CONTRIBUTIONS

EP planned and wrote whole manuscript and prepared table and figure.

# FUNDING

The work was supported by a Health and Medical Research Fund grant from the Hong Kong Government's Department of Health

## REFERENCES


(reference no. 02130636), a Research Grant Council-General Research Fund grant (reference no. 14113316; Hong Kong), and a Research Grant Council-Early Career Scheme Fund grant (reference no. 24100314; Hong Kong).


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

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

### *Rupesh K. Srivastava1,2\*, Hamid Y. Dar1 and Pradyumna K. Mishra3*

*1Department of Zoology, School of Biological Sciences, Dr. Hari Singh Gour University, Sagar, India, 2Department of Biotechnology, All India Institute of Medical Sciences (AIIMS), New Delhi, India, 3Department of Molecular Biology, ICMR-National Institute for Research in Environmental Health, Bhopal, India*

The role of immune system in various bone pathologies, such as osteoporosis, osteoarthritis, and rheumatoid arthritis is now well established. This had led to the emergence of a modern field of systems biology called as osteoimmunology, an integrated research between fields of immunology and bone biology under one umbrella. Osteoporosis is one of the most common inflammatory bone loss condition with more than 200 million individuals affected worldwide. T helper (Th) cells along with various other immune cells are major players involved in bone homeostasis. In the present review, we specifically discuss the role of various defined T lymphocyte subsets (Th cells comprising Th1, Th2, Th9, Th17, Th22, regulatory T cells, follicular helper T cells, natural killer T cells, γδ T cells, and CD8+ T cells) in the pathophysiology of osteoporosis. The study of the specific role of immune system in osteoporosis has now been proposed by our group as "immunoporosis: the immunology of osteoporosis" with special emphasis on the role of various subsets of T lymphocytes. The establishment of this new field had been need of the hour due to the emergence of novel roles of various T cell lymphocytes in accelerated bone loss observed during osteoporosis. Activated T cells either directly or indirectly through the secretion of various cytokines and factors modulate bone health and thereby regulate bone remodeling. Several studies have summarized the role of inflammation in pathogenesis of osteoporosis but very few reports had delineated the precise role of various T cell subsets in the pathobiology of osteoporosis. The present review thus for the first time clearly highlights and summarizes the role of various T lymphocytes in the development and pathophysiology of osteoporosis, giving birth to a new field of biology termed as "immunoporosis". This novel field will thus provide an overview of the nexus between the cellular components of both bone and immune systems, responsible for the observed bone loss in osteoporosis. A molecular insight into the upcoming and novel field of immunoporosis would thus leads to development of innovative approaches for the prevention and treatment of osteoporosis.

### Keywords: immunoporosis, osteoimmunology, bone loss, T lymphocytes, osteoporosis

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Thomas Ciucci, National Cancer Institute (NIH), United States Xing Chang, Shanghai Institutes for Biological Sciences (CAS), China*

*\*Correspondence: Rupesh K. Srivastava rksrivastava@aiims.edu,* 

*rupesh\_srivastava13@yahoo.co.in*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 05 December 2017 Accepted: 16 March 2018 Published: 05 April 2018*

### *Citation:*

*Srivastava RK, Dar HY and Mishra PK (2018) Immunoporosis: Immunology of Osteoporosis—Role of T Cells. Front. Immunol. 9:657. doi: 10.3389/fimmu.2018.00657*

**Abbreviations:** IFN, interferon; MCSF, macrophage colony-stimulating factor; NF-κB, nuclear factor κB; BM, bone marrow; TFH, follicular helper T; NKT cells, natural killer T cells; OPG, osteoprotegerin; TRAF, TNF receptor-associated factor; TRAF-6, TNF receptor-associated factor 6; RANK, receptor activator of NF-κB; RANKL, RANK ligand; MCP-1, monocyte chemoattractant protein-1; TGF-β, transforming growth factor beta; BMD, bone mineral density; TCR, T cell receptor; Th, T-helper; DC, dendritic cell; RA, rheumatoid arthritis; OA, Osteoarthritis.

# INTRODUCTION

Osteoporosis leads to enhanced rate of fractures and fragility of bones observed in both men and women. It has been estimated that more than 50% of women and 30% of men over the age of 50 years are susceptible for such fractures and bone loss (1). One-third of females and one-fourth of males will be suffering from osteoporosis leading to significant rise in mortality (20–30% associated with first hip fracture) and morbidity (2). Osteoporosis accounts for more than nine million of fractures annually (2). According to the latest International Osteoporosis Foundation report (3), it is estimated that by 2040 the number of osteoporotic patients above age of 50 years will double worldwide from that of 2010 figures of 158 million (3, 4). It has been estimated that by the end of 2025 the economic burden of osteoporosis will reach \$25.3 billion in the USA alone (5, 6).

Due to their common developmental niche, both the bone and immune systems work as a close knit functional unit (osteoimmune system), thereby leading to permanent interactions at various anatomical and vascular sites (7). This interaction of immune and skeletal system has now made it clear to the scientific fraternity that there do exist a nexus between these duo-systems. This intricate relationship between bone and immune systems has been fascinating scientists since the early 1970s, paving path for birth of a dedicated field of modern biology called as "osteoimmunology" (8). Dysregulation of immune system has already been related with initiation of different inflammatory autoimmune diseases leading to adverse effects on bone integrity (9). This affects the bone either in a localized way as in case of rheumatoid arthritis (RA) or *via* modulating bone metabolism which regulates key bone cell activities including differentiation. In other cases, immune cells induce changes in key factors or functional components of bone mass regulators, thereby affecting bone health. However, still the interaction between bone and immune system which is not unidirectional is largely unexplored. Indeed, during the recent past it has been observed in various studies that T lymphocytes play an important role in the process of bone remodeling (10).

Bone remodeling is a dynamic equilibrium occurring as a result of interaction between bone cells and bone marrow (BM) cells. Therefore, the lymphocytes residing within the BM form an important component for such process to occur. T cells which account for ~5% of total BM cells are found efficiently in both stromal and parenchymal parts of BM (11). T cells are represented by both CD4<sup>+</sup> T and CD8<sup>+</sup> T cell populations. CD4<sup>+</sup> T cells have a vital role in the function and maintenance of the immune system by helping B cells to enhance production of antibodies along with orchestrating CD8<sup>+</sup> T cells and other immune cell functions (12). Naive CD4<sup>+</sup> T cells differentiate into Th1, Th2, Th9, Th17, Th22, regulatory T (Treg) and follicular helper T (TFH) depending upon their respective environmental stimuli (13–16). Th17 cells are primarily responsible for initiating and stimulating bone resorption (osteoclastogenesis) (17, 18), while Treg cells are peculiarly associated with inhibition of bone resorption (18–21). Strikingly, not all T cells are osteoclastogenic, as CD8<sup>+</sup> T cells have recently been reported with bone protecting functions, thereby inhibiting bone loss. CD8+ T cells inhibit the process of osteoclastogenesis *via* secretion of various soluble factors, such as osteoprotegerin (OPG) (18) and interferon (IFN)-γ for regulating bone mass (22). Also, several studies have postulated that T cells may simultaneously function as an activator of bone formation (osteoblastogenesis), as they are associated with activation of Wnt signaling pathway in osteoblastic cells (18). In the present review, we will specially focus on the role of various subsets of T lymphocytes, their plasticity, and related unraveled opportunities for future clinical implications in various bone pathologies, with special emphasis on osteoporosis, i.e., "immunoporosis".

### BONE CELLS

Bone, a dynamic organ undergoes continuous remodeling throughout the life of an organism. This task of bone remodeling is meticulously achieved *via* the coordinated synergism between three different types of bone cells, *viz.* osteoclasts (bone eating cells), osteoblasts (bone forming cells) and osteocytes (bone deposition and resorption cells). Osteoblasts originate from mesenchymal stem cells (MSCs) in the BM which also gives rise to chondrocytes, myocytes and adipocytes. Runt-related transcription factor 2 and its target gene the Sp7 transcription factor (known as osterix) are primarily responsible for differentiation of MSCs into osteoblasts (23). In addition, Wnt signaling also plays an important role in the differentiation of osteoblasts (24). Osteocytes are produced as a result of matrix calcification of osteoblasts under the influence of enzyme alkaline phosphatase. Any mechanical strain on the bones is progressed by osteocytes by sending signals through cellular processes (canaliculi's) to interconnecting osteocytes and also to osteoblasts on surface of bones (25). Osteoclasts are multi-nuclear bone cells which are primarily responsible for bone resorption. They get differentiated from monocytic cell lineages like dendritic cells, macrophages, granulocytes, and microglia. Macrophage colony-stimulating factor (MCSF) is an essential factor required for proliferation and survival of osteoclasts, along with RANK ligand (RANKL) acting *via* its coupling molecule TNF receptor-associated factor 6 (TRAF-6) which leads to their final induction and differentiation (23). For maintaining bone integrity, a dynamic equilibrium is essential between bone forming osteoblasts and bone resorbing osteoclasts. Osteal macrophages (Osteomacs) on the other hand represent a special population of macrophage residing in bony tissues. The term "Osteomacs" was given by Australian researcher Allison Pettit. These are stellate shaped cells and are approximately one-sixth of the cells found in BM, giving rise to a complex networking system (26). They usually get originated from CD68<sup>+</sup> cell types of macrophage origin (27). Osteomacs are responsible for full functional differentiation and mineralization of osteoblasts during *in vitro* cultures and forms canopy at the site of bone remodeling during *in vivo* conditions. It has been observed that any reduction/alteration of macrophages leads to total loss of endosteal osteomacs and respective osteoblasts, thereby concluding that osteomacs have an important role in maintenance of osteoblast maturity (12, 26, 28–30).

### BONE REMODELING

Bone remodeling is a dynamic equilibrium resultant of various physiological/mechanical stress accompanied with differential functional adaptations of the bone (31). This dynamicity results in multi-mode interactions between bone resorbing osteoclasts and bone forming osteoblasts (32, 33). Remodeling leads to restoration of bone micro-damages and bone integrity *via* the balancing act in the release of calcium and phosphorus in host. In fact, remodeling corroborates important relationship between bone formation (osteoblastogenesis) and bone resorption (osteoclastogenesis) (**Figure 1**), regulated at various phases due to impact of immune system on bone cells, neuro-endocrine relationship with bone or a direct interface of osteoclasts and osteoblasts. Bone remodeling comprises of following fourstep, *viz.* activation, resorption, reversal and formation (34). Remodeling signals that arise from direct or indirect signals, due to action of hormones [estrogen and parathyroid hormone (PTH)] or structural damage leads to initiation of activation phase. This step is marked by higher apoptotic rate of osteocytes and increased osteoclastogenesis which enhances bone damage (35). The decreasing levels of transforming growth factor beta (TGF-β) resulting from osteoclast apoptosis enhances osteoclastogenesis many folds (36). During activation phase, osteoclast precursor activation occurs as a result of increased activity of protein kinase (C/A) and calcium signaling which is mainly responsible for bone resorption (37). This phase is followed by resorption phase, where osteoclast precursors are recruited by osteoblasts at the site of bone remodeling due to various endocrine signaling or by osteocytes. During this phase, osteoblasts overexpress monocyte chemoattractant protein-1 leading to upregulation of RANKL-induced osteoclastogenesis (38). RANKL is mainly responsible for differentiation and proliferation of osteoclast precursors into multinucleated osteoclasts, thereby enhancing life of mature osteoclasts (38). This results in emergence of an isolated environment called as "sealed zone" due to continuous adhesion of osteoclasts to integrin-binding sites (αvβ3 integrin molecules) on the bone surface (39). The overall result is formation of Howship's resorption lacunae with increased hydrogen ion concentration (acidic environment) facilitating dissolution of mineralized and organic components of bone (40). The accumulation of cathepsin K enzyme further enhances this rate of bone resorption (41).

Macrophages enhance the expression of osteopontin that leads to deposition of collagen matrix within Howship's lacunae (42). Next, the bone surface is prepared for bone formation by

osteoclasts. Both osteoclasts and osteoblasts lead to fine tuning of osteocytes, which in turn modulates remodelling. Osteoblasts are derivatives of MSCs and produce extracellular bone matrix of type I collagen and non-collagenous proteins, including osteocalcin, osteonectin, and osteopontin. Multinucleated osteoclasts are produced as a result of differentiation of macrophages and monocytes. RANKL in the presence of MCSF is primarily responsible for functioning and activation of osteoclasts leading to bone loss. On the other hand, osteoblasts produce OPG, which inhibits bone loss by inhibiting osteoclastogenesis. MSC, Mesenchymal stem cell; HSC, Hematopoietic stem cell; RANKL, Receptor activator for nuclear factor kappa; OPG, Osteoprotegerin.

osteoblasts and osteomacs, primarily responsible for removing different remnants of collagen. The different mechanisms responsible for this coupling phenomenon and for propagating this transition of bone formation to bone resorption has always been a subject of controversy, as earlier studies reported that bone functions as a store house for these coupling molecules and release them accordingly during various steps of bone resorption. During reversal phase, osteoblast precursors differentiate and secrete various molecules that are responsible for development of new bone surfaces. In this phase, various factors like IGF (I and II) and TGF-β are recruited by MSCs to the bone resorption sites (43). Finally, to attain final shape of newly formed bone, hydroxyapatite gets integrated at these newly developed osteoid (43). The bone remodeling phase comes to termination phase when equilibrium is attained between bone formation and bone resorption *via* signals initiated by osteocytes. The loss of sclerostin expression at the end of remodeling cycle instigates the osteoclastogenesis. The mature osteoblasts revert back to bone lining phenotype or undergo apoptosis and subsequently get differentiated into osteocytes. To maintain skeletal structural integrity, bones (both cortical and trabecular) need constant remodeling including repairing various micro-cracks developed due to normal wear and tear. Thus, the important aspects of bone remodeling are the repair, development and maintenance of bone along with functioning as a calcium store house of the host.

## T LYMPHOCYTES AND OSTEOPOROSIS (IMMUNOPOROSIS)

Activated T lymphocytes are primary sources of RANKL (44) and TNF-α (45) responsible for bone destruction observed during various pathological (45) and inflammatory conditions (2, 44) (**Figure 2**). Interestingly, T cells also possess anti-osteoclastogenic properties, as depletion of both CD4<sup>+</sup> T and CD8<sup>+</sup> T lymphocytes leads to decreased production of OPG (46–48). Different studies have revealed that T cell deficient nude mice have normal or elevated bone mineral density (45). These studies suggest that T cells have an important role in bone remodeling but the exact link between T cells and osteoclastogenesis particularly in osteoporosis is still not fully elucidated and needs more research thereby giving strong impetus for the emergence of a dedicated novel field to specifically study the role of immune system in osteoporosis, i.e., immunoporosis (with focused emphasis on the role of T lymphocytes).

### Th1 Cells and Osteoporosis

The term Th1 was first given by Tada in 1978 but the clear demonstration of the existence of Th1 cell was provided by Mosmann in 1986 (49, 50). Naive CD4<sup>+</sup> T cells differentiate into Th1 cells when stimulated by interleukin (IL)-12 resulting in the production of IFN-γ, IL-2, lymphotoxins, TNF-α and granulocytemacrophage colony-stimulating factor (38, 51). The production of high levels of IFN-γ by Th1 cells induces activation of both phagocytic activity and complement proteins thereby playing an important role in protection against various intracellular pathogens (52). In addition to protection from invading pathogens, Th1 lymphocytes have been associated with development of organ-specific autoimmune diseases (53). Initially, many inflammatory conditions, such as experimental autoimmune encephalomyelitis (EAE), inflammatory bowel disorders (IBD), autoimmune arthritis and collagen-induce arthritis were linked to unchecked Th1 responses (52, 54). Interestingly, Th1 cells do not possess such characteristics thereby making it clear that the osteoclastogenic Th cells might belong to a yet unknown subset (17, 54). Lately, it has been confirmed that yet another recently defined subset of helper T cells (*viz.* Th17 cells) are predominately responsible for causing these inflammation related bone pathologies (55, 56). This new subset thus became the fore runner for the modulation and regulation of bone health. Also, majority of cytokines produced from Th1 cells inhibit osteoclastogenesis (57), even small quantities of IFN-γ can inhibit osteoclastogenesis through degradation of TRAF6 molecule (9) thereby inhibiting bone loss. Together, these compelling findings establish the osteoprotective role of Th1 responses in the pathogenesis of osteoporosis.

### Th2 Cells and Osteoporosis

Like Th1, the term Th2 was also given by Tada in 1978, and later, Mosmann in 1986 (49, 50) provided a full description about Th2 cells. Th2 cells play an important part in host defense against multi-cellular parasites and in protection against allergies and atopic illnesses. Th2 cells are produced by stimulating naive CD4<sup>+</sup> T cells in the presence of IL-4, leading to production of IL-4, IL-5, IL-10 and IL-13 cytokines (12, 34, 51, 57). Th2 cells were initially believed to be responsible for anti-inflammatory activity in various Th1 cell mediated or Th1 model diseases (57). It was also observed that in case of severity of experimental autoimmune myocarditis inhibition of IL-4 with anti-IL-4 monoclonal antibody reduced the severity of disease (57). The Th2 signature cytokines IL-4, IL-5, and IL-13 have been reported to be associated with inhibition of osteoclastogenesis (58). It has been observed that Th2 cell activation enhances production of PTH, resulting in maintenance of anabolic activity of osteoblasts under various inflammatory conditions (59). Simultaneously, it has also been reported that mice lacking T lymphocytes are unprotected by catabolic activity of PTH (59). Interestingly, Th2 lymphocytes have also been reported to inhibit bone loss by significantly lowering the RANKL/OPG ratio (59). IL-4 has been observed to inhibit bone resorption under both *in vivo* and *in vitro* conditions (60, 61). Low concentrations of Th2 cytokines such as IL-4 and IL-10 have been observed in both the synovial fluid and the peripheral blood of osteoarthritis (OA) patients (62). These findings thus clearly establish the osteoprotective role of Th2 lymphocytes in the pathophysiology of osteoporosis.

### Th9 Cells and Osteoporosis

Th9 cells are recently defined subset of Th cells producing IL-9 and have been associated with immune responses against intestinal worms and immunopathology of various allergic and autoimmune disorders, *viz.* systemic sclerosis, systemic lupus erythematosus (SLE) and EAE (38, 63). Th9 cells in association with Th2 cells are believed to be promoting inflammation during allergic diseases but it still lacks full validation (64). In other cases, there has been a close association between IL-9

boxes. STAT, Signal transducer and activator of transcription; RORγ, RAR-related orphan receptor gamma; Foxp3, Forkhead box P3; BCL6, B-cell lymphoma 6.

with Th17 cells development, as IL-9 in the presence of TGF-β initiates differentiation of naive CD4<sup>+</sup> T cells into Th17 cells (65). In addition, IL-9 has also been reported to amplify the development of Th17 cells during positive autocrine loop (66). In addition, serum levels of IL-9 or total Th9 cell population have been found immensely increased in case of arthritis and OA (38). Together, these studies point towards the role of Th9 cells in various bone conditions like RA but evidences are still lacking for its correlation with osteoporosis, which warrants further research in the field.

### Th17 Cells and Osteoporosis

Th17 cells are associated with protection against bacterial infection and are primarily responsible for induction of various autoimmune diseases *via* recruitment of signatory cells especially neutrophils (34, 38). The differentiation and development of Th17 cells is mainly carried by TGF-β, IL-6, IL-1β and IL-23. Primary cytokines secreted by Th17 are IL-17A, IL-17F, IL-21 and IL-22 (67). Th17 cells have also been reported to be constitutively present across mucosal surfaces including lamina propria of intestine (68). Th17 cells play an important role in various inflammatory conditions, such as osteoporosis, psoriasis, periodontal disease, RA and IBD (55, 56). Th17 cells are thus now often labeled as osteoclastogenic subsets of T lymphocytes (17, 69). Th17 cells enhance osteoclastogenesis by secreting higher levels of IL-1, IL-6, IL-17, RANKL and TNF in addition to low levels of IFN-γ (18, 70, 71). IL-17 enhances osteoclastogenesis by stimulating osteocytes and osteoblasts and potentiates osteoclastogenic activity *via* upregulation of receptor activator of NF-κB (RANK) production to produce higher levels of RANKL (57, 72). IL-17 acts as a pivotal communication point between T lymphocytes and osteocytes by modulating production of RANKL (73). Although the profile of Th17 lymphocytes in different lymphoid tissues requires further analysis, its role in the pathogenesis of osteoporosis is now well documented, since in osteoporotic patients Th17 cell population has been found to be increased many folds (34, 38). In addition, Th17 cells in the blood and peripheral tissues can serve as an important marker for osteoporosis. Altogether, these observations and studies dissect a clear and comprehensive role of Th17 cells in osteoporosis.

### Treg Cells and Osteoporosis

Regulatory T cells (CD4+CD25+Foxp3+ T cells) represent a special subset of Th cells that are mainly accounted for the prevention of autoimmune diseases, maintaining peripheral tolerance and limiting chronic inflammatory diseases by suppressing and regulating the effector function of Th cells (74). It has been reported that Foxp3<sup>+</sup> Tregs are represented by both natural Treg (nTreg) and adaptive or induced Treg (iTreg) cell populations. nTreg cells develop in the thymus whereas iTreg cells are generated in the periphery (75, 76). nTregs represent population of CD4<sup>+</sup> T lymphocytes residing in the thymus and constitute approximately 5–10% of the peripheral naive CD4+ T lymphocyte pool in both mice and humans. They play a significant role in the maintenance of immunological self-tolerance and modulation of immune responses (77). iTregs are found in peripheral lymphoid tissues and are derived from naive T cells (78). IL-10 and TGF-β cytokines are mainly recruited for the development and differentiation of iTregs (79). Forkhead transcription factor (Foxp3) usually expressed by Tregs has an important role in the development and functions of Tregs (77). Treg cells presume an important role in immune regulation during various inflammatory and autoimmune diseases, since Tregs regulate secretion and expression of pro-inflammatory and anti-inflammatory cytokines (18, 34, 80).

Tregs also secrete cytotoxic T-lymphocyte antigen 4, which control immune function upon binding with CD80/CD86 present on mononuclear osteoclast cells thereby leading to inhibition of inflammatory responses (34). Furthermore, inhibition of collagen-induced arthritis in mice was reported to be suppressed by Treg cells (81). Treg cells have been reported to directly inhibit osteoclastogenesis by suppressing RANKL and M-CSF production leading to increased bone volume (20, 82). Recently, the CD8 counter part of Treg cells has also been discovered as osteoclast-induced FoxP3<sup>+</sup>CD8<sup>+</sup> Treg cells which suppress both the formation and activity of osteoclasts *via* suppression of actin ring formation leading to inhibition of osteoclastogenesis (22). Interestingly, this regulatory bi-directional mechanism does not require the presence of various pro-inflammatory cytokines. Altogether, it has now been well established beyond doubt that any dysregulation in the population or functioning of Tregs would result in enhanced bone loss reported in osteoporosis. Thus, exploring novel pathways and molecular mechanisms regulating the cross talk between Tregs and bone cells is highly desired for future clinical implications.

### TFH Cells and Osteoporosis

Follicular helper T lymphocytes are usually found in the follicles of lymphoid tissue and induce production of immunoglobulins from B cells (83). TFH cells express various distinctive genes such as C-X-C chemokine receptor type 5 (CXCR5), inducible T-cell costimulator (ICOS), CD40L, programmed cell death-1 (PD-1), B-cell lymphoma 6 protein and IL-21 (84). Growing evidence has shown that TFH cells influence the severity of various autoimmune diseases such as RA and SLE by enhancing generation of reactive autoantibodies from B cells (85). There has been reports of increased number of circulating TFH lymphocytes (*viz.* CXCR5<sup>+</sup>PD-1<sup>+</sup>CD4<sup>+</sup> or CXCR5<sup>+</sup>ICOS<sup>+</sup>CD4<sup>+</sup>) in SLE patients which correlates with the amount of autoantibodies and SLE severity (86). Specific immunohistochemistry analysis has further confirmed the presence of TFH lymphocytes (CD4<sup>+</sup>CXCR5<sup>+</sup>ICOS<sup>+</sup>) in the synovial tissues of RA patients (38). In addition, in RA patients, there was observed an elevation of CD19<sup>+</sup> B cells and increased serum IL-21 which is positively associated with disease scores and presence of anti-citrullinated antibodies (87, 88). There has been reports of elevated levels of TFH cells in both Sjogren's syndrome patients and RA patients (88, 89). These studies provide strong evidence for potentially important roles played by TFH cells in the pathogenesis and progression of autoimmune diseases through various pathways. Collectively, these studies establish the role of TFH cells in RA thereby opening new avenues in the field of immunoporosis for further dissecting and delineating the role of TFH lymphocytes in pathogenesis of osteoporosis.

### Natural Killer T (NKT) Cells and Osteoporosis

Natural killer T cells represent heterogeneous group of T lymphocytes which share properties of both natural killer cells and T cells. NKT cells are primarily involved in the clearance of transformed and virus-infected cells (90). NKT cells modulate initiation and development of immune responses mediated by both T and B cells *via* production of various growth factors and cytokines (16). NKT cells have been found to regulate development and function of macrophages, monocytes and myeloid dendritic cells (91, 92). A recently reported subset of NKT cells called as invariant NKT have been found to regulate the development and differentiation of osteoclasts (93). NKT cells have also been reported in the synovial tissues of RA patients where they constitute up to 20% of all the lymphocytes (94). The CD56bright subset of NKT cells have been reported with upregulated expression of various chemokine receptors and adhesion molecules responsible for enhanced recruitment of NKT cells (95) thereby engaging and activating monocytes for enhanced osteoclastogenesis in synovium of RA patients (94). T cells and macrophages activated by NKT cell-derived IFN-γ also leads to increased secretion of TNF-α (96), a strong pro-osteoclastogenic cytokine. TNF-α enhances osteoclastogenesis in a RANKL-dependent manner either directly or by promoting commitment of progenitors to osteoclast lineage and differentiation or indirectly by stimulating secretion of RANKL and M-CSF by osteoblasts (97, 98). In addition, RANKL and MCSF are also produced by NKT cells and thus induce osteoclastogenesis which is further upregulated by IL-15 (99). Altogether, these reports elaborate the important role of NKT in the pathogenesis of inflammatory bone diseases. Thus, regulation of NKT cells can be an important aspect for regulating bone loss in osteoporosis.

# **γδ**-T Cells and Osteoporosis

T cells present in circulation mainly express αβ-T cell receptor (TCR) chains along with a small subset of T cells which uniquely expresses TCRs, containing a gamma (γ) chain and a delta (δ) chain named as γδ-T cells. About 1–10% of T lymphocytes in periphery of human circulation comprise of γδ-T cells, however, skin and other tissues have more abundant population of γδ-T cells (99). γδ-T cells are more innate-like, unlike their counterpart αβ-T cell which are adaptive. Also, the TCR specificity of γδ-T cells is uniquely directed toward nonpeptide antigens. γδ-T cells are now increasingly being linked with autoimmunity, allergy, hematological tumors (100) and infectious diseases (101). γδ-T cells have recently been reported to produce factors which are important in the healing of various skeletal fractures (102). It has been observed that anti-CD3/ CD28-stimulated γδ T cells lead to inhibition of human osteoclast formation and simultaneous resorptive activity *in vitro*. In addition, stimulated γδ-T cells leads to increased production of IFN-γ and inhibits expression of IL-17 production (103). γδ-T cells are quite unique and heterogeneous population of

T lymphocytes and are easily lost in patients undergoing aminobisphosphonate treatment for various bone-related pathologies (104). γδ-T cells have been associated with peculiar property of immunomodulation and thus are a promising candidate for treatment of various inflammatory conditions including bone pathologies.

### CD8**+** T Cells and Osteoporosis

CD8<sup>+</sup> T cells are yet another established subset of T lymphocytes which play a key role in cell-mediated responses in the immune system. CD8<sup>+</sup> T cells are also referred as cytotoxic T lymphocytes which aid in protection of host from foreign organisms through both lytic and non-lytic means. CD8<sup>+</sup> T cells are responsible for regulating the immune responses and simultaneously eliminating transformed tumor cells (104, 105). Another recently defined regulatory subset of CD8<sup>+</sup> T cells called as Foxp3<sup>+</sup>CD8<sup>+</sup> Tregs has been found to suppress osteoclast formation and activity by secreting various anti-osteoclastogenic cytokines (22). Foxp3<sup>+</sup>CD8<sup>+</sup> Tregs cells not only regulate the survival of osteoclasts but also affect the maturation of osteoclasts by suppressing their actin ring formation. The recently discovered unique property of osteoclast inducing the generation of FoxP3+CD8+Treg cells and the ability of FoxP3<sup>+</sup>CD8<sup>+</sup>Treg cells to subsequently regulate osteoclast function establishes a bi-directional regulatory loop between these two cells types in the BM (22). CD8<sup>+</sup> T cells have also been reported to play an important role in the pathogenesis of OA (38). These reports thus clearly establish the important role of CD8<sup>+</sup> T cells in bone health, but still evidences for their role in osteoporosis is not well established and needs further research.

# CONCLUSION AND FUTURE PERSPECTIVES

Inflammatory bone conditions including RA, OA and osteoporosis arise due to dysregulation of the homeostatic nexus between bone and immune system thereby leading to enhanced bone loss. Various strategies are now being developed for inhibiting inflammation induced bone loss. One direct approach of inhibiting osteoclastogenesis could be interfering with inflammatory pathways thereby providing an alternate method for managing inflammatory bone loss/damage. The emergence of "immunoporosis" as an independent field of research would thus give a novel and unique insight into the underlying immune-skeletal interaction between both immune cells (T cells) and bone cells (**Figure 3**). Since the role of immune system, in general, has been implicated for elucidating the pathogenesis of numerous inflammatory diseases, however, the immunoporosis aspect, i.e., the role of specific subsets of T lymphocytes in osteoporosis is still unclear. Thus, it will be important to see the course of various subsets of T cells in the development and progression of osteoporosis. Both CD4<sup>+</sup> (Th1, Th2, Th9, Th17, Treg, NKT, γδ T cell subsets) and CD8<sup>+</sup> T cells play an important role in regulating bone health. Th17 cells are one of the major inducers of bone loss *via* expression of higher levels of RANKL and TNF-α (17, 106, 107). On the other hand, Tregs are major inhibitors of bone loss (108, 109) through production of IL-4, IL-10 and TGF-β1 cytokines (108, 109). Tregs also inhibit the effector functioning of Th17 cells in inflammation-induced bone loss (109, 110). Tregs also lead to suppression of bone loss by inhibiting differentiation


of monocytes into osteoclasts under both *in vitro* and *in vivo* conditions (20, 75, 111) (**Table 1**).

Since osteoporosis affects more than 50% of female population and one-fourth of males above age 50 (3, 20), osteoporotic patients are very higher in number than OA and RA (3). The current therapies employed for the treatment of osteoporosis, *viz.* bisphosphonates, strontium ranelate, selective estrogen receptor modulators, PTH (teriparatide), etc. provide remission of inflammation with little relief to the patients (112). Therefore, the need for new and effective future therapeutics are need of the hour for long-term relief from various bone loss mediated pathologies. This could be made possible only upon having a clear and better understanding of immune-skeletal biology which can be clearly defined and delineated under the novel field of immunoporosis. Thus, advanced exploration and research under the aegis of immunoporosis would lead to novel opportunities and avenues for development of enhanced therapies. Recently, the inclusion of probiotics (*via* modulation of host immune system) as a supplementary therapy for bone loss represents one such class (20, 113, 114). Thus, a better understanding of the nexus between both the systems should be at heart of future research in the area. The establishment of

### REFERENCES


"immunoporosis" as an independent field of modern biology catering to the recent developments in the field will thus provide new paradigms for development of focused novel therapeutic strategies for managing osteoporosis.

### AUTHOR CONTRIBUTIONS

RS suggested the focus and outline of the review along with writing the review. HD participated in the writing of the review along with creating figure illustrations. PM provided valuable inputs for manuscript preparation.

### ACKNOWLEDGMENTS

This work was financially supported by a project from UGC-FRPS (30–12/2014), Govt. of India, sanctioned to RS. RS and HD acknowledges Department of Zoology, Dr. Harisingh Gour Central University, Sagar (MP), India for providing infrastructural facilities. RS also thanks Department of Biotechnology, All India Institute of Medical Sciences (AIIMS), New Delhi, India for providing necessary facilities. HD thanks UGC for research fellowship.

distinct from the T helper type 1 and 2 lineages. *Nat Immunol* (2005) 6: 1123–32. doi:10.1038/ni1254


and *in vivo*. *J Immunol* (2008) 181:1232–44. doi:10.4049/jimmunol.181. 2.1232


maturation of dendritic cells in vivo and thereby acts as an adjuvant for combined CD4 and CD8 T cell immunity to a co-administered protein. *J Exp Med* (2003) 198:267–79. doi:10.1084/jem.20030324


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

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

# Emerging Functions of Regulatory T Cells in Tissue Homeostasis

### *Amit Sharma1,2 and Dipayan Rudra1,2\**

*1Academy of Immunology and Microbiology, Institute for Basic Science (IBS), Pohang, South Korea, 2Division of Integrative Biosciences and Biotechnology, Pohang University of Science and Technology (POSTECH), Pohang, South Korea*

CD4+Foxp3+ regulatory T-cells (Tregs) are a unique subset of helper T-cells, which regulate immune response and establish peripheral tolerance. Tregs not only maintain the tone and tenor of an immune response by dominant tolerance but, in recent years, have also been identified as key players in resolving tissue inflammation and as mediators of tissue healing. Apart from being diverse in their origin (thymic and peripheral) and location (lymphoid and tissue resident), Tregs are also phenotypically heterogeneous as per the orientation of ongoing immune response. In this review, we discuss the recent advances in the field of Treg biology in general, and non-lymphoid and tissue-resident Tregs in particular. We elaborate upon well-known visceral adipose tissue, colon, skin, and tumor-infiltrating Tregs and newly identified tissue Treg populations as in lungs, skeletal muscle, placenta, and other tissues. Our attempt is to differentiate Tregs based on distinctive properties of their location, origin, ligand specificity, chemotaxis, and specific suppressive mechanisms. Despite ever expanding roles in maintaining systemic homeostasis, Tregs are employed by large varieties of tumors to dampen antitumor immunity. Thus, a comprehensive understanding of Treg biology in the context of inflammation can be instrumental in effectively managing tissue transplantation, autoimmunity, and antitumor immune responses.

Keywords: immune tolerance, autoimmunity, regulatory T cells, regulatory T-cells, Foxp3, tissue Treg, tumor Treg, regeneration

# INTRODUCTION

Vertebrate immune and nervous system are two systems which are cognitive and under continuous interaction with the environment. This probably explains why both share common paradigms like recognition, learning or modulation, and memory. For long, immunology has been defined as a science of "self/non-self " discrimination (1). However, overtime, the immunological concept of "self " has evolved, where the very definition of an individual with defined anatomic borders, compatible balance between its parts, physiological autonomy, and ability to replicate as a unit is rapidly challenged by symbionts. Immune identity is now considered more fluid than restricted in strict borders (2). Hence, apart from generating a protective and offensive backdrop, the immune system must work to maintain an organismal identity by mediating dynamic exchange processes with the environment. This not only entails to generate robust defense against pathogens and toxins but also makes it rather paramount to suppress an overzealous immune response, curb autoimmune reactions, and maintain equilibrium toward commensals and food. Several mechanisms like clonal deletion, editing, anergy, ignorance, and immune deviation have evolved to safeguard against self-directed immunity (3). Specialized cells with immunosuppressive capabilities like

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Koji Yasutomo, Tokushima University, Japan Zhaocai Zhou, Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences (CAS), China*

### *\*Correspondence:*

*Dipayan Rudra rudrad@ibs.re.kr*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 28 February 2018 Accepted: 10 April 2018 Published: 25 April 2018*

### *Citation:*

*Sharma A and Rudra D (2018) Emerging Functions of Regulatory T Cells in Tissue Homeostasis. Front. Immunol. 9:883. doi: 10.3389/fimmu.2018.00883*

tolerogenic dendritic cells (DCs) (4–6), regulatory B cells (7, 8), regulatory innate lymphoid cells (9), type 1 regulatory (Tr1) T-cells (10), and Foxp3<sup>+</sup> regulatory T-cells (Tregs) (11–13) have also evolved.

Regulatory T-cells are arguably the most versatile immunosuppressive cells and work like immunological sentinels across various tissues. Both in mice and men, loss of these cells essentially results in breakdown of tolerance and multi-organ autoimmunity. Since their discovery, biology of Tregs has been a most dynamic field of immunological research and as a result, Tregs, which were once considered as a homogenous immunosuppressive population, have been found to be highly adaptable and diversified cell type. Their heterogeneity is now appreciated in the context of origin, localization, differentiation, and mechanisms of immunosuppression. In the first part of this review, we will briefly discuss the events which put Tregs to the center stage of immune research, following which we will attempt to elaborate on the various layers of Treg heterogeneity especially pertaining to non-lymphoid and tissue-resident Tregs.

## CONCEPT OF "DOMINANT" TOLERANCE AND THE EMERGENCE OF REGULATORY T-CELL RESEARCH

T-cell tolerance is pivotal for regulating adaptive immune responses as T-cell help is essential for mounting an antibody response *via* B cells (14). T-cell tolerance for long, was studied in light of "recessive tolerance," wherein T-cells with high affinity TCRs toward self-antigens are clonally deleted (15), or undergo "receptor editing" in thymus (16, 17). The runaway cells which escape these central processes encounter anergy or activation induced cell death in the periphery (15, 18). However, studies on tolerance ushered into an "active" or "dominant" era with the seminal discovery of suppressive CD4<sup>+</sup> T-cells expressing high levels of high efficiency α-chain receptor of IL2 (CD25) (19).

### The Outset of Treg Research

Preliminary evidences of suppressive cells maintained in thymus started emerging when several investigators reported that neonatal thymectomy (3 day postnatal, 3dTx) could induce various autoimmune diseases in suitable mouse strains (20–25). Even more astonishing was the fact that similarly induced disease processes in rats could be reversed by reconstitution with normal lymphoid cells (26). Several groups tried to identify specific markers to distinguish suppressive cells from pathogenic T-cells in the thymus. It was reported that T-cells depleted of CD4<sup>+</sup>CD5hi cells induced autoimmune phenotype akin to 3dTx in BALB/c and C3H mice (27). Two other groups demonstrated the capability of CD4<sup>+</sup>CD45RBhi T-cells in inducing inflammatory bowel disease in BALB/c SCID mice (28, 29) and its resolution upon reconstitution with total T-cells. While these studies demonstrated that phenotypically distinct subsets of T-cells are capable of mounting discrete immune responses, specific identity of tolerance inducing counterparts remained elusive. Sakaguchi et al. in 1995 (19) discovered high surface expression of CD25 on about 8–10% of CD4<sup>+</sup> T-cells, which were both CD5hi and CD45RBlo

in concordance with previous studies. Asano et al. (30) demonstrated that CD4<sup>+</sup>CD25<sup>+</sup> T-cells appear around day 3 postnatal and increase up to the adult levels by day 10. These authors were the first to propose the term "regulatory" for this subtype.

### Discovery of Foxp3

While subsequent studies involving numerous experimental models of autoimmunity established its functional existence (31), the usage of CD25 as a marker for Tregs remained controversial for a number of years due to its upregulation in all activated T-cells. Furthermore, it seemed possible that a subset of the activated T-cells, by virtue of marked upregulation of the IL2 receptor α on their surface, restrained immune response simply by competing for IL2.

A mouse line dubbed "scurfy," with spontaneous autoimmunity (originally appeared as a spontaneous mutation at the Oak ridge national laboratory, USA under the Manhattan project), was immunologically characterized in 1991. Scurfy mice have an X-linked recessive mutation which leads to scaly skin, lymphoproliferation, hypergammaglobulinemia, lymphadenomegaly, anemia, runting, and early death (32). Thymectomy reduced the severity of the disease but did not totally ameliorate it. However, crossing the strain with *nu/nu* mice totally prevented the disease, suggesting thymic origin of disease causing cells. Several other studies revealed scurfy to be mainly a T-cell dependent disorder (33–35) much similar to Cytotoxic T-Lymphocyte Associated Protein 4 (CTLA4) (36) and Transforming growth factor β1 (TGFβ1) deficient animals (37). These similarities instigated investigations to identify the gene responsible for scurfy phenotype. In 2001, Brunkow et al. (38) identified 20 putative genes in a 500-kb region of X-chromosome by sequencing four overlapping bacterial artificial chromosomes. Out of these, one possessed an ORF highly homologous with DNA-binding domain of the forkhead/HNF3/winged helix family of proteins. This gene in scurfy mouse was found to harbor a 2-bp insertion mutation, resulting in a truncated gene product, deleting the C-terminal forkhead domain (38). Investigators designated this gene as *Foxp3*. Functional complementation experiments by mating scurfy carrier females with *Foxp3* transgenic lines resulted in complete rescue of the scurfy phenotype, corroborating *Foxp3* mutation as the cause (38).

At around same time, mutations in *FOXP3* gene and its 3′ untranslated region were confirmed in human patients of IPEX syndrome (39, 40). IPEX syndrome is immunodysregulation polyendocrinopathy enteropathy X-linked, originally described in 1982 by Powell et al. (41). The striking similarity in autoimmune phenotype of IPEX patients, scurfy and 3dTx mice led several groups to examine the function of *Foxp3* in Tregs. Subsequently, in 2003, three studies reported that indeed *Foxp3* is uniquely expressed by CD4<sup>+</sup>CD8<sup>−</sup>CD25<sup>+</sup> thymocytes and CD4<sup>+</sup>CD25<sup>+</sup> peripheral regulatory T-cells (42–44) in mice. Retroviral transduction of *Foxp3* induced CD25 expression in CD4<sup>+</sup>CD25<sup>−</sup> T-cells which were functionally suppressive and expressed Treg associated molecules CTLA4 and GITR. Deletion of *Foxp3* in mice resulted in lymphoproliferative disorder identical to scurfy mice (43, 44). Mixed bone marrow chimera experiments demonstrated that indeed only *Foxp3*-sufficient bone marrows were capable of generating CD4<sup>+</sup>CD25<sup>+</sup> Tregs (44). Conclusive evidence for Foxp3 as lineage specific marker for mice Tregs came from Foxp3eGFP reporter mice (45) in which GFP expression was found only in TCRβ+ T-cells among all hematopoietic cellular compartments. Conditional deletion of *Foxp3* in CD4+ T-cells led to a lymphoproliferative disorder mirroring scurfy phenotype. These series of experiments established Foxp3 as the molecular identity responsible for implementing Treg transcriptional signature. Further investigations revealed that the *Foxp3* gene itself is regulated by three conserved noncoding sequences (CNS) 1–3. Detailed epigenetic analyses have identified CNS1 as the TGFβ responsive element which is required for peripheral generation of Tregs, CNS2 is involved with heritable maintenance of Foxp3 expression while CNS3 acts as a pioneering element for thymic induction of *Foxp3* (46). Proteomic analyses demonstrated Foxp3 to be interacting with more than 350 proteins in multiprotein complexes, many of which are transcription-related factors (47). A detailed review on regulation of *Foxp3* and Foxp3 mediated regulation of the Treg transcriptome can be found in Lu et al. (12).

### DEVELOPMENTAL AND PHENOTYPIC DIVERSITY IN Tregs

Neonatal thymectomy experiments in mice confirmed beyond doubt that early development of Tregs happens in thymus. A detailed discussion on thymic development of Tregs can be found in Ref. (13, 48, 49). Briefly, cell surface markers indicative of strength of TCR interaction (CD25, CD5, etc.) suggested the involvement of TCR signaling. TCR repertoires of Tregs have limited overlap with that of non-Tregs and are largely self-reactive (50). Nur77-GFP reporter mice which express GFP under *Nur77* gene locus, an early gene expressed upon TCR stimulation have higher GFP expression in thymic Tregs (tTregs) (51). With regard to cytokines, it was reported earlier that mice lacking either IL2 or CD25 (45) are able to generate Tregs, albeit at a reduced level. However, if common γ chain is deleted, Tregs are not formed (45). This suggests cooperation among γ chain cytokines in *Foxp3* expression and maintenance. Thus, the current model of Treg generation in thymus gravitates toward an instructive one wherein TCR signaling substantially above the strength required for positive selection and relatively near the strength that induces negative selection initiates specification toward pre-Treg state. In the second step, cytokines induce *Foxp3* expression. More recently, Satb1, a genome organizer and transcription factor was shown to at least partially mediate the genomic arrangement of super-enhancers responsible for Treg development (52). The Treg specific super-enhancer patterns were found "poised" for activation even in conventional peripheral T-cells.

Though thymic regulatory T-cells are adept at suppressing autoimmune responses against self-antigens, a reasonably tolerant immune environment cannot be developed if repeated immune responses are mounted against beneficial and innocuous microbes as well as food antigens. In part, this is achieved by generation of Tregs in the periphery. Indeed, initial evidences suggested that Tregs can be generated by oral antigen feeding (53, 54) as well as by antigen-specific APCs (55) in the absence of functioning thymus. Also, conventional CD4<sup>+</sup>CD25<sup>−</sup> naïve T-cells could be converted to CD4+CD25+CD45RB−/low suppressor cells by costimulation with TCR and TGFβ. TGFβ activates Smad2 and 3 transcription factors (56) which redundantly, help in peripheral Treg generation by initiating a cascade of interactions with specific enhancer regions within the *Foxp3* locus (56–58).

Contrary to initial interpretations of Tregs being a universal immunosuppressive population, diligent interrogations led to identification of a rather diverse and distinct pool of heterogenous subsets. Other than the site of induction, Tregs were classified into two separate populations: central Tregs (cTregs) and effector Tregs (eTregs) (59, 60). cTregs are comparatively quiescent Tregs in the lymphoid tissues. They express the lymphoid homing molecules CD62L and CCR7 and are dependent on IL2 secreted by Tconv in T-cell zones of lymphoid tissues (60). On the other hand, eTregs are primarily non-lymphoid Tregs which downregulate lymphoid homing molecules and upregulate CD44, ICOS, GITR, and other activation-induced markers. For maintenance, they are dependent on sustained ICOS signaling (60). Most of these cells express transcription factor BLIMP1 and produce high IL10, akin to a population of ICOS<sup>+</sup>IL10<sup>+</sup> Tregs in humans (61).

Investigations into Treg mediated suppression of distinct helper T-cell immune responses imply a contextual T helper–Treg coupled viewpoint of Treg heterogeneity. It was reported that Tregs express high amount of the interferon (IFN) regulatory factor IRF4, Treg specific ablation of which resulted in selective Th2 related pathologies (62). These findings initiated similar investigations in other helper T-cell contexts and indeed, now a well-established paradigm exists wherein, in a Th1 inflammatory context Tregs express transcription factor Tbet, responsible for Th1 speciation, and express CXCR3 to accumulate at such sites (63). Transcription factor STAT3 is expressed in Tregs in a Th17 context (64) which helps in upregulation of CCR6 to migrate to intestine and production of IL10 (65). Similarly, Bcl6 expression in Tregs was shown to be important for regulating Tfh cells and expression of CXCR5 (66, 67).

Moving a step further and increasing complexity and diversity among Tregs, several Treg subpopulations are discovered in non-lymphoid tissues. These Tregs are not only found to be instrumental in suppression of inflammatory responses but also integrate into a larger biological paradigm for the benefit of organ and organismal homeostasis. In the following sections, we will review the characteristics of non-lymphoid Tregs, how they originate and function in maintaining homeostasis of tissues. We will attempt to elucidate if, apart from location, these cell types can be further identified by virtue of tissue-specific phenotypic and functional characteristics. We will elaborately discuss the well-characterized Treg populations in adipose tissues (AT), intestines, and skin, as well as attempt to highlight upon some of the recently identified ones in other tissues like muscles, lungs, and placenta. While overzealous Tregs in malignancies not necessarily can be clubbed with normal tissue Tregs, but at certain level, as Tregs perceive the context of the environment they reside in and are hijacked by tumors for their benefit, we will also discuss tumor infiltrated Tregs in order to elucidate on the general principles that might integrate such heterogeneity.

# FAT Tregs: IMMUNOSUPPRESSION FOR METABOLIC HOMEOSTASIS

Adipose tissue is natural calorie reservoir of the body. Overall, the tissue is classified as two somewhat functionally antagonistic tissue types: white adipose tissue (WAT) and brown adipose tissue (BAT). WAT stores the excessive nutrients in the form of fat droplets during over-nutrition and releases it under energy deficit conditions. BAT, by virtue of higher expression of the uncoupling protein 1 (UCP1) protein, enhances energy utilization by non-shivering thermogenesis (68). WAT is abundantly available and found mainly in subcutaneous tissue, omenta, mesenteries, perirenal tissues, and bone marrow, while BAT has a restricted distribution occurring mainly in interscapular and inguinal regions.

### AT Architecture

Structurally, AT is a loose connective tissue comprising mainly of fat cells (adipocytes), each surrounded by its own basal lamina and separated by a thin layer of extracellular matrix composed of reticular and collagen fibers and supplied with numerous capillaries. Several other resident as well as transient cells are strewn around in AT namely fibroblasts, myofibroblasts, and immune cells like macrophages, mast cells, eosinophils, neutrophils, and T-cells. WAT are the professional fat depots (69) which store excess energy as triglycerides without the common lipotoxicity experienced by other cell types (70). WAT adipocytes are usually spherical with a single fat droplet occupying 90% cell volume and a thin elongated mitochondrion on one side. On the other hand, BAT integrates environmental conditions *via* brain adrenergic responses toward cold temperatures. Adipocytes in BAT are typically polygonal, containing triglycerides in multiple small vacuoles, and are characterized by numerous large, spherical mitochondria.

# Fat Treg Origin and Accumulating Factors

The plethora of adipokines and cross-talk with nervous as well as immune system has underscored the importance of AT as an endocrine organ with profound effects on body's metabolic homeostasis. However, as adipocyte increase in size due to excess calories, hypoxia sets in leading to accumulation of inflammatory macrophages (**Figure 1**) in obese adipose (71). Subsequent upregulation of several inflammatory adipokines (72) can activate CD4<sup>+</sup> T-cells independent of macrophages. Any immune response shall be regulated so that it does not outlive its utility; a task largely achieved by several anti-inflammatory immune cell types already resident in the adipose (73–77).

In a seminal study, Feuerer et al. (76) identified unique fatresiding Tregs in mouse and human ATs. Surprisingly, unlike peripheral lymphoid compartment where normally 10–15% of CD4<sup>+</sup> T-cells are Foxp3<sup>+</sup> Tregs, almost half of the fat CD4<sup>+</sup> T-cells are Foxp3<sup>+</sup> Tregs, which were found to accumulate primarily in visceral fat over a period of time since birth, peaking at around

CCR2. Tregs are abundant in lean adipose tissue of adult mouse; however, if there is a sustained positive energy balance as in high-fat diet-induced obesity animal model, then the Treg numbers decrease drastically. This is concomitant to a change in macrophage phenotype from anti-inflammatory M2 to inflammatory M1 macrophages and unhealthy hypertrophy of adipocytes. Sustained hypertrophy leads to adipocyte apoptosis and exacerbated inflammation. Overall, decrease in adipose Tregs is accompanied by obesity, insulin resistance, dyslipidemia, and chronic low-grade inflammation.

25 weeks of age. In one study, this characteristic accumulation of Tregs was found to be accompanied by a sudden drop to fifth week levels (78) in mice aged further, suggesting a negative correlation between the frequency of visceral WAT resident Tregs and age-related metabolic inflammation (**Figure 2**). In contrast to this finding, a recent study, reported even greater accumulation of Tregs in older mice, implicating an alternative viewpoint instead (79) (**Figure 2**). The cause of this discrepancy might be different colonies, husbandry practices, as well as microbial and dietary composition. Nonetheless, the aged visceral WAT Tregs were found to be functional (79).

As far as origin of AT Tregs are concerned, visceral WAT Tregs appear to be largely of thymic origin. A thymectomy beyond 3 weeks of life does not affect visceral WAT Tregs population in mice (80). TCRα sequencing experiments from "Limited" mice [mice with limited focused diversity restricted to CDR3α (81)] showed that TCR repertoire of fat Tregs are different from conventional fat T-cells. Furthermore, visceral fat Treg repertoire was only a restricted subset of lymph node (LN) Tregs TCR repertoire (76), suggestive of an abdominal WAT specific distinct TCR repertoire of Tregs. This indicates either a continuous supply of Tregs from peripheral LNs or an initial seeding of tTregs followed by selected clonal expansion in AT. Adoptive transfer of congenically marked Tregs confirmed that visceral WAT Tregs are not significantly derived from circulating Tregs (80). Also, very high expression of both Helios and Neuropilin1 (Nrp1) as well as transcriptomic analysis of visceral WAT Tregs suggest their thymic origin and little to no conversion of naïve T-cells into visceral WAT Tregs (80).

Origin of subcutaneous WAT Tregs and BAT Tregs have not been studied in much detail. Naïve T-cells isolated from both these tissues, however, produce significantly high number of induced Tregs than visceral WAT in *in vitro* conversion assays (82).

### Tissue Adaptation and Phenotype

That extralymphoid Tregs adapt to the context and microenvironment is best explained by higher expression of transcription factor PPARγ in fat Tregs (83, 84). Co-immunoprecipitation studies confirmed PPARγ interaction with Foxp3, suggesting a visceral WAT-specific Foxp3-PPARγ-mediated gene expression axis (83). Using a Blimp1-GFP reporter mouse, Vasanthkumar et al. (84) showed that most of the visceral WAT Tregs fall in the category of eTregs. For identifying the survival factors for eTregs, they reported that visceral WAT Tregs specifically express *Il1rl1*, which encodes alarmin IL33 receptor ST2. Both IL33 deficiency in general, as well as T-cell intrinsic ST2 deficiency resulted in specific reduction in number and percentage of visceral WAT Treg compartment (84). In *in vitro* settings, both IL2 and IL33

Figure 2 | Two contrasting scenarios of regulatory T-cells (Treg) numbers and their outcome in aged white adipose tissue (WAT). WAT Tregs increase with age reaching a plateau and then decrease abruptly in aged (~45 weeks) mouse (top). Whether this results in inflammation leading to age-associated insulin resistance is not explored; however, (bottom part) contrasting evidence suggest that white adipose Tregs keep on increasing even in the aged adipose tissue, which in concordance with "adipose tissue expandability hypothesis" (see text) results in suppression of healthy inflammation required for remodeling of adipose. This results in a storage space problem leading to ectopic deposition of fat in visceral organs, like liver and pericardium. This is accompanied by free fatty acid induced lipotoxicity and age-associated insulin resistance.

were able to induce proliferation and ST2 upregulation in a fraction of T-cells upon TCR stimulation which was dependent on MyD88 (84). Development and proliferation of visceral WAT Tregs appears to be dependent on two signals: (1) TCR crosslinking which induces PPARγ and ST2 expression *via* BATF and IRF4, both of which bind to the intronic regions of *Pparg* and *Il1rl1* genes (84) and (2) IL33 which *via* MyD88 feeds forward the expression of ST2 (84). In accordance to the concept of local adaptation, PPARγ in visceral WAT Tregs upregulates the expression of lipid metabolism genes like *Dgat1* (diacylglycerol *O*-acyltransferase 1), coding for an enzyme which catalyzes the terminal step in triacylglycerol synthesis by using diacylglycerol and fatty acyl CoA as substrates; and *Pcyt1a* (choline-phosphate cytidylyl transferase A), an enzyme involved in regulation of phosphatidylcholine biosynthesis (83). It is possible that these gene products enable visceral WAT Tregs to survive in a lipotoxic environment and/or enable the utilization of fatty acids as metabolic fuel. Other than these essential factors, visceral WAT Tregs also express high levels of GATA3, Klrg1, early activation marker CD69, and adipose signature chemokine receptor CCR2. BAT Tregs are found to be very similar to WAT Tregs at transcriptional level with higher expressions of PPARγ, IL10 and chemokine X ligands 1 and 2 (85). The under-expressed transcripts were those encoding for TCR signaling specific T Cell Factor 7 and cytokine IFNγ (85).

# WAT Tregs Are Important Players in Metabolic Syndrome

Adipose originated pro-inflammatory cytokines, like TNFα, IL6, and type1 IFNs, have been suggested as causative of insulin resistance and metabolic syndrome (86–88). Also, human obese subjects were found to be deficient in circulating Tregs, whose levels were inversely correlated with body weight and BMI (89). Considering the primarily immunosuppressive phenotype of Tregs, it is expected that fat Tregs play a major role in controlling adipose inflammation, and in turn, affect the overall metabolic homeostasis of the body. Indeed, in Foxp3DTR mice, in which the gene encoding diphtheria toxin receptor (DTR) is knocked into the *Foxp3* locus (90), Treg depletion upon diphtheria toxin (DT) administration, leads to visceral WAT tissue inflammation (76). However, total Treg deletion also initiates a strong systemic inflammatory response (90). Hence, a visceral WAT specific model for Treg deletion was required. This was achieved by Treg specific deletion of PPARγ, which resulted in more than 80% reduction in visceral WAT Tregs, without any effect on splenic Treg population (83). This decrease in Treg population was accompanied with marked inflammatory cell infiltration in visceral WAT (83). In a high-fat diet-induced obesity model, WAT Treg numbers are reduced drastically (**Figure 1**), which can be rescued by treatment with a synthetic PPAR ligand pioglitazone, which improves insulin sensitivity by working on PPARγ1 and 2 and affecting lipid metabolism (91). Its administration was able to improve the metabolic parameters in wild-type mice but not in mice harboring PPARγdeficient Tregs, suggesting a direct role of PPARγ expression in visceral WAT Tregs.

Interestingly, the role of visceral WAT Tregs in aging animals has been reported to be opposite to what was observed in comparatively young animals. Contrary to obese animals, depletion of visceral WAT Tregs in aged animals improved the metabolic parameters and rescued aging induced insulin resistance (79) (**Figure 2**). Treg specific deletion of PPARγ led to less increase in fat and more in lean weight with age (more than 45 weeks). The adipocyte size was less and hepatic triglyceridosis was decreased. An increase in total Tregs by IL2–IL2 antibody immune complex resulted in reduced glucose uptake by aged adipocytes, suggesting compromised storage function (79). Similar results were obtained upon external IL33 administration. Overall this study reveals an unexpected cooperative role of visceral WAT Tregs in age-associated insulin resistance and metabolic inflammation. One explanation of this seemingly counter-intuitive finding might be extended by the so called "adipose tissue expandability" hypothesis. This hypothesis posits that in a state of "positive energy balance" metabolic complications arise because WAT is not able to expand further and accommodate excess calories, essentially propounding that metabolic syndrome coming out of excessive nutrition and obesity is actually a storage space problem (92). Indeed, animal models where AT inflammation is controlled result in decreased AT hyperplasia in a high-fat diet-induced mouse model of obesity (93), which causes ectopic lipid depositions (hepatic steatosis, dyslipidemia, etc.) and worsened metabolic parameters (93). Also, obese ob/ob mice which were made transgenic for full-length adiponectin and thus had adiponectin levels equivalent to treatment with a PPARγ agonist, showed uninhibited WAT expansion leading to morbid obesity but improved insulin sensitivity and other metabolic parameters (94) owing to reduced ectopic lipid deposition in liver and muscles (94). Given these observations, whether and how age-related accumulation of the visceral WAT Tregs results in compromised AT hyperplasia, remains to be seen.

### BAT Tregs Help in Thermogenesis

A generalized Treg ablation alters the metabolic profile of mice with regard to BAT as well. DT mediated deletion of Tregs in Foxp3DTR mice resulted in reduced whole-body oxygen consumption in a short-term cold temperature exposure model (85). That the Tregs are at forefront of metabolic homeostasis and their interventions are highly context dependent is further strengthened by a recent study analyzing BAT Tregs in detail. While a short-term (2 weeks) high-fat high-sugar (HFHS) diet, promoting thermogenesis (95), increased BAT Tregs, it actually decreased visceral WAT Tregs in young adult mice (82). On the contrary, a long-term HFHS diet (16 weeks) significantly reduced visceral WAT Tregs but made no impact on percent of BAT Tregs. This suggests that the tenor of caloric intake can have specific effect on tissue Tregs. In accordance to the role of BAT Tregs in non-shivering thermogenesis, treatment with ADRB3 (β-3 adrenergic receptor) agonist increased Tregs in BAT. However, "betaless" mice which are deficient in all three (β-1, 2, and 3 adrenergic receptors) did not have a reduced percentage of BAT Tregs, suggesting a redundant role of adrenergic signaling in BAT Treg accumulation *per se*. Treg depletion, followed by β-3 stimulation on the other hand did result in reduced levels Sharma and Rudra Tregs in Tissue Homeostasis

of BAT thermogenic and lipolytic genes (*Ucp1*, *Ppargc1a*, *Pparg*, *Prdm16*, *Lpl*, etc.) confirming Treg functionality in BAT thermogenesis (82). It will be interesting to know if this results in reduced metabolic adaptation in cold exposure. Also, whether these effects are intrinsic to Tregs can only be ascertained with Treg specific *Adrb3* deletion. Thus, these studies confirm a role of Tregs in BAT which goes beyond immunosuppression and actively associates Tregs with cold adaptations of the body by regulating lipolysis and thermogenesis.

## INTESTINAL Tregs: PRESERVING THE HOLOBIONT

The mammalian digestive system performs two very vital functions—digestion and absorption of food and shaping a gut microbial ecosystem. According to recent estimates, human colon harbors about 4 × 1013 bacteria (96, 97), a density (1011/mL) highest among any microbial habitat (98). It has a perplexing task to efficiently implement a "goldilocks" balance between two seemingly opposite events; permit absorption of nutrients but check exposure to harmful substances, guard against invasive pathogens but facilitate colonization of commensals and help them thrive.

### Architecture

Incessant provision of food and microbial antigens has resulted in typical structural adaptations in gut (99, 100) and evolution of specialized gut immune cells for immunosurveillance and maintenance of tolerance (101). The small intestine (duodenum, jejunum, and ileum) with maximum absorptive surface created by large circular folds (plicae) and finger like projections (villi and microvilli), has evolved as the primary organ for food absorption. The folding results in formation of deep invaginations, crypts of Lieberkühn, which house Paneth cells that secrete antimicrobial molecules upon exposure to bacterial antigens. Several enteroendocrine cells and mucus producing Goblet cells are also interspersed among small intestinal epithelial cells. The large intestine (cecum, colon, and rectum) harbor majority of commensals and perform the vital functions of absorbing water and vitamins while converting the undigested food into feces. The large intestinal walls are protected from luminal contents by two layers of mucus, the outer (luminal) thin mucus layer which hosts most of the bacteria, and the inner thick mucus layer which is largely sterile (102, 103).

Histologically, intestinal tract contains four layers—mucosa, submucosa, muscularis propria, and adventitia or serosa. The mucosa is the layer where most of the immune processes take place. It consists of an epithelial layer, which has scattered intraepithelial lymphocytes (IELs), underlying lamina propria (LP) and a thin muscle layer muscularis mucosa. The LP consist of non-cellular connective tissue, like collagen and elastin, blood and lymphatics, myofibroblasts, and nerve endings, and is densely packed with immune cells including mononuclear cells, plasma cells, B and T lymphocytes including Tregs, eosinophils, macrophages, and mast cells (104).

Different parts of intestine drain into separate LNs, like duodenum drains into a LN embedded in pancreatic tissue; mesenteric LNs drain jejunum, ileum, cecum, and ascending colon; two small LNs in pancreatic tissue drain transverse colon and descending colon and rectum primarily drains to caudal LNs (101). Anatomic variations largely define the antigenic variations of intestine as well. While small intestine grapples for equilibrium against food antigens, large intestines have an overwhelming load of microbial antigens. As much as these are foreign to the body, they are equally essential. Thus, for the economy of immune response it is highly desirable to assimilate those in the immunological self. The intestinal population of Tregs is arguably the most important cell type to contain the immune response against both food and innocuous microbial antigens and maintenance of intestinal homeostasis (105, 106).

## Origin of Intestinal Tregs

Like other non-lymphoid tissues, Tregs are also enriched in intestinal LP with colon harboring 25–35% (107–109) and small intestine harboring 10–15% (109, 110) Tregs among total CD4<sup>+</sup> T-cells. Apropos to the prevalent inflammatory milieu and antigenic environment, the colonic Tregs are largely developed against microbial antigens. GF mice, devoid of any microbiota, have several folds less number of colonic Tregs compared to specific pathogen-free (SPF) mice (111, 112). A long-term broad-spectrum antibiotic treatment also reduces colonic Tregs in SPF mice (111). As small intestine is seat for nutrient absorption, most of the small intestine LP (siLP) Tregs do not develop against bacterial antigens as evident by comparable number of Tregs in SPF and GF mice (109). However, once the GF mice are brought up as antigen-free (AF) mice by provision of only elemental diet post-weaning, there is a marked decrease in siLP Tregs (109). A subset of Tregs, however, remains in colon as well as small intestine of GF and AF mice, presumably specific to gastrointestinal self-antigens.

Sequencing of colonic Treg TCRs of genetically engineered mice with limited polyclonal repertoires found that TCR usage of colonic Tregs was different from that of other tissues (113). Also, there was very little similarity of TCR usage between Tregs and naïve or effector Foxp3<sup>−</sup> T-cells (113). Stable chimeras made from retroviral transduction of colonic Treg TCRs to bone marrow progenitors resulted in induction of respective TCR expressing Tregs preferentially in colon while no specific TCR bearing cells were found in thymus (113). Adoptive transfer of naïve T-cells from the transgenic lines made from colonic TCRs resulted in very efficient conversion to peripherally induced Tregs (pTregs) in mesenteric LNs and colon (114). Other than the specific TCR repertoire, high surface expression of Helios, an Ikaros family transcription factor (115), and Nrp1, a membrane-bound co-receptor for vascular endothelial growth factor and semaphorin (116, 117), has been found to be associated with tTregs. Although the precise origin of Tregs based on these markers is debatable since their expression can be upregulated in inflammatory settings (118, 119), nevertheless, most of the studies have reported enrichment of Helios<sup>−</sup> and/or Nrp1<sup>−</sup> pTregs in colonic Treg compartment (107, 114). Also, the frequency of colonic Foxp3<sup>+</sup>Helios<sup>−</sup>Nrp1<sup>−</sup> Tregs is significantly reduced in GF mice compared to SPF mice (116). Furthermore, treatment of SPF mice with broad-spectrum antibiotics decreased the Helios<sup>−</sup> colonic Tregs (111). Molecular studies on *Foxp3* gene locus have identified an enhancer, CNS1, to have a prominent role in pTreg generation in gut-associated lymphoid tissues (46). CNS1 contains a TGFβ-NFAT response element (46) and binding sites for retinoic acid receptor (RAR) and retinoid X receptor heterodimer, receptor for retinoic acid (RA) (120). It has been shown that TGFβ and RA can induce *de novo* generation of pTregs upon antigen activation *via* CD103<sup>+</sup> DCs (121, 122) (**Figure 3**). Indeed, CNS1-deficient mice have fewer LP Tregs at weaning (46). However, it was recently reported that CNS1 deficiency in colonic TCR transgenic T-cells only delays the pTreg generation and ultimately T-cells do convert to Foxp3<sup>+</sup> Tregs (114).

While pTregs are largely accepted to be the primary source of intestinal Treg population, at least under an experimental setting where generation of extra-thymic Tregs are compromised, thymically generated Tregs can migrate to intestinal LP and proliferate to fill up the niche (**Figure 3**). It has been reported that in a limited TCR model, the TCR repertoire of tTregs and colonic Tregs overlap considerably (123). Another study utilizing K14- Aβb (Keratin 14 transgenic, K14) mice, which have restricted MHCII expression in thymic cortical epithelium and thus, cannot provide peripheral MHCII signals for extra-thymic Treg

Figure 3 | Subtype differentiation of regulatory T-cells (Tregs) in large intestine. Colonic Tregs differentiate into three different subtypes based on RORγt (Rorc) and GATA3 expression along with Foxp3. Foxp3+RORγt <sup>−</sup> cells comprise ~ 15% of colonic Tregs, their function in colon is not much elucidated but presumably these are responsible for establishing tolerance to dietary antigens. Second and major subset of colonic Tregs is Foxp3+RORγt <sup>+</sup> Tregs, constituting about 50% of total colonic Tregs. These cells are primarily involved in microbial tolerance and several mechanisms contributing to their generation has been described. From left, short chain fatty acids generated by microbial fermentation can differentiate naive T cells into Foxp3+RORγt <sup>+</sup> Tregs in lamina propria (LP). Butyrate, specifically, can contribute to Foxp3 expression and function by virtue of its histone deacetylase (HDAC) inhibitory function, as well as can turn dendritic cells (DCs) into Treg generating tolerogenic cells by suppressing their IL12 and IL6 levels and expression of NFκB subunit Relb. Vitamin D metabolite 1,25-dihydroxy vitamin D3 differentiates naive T cells into Tregs by activating its nuclear receptor. TGFβ along with DC generated vitamin A metabolite all-trans retinoic acid differentiates naive T cells into Tregs in presence of microbial antigens. Foxp3+RORγt <sup>+</sup> Tregs deploy various mechanisms toward establishing dominant tolerance with IL10 production and direct suppression of Th17 cells as prominent ones. A special subset of Foxp3+RORγt <sup>+</sup> Tregs express cMAF and is instrumental in establishing tolerance to pathobionts in homeostasis. Foxp3+RORγt <sup>+</sup>cMaf+ Tregs are identified against specific epitopes of *Helicobacter hepaticus* and suppress epitope specific Th17 cells. It has been shown that in absence of pTregs, early seeding of thymic Tregs (tTregs) happens in colonic LP. These tTregs expand into a niche which is independent of MHCII and IL2 signaling but depends on microbial antigens. These Tregs subsequently fill up the LP Treg compartment. However, whether these can differentiate into all Treg subtypes is not known. A third subtype of Tregs express GATA3 (not RORγt) and constitutes about a third of total Tregs. These Foxp3+GATA3+ Tregs express ST2 receptor which bind to tissue damage-induced alarmin IL33. Foxp3+GATA3+ Tregs are instrumental in suppressing inflammation and facilitating tissue repair by secretion of amphiregulin upon tissue damage.

generation, showed that while Treg population was significantly decreased in mLNs and spleen, the intestinal Treg compartment was rather intact (124). The LP Treg niche of these mice was found to be inhabited by Tregs of thymic origin at young age. This niche filling phenomenon was not IL2 dependent but was induced by microbial presence, as broad-spectrum antibiotic treatment decreased both large and siLP-Tregs (siTregs) (**Figure 3**). At an older age however; newly generated tTregs were excluded from the LP, presumably due to already occupied Treg niche (124). Taken together, therefore, the available data suggest a peripheral origin of majority of intestinal Tregs and thymic origin of a small subset. However, these mechanisms of origin and development cannot be mutually exclusive and contributions from different pathways are likely to fine tune the ultimate composition of the intestinal Treg compartment.

### Tregs Specialize into Multiple Subsets in Intestines

Transcriptomic and functional analyses of intestinal Tregs have largely identified three specialized subsets. Based on transcription factor and surface molecular expression these subsets are GATA3<sup>+</sup>Helios<sup>+</sup>(Nrp1<sup>+</sup>), retinoic acid receptor related orphan receptor γt (RORγt) expressing RORγt <sup>+</sup>Helios<sup>−</sup> and RORγt − Nrp1<sup>−</sup>(Helios<sup>−</sup>) subsets (**Figures 3** and **4A**).

GATA3 is expressed by about a third of intestinal Tregs and can be induced in CD4<sup>+</sup>Foxp3<sup>+</sup> Tregs upon TCR engagement (110). That most of these Tregs are Helios<sup>+</sup> and are unaffected by microbial presence suggests their thymic origin (107). However, upon TCR engagement, GATA3 can be induced in CD4<sup>+</sup>Foxp3<sup>−</sup> naïve T-cells as well, in both *in vivo* and Treg polarizing *in vitro* settings (110). While expression of GATA3 is not required under homeostatic conditions, under inflammatory conditions lack of GATA3 in Tregs hampers their accumulation at inflammatory sites (110). GATA3 and Foxp3 interact both at protein and gene levels in Tregs (47). GATA3 binds to *Foxp3* locus and its deletion in Tregs reduces Foxp3 expression (47, 125). It occupies significant number of genes, which are direct targets of Foxp3 and thus, collaborates with Foxp3 to establish Treg gene expression program. Indeed, Treg-specific GATA3 deletion results in intestinal pathology with heightened Th2 cytokine production from large intestinal effector T-cells (47). Most of the GATA3<sup>+</sup> Tregs in colon express ST2, which is regulated by GATA3 expression in a feed-forward manner (108). Alarmins like IL33 are produced upon local tissue damage (126, 127) and thus limit self-injury in part by activating GATA3<sup>+</sup>ST2<sup>+</sup> Tregs (108). ST2<sup>+</sup> Tregs show enhanced production and activation of IL10 and TGFβ (128). IL23 on the other hand appears to regulate the effect of IL33 in thymus-derived Tregs *via* STAT3 signaling (108).

CD4<sup>+</sup>Foxp3<sup>+</sup>RORγt <sup>+</sup> Tregs in the intestine are another subset of Tregs specialized toward microbial immunity. These comprise about 50% of total Tregs in colon, which are readily lost in GF mice or upon antibiotic treatment (107, 129) (**Figure 3**). A small population consisting of about 10–15% of Tregs is also found in small intestine (107) (**Figure 4**). Most of the RORγt <sup>+</sup> Tregs do not express Helios (107, 129) or Nrp1 (130), suggestive of their

Figure 4 | Subtype differentiation of regulatory T-cells (Tregs) in small intestine. (A) Small intestine lamina propria (LP) Tregs are differentiated into all three subtypes mentioned in Figure 3; however, Foxp3+RORγt <sup>−</sup> Tregs are primary population forming about 50% of siTreg compartment. They are primarily generated in response to dietary macromolecules and are proposed to be helpful in containing childhood allergies. Foxp3+RORγt <sup>+</sup> and Foxp3+GATA3+ Tregs are about 15% and 35% of total siTregs, respectively. (B) Some siTregs continuously traverse in and out of small intestinal epithelial compartment. While most of these Tregs return back to LP, a fraction of these lose their Foxp3 expression and repress ThPOK expression inside epithelial layer. They subsequently, de-repress *CD8a* locus to convert into CD4+CD8αα+ double positive intra epithelial lymphocytes.

extra-thymic origin. However, these cells have substantially reduced CpG methylation within the CNS2 enhancer region of *Foxp3* locus, which is known to be well correlated with stable Foxp3 expression (130, 131). Not all bacteria can elicit similar population of RORγt <sup>+</sup> Tregs, as it was found that a gradation of Treg inducing capacity exists (107). Mechanistic insights on why certain bacteria have superior capacity of inducing intestinal Tregs than others have started to emerge only recently. It has been reported that short chain fatty acids produced upon fermentation of starch and other dietary fibers by clostridia strains, mainly butyrate and propionate but not acetate, can contribute to colonic Treg generation. Mechanistically, this is attributed to their histone deacetylase inhibitory properties (132) resulting in increased acetylation of *Foxp3* locus (133, 134) (**Figure 3**). Apart from directly acting on T-cells, butyrate also affects DC ability to induce Treg differentiation. Knockdown of *Relb*, which encodes NFκB subunit, has been shown to generate tolerogenic DCs by inhibiting their maturation (135, 136). Indeed, *in vitro* treatment with butyrate represses lipopolysachharide response genes, including *Il12*, *Il6*, and *Relb* in DCs (133) (**Figure 3**). On a translational note, in human IBD patients, colonic butyrate producing bacteria are decreased (137) and mucosal butyrate transporter, monocarboxylate transporter 1, is downregulated (138). It is also possible that colonic Tregs are generated in an antigen-specific manner. Indeed, colonic Treg TCRs have been reported to interact with colonic bacteria *in vitro* (113). Very recently, colonic T cells with TCRs cognate to epitopes of a pathobiont *Helicobacter hepaticus* are shown to induce pTregs under homeostatic conditions (139). This study establishes the role of colonic pTregs in induction and maintenance of tolerance to pathobionts as well. Surprisingly, it was found that although such Tregs are RORγt <sup>+</sup>, their major functional suppressive capabilities are implemented by expression of the transcription factor cMAF (139) (**Figure 3**). cMAF offsets Th17 polarization by producing IL10 downstream to TGFβ1-STAT3 signaling (140). *H. hepaticus* colonized mice with Treg specific *Rorc* deletion had no significant increase in colonic Th17 cells while mice with Treg specific *Maf* deletion had significantly high Th17 cells (139). Thus, RORγt + pTregs in colon establish tolerance to commensals as well as pathobionts and suppress inflammation in a cMAF-dependent manner.

The third Treg subtype (CD4<sup>+</sup>Foxp3<sup>+</sup>RORγt <sup>−</sup>) was identified very recently. Most of these cells express low levels of Nrp1 and thus are, supposedly, pTregs. These cells constitute about 50% of siLP Tregs and 15% of colonic Tregs (109) (**Figures 3** and **4A**). Their localization suggests that these are primarily generated against dietary antigens. Indeed, long-term antibiotic treatment of SPF mice could not reduce their numbers in intestine while RORγt <sup>+</sup>Nrp1lo pTregs were reduced several folds. On the other hand, weaning SPF mice onto AF diet dramatically reduced RORγt <sup>−</sup>Nrp1<sup>−</sup> pTregs (109). Adoptive transfer of naïve OTII CD4<sup>+</sup> T-cells in mice on AF diet primarily elicits Th1 cell immune response while Th17 and Th2 responses were comparable to SPF animals. However, pTregs generated in this model were primarily RORγt <sup>+</sup>Nrp1<sup>−</sup> pTregs, suggesting that dietary antigens can also generate RORγt <sup>+</sup> pTregs in the absence of microbiota (109).

While TGFβ is by far the most important factor as far as generation and intestinal accumulation of pTregs is concerned (141–143), a generalized modulation of intestinal pTregs can be achieved by several other factors like dietary vitamin A, vitamin D, Niacin (Vitamin B3), Folic acid (Vitamin B9), and tryptophan [reviewed in Ref. (144)]. All-trans retinoic acid, a metabolite of Vitamin A, produced by DCs facilitates *de novo* generation of Foxp3<sup>+</sup> Tregs from naive CD4<sup>+</sup>CD25<sup>−</sup> T-cell populations in mice (121, 145) (**Figure 3**). It also plays an important role in upregulating gut-homing markers CCR9 and CD103 (integrin αE) on pTregs (146). Feeding mice with Vitamin A-deficient diet or treatment with RAR inhibitors reduces the RORγt <sup>+</sup> pTregs in colon while GATA3<sup>+</sup> Tregs are not affected (129). Similarly, RORγt <sup>−</sup> pTregs in small intestine are also not affected by vitamin A or RA (109). Vitamin D metabolite 1, 25-dihyroxyvitamin D3 binds to Vitamin D3 nuclear receptor in CD4+ T-cells and promotes *Foxp3* expression (147) (**Figure 3**). Recently, it was shown that in human patients of ulcerative colitis a Vitamin D3 agonist can convert CD4<sup>+</sup> T-cells to pTregs (148).

### Functions of Intestinal Tregs

Foxp3<sup>+</sup>GATA3<sup>+</sup> intestinal Tregs express high level of ST2 (108). These Tregs express high levels of tissue repair factor, an EGF like growth factor Amphiregulin (108, 149) (**Figures 3** and **4A**). It appears that Amphiregulin mediated tissue repair might be a generalized mechanism employed by tissue-resident Tregs as exemplified by its evolving role in lung resident as well as intratumoral Tregs (150, 151). Foxp3<sup>+</sup>RORγt <sup>+</sup> Tregs express increased levels of ICOS, CTLA4, and the nucleotidases CD39 and CD73 altogether, indicating robust regulatory functions (107, 129). Interestingly, Foxp3<sup>+</sup>RORγt + Tregs have been implicated in regulating both Th2 and Th1/Th17 mediated immunity in two independent studies, implicating animal housing conditions as an important determinant of the type of immune response (107, 129). siLP Foxp3<sup>+</sup>RORγt <sup>−</sup> Tregs primarily work toward containment of Th1 immunity. When OTII T-cells are transferred in AF mice, T-cells primarily convert to IFNγ+ OTII cells and induced Tregs are mainly Tbet<sup>+</sup>, although the extent of such RORγt <sup>−</sup> pTreg induction in AF conditions was severely compromised compared to SPF conditions (109). As a consequence, in an experimental model of food allergy, where SPF BALB/c mice were weaned onto an amino acid diet, higher diarrheal instances were reported than mice on normal chow (109). While these results underscore an important function of RORγt <sup>−</sup> pTregs in curbing food allergy, further transcriptomic and phenotypic analyses might provide additional clues to their functions.

Of note, one more function of siTregs was identified recently while looking at the IEL population. It was observed that LP Tregs migrate to epithelial compartment as well, where a fraction of them lose Foxp3 expression (152). Further, these Tregs then give up ThPOK expression leading to de-repression of *Cd8* locus and thus, get converted into a CD4<sup>+</sup>CD8αα+ IELs (152) (**Figure 4B**). IELs have cytotoxic as well as immunoregulatory machinery suggesting a role in both mucosal barrier maintenance and elimination of stressed intestinal epithelial cells (153).

# SKIN Tregs: KEEPS YOUR HAIR ON

Skin is the largest organ amounting to almost 15% of adult human body weight. Being our exterior, it is always exposed to environmental, microbial, physical, and chemical insults. Skin also harbors more than 1012 bacteria/m2 (154) in the surface intercorneocytic spaces.

### Skin Architecture

Anatomically, skin is composed of three layers, the outer epidermis, middle dermis, and inner subcutaneous tissue layer (155). The terminally differentiated keratinocytes in epidermis synthesize long, thread-like protective protein keratin and form a physical barrier. Products of various sweat and sebaceous glands interspersed at epidermal–dermal junction along with antimicrobial peptides develop an acidic hydrophilic skin which acts as a biochemical barrier (154). The epidermal–dermal junction also hosts hair follicles. Cellular component of epidermis comprises of Langerhans cells (specialized skin DCs) and T lymphocytes. The dermis is composed of layers of thick and thin collagen fibers which provide mechanical framework to host blood vessels and various immune cells like dermal DCs, αβ and γδ T-cells, NK cells, B cells, macrophages, and mast cells (154). Understandably, given the exposed nature of skin, it is highly vulnerable to overzealous immune responses against skin commensals and self-antigens. Tregs are an important component of establishing tolerance and homeostasis in the skin. Indeed, both scurfy mouse and human IPEX patients present fulminating immune responses in skin. A study, examining children with IPEX syndrome reported that more than 70% children presented Atopic Dermatitis and eczema with 1.5 months as median age of onset of symptoms (156).

### Skin Treg Origin and Accumulation

Normally, 30–50% and 20–30% of total CD4<sup>+</sup> T-cells are Tregs in mouse and human skin, respectively (157, 158). It is difficult to establish the origin of cutaneous Tregs, given the paucity of specific information. However, a wave of Tregs has been shown to populate the skin in early neonatal period in a skin bacterial colonization model in mice (159). Furthermore, restricting lymphoid emigration of T-cells by treating with sphingosine-1-phosphate receptor antagonist FTY720 resulted in preferential accumulation of Tregs in thymus instead of skin draining LNs, suggesting their migration directly from thymus (159). In humans, the cutaneous Tregs and Tconv cells share very few TCRβ sequences and these Tregs present a fully demethylated *FOXP3* CNS2 region, suggestive of their stability and thymic origin (157, 160, 161). This is little surprising as skin Tregs establish tolerance to not only self-antigens but also to commensals. How are then commensal antigen-specific Tregs generated in thymus? One possibility appears to be plasmacytoid DCs (pDCs) that have been shown to be able to take up innocuous peripheral antigens to thymus (162) and LNs (163) to induce tolerance. Generally, pDCs are not present in skin but can accumulate in the presence of inflammatory conditions (164, 165). Another probability is that some tTregs have TCRs with sufficient crossreactivity to microbial antigens and thus can specifically expand and accumulate at sites where antigen is present. However, if indeed skin Tregs are thymic by origin, the concerning mechanisms remain to be elucidated.

Modeling of inducible expression of a self-antigen, by fusing transferrin receptor transmembrane domain, GFP and amino acids 230–359 of chicken ovalbumin in mouse epidermis, revealed that circulating Tregs are not able to suppress primary immune response against OVA, though it initiated activation of Tregs (166). The inflammation resolved spontaneously and any subsequent antigen expression led to an attenuated and short immune response. Further analysis revealed that a fraction of Tregs persisted in the skin which expressed low level of CD25 but higher KLRG1, CTLA4, and CD127, akin to the memory T-cells (166). It is to be noted here that a recent study examining the transcriptional, epigenomic, and functional changes in inflammation experienced Tregs employing the Foxp3DTR system, presented that Tregs revert most of the activation induced changes and lose the accentuated suppressive ability over time (167). An earlier report revealed that cutaneous Tregs express CCR4 and adhesion molecule Integrin αE, CD103 (168). In a mixed bone marrow chimera study, CCR4-deficient Tregs could not reconstitute the skin Treg compartment (168). CCL17 and CCL22 are known chemokine ligands for CCR4 (169, 170), which are differentially expressed in inflamed skin mainly by endothelial cells and dermal DCs, respectively (171). These molecules sequentially manage T-cell homing to skin, where CCL17 promotes vascular recognition and extravasation and CCL22 guides subsequent migration in skin (172). More than 70% skin Tregs express GATA3, although its deletion in Tregs does not alter Treg profile or cause overt skin related phenotype under homeostatic conditions (110).

Since skin is heavily exposed to commensals as well as pathogens, it is imperative to speculate that a fine tuning of effector and suppressive immune responses has evolved. The commensal microbiota residing within skin has been shown to calibrate barrier immunity (173, 174). To identify mechanisms behind development of tolerance to skin commensals, Scharschmidt et al. performed some elegant experiments with model peptide antigen expressing *Staphylococcous epidermidis*, a human skin commensal which efficiently colonizes mouse skin (159). Surprisingly, skin colonization in adult mice did not evoke any tolerance to bacteria, as seen by inflammatory response upon re-challenge. However, when neonatal mice were colonized for a week during postnatal week 2, there was an appreciable attenuation of inflammatory response upon re-challenge after 3–4 weeks. Subdued response was associated with marked enrichment of antigen-specific Tregs in skin and draining LNs and the tolerance could be reversed upon FTY720 treatment (159). Another study has also shown that Tregs, generated very early in life in a defined perinatal window, play a very distinct role in maintaining selftolerance (175). Thus, in mice, an abrupt wave of Treg infiltration occurs in a defined early postnatal period to establish dominant tolerance toward skin commensals (159) (**Figure 5**).

Both Tregs and skin commensals are localized to hair follicles and in mouse, the time of Treg infiltration (week 2 postnatal) is coincident with hair follicle development (176, 177). Thus, it was speculated that hair follicles might have a role in ingress of

Tregs in skin. It has been shown that chemokines from different parts of hair follicles like CCL2 from isthmus, CCL20 from infundibulum, and CCL8 from bulge keratinocytes generate specific type of Langerhans cells (178). Similarly, in a mouse model of skin specific hair follicle morphogenesis arrest, skin Treg population was reduced without affecting Treg population in intestine or draining LNs in neonatal mice (179). Further investigations revealed that neonatal SPF mouse skin produces high amount of chemokine CCL20 whose receptor CCR6 was found to be enriched on cutaneous Tregs (**Figure 5**). It was confirmed by adoptive transfer experiments that CCR6-deficient Tregs were at a competitive disadvantage to reconstitute skin T-cell compartment (179). Tregs in neonates also express high CCR8 and its ligand CCL22 is, in turn, expressed in skin. However, their contribution in establishing and/or maintenance of skin Tregs is yet to be examined. CCL20 mRNA expression also increases in human fetal skin explants upon exposure to cutaneous commensals and bacterial components (179). Thus, tissue morphogenesis (hair

signaling triggers the active phase of hair growth cycle (anagen). Tregs are reduced in anagen phase around HFSCs.

follicle generation and subsequent chemokine production by the cells of hair follicle) and commensals are shown to cooperate in developing a tissue-specific immune tolerance. It would be interesting to extrapolate this model to other microbe inhabited organ systems. Whether, these mechanisms sustain throughout the life span particularly in adult life is not known. Earlier, 8- to 12-week-old GF mice were reported to have about twofold more cutaneous Tregs than SPF animals (180). It is possible that yet unknown factors other than commensals increase Treg population in adult GF skin or a lack of tuned immune response against commensals diverts immune resources toward self-antigens and in turn, progressively enrich self-antigen-specific Tregs in skin.

Short wavelength Ultraviolet (UVB) exposure also enriches and accumulates Tregs in mouse skin (181). UVB exposure damages self-RNA in keratinocytes which is sensed by TLR3 to generate an inflammatory response (182). Indeed, the Tregs accumulating post-UVB exposure were Nrp1<sup>+</sup> with highly demethylated *Foxp3* CNS2 (181). These Tregs highly expressed homing molecules like CD103, CCR4, and P-lig and thus, were able to migrate to the non-UVB exposed parts of the skin as well (181) (**Figure 6**). Interestingly, UVB phototherapy is an effective treatment in autoimmune skin conditions, like psoriasis (183) and atopic dermatitis (184).

### Tissue Adaptations and Homeostatic Functions

True to the nature of tissue-resident Tregs where they have been shown to not only develop immune tolerance but also get accustomed to and help local physiology, cutaneous Tregs in hair follicles have been found to modulate hair follicle stem cells (HFSCs) in addition to establishing tolerance to commensals. In mouse skin, an abundance of Tregs was found during telogen phase of hair follicle growth which is characterized by quiescence only to be followed by an active proliferation phase called anagen. Treg population decreases during anagen phase. And as these phases alternate in mouse skin so are the numbers of hair follicle associated Tregs (158) (**Figure 5**). Animals with transient depletion of Tregs during telogen, failed to progress to anagen phase and could not re-grow hairs (158). It was not inflammation related, as transient depletion of Tregs did not elicit any major inflammatory event in skin and co-depletion of either effector CD4<sup>+</sup>, CD8<sup>+</sup>, Gr-1 expressing neutrophils or CD11C<sup>+</sup> myeloid cells did not rescue HFSC proliferation (158). This suggested a non-immune role of Tregs in HFSC proliferation. Akin to Tregs in hematopoietic system which form an immune-privileged site to provide a protective niche to hematopoietic stem cells (185), hair follicle Tregs were found to co-localize with HFSCs. These Tregs highly express "*jagged1*" which encodes a ligand for notch signaling pathway responsible for HFSC proliferation (158) (**Figure 5**).

Regulatory T-cells play an important role in cutaneous wound healing as well. They accumulate in large numbers at the site soon after a wound injury in skin (186). These Tregs are of activated

Figure 6 | Cutaneous regulatory T-cells (Tregs) actively repair skin tissue damage. Tregs migrate to skin by virtue of sequential interactions of chemokines CCL17 and CCL22 with their receptor CCR4. CCL17 is secreted by endothelial cells in inflamed skin and helps in extravasation of Tregs, CCL22 manages subsequent migration of Tregs. In case of skin wound-induced inflammation, Tregs acquire a highly activated phenotype with higher surface expression of CTLA4 and ICOS. These Tregs actively suppress IFNγ production from inflammatory Ly6C+ macrophages. Also, Tregs express high amount of EGFR, helping in their tissue repair ability. In case of short wave UVB light induced skin damage, self mRNA from keratinocytes are taken up by CD103+ dendritic cells (DCs) which induce the cutaneous inflammation. These DCs are suppressed by Tregs which have high surface expression of CD103 and P-lig and thus can migrate into non-inflamed skin as well, subsequent to UV exposure.

Sharma and Rudra Tregs in Tissue Homeostasis

phenotype with high CD25, CTLA4, and ICOS expression and limit the IFNγ producing T-cells and inflammatory macrophages in wounds (186). The Tregs involved in cutaneous wound healing were shown to be dependent on EGFR pathway (186) as in lung and muscle-healing Tregs (150, 187, 188) (**Figure 6**).

# TUMOR-INFILTRATING REGULATORY T-CELLS (TI-Tregs): A BATTLE WON, WAR LOST

Tumors are wounds that do not heal (189–191). Solid tumors, in particular, heterogeneously indulge in various stages of wound healing, which provide essential growth factors for the tumor growth. This hijack of natural processes results in heightened inflammation and its subsequent resolution in the tumor microenvironment, presumably setting up a vicious cycle. Inflammation on one hand provides growth and metastasis opportunities; resolution of inflammation helps the tumor to escape antitumor immunity. Therefore, it is not surprising that many solid tumors including hepatocellular (192), gastric (193), lung (194), breast (195), ovarian (196), cervical (197), and melanomas (198) summon comparatively large numbers of Tregs which sometimes account for even more than 50% of CD4<sup>+</sup> T-cell compartment (199). For most of the tumors, the presence of high number of Tregs indicates a guarded to grave prognosis. However, several studies reported a favorable role of FOXP3<sup>+</sup> T-cells in colorectal carcinomas (CRC) (200–202). It is to be noted that FOXP3 expression is not exclusive to *bona fide* Tregs in humans, often, effector T-cells express FOXP3, albeit transiently (203, 204). Based on expression levels of FOXP3 and Protein tyrosine phosphatase isoform A (CD45RA), human peripheral blood FOXP3<sup>+</sup>CD4<sup>+</sup> T-cells can be classified into FOXP3hiCD45RA<sup>−</sup> *bona fide* Tregs which are highly suppressive and phenotypically eTregs; FOXP3loCD45RA<sup>+</sup> naïve T-cells and FOXP3loCD45RA<sup>−</sup> effector T-cells which are not suppressive in an *in vitro* suppression assay (203). Indeed, careful analysis of TILs in human CRC by Saito et al. (205) revealed the heterogeneity of FOXP3 expression. The authors identified that CRC, where higher expression of FOXP3 was associated with favorable outcomes, were actually infiltrated more with FOXP3loCD45RA<sup>+</sup> effector T-cells and upregulated inflammatory genes like *Il12a, Il12b*, *Tgfb1*, and *Tnf.* Higher infiltration of FOXP3hiCD45RA<sup>−</sup> cells resulted in poor prognosis and lower disease-free survival (205) as reported for other tumors.

### Origin and Accumulation of TI-Tregs

As tTregs and pTregs differ in their stability, conclusive information about origin of TI-Tregs can be very valuable to design TI-Treg specific therapies in cancers. Tumors drive immune responses against tumor-associated self-antigens as well as tumor-specific neo-self antigens. Thus, in theory, Tregs against self-antigens (tTregs) and pTregs against neo-self antigens are possible. Presence of high levels of TGFβ in most solid tumors reinforces the idea of generation of pTregs in tumor microenvironments (206, 207). However, TI-Tregs in several murine tumors have been shown to express high levels of Nrp1 and Helios proteins, suggestive of a thymic origin (208). Attempts for *in situ* conversion of conventional CD4<sup>+</sup> T-cells to Tregs against tumors expressing model antigens did result in intratumoral pTregs generation (209, 210), but monoclonal populations of antigen-specific T-cells do not recapitulate physiological conditions where antigen-specific T-cells represent less than 5% of TILs (211–213). TCR repertoire analyses have revealed almost no overlap in Foxp3<sup>+</sup> and Foxp3<sup>−</sup> T-cells in autochthonous prostate tumors (214), carcinogen-induced tumors (215), and in heterotopic transplanted tumors in mice (216). Further, these studies confirmed that there is enrichment and expansion of selective TCR bearing Tregs inside tumor microenvironment (214, 216). In human breast cancers (217) and hepatitis B virus positive hepatocellular carcinomas (HCC) (192), very low TCR repertoire overlap between TI-Tregs and conventional T-cells suggests little to no conversion of conventional CD4<sup>+</sup> T-cells into Tregs (**Figure 7A**). Malchow et al. (214) developed a transgenic mouse expressing the model oncogene SV40 T antigen in prostate and a fixed TCRβ chain (TRAMP-Foxp3eGFPTCRβTg). They found that Tregs expressing a single TCR (MJ23), reactive to a normally expressed prostate antigen, consistently populated the tumors (214). This TCR was able to drive a tTreg clone development. However, a deficiency of the transcription factor autoimmune regulator abolished development of these clones (214), suggesting that at least in these experimental settings, TI-Tregs are generated in thymus against a normal tissue expressed self-antigen (**Figure 7A**). Recently, two MHCII restricted natural self-antigen ligands of MJ23 Tregs were discovered. Both of these ligands are non-overlapping peptides originating from same prostatic protein (Tcaf3) and while one is expressed in mouse prostate tumors (MJ23), the other is associated with prostatic autoimmune lesions (SP33) (218). Another study focusing on epigenetic hallmarks of tTregs found that TI-Tregs had consistently hypomethylated *Foxp3* CNS2 in various orthotopic and heterotopic transplanted tumor models, even at different time points of tumor growth (219). These findings were further confirmed in TI-Tregs from different human tumors. It is to be noted that there have been equivocal reports about the demethylated CNS2 being specific for tTregs, since *Foxp3* CNS2 region in pTregs has also been shown to be demethylated (220, 221), most likely upon eventual stabilization following its *de novo* induction.

Overall, available evidences largely point toward higher enrichment and expansion of tTregs inside solid tumors. However, why should not higher TGFβ levels and an activating tumor environment drive pTreg generation and expansion is a baffling question. Probably more conclusive lineage tracing experiments across tumors can be proven more insightful.

## TI-Tregs Add Another Layer of Diversification

Recent studies analyzing transcriptome of TI-Tregs from human cancers have identified that though TI-Tregs are largely similar to normal tissue-resident Tregs, these have some specific characteristics and molecular patterns which can be utilized for selective therapy (192, 217, 222). Plitas et al. (217) found that breast cancers, with rather aggressive phenotype, were enriched

Figure 7 | Origin, accumulation, and functional potentiation of tumor-infiltrating regulatory T-cells (Tregs). (A) Tumor-specific antigens can be expressed by thymic epithelial cells in an *Aire*-dependent manner, which then select the tumor antigen-specific Tregs. These Tregs then expand in tumor draining lymph nodes (LNs) with the help of dendritic cells (DCs). Tregs can also be directly recruited to tumors and undergo expansion there. While intratumoral conversion of effector T-cells to pTregs is likely, the extent to which this occurs under physiological conditions is not completely understood. (B) A high number of TI-Tregs are apoptotic because of suppressed expression of *Nrf2*. These Tregs as well as dying tumor cells release copious amounts of ATP, which is converted to adenosine by Treg ectoenzymes CD39 and CD73 in a sequential manner. The resulting adenosine is highly suppressive to tumor-infiltrating CD8+ effector T-cells. Further, Tregs also produce Amphiregulin in certain tumor types, which help in tumor progression. (C) More than 50% of TI-Tregs express surface Neuropilin1 (Nrp1), which is a receptor for Semaphorin 4a. Upon ligation with Semaphorin4a, Nrp1 activates lipid phosphatase Phosphatase and tensin homolog (PTEN) which promptly dephosphorylates AKT rendering its sequestration in cytosol and nuclear retention of Foxo1/3 transcription factors which help in stabilization and survival of Tregs. A loss of Nrp1 renders Treg highly susceptible to IFNγ and such Tregs also produce high amount of IFNγ and HIF1α ("Fragile" Tregs). The resultant high IFNγ environment, reciprocally, can induce "fragility" in other Nrp1-sufficient Tregs as well, setting up a vicious cycle.

for Tregs, which were highly suppressive in a microsuppression assay and were highly proliferative based on increased expression of nuclear protein Ki67, a cellular marker for proliferation (223). Further transcriptome analysis of TI-Tregs from breast cancers and gastric cancers as well as brain metastases of NSCLC and liver metastasis of CRC, suggested a signature gene set for these cells (222). TI-Tregs expressed BATF, CCR8, CD30, IL1R2, IL21R, PDL1, and PDL2 along with FOXP3 and IL2Ra at a very high level (222). Recently, BATF was shown to be involved in context dependent gene set expression in tissue Tregs (224). In the IPEX patients, a gain-of-function mutation in FOXP3 locus (A384T) results in expanded DNA binding specificities of FOXP3. Its altered binding to *BATF* locus repressed BATF expression leading to repressed GATA3, ST2, and CCR4 expression in Tregs (224). These genes are significant in converting cTregs to eTregs, therefore, their decreased expression led to a widespread tissuespecific autoimmunity (224).

Genome-wide transcriptome analyses identified *MAGEH1*, Melanoma antigen family H1 gene, encoding a putative E3 ubiquitin ligase potentially regulating TI-Treg survival and function; Chemokine (C-C motif) receptor 8 (CCR8), known receptor for chemokines CCL1 (225), CCL8 (226), CCL16 (227), and CCL18 (228) in humans; CD177, a glycosyl-phosphatidylinositol-linked cell surface glycoprotein that can bind platelet endothelial cell adhesion molecule-1 and is known for neutrophil transmigration and survival (229); and *LAYN*, encoding a novel c-type lectin surface receptor layilin, a proposed receptor for Hyaluronan (230). However, subsequently, layilin has been found highly expressed on tumor-infiltrating cytotoxic CD8<sup>+</sup> T-cells as well, particularly those with an exhaustive (highly expressing *CTLA4*, *PDCD1*, and *HAVCR2*) phenotype (192). Enrichment of *LAYN*, *MAGEH1*, and *CCR8* in whole tumor samples correlated significantly with reduced 5-year survival rate of CRC and NSCLC patients (222). CCR8 expression was exclusively enriched on TI-Tregs, whereas CCR2, CCR4, and CCR5 expressions were found on other tumor-infiltrating and/or peripheral blood Tconv cells as well. Indeed, a CCR4 depleting antibody has been shown to deplete both Tregs and Tconv cells (231). Some Tregs in draining LNs also expressed high CCR8 which might be ones earmarked for tumor infiltration or Tregs in micrometastases inside LNs (217). As peripheral blood and/or LN Tregs do not express CCR8, its importance in recruitment of Tregs to tumors is not appreciated. It is possible that CCR8 is expressed to retain Tregs in tumors. Indeed, CCR8 ligands like CCL1 and CCL18 are highly transcribed in tumor-infiltrating myeloid cells (217). Whether, CCR8 expression is an indicator of highly suppressive TI-Tregs or it has further functional importance is not yet known. However, human Tregs exposed to CCR8 ligand CCL1 and not CCL8, CCL16, or CCL18 induce surface CCR8 expression *via* a STAT3-mediated pathway (232). Such cells, subsequently, upregulate their expression of FOXP3, CD39, Granzyme B, and IL10 and are functionally more suppressive in a microsuppression assay and in a mouse model of multiple sclerosis (232).

### Function of TI-Tregs

There is a repertoire of known and yet unknown mechanisms which Tregs utilize to suppress an immune response. TI-Tregs also use similar mechanisms which include production of immunosuppressive cytokines like IL10 and TGFβ (233, 234); sequestration of IL2 (235); direct cytolysis of target lymphocytes using granzyme B and perforin (236); contact based immunosuppression using surface inhibitory molecules like CTLA4 (237), PD1 (238, 239), LAG3 (240), TIM3 (241), and CD39/CD73-generated adenosine-mediated T-cell suppression *via* adenosine receptor 2A (242, 243). However, how the TI-Tregs have highly accentuated suppressive response is not very well understood. A large accumulation of Tregs might help in a collective exaggerated suppression, but it cannot explain individual potentiation. Recently, it was shown that TI-Tregs are highly apoptotic on account of comparatively low expression of the transcription factor Nuclear factor like 2 (NRF2) (244). NRF2 regulates antioxidant defense system in macrophages and epithelial cells (245). A lack of NRF2 makes TI-Tregs more apoptotic in high oxidative stress tumor microenvironment. But, owing to increased release of ATP and high CD73/CD39 expression, apoptotic TI-Tregs generate large amount of adenosine and thus, become even more suppressive (244) (**Figure 7B**). It is to be noted though that earlier Imatinib induced apoptosis of TI-Tregs was shown to enhance antitumor immunity (246).

TI-Tregs in human breast cancers (217) and HCC (192) highly express *Il1r2* gene encoding a decoy IL1 receptor. Also, TI-Tregs are found to be highly stable owing to the enhanced expression of lipid phosphatase Phosphatase and tensin homolog (PTEN) and VEGF receptor Nrp1 (199, 238, 247). Binding of Nrp1 to its ligand Semaphorin4a increased Foxo1 and Foxo3 nuclear localization by inhibiting AKT phosphorylation which stabilized Treg signature genes and antiapoptotic genes (247) (**Figure 7C**). AKT dephosphorylation was achieved by activation of PTEN by Nrp1. Indeed, mice with Treg specific PTEN deletion generate an accentuated antitumor immune response (238). Nrp1 expression is primarily important for TI-Tregs as the loss of Nrp1 even from a fraction of Tregs under appropriate experimental conditions, rendered all the Tregs (including those that were Nrp1 sufficient) "fragile" (199). This observation emphasizes that Tregs can not only modulate other immune cells but can phenotypically influence other Tregs as well. The TI-Treg fragility was shown to be induced by IFNγ production by Nrp1-deficient Tregs and exogenous IFNγ (199) (**Figure 7C**). The authors further show that HIF1α was a major factor induced in Nrp1-deficient and fragile Tregs, and both HIF1α and IFNγ can be induced by hypoxia (199). However, as most of the solid tumors become progressively hypoxic (248, 249), whether this phenomenon is prevalent in progressive tumors and if so, whether it is efficient for a significant regeneration of antitumor immune response, remains to be seen. TI-Tregs have been shown to highly express receptor activator of nuclear factor κB ligand (RANKL), which upon binding to its receptor RANK expressed in mammary carcinoma cells increases lung metastasis (250). RANKL has also been implicated in renewal of breast cancer progenitor cells (251) and metastasis of prostate cancers (252) by modulating protein kinase inhibitor of nuclear factor κB kinase α (IKKα). Overall, these findings suggest that there is a specific phenotypic and functional identity to TI-Tregs and thus, it is possible to selectively target them for triggering efficient antitumor immunity.

### OTHER Tregs IN TISSUE INFLAMMATION AND HOMEOSTASIS

As the diversity in characteristics and functions of tissue Tregs is being unraveled, several other interesting populations have been described which deserve more detailed phenotypic and functional characterization.

### Regeneration Powerhouse

It was reported earlier that in a non-inflammatory model of regenerative alveologenesis, Tregs enhanced epithelial proliferation. A Treg coculture with type II alveolar cells (AT2) increased their proliferation in CD103-dependent manner (253). In accordance to these findings, a distinct population of Tregs expressing high levels of pro-inflammatory cytokine IL18 receptor (IL18R) and ST2 has been described in lungs (150). IL18R<sup>+</sup> Tregs expand early in the course of a lung injury and enhance tissue repair by producing a large amounts of tissue repair protein amphiregulin in an "innate" manner, independent of TCR engagement (150) (**Figure 8A**). In animals with Treg specific amphiregulin deficiency, a rapid decline in lung functions was observed upon intranasal influenza virus infection, while antiviral immune

Figure 8 | Emerging evidences highlight a compulsory requirement of regulatory T-cells (Tregs) in tissue regeneration and repair. (A) Both lungs and muscles contain population of Tregs which proliferate vigorously upon tissue injury. Lung reparative Tregs respond to both inflammatory IL18 and alarmin IL33 and produce amphiregulin in a TCR-independent manner. Muscle Tregs respond to IL33 produced upon muscle damage and produce amphiregulin for subsequent repair. (B) In zebrafish, mammalian *Foxp3* ortholog *Foxp3a* expressing Tregs (zTregs) are primarily present in kidney. However, upon tissue injury, they soon accumulate at the site of the injury. Apart from anti-inflammatory IL10 production, zTregs co-localize with organ progenitor cells and provide tissue-specific growth factors to progenitors like neurotrophin3 for neural progenitors in spinal cord, neuregulin1 for cardiomyocytes in heart, and insulin-like growth factor1 to Müller glial cells in retina. response was intact (150). Transcriptomic analysis revealed that these "repair Tregs" have a distinct gene expression pattern indicating their proficiency in extracellular matrix remodeling and tissue repair (150).

Another unique tissue Treg population has been found in skeletal muscles, where by virtue of amphiregulin secretion, they help in muscle regeneration and healing. These cells, usually accounting for 10% of muscle T-cells at a steady state, proliferate vigorously after an intramuscular administration of cardiotoxin, which induces hypercontraction and myofibril death induced acute injury (254), reaching close to 50% of muscle T-cell population (188). High expression of Nrp1 and Helios and a unique and restricted TCR repertoire suggests thymic origin of muscle Tregs and reactivity to a local muscle antigen (188). These Tregs have a unique transcriptome compared to lymphoid organ Tregs with several differentially expressed genes. They have upregulated transcripts involved in Treg mediated suppression (*Il10*, *Gzmb*, etc.), tissue repair (*Il1rl1*, *Areg*, etc.), and muscle regeneration (255) (*Ccr2*) as well as genes encoding proteins found in contractile muscle function (256) like nebulin and nebulin-like proteins (*Neb, Nebl*). Depletion of Tregs during a muscle injury episode delays muscle healing, most probably because of loss of Treg generated amphiregulin. Also, fibro/ adipogenic progenitors in skeletal muscles produce high levels of IL33, whose receptor ST2 is highly expressed on muscle Tregs. Thus, muscle Tregs seem to be involved in an alarmin induced repair process (**Figure 8A**). Interestingly, muscle Treg population declines in old age mice which results in a deterioration of repair and regeneration process (257).

Recently, a very elegant and detailed (258) study in zebrafish has elaborated upon yet unknown and spectacular regenerative capabilities of Tregs (**Figure 8B**). The authors found that an ortholog of mammalian *Foxp3*, *Foxp3a*, which was exclusively expressed in a subpopulation of zebrafish T-cells, was upregulated most prominently in distinct regenerating organs. Zebra fish Tregs (zTregs) were predominantly found in kidneys but infiltrated and vigorously proliferated in regenerating tissues. As in the mammalian counterparts, these cells expressed high levels of *Nrp1a* and *Helios* in comparison to kidney zTregs (258). It has been reported that CNS1 region of *Foxp3* locus, responsible for pTreg generation, is not found in zebrafish (259). For identification of zTreg's role in organ regeneration, punctual and continuous deletion of *Foxp3a* resulted in reduced and delayed regeneration in heart, spinal cord, and retina injury models (258). Deletion of zTregs, in fact, reduced the tissue-specific precursor cells and subdued their proliferation (258). Indeed, zTregs were found near progenitors, sometimes even in close contact (258). However, the most striking finding of this study is that zTregs which presumably came from a common unbiased pool, became plastic in a tissue-specific regenerative context and produced tissue precursor cell specific regeneration factors like Neurotrophin 3 for neural progenitors in regenerating spinal cord, Neuregulin 1 for cardiomyocytes in injured heart and insulin-like growth factor 1 for retinal progenitor Müller glia cells (258) (**Figure 8B**). That zTregs are the primary source of these growth factors was confirmed by rescue of regeneration in zTreg depleted tissues by recombinant tissue-specific growth factors (258). The regeneration potential of zTregs was independent of their immunosuppressive potential or at least was not dependent on their IL10 production as IL10-deficient cells were fully capable of inducing precursor cell proliferation. However, regeneration potential was *Foxp3a*-dependent as regeneration process was significantly reduced in *Foxp3a*<sup>−</sup>*/*<sup>−</sup> tissues along with growth factor expression levels (258). On the other hand, *Areg* expression was not *Foxp3a* dependent and its role in regeneration was limited. It would be interesting to extrapolate and confirm similar findings in murine and human tissues.

### Feto-Maternal Tolerance

An equally fascinating population of Tregs which accumulates in murine placenta to induce maternal tolerance to fetus has been described (259). To say that the Tregs are extremely important from the outset of pregnancy will not be an overshoot [reviewed in Ref. (260, 261)]. Indeed, mating itself expands uterine Tregs and induces a transient "tolerance" to paternal alloantigens (260). In both humans and mice, seminal plasma contains TGFβ and prostaglandin E, which are potent Treg inducers. In fact, seminal fluid in humans and rodents contains highest measured TGFβ levels among biological fluids (260). Women with recurrent spontaneous abortions have reduced Treg population (262). Female decidual and uterus draining LN Treg generation is CNS1 dependent (259) and increased fetal resorption and placental T-cell infiltration was observed in CNS1-deficient mice. Apart from the their most likely peripheral origin, it is not known whether these Tregs have a distinct phenotypic and functional profile, elucidation of which could come very informative toward amelioration of infertility, pre-eclampsia, and other spontaneous abortive disorders. Very recently, an elegant study on human fetal antigen presenting cells (263) has found that fetal counterparts of DCs are primarily tolerogenic in their response. And, the primary response is generation of Tregs, even more than the adult DCs, in an *in vitro* Treg differentiation assay (263). These DCs were found across several fetal tissues, including spleen, thymus, skin, gut, and lungs (263). Unfortunately, the authors did not describe if Tregs were also present in these tissues. Earlier, it has been shown that human fetal Tregs promote tolerance to non-inherited maternal antigens (264) but only recently, it came to light that Tregs are required for suppression of *in utero* autoimmunity as well. Two children with IPEX syndrome, who died soon after birth, presented histological evidences of tertiary lymphoid structures, chronic inflammatory changes, and targeted exocrine pancreas autoimmunity (265). This signifies that in the perplexing settings of a pregnancy, Tregs are instrumental in establishing tolerance at both ends of maternal–fetal relationship.

# CONCLUSION AND PERSPECTIVES

Translational utility of many biological processes is marred by lack of specificity. A similar dilemma exists for Treg biologists as well; however, in case of Tregs selective therapeutic targeting appears to be achievable by virtue of harnessing their gradually established phenotypic and functional diversity. Recent studies have provided evidence that even for the organs like testes and eye, which are conventionally considered immune-privileged; there are populations of Tregs maintaining dominant tolerance and/or tissue homeostasis. While in testis, where otherwise privileged autoantigen escapes from the seminiferous tubules, only to generate systemic tolerance *via* Tregs (266), the retina actively recruits Tregs, which not only attenuate inflammation, but also repair the vasculature, saving blinding neovascular retinopathies (267). Another layer of specificity is added by discovery of tissueresident Tregs and their unique characteristics. However, most of the information except the recent reports on skin resident Tregs and TI-Tregs are from mouse tissues. There are several differences in structure as well as physiology between mice and humans. For example, mouse skin contains a thin muscle layer *panniculus carnosus*, which is vestigial in humans (268). This helps in contraction, revascularization and healing of wounds without scar formation in mouse. Human skin on the other hand heals by secondary intention leaving scar tissues. Thus, it is important to identify human tissue Tregs for an informed effort toward therapeutic usage.

There is a need to conclusively establish the origin and accumulating factors for tissue Tregs. One of the most pressing questions about almost all the tissue Tregs is identification of their natural ligands or tissue antigens. Although it has been demonstrated that in certain cases Tregs do not need TCR stimulation for some of their functions (150), Tregs with a smaller subset of specific TCR repertoire populate various tissues as well as malignancies. Therefore, cognate ligands that help in survival and proliferation of Tregs in these tissues are likely to have significant contributions in catering tissue-specific modulations. Proof of concept studies provide evidence that Tregs with defined antigen specificity (chimeric antigen receptor Tregs, CAR-Tregs) have potent immunosuppressive functions along with advantage of not inducing generalized immunosuppression (269).

Question remains as to how Tregs communicate with specific tissue cells-like adipocytes to establish a channel of communication with the environment. Beyond adaptation to inflammatory context, there are peculiarities of Treg biology, which modulate their effect temporally in life as well. Such as, Treg accumulation in aging WAT induces insulin resistance (79), whereas its accumulation in young obese WAT ameliorates it (83). The mechanisms that drive such specific outcomes need to be studied in detail. This accentuated capability to adapt sometimes becomes counterproductive too as seen in tumors where the suppressive capacity is enhanced even in comparison to normal tissue-resident Tregs and is, in turn, utilized by tumors for their

### REFERENCES


survival and immune escape. The mechanisms by which Tregs can push the limits of their functional capabilities are yet to be identified.

A major aspect of tissue adaptation is adjusting the cellular metabolism according to the tissue environment. There are huge gaps in our understanding of both lymphoid and tissue Treg metabolism. In *in vitro* differentiated Tregs (iTregs), it was shown that Foxp3 suppresses glycolysis by repression of Myc and helps in developing resistance to l-lactate (270). Similarly, Foxp3 counters PI(3)K-Akt-mTORC1 to diminish glycolysis in iTregs (271). Contrastingly, splenic and TI-Tregs were shown to uptake more 2NBDG, a fluorescent glucose analog, while intratumoral effector T cells showed glucose deprivation leading to reduced production of glycolytic metabolite phosphoenol pyruvate, resulting in compromised effector functions *via* reduced calcium-NFAT signaling (272). More recently, glycolysis was found to be instrumental in Treg trafficking and migration to inflamed tissues. The induction of the glycolytic enzyme glucokinase GCK and cytoskeletal rearrangement upon its association with actin was shown to be critical for the process (273). These findings underscore the need for extensive studies to delineate metabolic reprogramming in not only tissue Tregs but also lymphoid Tregs under steady state and activated conditions.

One can only be amazed by the diversity and functional plasticity of Tregs. A question, therefore, always comes up as to why Tregs are the chosen ones? Whether similar diversities among other immune cell types are still awaiting discoveries, or whether Foxp3 and presumably other unknown factors provide some degree of functional uniqueness to Tregs, remains to be seen. Nevertheless, looking at the diversity of responses ranging from maintaining immune tolerance to tissue repair, to becoming a major stakeholder in maintenance of physiological function of tissues, it would be apt to say that Tregs are the proverbial "Jack of all trades," and certainly, "master" of some.

### AUTHOR CONTRIBUTIONS

AS and DR planned and wrote the manuscript.

### FUNDING

This work was supported by Project IBS-R005 of the Institute for Basic Science, Korean Ministry of Science, Information/ Communication Technology and Future Planning.


associated insulin resistance. *Immunity* (2016) 45(2):428–41. doi:10.1016/j. immuni.2016.06.016


but not mucosa-generated induced Foxp3+ T reg cells. *J Exp Med* (2012) 209(10):1723–42, S1. doi:10.1084/jem.20120914


rejection. *Cell Immunol* (2012) 274(1–2):12–8. doi:10.1016/j.cellimm. 2012.02.012


polyendocrinopathy, enteropathy, X-linked (IPEX) syndrome. *Br J Dermatol* (2009) 160(3):645–51. doi:10.1111/j.1365-2133.2008.08835.x


melanoma. *Proc Natl Acad Sci U S A* (2007) 104(52):20884–9. doi:10.1073/ pnas.0710557105


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

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

# Clinical Remission of Sight-Threatening Non-Infectious Uveitis Is Characterized by an Upregulation of Peripheral T-Regulatory Cell Polarized Towards T-bet and TIGIT

*Rose M. Gilbert 1,2\*, Xiaozhe Zhang1,2, Robert D. Sampson3 , Michael R. Ehrenstein4 , Dao X. Nguyen4 , Mahid Chaudhry <sup>1</sup> , Charles Mein5 , Nadiya Mahmud5 , Grazyna Galatowicz1 , Oren Tomkins-Netzer 1,2, Virginia L. Calder1 and Sue Lightman1,2*

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Ashutosh Chaudhry, Memorial Sloan Kettering Cancer Center, United States Xing Chang, Shanghai Institutes for Biological Sciences (CAS), China*

*\*Correspondence:*

*Rose M. Gilbert rose.gilbert@gmail.com*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 30 September 2017 Accepted: 11 April 2018 Published: 03 May 2018*

### *Citation:*

*Gilbert RM, Zhang X, Sampson RD, Ehrenstein MR, Nguyen DX, Chaudhry M, Mein C, Mahmud N, Galatowicz G, Tomkins-Netzer O, Calder VL and Lightman S (2018) Clinical Remission of Sight-Threatening Non-Infectious Uveitis Is Characterized by an Upregulation of Peripheral T-Regulatory Cell Polarized Towards T-bet and TIGIT. Front. Immunol. 9:907. doi: 10.3389/fimmu.2018.00907*

*1Ocular Immunology, Institute of Ophthalmology, University College London (UCL), London, United Kingdom, 2Moorfields Eye Hospital NHS Foundation Trust, London, United Kingdom, 3 Flow Cytometry Core Facility, Institute of Ophthalmology, University College London (UCL), London, United Kingdom, 4Division of Medicine, Centre for Rheumatology, University College London (UCL), London, United Kingdom, 5Genome Centre, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom*

Background: Non-infectious uveitis can cause chronic relapsing and remitting ocular inflammation, which may require high dose systemic immunosuppression to prevent severe sight loss. It has been classically described as an autoimmune disease, mediated by pro-inflammatory Th1 and Th17 T-cell subsets. Studies suggest that natural immunosuppressive CD4+CD25+FoxP3+ T-regulatory cells (Tregs) are involved in resolution of inflammation and may be involved in the maintenance of clinical remission.

Objective: To investigate whether there is a peripheral blood immunoregulatory phenotype associated with clinical remission of sight-threatening non-infectious uveitis by comparing peripheral blood levels of Treg, Th1, and Th17, and associated DNA methylation and cytokine levels in patients with active uveitic disease, control subjects and patients (with previously active disease) in clinical remission induced by immunosuppressive drugs.

Methods: Isolated peripheral blood mononuclear cells (PBMC) from peripheral blood samples from prospectively recruited subjects were analyzed by flow cytometry for CD3, CD4, FoxP3, TIGIT, T-bet, and related orphan receptor γt. Epigenetic DNA methylation levels of FOXP3 Treg-specific demethylated region (TSDR), FOXP3 promoter, TBX21, RORC2, and TIGIT loci were determined in cryopreserved PBMC using a next-generation sequencing approach. Related cytokines were measured in blood sera. Functional suppressive capacity of Treg was assessed using T-cell proliferation assays.

Results: Fifty patients with uveitis (intermediate, posterior, and panuveitis) and 10 control subjects were recruited. The frequency of CD4+CD25+FoxP3+ Treg, TIGIT+ Treg, and T-bet+ Treg and the ratio of Treg to Th1 were significantly higher in remission patients compared with patients with active uveitic disease; and TIGIT+ Tregs were a significant predictor of clinical remission. Treg from patients in clinical remission demonstrated a

**131**

high level of *in vitro* suppressive function compared with Treg from control subjects and from patients with untreated active disease. PBMC from patients in clinical remission had significantly lower methylation levels at the FOXP3 TSDR, FOXP3 promoter, and TIGIT loci and higher levels at RORC loci than those with active disease. Clinical remission was also associated with significantly higher serum levels of transforming growth factor β and IL-10, which positively correlated with Treg levels, and lower serum levels of IFNγ, IL-17A, and IL-22 compared with patients with active disease.

Conclusion: Clinical remission of sight-threatening non-infectious uveitis has an immunoregulatory phenotype characterized by upregulation of peripheral Treg, polarized toward T-bet and TIGIT. These findings may assist with individualized therapy of uveitis, by informing whether drug therapy has induced phenotypically stable Treg associated with long-term clinical remission.

Keywords: uveitis, T-regulatory cells, TIGIT, T-bet, ocular inflammation, remission, Th1, Th17

### INTRODUCTION

Uveitis, defined as inflammation of the uveal inner layer of the eye, is the fifth commonest cause of visual loss in the developed world, accounting for about 10–15% of the cases of total blindness (World Health Authority definition) and up to 20% of legal blindness (1). Incidences vary between 38 and 200 per 100,000 (2). Uveitis can affect people of all ages but occurs most frequently in the working age population (20–50 years), with 35% affected remaining visually disabled (3). In the majority of cases, the inflammation in uveitis is non-infectious, of idiopathic and presumed autoimmune etiology. This type of disease may follow a relapsing, remitting course, which is clinically challenging to manage. Severe cases may require treatment with high dose oral corticosteroid drugs to minimize irreversible damage to the retina and visually important structures, which result in sight loss. Uveitis is a clinically heterogenous disease: at least 150 disorders are known to be associated with intraocular inflammation, although these disorders are individually rare. The Standardization of Uveitis Nomenclature (SUN) working group, therefore, developed clinical criteria to classify uveitis based on the anatomical localization of the inflammation (4). These criteria are useful but they do not define specific disease entities, phases of disease or include immunological biomarkers.

Our understanding of the immune mechanisms involved in non-infectious uveitis has derived from animal studies of experimental autoimmune uveitis (EAU) (5). Inflammation in EAU is mediated by Th1 and Th17 subsets of self-reactive CD4 T-lymphocytes. The Th1 and Th17 lineages are characterized by expression of T-bet (*Tbx21*) and retinoic-acid related orphan receptor (ROR) γ-t (*Rorc*) transcription factors and secretion of the "signature" pro-inflammatory cytokines IFNγ and IL-17, respectively. Phenotypically categorized CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> T-regulatory cells (Tregs) exist as a naturally occurring mechanism to suppress these autoreactive T-cells through production of anti-inflammatory cytokines such as IL-10 and transforming growth factor β (TGF-β) (6), in addition to many other mechanisms. Studies in rat models of EAU have shown that during resolution of the first acute attack of uveitis, the number of ocular Treg is increased (7). In those rats that went on to develop recurrent EAU, the suppressor function of Treg was found to be weaker (7). In murine models of EAU, a significantly increased frequency and immunoregulatory action of CD4<sup>+</sup>CD25<sup>+</sup> Treg cells has been associated with the regression of EAU, suggesting that CD4<sup>+</sup>CD25<sup>+</sup> Tregs are induced during EAU and may be involved in disease resolution (8). Further evidence in the mouse model demonstrates that retina-specific functionally suppressive FoxP3<sup>+</sup> Tregs accumulate in inflamed eyes and persist for several months after disease remission (9). Depletion of Treg at the peak of uveitis delayed resolution and, following resolution where mice displayed a low grade chronic inflammation, Treg depletion precipitated disease relapse (9).

Many human studies have examined the role of Treg in systemic autoimmune diseases, where defects in Treg numbers and/or suppressive function have been found, in addition to imbalances between pro-inflammatory T-effectors and immunosuppressive Treg (10, 11). Patients with active uveitis have decreased peripheral blood levels of Tregs compared with healthy control subjects, with levels being significantly upregulated during disease remission (12, 13). Low or absent levels of FoxP3 mRNA are found in Tregs from patients with severe, recalcitrant uveitis (13). In rheumatoid arthritis (RA), Tregs recover their suppressive function after immunomodulating treatment with anti-tumor necrosis factor (TNF) α (11) monoclonal antibody (Ab) therapy. More specifically, the anti-TNFα agent, adalimumab, has been shown to restore immune tolerance in RA through expansion of functional Foxp3<sup>+</sup> Treg cells, equipped to suppress Th17 effector cells (11, 14, 15). In clinical trials of systemically administered immunomodulating therapies in uveitis, induction of disease remission is associated with systemic upregulation of Tregs and modulation of cytokine production (16, 17). However, pharmacological effects on cytokines have also been demonstrated without a significant increase in the proportion of systemic FoxP3<sup>+</sup> Treg levels (18). Taken as a whole, these studies suggest that the mechanism of disease remission in uveitis may involve upregulation of systemic Treg and/or restoration of their suppressive function, the effects of which may be mediated through cytokines. Furthermore, "functional" Treg seem to be important for preventing disease relapse and maintaining long-term clinical remission.

A lack of consensus regarding immunophenotypic markers for functionally suppressive Treg has meant that their status in human autoimmune disease has not been fully understood. Inflammatory environments induce T-cell plasticity (19) in which peripherally induced CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Treg (pTregs) subsets may co-exist with more phenotypically stable thymically derived CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Tregs (nTregs). Foxp3 is considered the Treg "master transcription factor" because it is critically required for Treg-cell development and function and for suppressing autoimmunity (20–23). However, it is also known that activated human T-cells may transiently upregulate FOXP3 and that these FoxP3<sup>+</sup> T-cells are not effective at suppressing inflammation (24–27). Studies have revealed that a specific DNA hypomethylation pattern, the Treg-specific demethylated region (TSDR), a hypomethylated region within the FOXP3 gene*,* is associated with enhancement and stabilization of Foxp3 expression in Treg (28–31). In combination with CD25 and FoxP3 expression, the TSDR has allowed identification of Treg with a stable phenotype (32). The addition of a methyl group (–CH3) to DNA is a common epigenetic mechanism that cells use to switch genes "off." Extensive methylation of cytosine in DNA is known to correlate with reduced gene transcription. DNA methylation is generally thought to play an essential role in T-cell function and failure to maintain DNA methylation patterns in mature T-cells has been implicated in the development of autoimmune disease (33).

A recently discovered co-inhibitory molecule, T cell immunoreceptor with Ig and ITIM domains (TIGIT), is expressed by Treg (34) and suppresses a range of immune cells. Studies show that TIGIT ligation directly inhibits T cell proliferation and cytokine production in CD4+ T cells (35). Furthermore, increased expression of TIGIT, which delineates Treg from activated effector T cells, has been associated with hypomethylation and FOXP3 binding at the TIGIT locus (36). Treg cells expressing TIGIT were found to selectively inhibit Th1 and Th17, but not Th2 responses (37). Since it is known that the Th1 and Th17 subsets are pivotal to the pathogenesis of autoimmune disease, TIGIT may be a biomarker of stable, functionally suppressive Treg in these diseases.

Treatment paradigms of systemic autoimmune diseases, such as RA and inflammatory bowel disease, have now evolved beyond partial symptom control (38). The goal of inducing and maintaining sustained biological remission, to improve long-term disease outcomes, has been described as a "treat-to-target" strategy (38). A working definition of sustained "deep remission," which includes long-term biological remission and symptom control with defined patient outcomes, including no disease progression, has been proposed for Crohn's disease (38, 39). "Clinical remission" in uveitis, however, has not been defined or investigated as an immunologically distinct phase of disease. Understanding the immunological mechanisms underlying disease remission, in addition to the focus on inhibiting inflammation, could have significant implications for clinical phenotyping of uveitis and determining the best therapeutic approach to induce remission for each patient.

In this study, we investigate the hypothesis that there is a peripheral blood immunoregulatory T-cell phenotype associated with sustained clinical remission, using a prospective study design with recruitment of 50 non-infectious uveitis patients (without identifiable systemic disease) and 10 control subjects. Clinical remission is defined as a 6-month period of remission without treatment, in patients with previously active relapsing and remitting disease, which clinically responded to systemic immunosuppressive therapy. We evaluate the frequency of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Treg and, additionally, the frequency of TIGIT<sup>+</sup> Treg, which we propose to be a biomarker of functionally suppressive Treg. Co-expression of FoxP3, ROR γ-t, and T-bet, the ratio of Treg to putative T-effectors, and serum levels of cytokines associated with these T-cell subsets are also analyzed. Furthermore, we evaluate whether there is an epigenetic immune methylation pattern associated with clinical remission in uveitis, using a targeted next-generation sequencing (NGS) bisulfate sequencing approach toward T-cell genes of interest.

### MATERIALS AND METHODS

### Patient Recruitment and Specimen Collection

A total of 50 patients with a diagnosis of idiopathic noninfectious uveitis were consecutively recruited to the study from Moorfields Eye Hospital, London, UK between November 2014 and December 2016*.* Inclusion criteria for all patients were as follows: age 18–59 years; a current or previous diagnosis of non-infectious intermediate, posterior or panuveitis as per SUN (4); and the absence of known associated infectious disease or systemic inflammatory disease. Eligibility for recruitment was further determined by uveitic disease activity. To meet the clinical criteria for disease "in clinical remission," patients were required to have no signs of disease activity at the time of recruitment; have a previous diagnosis of active non-infectious chronic relapsing and remitting uveitis which was treated with oral corticosteroids ± second-line corticosteroid-sparing treatments before the period of clinical remission; and have discontinued all immunosuppressive treatment and remained "quiescent" (without reactivation of disease) for at least 6 months. To meet the criteria for "active" disease, patients were required to have new onset of uveitic disease occurrence at the time of recruitment (a new diagnosis or reactivation of existing disease), which was regardless of treatment status. Disease activity was determined by clinical symptoms, examination with slit lamp biomicroscopy and clinical imaging (optical coherence tomography, fundus autofluorescence, and fluorescein angiography). Patients who met the study inclusion criteria and who were clinically assessed to have active disease (i.e., anterior chamber cells ≥0.5+, vitreous cells ≥0.5+, inflammatory cystoid macular edema, optic disk edema, active vasculitis, or new/active chorio-retinal lesions) were considered eligible as "active" patients in the study. As controls, 10 age-matched subjects with no history of uveitis, systemic inflammatory disease or known infectious disease were enrolled onto the study.

A 30 mL sample of heparinized peripheral venous blood was obtained from all subjects. In the active cohort, samples were obtained before starting/modifying the immunosuppressive treatment regime. Additional peripheral venous blood samples were prospectively obtained from four active patients at 2, 6, and 12 months after starting/changing treatment.

This study was carried out in accordance with the recommendations of the UK National Research Ethics Service—London Harrow Committee and the Moorfields Eye Hospital National Health Service Foundation Trust, Department of Research and Development (13/LO/1653; 16039). All subjects gave written informed consent in accordance with the Declaration of Helsinki.

## Peripheral Blood Mononuclear Cells (PBMC) Isolation, Cell Culture, and Cryopreservation

Peripheral blood mononuclear cells were obtained from fresh heparinized blood (≤4 h following venupuncture) by Histopaque density gradient centrifugation (Histopaque-1077, Sigma-Aldrich, Gillingham, UK) and washed twice with RPMI 1640 with 10% human AB Serum (Sigma-Aldrich). 10 × 106 PBMC from each sample were stained for multi-color flow cytometry.

For experimental cell culture conditions requiring stimulated cells, freshly isolated PBMC were cultured in X-VIVO 15 medium with l-glutamine, gentamicin, and phenol red (Lonza, Cambridge, UK) for 5 days, with the addition of 5 µg/mL soluble anti-CD3 (HIT3a) NA/LE, 1 µg/mL anti-CD28 NA/LE (BD Biosciences), and 50 IU IL-2 IS/mL (Miltenyi Biotec, Woking, UK) to the culture media, unless otherwise specified.

The remaining PBMC were cryopreserved in human AB serum with 10% Hybri-Max Sterile-filtered DMSO (Sigma-Aldrich) at −20°C for 1 h before transfer to −70°C and then to liquid nitrogen for longer term storage, for subsequent DNA isolation and methylation analysis.

### Antibodies

For multi-color immunophenotyping by flow cytometry, the following Abs were used: RORγt-Alexa Fluor 488, T-bet-PerCP-Cy5.5, CD3-Vioblue, CD25-Brilliant Violet 711, CD4- Brilliant Ultraviolet 395, FoxP3-Alexa Fluor 647, TIGIT-PE, and Near Infrared live/dead fixable cell stain. For intracellular cytokine staining by flow cytometry, the following Abs were used: CD3-Vioblue, CD4-Brilliant Ultraviolet 395, IL-10-PE-CF594, IFNγ-BV605, IL-17A PE-Vio-770, and Near Infrared live/dead fixable cell stain. For live cell sorting, the following Abs were used: CD4-PerCP-Cy5.5, CD127-Brilliant Violet 605, CD3-Brilliant Violet 785, CD25-PE-Dazzle 594, CD14-APC-Fire, and SYTOX blue dead cell stain. Abs were provided by BioLegend (London, UK), BD Biosciences (Oxford, UK), eBiosciences (Hatfield, UK), Miltenyi Biotec, and ThermoFisher Scientific (Dartford, UK).

## Treg and Th1/Th17 Immunophenotyping by Flow Cytometry

The peripheral blood levels of the following cell subsets were determined by flow cytometry: CD25<sup>+</sup>FoxP3<sup>+</sup> Treg, CD25<sup>+</sup>FoxP3<sup>+</sup> TIGIT<sup>+</sup> Treg, T-bet<sup>+</sup> Th1, RORγt <sup>+</sup> Th17, T-bet<sup>+</sup>RORγt <sup>+</sup> Th1/17, Foxp3<sup>+</sup>T-bet<sup>+</sup> Treg, and Foxp3<sup>+</sup>RORγt <sup>+</sup> Treg in the CD3<sup>+</sup>CD4<sup>+</sup> subset. Single-cell suspensions were washed in PBS (Sigma-Aldrich) and stained with directly conjugated Abs for cell surface molecules (CD3, CD4, CD25, TIGIT), then stained with Zombie NIR (BioLegend) fixable dead cell stain, then fixed and permeabilized with a transcription factor fixation/permeabilization kit (eBiosciences), and subsequently stained using directed conjugated Abs recognizing transcription factors (FoxP3, RORγt, and T-bet). Subject-specific fluorescence-minus-one (FMO) controls for CD25, TIGIT, FoxP3, RORγt, and T-bet were included in each experiment.

Flow cytometric data were acquired using an LSRFortessa (BD Biosciences, San Jose, CA, USA) and BD FACSDiva version 6.1.3 software with a total of 50,000 events being recorded for each sample through a live, single-cell CD3<sup>+</sup>CD4<sup>+</sup> gate (gating strategy shown in **Figure 1**). Viable cells were identified by low uptake of the fixable dead cell stain. Single-stained OneComp and UltraComp beads (eBiosciences) were used to generate compensation matrices. FMO controls were used to identify gating boundaries where cell populations were ill-defined. Analysis of flow cytometry data was performed using FlowJo version 10 software (TreeStar Inc., Ashland, OR, USA).

### Cytokine Analysis

Whole blood samples were centrifuged following collection to obtain serum supernatant samples, which were stored in 1.5 mL aliquots at −70°C for subsequent multiplex cytokine analysis. Cryopreserved subject serum cytokine samples were thawed and immediately analyzed for cytokines IL-10, IFNγ, IL-17A, and IL-22 by multiplex bead arrays (R&D Systems) using a MAGPIX Luminex system (Luminex, TX, USA). The data were analyzed using xPONENT software (Luminex). Acidified serum aliquots from the same subject samples were analyzed for TGF-β by ELISA (R&D Systems) using the optical density setting of the Modulus II Microplate Multimode Reader (Promega Corporation). Minimum levels of detection based on standard curves for each array were as follows: IL-10 (7.30 pg/mL), IFNγ (17.30 pg/mL), IL-17A (2.65 pg/mL), IL-22 (6.10 pg/mL), and TGF-β (7.4 pg/mL).

For intracellular cytokine staining, freshly isolated PBMC from a patient with active disease (Patient 1) at 2 and 6 months after baseline were stimulated with Dynabeads Human T-Activator CD3/CD28 (ThermoFisher Scientific) for 24 h or a cell stimulation cocktail (containing PMA and ionomycin) 1:500 for 6 h as a positive control, with the addition of protein transport inhibitor (containing Brefeldin A) to the culture medium for the last 4 h. Staining and flow cytometric analyses were as described earlier, with the following modifications: directly conjugated Abs for cell surface molecules (CD3 and CD4); fixation and permeabilization with an intracellular cytokine fixation/permeabilization kit (BD Biosciences); and directly conjugated Abs recognizing intracellular cytokines (IL-10, IFNγ, and IL-17A). Subject-specific FMO controls for IL-10, IFNγ, and IL-17A were included in each experiment.

### Cell Sorting and T-Cell Proliferation Assay

Freshly isolated PBMCs from subject samples in each group were stained with CD3, CD4, CD25, CD127, and CD14 cell surface Abs. CD3<sup>+</sup>CD4<sup>+</sup>CD25<sup>+</sup>CD127lo, CD3<sup>+</sup>CD4<sup>+</sup>CD25<sup>−</sup>, and CD14<sup>+</sup> populations representing Treg, putative T-effectors, and

version 6.1.3 software with a total of 50,000 events being recorded for each sample through a live, single-cell CD3+CD4+ gate. Viable cells were identified by low uptake of the fixable dead cell stain. Single-stained OneComp and UltraComp beads (eBiosciences) were used to generate compensation matrices. Fluorescenceminus-one controls were used to identify gating boundaries where cell populations were ill defined.

monocyte (MC) subsets, respectively, were isolated directly from the stained PBMC by means of high speed Influx flow cytometric cell sorting (BD Biosciences, San Jose, CA, USA) (gating strategy shown in **Figure 2A**).

CD3<sup>+</sup>CD4<sup>+</sup>CD25<sup>−</sup> T-cells were washed and resuspended at 10 × 106 /mL in sterile PBS (Sigma-Aldrich) containing 1 µL of 1 mM violet proliferation dye (VPD) 450 stock solution (BD Biosciences) for each 1 mL of cell suspension for a final VPD450 concentration of 1 µM, according to the manufacturer's instructions (**Figure 2B**). The cells were stained by incubating the dyecell suspension in a 37°C water bath for 10 min. The reaction was quenched by adding 9× the original volume of PBS to the cells, followed by centrifugation, discarding the supernatant, and resuspending the cells in 10 mL of RPMI 1640 medium with 10% FBS before washing again.

The *in vitro* capacity of the Treg to suppress the proliferation of VPD450-labeled CD3<sup>+</sup>CD4<sup>+</sup>CD25<sup>−</sup> responding T-cells (Tresp) was assessed in 96-well plates (1 × 106 per well density) in a classical 5-day coculture assay, as follows: VPD450-labeled CD3<sup>+</sup>CD4<sup>+</sup>CD25<sup>−</sup> Tresp were cocultured with sorted CD14<sup>+</sup> MCs at 1:1 ratio and sorted CD3<sup>+</sup>CD4<sup>+</sup>CD25<sup>+</sup>CD127lo Tregs were cocultured with Tresp and MC at a 1:3:3 ratio. Cell culture conditions were as previously described. Data were acquired by flow cytometry and analyzed using the cell tracking function of Modfit LT modeling software (Verity Software House, ME, USA) to generate a statistic termed proliferation index (PI) (**Figure 2C**). Percentage (%) suppression was determined for each subject sample as follows, adapted from previously described methods for conducting suppression assays from small numbers of isolated T-cells (15, 40):

$$\% \text{ suppression} = \frac{(\text{Tresp} + \text{MC}) - (\text{Tresp} + \text{MC} + \text{Treg})}{(\text{Tresp} + \text{MC})} \times 100.1$$

### DNA Isolation and Methylation Analysis

Genomic DNA extraction was performed directly on samples of thawed cryopreserved PBMC from each subject group using the DNeasy Blood and Tissue Kit (Qiagen, Manchester, UK). The samples were then analyzed at five DNA sites of cytosine (CpG) methylation (FOXP3 TSDR, FOXP3 promoter, TBX21, RORC2, and TIGIT) with bisulfite Amplicon Sequencing using an NGS approach on the Illumina sequencing platform (Fluidigm, CA, USA) (**Figure 3**; bisulfite Amplicon target sites are shown in Figure S1 in Supplementary Material).

proliferation index using VPD 450. (D) Comparison of % suppression of T-cell proliferation by Treg isolated from the different subject groups.

Before DNA bisulfite treatment, the integrity of the DNA was assessed using the Agilent 2100 Bioanalyser Tapestation (Agilent Technologies, Waldbronn, Germany), and the DNA concentration was measured using Qubit Fluorometric Quantification (ThermoFisher Scientific). 500 ng of normalized DNA was bisulfite converted using the EZ DNA Methylation kit (Cambridge BioSciences, Cambridge). The incubation conditions recommended by the manufacturers guide [95°C for 30 s, 50°C for 60 min] × 16 cycles, 4°C hold were used for the conversion reaction. The yield of bisulfite converted DNA was estimated using the RNA-40 setting on the NanoDrop 8000 Spectrophotometer V2.0 (ThermoScientific, USA). Primers for the target sites were designed using Pyromark Assay Design 2.0 software (Qiagen) and synthesized with Fluidigm Universal CS1 and CS2 tags (Fluidigm, San Francisco, USA) to allow for the construction of DNA libraries using Fluidigm proprietary indexes (see Table S2 in Supplementary Material for primer sequences *Fluidigm CS1 and CS2 tags are highlighted in bold*). MyTaq™ HS DNA Polymerase (Bioline, USA) was used to optimize the assays and amplify the DNAs. The PCR products were checked for amplification on an Agilent 2100 Bioanalyser D1K Tape (Agilent Technologies, Waldbronn, Germany) after which unique Fluidigm Barcodes (Table S3 in Supplementary Material) were added to the PCR product under the following PCR cycling conditions: 95°C for 10 min (95°C for 15 s, 60°C for 30 s, and 72°C for 1 min) cycle 14 times; 4°C hold. To check that the barcodes had been successfully incorporated onto the PCR products, the barcoded PCR product was run on an Agilent D1K tape. If the barcodes had attached, then a shift of ~60 bp was observed between the PCR product and the barcoded products. Equal volumes of barcoded products were pooled and primer dimer was removed using Agencourt Ampure XP beads (Beckman Coulter Life Sciences) at a volume of 1:1. The pool was quantified using High Sensitivity Qubit Reagents (Qiagen), and the library size was measured using the Agilent 2100 High Sensitivity D1K (Agilent Technologies, Waldbronn, Germany) to calculate the molarity of the library before sequencing. 150-bp Paired-end sequencing was performed using the Miseq Illumina sequencer (Illumina, Inc.) to a total of 400,000 reads passing filter per sample. Bismark (41) was used to align bisulfite-treated reads to a human reference genome and calculate the cytosine methylation calls at the target sites. Biostatistical analysis generated beta (%) methylation values as a ratio of methylated to unmethylated DNA for each CpG site of interest.

### Statistical Analysis

Statistical analysis of final datasets was performed with SPSS version 24 (IBM Software) and GraphPad Prism version 7 (GraphPad Software). Differences between groups were determined by the Kruskal–Wallis test with *post hoc* Bonferroni correction for multiple comparisons. Bivariate correlations between immunological variables were calculated using Spearman's test. Relationships between selected variables, which had clinically relevant associations, were modeled using multiple linear regression and logistic regression, using "stepwise" or "enter" variable entry, respectively. Where possible, variables with a non-normal distribution were transformed to a normal distribution, using a log transformation, to include in the multiple regression model. All significance tests were two-tailed. *P*-values < 0.05 were considered significant. Results are expressed as frequency *n* (%) or median (IQR), unless otherwise stated.

### RESULTS

### Subject Characteristics

A total of 50 uveitis patients and 10 control subjects were recruited to the study (**Table 1**). Of the 50 patients recruited, 37 were in clinical remission and 13 had active disease at the time of recruitment. 22 (59%) of the patients in the clinical remission group received previous therapy with corticosteroids only, the remaining patients received additional second-line oral immunosuppressive treatment (**Table 2**). Of the active patients, who all had intermediate uveitis, posterior uveitis or panuveitis, 11 (85%) had cells and/or haze in the vitreous,

Table 1 | Subject demographics and clinical characteristics.


*Median values are in bold font.*

7 (54%) had cells in anterior chamber, 6 (46%) had macular edema, 3 (23%) had optic disk edema, two (15%) had vasculitis, and 1 (8%) had chorio-retinal lesions at baseline. Six (46%) of these patients with active disease were on systemic immunosuppression at the time of recruitment. A subgroup analysis comparing the active patients on therapy and active patients not on therapy at baseline did not reveal any differences between the groups. Four of the 13 active patients underwent additional immunophenotyping at 2, 6, and 12 months after starting or changing treatment. The immunological marker levels and comparison across the three subject groups (active, remission, and controls) with summary *P*-values are shown in **Table 3** and **Figure 4**, with the *post hoc* pairwise comparisons and adjusted *P*-values shown in Table S1 in Supplementary Material.

### Clinical Remission of Uveitis Is Associated With Higher Levels of Treg Polarized Toward T-bet and TIGIT Compared With Active Disease

To determine whether there was a difference in peripheral blood Treg levels between control, active and quiescent subjects in disease remission, their PBMC were analyzed by flow cytometry for levels of CD25<sup>+</sup>FoxP3<sup>+</sup> Treg, FoxP3<sup>+</sup>TIGIT<sup>+</sup> Treg, FoxP3<sup>−</sup>TIGIT<sup>+</sup> T-cells, and FoxP3<sup>+</sup>T-bet<sup>+</sup> Treg in the CD3<sup>+</sup>CD4<sup>+</sup> lymphocyte subset (**Figure 5**). Highly statistically significant differences were found between the active and remission groups in the levels of Treg (4.4 ± 1.2 vs 7.1 ± 1.3, *P* = 0.000), TIGIT<sup>+</sup> Treg (2.5 ± 1.3 vs 5.0 ± 0.9, *P* = 0.000) and T-bet<sup>+</sup> Treg (0.7 ± 0.4 vs 1.1 ± 0.3, *P* = 0.000) (**Figure 4**). CD25<sup>+</sup>FoxP3<sup>+</sup> Treg, FoxP3<sup>+</sup>TIGIT<sup>+</sup> Treg, and FoxP3<sup>+</sup>T-bet<sup>+</sup> Treg were also observed to increase as disease resolved in patients with active uveitis (**Figure 6**). The ratio of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup>TIGIT<sup>+</sup> Treg to CD4<sup>+</sup>FOXP3<sup>−</sup>TIGIT<sup>+</sup> T-cells was significantly higher in the remission group compared with the active group (0.7 ± 0.2 vs 0.3 ± 0.2, *P* = 0.003). T-bet<sup>+</sup>FoxP3<sup>+</sup> cell levels positively correlated with TIGIT<sup>+</sup>FoxP3<sup>+</sup> cell levels across the three subject groups (*r* = 0.431, *P* = 0.001). There

Table 2 | Subject immunosuppressive treatment characteristics.


*Median values are in bold font.*

Table 3 | Comparison of immunological markers across the three subject groups.


*Median values are in bold font.*

*\*Significant at the 5% level.*

*\*\*Significant at the 1% level.*

*\*\*\*Significant at the 0.1% level.*

were no statistically significant differences between Treg levels in control and active uveitis patient groups.

### FoxP3**+**ROR**γ**t**+** Tregs Are Not Associated With Clinical Remission but May Have a Role in Clinical Resolution of Non-Infectious Uveitis

Peripheral blood levels were analyzed for CD3<sup>+</sup>CD4<sup>+</sup>FoxP3<sup>+</sup> RORγt <sup>+</sup> Treg (**Figure 5**). Although the remission and control groups had higher median levels of FoxP3<sup>+</sup>RORγt + Treg compared with the active group, these differences did not reach statistical significance (*P* = 0.363) (**Figure 4**). However, an overall increase in FoxP3<sup>+</sup>RORγt <sup>+</sup> Treg levels was observed over 12 months in all four patients with active disease at baseline (**Figure 6**).

## Clinical Remission of Uveitis Is Associated With Overall Lower Levels of Th1 Transcription Factors and a Higher Ratio of Treg to Th1 Compared With Active Disease

To determine whether there was a difference in Th1 (T-bet transcription factor expression) levels and ratios of Treg to Th1 between control, active and quiescent subjects in disease remission, their PBMC were analyzed by flow cytometry for levels of CD25<sup>+</sup>FoxP3<sup>+</sup> Treg (**Figure 5**) and T-bet<sup>+</sup> T-cells in the CD3<sup>+</sup>CD4<sup>+</sup> lymphocyte subset. Significantly lower levels of T-bet transcription factor levels (48.5 ± 6.8 vs 58.6 ± 11.7, *P* = 0.024) and higher ratios of Treg:Th1 (0.7 ± 0.2 vs 0.4 ± 0.1, *P* = 0.001) were found in remission patients compared with active patients (**Figure 4**). Active patients also had significant lower ratios of Treg:Th1 compared with control subjects (0.4 ± 0.1 vs 0.7 ± 0.1, *P* = 0.013) (**Figure 4**). Disease resolution in active patients over the course of 12 months appeared to be associated with an increase in levels of Treg and decrease in levels of Th1 (**Figure 6**).

### Clinical Remission of Uveitis Is Not Associated With Overall Lower Levels of Th17 Transcription Factors or a Higher Ratio of Treg to Th17 Compared With Active Disease

To determine whether there was a difference in Th17 (RORγt + transcription factor expression) levels and Treg:Th17 ratios between control, active and quiescent subjects in disease remission, their PBMC were analyzed by flow cytometry for levels of CD25<sup>+</sup>FoxP3<sup>+</sup> Treg (**Figure 5**) and RORγt <sup>+</sup> T-cells in

CD3+CD4+ T-cell compartment, with comparisons across the three subject groups (active, remission, and controls). (A) % CD25+FoxP3+ Treg. (B) % FoxP3+TIGIT<sup>+</sup> Treg. (C) Ratio of FoxP3+TIGIT+ Treg to TIGIT+FoxP3− T-cells. (D) % FoxP3+T-bet+ Treg. (E) % T-bet+ Th1. (F) Ratio of Treg to Th1. (G) % FoxP3+RORγt <sup>+</sup> Treg. (H) % RORγt <sup>+</sup> Th17. (I) Ratio of Treg to Th17. (J) % T-bet+RORγt <sup>+</sup> "double positive" Th1/Th17.

the CD3<sup>+</sup>CD4<sup>+</sup> lymphocyte subset. No significant difference was found in Th17 levels between the three groups (*P* = 0.564), although it was noted that both active (2.0 ± 1.3) and remission (2.8 ± 1.4) patients had significantly higher Treg:Th17 ratios in comparison with control subjects (0.4 ± 0.3, *P* = 0.014; *P* = 0.000 respectively) (**Figure 4**). Increased Th17 levels in three out of four patients with active disease were noted as the disease clinically resolved over 12 months (**Figure 6**).

## Clinical Remission of Uveitis Is Not Associated With Decreased Levels of "Double Positive" Effector ROR**γ**t**+**T-bet**<sup>+</sup>** T-Cells

Peripheral blood levels were analyzed for "double positive" RORγt <sup>+</sup>T-bet<sup>+</sup> T-cells in the CD3<sup>+</sup>CD4<sup>+</sup> subset. There was no significant difference in levels of RORγt <sup>+</sup>T-bet<sup>+</sup> T-cells between the three groups (*P* = 0.183) (**Figure 4**).

## Clinical Remission of Uveitis Is Associated With Higher Serum Levels of TGF-**β** and IL-10 and Lower Serum Levels of IFN**γ**, IL-17A, and IL-22 Compared With Active Disease, and Serum TGF-**β** and IL-10 Levels Positively Correlate With Treg Levels

Cytokine levels of IL-10, TGF-β, IFNγ, IL-17A, and IL-22 were assayed in serum from 10 remission patients, 8 active patients and 5 control subjects from the above groups (**Figures 7A–E**). IL-10 levels were significantly higher in serum from remission patients compared with active patients (21.5 ± 2.4 vs 11.6 ± 1.5, *P* = 0.011), and lower in active patients compared with controls (11.6 ± 1.5 vs 24.2 ± 2.3, *P* = 0.000). TGF-β levels were higher in remission patients than in active (161.1 ± 81.4 vs 62.7 ± 24.0, *P* = 0.007) and control (161.1 ± 1.5 vs 56.9 ± 24.0, *P* = 0.004) subjects. Clinical remission was also associated with significantly lower serum levels of IFNγ (*P* = 0.015), IL-17A (*P* = 0.002), and IL-22 (*P* = 0.001) compared with patients with active disease. Increased intracellular IL-10 and IL-17A and decreased intracellular IFNγ levels were found in CD4<sup>+</sup> T-cells from active uveitis, 6 months after starting treatment, when the disease had clinically resolved (**Figure 7F**). Serum TGF-β (*r* = 0.752, *P* <0.0001) and IL-10 (*r* = 0.667, *P* = 0.0005) levels positively correlated with Treg levels across all subjects (**Figures 7G,H**).

### Clinical Remission of Uveitis Is Associated With Treg Which Demonstrate a High Capacity to Suppress Proliferating T-Effectors

Cell-sorted populations of Treg, putative T-effectors and macrophages were set up in *in vitro* culture and the T-cell PI was modeled by Modfit software, which was used to calculate% suppression by Treg, as described earlier. Treg from active uveitis without treatment failed to suppress proliferation of T-cells *in vitro* (**Figure 2D**). By contrast, Treg from active uveitis treated with oral immunosuppression, Treg from control subjects and Treg from uveitis in clinical remission all showed *in vitro* functional suppression, with the highest suppressive capacity demonstrated by Treg from uveitis in clinical remission (**Figure 2D**). Data are shown from the following subjects: Patients 1 and 50 (in active disease group), Patient 49 (in clinical remission group), and

Volunteer 2 (in control group). Patient 50 had active disease and had not yet started oral immunosuppression, whereas Patient 1 had a diagnosis of active disease at baseline and had received 6 months of oral immunosuppression.

## Lower Levels of Methylation at Key Epigenetic CpG Sites Determining Treg Function in PBMC Are Associated With Clinical Remission but Are Not Consistently Observed in Clinical Resolution of Uveitis

To determine whether there was a difference in methylation at epigenetic CpG sites associated with Treg function (FoxP3 and TIGIT), bisulfite Amplicon Sequencing using an NGS approach was performed using cryopreserved PBMC from 8 active patients, 8 control subjects and 12 remission patients. Cryopreserved served samples, at 12 months following treatment, were also obtained from four of the eight active patients. The sample consisted of 17 female subjects (61%) and 11 male subjects (39%). Following biostatistical analysis of the epigenetic percentage (%) methylation data, a single CpG site demonstrating a high level of differential methylation between samples was selected for statistical comparison between the three groups at that site, as follows: FoxP3 TSDR (49117116), FoxP3 promoter (49121204), and TIGIT (114012658) (**Figure 8**). The median CpG site% methylation levels of the FoxP3 TSDR (48 ± 3.6 vs 59 ± 7.4, *P* = 0.003) (**Figure 8A**), FoxP3 promoter (53 ± 11 vs 72 ± 5.3, *P* = 0.036) (**Figure 8B**), and TIGIT (48 ± 3.3 vs 59 ± 3.3, *P* = 0.003) (**Figure 8C**) CpG sites were significantly lower in patients in clinical remission compared with active patients. When comparing% CpG site methylation within individual patients with active disease at 0 month and resolved disease at 12 months, a decrease in methylation at the FoxP3 promoter was

Figure 7 | Serum cytokines levels (pg/mL) of IL-10, IFNγ, IL-17A, IL-22, and transforming growth factor β (TGF-β) in the three subject groups (active, remission, and controls), intracellular levels of IL-10, IFNγ, and IL-17A, and correlation of serum IL-10 and TGF-β levels with T-regulatory cell (Treg) levels. (A) Serum IL-10 levels. (B) Serum IFNγ levels. (C) Serum IL-17A levels. (D) Serum IL-22 levels. (E) Serum TGF-β levels. (F) Intracellular levels of IL-10, IFNγ, and IL-17A in the CD3+CD4+ compartment of T-cells in a patient with active disease at 2 and 6 months after starting immunosuppressive treatment, following 24 h *in vitro* culture with anti-CD3 and anti-CD28 bead stimulation. (G) Bivariate correlation of serum TGF-β (pg/mL) and % Treg frequency, across all subject groups. (H) Bivariate correlation of serum IL-10 levels (pg/mL) and % Treg frequency, across all subject groups.

observed over time whereas the methylation of FoxP3 TSDR and TIGIT appeared more stable (**Figures 8F–H**).

### Increased Levels RORC Methylation May Be a Biomarker of Disease Response to Treatment

To determine whether there was a difference in methylation at epigenetic CpG sites associated with the T-effector transcription expression levels (RORγt and T-bet) assessed by flow cytometry in this study, bisulfite Amplicon Sequencing using an NGS approach was performed using cryopreserved PBMC as described earlier. Following biostatistical analysis of the epigenetic percentage (%) methylation data, a single CpG site demonstrating a high level of differential methylation between samples was selected for statistical comparison between the three groups at that site, as follows: RORC2/RORγT (151798858) and TBX21/T-BET (45810951) (**Figure 8**). The median CpG site% methylation level of RORC2 was higher in patients in clinical remission (55 ± 3.3 vs 46 ± 3.4, *P* = 0.016) and in control subjects compared with active (55 ± 4.3 vs 46 ± 3.4, *P* = 0.014) patients (**Figure 8D**). Within the active patient cohort, higher RORC% CpG methylation levels were observed at 12 months compared with 0 month after starting/ changing immunosuppressive treatment (**Figures 8I,J**). TBX21 CpG sites had overall low levels of methylation compared with the other T-cell-associated epigenetic methylation sites and no significant difference in the median% methylation levels was found between the three cohorts (**Figure 8E**).

(45810951). (F) Change in individual % methylation at FOXP3 TSDR (49117116). (G) Change in individual % methylation at FOXP3 promoter (49121204). (H) Change in individual % methylation at TIGIT (114012658). (I) Change in individual % methylation at RORC2/RORγT (151798858). (J) Change in individual %

### Previous Duration of Oral Immunosuppressive Treatment Is a Significant Predictor of Methylation at FoxP3 TSDR

methylation at TBX21/T-BET (45810951).

A bivariate analysis of duration of previous oral treatment (months) and CpG methylation at FoxP3 TSDR showed that CpG methylation levels at FoxP3 TSDR negatively correlated with previous duration of oral immunosuppressive treatment (simple linear regression shown in **Figure 9**). A multivariate linear regression model of methylation of the FoxP3 TSDR, which included the duration of previous oral immunosuppressive treatment (log transformed variable) as a predictor, was significant (*P* = 0.002) and explained around 42% of the variance in FoxP3 TSDR. Additional predictive variables, including TIGIT<sup>+</sup>, FoxP3<sup>+</sup>, IL-10, and TGF-β levels, added stepwise to the multiple regression model were not significant and were, therefore, removed from the final model. Subject age (years) was added to the regression model as a co-variate to adjust for its effect on methylation and the model remained significant (*P* = 0.006), explaining around 41% of the variance in FoxP3 TSDR (**Table 4**).

### TIGIT**+** Tregs Are a Sensitive but Not Specific Biomarker of Clinical Remission in Sight-Threatening Non-Infectious Uveitis

A logistic regression model that included TIGIT<sup>+</sup> Treg levels and subject age (years) as predictors of clinical remission was significant [χ<sup>2</sup> (2) = 18.402, *P* < 0.001] and a good fit for the data [Hosmer and Lemeshow test; χ<sup>2</sup> (8) = 5.786, *P* = 0.671] (**Table 5**). It correctly classified 84% subjects and explained around 45% of the variance in clinical remission of sight-threatening noninfectious uveitis. TIGIT was a significant predictor of clinical remission, whereas age was not significant. Peripheral blood TIGIT<sup>+</sup> Treg levels in the study population of patients with uveitis had a sensitivity of 92%, a specificity of 62%, a false positive rate

Table 4 | Summary of multiple regression model of methylation of FoxP3 Treg-specific demethylated region.


FoxP3 TSDR, and this relationship was significant (*P* = 0.012) when modeled

*\*\*Significant at the 1% level.*

with a simple linear regression.

Table 5 | Summary of logistic regression model of clinical remission.


*\*Significant at the 5% level.*

of 38% and a false negative rate of 8% for detecting clinical remission (Table S4 in Supplementary Material).

### DISCUSSION

In this study, we investigated the hypothesis that phenotypically stable Treg induced by immunosuppressive drugs are associated with sustained clinical remission of non-infectious uveitis. We did this by performing peripheral blood immunophenotyping of large group of patients, who had previously received immunosuppressive drug therapy and were in sustained clinical remission, and comparing these results with those from patients with active uveitis and control subjects. Our data show that the frequency of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Treg, TIGIT<sup>+</sup> Treg, T-bet<sup>+</sup> Treg and the ratio of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Treg to Th1 (T-bet<sup>+</sup>CD4<sup>+</sup>) T-cells are higher in patients in clinical remission compared with patients with active disease. T-bet<sup>+</sup>FoxP3<sup>+</sup> cell levels showed significant positive correlation with TIGIT<sup>+</sup>FoxP3<sup>+</sup> cell levels. The levels of CD4<sup>+</sup>FoxP3<sup>+</sup>TIGIT<sup>+</sup> Treg, CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Treg and CD4<sup>+</sup>FOXP3<sup>−</sup>TIGIT<sup>+</sup> T-cells in each subject sample were analyzed as ratios of TIGIT<sup>+</sup> Treg to TIGIT<sup>+</sup> T-cells (T-cells which expressed TIGIT but not FoxP3). These ratios were significantly higher in the remission group compared with the active group. This suggested that TIGIT expression proportionally increased in both the CD4<sup>+</sup>FoxP3<sup>+</sup>and the CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> compartment (with the highest levels TIGIT expression in the CD25<sup>+</sup> subset of FoxP3<sup>+</sup> cells) of T-cells in the clinical remission group compared with the active group, rather than increasing on all T-cells. This provides some supporting evidence for the utility of TIGIT as a marker of functional Treg, as opposed to being a marker of T-cell activation.

We utilized a novel approach of NGS bisulfate sequencing targeted toward T-cell genes of interest, to identify whether there was an epigenetic immune methylation pattern associated with clinical remission in uveitis. In this study, PBMCs from patients in clinical remission had lower levels of DNA methylation at CpG sites within the FOXP3 TSDR, FOXP3 promoter, and TIGIT loci compared with patients with active disease, supporting the higher levels of FoxP3 and TIGIT expression detected by flow cytometry in these patients. Through *in vitro* functional studies, we demonstrate that Treg from patients in clinical remission, a population in which there were high levels of Treg and TIGIT expression, are effective at suppressing T-cell proliferation. Together, these findings suggest that Treg from patients in clinical remission are polarized toward Type 1 inflammation and have a stable highly suppressive phenotype, characterized by intracellular T-bet and extracellular TIGIT expression, respectively, in comparison with subjects with active disease.

The cytokine milieu in non-infectious uveitis has been previously been investigated by our group and others (42–47). In this study, our focus was on investigating serum levels of specific cytokines associated with the Treg, Th1 and Th17 subsets identified in the flow cytometric analysis. Intracellular cytokine levels of IL-10, IFN-γ, and IL-17A were also evaluated in active disease at 0 and 6 months after starting therapy, at which time, disease had resolved. We found that patients in clinical remission had higher serum levels of IL-10 and TGF- β than patients with active disease. While control subjects also had higher levels of IL-10 than active patients, the higher serum TGF- β differentiated remission patients from controls and were perhaps related to high levels of Treg with suppressive function in this group. Tregs levels positively correlated with serum IL-10 and TGF- β levels, suggesting that Tregs may directly or indirectly influence the systemic cytokine milieu. Furthermore, increased intracellular expression of IL-10 by T-cells from active disease was increased 6 months after starting treatment, which corresponded with clinical resolution of disease.

Subjects' levels of Th1 and Th17 were evaluated by flow cytometric detection of their respective transcription factors, T-bet and RORγt, along with the corresponding serum cytokine levels of IFN-γ, IL-17, and IL-22 from peripheral blood samples. Upregulated T-bet expression in association with significantly increased IFN-γ levels has been previously demonstrated in patients with active uveitis secondary to Vogt–Koyanagi–Harada (VKH) syndrome, a bilateral chronic granulomatous panuveitis, which tends to be sight-threatening (48). Th17 cells have also been shown to be involved in ocular inflammation (49) and, more recently, B27<sup>+</sup> anterior uveitis (50). Paradoxical expression of effector CD4 T-cell transcription factors, such as T-bet and RORγt, by Treg has, however, been suggested to enhance Treg suppressive capacity in murine models of inflammation (51–54). These models demonstrate that intestinal RORγt <sup>+</sup>FoxP3<sup>+</sup> Treg induced *in vivo* by the local microbiota display a stable suppressive phenotype and exist in dynamic balance with pathogenic Th17 (53, 55). The local microenvironment, for example, the cytokine milieu and/or the microbiome, have been shown to regulate the Treg/Th17 balance and influence cell plasticity; disruption of this balance may lead to the development of inflammatory disease (54–57). Furthermore, during Type 1 inflammatory responses, Treg may upregulate T-bet in response to IFN-γ production (51). These T-bet<sup>+</sup> Treg have been observed to be phenotypically stable and selectively suppressive of Th1 (52). However, it is noted that other studies have demonstrated that T-bet expression is not required for Treg to suppress inflammation, in the context of CNS inflammation and colitis (58).

In this study, we show that in clinical remission and control groups, the levels of T-bet<sup>+</sup> Treg and ratios of Treg to Th1 are higher and serum levels of IFN-γ levels are lower compared with active uveitis patients. We also found that T-bet<sup>+</sup> Th1 levels in uveitis patients were overall lower in clinical remission compared with active and control subjects. Disease resolution in active patients over the course of 12 months appeared to be associated with an increase in levels of Treg and a decrease in levels of Th1. Intracellular expression levels of IFN-γ by CD4<sup>+</sup> T-cells were also decreased at 6 months after starting treatment. However, RORγt <sup>+</sup> Th17 and RORγt <sup>+</sup> Treg levels did not significantly differ between the three subject groups. Tregs are thought to regulate the expression of IL-22, a Th17 cytokine which facilitates inflammatory cell infiltration in uveitis (59, 60). Although the ratio of Treg to Th17 was not significantly greater in clinical remission, it was noted that serum IL-17A and IL-22 levels were significantly lower in the patients in clinical remission, compared with active and control subjects. RORγt <sup>+</sup> Treg levels were increased in all four active patients at 12 months compared with baseline, as the disease clinically resolved. However, unlike Th1 cytokine levels, Th17 cytokine levels did not appear to consistently decrease when disease resolved in active patients. Three out of four active patients had increased serum levels of IL-17A at 12 months after starting treatment, compared with baseline. Intracellular expression levels of IL-17A of by CD4<sup>+</sup> T-cells in active disease also increased at 6 months after starting treatment. Our results show that Th17 express FoxP3+ during disease resolution, in response to immunosuppressive treatment, but suggest that Th17-mediated inflammation is not driving inflammation and that FoxP3<sup>+</sup>RORγt <sup>+</sup> Tregs are not significantly associated with clinical remission.

It was further investigated whether clinical remission in non-infectious sight-threatening uveitis could be modeled by a linear relationship with predictive immunological variables, after adjusting for co-variates in the model. Multiple regression analysis showed that the duration of previous oral immunosuppressive treatment was the strongest predictor of hypomethylation at the FoxP3 TSDR (with the simple linear regression analysis demonstrating an observable hypomethylation effect of oral immunosuppression at the FoxP3 TSDR over several months of treatment). Hypomethylation at the FoxP3 TSDR has been shown to be important for Treg suppressive function and phenotypic stability and was significantly associated with clinical remission in our analysis. The association between previous duration of oral immunosuppression and methylation at FoxP3 TSDR remained significant after adjusting for subject age. This suggests that the duration of oral immunosuppression is an important factor in determining the phenotypic stability of Treg and supports the role of immunosuppressive drug induced Treg in the maintenance of clinical remission. Logistic regression showed that peripheral blood TIGIT levels, but not subject age, were a significant predictor of clinical remission in sight-threatening non-infectious uveitis. TIGIT<sup>+</sup> Treg levels had high sensitivity (92%) but low specificity (62%) for clinical remission. This infers a high false positive rate for using TIGIT+ Treg as a predictor of clinical remission, that is, a patient could test positive for disease remission when they actually have subclinical active disease. TIGIT<sup>+</sup> Treg levels detected by flow cytometry explained almost half of the variance observed in this model of clinical remission. However, many other factors not accounted for in this analysis, for example, posttranslational protein modifications and other immune markers such as chemokines, could also influence clinical remission.

Previous studies have compared Treg levels between active uveitis, inactive uveitis and/or healthy control subjects but there has been variation in subject groups and clinical immunophenotyping. A study from Chen et al. included 49 patients with VKH syndrome with uveitis and without uveitis, and showed a decreased percentage of CD4<sup>+</sup>CD25high T-cells, a decreased frequency of Foxp3+ expression in CD4+CD25high T-cells and reduced functionality of CD4<sup>+</sup>CD25high "Treg" in patients with active uveitis (61). Yeh et al. also demonstrated that patients with active uveitis (*n* = 8) have lower percentages of CD4<sup>+</sup>FoxP3<sup>+</sup> lymphocytes than patients with inactive disease (*n* = 12) (13). In contrast to this study, several patients in their series demonstrated evidence of systemic autoimmune disease, such as sarcoidosis and multiple sclerosis, which are, of themselves, associated with abnormal Treg populations or deficits in Treg suppressive function. Another study compared Treg levels in active and inactive uveitis patients and found that patients in remission on treatment (*n* = 25) had significantly increased levels of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Treg compared with active patients on treatment (*n* = 6) (12). However, over half of the patients (53%) recruited to their study had a diagnosis of anterior uveitis, which is not usually sight-threatening and does not require systemic immunosuppression.

Regarding the comparison of Treg levels in active patients with control subjects, the results from the literature are mixed. Two of aforementioned studies show a higher level of Treg in control subjects compared with uveitis patients (12, 61); however, this could be due to the inclusion of patients with systemic disease in the uveitis groups. By contrast, Yeh et al. comment that they have previously observed that the percentages of CD4+FoxP3+ lymphocytes do not differ between patients with uveitis and control subjects (unpublished data) (13). Molins et al. compared Treg levels and cytokine production in 21 patients with active non-infectious uveitis with 18 controls (18). They found that PBMCs from uveitis patients produced lower levels of IL-10 than those from controls but no differences were observed in Treg levels. TGF-β levels were not measured and functional assays were not performed in this study. Recently, Zhuang et al. compared 20 patients with active B27<sup>+</sup> anterior uveitis with healthy controls and observed an increase in CD4<sup>+</sup>IL-17<sup>+</sup> T cells, a decrease of CD4<sup>+</sup>CD25<sup>+</sup>Foxp3<sup>+</sup> Treg and a higher ratio of Th17/Treg in peripheral blood of patients compared with controls (50). The authors conclude that the imbalance of Th17 and Treg cells may play a vital role in the pathogenesis of B27<sup>+</sup> anterior uveitis. Of note, patients with B27<sup>+</sup> uveitis and anterior uveitis were excluded from this study due to the typically non-sight-threatening anterior anatomic localization of inflammation and systemic disease associations.

There has been interest in glucocorticoid-resistant Th17 cells which are refractory to Treg mediated suppression (62, 63), and these have been previously investigated in sight-threatening uveitis (64). Furthermore, a subset of "non-classic" Th17-derived Th1 have been described, which are demethylated at RORC, and are thought to play a pivotal role in the establishment and persistence of inflammatory disease (65). These non-classic Th1 express CD161, a marker of steroid-resistant human Th17, induced by RORC (66). In the clinical setting, however, most cases of uveitis show a good clinical response to high dose corticosteroids. Second-line immunosuppressive drugs are usually prescribed because the dose of corticosteroids required to suppress disease activity is very high and disease recurs on tapering the dose, rather than because the disease is refractory to corticosteroids. Only one patient in this study received biological therapy (Adalimumab), as the rest were clinically responsive to corticosteroids, with or without addition of conventional steroid-sparing agents. Twenty percent of the clinical remission group received previous therapy with cyclosporine, which has been shown to selectively attenuate steroid-resistant Th17 and to be efficacious in the treatment of steroid-refractory disease (67, 68). It is of note, in this study, that the clinical remission and control groups had significantly higher levels of RORC CpG methylation than the active group. Furthermore, within the active patient cohort, higher RORC% CpG methylation levels were observed at 12 months after starting/changing immunosuppressive treatment, when disease had resolved. RORC CpG hypermethylation could be a biomarker of clinical responsiveness to corticosteroid treatment; however, this requires further investigation.

The findings of this study are consistent with previous studies in uveitis demonstrating that Treg levels are higher and more functional in patients with inactive disease, compared with both patients with active and control subjects. We have previously observed that in ocular Behçet's disease, peripheral blood Treg levels are increased by 3 months' treatment with the immunomodulating therapy, pegylated interferon-α (pegIFNα), and that the increased levels of Tregs are still detectable at 12 months, 6 months after cessation of therapy (16). However, the expression of the co-inhibitory molecule TIGIT by Treg from patients in sustained clinical remission of disease, who are no longer on therapy, has not been previously investigated in uveitis. It is known that stable FoxP3 expression is ensured, at least partly, by DNA demethylation at the FOXP3 TSDR and that this infers commitment to the Treg lineage (28–31). In a genome-wide methylation analysis of nTregs from healthy individuals, it was found that hypomethylation at the TIGIT locus was one of the most significantly differentially methylated regions that distinguished naïve T-cells and nTreg; that it was not altered by activation and is required for FoxP3 binding (36). Other studies have demonstrated a key role for TIGIT in Treg-mediated suppression of Th1 and Th17 subsets (35, 37) and that TIGIT signaling in Treg directs their immunoregulatory phenotype in chronic disease settings (69). Our data show that hypomethylation at FOXP3 TSDR, FOXP3 promoter, and TIGIT CpG sites and higher levels of TIGIT<sup>+</sup> Tregs are associated with clinical remission, and may have value in determining remission, of relapsing and remitting sight-threatening non-infectious uveitis.

Our study addressed a need to define and investigate peripheral biomarkers for sustained clinical and biological disease remission in sight-threatening non-infectious uveitis. The total number of subjects recruited to this study (*n* = 60) represents one of the largest sample sizes of published clinical immunology studies of sight-threatening non-infectious uveitis. To improve the specificity of this study toward detecting peripheral blood biomarkers associated with ocular inflammation, we excluded patients with known systemic diagnoses, for example, sarcoidosis or Behçet's disease, and any other identifiable systemic inflammation. We were rigorous in our clinical phenotyping of the recruited subjects and undertook a comprehensive immunophenotypic analysis of CD4<sup>+</sup> T-cells derived from these groups, including functional and epigenetic methylation studies, which has not been previously performed in uveitis patients. Our group has previously analyzed aqueous humor samples from the eyes of patients with uveitis (47). However, the procedure of direct intraocular sampling carries risks of introducing infection, triggering or exacerbating inflammation and causing retinal detachment to the eye being sampled. Even when it is deemed clinically necessary, sequential/ repeated intraocular sampling should be avoided. Therefore, T-cell immunophenotyping by flow cytometry was performed on freshly isolated subject PBMCs (within 4 h of venepuncture), but additional intraocular sampling was not undertaken as part of this study. The 37 patients in clinical remission recruited to the study demonstrated a good clinical ocular response to immunosuppressive therapy given systemically, supporting the fact that systemic Treg and their associated microenvironment influence ocular immunity and that peripheral blood biomarkers have utility in monitoring ocular disease. Phenotypic analysis of Treg, Th1, and Th17 subtypes was based on "master transcription factor" expression, with prospective intracellular cytokine analysis only in selected subjects with active disease, followed up over a 12-month period. An in-depth analysis of intracellular cytokine production and methylation at signature cytokine loci could be considered for a future study. The number of Treg functional assays that were performed as part of this study was limited by the remaining PBMC numbers after the immunophenotypic analysis, rendering firm conclusions from these results difficult. The assay findings were in keeping with those from previous studies, which have demonstrated that in the inflammatory setting of autoimmune disease, Treg have reduced suppressive function (11, 70). However, they could be confirmed as part of a further study.

The clinical utility of TIGIT<sup>+</sup> Treg as a single biomarker for remission is limited by its potentially high false positive rate in this regard. It is possible that the calculated test sensitivity, specificity, false positive and negative rates were skewed by the overweighting of the sample toward remission patients rather than active patients. TIGIT levels explained almost half of the variance observed in the model of clinical remission, but there may be other clinically relevant factors, not accounted for in the model, which were also influencing clinical remission. These would need to be further investigated to develop better models for predicting clinical remission for individual patients, which could augment standard clinical assessment of patients. However, our results suggest that low levels of TIGIT may be a biomarker for increased risk of disease relapse in sight-threatening noninfectious uveitis. This could help to inform clinical management decisions, by indicating whether immunosuppressive drugs have induced functionally suppressive Treg.

In summary, our data show that CD4<sup>+</sup> T-cells in chronic relapsing, remitting sight-threatening non-infectious uveitis deviate toward the Th1 type and that Treg with upregulated levels of T-bet and TIGIT expression are associated with disease remission. These data provide supporting evidence that Tregs are part of the mechanism of clinical remission in non-infectious uveitis, which has been previously demonstrated in EAU model studies only (7–9). We suggest that the overall lower T-bet expression and IFN-γ levels detected in patients who achieve sustained clinical remission are related to potent suppression of Type 1 inflammation by phenotypically stable Treg induced by systemic immunosuppressive therapy. This is an important finding because a recent study in mice has shown that Treg expressing T-bet have a stable and functional phenotype, which potentiates suppression of Th1 autoimmunity (52), but it has not yet been demonstrated in clinical studies. Our results are also consistent with evidence from EAU that relapsing, remitting uveitic disease is closely associated with a Th1-like phenotype whereas monophasic inflammation is associated with a Th17-like phenotype (71).

Finally, these data are relevant to the recent interest in quantitative identification and isolation of viable, functional Treg for downstream clinical purposes, which include their generation *ex vivo* for immunomodulating therapies (72, 73). The results of this study provide evidence that TIGIT has utility as a marker of functional Treg for these therapies.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the UK National Research Ethics Service (NRES)—London Harrow Committee (13/LO/1653; 16039) with written informed consent from all subjects, in accordance with the Declaration of Helsinki. The protocol was approved by the Moorfields Eye Hospital National Health Service (NHS) Foundation Trust Department of Research and Development.

### REFERENCES


# AUTHOR CONTRIBUTIONS

RG, SL, VC, RS, XZ, ME, OT-N, DN, GG, CM, and NM contributed to the design of the study and experiments. RG, XZ, RS, MC, and NM performed the experiments, data capture, and analysis. RG, VC, SL, and ME performed the data interpretation. All the authors contributed to the critical revision of the manuscript for intellectual content and final approval of published version and agreed to be accountable for the work.

### ACKNOWLEDGMENTS

The authors thank all of the study participants for their contribution to the research. Additional thanks to Ms. Thurka Poobalasingam for technical advice regarding laboratory assays, flow cytometry Abs, and panel design; to Ms. Emma Bourne for assistance with sample preparation for NGS sequencing; to Dr. Dirk Paul and Dr. Amy Webster for advice on epigenetic analysis techniques; and to Mrs. Hazel Lawrence for assistance with arranging patient appointments and other clinical administration. An oral presentation of the research was made at the Symposium German/UK/Irish Meeting of the Young Investigators Network at the Deutsche Ophthalmologische Gesellshaft (DOG) Scientific Association of Ophthalmology in September 2016, and posters were presented at the Association for Research in Vision and Ophthalmology (ARVO) Annual Congress 2016 and the Rosetrees PhD Research Symposium 2016.

## FUNDING

This study was supported by a Pilot Project research grant awarded to SL and a continuation of Ph.D. research grant awarded to RG; both from the Rosetrees Trust (M363).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00907/ full#supplementary-material.

Figure S1 | Bisulfite amplicon epigenetic CpG methylation target sites for FOXP3 TSDR, FOXP3 promoter, TBX21, RORC2, and TIGIT loci.

Table S1 | Kruskal–Wallis *post-hoc* test pairwise comparisons of immunological markers across the three subject groups.

Table S2 | Primer sequences used in the Fluidigm assay for DNA methylation analysis (the Fluidigm CS1 and CS2 tags are highlighted in bold).

Table S3 | Unique Fluidigm barcodes added to the PCR products in the Fluidigm assay for DNA methylation analysis.

Table S4 | TIGIT+ Treg as a biomarker of clinical remission.


the role of IL-2, TGF-beta, and IL-10. *J Immunol* (2004) 172:5213–21. doi:10.4049/jimmunol.172.9.5213


profiles in uveitis. *Invest Ophthalmol Vis Sci* (2006) 47:272–7. doi:10.1167/ iovs.05-0790


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

*Copyright © 2018 Gilbert, Zhang, Sampson, Ehrenstein, Nguyen, Chaudhry, Mein, Mahmud, Galatowicz, Tomkins-Netzer, Calder and Lightman. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Cell-Intrinsic Roles for Autophagy in Modulating CD4 T Cell Functions

### *Elise Jacquin1,2\* and Lionel Apetoh1,2\**

*<sup>1</sup> INSERM, U1231, Dijon, France, 2Université de Bourgogne Franche-Comté, Dijon, France*

The catabolic process of autophagy plays important functions in inflammatory and immune responses by modulating innate immunity and adaptive immunity. Over the last decade, a cell-intrinsic role for autophagy in modulating CD4 T cell functions and differentiation was revealed. After the initial observation of autophagosomes in effector CD4 T cells, further work has shown that not only autophagy levels are modulated in CD4 T cells in response to environmental signals but also that autophagy critically affects the biology of these cells. Mouse models of autophagy deletion in CD4 T cells have indeed shown that autophagy is essential for CD4 T cell survival and homeostasis in peripheral lymphoid organs. Furthermore, autophagy is required for CD4 T cell proliferation and cytokine production in response to T cell receptor activation. Recent developments have uncovered that autophagy controls CD4 T cell differentiation and functions. While autophagy is required for the maintenance of immunosuppressive functions of regulatory T cells, it restrains the differentiation of TH9 effector cells, thus limiting their antitumor and pro-inflammatory properties. We will here discuss these findings that collectively suggest that therapeutic strategies targeting autophagy could be exploited for the treatment of cancer and inflammatory diseases.

### Keywords: autophagy, T cell, CD4, differentiation, adaptive immunity, immunotherapy

### INTRODUCTION

Macroautophagy, hereafter referred to as autophagy, is an evolutionary conserved catabolic pathway that ensures the degradation and recycling of intracellular components. During autophagy, cytosolic proteins and organelles are sequestered in a double membrane-bound structure called the autophagosome and eventually degraded upon fusion of the autophagosome with the lysosome (1). The autophagic flux accordingly refers to a complete catabolic process that ensures the breakdown of cargos and the release of the resulting macromolecules in the cytosol (2). Many autophagy-related (Atg) proteins have been found essential to orchestrate the formation of autophagosomes. This includes the upstream ULK complex (Ulk1/2, Rb1cc1/Fip200, Atg13, and Atg101) which regulates the induction of autophagosome formation. The class III phosphatidylinositol (PtdIns) 3-kinase (PIK3C3/Vps34) complex is then essential for the initial curvature of the phagophore and the recruitment of two ubiquitin-like conjugation systems (Atg4, 3/7/10, and 16l1/5/12) which conjugate Atg8 homologs (microtubule-associated protein 1A/1B-light chain 3/LC3, GABARAPs) to phosphatidylethanolamine and thus ensure the elongation of the phagophore and its closure (**Figure 1**) (1, 3).

While autophagy was initially described as a nonselective process induced under various stress conditions including nutrient deprivation (8, 9), it is now clear that autophagy can also specifically target organelles, proteins, or pathogens for degradation. This selective autophagy process requires autophagy receptors such as p62/Sqstm1 (5). p62/Sqstm1 recognizes poly-ubiquitinated cargos

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Ashutosh Chaudhry, Memorial Sloan Kettering Cancer Center, United States Cosima T. Baldari, University of Siena, Italy*

### *\*Correspondence:*

*Elise Jacquin elise.jacquin@inserm.fr; Lionel Apetoh lionel.apetoh@inserm.fr*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 30 January 2018 Accepted: 24 April 2018 Published: 09 May 2018*

### *Citation:*

*Jacquin E and Apetoh L (2018) Cell-Intrinsic Roles for Autophagy in Modulating CD4 T Cell Functions. Front. Immunol. 9:1023. doi: 10.3389/fimmu.2018.01023*

**150**

through its ubiquitin-associated domain and targets them for autophagic degradation through its LC3-interacting region motif (**Figure 1**). The levels of p62 are thus used as an index of autophagic degradation in combination with LC3 lipidation analysis (2).

Autophagy plays important functions in inflammatory and immune responses (10, 11). Atg proteins are crucial actors of cell-autonomous innate immunity. They contribute to pathogen elimination by antimicrobial selective autophagy (12) and LC3-associated phagocytosis (LAP), a recently identified form of noncanonical autophagy that directs LC3 lipidation onto endolysosomal compartments such as pathogen-containing phagosomes (13, 14). Moreover, both autophagy and LAP indirectly modulate adaptive immunity by contributing to antigen processing and major histocompatibility complex-restricted presentation to T cells (15). Over the last decade, accumulating evidence revealed a cell-intrinsic role for autophagy in CD4 T cell differentiation with direct consequences on physiopathology.

Here, work describing the regulation of autophagy in immune cells will be reviewed, with a focus on the functions of autophagy in CD4 T cell homeostasis and activation. We will also discuss recent findings showing that autophagy modulates the differentiation as well as the effector and regulatory functions of CD4 T cells and how this affects anticancer immune responses.

# AUTOPHAGY IS INDUCED IN CD4 T CELLS UPON ACTIVATION

The first evidence of autophagosome formation in CD4 T cells was reported 10 years ago by Li and colleagues (16–18). The authors assessed the presence of double membrane-bound autophagosomes in mouse primary CD4 T cells differentiated *in vitro* into effector T helper (TH) cells. They focused on TH1 and TH2 cells, which are, respectively, essential for cell-mediated and humoral immunity (19). Using transmission electron microscopy, they detected autophagosomes in about 20% of TH1 and TH2 cells activated *in vitro* with anti-CD3 and anti-CD28 antibodies, whereas they did not observe autophagosome in naïve resting CD4 T cells. These findings were confirmed by the expression of exogenous green fluorescent protein (GFP)–LC3 fusion protein Jacquin and Apetoh Autophagy in CD4 T Cells

in effector T cells and monitoring of GFP–LC3 puncta formation by fluorescence microscopy. With this method, the authors measured the proportion of TH1 cells undergoing autophagy in various culture conditions and determined that T cell receptor (TCR) signaling can sustain autophagy in effector CD4 T cells (17). Shortly after, a study conducted by Pua and colleagues gave further support to these data by detecting increased levels of LC3 lipidation by Western blot in primary mouse CD4 T cells cultured in the presence of anti-CD3 antibodies (18). Accordingly, both reports showed for the first time that key autophagy genes Atg5, Atg7, Beclin1, and LC3 are expressed in CD4 T cells (17, 18). They also found that downregulation of the expression of these genes and inhibition of Jun amino-terminal kinase (JNK)/mitogen-activated protein kinase pathways or PtdIns-3 kinase (PI3K) could inhibit autophagy in CD4 T cells, whereas the inhibition of mammalian target of rapamycin (mTOR) led to autophagy induction (17). These two initial reports, which evidenced that autophagy is induced in CD4 T cells upon TCR activation, were confirmed by several later studies conducted in mouse (4, 7, 20–22) and human primary CD4 T cells (23). In line with these studies, the expression of some autophagy proteins increases upon TCR activation. The activation of primary mouse CD4 T cells results in increased Beclin1 protein levels possibly after the activation of Becn1 promoter by p65/NF-κB (24). Upregulation of LC3 protein levels upon the activation of naïve CD4 T cells and the reactivation of differentiated effector CD4 T cells has also been reported. Collectively, these studies indicate that the molecular mechanisms of autophagy in CD4 T cells are similar to those described in other cell types and that this pathway can be modulated by pharmacological and genetic approaches.

### MOLECULAR MECHANISMS REGULATING AUTOPHAGOSOME FORMATION IN CD4 T CELLS

While TCR activation activates autophagosome formation in CD4 T cells, it has also been shown to induce mTOR activation (25). Botbol and colleagues have interrogated the involvement of mTOR in TCR-induced autophagy. To measure autophagic flux, the authors monitored LC3 lipidation in effector TH1 and TH2 cells cultured in various conditions in the presence of the inhibitors of lysosome function ammonium chloride (NH4Cl) and leupeptin. Surprisingly, effector TH1 and TH2 CD4 T cells reactivated with anti-CD3 and anti-CD28 antibodies did not display an increased autophagic flux upon mTOR inhibition with rapamycin, suggesting that this process is mTOR-independent. However, it cannot be excluded that TH1 and TH2 CD4 T cell reactivation on its own increased autophagic flux to its maximal level. This result may also suggest that TCR-induced autophagy signaling pathways other than mTOR can be involved in the regulation of autophagy in CD4 T cells such as the Janus tyrosine kinase (JAK)/ signal transducer and activator of transcription (STAT) signaling pathway. Indeed, the γ-chain cytokines interleukin (IL)-2 and IL-4, which are, respectively, produced by TH1 and TH2 cells upon reactivation, have been shown to contribute to autophagy induction in effector CD4 T cells in an autocrine/paracrine and JAK3-dependent manner (**Figure 1**) (4).

Data from the literature collectively suggest that autophagosome formation in CD4 T cells requires the canonical steps and molecules previously described in other cell types. For instance, overexpression of a kinase-dead mutant of the upstream autophagy protein ULK1 (ULK1 K461) in human naïve CD4 T cells impairs LC3 lipidation and autophagy (23). Likewise, reduced levels of autophagy have been described in CD4 T cells lacking the PI3K complex component Beclin1 (26) or the autophagy lipidation machinery proteins Atg7 (17, 18, 27), Atg3 (20), and Atg5 (28) (**Figure 1**). The requirement of PtdIns-3-phosphate (PI3P) formation by Vps34 during autophagy in CD4 T cells remains, however, elusive. Conditional deletion of Vps34 in mouse CD4 T cells does not completely suppress LC3 lipidation, and the PI3K inhibitor 3-methyladenine inhibits LC3 lipidation in Vps34-deficient cells, underscoring a possible contribution of Vps34-independent additional sources of PI3P in these cells (29–31). These findings can also indicate that noncanonical autophagy leading to endolysosomal LC3 lipidation may be active in CD4 T cells, a question that remains to be addressed.

Over the last decade, studies conducted in various models of autophagy deletion allowed for a better understanding of the molecular signals which control autophagy in CD4 T cells. These studies and others also strongly contributed to the understanding of the functions of autophagy in CD4 T cells. Indeed, while the early studies, which elegantly uncovered autophagosome formation in CD4 T cells, also suggested that autophagy could contribute to CD4 T cell death *in vitro* in response to specific stimuli (16, 17), they did not clearly ascribe a physiological function to autophagy in these cells.

## AUTOPHAGY MAINTAINS CD4 T CELL HOMEOSTASIS

The first study aiming at exploring the role of autophagy in CD4 T cell biology was conducted by Pua and colleagues in 2007. The authors generated autophagy-deficient T cells by transferring Atg5-deficient fetal hematopoietic progenitors into lethally irradiated wild-type congenic hosts and investigated how autophagy contributes to T cell development and functions. The chimeric mice generated by Pua and colleagues displayed a reduced thymus cellularity as well as severely altered CD4 and CD8 T cell compartments in peripheral lymphoid organs. The authors attributed these results to the defective capacity of Atg5−/− lymphoid progenitors to reconstitute lymphoid compartments and the impaired survival of Atg5<sup>−</sup>/<sup>−</sup> CD8 and, to a lesser extent, CD4 T cells in the periphery (18). Since Atg5 was known to interact with proteins from the apoptosis pathway (32), Pua and colleagues generated a second autophagy-deficient mouse model. To overcome the possible effect of autophagy deletion on hematopoietic progenitor cell proliferation capacity, they used an Atg7fl/fl:Lck-Cre system which allows for the deletion of Atg7 from the double-positive stage of thymocyte development. Analyses of thymocytes and mature T cells revealed a decrease in CD4 and CD8 single-positive thymocyte numbers and a dramatic loss of CD4 and CD8 T cells in the peripheral lymphoid organs of Atg7fl/fl:Lck-Cre animals, indicating a critical role for autophagy in maintaining peripheral CD4 T cell homeostasis (33). Along with lymphopenia in peripheral lymphoid organs, Pua and colleagues noted the acquisition of a "memory-like phenotype" by CD4 T cells characterized by a reduction in CD62L membrane levels and an increase in CD44 membrane levels (CD44high/CD62Llow) (33). These findings, which were further supported by several studies conducted in other models of autophagy deletion in mouse T cells, also demonstrated that autophagy is required for peripheral CD4 T cell survival (**Table 1**) (7, 20, 28–31, 34–36). Indeed, CD4 T cells isolated from Atg7fl/fl: Lck-Cre mice displayed increased apoptosis levels as shown by Annexin V and active caspase staining together with an imbalance of anti- and pro-apoptotic proteins. Consistent with the role of autophagy in eliminating damaged organelles and protecting from cell death, Atg7-deficient CD4 T cells displayed an increased mitochondrial content and levels of reactive oxygen species (ROS) as well as an impaired regulation of mitochondrial number during their development (33). Supporting this idea, transcriptomic analyses conducted in Atg5-deficient thymocytes revealed an enrichment of transcripts encoding mitochondrion-associated proteins which can account for the increased mitochondrial mass observed in peripheral mature autophagy-deficient CD4 T cells (28). Furthermore, *in vitro* deletion of Atg3 in splenic T cells from Atg3fl/fl estrogen receptor-Cre mice had no acute effect on organelle homeostasis and CD4 T cell survival but induced temporal accumulation of mitochondria and endoplasmic reticulum cell death after long-term culture. This suggested that the effect of autophagy deletion on peripheral CD4 T cell homeostasis is due to an accumulation of defects during their development rather than an acute phenomenon (20).

While further studies based on the deletion of Vps34 in CD4 T cells confirmed the critical role of autophagy in maintaining CD4 T cell survival during their development (30, 31), it was proposed that autophagy and autophagy proteins could favor CD4 T cell survival upon TCR stimulation. While Pua and colleagues initially suggested that *in vitro* TCR activation with anti-CD3 antibodies leads to the apoptotic cell death of Atg5-deficient CD4 T cells isolated from Atg5<sup>−</sup>/<sup>−</sup> chimeric mouse splenocytes (18), further work conducted in mouse models of Atg7 deletion found similar levels of apoptosis in Atg7-competent and -deficient CD4 T cells upon *in vitro* TCR stimulation (6, 7). Conversely, targeting upstream components of the autophagy pathway seems to impair CD4 T cell survival upon TCR stimulation. Indeed, Kovacs and colleagues showed that the deletion of Beclin1 in CD4 T cells led to increased apoptosis upon TCR activation with anti-CD3 and anti-CD28 antibodies as indicated by increased levels of the pro-apoptotic genes Bim and pro-caspases 8 and 3 as well as DNA fragmentation. The authors proposed that the degradation of pro-caspase 8 by autophagy prevents its accumulation in protein complexes that function as signaling platforms to activate apoptosis. The ability of caspase inhibitor addition and modulation of pro- and antiapoptotic protein expression levels to prevent cell death further supported the apoptosis-mediated death of autophagy-deficient cells (34). Similarly, overexpression of a kinase-dead mutant of ULK1 in human CD4 T cells has been shown to induce mitochondria and ROS accumulation leading to CD4 T cell apoptosis upon TCR stimulation (23). It is worth noting that autophagy-independent functions Vps34 have also been shown to modulate CD4 T cell survival. Indeed, Vps34 deletion leads to CD4 T cell death, independently of autophagy and rather through impairment of trafficking and surface expression of IL-7 receptor, and regulation of IL-7 signaling (29).


*Phenotypic observations made from different mouse models of autophagy deficiency reported in the literature and identifying a role for autophagy in CD4 T cell homeostasis. SP, single positive; Lck, lymphocyte protein tyrosine kinase; ER, estrogen receptor; Cre, Cre recombinase; vav, vav guanine nucleotide exchange factor 1.*

By controlling activation-induced death of CD4 T cells but also their proliferation, differentiation, and cytokine production, TCR engagement is crucial for CD4 T cell homeostasis. Early after the first observation of autophagosomes in CD4 T cells, autophagy has not only been shown to be activated upon TCR engagement but also to modulate CD4 T cell responses to this signal.

## AUTOPHAGY CONTROLS CD4 T CELL PROLIFERATION IN RESPONSE TO TCR ACTIVATION

Along with demonstrating the crucial role of autophagy for CD4 T cell homeostasis and survival in peripheral organs, Pua and colleagues also showed that autophagy induction following TCR activation promotes TCR-driven proliferation of CD4 T cells. Analysis of cell proliferation by 5(6)-carboxyfluorescein N-hydroxysuccinimidyl ester dilution revealed that Atg5 deficient CD4 T cells display impaired proliferation following *in vitro* TCR activation with anti-CD3 antibodies, anti-CD28 antibodies, and IL-2. However, the normal levels of membrane TCR and activation markers CD69 and CD25 indicated that autophagy inhibition does not alter TCR-driven activation of CD4 T cells (18). While these findings were confirmed later in other mouse models of Atg3, Atg5, Atg7, and Vps34 deletion in CD4 T cells (6, 7, 20, 28, 31), several groups aimed at uncovering the molecular mechanisms linking autophagy and CD4 T cell responses to TCR activation.

Jia and colleagues proposed that the selective degradation of the cell cycle inhibitor CDKN1B/p27Kip1 by p62-dependent autophagy may account for the TCR-driven proliferation defect observed in autophagy-deficient cells, as Atg3- and Atg7-deficient CD4 T cells accumulate CDKN1B/p27Kip1 and fail to enter S phase after *in vitro* TCR stimulation (**Figure 1**) (6). In an earlier study, the same group demonstrated that Atg7 deletion in CD4 T cells led to the accumulation of endoplasmic reticulum and impaired calcium mobilization upon *in vitro* TCR stimulation associated with an increased endoplasmic reticulum stress (37). However, this calcium mobilization defect does not seem to affect CD4 T cell activation since Atg7-deficient CD4 T cells display intact proximal TCR signaling and NF-κB pathway and maintain IL-2 production upon *in vitro* TCR activation (37).

The induction of autophagy upon TCR activation has also been proposed to regulate the energy metabolism changes required for CD4 T cell activation. Indeed, in TH1 cells, both pharmacological and genetic inhibition of autophagy impair the production of ATP, interferon-gamma (IFN-γ), and IL-2 following TCR activation. These defects of autophagy-deficient TH1 cell functions can be reversed by the addition of methyl pyruvate, a cell-permeable glucose metabolism intermediate that can restore electron transport chain, and oxidative phosphorylation activity. Interestingly, the analysis of autophagosome content in TH1 cells revealed a change in autophagic cargos from mitochondria to soluble cytosolic component upon TCR activation (**Figure 1**). The authors proposed that this selective exclusion of mitochondria from autophagic degradation in response to TCR engagement may contribute to signal transduction and adaptation to energy requirement modification (7). Mitochondrion remodeling has been shown to control T cell metabolic adaptation, driving their response and fate following TCR activation, reinforcing the central role of this organelle in CD4 T cell biology (38).

A contrasting study conducted in effector TH2 cells has proposed that selective autophagy may prevent a sustained TCR activation by targeting the adaptor protein B-cell CLL/lymphoma 10 for degradation and thus limiting TCR-driven NF-κB activation. Nevertheless, this process does not seem to contribute to naive CD4 T cell response to TCR activation and has not been tested in other effector CD4 T cell subsets, suggesting that it may be limited to TH2 cells (39).

Collectively, these data demonstrated the importance of autophagy for CD4 T cell homeostasis not only during their development but also upon activation. These results are in line with increased levels of autophagy detected in the CD4 T cells of rheumatoid arthritis that may account for their hyperactivation and resistance to apoptosis (22). Until now, the specific role of autophagy in the different subtypes of effector and regulatory T cells has received less attention. Recent studies have shown that the function of autophagy may vary according to the subsets of CD4 T cells and most importantly that autophagy modulates CD4 T cell differentiation and functions by regulating energy metabolism or intracellular component levels.

### AUTOPHAGY REGULATES EFFECTOR AND REGULATORY CD4 T CELL DIFFERENTIATION AND FUNCTIONS

In 2016, Wei and colleagues reported that autophagy is essential to maintain the functional integrity of the suppressor CD4 Foxp3<sup>+</sup> Treg cells (40), which suppress effector T cell functions (41, 42). The specific deletion of Atg7 in Foxp3<sup>+</sup> CD4 T cells leads to reduced proportions of Foxp3<sup>+</sup> CD4 T cells in peripheral organs and a higher active caspase-3 staining of the remaining cells, indicating a survival defect of autophagy-deficient Treg cells (40). The authors reported an increased proportion of Ki67-positive CD4 Treg cells purified from Atg7fl/fl:Foxp3Cre mice, probably reflecting a niche-filling behavior of the surviving cells. However, Atg7-deficient Treg cells proliferate normally in response to *in vitro* TCR stimulation with anti-CD3 antibodies, anti-CD28 antibodies, and IL-2 and after adoptive transfer into Rag1<sup>−</sup>/<sup>−</sup> mice. This suggests that Atg7 may be dispensable for TCR-induced proliferation of Treg cells (40). Compared to autophagy-competent cells, Atg7-deficient Treg cells display an increased glycolytic activity upon TCR activation. This reveals an important role for autophagy in negatively regulating glucose metabolism in Treg cells by restraining the mTOR complex 1 (mTORC1)–c-Myc pathway. Indeed, the high mTORC1 activity observed in activated Atg7-deficient Treg cells and characterized by high phosphorylation levels of ribosomal protein S6 (S6) and eukaryotic translation factor 4E binding protein 1 (4EBP1) has been shown to rely on increased PI3K and pyruvate dehydrogenase kinase 1 abundance and activation and to be responsible for upregulated c-Myc expression and altered transcription programs (40). Importantly, autophagy controls transcriptional programs in Treg cells in a similar mTORC1-dependent manner. Atg7-deficient Treg cells display a reduced expression of Foxp3, Foxo, and Bach2, as well as an enrichment of effector T cell differentiation pathways that can be rescued by mTORC1 inhibition with rapamycin. This indicates a central role for autophagy in negatively regulating effector programs and maintaining Treg cell stability. These results are in line with the loss of Foxp3 expression and the aberrant production of IFN-γ and IL-17 observed in Atg7-deficient Treg cells *in vitro* and *in vivo* (**Figure 2**). Moreover, autophagy is required for the ability of Treg cells to suppress antitumor responses *in vivo* as illustrated by an impaired tumor growth of MC38 colon adenocarcinoma cells in Atg7fl/fl:Foxp3Cre mice associated with a loss of Treg cells and an increased expression of IFN-γ from T cells at the tumor site. Although Atg7 deficiency also leads to survival defects of Treg cells, it is worth noting that autophagy seems to repress Treg cell apoptosis through a mechanism that does not solely rely on mTORC1 and which may thus be distinct from its role in Treg cell stability and functional integrity (40) (**Figure 2**).

In line with this, Le Texier and colleagues reported that the specific deletion of Atg7 in Foxp3 + CD4 T cells leads to the profound loss of the Helios + TIGIT + subset of Treg cells in the spleen and the bone marrow. The authors showed that autophagy is required for the survival of these Helios + TIGIT + Foxp3 + CD4 T cells and the maintenance of their immunosuppressive functions. Indeed, aged Atg7fl/fl:Foxp3Cre mice develop spontaneous T cell activation and multi-organ inflammation, suggesting an important role for autophagy-dependent Treg cells in the suppression of autoimmunity. Importantly, Treg-intrinsic autophagy promotes Treg reconstitution upon stem cell transplantation and attenuates graft versus host disease, supporting the idea that autophagydependent Treg cells are critical for tolerance (27).

In another model, the specific deletion of Atg16l1 in both CD4 T cells and Foxp3<sup>+</sup> CD4 T cells results in a loss of Treg cells in mouse intestine and aberrant expression of the effector cytokines IFN-γ and IL-17 by the remaining Treg cells, associated with TH2-driven intestinal inflammation toward dietary antigens and commensal microbiota (36). In this study, no evidence of defective *in vitro* Treg cell differentiation or stability was detected. Thus, Kabat and colleagues attributed the impairment of Atg16l1-deficient Treg cell survival to their aberrantly high glycolytic metabolism which prevents their metabolic adaptation to the intestinal mucosal environment. Indeed, the authors found that glycolysis gene expression levels are higher in Atg16l1-deficient Treg cells compared to their autophagy-competent counterparts, especially

maturation of CD4 Treg cells and activate both mammalian target of rapamycin (mTOR)C1 and autophagy. Autophagy-deficient CD4 Treg cells lacking Atg5 or Atg7 display an increased mTORC1 activity and upregulation of c-Myc and glycolytic metabolism (36, 40). This dysregulation of mTORC1/c-Myc pathway and energy metabolism leads to the loss of Foxp3, Foxo, and Bach2 which are essential for CD4 Treg cell differentiation and maintenance, and to an aberrant production of inflammatory cytokines (40). Furthermore, autophagy deficiency impairs CD4 Treg cell survival through a mechanism that seems to only partly rely on mTORC1 activity (36, 40). Autophagy thus plays a crucial role in maintaining the stability and functional integrity of CD4 Treg cells by restraining mTORC1–c-Myc pathway and glycolytic metabolism as well as promoting survival. Importantly, autophagy is required to maintain the ability of CD4 Treg cells to suppress antitumor immune responses *in vivo* (40).

in Foxp3<sup>+</sup> Treg cells sorted from the colon lamina propria of young Atg16l1-deficient mice. Furthermore, Treg cells differentiated *in vitro* from Atg16l1-deficient naive CD4 T cells display higher expression levels of c-Myc and a panel of glycolysis genes, reduced expression levels of lipid metabolism-associated genes as well as higher rates of glycolysis and oxidative phosphorylation than autophagy-competent Treg cells, confirming that autophagy negatively controls glucose metabolism in Treg cells (36).

This study also revealed that autophagy deletion has an opposite effect on TH2 cells, promoting their expansion in peripheral tissues through both Treg cell-mediated control and cell-intrinsic mechanism. Glucose metabolism is not altered by autophagy deficiency in effector TH2 cells which display high levels of c-Myc and glycolysis gene expression as well as glycolytic energy metabolism, irrespective of their Atg16l1 genotype (36), consistent with their high glycolytic rate compared to other CD4 T cell subsets (43). The constitutively high levels of glycolysis displayed by TH2 cells may allow them to adapt to the metabolic switch toward glycolysis induced by autophagy deficiency. GATA-3 having been previously associated with c-Myc-driven metabolic reprogramming leading to glycolysis induction upon TCR activation (44), the authors proposed that this transcription factor essential to TH2 cell differentiation may orchestrate the metabolic adaptations induced by autophagy deficiency. This suggests that the differential expression of transcription factors such as GATA-3 in the various subsets of regulatory and effector CD4 T cells may account for the different metabolic adaptations to autophagy deficiency (36). Together, these studies emphasize how the contribution of autophagy in energy metabolism varies according to the subsets of CD4 T cells, explaining, at least partly, the differential impact of autophagy deletion on CD4 T cell subset differentiation, stability, and functions.

In line with this, our recent study focused on the role of autophagy in the differentiation and functions of TH9 cells, characterized as IL-9 producing effector T cells which contribute to autoimmune diseases (45, 46) and exert potent anticancer functions (47–51). Using pharmacological and genetic approaches of autophagy inhibition in CD4 T cells, we showed that autophagy restrains TH9 cell differentiation and effector functions through a cell-intrinsic mechanism independent of energy metabolism modulation (52, 53). While we confirmed that mTORC1 activity is enhanced in Atg5-deficient *in vitro* differentiated Treg cells compared to autophagy-competent controls, we observed no difference in S6 and 4EBP1 phosphorylation in Atg5<sup>−</sup>/<sup>−</sup> CD4 T cells differentiated *in vitro* in effector TH9 cells, indicating that autophagy does not control mTORC1 signaling and thus glycolytic metabolism in TH9 cells. We proposed that TH9 cells display similar responses than TH2 cells regarding glycolytic metabolism in the absence of autophagy (53).

Although the inhibition of autophagy through genetic (Atg5 deletion in CD4 T cells) and pharmacological chloroquine approaches reduces the viability of *in vitro* differentiated effector and regulatory CD4 T cells, it specifically enhances IL-9 production and TH9 cell differentiation *in vitro*, without affecting the differentiation of other effector T cell subsets or skewing them toward TH9 differentiation program (52, 53). This role for autophagy in repressing IL-9 production by differentiating TH9 cell relies on the degradation of PU.1, the master TH9 cell transcription factor (54), by selective autophagy. Indeed, upon TH9 cell differentiation, K63 ubiquitination of PU.1 leads to its specific recruitment by the autophagy receptor p62 *via* its ubiquitin-associated domain and the subsequent degradation of PU.1 in autophagosomes (**Figure 3**). TH9 cells treated with the inhibitor of lysosome function chloroquine (CQ) display increased IL-9 secretion and antitumor properties upon adoptive transfer. Indeed, CQ-treated *in vitro* differentiated TH9 cells have enhanced suppressive activity against melanoma tumor growth compared to TH9 cells displaying intact autophagic flux, suggesting that CQ may be considered in the context of adoptive T cell therapy of cancer (52, 53). Furthermore, genetic inhibition of autophagy in T cells leads to IL-9-dependent inhibition of tumor growth in mice bearing MC38 colon cancer and B16 melanoma and enhanced IL-9-producing CD4 tumor-infiltrating lymphocytes. While these results lend further support to the potent anticancer functions of TH9 cells (**Figure 3**), they suggest that autophagy induction in tumor microenvironment represses TH9 cell development and/or function, potentially explaining the low frequencies of TH9 cells detected in tumor lesions (48). This also suggests that inhibiting autophagy may restore antitumor immunity by concomitantly promoting TH9 cell-dependent antitumor immunity and impairing Treg cell stability (40), providing new opportunities for cancer immunotherapy.

Figure 3 | Selective autophagy limits TH9 cell differentiation and antitumor functions. Upon TH9 cell differentiation, K63 ubiquitination of PU.1, the master transcription factor of TH9 cells, leads to its specific recruitment by the autophagy receptor p62 *via* its ubiquitin-associated domain and the subsequent degradation of PU.1 in autophagosomes. Genetic (Atg5−/−) and pharmacological inhibition of autophagy with chloroquine (CQ) prevent PU.1 degradation and enhance IL-9 secretion from TH9 cells, thus increasing their antitumor functions *in vivo* (53).

Jacquin and Apetoh Autophagy in CD4 T Cells

Through their pro-inflammatory functions, TH9 cells play an important role in the development of inflammatory bowel diseases (IBDs). Indeed, IL-9-secreting TH9 cells induce colitis in mice (45, 55), and PU.1-expressing cells as well as target cells expressing IL-9 receptor are frequently detected in the gut mucosa of IBD patients (56, 57). Autophagy contributes to inflammation, and IBD pathogenesis through multiple processes (58, 59) and mutations in Atg genes such as *NOD2*, *ATG16L1,* and *ULK1* leading to autophagy deficiency have been associated with IBD susceptibility in humans (59). Our findings raise the hypothesis that autophagy may limit inflammation by repressing TH9 cells' inflammatory properties. Combined with the evidence that autophagy plays a central role in Treg cell-mediated immune tolerance in peripheral tissues (36, 40, 60) and limits TH2 expansion in intestinal mucosa, they provide new insights that could be exploited for therapies against allergies as well as inflammatory and autoimmune diseases.

Collectively, investigators have demonstrated that autophagy levels in CD4 T cells are regulated in response to environmental signals and that autophagy controls CD4 T cell homeostasis and functions. The molecular mechanisms, which contribute to autophagy-driven modulation of effector and regulatory T cell functions, have also been clarified. The role for autophagy inhibition in restraining Treg cell functions and promoting TH9 cell differentiation evidenced in recent studies opens

### REFERENCES


new therapeutic perspectives regarding the combination of autophagy inhibitors with anticancer immunotherapies. Recent cancer clinical trials have suggested that the clinically approved autophagy inhibitor hydroxychloroquine may be considered for cancer therapy although there is still a need for the development of more specific and potent autophagy inhibitors (61). Further work is now required to determine the effect of such treatments on CD4 T cell autophagy, differentiation, and functional integrity *in vivo*.

# AUTHOR CONTRIBUTIONS

EJ and LA wrote the manuscript and EJ designed the figures.

### FUNDING

The authors are supported by grants from the H2020 European Research Council (677251) (to LA) and the Fondation pour la Recherche Médicale (ARF20170938687) (to EJ). This work was supported by a French Government grant managed by the French National Research Agency under the program "Investissements d'Avenir" with reference ANR-11-LABX-0021 (Lipstic Labex), by the Conseil Régional de Bourgogne and the European Union through the PO FEDER-FSE Bourgogne 2014/2020 programs.

targeting single membranes. *Nat Cell Biol* (2011) 13:1335–43. doi:10.1038/ ncb2363


generates IL-9+ IL-10+ Foxp3(-) effector T cells. *Nat Immunol* (2008) 9:1347–55. doi:10.1038/ni.1677


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

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

# Transcriptional Control of Th9 Cells: Role of Foxo1 in Interleukin-9 Induction

### *Sakshi Malik and Amit Awasthi\**

*Immuno-Biology Laboratory, Center for Human Microbial Ecology, Translational Health Science and Technology Institute, Faridabad, India*

Interleukin (IL) 9-producing helper T (Th) 9 cells play a major role in contributing immunity against extracellular pathogens. In addition, the role of Th9 cells was demonstrated in the pathogenesis of allergic, skin, and intestinal inflammation. The functions of Th9 cells were further extended in antitumor immune response, as Th9 cells were suggested to be potent antitumor Th cells. Given the pleotropic functions of IL-9 in various pathophysiological conditions, it is essential to understand the differentiation and stability of Th9 cells and other IL-9-producing T cells. In addition to Th9 cells, Th2 and Th17 cells as well as induced Foxp3+ regulatory T cells (iTregs) cells also produce IL-9, but how IL-9 production is regulated in these cell types is not yet clearly defined. Although Th2, Th9 and Th17 cells as well as iTregs develop in the presence of distinct differentiating factors, yet they all express IL-9 together with their own lineage specific cytokines. Here, in this review, we summarize the current understanding of signaling pathways that lead to the promotion of differentiation of Th9 cells and IL-9 induction in Th2 and Th17 cells, as well as in iTregs. We further discuss the transcriptional regulation of Th9 cells in context of Foxo1, as an essential transcription factor required for the development and functions of Th9 and other IL-9-producing T cells.

Keywords: T helper cells, T helper 17 cells, inflammation, Foxo1, interleukin-9

### INTRODUCTION

Almost more than two decades ago, interleukin (IL)-9 was described as T cell growth factor, which was later categorized as one of the Th2 cytokine (1, 2). After the association of IL-9 with Th2 cells was established, much of its functions was tested in Th2-biased mouse models of allergic inflammation and *Leishmania* infection, which further reinforced its classification as Th2 cytokine (3, 4). The functions of IL-9 was not greatly discussed separately, as it was thought to be enhanced during disease pathology induced by Th2 cells. Nonetheless, the genetic association studies identified the association of IL-9 and IL-9R with human asthma, which was further validated in mouse model of allergic inflammation in asthma (5, 6). Pulmonary overexpression of IL-9 was seen to be associated with inflammatory infiltration of eosinophils and lymphocytes (7). One of the striking findings in this model was greatly enhanced mast cell infiltration within the airway epithelium. This was in agreement with other findings which identified that lung-expression of IL-9 increased IgE-mediated disease pathology and mucus production in mouse model of asthma. These observations were further validated in transgenic mice in which lung-specific inducible IL-9 production was controlled by doxycycline (8). Consistent with constitutive expression of IL-9, doxycycline inducible IL-9 production in the lung promotes lymphocytic and eosinophilic infiltration with mucus production and mast cell hyperplasia, which leads to lung immune-pathology (8). In addition, IL-9 overexpression further enhanced the production of Th2 cytokines such as IL-4, IL-5, and IL-13. Strikingly, neutralization of IL-13 leads to inhibition of both lung inflammation and mucus

### *Edited by:*

*Paulo Vieira, Institut Pasteur, France*

### *Reviewed by:*

*Pablo Pereira, Institut Pasteur, France Jinfang Zhu, National Institute of Allergy and Infectious Diseases (NIAID), United States*

> *\*Correspondence: Amit Awasthi aawasthi@thsti.res.in*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 27 February 2018 Accepted: 20 April 2018 Published: 09 May 2018*

### *Citation:*

*Malik S and Awasthi A (2018) Transcriptional Control of Th9 Cells: Role of Foxo1 in Interleukin-9 Induction. Front. Immunol. 9:995. doi: 10.3389/fimmu.2018.00995*

production resulting in suppression of lung immune-pathology in allergic inflammation. In order to further refine the functions of IL-9 in comparison to other Th2 cytokines, IL-9-deficient mice were generated. IL-9-deficient mice manifest highly defined phenotype of Th2 responses such as mast cells proliferation and mucus production without affecting worm expulsion (6).

The clarity in IL-9 functions in immune responses came with identification and discovery of IL-9-producing Th9 cells (9, 10). It was identified that the activation of naïve T cells in the presence of TGF-β1 together with IL-4 induced the generation of IL-9 producing helper T (Th) cells, and therefore these cells were referred to as Th9 cells (9, 10). While TGF-β1 alone induces Foxp3 expression and generated immunosuppressive Foxp3<sup>+</sup> induced Tregs (iTregs), addition of IL-4 suppressed TGF-β1 induced Foxp3 expression (9). On the other hand, TGF-β1 suppressed IL-4 functions, which is otherwise known to induce the differentiation of Th2 cells. While TGF-β1 and IL-4 suppressed each other's respective functions such as Foxp3 induction and Th2 differentiation, but two cytokines together induced a new pathway of Th9 cell differentiation. GATA3 is a common transcription factor of two IL-9 producing sister populations, i.e., Th2 and Th9 cells and one of the major function of GATA-3 in Th9 cells is to counteract the TGF-β1-induced Foxp3 expression, which in turn limit the ability of GATA-3 to induce *Il4* expression (9). Later on, it was identified that other cytokines such as IL-2, IL-1, IL-25, IL-33, IL-7, and TSLP further enhanced the differentiation of Th9 cells induced by TGF-β1 and IL-4 (11–16).

### DIFFERENTIATION AND TRANSCRIPTIONAL REGULATION OF Th9 CELLS

The regulatory network of transcription factors in Th9 cells seems to be quite complex, as Th9 cells express number of transcription factors. Nonetheless, classification of a unifying master transcription factor is still ambiguous, as most of the transcription factors expressed in Th9 cells is also co-expressed by other T helper lineages. In order to simplify the complex network of Th9 cell transcription factors, the different transcription factor involved in Th9 cells development can be distributed into different groups depending upon their priming signals. For example, downstream of TGF-β1, Smad-dependent pathway majorly regulates RBP/ Notch signaling while TAK1-mediated Smad-independent pathways control the induction of Id3 and HIFα in Th9 differentiation (17–19). PU.1, which is one of the major transcription factor, is regulated by TGF-β1, and is not dependent on Smad2/3 (20). Although IL-4–STAT6 signaling seems to regulate BATF/ IRF-4 and ETV5 in Th9 cells, TGF-β1 also enhances binding of IRF-4 to *Il9* locus (21–23). In addition to IRF-4, other interferon regulatory factors such as IRF-1 and IRF-8 are also involved in IL-9 regulation in Th9 cells (24–26). While IL-1β induces IRF-1, TGF-β/Smad3 pathway induces IRF-8 in Th9 cells (24, 26). T cell receptor (TCR)-dependent signals regulate the function of NFAT, TNF superfamily, NF-κB, and Foxo family members in various T cell subsets (27–30, 62). It may be possible that these factors work in a concerted manner to drive optimal differentiation of Th9 cells, however, complex regulatory network of Th9 cells is

2 **160** not yet defined. Nonetheless, a recent report has identified as to how IRF-8 cooperatively interacts with IRF-4/BATF/PU.1 to promote Th9 development while simultaneously repressing *Il4* transcription, suggesting the involvement of large molecular transcriptional complexes in Th9 differentiation akin to Tregs (26, 31).

PU.1, an ETS family transcription factor is one of the first factors that were seen to be specifically associated with IL-9 induction in Th9 cells (32, 33). PU.1 imprints heterogeneity in Th2 cells in respect of IL-9 induction, as overexpression of PU.1 increases IL-9 production with concomitantly decreases type-2 cytokines production. Molecularly, PU.1 restricts the ability of GATA3 and IRF-4 to induce Th2 cytokines signature, and thereby promoting the differentiation of low level of IL-4 production in Th2 cells. Furthermore, mice with PU.1 deficiency in T cells develop attenuated allergic inflammation in lungs in response to OVA, which is found to be associated with reduced amounts of IL-9 production and Th2 cytokines. In addition, PU.1-dependent IL-9 induction was linked to the pathology of intestinal inflammation, as IL-9 deficient as well as *Spi* conditional deficiency in T cells were seen to have reduced clinical and histological signs in oxazolone-induced colitis model (33).

In addition to PU.1, another ETS family member, ETV5, exerts dominant effect on IL-9 induction in Th9 cells, as ETV5-deficient T cells have shown reduced *Il9* expression. Consistently, ectopic expression of ETV5 enhanced induction of IL-9 in Th9 cells (23). Interestingly, IL-9 production from Th9 cells was found to be further suppressed upon a combined deficiency of both PU.1 and ETV5 as compared to either PU.1 or ETV5 single deficiency, suggesting that PU.1 and ETV5 work in concerted manner to induce optimal IL-9 induction and Th9 cell differentiation. Although ETV5 and PU.1 belong to the same family, their induction and functions differ in Th9 cells. While PU.1 was shown to be induced by TGF-β1 signaling, ETV5 was found to be induced and essential for IL-4–STAT6 axis in Th9 differentiation. Mechanistically, ETV5 physically binds to *Il9* locus at sites that are distinct from PU.1 DNA-binding motif, and transactivate IL-9 induction in Th9 cells. The functions of other ETS family member such as Elk3 and Etv6 were also tested in Th9 cells differentiation, but found not be essential for IL-9 induction in Th9 cells. This implies that among other ETS family member, ETV5 and PU.1 play selective and specific role of in Th9 differentiation.

Although the role of IRF-4 has been identified in the development and functions of Th2 and Th17 cells, it has been shown that IRF-4 is essential for differentiation of Th9 cells (22). In fact, both BATF and IRF-4 have been shown to work cooperatively in Th9 cells development, as deficiency of BATF have shown to reduce the binding of IRF-4 to *Il9* promoter or *vice-versa* (21). Consistently, ectopic expression of BATF failed to rescue IL-9 production in the absence of IRF-4 in Th9 cells (21). Similar to PU.1-deficient mice, IRF-4-deficient mice display attenuated signs of development of IL-9-dependent OVA-induced allergic inflammation in lungs in mouse model of asthma (21). Molecularly, chromatin immunoprecipitation (ChIP) sequencing analysis combined with proximity ligation assays have identified that, in addition to BATF-IRF-4 complex, IRF-8 interact and form large transcriptional complexes with IRF-4/BATF/PU.1 to induce the development of Th9 cells (26). Interestingly, it has been proposed that IRF-8 executes dual functions in Th9 cells differentiation, while on the one hand, it partners with other transcription factors to form large transcriptional complex to optimally induce IL-9, on the other hand, IRF8/Etv6 heterodimer represses *Il4* transcription. Taken together, these observations clearly indicate the dual functions of IRF-8 in promoting Th9-exclusive gene signature. In addition to IRF-4 and IRF-8, the involvement of IRF-1 and its functions in Th9 differentiation remains unclear, as two independent studies have reported contrary functions of IRF1 in Th9 development (24, 25). While Végran et al. have shown that IRF1-deficient CD4<sup>+</sup> T cells have profound defect in Th9 differentiation, on contrary to this Campos Carrascosa et al. have identified that IFN-γ-induced IRF1 promotes transcriptional shift of Th9 cells to Th1 cells, as IRF1 outcompetes IRF4 binding at *Il9* promoter. The discrepancy in these two studies could be accounted to the cytokines used for inducing Th9 differentiation. Végran et al. used IL-1β together with TGF-β plus IL-4 to induce Th9 differentiation, which leads to the induction of IRF-1. Campos Carrascosa et al demonstrated that IFN-γ/IRF1 axis negatively regulates the differentiation of Th9 cells induced by TGF-β plus IL-4 (24, 25).

In addition to the cytokines induced transcription factors, TCR-stimulated activation of NFAT and NF-κB contributes rapid induction of IL-9 in Th9 cells. Both NFAT1 and NF-κB work together synergistically in Th9 differentiation. NFAT1 facilitates the binding of NF-κB p65 to *Il9* promoter by actively remodeling chromatin, as T cells-deficiency of NFATc1/NFATC2 produces attenuated IL-9 in mouse model of allergic inflammation (27). Two major components of NF-κB pathway, i.e., RelB-p52 and p50 are induced upon OX-40 and GITR ligation, respectively, in Th9 cells (28–30). In addition to TCR-mediated activation of transcription factors, ligation of secondary co-stimulatory checkpoint molecules on Th9 cell enhance the development of Th9 cells by further promoting the activation of transcriptional network that support Th9 differentiation. While OX-40, a member of TNFR superfamily of receptor, induced IL-9 is dependent on STAT6, GITR induces the activation of STAT6, BATF, PU.1, and IRF-4 in Th9 cells (28–30). Interestingly, GITR ligation enhances IL-9 expression in Th9 cells in the absence of IL-4 receptor signaling but not in STAT6 deficient mice, as induction of STAT6 under GITR stimulation is IL-4–IL-4R is independent and is required for Foxp3 repression (29). Surprisingly, other TNFRSF member, TLA1/DR3 requires functional IL-2/STAT5 pathway but is independent of NF-κB or STAT6 in Th9 cells (34).

Engagement of TGF-β with its receptor activates Smaddependent and -independent pathways that leads to Th9 differentiation. It has been shown that T cells lacking Smad2, Smad3, or Smad4 (Smad2fl/flCD4cre, Smad3−/− T cells and Smad4fl/fl CD4cre animals) have reduced IL-9 production in Th9 cells (20). Mechanistically, Smad-deficiency leads to modifications in histone acetylation/deacetylation and methylation at *Il9* promoter or CNS regions, suggesting that Smads might be essential for favorable epigenetic modifications of *Il9* locus in Th9 cells (20). In fact, as compared to single gene deficiency of either Smad2 or Samd3, double deficiency of Smad2/Smad3 leads to profound reduction in IL-9 production in Th9 cells, which found to be associated with reduced histone acetylation marks in *Il9* locus (20). In addition, it has been found that Smad3 bind at a site near to recombinationsignal-binding protein for immunoglobulin-κ-J region (RBP-Jκ) and the Notch intracellular domain in IL-9 locus to further positively regulate the differentiation of Th9 cells (17). In addition to Smad dependent pathway, TGF-β1-induced Smad-independent pathways are primarily coordinated *via* TAK1, as inhibition of TAK1 suppresses two major transcriptional repressors, Id3 and SIRT1, in Th9 cells developmental pathway (18, 19). Id3, an E-box transcription factor inhibitor, act as a negative regulator in Th9 differentiation. Molecularly, Id3-deficient T cells have shown an increased binding of E2A and GATA-3 at *Il9* promoter, suggesting that the absence of Id3 promotes accessibility of *Il9* locus to Th9-associated transcription factors leading to Th9 differentiation. Similar to Id3, SIRT1-deficient T cells have increased IL-9 production due to enhanced HIF-1α-dependent glycolysis in Th9 cells (19).

In addition to factors induced by TGF-β1, IL-4, and TCR, IL-2–IL-2 receptor pathway plays a critical role in enhancing IL-9 production and Th9 development (11). Upon binding to its receptor, IL-2 induces STAT5 activation that lead to differentiation of Th9 cells. Moreover, nitric oxide, Itk activation, TSLP, and TL1A enhanced IL-9 induction in Th9 cells is dependent on IL-2 (34–36). Although activated STAT5 binds directly to *Il9* promoter to induce IL-9 gene expression, however, the functions of STAT5 in Th9 cells are suppressed by BCL-6. Mechanistically, BCL6 competes with STAT5 for binding at the *Il9* promoter in Th9 cells, thus suppresses the development of Th9 cells (11). In addition to STAT5, IL-1β-induced STAT1 was also found to amplify IL-21 and IL-9 production *via* STAT-1/IRF-1 module to enhance antitumor functions of Th9 cells (24).

### PHYSIOLOGICAL IMPORTANCE OF Th9 CELLS

Physiologically, Th9 cells were shown to play crucial roles in aggravating inflammation in disease like asthma, EAE, colitis, and skin inflammation (37, 38). IL-9 and IL-9R single nucleotide polymorphisms (SNPs) have been associated with allergen sensitization in allergy. Other Th9-associated genes such as IL-4RA, STAT6, and IL-33 are also found to be associated with allergic inflammation in human diseases (5, 39–41). Consistently, administration of anti-IL-9 neutralizing antibody in murine models of asthma decreased the severity of disease associated with attenuated infiltration of eosinophils and AHR, suggesting a crucial role of IL-9 in progression of allergic inflammation in asthma (15, 21, 22). Nevertheless, humanized anti-IL-9 neutralizing antibody (MEDI-528) clinical trial could not report any improvement in subjects as compared to control treatment (42). This could be attributed to the polygenic nature of asthma and genetic basis of heterogeneity in different individuals. Since asthma is contributed by waves of different cytokines such as IL-4, IL-5, IL-13, and IL-9, hence it is needful to define and identify the predominant cytokine signature and subtype that is inducing the disease.

IL-9 and Th9 cells were found to be associated with skin inflammation. Human IL-9-producing Th9 cells were found to be skin tropic and express cutaneous lymphocyte antigen (CLA), which by virtue makes them skin tropic (38). Further analysis identified that skin tropic CLA<sup>+</sup> human Th9 cells were found to be independent of TGF-β1 and IL-2, and were accumulated in psoriatic lesions (38). It is, however, intriguing as to what is the functions of Th9 cells in skin under homeostatic conditions. Whether Th9 cells are required for the maintenance of barrier functions in the skin surfaces are not yet elucidated. Based on the observations that IL-9 can acutely stimulate mast cells, the constant presence of Th9 cells under the skin might potentially activate innate immune cells including mast cells upon skin infection to contain pathogens. In fact, IL-9 is shown to induce IL-8 production from keratinocytes, which promote the influx of neutrophils at the site of fungal infections (43, 44). Emerging literature is suggesting that IL-9 potentially contributes to different types of skin disorders such as atopic dermatitis, allergic contact dermatitis, allergen-induced delayed type hypersensitivity, psoriasis, and cutaneous T cell lymphoma (45).

In the gut inflammation in inflammatory bowel diseases (IBD), IL-9-producing CD4+ T cells were found to be colitogenic, as gut epithelial cells of ulcerative colitis patients expressed elevated levels of IL-9R. Moreover, lamina propria T cells from IBD patients were found to have an increased frequency of CD4<sup>+</sup>PU1<sup>+</sup>IL-9<sup>+</sup> and CD4<sup>+</sup>IRF4<sup>+</sup>IL-9<sup>+</sup> T cells, suggesting a strong association of IL-9 with disease severity in IBD (33, 46). In mouse models of colitis, adoptive transfer of *in vitro* differentiated Th9 cells into Rag-deficient hosts led to the development of severe colitis in an IL-9 dependent manner (33, 47, 48). In chemically induced model of colitis, Th9 cells were found to be one of the major effector T cells population that induced disease pathogenesis, as both PU.1- and IL-9 deficient mice were found to have reduced incidence of colitis and reduced inflammatory score as compared to wild-type mice (33). Consistently, treatment with anti-IL-9 neutralizing antibody was found to effectively control tissue inflammation of intestine in colitis (33). Mechanistically, IL-9 was found to suppress epithelial cell proliferation thereby affecting mucosal wound healing in IBD. In fact, topical administration of IL-9 was found to repress epithelial cells tissue repair mechanism *in vivo* (33). Nonetheless, the protective role of IL-9 was also attributed in DSS-induced colitis, as NKT cells-driven IL-9 is found to protect gut inflammation in DSS-induced colitis (46).

In addition to the intestinal inflammation, Th9 cells were found to be associated with tissue inflammation in EAE, a mouse model of human multiple sclerosis. IL-9<sup>+</sup> T cells can be isolated from the draining lymph nodes of mice that develop EAE. In fact, similar to Th1 and Th17 cells, adoptive transfer of MOG-specific Th9 cells into Rag-deficient mice induced the development of EAE (49, 50). Nonetheless, CNS lesions induced in Th9 transfer model were quite different in their appearances as compared to Th1 and Th17 cells transfers (51). In addition to inducing tissue inflammation in autoimmune diseases, IL-9 plays a pivotal role in providing immunity against helminth infections by expelling worms *via* enhanced intestinal muscle contraction, mucus production, and increased mast cell activity (52). Th9 cells provide immunity against *Nippostrongylus brasiliensis*, as adoptive transfer of Th9, but not Th2, cells into Rag-deficient hosts provided long-lasting immunity against worms (53, 54). Similarly, animals expressing dominant negative form of TGFβRII in CD4<sup>+</sup> T cells were found to have reduced levels of IL-9 associated with enhanced parasitic burden in *T. muris* infection (10).

In addition to inducing immunity against extracellular pathogens as well as tissue inflammation in organ-specific autoimmunity, Th9 cells were also found to be associated in mounting superior antitumor response as compared to Th1 and Th17 cells (24). In fact, *Il9* SNP was found to be linked with enhanced risk of cutaneous malignant melanoma (55). In a tumor microenvironment, Th9 cells were found to produce CCL20, which facilitate the migration of CCR6<sup>+</sup> leukocytes in the tumor tissue. Moreover, IL-3 and IL-21 produced by Th9 cells, respectively, promote dendritic cells survival and functions as well as CD8<sup>+</sup> CTLs (56). Adoptive transfer of antigen-specific Th9 cells in B16-F10 melanoma models reduces tumor burden and severity, and this antitumor effect of Th9 cells is found to be dependent on IL-9, as neutralization of IL-9 suppressed the antitumor functions of Th9 cells (24). On an intriguing note, adoptive transfer of IL-1β preconditioned Th9 cells retained their antitumor functions in IL-9R-deficient mice or upon IL-9 neutralization in wild-type mice *in vivo* (24). Although multiple studies have demonstrated the potent antitumor functions of Th9 cells, however, the protective effect of Th9 cells in tumor was found to be restricted to solid tumors such as melanoma and lung-adenocarcinoma (57, 58).

Although Th9 cells were known to be involved in multiple diseases, the *in vivo* differentiation and development of Th9 cells is not well defined. Most of the studies have defined the early events in Th9 cells, suggesting a caveat in understanding of the precise genetic programming involved in stably developed Th9 cells. Recently, it has been shown that *in vitro* differentiated Th9 cells lose their ability to secrete IL-9 with chronic stimulation, suggesting that Th9 cells transiently produce IL-9 during *in vitro* differentiation (59). The loss of IL-9 production in Th9 cells could also be explained in the context of Th plasticity, as Th9-Th1 plasticity was suggested (51). Adoptively transferred MOG-specific Th9 cells were found to be converted into IFN-γproducing cells at sites of target tissues (51). Similarly, Th9 cells produce copious amounts of IFN-γ in B16F10 melanoma model. These observations clearly suggest that Th9 to Th1 plasticity may be crucial for inducing effector functions in these disease models (51). However, none of the studies molecularly defined the Th9 plasticity in greater details.

Although IL-9 is predominantly produced by Th9 cells, however, the production of IL-9 is not restricted to Th9 only. Other effector and regulatory T cells such as Th2, Th17, Tfh, and Tregs cells also produce IL-9 (3, 4, 60–64). It is possible that, in general, during TCR-dependent stimulation, epigenetic modifications keep *Il9* locus accessible for regulators for a narrow time frame. It can also be suggested that a common transcriptional signature might be shared by all these subsets leading to expression of *Il9* by T cells. The transcriptional profiling of Th9 cells suggests the involvement of multiple transcription factors, however, Th9 cells is still devoid of its master regulator as compared to other Th subsets (20, 23, 62). Nonetheless, the presence of IL-9<sup>+</sup> T cells in patients and target tissues suggests their importance and development *in vivo*.

### Table 1 | Role of Foxo1 in effector and regulatory T cells.


## PHOSPHOINOSITIDE 3-KINASE (PI3K) SIGNALING AND ROLE OF Foxo1 TRANSCRIPTION FACTOR IN CD4**+** T CELLS

Phosphoinositide 3-kinase signaling plays crucial role in integrating diverse biological functions ranging from cell survival, metabolism to tolerance and aging. In response to growth factor stimulation, PI3K activation transmit cellular signals and further activate Akt and mTOR signaling pathway, which together contributes to different biological processes such as tumor survival including angiogenesis and recruitment of inflammatory cells (65–67). Though there are four different classes of PI3K, but class IA and class IB are elucidated in detail in activation and functions of T cells (67). The major function of these kinases is to activate PLCγ to generate PI3K effector molecules of cell signaling such as diacyl glycerol and inositol-1, 4, 5-triphosphate (IP3), which induces calcium mobilization, and thus leading to PKC activation and NF-κB nuclear translocation. PI3K also phosphorylate PtdIns (4,5) P2 to generate PtdIns (3,4,5) P3, which in turn activates downstream kinases specifically with PH domains such as Akt. Upon T cell activation, Akt, a serine threonine kinase, gets phosphorylated at Thr308 and Ser473 by PDK1 and mTORC2, respectively, to attain its complete activation (65–67). Although PI3K axis has been shown to be essential for clonal expansion and differentiation of Th1 and Th17 cells, however, PI3K negatively regulates regulatory T cells development and function, as TGFβ1-induced Foxp3 expression is impaired upon constitutive AKT activation (68–70). Transcription factor Foxo1 is one of the major downstream targets of Akt activation and regulates cell cycle, cell survival, and energy generation. Foxo transcription factors were shown to respond to variety of physiological stimuli and physiological conditions including oxidative stress, mitogenic factors, and inflammation (71). In lymphocyte compartment, Foxo factors regulate T cell homing and homeostasis, formation of memory T cells, and process of T cell differentiation (72, 73). Transcriptional activity of Foxo factors is regulated by various posttranslational modifications which are collectively known as Foxo code (71). In a simplified context, upon growth factor stimulation, Foxo gets localized to cytoplasm subsequent to phosphorylation at three conserved residues (T24, S256, and S319) by Akt or serum glucocorticoid kinase-1 (SGK1) kinase and subsequently leads to nuclear export and degradation (71).

Due to their dominant role in controlling T cell survival, migration, and metabolism, Foxo1 is studied in most of the effector and Tregs as well as in CD8<sup>+</sup> T cells (**Table 1**). Within CD8+ T cells, Foxo1 suppresses IL-12-dependent T-bet expression while promotes memory CD8+ T cells phenotype by inducing expression of Eomes. Although Foxo1 have binding sites in T-bet promoter, it does not bind to T-bet promoter and exerts its functions in DNA-binding independent manner (74, 75). Foxo1 deficient CD8<sup>+</sup> T cells expand normally and form effectors, but failed to make pool of memory cells (74–76). Inhibition of glycolysis in CD8<sup>+</sup> T cells facilitates nuclear localization of Foxo1 and enhanced expression of Foxo1 target genes such as *Klf2*, *Cd62l* (L-selectin), *Ccr7* (chemokine receptor), and *S1p1r* (sphingosine-1-phosphate receptor 1) (77). It was shown that PI3K/Akt/Foxo1 axis is considered to be a major pathway involved in Tregs to Th1 reprogramming (68). Metabolic regulator PPARγ is also known to stabilize Foxo1 functions, as PPAR-γ-deficient CD4<sup>+</sup> T cells were found to produce enhanced levels of pro-inflammatory cytokines such as IFN-γ and IL-17 due to inhibition of Foxo1 functions (78). In addition to Th1 and Th17 cells, Foxo1 has been shown to negatively regulate the development of T follicular (Tfh) cells, which are marked by the expression of *Bcl6* and CXCR5 (79).

Foxo1 is a potent suppressor of both human and mice Th17 differentiation (69, 80–82). γc cytokines such as IL-2, IL-7, or IL-15 are known to drive IL-17 and IL-22 expression in CCR6<sup>+</sup> human memory T cells as compared to CCR6<sup>−</sup> Tm cells. Furthermore, γ<sup>c</sup> cytokines activate PI3K axis and represses Foxo1 thereby promoting human Th17 differentiation (69). Consistently, ectopic expression of Foxo1 suppresses γc-mediated IL-17/IL-22 expression in CCR6<sup>+</sup> Tm cells (69). In addition to PI3K, Foxo1 is also regulated by SGK1 *via* regulating IL-23–IL-23R signaling, which is essential for stabilizing and acquiring the pathogenic functions of Th17 cells (80). IL-23R–SGK1 axis has been shown to suppress Foxo1 transcriptional activity by inducing its phosphorylation. In addition, Foxo1 is not only a potent inhibitor of Rorγt-mediated transactivation of *Il23r* expression but also binds directly to RORγt *via* DNA-binding domain (DBD) thereby suppressing Rorγt dependent transcriptional program of Th17 cells (81). In fact, Foxo1 T cell conditional deficient mice have shown increased numbers of Th17 cells in thymus and periphery as compared to wild-type mice (81). Furthermore, in a mixed bone marrow chimera experiment, Foxo1 deficiency is sufficient to drive Th17 differentiation *in vivo* upon antigen challenge as compared to wild-type cells (81). It has been demonstrated that dicer-regulated microRNA-183-96-182 (mir-183C) regulates pathogenic Th17 differentiation *via* suppressing Foxo1 (82). Hence, proposing that factors which are positively associated with Foxo1 are the negative regulators of Th17 differentiation.

As compared to its role in effector CD4+ T cells, Foxo1 functions are well established in induction and functions of Tregs (83–86). Foxo1-deficient mice were shown to have expanded population of CD4+CD44hi T cells as well as hyper B cell activation leads to hyper gammaglobulinemia and expansion and increased number of follicular T cells. Although the frequency of thymic Foxp3<sup>+</sup> Tregs was found to be decreased in mice that harbor Foxo1 conditional deficiency in Foxp3+ Tregs, the frequency of Foxp3+ Tregs in the periphery remain normal in these mice. Nonetheless, Foxo1 deficiency in Tregs leads to lose their suppressive ability (83, 84). In addition to Foxo1, Foxo3 is also expressed in immune cells and Foxo3<sup>−</sup>/<sup>−</sup> mice do not develop spontaneous autoimmunity or purified T cells have no defect in proliferation or survival. Combined deletion of both Foxo1 and Foxo3 in T cells leads to fatal systemic inflammatory disease due to defective Foxp3 Tregs development (85), suggesting a specific involvement of Foxo1 in regulating Foxp3 Tregs functions and development. Furthermore, Foxo1 facilitate the binding of other transcription factors at Foxp3 locus thereby regulating complete Foxo1-mediated transcriptional gene program (83, 85, 87). Overexpression of Akt, a negative regulator of Foxo1, has been shown to inhibit suppressive function of Tregs (68, 84). Functionally, Tregs are also classified as resting Tregs, found in spleen and lymph nodes, and activated Tregs found in lymphoid organs and non-lymphoid tissues. Recently it has been demonstrated that Foxo1 repression is associated with enhanced migration of activated Tregs to tumor sites while Foxo1 gain of function leads to quick depletion of activated Tregs, resulting in effective tumor immunity (86). Taken together, Foxo1 have a discrete role in effector and regulatory T cells development and functions.

# Foxo1 REGULATES IL-9 PRODUCTION AND DEVELOPMENT OF Th9 CELLS

Th9 cells are now unraveled as a separate subset of effector CD4<sup>+</sup> T cells and one of the dominant producers of IL-9 (8, 9). Emerging literature has suggested the involvement of various transcription factors in the development of Th9 cells, as Th9 cells emerges as an effector Th population involved in the pathogenesis of many diseases like allergy, asthma, IBD, and antitumor immunity (20, 23, 62). Although previous reports have demonstrated the involvement of IRF-4, PU.1, BATF, and IRF-1 in induction and functions of Th9 cells, none of these factors determine the lineage-specificity Th9 cells. In fact, most of these transcription factors (IRF-4, PU.1, BATF, and IRF-1) are also shared by other Th cells—for example, IRF-4 and PU.1 expressed in Th2 cells, IRF-4 and BATF expressed in Th17 cells, and IRF-1 expressed in Th1 and Tr1 cells (88–93). We and others have recently characterized that Foxo1 is essential for Th9 and other IL-9-producing T cells (16, 62, 94). Though Foxo1 is also co-expressed or shared by regulatory T cells, nonetheless Foxo1 generally suppresses other effector lineages such as T-bet/ Th1 cells and Rorγt/Th17 cells. Unlike other transcription factors of Th9 cells (IRF-4, BATF) which also promote the production of IL-4, IL-17 by T cells, Foxo1 negatively regulates the production of IFN-γ and IL-17 with no observable effect on IL-4 induction. This suggests that Foxo1 may not be a unique transcription factor of Th9 cells but does impart specificity in promoting IL-9 induction in Th9 cells as well as in other T cells (75, 77, 78, 80, 81, 93).

As discussed above that PI3K/Akt axis generally promotes the effector function of CD4<sup>+</sup> T cells while suppresses regulatory T cell development, however, its role in IL-9 induction is lately elucidated. Inhibition of upstream PI3K/AKT by pan-PI3K inhibitor (LY294002) pathway enhanced the induction of IL-9 in Th9 cells. Since Foxo1 tends to be one of the major downstream cellular targets of PI3K/Akt axis, therefore inhibition of PI3K axis results in reduced levels of phospo-Foxo1, mark of inactive cytosolic form of Foxo1 (62). Furthermore, inhibition of Foxo1 reversed the effects of PI3K/AKT inhibition on Th9 cells, suggesting the previously uncharacterized involvement of PI (3)K/AKT-Foxo1 axis in inducing the development of Th9 cells (62). In fact, time kinetics of Foxo1 expression in Th9 cells suggests that it is induced early starting from 24 h of differentiation and is maintained till 72 h (16). Akin to Foxo1, mTOR is also targeted by PI3K/Akt. Suppression of mTOR by rapamycin or stimulation of mTOR by mTOR activator MHY1485 substantially inhibited or promoted Th9 differentiation, respectively (16). Similar to overexpression of constitutive form of Akt, Th9 differentiation was found to be compromised in Foxo1-deficient CD4<sup>+</sup> T cells or upon direct inhibition of Foxo1 by chemical or genetic approaches (16, 62, 94). In addition of PI3K/Akt, TGF-β1-induced Smad3 pathway is implicated to be involved in Foxo1 induction in Th9 cells (94).

Similar to IL-1β-induced IRF-1 expression in Th9 cells, Foxo1 is seen to be induced upon IL-7 stimulation during Th9 differentiation (16, 24). Priming of naïve CD4<sup>+</sup> T cells with IL-7 not only enhances *Il9* and *Il21* expression but is also essentially required for repressing the Th9 repressor-Foxp1 (16). During Th9 differentiation, both Foxo1 and Foxp1 are competitively regulated as both these Fox members have similar conserved binding site on IL-9 promoter. IL-7 priming of Th9 cells induces p300 which is both a co-activator and stabilizer of Foxo1 protein (16). This not only enables Foxo1 to outcompete Foxp1 for binding to IL-9 promoter but also leads to decreased amounts of phosphorylated Foxo1 protein in CD4<sup>+</sup> T cells thereby increasing the relative amounts of total Foxo1 protein (**Figure 1**). Due to redundancy of Foxo proteins, in addition of Foxo1, Foxo4 was also found to augment Th9 differentiation and provides a link to the observation as why noticeable Th9 differentiation is still observed in Foxo1 conditional deficient CD4<sup>+</sup> T cells (16).

While being in nucleus, Foxo1 not only targets the cytokine gene locus in order to directly regulate the lineage specific cytokines but also auto-amplifies itself, as Foxo1 can bind to its own promoter (95). Since Foxo transcription factors have both DBD and protein interaction domain, therefore they mediate their function in both DNA dependent and independent manner (71). Foxo1 physically interact with IL-9 and IRF-4 promoters in Th9 other IL-9-producers such as Th17 and iTregs (62). Moreover, DNA-binding activity of Foxo1 is required for IL-9 induction in Th9 cells, as ectopic expression of Foxo1 that lacks DBD fails to enhance IL-9 in Th9 cells. Interestingly, both overexpression and inhibition of Foxo1 in Th9 cells, respectively, enhances or suppresses other Th9-associated genes such as IRF-4, PU.1, BATF, and IRF1, indicating a potential involvement of Foxo1 partner complexes in regulating IL-9 induction. In fact, in an *in vitro* protein interaction assay Foxo1 was found to interact with IRF-4, suggesting that Foxo1–IRF4 protein complex might be essential for the induction of Th9 cells (62). Since modulation of Foxo1 also resulted in changes in Th9 genetic program, hence it could be hypothesized that Foxo1 is induced early during Th9 differentiation and may act upstream of other factors, which are crucial for development of Th9 cells. Corroborating this, Foxo1 directly induces and acts upstream of IRF4 in Th9 cells and thereby potentiating the development of Th9 cells (62, 94). It could be suggested that similar to IRF-8, which participate in formation of transcriptional complexes, Foxo1 might also work collectively by recruiting other transcription factors and form large transcriptional complex, which drive optimal *Il9* transcription in Th9 cells (26). In nutshell, above mentioned observations

Th9 cells. Foxp1 is enriched and bound to interleukin (IL)-9 promoter in naive CD4+ T cells. Activation of naive T cells in the presence of Th9 polarizing cytokines induces Foxo1-co-activator p300 which recruits Foxo1 from cytoplasm to nucleus leading to displacement of Foxp1 from Foxo1 binding sites as both these factors bind to the same region in IL-9 locus, and therefore induces the development of Th9 cells.

suggested a crucial role of Foxo1 in inducing IL-9 during *in vitro* differentiation of Th9 cells.

As discussed above, IL-9 plays an indispensable role in inducing and promoting atopic diseases such as dermatitis and allergic asthma. Both IL-9 and IL-9R polymorphism are found to be genetically associated with asthma. Consistently, circulating T cells from allergic patients have enhanced capacity to produce IL-9 in response to pollen or house dust mite extract (5, 37). As reported by fate reporter system, Th9 cells are one of the major IL-9 producers in OVA sensitized mice models. In fact, administration of anti-IL-9 neutralizing antibody after allergen sensitization reduced allergic inflammation in murine model of asthma associated with attenuation in inflammatory cell infiltration, indicating the crucial role of IL-9 in development of asthma. Since our *in vitro* data suggests a strong association of IL-9 and Foxo1, hence we speculated if Foxo1 also regulates IL-9 *in vivo* during the pathogenesis of asthma. Corroborating *in vitro* data, *in vivo* blockade of Foxo1 resulted in reduced signs of asthma as measured by AHR and associated with reduced frequency of IL-9<sup>+</sup>CD4<sup>+</sup>T cells in the lungs (62, 94). In a B16-OVA tumor model, IL-7-treated OT-II Th9 cells were found to mount potent antitumor activity, which is suppressed in the absence of Foxo1, suggesting that Foxo1-mediated IL-9 induction is essential in mounting the potent antitumor immunity by Th9 cells (16).

Since Foxo1 inhibition in asthma model reduced IL-9 dependent inflammation and antitumor potential of Th9 cells, hence targeting Foxo1 could provide a potential therapeutic advantage in these diseases. Nonetheless, due to lack of firm understanding of the time window at which IL-9 appears *in vivo,* it is difficult to extrapolate when and how Foxo1 controls IL-9 appearance *in vivo*.

### Foxo1 TUNES IL-9 INDUCTION BY Th17 CELLS AND REGULATORY T CELLS

Interleukin-9 is a pleotropic cytokine and its expression is not confined to one particular T cell subset. IL-9 is shown to be produced by Th2, Th9, Th17 cells, and Tregs. Interestingly, IL-9 is produced by Th17 cells induced by TGF-β1/IL-6. Further detailed analysis of pathogenic and non-pathogenic Th17 cells revealed that IL-9 production is restricted to non-pathogenic Th17 cells (polarized with TGF-β1/IL-6), as pathogenic Th17 cells induced by IL-23 exposure lose IL-9 production (63, 80). In addition to IL-23, pathogenic Th17 cells are also induced by TGF-β3 and IL-1β combined separately with IL-6. Interestingly, TGF-β3 and IL-6 polarized Th17 cells express less *Il9* as compared to TGF-β1 and IL-6 (62). Cogently, pathogenic Th17 cells maintain high levels of phospho-Foxo1 during their differentiation suggesting the sequestration and degradation of Foxo1 in cytoplasm. Though cytosolic Foxo1 is suggested to be marked for degradation, however, cytosolic Foxo1 regulates autophagy and is not truly nonfunctional in cytoplasm (96). In addition of PI3K/Akt, Foxo1 is also negatively regulated by salt sensing kinase SGK1. Strikingly, IL-9 is one of the highly expressed genes in SGK-deficient Th17 cells as compared to wild-type Th17 cells. Interestingly, SGK1 also promotes the generation of pathogenic Th17 cells thus endorsing cells.

the fact that non-pathogenic Th17 cells not only express *Il9* but also *Foxo1* while acquisition of pathogenicity by Th17 cells leads to concomitant disappearance of both *Il9* and *Foxo1* (62).

Since Foxo1 is generally controlled by PI3K/Akt axis, hence we employed a reverse approach where inhibition of PI3K by paninhibitor or overexpression of dominant form of Akt not only suppressed IL-17 and Th17 genetic program but also enhanced the induction of IL-9 in Th17 cells. Similarly, inhibition of Foxo1 in Th17 cells also restrains the production of IL-9 (**Figure 2**). While suppression of IL-17 by Foxo1 is not surprising, as Foxo1 is previously known to directly inhibit RORγt (81). However, what is truly compelling is the ability of Foxo1 to reciprocally regulate the balance of IL-9 and IL-17 in Th17 cells (62). Though it is not known what drives the expression of IL-9 in Th17 cells, physiological need to secrete IL-9 or if IL-9 production by Th17 cells is a transient phenomenon. Nonetheless, current study suggests that Foxo1 can discern *Il9* expression over *Il17* in Th17 cells (62).

Strikingly, not only effector T cells but regulatory T cells also produce IL-9. While molecular mechanisms driving the induction of IL-9 in Tregs is not known, however, gene expression analysis have shown the transcriptional similarities between Th9 and Tregs due to the presence of shared differentiation factor TGFβ1 in culture conditions (21). Functionally, IL-9 production by regulatory T cells is seen to mediate protective immunity against nephritis in a model of nephrotoxic serum nephritis (NTS) and skin allograft tolerance *via* Treg–mast cell cooperation in the target organ during inflammation (64). Moreover, IL-9R-deficient mice immunized with MOG develops severe EAE and Tregs isolated from IL-9R-deficient animals were shown to have poor suppressive function, suggesting that IL-9 regulates Tregs suppressive functions in autoimmunity (60). Since we have found an association of Foxo1 with IL-9 in effector T cells, therefore we speculated whether Foxo1 is also required for IL-9 induction in Tregs. We have reported that TGF-β1-induced iTregs not only express *Il9* but also other Th9-associated factors such as *Batf*, *Irf4* and *Spi1* and *Klf2*, a Foxo1 target gene, and inhibition of Foxo1 in iTregs leads to IL-9 suppression. Interestingly, Foxo1 physically binds to IL-9P in iTregs thereby regulating *Il9* transcription directly. The ChIP binding data also suggests that Foxo1 has accessibility to IL-9P during TGF-β1 permissive milieu, however, if Foxo1 is involved in directly modulating chromatin dynamics at IL-9P is not known. Though this preliminary data only suggests that Foxo1 plays a functional role in IL-9 induction by iTregs. Furthermore, in order to prove an essential requirement of Foxo1 for IL-9 induction in Tregs, experiments employing Tregs isolated from Foxp3creFoxofl/fl conditional deficient mice or Foxp3creFoxo1AAA (Akt-mediated phosphorylation site is mutated) should be implemented. Since Tregs express other Th9-associated factor such as IRF-4, hence we do not exclude the possibility of involvement of other transcription factors in regulating IL-9 induction by Tregs. The production of IL-9 by regulatory T cells and IL-10 producing Th17 cells suggests that IL-9 might play a significant role in regulating inflammation *via* immune-suppression. In a striking contrast, murine Th9 cells also produce IL-10 but lacks immune-suppressive capability thereby enforcing the idea of a subset specific role of IL-9 in inducing or ameliorating tissue inflammation.

### CONCLUSION AND PERSPECTIVE

Over the last one decade since the discovery of Th9 cells in 2008, an extensive array of signaling axis, transcription factors, and physiological functions that are involved in development and amplification of IL-9<sup>+</sup> T cells have been unraveled. Very recently, growth factor-dependent PI3K/Akt axis *via* Foxo1 is seen to control the development of IL-9 production by effector T cells. Strengthening the role of Foxo1, three independent groups have summarized the essential requirement of Foxo1 in regulating IL-9 in Th9 cells. It is interesting to note that in general Foxo1 negatively regulates other effector subsets but is seen to be a positive regulator of IL-9 expression. Physiologically, Foxo1-dependent functions are critically required for inducing allergy asthma and antitumor immunity in mouse models. Due to our limited understanding for pathways involved in development of human Th9 cells, ambiguity and overlap of current mechanisms required for Th9 transcriptional regulation with other T helper subsets, the therapeutic exploitation of Th9 cells for targeted therapy against various diseases is still not achieved. Furthermore, a lot more efforts need to be invested in understanding the requirements for maintaining stability of Th9 cells *in vivo*, which will provide a better template for manipulating Th9 cells therapeutically.

### AUTHOR CONTRIBUTIONS

AA and SM have written and edited the review.

# FUNDING

This work was supported by Wellcome Trust/DBT India Alliance Intermediate fellowship (IA/I/12/1/500524), Department of Biotechnology, Government of India and Core grant of Translational Health Science and Technology Institute. SM was supported by a PhD fellowship from Council of Scientific and Industrial Research (CSIR).

# REFERENCES


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

The reviewer PP and handling Editor declared their shared affiliation.

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

*Anke Fähnrich1†, Sebastian Klein1†, Arnauld Sergé2 , Christin Nyhoegen3 , Sabrina Kombrink <sup>3</sup> , Steffen Möller <sup>4</sup> , Karsten Keller <sup>3</sup> , Jürgen Westermann1 and Kathrin Kalies <sup>1</sup> \**

*<sup>1</sup> Institute of Anatomy, University of Luebeck, Luebeck, Germany, 2Centre de Recherche en Cancérologie de Marseille (CRCM) U1068 INSERM – UMR7258 CNRS – Institut Paoli Calmette, Aix-Marseille University, UM105, Marseille, France, <sup>3</sup> Institute of Mathematics, University of Luebeck, Luebeck, Germany, 4 Institute for Biostatistics and Informatics in Medicine and Ageing Research, Rostock, Germany*

### *Edited by:*

*Amit Awasthi, Translational Health Science and Technology Institute, India*

### *Reviewed by:*

*Paul E. Love, National Institutes of Health (NIH), United States Christoph Wülfing, University of Bristol, United Kingdom*

*\*Correspondence: Kathrin Kalies kalies@anat.uni-luebeck.de These authors have contributed* 

*†*

### *Specialty section:*

*equally to this work.*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 14 February 2018 Accepted: 24 April 2018 Published: 17 May 2018*

### *Citation:*

*Fähnrich A, Klein S, Sergé A, Nyhoegen C, Kombrink S, Möller S, Keller K, Westermann J and Kalies K (2018) CD154 Costimulation Shifts the Local T-Cell Receptor Repertoire Not Only During Thymic Selection but Also During Peripheral T-Dependent Humoral Immune Responses. Front. Immunol. 9:1019. doi: 10.3389/fimmu.2018.01019*

CD154 is a transmembrane cytokine expressed transiently on activated CD4 T cells upon T-cell receptor (TCR) stimulation that interacts with CD40 on antigen-presenting cells. The signaling via CD154:CD40 is essential for B-cell maturation and germinal center formation and also for the final differentiation of CD4 T cells during T-dependent humoral immune responses. Recent data demonstrate that CD154 is critically involved in the selection of T-cell clones during the negative selection process in the thymus. Whether CD154 signaling influences the TCR repertoire during peripheral T-dependent humoral immune responses has not yet been elucidated. To find out, we used CD154 deficient mice and assessed the global TCRβ repertoire in T-cell zones (TCZ) of spleens by high-throughput sequencing after induction of a Th2 response to the multiepitopic antigen sheep red blood cells. Qualitative and quantitative comparison of the splenic TCZ-specific TCRβ repertoires revealed that CD154 deficiency shifts the distribution of Vβ-Jβ genes after antigen exposure. This data led to the conclusion that costimulation via CD154:CD40 during the interaction of T cells with CD40-matured B cells contributes to the recruitment of T-cell clones into the immune response and thereby shapes the peripheral TCR repertoire.

Keywords: CD154 costimulation, T-cell repertoire, humoral immune response, sheep red blood cells, spleen, T:B-cell interaction

# INTRODUCTION

The T-cell receptor (TCR) repertoire is shaped during negative, positive, and agonist selection in the thymus and by inter-clonal and intra-clonal competition in the periphery in adults after thymic involution. The latter is mainly triggered by homeostatic proliferation of naive T cells and by expansion of individual T-cell clonotypes after antigen exposure (1, 2). Clonal expansion and the resulting numbers

**Abbreviations:** BCZ, B-cell zones; CDR3, complementary determining region 3; GC, germinal centers; SRBCs, sheep red blood cells; TCR, T-cell receptor; TCZ, T-cell zones.

of progeny depend on the strength of signals transmitted via the TCR upon ligation to its cognate peptide-MHC ligand. Thereby T cells with stronger TCR signaling generate bigger burst sizes (3–5). In addition to TCR signaling, the activation and lineage decision of CD4 T cells is regulated by costimulatory pathways. In particular, costimulation via CD154:CD40 that takes place during the interaction of antigen-activated T and B cells is critical for the differentiation of CD4 T cells into cytokine-producing effector T cells (6–8). The assumption that CD154 costimulation contributes to TCR signaling intensity leads to the hypothesis that it could provoke the enrichment or diminishment of individual T-cell clones during T-dependent humoral immune responses. Thus, the peripheral TCR repertoire could become narrower, shifted or alternatively broader due to CD154:CD40 costimulation supporting the clonal expansion of additional T-cell clonotypes that would otherwise be outnumbered by inter-clonal competition.

It has been shown previously that CD154 costimulation contributes to the TCR repertoire diversity during thymocyte development. Here, during the process of negative selection, CD154 deficiency permits the survival of T cells that bear specific Vβ segments, which are normally deleted in wild-type (WT) mice due to the recognition of superantigens presented in MHCII (9). Clear differences in the thymic T-cell repertoire have been described between WT and CD154-deficient mice, which were especially prominent in mice expressing the H-2E molecule (such as BALB/c mice) but also to a minor extent in mice expressing only the H-2A allele (such as C57BL/6 mice) (10–12). Whether CD154 costimulation affects the TCR repertoire during peripheral immune responses has not yet been clarified.

Here, we applied a high dose of sheep red blood cells (SRBCs) i.v. for induction of a strong local Th2 response in the spleen and isolated two individual T-cell zones (TCZ) by laser-microdissection (13) for analysis of the TCRβ repertoire by high-throughput sequencing. To compare the TCRβ repertoire between WT and CD154-deficient mice, we assessed typical features such as: (i) the number of TCRβ clonotypes, (ii) the percentage of identical TCRβ clonotypes between the groups, (iii) the frequency of individual TCRβ clonotypes, (iv) the length of their complementary determining region 3 (CDR3), and (v) the distribution of the V-J gene usage (14–16). Most of the differences were observed in both unimmunized and immunized mice, which clearly confirm the influence of CD154 costimulation during T-cell development in the thymus in C57BL/6 mice. However, the distribution of the V-J genes shifted differently after immunization in CD154-deficient mice compared to WT. These data demonstrate that CD154 costimulation influences the TCRβ repertoire not only during thymocyte development but also during T-cell differentiation in the periphery. Further studies are required to delineate whether targeting CD154 could be a therapeutic option to shape the TCR repertoire in a beneficial way in patients suffering from severe immune disorders.

### MATERIALS AND METHODS

### Mice and Injections

8–12-week-old female C57BL/6J WT mice were obtained from Charles River Laboratories (Sulzfeld, Germany). CD154 (CD40L)-deficient mice (C57BL6; 129S2-Cd40lgtm1Imx/J; provided by D. Gray, Edinburgh, UK), and JHT (*gh-Jtm1Cgn,* provided by Klaus Rajewsky, MDC Berlin) were bred in our animal facility (17, 18). Animal experiments were approved by local authorities of the Animal Care and Use Committee Kiel, Germany [V# 242-7224 122-1 (120-8/13) and (112-9/14)] and performed by certified personnel. A total of 109 SRBCs (Labor Dr. Merk, Ochsenhausen, Germany) were prepared and injected into the tail veins as described (13). The spleens were removed before and 72 h after injection, snap frozen and stored at −80°C.

### Histological Analysis

Serial cryo-sections of spleens (10 µm thick for histology, 12 µm thick for laser-microdissection, and 14 µm thick for 3D reconstruction) were mounted on plain glass slides for histology and 3D model reconstruction or on membrane-covered slides (Palm Membrane Slides, PEN membrane, 1 mm; Carl Zeiss AG, Germany) for laser-microdissection. T- and B-cell compartments of spleens were analyzed by immunohistochemical staining with biotinylated mAbs against TCRβ and B220 (both from BD Biosciences). Alkaline phosphatase goat anti-rat IgG (Roth, Karlsruhe, Germany) and goat anti-hamster IgG (Abcam, Berlin, Germany) were used as respective secondary Ab. Alkaline phosphatase activity was visualized with Fast Red (BB Salt, Sigma-Aldrich Chemie, Steinheim, Germany). Proliferating cells were identified by staining with rat anti-mouse Ki-67 mAb as primary antibody (BioLegend, Koblenz, Germany) and biotinylated rabbit anti-rat IgG (Dako, Glostrup, Denmark) as secondary antibody as described (19). For laser-microdissection and subsequent RNA analysis, the staining with toluidine blue was performed as described (13).

### 3D Reconstruction

For each condition, a complete collection of serial cryo-sections from half a spleen was imaged by automatic scanning microscope and processed by For3D as described (20, 21) using a Miraxmidi slide scanner (Zeiss, Jena, Germany). ImageJ and homemade Matlab functions were used to render spleen sections into 3D. TCZ were segmented by filtering, thresholding and soothing the stack of spleen section images. Matlab was used to identify individual volumes of the 3D structures within the spleen as described (20, 21).

### Laser-Microdissection

To obtain sufficient concentration of TCR-specific RNA, it was important to carefully select the largest TCZ. A total of 10–15 serial sections had to be prepared for the isolation of whole large TCZ by laser-microdissection (**Table 1**; **Figures 2A–C**). Therefore, the two largest TCZs were chosen carefully and isolated using a pulsed UV laser (Palm Microbeam; Zeiss microImaging GmbH, Germany). To estimate the TCZ volumes, the isolated TCZ areas were determined by the Palm Microbeam software (Zeiss microImaging GmbH, Germany) and multiplied by the section thickness (12 µm) (**Table 1**). In order to prevent any degradation of RNA, the tissues were shock frozen immediately after isolation and not allowed to thaw during their preparation. All specimen


Table 1 | TCZ volumes, raw reads, total, and unique TCRβ sequences obtained from laser-captured splenic TCZ in wild-type (WT) and CD154-deficient (KO) mice.

*Two individual TCZ per spleen were isolated by laser-microdissection and subjected to deep sequencing.*

*a A total of 2–3 TCZ were pooled for one analysis.*

were treated identically in order to exclude any biases between the mice.

# Gene Expression Analysis

Five serial cryo-sections of spleens were prepared for isolation of the total RNA with the innuPREP RNA Mini Kit (Analytik Jena, Hildesheim, Germany). After reverse transcription, the cDNA and the respective primers were added to the Taq Man PCR Master Mix (Applied Biosystems) and amplified. The optimal primer concentrations used were 900 nM each for the forward and reverse primers and 200 nM for the TaqMan probe (Biomers, Ulm, Germany): IFN-γ (for: 5′GCAAGGCGAAAAAGGATGC, rev: 5′GACCACTCGGATGAGCTCATTG, probe: 5′TGCCAAGTT TGAGGTCAACAACCCACAG); IL-4 (for: 5′GAGACTCTTTC GGGCTTTTCG, rev: 5′AGGCTTTCCAGGAAGTCTTTCAG, probe: 5′CCTGGATTCATCGATAAGCTGCACCATG); and MLN51 (for: 5′CCAAGCCAGCCTTCATTCTTG, rev: 5′TAACG CTTAGCTCGACCACTCTG, probe: 5′CACGGGAACTTCGAG GTGTGCCTAAC). For signal detection, the ABI Prism 7000 sequence detector (Applied Biosystems, Darmstadt, Germany) was used. The amount of cDNA copies was normalized to the housekeeping gene MLN51 according to the 2**ΔΔ**ct method (13, 22).

## Identification of TCR**β** Clonotypes Within Splenic TCZ by High-Throughput Sequencing

The RNA from TCZ was isolated as described above. The preparation of cDNA and amplification of the antigen-binding site (CDR3β region) of the TCRβ chains were performed according to the manufacturer's protocol (iRepertoire, patent 7999092, 2011, Huntsville, USA) and prepared for pair-end sequencing with the Illumina Miseq system as described (23). CDR3 identification, clonotype clustering, and correction of PCR and sequencing errors were performed using ClonoCalc wrapping MiTCR software according to the IMGT nomenclature (16, 24, 25). To avoid unpredictable PCR and sequencing errors, the default parameters ("eliminate these errors") were used. Additionally, to avoid artificial diversity due to PCR errors, all TCRβ clonotype sequences that appeared only once were removed (on average 4% of all sequencing reads). To compare similarity or diversity among the groups we calculated the Jaccard Index. Therefore, we asked how many TCRβ clonotypes that exist in one TCZ would be present also in all other TCZ from the other mice (excluding the one TCZ from the same mouse). By doing this for each TCZ 12 values (Jaccard indices) for each group were determined (Table S1 in Supplementary Material). These 12 values were taken for statistical analysis (two-way repeated measures ANOVA with Tukey's multiple comparison test). Further data analysis [frequency distribution (**Figure 3**), CDR3 length (**Figure 4**), principal component analysis (PCA) of V-J usage (**Figure 5**)] was performed after normalization of TCRβ clonotypes to the total number of TCRβ sequences (**Table 1**) using the R programming language, including the tcR package (15).

### Statistical Analysis

Statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Software Inc., La Jolla, USA). Statistical significance was assessed by Kruskal–Wallis test, Mann–Whitney *U*-test, two-way repeated measures ANOVA with Tukey's multiple comparison test, and multiple *t*-tests, one per row corrected for multiple comparisons with the Holms Sidak method. A *p* value of less than 0.05 was considered statistically significant.

### RESULTS

### CD154 Costimulation Is Essential for CD4 T Helper Cell Differentiation into Th2 Cells and B-Cell Maturation

It has been shown previously that CD154 deficiency has bidirectional effects during T-dependent humoral immune responses: (i) it impairs the differentiation of CD4 T cells despite normal T-cell expansions and (ii) it abolishes germinal centers (GC) formation and affinity maturation of B cells (26–28). However, some reports demonstrated that primary GC could appear even under CD154-deficient conditions (29). To investigate whether a high dose of SRBC induces GC in CD154-deficient mice we monitored B-cell proliferation immunohistochemically 10 days after injection. GC were observed in WT mice but not in CD154 deficient mice (**Figure 1A**).

Next, we quantified proliferating T cells and determined respective mRNA expression levels of the Th1 cytokine IFNγ and the Th2 cytokine IL-4. At 3 days after injection, the peak of T-cell proliferation, we observed a three- to fivefold increase of proliferating cells in the TCZ of both groups (**Figure 1B**). In contrast, the IL-4 mRNA expression increased only in WT mice and was completely abolished in CD154-deficient mice (**Figure 1C**) whereas the expression of IFNγ did not change in either group (**Figure 1D**). To find out whether DC or B cells mediate the effects of CD154 costimulation additional experiments with B-cell-deficient (JHT) mice were performed. The result revealed that B cells are required for the induction of IL-4 mRNA expression (Figure S1 in Supplementary Material). The crucial role for B cells in this model was further supported by their increased expression of MHCII and their uptake of CFSE-labeled SRBC *in vivo* (Figures S2 and S3 in Supplementary Material). In conclusion, our data show that CD154 deficiency impairs GC formation and Th2 differentiation but has no effect on T-cell proliferation in response to SRBC.

# Laser-Microdissection Allows the Isolation of Complete TCZ

It is well known that TCZ are located around the splenic arteries in periarteriolar lymphoid sheaths (30). However, the organization of these structures in whole spleens is not well described. Most current data were obtained and extrapolated from twodimensional tissue sections. Here, we performed a 3D reconstruction from half of the spleens (20, 21). Splenic TCZ appear as individual entities of highly diverse shape and size scattered throughout the spleen in transversal and longitudinal directions (**Figure 2A**; Video S1 in Supplementary Material). The volumes of the 20 largest TCZ range from 17 × 106 to 290 × 106 µm3 in naive and immunized spleens (**Figure 2C**). Due to the irregular shapes, it appears difficult to laser-capture a TCZ completely from two-dimensional cryo-sections. Therefore, only the two largest TCZ of one spleen were selected for isolation. Estimation of the laser-captured TCZ volumes revealed sizes of on average 53 ± 2 × 106 µm3 (mean ± SD) (**Table 1**), which is in the range of an entire TCZ. In conclusion, through the use of a stack of serial sections, an almost complete TCZ can be harvested by laser-microdissection (**Figure 2C**).

## CD154 Deficiency Increases the TCR Diversity in Splenic TCZ

Next, we isolated TCZ from WT and CD154-deficient mice, which were immunized or not. To exclude the possibility that CD154 deficiency influences the structure of the spleen and thereby the sizes and organization of the TCZ and B-cell zones (BCZ) a quantitative analysis of splenic compartments was performed (31). TCZ made about 40% and BCZ about 50% of the splenic area in both groups (Figure S4 in Supplementary Material). Due to the fact that no difference was observed regarding the TCZ and BCZ, we collected an identical area of TCZ from WT and CD154-deficient mice and analyzed their TCZ-TCRβ repertoire by high-throughput sequencing. We obtained between 0.8 and 1.88 × 106 total TCRβ sequences for TCZ of naive spleens and from 0.7 to 3.2 × 106 for TCZ of activated spleens, which contained between 10951 and 54652 unique TCRβ sequences (here referred to as TCRβ clonotypes) before immunization and from 12371 to 65306 after immunization, respectively, regardless whether the TCZ derived from WT or CD154-deficient mice (**Table 1**). The diversity occurring within each of the four groups (WT and CD154-deficient mice; unimmunized and immunized mice) was assessed as Jaccard index (**Figure 3A**; Table S1 in Supplementary Material). It provides a measure of similarity of samples and ranges from 0 to 1 as described in the method section (0, 100% different; 1, 100% identical). By contrast the Jaccard index is significantly lower in CD154-deficient mice compared to WT regardless of whether the mice were immunized or not. These data indicate that CD154 deficiency increases the number of distinct TCRβ clonotypes within the group of CD154-deficient mice. However, the Jaccard index does not differ significantly between unimmunized and immunized mice of either group, which indicates that CD154 deficiency influences the TCR diversity during thymic selection but not during the primary immunization with SRBC.

individual TCZ from spleens of wild-type mice before (A) and after immunization (B) are shown. Please see Videos in Supplementary Material. (C) 3D reconstructed spleens were used to determine the volumes of the 20 largest TCZ with Matlab (*n* = 3, 20 TCZ per spleen). The volumes of three naive and three immunized spleens from wild-type mice are shown and compared to the volumes of TCZ that were harvested by laser-microdissection and estimated from 2D cryo-sections (*n* = 12, two TCZ per mice).

Figure 3 | The diversity of TCRβ clonotypes is higher among the group of CD154-deficient mice compared to wild-type. Two TCZ per spleen of wild-type and CD154-deficient mice before and after immunization were laser-captured (two TCZ, *n* = 3 mice per group). TCRβ clonotypes were identified for each TCZ (six TCZ per group). (A) For every pair of TCZ within one group (excluding the TCZ of the same mouse) the Jaccard index, that is the number of TCRß clonotypes shared between the two TCZ relative to the total number of TCRß clonotypes occurring in the two TCZ together, is displayed (see Table S1 in Supplementary Material). In addition, the respective means are shown (\*\**p* ≤ 0.01, two-way repeated measures ANOVA with Tukey's multiple comparison test). (B) The relative frequency of TCRβ clonotypes of two representative TCZ from one spleen per group [wild-type (blue) and CD154-deficient mice (red) before and after immunization] is shown. (C) Box plot analyses were performed to compare the frequency distribution of TCRβ clonotypes. Copy numbers relative to the number of all TCRβ sequences obtained are displayed (*n* = 3, two TCZ per spleen). Extremely high-frequent TCRβ clonotypes (outliers) that are outside an IQR of 3 are displayed as single dots. The following relative median clonotype frequencies were found: 14.3 × 10−<sup>6</sup> for naive wild-type mice, 10.8 × 10−<sup>6</sup> for wild-type mice after exposure to sheep red blood cell (SRBC), 26.87 × 10−<sup>6</sup> for naive CD154-deficient mice, and 17.4 × 10−<sup>6</sup> for CD154-deficient mice after exposure to SRBC. (D) Absolute numbers of high frequent outliers were compared between groups. Indicated are mean and standard error (*p* value is displayed for difference between indicated groups, Mann– Whitney *U*-test, *n* = 3, two TCZ per spleen).

# CD154 Deficiency Marginally Reduces the Number of High Frequent Outliers

For comparison between the groups we elucidated the frequency of each individual TCRβ clonotype. Frequency plots were created in which each TCRβ clonotype is displayed as a single dot (x-axis) and arranged according to its relative frequency (y-axis) after normalization to the total number of TCRβ clonotypes in the respective TCZ. As shown in **Figure 3B**, a minority of TCRβ clonotypes appear at higher frequencies than the vast majority. Consistent with the fact that antigen-specific TCRβ clonotypes expand clonally after antigen exposure, the frequency of some high-expanded TCRβ clonotypes increases in both groups (**Figure 3B**; triangles compared to circles). For quantification, box plot analyses were performed. As **Figure 3C** shows, the distribution of copy numbers of individual TCRβ clonotypes is far from a Gaussian distribution in all four groups. Between 500 and 1,500 of the high frequent TCRβ clonotypes were identified as outliers irrespective of whether 1.5 (data not shown) or 3 was chosen as the interquartile range (**Figures 3C,D**). Enumeration of TCRβ outliers revealed that CD154 deficiency significantly reduces the accumulation of outliers compared to WT mice (**Figure 3D**). This reduction is found in both unimmunized and immunized mice, which indicates that CD154 costimulation regulates the number of outliers mainly during thymic selection but does not influence the expansion of TCRβ clonotypes during the immune response to SRBC.

## CD154 Deficiency Selects T Cells With Shorter CDR3 Regions

To further compare the TCZ-TCR repertoires between WT and CD154-deficient mice, we compared the length of their Fähnrich et al. CD154 Costimulation Affects the TCR-Repertoire

antigen-binding sequence, the CDR3 region. Given that the outliers better survive the inter-clonal competition and expand preferentially during thymic and peripheral selection, we assumed that they might represent the TCRβ clonotypes that are most affected by CD154 deficiency and therefore compared the length of the CDR3 region exclusively between the outliers of both groups (see **Figures 3C,D**). However, the CDR3 length of TCRβ outliers from CD154-deficient mice is shorter compared to WT mice. This applies in particular for the TCRβ clonotypes with a CDR3 length of 11 and 12 amino acids (**Figure 4**). For example, about 28% of the high frequent TCRβ outliers from WT mice have a CDR3 length of 11 AA and about 31.4% from CD154-deficient mice. In turn, a higher percentage (29.9%) of the high frequent outliers from WT mice have a length of 12 AA compared to only 27.2% in CD154-deficient mice (first and third columns, **Figure 4**). These shifts were found in both unimmunized and immunized mice, which indicate that the enrichment of TCRβ clonotypes with shorter CDR3 regions in CD154-deficient mice takes place during thymocyte development. However, marginal shifts in the CDR3 length due to the immune response to SRBC were found in WT outliers with a CDR3 length of 13 amino acids but not in CD154 deficient mice (**Figure 4**). The CDR3 length of the low or medium abundant T-cell clonotypes did not differ (data not shown).

## CD154 Signaling Controls the Selection of TCR**β** Clonotypes During Immunization

We compared the V-J gene usage between splenic TCZ-TCRβ clonotypes from WT and CD154-deficient mice by PCA (**Figure 5A**). The TCRβ clonotypes from both groups localize in distinct clusters. The result that the TCRβ clonotypes of WT and CD154-deficient mice remain separated after immunization reveals that shifts in V-J gene usage are induced during thymic selection, which has been described before (**Figure 5A**, left and right panel) (12).

In more detail, distribution analysis of the individual Vβ and Jβ gene segments revealed that WT mice have significantly less TCRβ clonotypes with TCR expressing Vβ12-2 and Vβ3.

Figure 4 | TCRβ outliers of CD154-deficient mice have shorter complementary determining region 3 (CDR3) regions. Relative length distribution of CDR3 regions of the high-frequent TCRβ outliers as described in Figures 3C,D were compared between wild-type and CD154-deficient mice before and after immunization. Mean ± SEM is shown, and bars above columns indicate significant differences between groups multiple *t*-tests, one per row corrected for multiple comparisons with the Holms Sidak method; \*\**p* ≤ 0.01.

Conversely, TCRβ clonotypes carrying the Vβ13-2 and Vβ19 genes were significantly enriched (**Figure 5B**). No significant difference in the Jβ gene usage was observed. After removal of all 4 Vβ genes (Vβ12-2, Vβ13-2, Vβ19, and Vβ3), which are sensitive to CD154 signaling in naive mice, the two groups are not separated anymore. (**Figure 5C**, left). However, distinct clusters of TCRβ clonotypes of WT and CD154-deficient mice reappear after exposure to SRBC (**Figure 5C**, right). Here, a significantly higher percentage of Vβ15 and Jβ2-1 was found in CD154-deficient mice compared to WT (data not shown). This data indicates that the lack of CD154 costimulation leads to an accumulation of TCRβ clonotypes expressing Vβ15 and Jβ2-1 segments whereas the presence of CD154 costimulation supports a more uniform distribution of the TCRβ clonotypes in regard to their V-J gene usage during the response to SRBC.

### DISCUSSION

A highly diverse peripheral TCR repertoire is a prerequisite for successful responses to infections or vaccinations (32). A decreased diversity has been linked to chronic infections (33), aging (2) and several autoimmune diseases (34, 35). The factors that form the peripheral TCR repertoire during immune responses as well as during lifetime are poorly defined. The current available data allow catching a first glimpse only on the extreme diversity of the TCR repertoire and its regulation. Here, we asked whether T-cell costimulation during peripheral immune responses contributes to the diversity of the TCRβ repertoire.

CD154, a member of the TNFR superfamily, is transiently expressed on antigen-activated T cells. Its ligand CD40 is found on antigen-presenting cells such as dendritic cells, B cells and macrophages but also on thymic epithelial cells. CD154 is a key molecule for T-cell survival in the thymus and CD4 T-cell differentiation in the periphery (36, 37). The effects of CD154 deficiency on the TCR repertoire of thymocytes have been described previously. In these studies monoclonal antibodies directed against certain Vβ segments from thymocytes of BALB/c mice have been used (11, 12).

Our study provides for the first time high-quality data on global TCR repertoires of the spleen as a central peripheral secondary lymphoid organ in unimmunized and immunized WT and CD154-deficient mice. To focus on the local TCR repertoire we isolated two TCZ per spleen by laser-microdissection. Besides the advantage that the local distribution of T-cell clonotypes remains undisturbed, this approach avoids any potential loss of cells during isolation procedures. This is of importance because a significant number of splenic T cells get lost by conventional isolation techniques (38). One could assume that the TCR repertoires could be biased because especially the highly activated T cells are prone to undergo apoptosis during the isolation steps compared to their resting counterparts. To find out it will be necessary to compare the TCR repertoires *in vivo* within the tissues and *in vitro* after isolation.

Analysis of the diversity of the TCZ-TCRβ repertoire revealed that the group of CD154-deficient mice shares less TCRβ clonotypes than the WT group, which indicates that splenic TCZ of CD154-deficient mice harbor a higher number of different TCRβ

clonotypes (**Figure 3A**). This observation is consistent with the impaired deletion of T-cell clonotypes during thymic negative selection in CD154-deficient mice (10, 11).

between wild-type and CD154-deficient mice before (left panel) and after immunization (right panel).

In general, our data reveal that CD154 costimulation has a strong impact on the selection of T-cell clones during T-cell development in the thymus. In the thymus, an increased TCR signaling strength during CD154 costimulation leads to the induction of apoptosis of those T-cell clones that bind with the strongest avidity to self-peptide MHCII complexes during the negative and agonist selection process. It is, therefore, expected that more T cells survive and diversity increases under CD154 deficient conditions. This higher diversity comes along with fewer outliers (**Figure 3D**), small shifts toward a shorter CDR3 region (**Figure 4**) and a higher number of T-cell clones that express Vβ12.2 and Vβ3 in CD154-deficient mice (**Figure 5**).

The situation is different during peripheral immune responses. Here, an increased TCR signaling strength should increase the progeny of the antigen-specific T-cell clones. Unexpectedly, our findings do not confirm this assumption. By analyzing the global TCZ-TCR repertoires, we find that CD154 costimulation has no effect on the diversity and the number of outliers during the primary immune response to SRBC. This finding might be explained by the fact that during T-cell development in the thymus each thymocyte is activated to express CD154 and selected for binding to CD40, whereas during peripheral immune responses only those T cells that are specific for the antigen have the chance to contact cognate CD40-matured B cells. It could be that the effects of CD154 costimulation would become more obvious if analysis was restricted to the antigen-specific CD4 T cells only, for example with the use of MHCII tetramers (4), instead of the bulk analysis performed here. However, recent studies indicated that the TCR sequences of antigen-specific T cells are extremely divers even in genetically identical and cage-matched mice. Many different T-cell clonotypes of both high and low frequency, rather than the dominant expansion of a few dominating antigen-specific clones, contribute to the immune response (39). In addition, it could be that the effects of CD154 costimulation could become stronger after repeated immunization with the same antigen. Here, we chose SRBC for immunization because it has multiple epitopes and induces CD4 T-dependent humoral immune responses without the need for adjuvants (40, 41). We assumed that SRBC would be recognized not only by the toll-like receptors of professional antigen-presenting cells due to their RNA content (42) but also by B cells due to the carbohydrate structures present on the surface of each red blood cell. The fact that the activation of B cells is critical for CD4 T-cell differentiation into Th2 cells was shown in B-cell-deficient mice (Figures S1–S3 in Supplementary Material). In addition, it has been demonstrated previously by administration of low or high doses of SRBC (13, 43). The major role of CD40–CD154 signaling during T–B interaction is further underlined by previous *in vitro* studies that demonstrated that this lack of IL-4 expression is not caused by an intrinsic inability of CD154-deficient T cells to express IL-4 (6). The level of TCR signaling strength might be increased under *in vitro* conditions due to a boosted peptide:MHCII density or a shifted ratio of antigen-presenting cells to T cells compared to the *in vivo* situation, which could enable the expression of IL-4 even without CD154 costimulation (6, 44). However, even though SRBC induced a strong polyclonal T-cell response in WT and CD154-deficient mice (**Figure 1B**) *in vivo*, the expression of IL-4 was impaired in CD154-deficient mice (**Figure 1C**). In addition, the obtained TCRβ repertoire data revealed that CD154 deficiency had no effect on the diversity and the number of outliers. Further studies are required to address the differences between TCR repertoire data obtained by analysis of antigen-specific T cells versus: (i) bulk analysis, (ii) the impact of secondary and tertiary immunizations, and (iii) the role of the nature of the antigen.

Unexpectedly, we detected shifts in the V-J gene usage due to the immune response to SRBC. This difference becomes visible only after exclusion of those V-J genes that were affected by CD154 deficiency during the thymic selection process. The question arises: what causes the distinct enrichment of T-cell clones according to their V-J genes? In the thymus the shifts in V gene usage have been linked to the expression of superantigens such as mouse mammary tumor virus, which are recognized predominantly by TCR expressing specific Vβ segments (9). Accordingly the shifts in V-J gene usage during the peripheral immune response could be due to differences in the presentation of SRBC-specific epitopes under CD154-deficient conditions. One possible explanation that would support this scenario is the finding that the costimulatory signals CD28:CD80/86 and CD154:CD40, which are crucial for the differentiation of CD4 T cells, precede as segregated events on distinct cells (8). This data leads to the assumption that B cells mount their own antigenspecific response, which is different from that of DC, and thereby recruit their own T-cell clones into the immune responses. Thus, in the case of CD154 deficiency, which impairs the interaction of CD4 T cells with B cells and prevents the antigen presentation by B cells, those T-cell clones that were activated preferentially by dendritic cells would accumulate more compared to the WT. Conversely, in WT mice with an intact antigen presentation by B cells, those CD4 T cells that were preferentially activated by dendritic cells would have to compete for interaction with CD40 matured B cells, which would clearly impact the composition of the individual T-cell clonotypes. A specific modulation of the B-cell response could therefore be a promising target for the modulation of the CD4 TCR repertoires.

Alternatively, the observed shifts in V-J segment usage could be explained by distinct migration behaviors. The lack of CD154:CD40 signaling during CD4 T-B interaction could prevent the migration into the BCZ of CD4 T cells that express the V-J segments that recognize SRBC-specific epitopes. To further clarify the effects of peripheral CD154 signaling it would be preferable to use CD154 conditional knockout mice. In addition, it will be interesting to find out whether CD40-deficient mice show a similar phenotype as observed under CD154-deficient conditions. CD40 is expressed on thymic epithelial cells before birth whereas CD154 expression was found only in neonatal mice (45). One could speculate that CD40-deficient mice would have a more skewed TCR repertoire than CD154-deficient mice.

In summary, in this study we provide evidence that CD154 signaling controls the selection of TCR clonotypes during a T-dependent humoral immune response. Further studies are required to investigate whether a variation of CD154 signaling could be used as a therapeutic option to modulate the TCR repertoire in a controlled manner. This could help to improve vaccines, treat autoimmune conditions, or prevent rejections after organ transplants. Due to this fundamental role of CD154 in adaptive immunity, CD154 signaling pharmacology for transplantation medicine and the treatment of autoimmune disorders is already being subjected to clinical trial (46, 47).

## ETHICS STATEMENT

Animal experiments were approved by local authorities of the Animal Care and Use Committee Kiel, Germany [V# 242-7224 122-1 (120-8/13) and (112-9/14)] and performed by certified personnel.

# AUTHOR CONTRIBUTIONS

AF designed experiments and analyzed data. SeK designed and performed experiments. AS performed experiments and analyzed data. CN, SaK, SM, and KKe analyzed data and contributed to manuscript writing, JW designed parts of the study and contributed to manuscript writing. KKa directed the study and wrote the manuscript.

# REFERENCES


## ACKNOWLEDGMENTS

We thank L. Gutjahr, P. Lau, Rene Pagel, and Daniela Rieck for their technical assistance and Felix Dino Lange for computational support. Richard Holland is thanked for language editing. This study was funded by the German Research Foundation (GRK 1727/2 project TP1), the Schleswig-Holstein Excellence Cluster Inflammation at Interfaces (EXC306/XTP4) and the TR-SFB654 project C4 and Focus Program "Modulation of Infection- and Allergy" at the University of Lübeck. We acknowledge financial support by Land Schleswig-Holstein within the funding programme Open Access Publikationsfonds.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01019/ full#supplementary-material.


antibody production delineates sites of cognate T–B cell interactions. *J Exp Med* (1993) 178:1555–65. doi:10.1084/jem.178.5.1555


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

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

# A Decade of Th9 Cells: Role of Th9 Cells in Inflammatory Bowel Disease

### *Shachi Pranjal Vyas and Ritobrata Goswami\**

*School of Bioscience, IIT Kharagpur, Kharagpur, India*

T helper cell subsets play a critical role in providing protection against offending pathogens by secreting specific cytokines. However, unrestrained T helper cell responses can promote chronic inflammation-mediated inflammatory diseases. Dysregulated T helper cell responses have been suggested to be involved in the pathogenesis of multiple inflammatory diseases, including allergic airway inflammation, rheumatoid arthritis, and inflammatory bowel disease (IBD) among others. Aberrant proinflammatory responses induced by Th1, Th2, and Th17 subsets are known to trigger IBD. IBD is a chronic inflammatory disease characterized by weight loss, diarrhea, pain, fever, and rectal bleeding. It poses a major health burden worldwide owing to the increased risk of colorectal cancer development. Despite numerous therapeutic advancements, IBD still remains a major health burden due to the inefficiency of the conventional therapies. Recently, IL-9-secreting Th9 cells are known to be involved in the pathogenesis of IBD. However, the role of Th9 cells and their secretory cytokine IL-9 in IBD is unclear. The functional relevance of Th9 cells is also relatively understudied in IBD. Thus, investigating the actual role of various T helper cell subsets including Th9 cells in IBD is essential to develop novel therapies to treat IBD. Here, we highlight the role of Th9 cells in promoting IBD. We discuss the mechanisms that might be employed by Th9 cells and IL-9 in promoting IBD and thereby propose potential targets for the treatment of Th9 cell-mediated IBD.

### *Edited by:*

*Loretta Tuosto, Sapienza Università di Roma, Italy*

### *Reviewed by:*

*Ashutosh Chaudhry, Memorial Sloan Kettering Cancer Center, United States Lorenzo Cosmi, Università degli Studi di Firenze, Italy*

> *\*Correspondence: Ritobrata Goswami ritobrata@gmail.com*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 28 February 2018 Accepted: 07 May 2018 Published: 24 May 2018*

### *Citation:*

*Vyas SP and Goswami R (2018) A Decade of Th9 Cells: Role of Th9 Cells in Inflammatory Bowel Disease. Front. Immunol. 9:1139. doi: 10.3389/fimmu.2018.01139*

Keywords: T helper cells, cytokines, Th9 cells, IL-9, inflammatory bowel disease, ulcerative colitis, claudin, occludin

# INTRODUCTION

T helper cells (CD4<sup>+</sup> T cells) constitute one of the key components of adaptive immune system. They play a crucial role in imparting protective immunity against a wide range of pathogens by secreting specific cytokines and chemokines (1). CD4<sup>+</sup> T cells attain the potential to secrete specific cytokines in response to environmental signals including strength of TCR-peptide engagement, co-stimulatory signaling, cytokine signals and expression of transcription factors in a process known as T helper cell differentiation (2, 3). Differentiated T helper cells play a myriad of functions in obliterating the infecting pathogens. However, when these cell-mediated responses are not properly regulated, unrestrained CD4<sup>+</sup> T cell responses lead to chronic inflammation and tissue damage. Exactly one decade back, another subset of CD4<sup>+</sup> T cells, which predominantly secrete the pro-inflammatory cytokine IL-9 was identified and christened as Th9 cells (4). Th9 cells have been known to provide immunity against helminth parasites (4–6). Moreover, they also play a vital role *in vivo* by providing antitumor immunity by secreting cytokines such as IL-9, IL-3, and IL-21 (7). Interestingly, Th9 cells are also known to induce inflammation and thereby exacerbate inflammatory diseases, including allergic asthma, multiple sclerosis, rheumatoid arthritis, among others (8, 9). Recent studies suggest that Th9 cells and their secretory cytokine IL-9 can also promote inflammatory bowel disease (IBD), which poses a significant risk factor for colon cancer development (10). However, the functional relevance of Th9 cells in IBD remains underappreciated and needs to be further investigated. Here, we briefly discuss how Th9 cells may be critical for the development of IBD.

### Th9 CELLS IN INFLAMMATION: ROLE IN IBD

Th9 cells are known to induce pathogenic responses in numerous inflammatory diseases including IBD, rheumatoid arthritis, allergic asthma among others. In this section we discuss the potential role of Th9 cells in IBD. IBD is a gastrointestinal tract disorder arising due to unrestrained gut inflammation and is divided into Crohn's disease (CD) and ulcerative colitis (UC) (11). Despite numerous therapeutic advancements, many patients continue to suffer from the complications of IBD due to lack of proper understanding of key immune players responsible for the pathogenesis of IBD.

Hallmark of IBD is chronic inflammation of the gut (9). Several T helper cell subsets are known to promote inflammatory responses in the gut. Previous studies have indicated that Th1 and Th2 cells are responsible for the pathogenesis of CD and UC, respectively (12). Moreover, Th17 cells were also demonstrated to induce chronic intestinal inflammation and promote IBD by secreting IL-17A (13). Additionally, it also known that in CD, Th17 cells secrete key signature cytokines of other T helper cell subsets such as IFN-γ (Th1) and IL-4 (Th2) resulting in the development of pathogenic Th17/Th1 or Th17/Th2 phenotype. Th17/Th1 cells are known to be more pathogenic than Th17 cells and are a promising target for the treatment of inflammatory conditions including CD (14, 15). Recently, it was observed that the transfer of Th9 cells resulted in the aggravation of UC in the gut mucosa of RAG-deficient mice indicating a crucial role of Th9 cells in IBD progression (16). Moreover, a correlation between UC progression and IL-9 secreted by Th9 cells in UC patients has also been demonstrated recently (16, 17).

Several murine models of chronic inflammation have been generated to understand the role of various immune players in IBD pathogenesis (18, 19). In TNBS-induced colitis model, there was an increase in the number of PU.1-expressing T cells and IL-9 secretion in the intestinal epithelial cells; underlying the importance of the transcription factor PU.1 in Th9 cell development (2, 20). IL-9 induced inflammation in mucosal epithelial cells and promoted colitis upon treatment with TNBS (20). Moreover, IL-9 deficiency resulted in reduced number of PU.1+ T cells and protected mice from colitis in TNBScolitis model indicating a role of Th9 cells and their secretory cytokine in the regulation of mucosal inflammation-mediated colitis (20). In oxazolone-mediated colitis model also there was an increase in the expression of IL-9 and IL-9R by intestinal epithelial cells (16). Moreover, deficiency of PU.1<sup>+</sup> IL-9<sup>+</sup> T cells resulted in suppression of experimental colitis upon oxazolone treatment indicating a critical role of PU.1+ T cells in promoting UC in mice. The role of IL-9 in promoting UC was also corroborated by investigating the intestinal mucosa of UC patients and healthy volunteers. There was enhanced number of mucosal T cells that express PU.1. Moreover, abundant IL-9R expressing intestinal epithelial cells were observed in the gut mucosa of UC patients indicating the role of IL-9 and Th9 cells in promoting colitis (16). These indicate that IL-9<sup>+</sup>, PU.1<sup>+</sup> Th9 cells play a vital role in the progression of UC and their differentiation needs to be tightly regulated to prevent the disease progression. Interestingly, a contradictory observation was demonstrated in DSS-induced model of UC, wherein IL-9 secretion by invariant natural killer T (iNKT) cells resulted in resolution of intestinal inflammation instead of promoting inflammation (21). Overall these studies suggest that the role of IL-9 in disease outcome depends on both the cytokine and the source of its secretion.

The potential mechanism for Th9-mediated colitis has been unraveled by investigating the function of Th9 cells and IL-9 in regulating the expression of tight junction proteins essential for maintaining intestinal barrier integrity (20). Tight junction proteins including claudins and occludin are essential for maintaining intestinal barrier functions and alteration in their expression is known to be responsible for numerous inflammatory disorders (22). In IL-9 knockout mice treated with TNBS, claudin 1 expression was downregulated; however, the expression of claudin 2 remained unchanged. Interestingly, in oxazolone colitis model, there was a downregulation in the expression of claudin 2 in IL-9-deficient mice than wild-type mice. This indicates that IL-9 might disrupt intestinal permeability by enhancing the expression of claudin 2 in oxazolone-mediated colitis (20). Thus, it could be suggested that IL-9 promotes colitis by regulating the expression of different tight junction molecules in different inflammatory conditions (20). Occludin, another tight junction molecule is also known to be upregulated in UC patients, but the role of IL-9 in regulating the expression of occludin in UC patients is not known so far (23). Disruption of intestinal barrier by IL-9 resulted in enhanced bacterial translocation in the mucosa of wild-type mice than IL-9-deficient mice in oxazolone-colitis model. This suggests that the potential mechanism for IL-9-mediated colitis involves disruption of the intestinal barrier culminating in increased bacterial entry into the mucosa and associated pro-inflammatory responses (16). The role of Th9 cells in regulating colitis in various mice models have been mentioned in **Table 1**.

IL-9 secreted by Th9 cells could be directly or indirectly involved in the progression of IBD. IL-9 is a pleiotropic cytokine involved in the activation and migration of various immune cells to the site of inflammation by inducing the production of chemokines. IL-9 has been reported to augment the expression of the chemokine eotaxin in smooth muscle cells (24). IBD has

Table 1 | Mechanisms employed by Th9 cells in regulating ulcerative colitis in various mice models.


been associated with the expression of multiple chemokines including RANTES, CCL19, and CCL21 (25). Thus, IL-9 produced by Th9 cells might also possibly induce the expression of such chemokines in intestine during IBD to promote inflammatory responses. It is also possible that the *in vivo* effects of IL-9 in IBD could depend on other immune cells and their secretory cytokines. IL-9 secreted by Th9 cells is known to promote Th2 skewed immune responses, thereby inducing UC in the process (16). IL-9 is also known to recruit and activate mast cells which secrete histamine, pro-inflammatory cytokines, and mast cell proteases to the intestine resulting in increased intestinal permeability and intestinal anaphylaxis (26). Thus, it is possible that IL-9 might recruit mast cells to intestinal epithelial cells, which in turn alter the intestinal barrier culminating in inflammation in the gut and UC.

Moreover, there is lack of evidence regarding the plasticity of Th9 cells in IBD. Th9 cells could possibly switch to an intermediate Th17 or Th1 cell type that would lead to the inflammation. Thus, blocking either Th17 or Th1-specific cytokine in such conditions could be of therapeutic interest. Therefore, it is critical to determine the actual role of Th9 cells and their secretory cytokine IL-9 in IBD.

## WHAT LIES AHEAD IN THE FUTURE: THERAPIES TARGETING CYTOKINES SIGNALING IN INTESTINAL MUCOSA FOR IMPROVED TREATMENT OUTCOMES?

The actual role of IL-9 secreting Th9 cells in various inflammatory conditions is not clear. The role of IL-9 in various inflammatory conditions also depends on the *in vivo* mice models. In OVA-induced model of allergic inflammation, Th9 cells were the main source of IL-9; while, in papain model of allergic inflammation ILC2s were the key source of IL-9 (2). Since observations from T cell transfer, TNBS and oxazolone-induced UC indicate a pathogenic role of IL-9 secreted by Th9 cells in the progression of UC; it would be benefical to determine whether UC progression can be curbed by preventing the expansion of Th9 cells in gut. Several cytokines and signaling molecules are known to promote Th9 cell development and proliferation. A recent study has demonstrated that Th9 cell differentiation can be promoted by IL-36γ (IL-1 family cytokine) signaling (27). In oxazolone-induced colitis, IL-36γ signaling was observed to promote naïve CD4+ T cell differentiation to Th9 cells by inducing phosphorylation of STAT5 and STAT6. Moreover, IL-36γ secretion also impaired Treg cell differentiation (27). IL-36γ signaling, therefore, abrogates Treg cell development and alters the balance between Th9 and Treg cells resulting in aggravated intestinal inflammation and progression of IBD (27). Thus, targeting IL-36 binding with its receptor in intestinal mucosal cells might be a promising strategy to control Th9 cell-mediated colitis (27). It has also been demonstrated that there is an increase in the production of IL-33 in UC patients. IL-33 is involved in promoting Th9 cell development independent of IL-4 signaling and blocking IL-33 could abrogate Th9 cell development (28, 29). Thus, these signaling molecules and cytokines could serve as potential targets for regulating Th9 cell differentiation and associated diseases. The potential mechanisms for regulating Th9 cell differentiation and IL-9 secretion have been depicted in **Figure 1**.

# FUTURE DIRECTIONS

The role of IL-9 in propagating UC is complex. IL-9 is known to prevent the complications of UC disease by suppressing the secretion of IFN-γ and IL-17A (21). Moreover, IL-9 secreted by iNKT cells in the DSS-mediated colitis model also dampened the immune response by inducing the release of IL-10 and TGF-β (anti-inflammatory cytokines) (21). Therefore, further studies involving determination of the source of IL-9 secretion in UC could potentially improve the therapeutic outcome among UC-infected patients. It has been observed that neutralization of IL-9 inhibits intestinal inflammation in oxazolone-induced

Figure 1 | Complex role of IL-9-secreting Th9 cells in inflammatory bowel disease (IBD). Interleukin-9, a pleiotropic cytokine, is known to play dichotomous role in IBD. IL-9 secretion by Th9 cells disrupts intestinal barrier permeability resulting in the entry of innocuous antigens into the gut mucosa. This leads to antigen presentation by the dendritic cells to naïve CD4+ T cells culminating in T helper differentiation. Several cytokines are known to promote Th9 cell differentiation. IL-33 and IL-36 play a key role in Th9 cell proliferation in the gut mucosa culminating in aggravated colitis. IL-9 secreted by Th9 cells is also known to promote Th2-like responses culminating in the progression of ulcerative colitis (UC). However, IL-9 secreted by cells other than Th9 cells such as invariant natural killer T (iNKT) cells are known to dampen inflammatory responses rather than promoting them. Thus, anti-IL-9 antibody might not be the "magic bullet" to treat IBD. Instead, targeting the cytokines involved in Th9 cell expansion in the gut mucosa could be a possible strategy to maintain optimal Th9 cell development in the gut and thereby restrain Th9 cell-mediated UC.

colitis model (16). However, it is not known whether blockade of IL-9 in other mice models of colitis would also give similar results or not. Moreover, the effect of IL-9 neutralization in humans is also unknown. Thus, deficiency of IL-9 might aid in controlling inflammation in UC, but it might also impair the protective immunity imparted by IL-9 secretion. Therefore, due to contradictory observations using various mice models of IBD, there is an unmet need to determine the actual role of Th9 cells and IL-9 in humans suffering from IBD. Additionally, it is essential to unravel the mechanisms employed by IL-9 secreting Th9 cells in the progression of IBD to develop novel therapeutic strategies and thereby curb intestinal inflammation and IBD.

### REFERENCES


### AUTHOR CONTRIBUTIONS

SV wrote the manuscript. RG conceptualized the idea and co-wrote the manuscript.

### FUNDING

This study received funding from Young Scientist Scheme, Department of Science & Technology, Government of India YSS/2015/001147/LS is awarded to RG. SV would like to acknowledge Council of Scientific & Industrial Research-University Grants Commission, India for providing Senior Research Fellowship.


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

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

*Tanbeena Imam1 , Sungtae Park2 , Mark H. Kaplan1 \* and Matthew R. Olson1,2\**

*1Department of Pediatrics and Herman B Wells Center for Pediatric Research, Indiana University School of Medicine, Indianapolis, IN, United States, 2Department of Biological Sciences, Purdue University, West Lafayette, IN, United States*

The gastrointestinal tract is a site of high immune challenge, as it must maintain a delicate balance between tolerating luminal contents and generating an immune response toward pathogens. CD4+ T cells are key in mediating the host protective and homeostatic responses. Yet, CD4+ T cells are also known to be the main drivers of inflammatory bowel disease (IBD) when this balance is perturbed. Many subsets of CD4+ T cells have been identified as players in perpetuating chronic intestinal inflammation. Over the last few decades, understanding of how each subset of Th cells plays a role has dramatically increased. Simultaneously, this has allowed development of therapeutic innovation targeting specific molecules rather than broad immunosuppressive agents. Here, we review the emerging evidence of how each subset functions in promoting and sustaining the chronic inflammation that characterizes IBD.

### *Edited by:*

*Ritobrata Goswami, Indian Institute of Technology Kharagpur, India*

### *Reviewed by:*

*Lewis Zhichang Shi, Case Western Reserve University, United States Benno Weigmann, Universitätsklinikum Erlangen, Germany*

*\*Correspondence:*

*Mark H. Kaplan mkaplan2@iu.edu; Matthew R. Olson olson126@purdue.edu*

### *Specialty section:*

*This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 28 February 2018 Accepted: 15 May 2018 Published: 01 June 2018*

### *Citation:*

*Imam T, Park S, Kaplan MH and Olson MR (2018) Effector T Helper Cell Subsets in Inflammatory Bowel Diseases. Front. Immunol. 9:1212. doi: 10.3389/fimmu.2018.01212*

Keywords: inflammatory bowel disease, Crohn's disease, ulcerative colitis, T helper cells, inflammatory cytokines, transcription factors

# CD4**+** T CELLS IN INFLAMMATORY BOWEL DISEASE (IBD)

Inflammatory bowel disease is a complex set of diseases that includes Crohn's disease (CD) and ulcerative colitis (UC), each with multiple bacterial, immune, and non-immune cell types contributing to inflammation. However, there are a number of lines of evidence that suggest that CD4<sup>+</sup> T helper cells are major initiators in the disease process. CD4<sup>+</sup> T cells are enriched in lesional tissue from patients with CD and UC and blockade or depletion of CD4+ T is effective in treating patients with IBD. In these studies, CD4+ T cell-depleting and blocking antibodies caused remission from disease in a number of CD and UC patients examined, suggesting a prominent role of CD4<sup>+</sup> T cells in propagating disease (1, 2). Interestingly, IBD patients with concurrent HIV infection also exhibit a greater incidence of remission as compared to non-HIV controls, correlating with decreased total blood CD4<sup>+</sup> T cell counts (3, 4). Finally, a number of biologics that target CD4<sup>+</sup> T cell differentiation into inflammatory subsets or their byproducts (i.e., cytokines) have shown efficacy in treating patients with IBD (5–7). Given the importance of CD4<sup>+</sup> Th cells in the disease process, this review will focus on how Th cells differentiate in the inflamed intestinal tract during IBD and how the Th lineage-specific cytokines and transcription factors (TFs) contribute to disease.

# Th1 CELLS

Th1 cells are important for protecting against infectious pathogens. These cells primarily produce interferon (IFN)-γ and tumor necrosis factor (TNF) that, respectively, activate macrophages and direct cytotoxic CD8<sup>+</sup> T cell responses, that in turn promote elimination of intracellular pathogens such as viruses and bacteria (8). In IBD, however, Th1 cells accumulate in the intestinal tract of individuals with CD and are directly associated with disease. Interleukin (IL)-12, which is secreted by

**185**

antigen-presenting cells, acts *via* signal transducer and activator of transcription (STAT)4 to promote the differentiation of naïve T cells into Th1 cells (9–11). STAT4 also signals activation of the TF T-bet, a lineage-defining factor for Th1 differentiation, which upregulates the IL-12 receptor, IFN-γ expression, and causes further expansion of Th1 cells (**Figure 1**) (12).

## Th1-Associated Cytokines

### Interferon-**γ**

Interferon-γ is the defining cytokine produced by Th1 cells and is used almost exclusively to identify Th1 cells in settings of disease. During intestinal inflammation, IFN-γ in combination with another Th1-associated cytokine, TNF, was proposed to drive intestinal epithelial cell beta catenin signaling and limit

Figure 1 | Critical factors in the differentiation of effector Th cells during inflammatory bowel disease (IBD). T helper cells recognize antigen presented in the context of major histocompatibility complex II on antigen-presenting cells in a T cell receptor-dependent fashion (not shown). In conjunction with assorted co-stimulatory signals (i.e., CD80/86–CD28 interaction and others), these signals initiate a program of cell division and differentiation. This differentiation program can be profoundly influenced based on the cytokines present in the environment in which they are initiated. Interleukin (IL)-12 and IL-23, cytokines induced during early stages of IBD, play important roles in differentiation of interferon (IFN)-γ/tumor necrosis factor (TNF)-producing Th1 cells as well as IL-17-producing Th17 cells. However, Th17 cells require additional signals including IL-6 and TGF-β for full induction of their differentiation. IL-6, in combination with TNF-α and tryptophan metabolites, initiates differentiation of protective IL-22-producing Th22 cells. Th1 differentiation is initiated and stabilized by transcription factors signal transducer and activator of transcription (STAT)4 and T-bet while Th17 cells require a combination of transcriptional regulators including STAT3, SMAD proteins, and RORγt. IL-4, IL-5, and IL-13-secreting Th2 and IL-9-secreting Th9 cells require IL-4 and STAT6 for their differentiation. Similar to Th17 cells, Th9 cells additionally require TGF-β, SMAD proteins, and a TGF-β/ SMAD-induced transcription factor PU.1 for their development. As a whole, the inflammatory mediators produced by Th cells in IBD play a role in the maintenance or breaking down gut epithelial barriers or in recruiting unique cells types to the intestines that further promote inflammation. Factors in red indicate genes involved in Th cell differentiation or function that contain single nucleotide polymorphisms that are associated with increased disease susceptibility or severity in humans (see Table 2).

their differentiation and proliferation (13). Despite this proposed model, the role of IFN-γ in murine IBD is controversial. Powrie et al. (14) and Ito et al. (15) both demonstrate that IFN-γ is required for disease development in a CD45RBhi RAG adoptive transfer model and in a DSS model of IBD (see **Table 1**), respectively. Nava et al. (13) also observed moderately reduced disease in IFN-γ-deficient mice in a DSS model of disease. Loss of IFN-γ in these reports correlated with overall reduced inflammation and tissue damage as well as reduced type-1-associated chemokine expression that would recruit other inflammatory cells to the intestinal tract. In contrast to these studies, Simpson et al. (16), using a adoptive transfer model of colitis modified from Powrie et al. (14), demonstrated that IFN-γ was not required for disease in this setting. Furthermore, Muzaki et al. (17) showed that IFNγ-deficient mice were in fact more susceptible to DSS-induced colitis, suggesting a protective role for IFN-γ. In a TNBS model of colitis (**Table 1**), IFN-γ was neither protective nor required for disease (18, 19).

The seemingly contradictory nature of the above studies suggests that other factors may be involved or compensate for the loss of IFN-γ in some models. Alternatively, the dependence of disease on IFN-γ may be a factor of the exact model system used or the differences in gut microbiomes within populations of mice across institutes or may be attributable to different mouse strains used. However, in human GWAS studies, there is a clear enrichment of SNPs in a CD- and UC-associated risk area comprised of several regions up and downstream of the human *IFNG* gene (45, 46). Furthermore, a particular IBD-associated SNP within the *IFNG* gene (rs1861494) is functionally linked with elevated IFN-γ expression in IBD patients (**Figure 1**; **Table 2**) (47). These data suggest at least a possible role of IFN-γ in promoting IBD in humans.

### Tumor Necrosis Factor

A wide array or both immune (i.e., Th1 cells, **Figure 1**) and nonimmune cells produce TNF during inflammation that induces signaling through two distinct receptors, TNFR1 and TNFR2. Similar to IFN-γ, TNF has also been demonstrated to be involved in intestinal barrier dysregulation during IBD (13) and has a varied role in mouse models of intestinal inflammation. Blockade of TNF in mice undergoing TNBS-induced disease, exhibited reduced weight loss and inflammation (24). Similarly, Yang et al. (25) demonstrated that TNFR1-, TNFR2-, and TNFR1/TNFR2 double-deficient mice exhibited similarly reduced TNBS-induced IBD as compared to control mice, suggesting a non-redundant role each receptor in this model. However, these findings were challenged by other studies in the TNBS model system where Ebach et al. (70) demonstrated that TNFR1-deficient mice had exacerbated disease and TNFR2-deficient mice had attenuated disease compared to controls, suggesting a protective and exacerbating role for each receptor, respectively.

Strikingly similar to the role of IFN-γ in murine IBD, the effect of TNF also varies greatly between model systems. Despite the protective role of TNF blockade in the TNBS model, TNF blockade in oxazolone-treated mice had little effect (24). In addition, Stillie and Stadnyk (71) showed that both TNFR1 and TNFR2-deficient mice developed significant disease after DSS treatment. Even


*Th-association was determined by induced expression in a particular model or by mechanistic studies, where cytokines/TFs were manipulated by gene deletion or blocking antibody. Acute, acute exposure to chemical; Chronic, repeating exposure to chemical; STAT, signal transducer and activator of transcription; IL, interleukin; TNF, tumor necrosis factor; IFN, interferon; TFs, transcription factors.*

more surprisingly, Wang et al. (72) reported exacerbated disease in DSS-treated TNFR1- and TNFR2-deficient mice as compared to controls, suggesting a protective role of TNF signaling during murine IBD. In these last experiments, lack of TNF signaling resulted in production of another auto-immune-associated cytokine, GM-CSF that played a critical role in the exacerbated disease phenotype, suggesting a role for compensating cytokines in some of these model systems. As a whole, these data indicate a need for standardization of model systems and sharing of reagents between laboratories to uncover potential discrepancies in these studies.

In humans, there are several SNPs in the TNF gene that are associated with IBD in a number of populations (**Figure 1**; **Table 2**). In addition, anti-TNF therapy is currently used in treatment of both CD and UC and is effective in a number of patients. Interestingly, while anti-TNF therapy was sufficient for complete remission of IBD-related symptoms (5, 6, 73), only ~24–50% of patients exhibited mucosal wound healing depending on the type of anti-TNF antibody used (74, 75). These data indicate a disconnect between tissue damage and what generally makes IBD patients feel "sick" (i.e., possibly more systemic effects). While both factors are relevant end points, mucosal wound healing is potentially more meaningful as these patients may be less likely to relapse after ceasing anti-TNF therapy.

### Th1-Associated TFs

### T-bet

T-bet is required for the differentiation of Th1 cells both *in vitro* and *in vivo* in the context of IBD. *Tbx21*-deficient CD4 T cells, isolated from BALB/c mice, were unable to induce colitis in a SCID adoptive transfer model of IBD. Furthermore, overexpression of T-bet in Th cells exacerbated experimental colitis in this same model (23). By contrast, a more recent study demonstrates that *Tbx21*-deficient CD4 T cells, this time from C57BL/6 mice, produced similar pathology as WT CD4 T cells (76). However, while *Tbx21*-deficient cells exhibited a dramatic decrease in IFN-γproduction, there was a corresponding increase in the proportion of cells that produced IL-17 and other Th17-associated cytokines in the inflamed intestines. Further mechanistic work demonstrated that *Tbx21*-deficient cells had heightened sensitivity to IL-23 signaling and increased RORγt expression, a TF associated with Th17 differentiation. As a whole, these data demonstrate that T-bet plays a pivotal role in Th1 differentiation in murine models of IBD, however, in some cases, the loss of Th1 cells is compensated for by an increase in pathogenic IL-17-producing cells resulting in a similar disease phenotype.

T-bet can also play a protective role in murine IBD. *Tbx21* deficient mice develop more severe colitis than control animals after treatment with DSS. This elevated disease is marked by enhanced immune cell infiltration into the intestines, ulceration, and loss of crypts (77). Interestingly, this disease became a spontaneous disease when crossed to Rag2-deficient mice that lack T and B cells, suggesting that CD4<sup>+</sup> T cells do not drive this disease. Instead, Garrett et al. (77) and Powell et al. (78) demonstrated that colitis in these mice was induced by exaggerated TNF production in dendritic cells (DCs) and elevated numbers of IL-17-producing innate lymphoid cells (ILCs) in the intestinal tract that causes a microbial breach in the intestinal lining.

Human GWAS studies have identified several SNPs that were associated with development of UC and CD (46, 79). Interestingly, a number of these SNPs sites were enriched for T-bet binding. Although T-bet bound a number of these intergenic regions in Gene SNP Th subset association Population studied Disease type Reference *IL12B* rs3212227 Th1 – Caucasian – Spanish Crohn's disease (CD), ulcerative colitis (UC) (48–50) rs6887695 Th1 – Spanish CD, UC (49) rs2288831 Th1 – Korean CD (51) rs10045431 Th1 – British (Caucasian) CD, UC (52, 53) rs6871626 Th1 – British (Caucasian) CD, UC (54) rs6556412 Th1 – British (Caucasian) CD, UC (54) *IFNG* rs1861494 Th1 – Caucasian CD, UC (47) rs7134599 Th1 – British (Caucasian) UC (54) *TNF* rs1800629 Th1 – Spanish – Portuguese Inflammatory bowel disease (55, 56) rs1799964 Th1 – Iranian – French Canadian CD (57, 58) *STAT4* rs7574865 Th1 – Spanish UC (59) rs925847 Th1 – Korean UC (60) *IL23R* rs11805303 Th17 – British (Caucasian) CD, UC (48) rs11209026 Th17 – Dutch(Caucasian), – Hungarian(Caucasian), – New Zealanders(Caucasian), – British (Caucasian) – Spanish CD, UC (49, 61–64) rs7517847 Th17 – British (Caucasian) – Spanish CD (49, 65) rs1004819 Th17 – German/Caucasian CD (50) rs11465804 Th17 – Mixed CD (52) *STAT3* rs12948909 Th17 – British (Caucasian) – German (Caucasian) UC (53, 66, 67)

rs744166 Th17 – Caucasian CD, UC (67)

rs2243248 Th2/Th9 – Iranian CD, UC (69)

– Mixed

*STAT6* rs324015 Th2/Th9 – British (Caucasian) CD (68) *IL4* rs2243250 Th2/Th9 – Iranian CD, UC (69)

Table 2 | Single nucleotide polymorphisms (SNPs) associated with Th-associated cytokines or transcription factors.

healthy individuals, several of these SNPs lead to reduced T-bet binding at these sites and altered gene expression of a number of Th1 related factors (*IL18RAP and TRIB1*) (80). Although both IL-18 and IL18RAP have been previously shown to be also associated with IBD, the role of TRIB1 in this setting is unclear. While these data need to be further verified using *in vivo* models of IBD, it might suggest that T-bet serves more of an inflammatory role than a protective role in human disease.

*JAK2* rs10758669 Th17 – British (Caucasian)

### Signal Transducer and Activator of Transcription 4

Binding of IL-12, produced by DCs and other innate immune cells, to its receptor triggers the phosphorylation of STAT4 which then translocates to the nucleus and initiates the transcriptional network associated with Th1 differentiation (81, 82). STAT4 binds a number of distinct regions within the genome including a number of key Th1-associated genes [i.e., *Ifng* and *Tbx21* (82, 83)]. Given its initiator role, STAT4-deficiency leads to a major impairment of Th1 differentiation *in vitro* as well as in mouse models of IBD (9, 10, 16). In support of these initial findings, mice engineered to overexpress STAT4 (STAT4-transgenic mice) developed an IBD-like disease after administration of dinitrophenyl-conjugated keyhole limpet hemocyanin that correlated with increased numbers of IFN-γ and TNF-producing CD4<sup>+</sup> T cells in the intestines (84). Interestingly, IFN-γ, a key Th1-associated gene downstream of STAT4 signaling, was not required for disease in an adoptive transfer model of IBD (16). These data suggest that although STAT4 is critical for Th1 differentiation and disease development in murine IBD, it may promote disease IFN-γ-independent manner.

CD, UC (52–54)

Interleukin-12 and STAT4 also appear to be linked to IBD in humans. There are a number of IBD-associated SNPs in the *IL12B* gene in Caucasian populations that correlate with enhanced disease (**Figure 1**; **Table 2**). STAT4 SNPs are associated with UC, but not CD, in some populations (51, 59, 85), suggesting a restricted role for these polymorphisms in human disease. More strikingly, STAT4 isoform ratios appear to be a better predictor of disease in IBD patients. STAT4 is produced as two distinct isoforms (α and β), where the STAT4β variant lacks 44 amino acids at the C-terminal end of the protein spanning the transactivation domain. Patients with UC or CD have elevated STAT4β:STAT4α ratios as compared to non-IBD controls and this ratio is normalized after steroid treatment (86). In both murine models of colitis and experimental autoimmune encephalomyelitis, CD4 T cells only expressing the STAT4β isoform drive exacerbated disease as compared to cells expressing the STAT4α isoform (87, 88). In the murine IBD model, STAT4α- and STAT4β-expressing cells had a similar capacity to produce IFN-γ and IL-17, but STAT4β preferentially drove expression of TNF and GM-CSF that have been also described to play a role in IBD severity (25, 88). Again, these data indicate that while STAT4 in CD4 T cells is critical for development of IBD, it might do so independently of IFN-γ.

### Th17 CELLS

The traditional Th1/Th2 paradigm of CD and UC was challenged with the discovery of a unique subset of IL-17-producing, termed Th17 cells (89). Th17 cells play an important role in maintaining commensal populations at important barrier sites (i.e., skin and gut), but when triggered in settings of autoimmunity can often exacerbate disease (89). The Th17 differentiation program is driven primarily by IL-6 and TGF-β (90) and further stabilized by signals from accessory cytokines including IL-23 and IL-1β (**Figure 1**) (91, 92). Interestingly, Th17 differentiation is inhibited by cytokines produced by other Th lineages including IFN-γ, IL-4, and IL-2 (93, 94). As IL-6 induces STAT3 phosphorylation, STAT3 and other downstream TFs, retinoic-acid-receptor-related orphan receptor (ROR) gamma and ROR alpha, are critical for Th17 differentiation *in vitro* and *in vivo* (95, 96). After differentiation, Th17 cells primarily express IL-17A and IL-17F; however, these cells can also co-produce signature cytokines from other Th cell lineages in particular settings of autoimmune inflammation (i.e., IFN-γ: Th1, IL-22: Th22, IL-9: Th9).

# Th17-Associated Cytokines

### Interleukin-17

Th17 cells are largely defined by their capacity to produce IL-17 (IL-17A and/or IL-17F) in settings of inflammation. IL-17 signals through a heterodimeric receptor (IL-17RA and IL-17RC) that is expressed on many non-hematopoietic cells including intestinal epithelial cells and on some activated T cells (97). Signaling through these receptors plays an important role in epithelial cell barrier function, and in the production of inflammatory chemokines and cytokines by target cells. In Th cells, IL-17 can also dampen the production of IFN-γ thereby possibly enhancing the stability of the Th17 phenotype by limiting Th1 differentiation (98).

Interleukin-17A protein in the serum and IL-17-producing Th cells in the gut-draining lymph nodes were elevated in patients with CD, but not UC, over non-IBD patients (99). In addition, the numbers of IL-17<sup>+</sup> cells in the intestines of patients with active CD and UC were also elevated as compared to healthy controls and patients with inactive CD or UC (100). While IL-17A blockade is successful in diminishing disease in patients with plaque psoriasis and ankylosing spondylitis, treatment of CD patients with IL-17A-blocking antibodies surprisingly enhanced disease severity resulting in a premature end to these clinical trials (101, 102). These adverse effect studies were substantiated in mice where IL-17 blockade in a DSS model of colitis exacerbated disease symptoms as well as immune cell infiltration into the mucosa (103). This result was recapitulated in IL-17-deficient mice treated with DSS (104). Furthermore, transfer of IL-17A- or IL-17RAdeficient Th cells into RAG recipient mice lead to an enhanced colitis-like wasting disease (98). Interestingly, IL-17R-deficient and IL-17R-Ig fusion protein-treated mice were protected from disease in TNBS model of UC, suggesting that the protective role of IL-17 may depend on the particular model of colitis used (30).

One explanation for the enhanced disease observed in anti-IL-17A-treated or IL-17A-deficient mice may be compensatory effects of other Th subsets or inflammatory cytokines. Indeed, O'Connor et al. (98) observed increased intestinal *Ifng* mRNA and increased Th1 polarization in the absence of IL-17A. Furthermore, addition of IL-17A to developing Th1 cells in culture suppressed IFN-γ production, suggesting that IL-17A limits Th1 differentiation and Th1-mediated immunopathology. Unfortunately, the authors did not neutralize IFN-γ in these mouse experiments to determine if elevated IFN-γ in mice that had received IL-17Adeficient Th cells was the cause of exacerbated disease. However, in *Abcb1a*-deficient mice that also develop enhanced colitis after *Helicobacter* infection and IL-17RA blockade, co-blockade of IFN-γ did not decrease disease severity (42), suggesting that increased IFN-γ may not be causative for the enhanced disease observed in the absence of IL-17. Leppkes et al. (105) demonstrated a different type of immune compensation in the absence of IL-17A. This study observed striking increases in IL-17F after adoptive transfer of IL-17A-deficient cells into RAG-deficient hosts. Interestingly, antibody blockade of IL-17F in mice that had received IL-17A-deficient Th cells ablated disease. Similarly, Wedebye Schmidt et al. (106) also found that antibody-mediated blockade of either IL-17A or IL-17F alone was insufficient to reduce disease severity, whereas blockade of both IL-17A and IL-17F completely abrogated disease. These studies highlight the compensatory nature of the intestinal immune response during IBD and indicate that multiple Th cell types and cytokines likely work in concert to promote autoimmunity in this setting.

Another theory of why the absence of IL-17A enhances intestinal inflammation lies within its ability to regulate epithelial barrier function and gut homeostasis. Maxwell et al. (42) observed increased gut permeability in the *Abdcb1a*-deficient colitis model when IL-17RA signaling was abrogated *via* blocking antibodies as compared to control mice. This was further substantiated in a DSS model of colitis where antibody-mediated blockade of IL-17 also resulted in enhanced "leakiness" of the intestinal epithelial barrier (107). In both cases, the increase in permeability correlated with changes in epithelial tight junction gene expression and changes in occludin positioning within the damaged epithelial layer (42, 107). Interestingly, Maxwell et al. (42) demonstrated that IL-17 signaling was also critical for production of anti-microbial peptides (AMPs) that may influence bacterial populations within the intestines during colitis. Song et al. (104) also showed that IL-17, in concert with fibroblast growth factor 2, regulated both epithelial barrier function and bacterial homeostasis in the gut. As a whole, these data indicate that although IL-17 has bona fide inflammatory properties, the barrier maintenance and microbial "grooming" function of IL-17 is likely dominant and is crucial for protection against intestinal barrier breach and enhanced inflammation during colitis.

### Interleukin-23

Interleukin-23 is composed of the IL-12p40 and IL-23p19 subunits and is produced primarily by DCs and monocytes in IBD (108, 109). IL-23 signals through the IL-23 receptor (IL-23R) which is induced by TGF-β, IL-6, and STAT3 signaling on activated T cells (110, 111). IL-23 distinctly enhances Th17 differentiation *in vitro* and lineage commitment *in vivo* (112, 113). Furthermore, IL-23 levels are elevated in intestinal biopsies taken from patients with IBD (109) and SNPs in the IL23R locus have been associated with increased risk for IBD (**Table 2**) (79, 114).

Interleukin-23 is also elevated in mice with colitis and largely appears to be required for disease progression. Interestingly, IL-23 plays an important role in both innate and adaptive immunedriven disease. Hue et al. (39) demonstrated that treatment with neutralizing antibodies to IL-23p19 blocked the development of innate-driven infectious colitis in a 129SvEvRAG<sup>−</sup>/<sup>−</sup> mice. Furthermore, adaptive immune-driven colitis in an adoptive transfer RAG model of disease was abolished when CD45RBhi cells were transferred into mice lacking both IL-12p40 and IL-23p19, but not in mice lacking individual cytokines (39). These findings were confirmed in similar models where either IL-23Rdeficient donor cells (115) were transferred or IL-23 signaling was blocked with neutralizing antibodies (116). There was also a reduction in disease in IL-23p19-deficient mice treated with anti-CD40 agonist antibody as compared to anti-CD40-treated WT controls (**Table 1**) (36). In essentially all situations, reduced disease in these animals correlated with a reduction in IL-17<sup>+</sup> CD4 T cells, particularly those that co-produce IFN-γ, suggesting that these IL-17<sup>+</sup>/IFN-γ+ cells may be the major inducers of disease. Interestingly, both Ahern et al. (115) and Uhlig et al. (36) observed that while IL-23p19-deficient mice lacked intestinal pathology there was very little difference in the weight loss that is often associated with murine models of colitis. By contrast, IL-12p40-deficient mice exhibited reduced weight loss, but minimal reductions in intestinal pathology (36), suggesting that weight loss and intestinal pathology are controlled by divergent mechanisms.

Similar to IL-17, IL-23 is not always pathogenic, and in some cases, can have a protective role. Becker et al. (37) demonstrated in TNBS and DSS models of colitis that IL-23p19-deficient mice were much more susceptible to developing colitis as compared to their WT counterparts. This phenomenon was also observed in a model of infectious colitis, using *C. rodentium*, where antibody-mediated blockade of IL-23 or IL-23-deficiency exacerbated disease and enhanced mortality compared to controls (117). Both Aychek et al. (117) and Becker et al. (37) noted increased IFN-γ in the absence of IL-23 and enhanced disease could be ameliorated by blockade of IL-12 or IFN-γ. This is consistent with several reports noted above where lack of IL-17 may result in enhanced Th1 responses and suggests a delicate balance in mechanisms that control development of intestinal inflammation.

### Th17-Associated TFs

### Signal Transducer and Activator of Transcription 3

STAT3 is activated *via* phosphorylation by kinases associated with the IL-6 and IL-21 receptor and is critical for *in vitro* and *in vivo* differentiation of Th17 cells (96, 118). In IBD patients, there is an increase in both total STAT3 protein as well as phosphorylated STAT3 in the inflamed colon over non-IBD controls which correlated with disease severity (119, 120). There is also a *STAT3* SNP associated with enhanced IBD susceptibility or disease severity in different populations, further implicating a role for STAT3 in both CD and UC (**Figure 1**; **Table 2**) (67, 121, 122). In line with these observations in humans, STAT3 signaling also plays a role in a mouse model of IBD. In a RAG adoptive transfer model of IBD, Durant et al. (40) demonstrated that STAT3 deficient CD45RBhi CD4 T cells were unable to differentiate into IL-17-producing cells and were unable to initiate inflammation in the intestines. Interestingly, lack of STAT3 in CD4 T cells had little effect on their capacity to produce IFN-γ, but significantly enhanced the frequency of these cells that became FOXP3<sup>+</sup> T regulatory cells. STAT3-deficient CD4 T cells also lacked the ability to proliferate or survive in the lymphopenic environment present in RAG-deficient hosts. The authors further demonstrated *via* CHIP-seq analysis that STAT3 binds and promotes epigenetic remodeling of both Th17- and anti-apoptotic/survival-associated genes suggesting a complex role for T cell-intrinsic STAT3 in the progression of colitis.

### SMADS

TGF-β binding to its receptor triggers activation and phosphorylation of SMAD proteins that translocate to the nucleus and influence transcription of a variety of genes involved in tumor metastasis and T helper cell differentiation. In T cells, TGF-β induces SMADs 2 and 3 that form a complex with SMAD4 and activate transcription of a number of genes including FOXP3 and RORγt (123). Furthermore, SMAD-induced TGF-β signaling also results in increased expression of SMAD7 that acts as an inhibitor of TGF-β-induced signaling pathway (124).

Although TGF-β is critical in the differentiation of Th17 cells, each SMAD protein has a unique and sometimes antagonizing role in this process. SMAD2 promotes Th17 differentiation *in vitro* by enhancing IL-6R expression or by physically interacting with the Th17-associated TF RORγt (41, 125). Furthermore, SMAD2-deficient CD4 T cells were unable to initiate colitis in a RAG adoptive transfer model and in a *C. rodentium* model of infectious colitis (**Table 1**) (41). By contrast, SMAD3-deficiency resulted in enhanced Th17 differentiation *in vivo* (125) and SMAD4-deficiency had little effect on Th17 differentiation (96). However, more recent reports indicate that SMAD4 degradation might play an important role in Th17 differentiation. Zhang et al. (126, 127) demonstrated that SMAD4-deficient cells produced IL-17 in the absence of TGF-β in culture, an over-looked point in the previous study, and ectopic SMAD4 expression suppressed Th17 differentiation. Although SMADs 2, 3, and 4 have been proposed to work together to drive transcription, it is clear from these reports that each SMAD protein has a unique role in Th17 differentiation.

SMAD7 expression is increased in lesional tissue isolated from patients with UC and CD, whereas phosphorylation of SMADs 2 and 3 is reduced as compared to non-IBD controls (128), suggesting a dominant role for SMAD7 in IBD. Blockade of SMAD7 activity with an anti-sense oligonucleotide enhanced SMAD2 activation and reduced inflammatory cytokine production in intestinal explants (128) and severity of colitis in mice sensitized and treated with TNBS or oxazolone (29). Furthermore, SMAD7-overexpressing T cells enhanced development of colitis in a RAG adoptive transfer model of colitis even in the presence of co-transferred T regulatory cells (129). Interestingly, SMAD7 transgenic CD4<sup>+</sup> T cells isolated from the intestines of these mice are not enriched for IL-17-producing cells, but instead have more IFN-γ-producing Th1 cells and have a lesser capacity to express FOXP3. These data are consistent with previous reports that demonstrated that TGF-β signaling in T regulatory cells induced FOXP3 which in turn limits SMAD7 expression (124). As a whole, these data indicate that an imbalance in T cell-intrinsic SMAD signaling does not promote Th17 differentiation *in vivo*, but rather enhances the ability of T cells to avoid Treg-mediated suppression.

### ROR**γ**t

An immune-specific isoform of retinoic acid receptor-related orphan nuclear receptor gamma (RORγt), a type of nuclear hormone receptor (NHR), is critical for the differentiation of Th17 cells (95). RORγt is induced by both IL-6/STAT3 and in some cases TGF-β, two cytokines that drive Th17 differentiation (130). Currently, there has been no identified IBD-associated *RORC* SNPs suggesting that while it is critical for Th17 differentiation, variations in this gene are not associated with enhanced disease.

Nuclear hormone receptors, like RORγt, are prime candidates for therapeutic intervention based on their mechanism of action where binding of a particular ligand controls their transcriptional activity. Natural ligands for RORγt have been recently identified as being a cholesterol biosynthetic intermediates (CBIs) and deletion of enzymes that generate these CBIs appear to be involved in Th17 differentiation (131). There has been a great deal of interest in generation of small molecule inhibitors that would be able to block or displace binding of natural RORγt ligands to therapeutically inhibit Th17 responses during IBD. Withers et al. (132) demonstrated that therapeutic delivery of an oral-available Rorγt inhibitor (GSK805) was able to significantly reduce disease in a number of murine IBD models. Interestingly, GSK805 treatment was more effective in treating disease that IL-17 blockade or deficiency and was not marked by compensatory increases in IFN-γ production as observed in studies mentioned above. These data suggest that either RORγt small molecule inhibitors target additional RORγt-expressing cells types that contribute to inflammation (i.e., ILC3s) or inhibit other Th17-associated functions besides IL-17 production.

### Th22 CELLS

Th22 cells are characterized by their production of IL-22, but not IFN-γ or IL-17. Differentiation of Th22 cells occurs in the presence of IL-6, TNF-α, and IL-1β, and suppressed by TGF-β (**Figure 1**) (133, 134). Similar to Th17 cells, IL-21 also can enhance the differentiation of IL-22-producing T cells (135). Although Th22 cells do not produce IFN-γ, they express high levels of T-bet that is critical for their differentiation (136). Furthermore, while expression of the aryl hydrocarbon receptor (AhR) TFs is low in Th22 cells as compared to Th17 cells, it is also critical for IL-22 production from CD4 T cells (133, 136).

### Th22-Associated Cytokines Interleukin-22

Interleukin-22 is a member of the IL-10 family of cytokines and binds to IL-22R which is complex of IL-22R1 and IL-10R2 that signals primarily through STAT3 (137). IL-22 is elevated in both mice and humans with IBD (138, 139) and is known to have a protective effect against gut inflammation, tissue damage, and bacterial infection (136, 140–142). IL-22 producing Th cells were significantly reduced in inflamed tissues from UC patients and were replaced by IL-17-producing Th17 cells, suggesting that the balance of inflammatory and anti-inflammatory cells is disrupted during IBD.

In IBD, IL-22 ameliorated intestinal inflammation in a STAT3 dependent manner that correlated with enhanced mucus production by colonic epithelial cells (143). Mice deficient in IL-22 are susceptible to *C. rodentium* infection (136) and mice that lack STAT3 specifically in T cells exhibited enhanced susceptibility to infectious colitis during *C. rodentium* infection that was rescued by intestinal overexpression of IL-22 (144). Furthermore, adoptive transfer of Th cells cultured under Th22 conditions was sufficient to protect from *C. rodentium* infectious colitis (136). In both a DSS and RAG adoptive transfer model of colitis, IL-22-deficient mice or mice receiving IL-22-deficient Th cells exhibited enhanced weight loss and colon inflammation as compared to controls indicating that IL-22 has a protective role in multiple models of murine colitis (145, 146). Together, these data suggest that Th22 cells are a relevant protective source of IL-22 during intestinal inflammation.

Interleukin-22 mediates its protective properties by inducing production of mucus and AMPs that have direct anti-microbial activity (141, 147). In addition, injection of IL-22 into mice induced the expression of LPS binding protein by hepatic cells thereby increasing the capacity to neutralize systemic LPS and reduce inflammation (138).

### IL-22 Binding Protein (IL-22BP)

IL-22 binding protein exists as soluble inhibitory receptor and it is expressed at high levels in the PP and colon by DCs in the steady state and by CD4<sup>+</sup> T cells during inflammation (146, 148–150). IL-22BP has higher affinity than IL-22R1 and competes for IL-22 binding with the cell-associated receptor, and interferes with the protective role of IL-22. Both UC and CD patients exhibit increased IL-22BP expression and IL-22BP<sup>+</sup> cells in inflamed intestinal tissue as compared to healthy controls (149). Importantly, transfer of IL-22BP-deficient Th cells into either IL-22BP-sufficient or -deficient RAG mice resulted in less disease as compared to mice receiving WT Th cells, indicating that Th cell-derived IL-22BP was the biologically dominant source for pathogenic IL-22BP production during IBD (149). In future studies, it will be important to determine what Th cells are producing IL-22BP (i.e., Th1 and Th17) or if these cells are a unique Th subset. Furthermore, the factors that result in IL-22BP production by Th cells is unknown. Understanding of this pathway will have clear implications for treatment of IBD.

### Th22-Associated TFs Aryl Hydrocarbon Receptor

STAT3 signaling and T-bet are required for the differentiation of Th22 cells (136) and the role of these factors in IBD have been Imam et al. Effector T Helper Cells in IBD

discussed above in terms of Th17 and Th1 differentiation, respectively. Although Th22 cells do not express elevated levels of AhR as compared to Th17 cells, it is critical for their differentiation (136). As Th22 cells are associated with protection against intestinal disease, it is not surprising that AhR is also associated with protection from IBD. Patients with IBD exhibit significantly higher IL-22 and AhR expression as compared to healthy individuals (151), presumably to counteract enhanced inflammation within the tissue. In mice, transfer of AhR-deficient Th cells in a RAG model of colitis induced more severe disease as compared to mice that received WT Th cells (135). In a murine TNBS colitis model, an AhR antagonist induced more severe colitis by suppressing synthesis of IL-22. By contrast, treatment of AhR agonist 6-formylindolo (3, 2-b)carbazole (FICZ), a tryptophan derivative, reversed relapsing TNBS and DSS-induced colitis (151). Mechanistically, tryptophan metabolites can alter the intestinal microbiota in favor of microbes that induce AhR production and AhR-dependent IL-22 transcription (152). As a whole, these data indicate that AhR expression in Th cells is required for IL-22 production and protection from IBD.

# Th2 CELLS

Th2 cells classically function to provide anti-parasite immunity, but are also known to be effector cells in asthma and differentiate in response to IL-4 (153). IL-4 signaling in Th2 cells leads to activation of the receptor-associated signaling molecule STAT6 and downstream induction of the Th2 defining TF GATA3 (**Figure 1**) (153). GATA3 is able to further polarize the differentiation of Th2 cells through a positive autoactivation pathway (154) and can also convert committed Th1 cells can to Th2 cells when ectopically expressed (155). While Th1 cells are indicative of CD, Th2 or Th2-like cells are more associated with UC. This Th1/Th2 paradigm, although controversial, has been recently supported by the development of an equation that can predict CD vs. UC based on cell populations with 83% accuracy (156).

### Th2-Associated Cytokines Interleukin-4

The role of IL-4 in perpetuating IBD is controversial. T cells isolated from UC biopsies do not exhibit significant production of IL-4, which has shown to be vital in the differentiation of Th2 cells and their defining cytokine (157). Similarly, *IL4* mRNA expression in intestinal mucosa in both CD and UC patients was undetectable (158). IL-4, in combination with IL-10 (another cytokine enriched in Th2 cells), has also been shown to synergistically inhibit the pro-inflammatory cytokines TNF-alpha and IL-1β that are associated with IBD (159). Furthermore, SNPs present in the IL-4 gene are overrepresented in UC patients and are presumably loss of function mutations, suggesting a possible regulatory role of these cytokines in IBD patients (**Figure 1**; **Table 2**) (69). However, anti-IL-4 treatment of mice undergoing oxazolone-induced colitis or TCRα−/<sup>−</sup> adoptive transfer model prevented the majority of disease (31, 43, 160). These data indicate that IL-4 may have anti- or pro-inflammatory roles based on the particular type of IBD or mouse model used.

### Interleukin-5

Stimulated lamina propria (LP) T cells isolated from colonic biopsies from UC patients had increased expression of IL-5 compared to CD and control patients (161). An assessment of cytokine transcripts in UC and CD patients found IL-5, IL-13, IL-15, and IL-33 mRNA levels to be increased in UC patients (162). Although IL-5 is produced in these tissues, its exact contribution to UC is still unclear. IL-5 is a potent inducer of eosinophils from the bone marrow and there is some evidence that elevated IL-5 levels in UC promotes the recruitment of eosinophils into the inflamed intestine (163, 164). The role of these eosinophils in UC, however, remains unclear.

### Interleukin-13

Similarly, LP mononuclear cells from UC produced much larger amounts of IL-13 than CD or control patients (33). However, the source of IL-13 was not always T cell derived. In human studies, both LP Th cells and invariant natural killer T cells accounted for the majority of the IL-13 production (161, 165). IL-13 is also produced in mice by intestinal ILCs (166, 167). Interestingly, IL-13 signaling through the once thought "decoy" IL-13Rα2 receptor induces TGF-β and fibrosis of intestinal tissue in a chronic model of TNBS colitis (26, 27). Given the apparent importance of IL-13 in both human IBD and murine model studies, IL-13-specific antibodies (anrukinzumab and tralokinumab) have been trialed as therapy for UC patients. Unfortunately, anrukinzumab did not show significant therapeutic effect in UC patients with active disease (168). Tralokinumab showed some improved clinical remission rates and mucosal healing, but did not have significant clinical response when compared to placebo, which may indicate limited therapeutic benefit (169).

# Th2-Associated TFs

### Signal Transducer and Activator of Transcription 6

STAT6 is activated by the cytokines IL-4 and IL-13 and binds a number of IL-4-responsive promoters (170). Activated STAT6, as measured by phosphorylation, was also enhanced in intestinal tissues take from UC patients and correlated with disease severity (171). There is also one reported SNP in the STAT6 gene that is associated with CD (**Figure 1**; **Table 2**) (68). In models of UC (e.g., DSS and oxazolone-induced UC), STAT6 can play an important role in IBD pathogenesis through changes in inducible NO synthase or tight junction proteins and Th2 cytokine production (20, 172). However, STAT6 was found to be dispensable in a TCRα−/<sup>−</sup> colitis model (173).

### GATA3

STAT6 binds the promoter and activates transcription of GATA3; GATA3 can subsequently induce its own expression and inhibit Th1 production. Thus, GATA3 is a lineage-defining factor for Th2 cells. GATA3 was expressed at higher levels in colonic tissue from UC patients compared to ileal CD patients and correlated with disease severity. Furthermore, UC patients had an increased number of mucosal CD4<sup>+</sup>/GATA<sup>+</sup> cells (35). This has also been shown in pediatric UC patients, those with active disease had increased expression of mucosal *GATA3* compared to age-matched controls (174). In a murine models of UC (DSS and oxazolone-induced UC), overexpression of GATA3 accelerates acute colitis in contrast to overexpression of T-bet or RORγt (21, 35). Furthermore, T cell-specific GATA3 deficient mice were resistant to oxazolone-induced disease and treatment with hgd40 DNAzyme that cleaves *GATA3* mRNA also exhibit a significant decrease in disease severity (35). Currently, SB012, which contains hgd40, is being investigated for use in UC patients (https://www.clinicaltrials.gov/show/ NCT02129439).

### Others

c-MAF is a TF that is associated with Th2 and Th17 differentiation and aids in both IL-4 and IL-17 production (175, 176). CD and UC patients have increased numbers of c-MAF<sup>+</sup> T cells in the inflamed intestine. Interestingly, c-MAF<sup>+</sup> cells were also T-bet<sup>+</sup>, a Th1-associated factor, suggesting that these cells may have a Th1-like phenotype (177). Furthermore, c-MAF overexpressing naïve CD4 T cells were unable to induce colitis as compared to WT controls. However, c-MAF overexpression within memory/ effector CD4 T cells, normally containing T regulatory cells, augmented colitis when co-transferred with naïve Th cells (177). Deletion of c-MAF in T regulatory cells also limited their ability to produce the anti-inflammatory cytokine IL-10 and regulate microbiota-specific Th cells (178). These data indicate that there is a fine balance in c-MAF expression in T regulatory cells that controls their regulatory capacity.

The TF NFAT is also involved in Th2 differentiation and NFAT family member expression was enriched in the inflamed intestine during IBD (179). A gene controlling NFAT nuclear translocation was identified as a key IBD susceptibility gene in human IBD GWAS studies (180). Furthermore, genetic deletion of NFATc2 results in reduced disease in an oxazolone model of murine colitis by modulating the ability of Th cells to produce IL-6, a cytokine normally associated with Th2 cells (181).

# Th9 CELLS

Interleukin-9 producing T cells were initially thought to be a subset of the Th2 population. This changed after the finding that TGF-β, together with IL-4 reprogram Th2 cells to become Th9 cells that secrete IL-9 and IL-10 and have a unique transcriptional profile (**Figure 1**) (182, 183). In these cells, IL-9 production is under the control of PU.1, STAT6, BATF, GATA3, and IRF4 (183, 184). TGF-β, in particular, stimulates the production of TF PU.1, which in turn induces IL-9 expression (185).

# Th9-Associated Cytokines

### Interleukin-9

Recent studies both of CD and UC patients have shown IL-9 production to be enhanced in these disease states and correlates with endoscopic Mayo scores (34). Further studies have also revealed that in UC patients, activated peripheral blood lymphocytes produced increased amounts of IL-9 (186). Clinically, patients with increased serum IL-9 levels have been shown to have a worsened prognosis (187) and higher levels of systemic IL-9 were also associated with cachexia and lower hemoglobin concentrations in IBD patients (188).

In mouse studies, mice with oxazolone-induced colitis exhibited increased IL-9 expression and IL-9-deficient mice were resistant to development of colitis (34). Interestingly, therapeutically targeting IL-9 with a neutralizing antibody in these studies also significantly reduced disease over controls, and this was also verified in a DSS model of UC (22). In a TNBS-induced colitis model, IL-9-deficient mice were also resistant to colitis development and suggested that IL-9 may have a role in regulating intestinal barrier function (28). In additional studies using the RAG adoptive transfer model of colitis found that when IL-9 producing T cells were transferred into RAG 1-deficient mice, the mice exhibited increased degree of colitis (189). As compared to other cytokines that have model-dependent protective or pathogenic roles in IBD (i.e., IFN-γ and IL-17), IL-9 appears to have a consistent pathogenic role across disease models making it a potential druggable target in IBD.

Tofacitinib, one of the newest approved therapies for IBD, is a small molecule Janus kinase (JAK) inhibitor that has shown promising therapeutic potential in UC (190). The JAK family includes intracellular tyrosine kinases that activate STATs to control multiple cytokines including IL-9. However, IL-9 is one of the many cytokines that are inhibited; perhaps newer therapies with more targeted IL-9 activity may be beneficial in patients who have failed other forms of treatment.

### Interleukin-10

Th9 cells also produce IL-10, which is classically known to be immunosuppressive (184). However, as above, experiments by Dardalhon et al. showed that transferring IL-9 and IL-10 double positive cells into RAG 1-deficient mice worsened colitis in these mice (189). However, IL-10 and its role in IBD has been studied extensively. IL-10-deficient mice irrefutably develop colitis spontaneously and pathologically resemble human IBD (191). The role of IL-10 in relation to Th9 cells, however, needs to be further elucidated.

### Th9-Associated TFs PU.1

PU.1 is required for Th9 development and is produced more in these cells compared to the other subsets (185). In Th9 cells, PU.1 binds to the *Il9* promoter and alters histone acetylation (192). In human IBD, the number of PU.1<sup>+</sup> Th cells is higher in mucosal biopsies of patients with active IBD (34). This was also shown using immunohistochemistry to assess the density of PU.1 expressing cells in both human UC and murine DSS colitis (22). Moreover, mice with PU.1-deficient T cells had diminished pathology in the oxazolone model (34).

### Others

STAT6, GATA3, and SMAD proteins are also required for Th9 differentiation and may be involved in Th9-mediated colitis. Indeed, IL-9 is significantly reduced in oxazolone-induced colitis when GATA3 is deficient in T cells (35). However, the role of STAT6 and SMADs on IL-9 production in colitis has not been examined. IRF4 is also required for differentiation of Th2, Th9, and Th17 cells (193–195) and is required for inflammation in a RAG adoptive transfer model of colitis (196, 197). Therefore, a number of TFs that play a role on other Th lineages can also impact Th9 function during intestinal inflammation.

### OUTLOOK ON IBD THERAPY

Over the last several decades, researchers have generated a tremendous amount of information regarding how the CD4<sup>+</sup> Th response influences the outcome of IBD in both mice and humans. This has already led to the development of a number of biologics (i.e., anti-TNF) that are effective in causing at least temporary remission from disease. However, these therapies are only effective in causing endoscopic and microscopic remission in a subset of patients. In this review, we have highlighted a number of instances where depletion or deletion of particular Th-associated cytokines can have deleterious effects. In the majority of these instances, there are compensatory immune mechanisms that either render treatment ineffective or make disease considerably worse. Clearly, more work must be done to identify these compensatory mechanisms and be able to predict them in both murine models of disease and in human patients. Broad-scale successful treatments of IBD will likely require targeting multiple arms of the Th cell response to account for compensating mechanisms that may be unique to each individual. These personalized treatments might include bi-specific antibodies or multiple biologics used in conjunction to account for predicted compensatory responses.

There are, however, some inherent drawbacks to this "multifaceted" approach. Targeting multiple immune mechanisms will likely occur at increased cost to the patient. CD treatment is already ~3× more expensive than UC therapy with ~64% of that cost being driven by anti-TNF therapy (198), a cost that will likely increase when multiple biologics are used. One approach to this problem is to utilize a broader pan-Th inhibitor that might target multiple cytokines that are produced by a number of Th subsets. As the majority of these cytokines signal through cytokine receptor-associated JAKs, JAK inhibitors are an attractive target

### REFERENCES


for limiting autoimmune inflammation. The JAK inhibitor tofacitinib has been approved for treatment of rheumatoid arthritis and has shown promise in phase 3 clinical trials in IBD (199). Beyond cost, targeting multiple arms of the immune system may also lead to increased susceptibility to infection or cancer. Although a number of Th-associated cytokines are involved in IBD pathogenesis, almost all of these have evolved for protection from infection and for regulating gut homeostasis and microbiota. Patients on anti-TNF therapy are more susceptible to a number of infections and development of particular types of tumors (200, 201). It is easy to imagine that combination of anti-TNF therapy with anti-IL-23 or a JAK inhibitor would further increase susceptibility and thereby limit the efficacy or desirability of treatment. Ideally, future treatments would target offender-only molecules that do not also play protective or homeostatic roles in the immune system. A recent study by van Unen et al. (202) used high parameter time of flight mass cytometry (CyTOF) to identify a number of novel immune cell populations that correlate with disease in IBD patients. This type of data will be instrumental in identifying possible "offenderonly" immune cell populations that may be targeted therapeutically without hampering normal host immune responses.

## AUTHOR CONTRIBUTIONS

TI, SP, MK, and MO wrote the manuscript and designed figures and tables.

# FUNDING

The work in the study was supported by NIH grants R01 AI057459, R01 AI129241, R21 AI117380, and R01 AI095282 to MK. Support provided by the Herman B Wells Center was in part from the Riley Children's Foundation. TI was supported by the Department of Pediatrics through the Project Development Team within the ICTSI NIH/NCRR Grant Number UL1TR001108. MO had support from Purdue University.

in patients with moderate-to-severe Crohn's disease: a randomised, double-blind, placebo-controlled phase 2 study. *Lancet* (2017) 389(10080): 1699–709. doi:10.1016/S0140-6736(17)30570-6


evidence of interaction. *Inflamm Bowel Dis* (2008) 14(9):1192–6. doi:10.1002/ ibd.20463


through Foxp3 induction and down-regulation of Smad7. *J Immunol* (2004) 172(9):5149–53. doi:10.4049/jimmunol.172.9.5149


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

The handling Editor declared a past co-authorship with one of the authors [MK].

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

# Beyond Type 1 Regulatory T Cells: Co-expression of LAG3 and CD49b in IL-10-Producing T Cell Lineages

Weishan Huang1,3 \*, Sabrina Solouki <sup>3</sup> , Chavez Carter <sup>3</sup> , Song-Guo Zheng<sup>2</sup> and Avery August <sup>3</sup> \*

<sup>1</sup> Department of Pathobiological Sciences, School of Veterinary Medicine, Louisiana State University, Baton Rouge, LA, United States, <sup>2</sup> Department of Medicine, Milton S. Hershey Medical Center, Pennsylvania State University, Hershey, PA, United States, <sup>3</sup> Department of Microbiology and Immunology, College of Veterinary Medicine, Cornell University, Ithaca, NY, United States

### Edited by:

Amit Awasthi, Translational Health Science and Technology Institute, India

### Reviewed by:

Silvia Deaglio, Università degli Studi di Torino, Italy Ashutosh Chaudhry, Memorial Sloan Kettering Cancer Center, United States

### \*Correspondence:

Weishan Huang weishan.huang@cornell.edu; huang1@lsu.edu Avery August averyaugust@cornell.edu

### Specialty section:

This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology

Received: 28 February 2018 Accepted: 25 October 2018 Published: 19 November 2018

### Citation:

Huang W, Solouki S, Carter C, Zheng S-G and August A (2018) Beyond Type 1 Regulatory T Cells: Co-expression of LAG3 and CD49b in IL-10-Producing T Cell Lineages. Front. Immunol. 9:2625. doi: 10.3389/fimmu.2018.02625 Type 1 regulatory CD4<sup>+</sup> T (Tr1) cells express high levels of the immunosuppressive cytokine IL-10 but not the master transcription factor Foxp3, and can suppress inflammation and promote immune tolerance. In order to identify and obtain viable Tr1 cells for research and clinical applications, co-expression of CD49b and LAG3 has been proposed as a unique surface signature for both human and mouse Tr1 cells. However, recent studies have revealed that this pattern of co-expression is dependent on the stimulating conditions and the differentiation stage of the CD4<sup>+</sup> T cells. Here, using an IL-10GFP/Foxp3RFP dual reporter transgenic murine model, we demonstrate that co-expression of CD49b and LAG3 is not restricted to the Foxp3<sup>−</sup> Tr1 cells, but is also observed in Foxp3<sup>+</sup> T regulatory (Treg) cells and CD8<sup>+</sup> T cells that produce IL-10. Our data indicate that IL-10-producing Tr1 cells, Treg cells and CD8<sup>+</sup> T cells are all capable of co-expressing LAG3 and CD49b in vitro following differentiation under IL-10-inducing conditions, and in vivo following pathogenic insult or infection in the pulmonary mucosa. Our findings urge caution in the use of LAG3/CD49b co-expression as sole markers to identify Tr1 cells, since it may mark IL-10-producing T cell lineages more broadly, including the Foxp3<sup>−</sup> Tr1 cells, Foxp3<sup>+</sup> Treg cells, and CD8<sup>+</sup> T cells.

Keywords: IL-10, Foxp3, T cell, parasitic infection, house dust mite, influenza infection, farmer's lung disease, lung inflammation

### INTRODUCTION

The mammalian immune system has evolved both effector and regulatory immune axes to protect the host from invading pathogens, along with a control mechanism to tune the level of immune reactivity against self- and non-self- agents to prevent host tissue damage. Interleukin-10 (IL-10) is a regulatory cytokine with a demonstrated anti-inflammatory function and plays an essential role in preventing allergic inflammation (1), autoimmunity (2), and pathogen-induced immunopathology (3, 4), but can also promote the establishment and maintenance of chronic infection (5, 6). IL-10 has been reported as a product of activation of multiple immune cell lineages. Innate immune cells including dendritic cells (DCs) (7), macrophages (8), neutrophils (9), and innate lymphoid cells (ILCs) (10) have been reported to express IL-10 in vivo and in vitro. IL-10 is also expressed by many cell subsets of the adaptive immune system, including B cells (11) and T cells comprising the Foxp3<sup>−</sup> CD4<sup>+</sup> (12), Foxp3<sup>+</sup> Treg (13), and CD8<sup>+</sup> T cell subsets (14). Regulatory T cells are defined by their immunosuppressive function, and the three aforementioned subsets of IL-10-producing T cells have been reported as phenotypically distinct regulatory T cell subsets, playing important roles in promoting immune tolerance and/or suppressing inflammation in both mouse and human (15–20).

Among the IL-10-producing T cells, the Foxp3<sup>−</sup> CD4<sup>+</sup> T cell subset, also known as type 1 regulatory T cells (Tr1 cells), are inducible in the periphery and have a pivotal role in limiting inflammation (15, 21–23). Tr1 cells have been shown to prevent allergic asthma (24) and atopic dermatitis (25) in murine models. In both mouse models and humans, induction of tolerance via specific antigen immunotherapy (SIT) is accompanied by induction of Tr1 cells (26, 27). Therefore, Tr1 cells have strong promise as a potential therapeutic approach for inflammatory diseases. Tr1 cells can be differentiated from naïve CD4<sup>+</sup> T cells upon TCR engagement in the presence of IL-27 in vitro (28), and in order to identify and obtain viable Tr1 cells for clinical application, co-expression of LAG3 and CD49b has been recently proposed to be a cell surface signature of the Foxp3<sup>−</sup> IL-10high Tr1 cells (15). LAG3 is a structural homolog of the CD4 molecule and can bind to MHC class II with high affinity (29, 30). LAG3 is highly expressed by IL-10+CD4<sup>+</sup> T cells (31), as well as by activated effector T cells (32) and Foxp3<sup>+</sup> Treg cells (33). CD49b is the α2 integrin subunit, highly expressed by NK cells (34). CD49b is up-regulated in T cells that may produce IL-10 and/or pro-inflammatory cytokines (35–37). In addition to Foxp3<sup>−</sup> Tr1 cells, IL-10 can be highly up-regulated in activated Foxp3<sup>+</sup> Treg and CD8<sup>+</sup> T cells under inflammatory conditions and/or upon TCR activation. Given the importance of being able to identify Foxp3<sup>−</sup> Tr1 cells, including under clinical conditions, and to gain a better understanding of the selectivity of co-expression of LAG3 and CD49b as a cell surface signature for IL-10-producing cells, we sought to determine whether co-expression of LAG3 and CD49b can mark a broader range of T cell subsets that are actively producing high levels of IL-10.

Using a murine model carrying an IL-10GFP/Foxp3RFP dual reporter system, we find that co-expression of LAG3 and CD49b is a generic feature of the IL-10-producing Foxp3<sup>−</sup> CD4+, Foxp3<sup>+</sup> CD4+, and CD8<sup>+</sup> T cell subsets. The capacity of co-expression of LAG3 and CD49b in marking IL-10high T cell subsets is dependent on the disease conditions and anatomical location of the cells. Furthermore, co-expression of LAG3 and CD49b is also a shared feature of human IL-10-producing FOXP3<sup>−</sup> CD4+, FOXP3<sup>+</sup> CD4+, and CD8<sup>+</sup> T cell subsets. Our data reveal that co-expression of LAG3 and CD49b is a generic signature of IL-10-producing T cells, which is broader than previously appreciated.

# MATERIALS AND METHODS

### Mice and Human Blood Samples

All mice were on the C57BL/6 background. Rag1−/<sup>−</sup> (B6.129S7- Rag1tm1Mom/J), IL-10GFP (B6(Cg)-Il10tm1.1Karp/J) (38), and Foxp3RFP (C57BL/6-Foxp3tm1Flv/J) (39) reporter mice were from the Jackson Laboratory (Bar Harbor, ME). Single reporter strains were crossed to generate an IL-10GFP/Foxp3RFP dual reporter strain as we recently reported (40). Human peripheral blood samples were procured from the New York Blood Center collected from healthy cohorts. All experiments were approved by the Office of Research Protection's Institutional Animal Care and Use Committee and Institutional Review Board at Cornell University.

### Antibodies and Other Reagents

All fluorescent antibodies are listed in "fluorochrome-target (clone; annotation if desirable)" format below.

### Mouse Antibodies

Purified anti-CD16/32 (93; Fc block), CD3ε (145-2C11), CD28 (37.51), IFN-γ (XMG1.2), and IL-12 (C17.8) antibodies were from BioLegend (San Diego, CA); Pacific Blue-CD90 (53- 2.1), FITC-TCRβ (H57-597), APC-LAG3 (C9B7W), PE-Cy7- CD49b (HMα2), and PE-Cy7-CD62L (MEL-14) were from BioLegend; eFluor 450-CD4 (GK1.5); Alexa Fluor 700-CD4 (GK1.5) were from eBioscience; BD Horizon V500-CD44 (IM7), PE-CD44 (IM7), and APC-Cy7- TCRβ (H57-597) were from BD Biosciences; PerCP-Cy5.5-CD8α (2.43) was from Tonbo Biosciences.

### Human Antibodies

Purified anti-CD3ε (OTK3) and CD28 (28.2), eFluor 450-CD8α (RPA-T8) FITC-CD4 (OKT4), and APC-FOXP3 (236A/E7) were from eBioscience; PE-IL-10 (JES3-19F1), Alexa Fluor 647- LAG3 (11C3C65), and PerCP-Cy5.5-LAG3 (11C3C65) were from BioLegend; FITC-CD49b (AK-7) and Alexa Fluor 700-CD4 (RPA-T4) were from BD Biosciences.

### Other Reagents

Human TruStain FcX (Fc receptor blocking solution) was from Biolegend; cell fixable viability dye eFluor 506 was from eBiosciences.

# Cell Isolation From Various Organs

Cells from various organs were isolated as we recently described (40). Briefly: blood cells were collected through cardiac puncture, and red blood cells were lysed before analysis; lungs were minced and digested in 0.2 mg/ml Liberase TL (Sigma, St. Luis, MO) in 37◦C for 15–30 min, then filtered and red blood cells were lysed before analysis; intestines were flushed, opened longitudinally, and inner contents removed with the blunt end of scissors, then cut into 0.5-cm fragments, followed by digestion in 100 U/ml collagenase VIII (Sigma) in 37◦C for 1 h, filtered, and lymphocytes isolated using gradient separation by 40% and 80% Percoll (GE Healthcare, Wilkes-Barre, PA) solutions; perigonadal adipose tissues were minced and digested in 500 U/ml collagenase I (Worthington Biochemical Corp., Lakewood, NJ) in 37◦C for

**Abbreviations:** Nb, Nippostrongylus brasiliensis; Tr1 cell, Type 1 regulatory T (CD4+TCRβ <sup>+</sup>Foxp3−IL-10+) cell; Treg cell, Foxp3-expressing regulatory T cell; HDM, house dust mite; SR, Saccharopolyspora rectivirgula; WSN, Influenza A/WSN/1933 (H1N1).

30 min, filtered and red blood cells were lysed before analysis. 50– 150 U/ml DNase I (Sigma) were added during digestion to reduce cell death triggered by free DNA.

## In vivo Induction of IL-10-Producing T Cells by TCR Activation

Foxp3RFPIL-10GFP dual reporter mice were injected with 15 µg/mouse anti-CD3ε (145-2C11) intraperitoneally on day 0 and 2, and analyzed on day 4, as previously described (23).

## Nippostrongylus brasiliensis (Nb) Infection

Mice were given 500 L3 Nb larvae per mouse through subcutaneous injection, as we previously described (40). Cells from the lungs were analyzed 7 days post infection (7 dpi).

### House Dust Mite (HDM)-Induced Allergic Disease Model

Mice were given daily intranasal exposures of 10 µg house dust mite (Dermatophagoides pteronyssinus) protein extract (XPB82D3A2.5 from Greer) in PBS, for 10 consecutive days. Cells from the lungs were analyzed 24 h post the last treatment.

### Farmer's Lung Disease (Hypersensitivity Pneumonitis) Model

Mice were intranasally exposed to 150 µg Saccharopolyspora rectivirgula (SR, ATCC 29034) extract on 3 consecutive days each week as previously described (41), for 4 weeks. Cells from the lungs were analyzed on the last day of the fourth week.

### Influenza A/WSN/1933 (WSN) Infection

Mice were intranasally infected with 1 LD<sup>50</sup> (10<sup>4</sup> PFU) WSN per mouse, as we previously described (40). Cells from the lungs were analyzed 7 days post infection (7 dpi).

### Differentiation of IL-10-Producing T Cells Mouse

TCRβ <sup>+</sup>Foxp3RFP<sup>−</sup> CD44−CD62L<sup>+</sup> splenic naïve T cells were sorted on BD FACS Aria II or Fusion systems (BD Biosciences, San Jose, CA), then cultured with Mitomycin-C (Sigma, 50µg/ml) treated antigen-presenting cells (APCs; Rag−/<sup>−</sup> splenocytes) at 1:2 ratio in the presence of anti-CD3ε (1µg/mL), anti-CD28 (1µg/mL), recombinant murine (rm) IL-27 (R&D Systems, 20 - 25 ng/ml), anti-IFN-γ and anti-IL-12 (10µg/mL) for 3 days.

### Human

Human peripheral blood mononuclear cells (PBMCs) were isolated from blood (New York Blood Center, Long Island, NY) using gradient separation in Ficoll-Paque PLUS (GE Healthcare). PBMCs were cultured in full RPMI-1640 medium for 30 min in 37◦C, then non-adherent cells were used to enrich for CD4<sup>+</sup> T cells using anti-human CD4 microbeads (Miltenyl Biotec, San Diego, CA) or CD8<sup>+</sup> T cells using a human CD8 isolation kit (BioLegend). Adherent cells were treated with Mitomycin-C (Sigma, 50µg/ml) in 37◦C for 30 min and used as APCs. Anti-human CD3ε(1µg/ml) and CD28 (CD28.2, eBioscience, 1– 3µg/ml), recombinant human (rh) IL-2 (PeproTech, 200 U/ml), IL-10 (PeproTech, 100 U/ml), IL-27 (R&D System, 25 ng/ml), and IFN-α2b (R&D System, 10 ng/ml) were added to differentiate human IL-10-producing T cells. Three days after cultures were set up, cells were stimulated with PMA (100 ng/ml, Sigma-Aldrich), Ionomycin (0.5µM, Sigma), Brefeldin A (5µg/ml), and GolgiStop (0.5 µl/ml, BD Biosciences) for 4 h as we previously described (42), and subjected to surface staining and intracellular staining (see details below).

# Flow Cytometry

Surface staining of live cells were done in the presence of Fc block and fixable viability dye. To detect human CD4 in activated T cells, anti-CD4 antibody was added into the intracellular staining panel. For intracellular cytokine staining, cells were fixed with 2% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA), permeabilized and stained with anti-cytokine antibodies in PBS/0.2% saponin (Sigma). Staining for human transcription factor FOXP3 was performed with a Foxp3 staining buffer kit (eBioscience). Flow cytometry data were acquired on LSRII, FACS Aria II or Fusion systems (BD Biosciences), and analyzed in FlowJo (Tree Star, Ashland, OR). All analyses were performed on fixable viability dye negative singlet population.

## Statistical Analysis

Non-parametric Mann-Whitney tests and one-way ANOVA were performed using GraphPad Prism v5.00 (GraphPad, San Diego, CA), with p ≤ 0.05 considered statistically significant. "NS" refers to "No Significance."

# RESULTS

### Co-expression of LAG3 and CD49b Marks Both IL-10-Producing CD4<sup>+</sup> and CD8<sup>+</sup> T Cells

LAG3 and CD49b co-expression was previously reported to be a cell surface signature for both mouse and human IL-10-producing CD4<sup>+</sup> T cells that lack the expression of Foxp3 (also known as type 1 regulatory T cells, Tr1 cells) (15). We and others have previously reported that co-culturing murine naïve CD4<sup>+</sup> T cells with antigen presenting cells (APCs) in the presence of anti-CD3, anti-CD28, anti-IFN-γ, anti-IL-12, and IL-27 can efficiently induce the differentiation of Tr1 cells (28, 40, 43), which express high levels of LAG3 and CD49b. Our recent data also demonstrated that this protocol can induce IL-10 production in bulk T cell populations that include both CD4<sup>+</sup> and CD8<sup>+</sup> T cells (**Figure 1A**). Surprisingly, the resultant IL-10-producing CD8<sup>+</sup> T cells induced in vitro through this protocol also exhibited high levels of LAG3/CD49b co-expression (**Figure 1A**, last plot). In fact, IL-10-producing CD8<sup>+</sup> T cells can express higher levels of LAG3 and CD49b than their IL-10 producing CD4<sup>+</sup> counterparts induced in the same cell culture (**Figure 1B**). These data suggest that co-expression of LAG3 and CD49b is not an exclusive cell surface signature of the Tr1 cells, and may be a shared feature of both IL-10-producing CD4<sup>+</sup> and CD8<sup>+</sup> T cells.

FIGURE 1 | IL-10-producing LAG3+CD49b<sup>+</sup> T cells include both CD4<sup>+</sup> and CD8<sup>+</sup> subsets. All experiments were performed with cells carrying the IL-10GFP reporter system for live cell analysis. Naïve T cells were cultured under IL-10-inducing conditions for 3 days. (A) Gating strategy to identify IL-10-expressing CD4<sup>+</sup> and CD8<sup>+</sup> T cells, and representative FACS plots showing the co-expression of LAG3/CD49b by IL-10<sup>+</sup> CD4<sup>+</sup> or IL-10<sup>+</sup> CD8<sup>+</sup> T cells. Gray backgrounds show naïve T cells as negative control for LAG3/CD49b quadrant gating. (B) Summary of percentage of LAG3/CD49b double positive population and geometric mean fluorescence intensity (gMFI) of LAG3 and CD49b in IL-10-producing CD4<sup>+</sup> and CD8<sup>+</sup> T cells. N = 4. Data represent results of more than three experiments. \*p ≤ 0.05, by non-parametric Mann-Whitney test. Data presented as Mean ± S.E.M.

# Co-expression of LAG3 and CD49b Marks Both IL-10-Producing Tr1 and Treg Cells

IL-10 production can be significantly elevated in the pulmonary mucosa during the late stages of parasitic infection by Nippostrongylus brasiliensis (Nb), predominantly by CD4<sup>+</sup> T cells that are LAG3/CD49b double positive (15, 44). We recently observed that both Foxp3<sup>−</sup> and Foxp3<sup>+</sup> T cells are capable of producing IL-10 in the pulmonary tissues post Nb infection (40). To determine whether co-expression of LAG3 and CD49b is exclusive to Foxp3<sup>−</sup> Tr1 cell subset, we infected IL-10GFP/Foxp3RFP dual reporter mice with Nb, and analyzed the IL-10-producing T cells. We found that, as previously described, a large majority of the IL-10-producing T cells in the lungs of Nb-infected mice express high levels of both LAG3 and CD49b (**Figures 2A,B**), and are predominantly CD4<sup>+</sup> T cells (**Figure 2C**). However, interestingly, these IL-10-producing LAG3+CD49b<sup>+</sup> CD4<sup>+</sup> T cells included both Foxp3<sup>+</sup> and Foxp3<sup>−</sup> CD4<sup>+</sup> T cells subsets (**Figure 2C**, last plot; and **Figure 2D**). To further determine whether LAG3/CD49b co-expression is correlated with IL-10 and/or Foxp3 expression in CD4<sup>+</sup> T cells, we compared the percentage of LAG3/CD49b double positive populations in Foxp3−IL-10−, Foxp3−IL-10+, Foxp3+IL-10−, and Foxp3+IL-10<sup>+</sup> CD4<sup>+</sup> T cells isolated from the lungs of Nb-infected mice. We found that regardless of expression of Foxp3, Nb infection did not lead to significant up-regulation of LAG3/CD49b co-expression on IL-10<sup>−</sup> CD4<sup>+</sup> T cells (**Figures 2E,F**). However, the percentage of the LAG3+CD49b<sup>+</sup> population of IL-10<sup>+</sup> CD4<sup>+</sup> T cells is significantly higher than their IL-10<sup>−</sup> counterparts (**Figure 2F**). Moreover, the levels of LAG3 and CD49b expression are similar between the Foxp3<sup>+</sup> and Foxp3<sup>−</sup> counterparts of IL-10<sup>+</sup> CD4<sup>+</sup> T cells (**Figure 2G**). Therefore, IL-10-producing CD4<sup>+</sup> T cells, regardless of Foxp3 expression, have a high capacity of coexpressing LAG3 and CD49b. Together with the data shown in **Figure 1**, these data suggest that co-expression of LAG3 and CD49b is a generic feature of IL-10-producing T cells, including Foxp3<sup>−</sup> Tr1 cells, Foxp3<sup>+</sup> Treg cells and CD8<sup>+</sup> T cells.

# The Composition of LAG3+CD49b<sup>+</sup> IL-10-Producing T Cells Differs in Different Disease Models

IL-10 plays an essential role in pulmonary inflammatory diseases, which has been reported in multiple murine models of lung disease, including allergic asthma (45), hypersensitivity pneumonitis (HP) (46) and influenza pneumonia (20). We examined whether the expression of LAG3 and CD49b would differ based on the inflammatory response in three mouse models of lung inflammation. Mice carrying the IL-10GFP/Foxp3RFP dual reporters were exposed intranasally to house dust mite (HDM) protein extract (as a model of allergic asthma), Saccharopolyspora rectivirgula (SR) (as a model of HP/farmer's lung disease), or infected intranasally with WSN/flu virus (as a model of influenza infection). We observed significant percentages of IL-10-producing T cells in the lung tissue of mice exposed to HDM (**Figure 3A**), SR (**Figure 3C**), or WSN/flu virus (**Figure 3E**). These IL-10-producing T cells all co-expressed high levels of LAG3 and CD49b, and include Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4−, and CD8<sup>+</sup> subsets in all disease models analyzed (**Figure 3**). However, the relative proportions of IL-10-producing Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4−, and CD8<sup>+</sup> subsets differed in the various disease models (**Figure 3**). In contrast to the composition of IL-10-producing LAG3+CD49b<sup>+</sup> T cells induced by Nb infection, in which the Foxp3<sup>−</sup> CD4<sup>+</sup> subset is the majority (72% in **Figure 2D**), in HDM-induced allergic asthma model, the largest subset of the IL-10-producing LAG3+CD49b<sup>+</sup> T cells in the lungs are Foxp3<sup>+</sup> CD4<sup>+</sup> T cells (60%), followed by Foxp3<sup>−</sup> CD4<sup>+</sup> T cells (32%), while CD8<sup>+</sup> T cells are only around 1.4% (**Figure 3B**). In the SR-triggered farmer's lung disease model, Foxp3<sup>+</sup> CD4<sup>+</sup> T cells are the majority of the IL-10 producing LAG3+CD49b<sup>+</sup> T cell subset in the lungs, however, there are similar percentages of Foxp3<sup>−</sup> CD4<sup>+</sup> and CD8<sup>+</sup> T cells (16% each) (**Figure 3D**). Strikingly in the murine model of influenza infection, CD8<sup>+</sup> T cells are the largest subset of the IL-10-producing LAG3+CD49b<sup>+</sup> T cells in the lungs (86%),

were infected with 500 L3 Nippostrongylus brasiliensis (Nb) (or PBS as control) and lungs analyzed 7 days post infection (dpi). (A) Gating strategy to identify live singlet lymphocytes from cells isolated from the lungs of mice for analyses. (B) Gating strategy to identify IL-10-producing T cells (with IL-10<sup>−</sup> T cells as control), and representative FACS plots for LAG3/CD49b co-expression in IL-10<sup>−</sup> vs. IL-10<sup>+</sup> T cells. (C) Representative FACS plots showing that the majority of IL-10<sup>+</sup> LAG3/CD49b co-expressing T cells in Nb-infected mouse lungs are CD4<sup>+</sup> T cells, including both Foxp3<sup>+</sup> Treg and Foxp3<sup>−</sup> Tr1 cells. (D) Pie chart summarizing the proportion of Foxp3<sup>+</sup> vs. Foxp3<sup>−</sup> IL-10-producing LAG3/CD49b double positive CD4<sup>+</sup> T cells. (E) Gating strategy identifying CD4<sup>+</sup> T cells that are Foxp3−IL-10−, Foxp3−IL-10+, Foxp3+IL-10<sup>−</sup> and Foxp3+IL-10+. (F) Summary of percentage of LAG3/CD49b double positive population of Foxp3−IL-10−, Foxp3−IL-10+, Foxp3+IL-10−, and Foxp3+IL-10<sup>+</sup> CD4<sup>+</sup> T cells. (G) Summary of gMFI of LAG3 and CD49b in Foxp3<sup>+</sup> vs. Foxp3<sup>−</sup> IL-10-producing CD4<sup>+</sup> T cells. N = 4. Data represent results of three experiments. \*\*\*p ≤ 0.001, NS = No significance, by non-parametric Mann–Whitney test. Data presented as Mean ± S.E.M.

followed by Foxp3<sup>−</sup> CD4<sup>+</sup> T cells, while Foxp3<sup>+</sup> CD4<sup>+</sup> are a minority (**Figure 3F**). These data compared the composition of IL-10-producing LAG3+CD49b<sup>+</sup> T cells in various murine models of pulmonary inflammatory diseases. Along with the model of parasitic infection shown in **Figure 2**, our data suggest that co-expression of LAG3 and CD49b marks all IL-10-producing T cell lineages in the pulmonary system, and relative abundance of the marked T cell subsets is dependent on the type of immune response as shown in the disease models.

## The Composition of LAG3+CD49b<sup>+</sup> IL-10-Producing T Cells Differs in Different Organs

As discussed above, we demonstrated that co-expression of LAG3 and CD49b is a generic feature of IL-10-producing T cells in vivo in pulmonary tissues under multiple inflammatory conditions (**Figures 2**, **3**). To determine whether this feature is applicable to IL-10-producing cells in other organs, we injected IL-10GFP/Foxp3RFP dual reporter mice with an anti-CD3ε antibody that has been shown to stimulate pronounced IL-10 production by T cells through TCR activation in vivo (23, 40). We analyzed IL-10-producing T cells in blood, lymph nodes (LN), lung, fat and small intestine, and found that coexpression of LAG3 and CD49b marked a portion of the IL-10 producing T cells following TCR activation in vivo, which again included Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4+, and CD8<sup>+</sup> T cell subsets (**Figure 4A**) in all organs analyzed. An interesting note is that the relative abundance of Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4+, and CD8<sup>+</sup> subsets among the IL-10-producing LAG3+CD49b<sup>+</sup> T cells vary significantly in different organs of the same mice (**Figure 4B**). Upon TCR activation in vivo, Foxp3<sup>+</sup> CD4<sup>+</sup> T cells are the major population that are IL-10+LAG3+CD49b<sup>+</sup> in the blood, lymph nodes and lungs, while CD8<sup>+</sup> T cells are the majority of IL-10+LAG3+CD49b<sup>+</sup> T cells in the perigonadal fat and small intestine (**Figure 4B**). These data suggest that co-expression of LAG3/CD49b marks all three IL-10-producing T cell subsets in

multiple organs and the relative abundance of the Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4<sup>+</sup> and CD8<sup>+</sup> T cell subsets in IL-10-producing LAG3+CD49b<sup>+</sup> T cells is dependent on the anatomical location of the cells.

# Human IL-10-Producing CD4<sup>+</sup> and CD8<sup>+</sup> T Cells Exhibit a LAG3+CD49b<sup>+</sup> Phenotype

In murine models, we demonstrated that co-expression of LAG3 and CD49b marks IL-10-producing Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4+, and CD8<sup>+</sup> T cells under different inflammatory conditions in the lungs (**Figures 2**, **3**), as well as in different organs when TCR is activated in vivo (**Figure 4**). To further determine whether different subsets of IL-10-producing T cells exhibit the shared feature of co-expression of LAG3 and CD49b in humans, we isolated CD4<sup>+</sup> and CD8<sup>+</sup> T cells from human peripheral blood and cultured them under IL-10-inducing conditions. We found that as in the mouse, human IL-10 producing FOXP3<sup>+</sup> CD4+, FOXP3<sup>+</sup> CD4+, and CD8<sup>+</sup> subsets all up-regulated both LAG3 and CD49b expression, with a significant LAG3/CD49b double positive population (**Figure 5**).

### DISCUSSION

Our data presented in this report demonstrated that, as previously reported (15), co-expression of LAG3 and CD49b identifies Foxp3<sup>−</sup> IL-10high Tr1 cells, but that these markers are not exclusive for Tr1 cells in human and mouse. Furthermore, we find that the IL-10-producing LAG3+CD49b<sup>+</sup> T cell population is composed of Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4+, and CD8<sup>+</sup> T cell subsets, and this composition varies depending on the disease conditions and anatomical locations of the cells. The importance of this work is emphasized by the need to specially identify Foxp3<sup>−</sup> Tr1 cells, especially under clinical conditions, and our findings suggest that the use of LAG3 and CD49b should be combined with other markers to uniquely identify these cells.

Using Tr1 cell clones derived from human naïve CD4<sup>+</sup> T cells and purified by an IL-10 secretion assay, Gagliani, Roncarolo and colleagues identified co-expression of LAG3, CD49b, and CD226 as cell a surface signature of IL-10-producing CD4<sup>+</sup> T cells, and demonstrated that co-expression of LAG3 and CD49b is sufficient to distinguish Foxp3<sup>−</sup> IL-10high Tr1 cells from T helper and/or regulatory subsets that expressed lower levels of or no IL-10 in both mouse and human (15). However, it was recently reported that IL-10-producing T cells derived from human CD4<sup>+</sup> memory T cells exhibit low levels of surface expression of LAG3 and CD49b (47), suggesting that the pattern of co-expression of LAG3 and CD49b may vary in human IL-10-producing CD4<sup>+</sup> T cells, which may be associated with whether they were derived from naïve precursors vs. memory cells. Indeed, it is clear that co-expression of LAG3 and CD49b can mark IL-10-producing Tr1-like cells, and serve to help eliminate those T cells lineages that are not capable of producing IL-10 for potential clinical interest. However, other non-Tr1-like IL-10-producing T cells can also co-express these markers. Whether co-expression of LAG3 and CD49b allows efficient recovery of IL-10high T cells

may be dependent on the proportion of IL-10high T cells that are co-expressing LAG3 and CD49b, which may differ depending on whether the relevant IL-10-producing cells had different origins and/or underwent different activation regimes.

Both LAG3 and CD49b can individually be up-regulated in activated T cells, regardless of their production of the anti-inflammatory IL-10 or pro-inflammatory cytokines (31–33, 35–37). Co-expression of LAG3 and CD49b is more restricted to IL-10-producing subsets, as previously described in Foxp3<sup>−</sup> CD4<sup>+</sup> T cells (15). Our data here demonstrated that this is a generic feature of IL-10-producing cells, including Foxp3<sup>−</sup> CD4+, Foxp3<sup>+</sup> CD4+, and CD8<sup>+</sup> subsets. The cooccurrence of IL-10 production and LAG3+CD49b<sup>+</sup> may be explained in two ways. The first explanation is that IL-10

stimulation through autocrine and/or paracrine may induce the expression of LAG3 and CD49b in the stimulated cells, therefore, IL-10-capturing T cells [cells detected by Roncarolo's group in ref (15)] exhibited high levels of LAG3/CD49b coexpression. This hypothesis would place IL-10 up-stream of LAG3 and CD49b expression, which is unlikely, as recent studies by Flavell and Huber's groups showed that the IL-10 receptor is dispensable for Tr1 cell differentiation, including the LAG3/CD49b double positive feature; instead, IL-10 receptor signaling is critical for maintaining the cell fate commitment and functional performance of the differentiated Tr1 cells (48). Therefore, it is more likely that LAG3 and CD49b signaling pathways function cooperatively to activate the expression of IL-10. This hypothesis is more reasonable, given our data that LAG3 and CD49b co-expression is a generic feature of IL-10 producing cells in multiple subsets. LAG3 is a structural homolog of the CD4 molecule, and can bind to MHC class II with higher affinity than CD4 (29, 30). In unstimulated T cells, LAG3 is retained in the intracellular compartment and degraded in the lysosome. Upon T cell activation, LAG3 traffics from the lysosomal compartment to the cell surface through a protein kinase C (PKC) dependent pathway (49). CD49b is the integrin α2 subunit and plays a critical role in cell-cell interaction and adhesion (50). The CD49b pathway activates multiple downstream effector signaling pathways, among which is the RAS/MAPK signaling cascade (51, 52). We recently reported that the RAS/MAPK signaling pathway functioning downstream of the TCR is indispensable for IL-10 production by Foxp3<sup>−</sup> CD4<sup>+</sup> cells, through activation of the expression of the transcription factor interferon regulatory factor 4 (IRF4) (40). PKC can regulate RAS signaling to downstream effectors, and can activate MAPK signaling in both RAS-dependent and independent manners (53, 54). The interplay between PKC and RAS may regulate signals that connect LAG3 and CD49b downstream pathways, leading to up-regulation of IL-10. Given the complexity of these pathways, comprehensive understanding of the mechanism(s) underlying the co-expression of these markers with IL-10 will require significantly more in-depth analyses. Regardless of the mechanism, our findings indicate that LAG3/CD49b co-expression does not uniquely identify Foxp3<sup>−</sup> Tr1 cells, but is a more general indicator of IL-10 production in T cell lineages.

Despite the shared feature of co-expression of LAG3 and CD49b by IL-10-producing Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4+, and CD8<sup>+</sup> T cells, we also observed interesting discrepancies in the proportional composition of these three IL-10high T cell subsets that are all LAG3/CD49b double positive in the lung mucosa of different pulmonary inflammatory disease models, as well as in different anatomical locations in the same mice upon TCR activation in vivo. For example, parasite infection (Nb) and mite allergen (HDM) induced predominantly IL-10 producing LAG3+CD49b<sup>+</sup> T cells that are CD4<sup>+</sup> with very few that are CD8+. In the Nb infection model, the majority of IL-10<sup>+</sup> LAG3+CD49b<sup>+</sup> CD4<sup>+</sup> T cells are Foxp3<sup>−</sup> Tr1 cells, while in HDM-exposed mice, Foxp3<sup>+</sup> Treg cells are the majority. In the bacterial exposure (SR) and viral infection (Flu) models, we observed an increased proportion of IL-10<sup>+</sup> LAG3+CD49b<sup>+</sup> T cells that are CD8+, which is the predominant population in the Flu model (**Figure 3F**). This discrepancy in the composition of IL-10-producing LAG3+CD49b<sup>+</sup> T cells may be due to the difference of the microenvironment in which the IL-10 producing cells are being induced. Factors that may affect the different composition of the IL-10high LAG3+CD49b<sup>+</sup> T cells may include the type of immune response, abundance and affinity of the TCR ligands, the cytokines induced and orchestrated by the stimuli, and the nutritional microenvironments that favor different subsets of the T cells. For example, type I IFN can facilitate the preferential induction of IL-10-producing effector CD8<sup>+</sup> T cells through inducing and sustaining expression of the IRF4 and Blimp1 transcription factors (55). Type I interferon is significantly elevated during influenza infection (56) but much less so by HDM exposure (57). Another possible orchestrator could be the cytokine IL-27, which has been reported to be directly required for IL-10 induction in CD8<sup>+</sup> T cells (58) and CD4+Foxp3<sup>−</sup> (59) but not CD4+Foxp3<sup>+</sup> T cell subsets (60). Transcription factors and their interacting molecular networks that exhibit differential functions for IL-10 induction in different T cell lineages might provide an answer as well. For example, Blimp-1 is indispensable for IL-10 induction in Foxp3<sup>+</sup> and Foxp3<sup>−</sup> CD4<sup>+</sup> T cells (61, 62), as well as in CD8<sup>+</sup> T cells (63). AhR interaction with cMAF is critical in IL-10 induction in Foxp3<sup>−</sup> Tr1 cells in response to an IL-27-supplemented environment, but is insufficient in inducing IL-10 production in Foxp3<sup>+</sup> Treg cells under the same conditions (43); AhR expression was not critical in IL-10-producing CD8<sup>+</sup> T cells either (64), suggesting that AhR signaling has a multifaceted function in regulating the level of IL-10 expression in different T cell lineages. A more comprehensive understanding of the T cellintrinsic molecular features that are shared or distinct among the IL-10-producing CD8+, Foxp3−CD4<sup>+</sup> and Foxp3+CD4<sup>+</sup> T cells awaits further investigation.

### REFERENCES


Our data reported here demonstrate that co-expression of LAG3 and CD49b marks IL-10high T cell subsets that are Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4+, or CD8<sup>+</sup> in both human and mouse, and thus does not uniquely identify Foxp3<sup>−</sup> Tr1 cells. However, this finding does not negate the feasibility of utilizing co-expression of LAG3 and CD49b in marking a broader range of immunosuppressive IL-10high T cell populations that have potential therapeutic effects for clinical application. Further investigations are required to determine the levels of regulatory and pro-inflammatory cytokine production in the bulk LAG3+CD49b<sup>+</sup> T cells, and to determine and compare the ability of the Foxp3<sup>+</sup> CD4+, Foxp3<sup>−</sup> CD4+, and CD8<sup>+</sup> subsets of the LAG3+CD49b<sup>+</sup> T cells in suppressing effector immunity and inflammation in vivo.

### AUTHOR CONTRIBUTIONS

WH and AA conceived research, designed experiments, analyzed and interpreted data, and wrote the manuscript. WH, SS, and CC performed experiments; S-GZ contributed reagents and intellectual input.

### ACKNOWLEDGMENTS

We thank A. Redko for animal care and L. Zhang for technical assistance; Dr. E. Tait Wojno for N. brasiliensis; and Drs. D. Topham, G. Whittaker, and M. Straus for influenza A virus. This work was supported in part by grants from the National Institutes of Health (AI120701, AI138570 and AI126814 to AA; AI129422 and AI138497 to AA and WH, and AI137822 to WH), a Careers in Immunology Fellowship from the American Association of Immunologists (to WH), a Pilot Grant from the Center for Experimental Infectious Disease Research (funded by NIH P30GM110760), the Faculty Development Program and an award from Competitive Research Programs of the Louisiana State University (to WH).


binding site on LAG-3 protein. Proc Natl Acad Sci USA. (1997) 94:5744–9. doi: 10.1073/pnas.94.11.5744


differentiation of IL-10-producing Tr1 cells. J Immunol. (2009) 183:797–801. doi: 10.4049/jimmunol.0901233


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

Copyright © 2018 Huang, Solouki, Carter, Zheng and August. 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.

# Metabolic Checkpoints in Differentiation of Helper T Cells in Tissue Inflammation

Suyasha Roy, Zaigham Abbas Rizvi and Amit Awasthi\*

Immuno-Biology Lab, Translational Health Science and Technology Institute, Faridabad, India

Naïve CD4<sup>+</sup> T cell differentiate into effector and regulatory subsets of helper T (Th) cells in various pathophysiological conditions and modulate tissue inflammation in autoimmune diseases. While cytokines play a key role in determining the fate of Th cells differentiation, metabolites, and metabolic pathways profoundly influence Th cells fate and their functions. Emerging literature suggests that interplay between metabolic pathways and cytokines potentiates T cell differentiation and functions in tissue inflammation in autoimmune diseases. Metabolic pathways, which are essential for the differentiation and functions of Th cell subsets, are regulated by cytokines, nutrients, growth factors, local oxygen levels, co-activation receptors, and metabolites. Dysregulation of metabolic pathways not only alters metabolic regulators in Th cells but also affect the outcome of tissue inflammation in autoimmune and allergic diseases. Understanding the modulation of metabolic pathways during T cells differentiation may potentially lead to a therapeutic strategy for immune-modulation of autoimmune and allergic diseases. In this review, we summarize the role of metabolic checkpoints and their crosstalk with different master transcription factors and signaling molecules in differentiation and function of Th subsets, which may potentially unravel novel therapeutic interventions for tissue inflammation and autoimmune disorders.

### Edited by:

Loretta Tuosto, Sapienza University of Rome, Italy

### Reviewed by:

Lawrence Kane, University of Pittsburgh, United States Silvia Deaglio, Università degli Studi di Torino, Italy

> \*Correspondence: Amit Awasthi aawasthi@thsti.res.in

### Specialty section:

This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology

Received: 25 September 2018 Accepted: 07 December 2018 Published: 14 January 2019

### Citation:

Roy S, Rizvi ZA and Awasthi A (2019) Metabolic Checkpoints in Differentiation of Helper T Cells in Tissue Inflammation. Front. Immunol. 9:3036. doi: 10.3389/fimmu.2018.03036 Keywords: T cell, cytokines, inflammation, metabolism, checkpoint, transcription factor

# INTRODUCTION

Nutrients, water, and oxygen are the fundamental constituents that are required for all the living cells, even more so for the cells of the immune system, which are metabolically hyper active during immunological reactions (1). In general, the energy requirement of resting naïve T cells is fulfilled by aerobic metabolism of glucose via oxidative phosphorylation. However, upon antigenic stimulation, naïve T cells get activated, and metabolically shift toward aerobic glycolysis (1, 2). The nutrients in the form of metabolites are not only essential for maintaining T cell homeostasis but are also essential for generating precursors of bio-molecules, which support rapid proliferation of activated T cells essential for their functions. The metabolic changes occurring within T cells influence the fate of diverse immune responses during infection, inflammation, and autoimmunity (**Figure 1**). In order to control hyper activation of T cells in an immune reaction, various regulatory mechanisms have been identified. Metabolic checkpoints were identified as one of the key regulator of T cell responses (3). Classically, the activity of metabolic enzyme or concentration of a specific metabolite was suggested to be important checkpoint in immune response in infection and inflammation in autoimmunity. However, emerging literature indicates that metabolic checkpoint

are characterized by increased glucose transporter 1 (GLUT 1) expression needed for higher influx of Glucose. The influxed glucose is then catabolized through increased rate of glycolysis. The intermediate products of glycolysis are utilized for the generation of biosynthetic precursor through pentose phosphate pathway (PPP) and amino acid synthesis. Also, there is increased fatty acid oxidation (FAO) and in some cases increased Fatty acid synthesis (FAS). The memory T cells and the Tregs behave similar to the naïve T cells metabolically. They maintain a steady rate of glycolysis, TCA cycle, and oxidative phosphorylation (OXPHOS) producing ATP. The striking characteristic of memory T and Treg cells is high FAO and FAS.

could be an enzyme, metabolite, a signaling molecule, and/or even a transcription factor that could potentially regulate T cells differentiation and functions (**Figure 1**).

Metabolism is a dynamic process, which provides essential building blocks for diverse cellular processes and fulfills energy requirements of cells. In addition to maintaining essential functions of cells, metabolites regulate cellular differentiation, and functions (4). The metabolic reprogramming of an activated T cell is essential for its rapid proliferation and acquisition of effector functions (2). Published literature have clearly indicated the distinct metabolic requirements of undifferentiated naïve vs. activated and differentiated T cells, as undifferentiated naïve T cells predominantly rely upon ATP produced via oxidative phosphorylation and β-oxidation of fatty acids while activated/differentiated T cells met their energy demands primarily by glycolytic, pentose-phosphate, and glutaminolysis pathways (5).

Activated CD4<sup>+</sup> T cells proliferate and acquire distinct effector phenotypes such as Th1, Th2, Th9, and Th17 cells, which contribute to specialized functions in eliminating intra and extracellular pathogens as well as inducing tissue inflammation in autoimmunity and allergic inflammation (6). On contrary, regulatory subsets of CD4<sup>+</sup> T cells, which include Foxp3<sup>+</sup> regulatory T cells (Tregs) and type 1 regulatory T (Tr1) cells, suppress effector T cell functions and contribute to resolution of tissue inflammation in autoimmune diseases (6) (**Figure 2**). Each of these effector and regulatory subsets of T cells differentiate in the presence of specific differentiating cytokines, cell signaling modules, and unique sets of transcriptional network which regulates the distinct metabolic reprogramming in both the effector and regulatory T cells. For example, Foxp3<sup>+</sup> Tregs, and memory T cells primarily rely on lipid oxidation while effector T cells utilizes glycolytic and glutaminolytic pathway to support their survival (6). In this review, we discuss and summarize important metabolic checkpoints of T helper cell differentiation and function in immunity and autoimmunity. We further describe the current advances as to how metabolic reprogramming of T cells regulates their effector functions in an immune response.

### METABOLITES AND METABOLIC SIGNALS ACTS AS A CHECKPOINT IN T HELPER CELL DIFFERENTIATION

Emerging data in cancer biology has established that specific metabolites may act as signaling components of specific metabolic pathways, which affect the physiology of the cancer cells. Similar to cancer cells, T cells, upon antigenic stimulation, rapidly proliferate, suggesting a possibility of involvement of metabolites in T cells activation and proliferation. The metabolic components, ranging from

metabolites, enzyme complexes to signaling molecules, regulate T cell activation, differentiation, and functions in various patho-physiological conditions. Many of these metabolites could act as checkpoints during Th cell differentiation in tissue inflammation. Here we discuss some of the examples as to how metabolites influence T cells fate and their effector functions.

### Glucose

Glucose, the primary source of cellular energy, feeds into pathways that generate metabolites essential for cellular growth, and functions. Glucose provides instant energy for activated and proliferating T cells as well as metabolic precursors such as nucleic acids, proteins, lipids, and carbohydrates. The increased energy demand of rapidly dividing cancer cells is fulfilled by the upregulation of glycolysis (7), and this phenomenon, which was initially described in cancer cells, is known as "Warburg Effect" (8). Similar to tumor cells, rapidly proliferating activated T cells also rely on aerobic glycolysis, instead of oxidative phosphorylation, to support their cellular energy requirements as well as biosynthetic metabolic demands (8). A differentiating pro-inflammatory T cells shift to aerobic glycolysis, which lead to lactate production even in the abundance of oxygen (8, 9). The "Warburg Effect" in T cells begins, with the engagement of T cell receptor (TCR) with its ligand that leads to T cells activation, and triggers rapid glucose uptake via upregulation of glucose transporter, Glut1 (8, 9). In addition to TCR activation and costimulatory signals of CD28, which activate PI3K/AKT-mediated translocation of Glut1 from the intracellular vesicles to the cell surface of activated T cells (10). Upon translocation through its transporter, glucose is catabolized to provide the building blocks for cellular proliferation and generate energy via TCA cycle, OxPhos, or Warburg metabolism.

The critical function of glucose in T cell activation and differentiation was identified in a study where deprivation of glucose leads to impairment of T cell activation and survival (10). Interestingly, T cell survival and functions could not be rescued by an alternate energy source to glucose such as glutamine (10, 11). In addition to T cell survival, the effect of glucose deprivation was also found to be associated with cytokines production by T cells, as the absence of glucose in T cell cultures results in decreased IFN-γ production (10, 12). On contrary, transgenic expression of Glut1 in T cells induces enhanced production of IL-2 and IFN-γ, suggesting a critical role of glucose in T cell activation and functions (12). Consistently, it was demonstrated that the deficiency of Glut1 profoundly suppressed glucose metabolism that lead to a decrease in T cells differentiation and functions (12). Interestingly, in contrast to effector T cell subsets, Tregs functions remain intact in Glut1 deficiency (12). Taken together, these observations indicate that glucose metabolism is essential for effector T cell functions but not for Tregs functions. The role of glucose and its metabolism has also been implicated in T cell dependent function of B cell activation and antibody production (13).

Activated T cells prefer glycolysis over OxPhos despite glycolysis provides lesser ATP than Oxphos (14). The advantage of glycolysis over Oxphos is to provide biosynthetic precursors that are essential for the generation of amino acids, lipids and nucleic acids for rapidly proliferating T cells. The rate of glycolysis in T cell activation determine the outcome of the generation of effector and memory T cells repertoire. Further, it is known that metabolic status of Tr1 is similar to the effector T cells, however Tregs are known to adopt low glycolysis and high oxidative metabolism (14–16). Effector T cell populations such as Th1, Th2, Th9, and Th17 were shown to have enhanced glycolytic activity while memory T cells and Tregs rely primarily on fatty acid oxidation for their survival (5, 14, 15, 17). Consistently, blocking glycolysis using 2-DG, a prototypical inhibitor of hexokinase (first rate-limiting enzyme of glycolysis), was found to suppress the differentiation of both CD4<sup>+</sup> and CD8<sup>+</sup> T cells into their effector subsets, indicating the importance of glucose metabolism and glycolysis in T cells activation and functions (18). Mechanistically, glucose metabolism is primarily regulated by PI3K/AKT/mTOR pathway, signal transducer and activator of transcription 5 (STAT5), extracellular signal-related kinase (ERK) and Mitogen-Activated Protein Kinases (MAPKs) (19, 20). Consistently, high glucose level or hyperglycemia has long been associated with the onset of chronic inflammatory diseases such as rheumatoid arthritis (RA), multiple sclerosis (MS), as proteomic analysis of synovial fluid has identified the proteins critical for glycolytic pathways are overexpressed in RA patients (21). This observation is in accordance with upregulation of glycolytic flux in synovial lesions. In patients with such conditions, glucocorticoid therapy has helped conventionally to overcome inflammatory conditions associated with inflammatory T cells (21). In addition, recent studies with exogenous insulin treatment have also shown promising result in lowering the glucose level to normal (22).

# Lipids

Fatty acids composition, particularly of the cell membrane, regulates variety of physiological processes including cellular signaling, which ultimately leads to modulation of inflammatory response and T cells functions in autoimmune diseases (23). The role of glucose and glutamine in sustaining T cells survival and effector functions is well-understood, however little is known about fatty acids metabolism and their effect on regulating T cell differentiation and functions. Dietary fatty acids, omega 6, and omega 3 polyunsaturated fatty acids (PUFAs), are known to regulate inflammatory conditions, as omega 3 PUFAs act as anti-inflammatory mediators in inflammatory diseases such as psoriasis, RA, and IBD in mouse model of TNBS-induced ulcerative colitis (24).

Fatty acid metabolism remains an integral part of T cell differentiation in tissue inflammation, as T cell activation induces cellular lipid biosynthetic pathways that are linked to glucose metabolism (25). β-oxidation of fatty acids is essential for the generation and functions of Foxp3<sup>+</sup> iTregs, and is required for the membrane synthesis of proliferating effector T cell populations such as Th1, Th2, Th9, and Th17 cells (17, 26). Proliferating activated T cells augment fatty acid synthesis with a concomitant reduction in fatty acid oxidation. Fatty acid synthesis and fatty acid oxidation occur in cytosol and mitochondria, respectively. While fatty acid synthesis requires ATP consumption, fatty acid oxidation generates ATP via acetyl-CoA-TCA-OXPHOS cycle.

Fatty acids metabolic pathways are crucial in the interplay between Th17 and Foxp3<sup>+</sup> Tregs functions and their generations (26). The generation of Th17 and Foxp3<sup>+</sup> Tregs are reciprocally regulated, as TGF-β1 induces the generation of Foxp3<sup>+</sup> Tregs cell while IL-6 together with TGF-β1 not only suppresses the generation of Tregs but induces the differentiation of Th17 cells (27). It was demonstrated that Fatty acids contribute in generation and functions of Tregs cells while their intracellular accumulation was found to modulate the pathogenicity of Th17 cells.

Acetyl-CoA carboxylase 1 (ACC1) plays a critical role in fatty acid synthesis by converting acetyl-CoA to malonyl-CoA (26). ACC1-mediated fatty acid synthesis is crucial for the generation of Th17 cells, as genetic deficiency of ACC1 leads to defective Th17 cell response (26). In contrast, ACC1 deficiency induces the generation of Foxp3<sup>+</sup> Tregs cells (26). Consistent with these findings, adoptive transfer of ACC1-deficient T cells in mouse model of GVHD reduced mortality associated with enhanced frequency of Foxp3<sup>+</sup> Tregs cells, suggesting that fatty acid regulates effector T cells while enhances Tregs functions. Furthermore, oxidation of fatty acid is essential for the generation of iTregs via upregulation of AMPK, which is further shown to phosphorylate ACC, the key enzyme for fatty acid synthesis (17, 26, 28, 29). AMPK inactivates both the isozymes, ACC1 and ACC2, and thus downregulates fatty acid synthesis. Moreover, AMPK upregulates the expression of carnitine palmitoyltransferase I (CPT1), the rate-limiting enzyme in fatty acid oxidation and promoting the differentiation of Tregs. Consistently, mice deficient in AMPK were shown to have severe tissue inflammation due to an impaired development and functions of Tregs, which in turn induces exaggerated differentiation of effector T cell populations such as Th1, Th2, Th17 (17, 26, 28, 29).

Using the single cell RNA sequencing, the role of fatty acid in the pathogenicity of Th17 cells was further identified (30). The expression of CD5L (CD5 antigen-like) in Th17 cells regulate the pathogenic functions of Th17 cells by modulating the functions of RORγt, a master transcription factor of Th17 cells (30). The functions of CD5L was found to be associated with lipid metabolism, as it suppresses fatty acid synthase (30). It was demonstrated that CD5L promotes the binding of RORγt to Il10 gene locus while suppressing RORγt binding to Il23r and Il17 gene locus in Th17 cells, thus enhancing the development of non-pathogenic Th17 cells (30).

Factors that affect lipid synthesis were also found to be associated in regulation of interplay between effector and regulatory T cells in tissue inflammation. Lipid synthesis was shown to be regulated by transcription factor Myc, as Mycdeficient cells was found to have lower levels of lipid synthesis, which leads to reciprocal regulation of effector and regulatory T cells in tissue inflammation (25). In addition, cell signaling kinases such as mTOR are also critical for lipid biosynthesis, as inhibition of mTOR using rapamycin substantially reduced fatty acid synthesis upon T cell activation due to impairment of Myc induction (25). Upon T cell activation, PI3K and mTOR induces the expression of sterol regulatory element-binding proteins (SREBPs), which bind to the promoter of fatty acid synthesis (FAS)-specific genes (31). Taken together, the role of fatty acid is clearly implicated in regulation of tissue inflammation by enhancing the generation and functions of Tregs.

In addition to fatty acids, cholesterol, an essential component of cellular membranes, is required for T-cell activation and proliferation (32). It was suggested that an increase in cellular cholesterol helps in fighting bacterial infection by promoting inflammation (32, 33). However, in chronic metabolic inflammatory conditions such as obesity and atherosclerosis hypercholesterolemia, cholesterol is known to worsen the disease conditions (33). Similarly, an increased level of cholesterol was found in sera samples of RA patients, suggesting a pathogenic role of cholesterol in promoting tissue inflammation in RA (34–36). In autoimmune diseases like RA and systemic lupus erythematosus (SLE), a disturbed cholesterol efflux homeostasis results in worsening of the disease, and such patients were shown to have therapeutics effects by administration of high-density lipoproteins (37, 38). Cholesterol promotes the activation, differentiation, and proliferation of both CD4<sup>+</sup> and CD8<sup>+</sup> T cells via suppression of LXRβ and activation of sterol responseelement-binding protein-2 (SREBP2) (39). Furthermore, SREBP2 increases cholesterol synthesis, activating PI3K-mTOR pathway, which is crucial for T cell activation and differentiation; while LXRβ inhibits the cholesterol deposition thereby suppressing the T cell activation and proliferation (31, 39). Molecularly, cholesterol regulates TCR signaling by binding to the TCRβ chain, enhancing its avidity for MHC-Peptide complex through the formation of membrane raft (32). It has been recently reported that accumulation of intracellular cholesterol through mevalonic acid pathway drives Th17 cell differentiation (40). Interestingly, oxysterols such as 7α,27-OHC and 7β,27-OHC acts as RORγt agonists that binds to ligand binding domain of RORγt further activate its binding to Il17 gene locus as well as other Th17 cells promoting factors to potentiate Th17 cell differentiation (40). Moreover, LXR inhibits Th17 cell differentiation by interfering with the aryl hydrocarbon receptor mediated IL-17 transcription (41). Blocking of mevalonate pathway for cholesterol biosynthesis by atorvastatin inhibits Th1 cell differentiation and pro-inflammatory response during experimental autoimmune encephalomyelitis (EAE) (42).

### Nitric Oxide

Nitric oxide (NO) is a highly reactive free radical, which plays an important role in mediating numerous biological functions such as vasodilation, platelet aggregation, smooth muscle cell proliferation, superoxide radical generation, monocyte adhesion, LDL oxidation, and immune regulation etc. Briefly, NO is derived from L-Arginine in a reaction catalyzed by nitric oxide synthase (NOS). There are three different forms of NOS: neuronal nitric oxide synthase (nNOS or NOS1), inducible nitric oxide synthase (iNOS or NOS2), and endothelial nitric oxide synthase (eNOS or NOS3). Nitric oxide production in immune cells is primarily regulated by inducible NOS or iNOS, which is activated by different immunological stimuli such as lipopolysaccharide (LPS), interferon-γ (IFN-γ), tumor necrosis factor α (TNFα), interleukin 1β (IL-1β) generated during immune response (43–45).

The level of NO acts as a disease index in inflammation in many diseases. For example, in RA or osteoarthritis patients, increased levels of NO were found in the synovial fluid and serum of the inflamed joints, suggesting an association of NO with disease pathogenesis (46–48). Similarly, non-steroidal antiinflammatory drugs (NSAIDs) are helpful in treating high levels of excretory urinary nitrate (an indicator of NO in kidney during arthritis) (49, 50).

In addition, NO also acts as a mediator during tissue inflammation in IBD, as production of NO was found to be increased in IBD patients, and therefore found to be associated with the enhanced activity of iNOS in the inflamed mucosa of the gut (51). The primary source of enhanced NO in IBD patients was monocytes, lymphocytes, macrophages, and neutrophils, and these patients were characterized by enhanced urinary nitrate excretion (51, 52). Moreover, treatment with NSAIDs is found to have similar reducing results for nitrate excretion as in RA. Based on the clinical observations, iNOS has become one of the most prominent therapeutic target for IBD (52).

Since the role of NO was well-established in inflammatory conditions of arthritis and IBD, therefore it is essential to understand its influence on effector functions and differentiation of T cell subsets. Similar to cytokines, NO is a soluble factor that was shown to influence the differentiation of Th cell. The expression of iNOS is induced in activated CD4<sup>+</sup> T cells through distinct signaling pathways triggered by micro-environmental cues in the extracellular milieu. The role of NO was extensively studied in reciprocal regulation of Th17 and Tregs cells, as NO negatively regulates Th17 cell differentiation by nitrating the tyrosine residues of RORγt, which leads to inhibition of the transcriptional activity of RORγt in Th17 cells (53). Consistently, iNOS deficient mice were found to be more susceptible for EAE and experimental colitis associated with accumulation of higher frequency of Th17 cells in their target tissues (53). Contrary to iNOS deficiency, increased NO using NO donor, NOC-18, found to inhibit the development of Th17 cells by suppressing aryl hydrocarbon receptor (AHR) expression, and thus inhibiting Th17 cell differentiation (54). In addition to blocking RORγt functions, NO also found to block Th17 cell differentiation by antagonizing the functions of IL-6, as IL-6 is one of the key differentiating factor that induces de novo differentiation of Th17 cells (55).

NO modulates the differentiation of Th1, Th17, and Tregs by modulating their respective differentiating cytokines such as IL-12, IL-6, and TGF-β1. NO, together with IL-6, was found to potentiate the suppression of Tregs development induced by TGF-β1 and retinoic acid, as retinoic acid is known to enhance Foxp3 expression in the presence of TGF-β1, therefore stabilizing the functions of Tregs (55). It was suggested that higher amounts of retinoic acid overcome the repression of Foxp3 induced by NO and IL-6 and leads to the predominance of Foxp3<sup>+</sup> Tregs over Th1/Th17 cells (55). Although much of the published literature suggested that NO suppresses the differentiation of Th17 cells, it has also been shown that NO is critically required for the induction and stability of human Th17 cells (56). Physiological concentration of NO was shown to promote the generation of Th17 cells (56), as blocking of NOS2, or cGMP–cGK signaling pathway suppresses the de novo induction of Th17 cells in ovarian cancer patients (56).

In addition to Th17 cells, the effect of NO was shown to modulate TGF-β1 activity in potentiating Th1 cell differentiation in both IL-12-dependent and -independent manner (55). Although NO potentiates Th1 cell differentiation induced by IL-12 in an inflammatory environment, it was also suggested that NO maintains Th1 response even in the absence of IL-12 (57). NO was shown to enhance IFN-γ-dependent expression of T-bet, which further promote Th1 differentiation even in the absence of IL-12 (57). Even though the effect of TGF-β1 is dominant on IFN-γ signaling that result in the shift from Th1 to iTregs, however presence of NO potentiates STAT1-mediated IFN-γ signaling, inhibiting TGF-β1-induced Foxp3 expression, and thereby reinforces Th1 development (57).

In addition to Th1 and Th17 cells, the effect of NO was also shown on the development of Th9 cells. It was demonstrated that NO enhances differentiation of Th9 cell, and therefore exacerbates IL-9, and Th9 dependent allergic inflammation in asthma (58). Mechanistically, NO was shown to enhance both TGF-β1, and IL-4 signaling pathways to promote the development of Th9 cells. Moreover, NO nitrosylates MDM2 at cysteine residue thereby derepressing p53 from MDM2-p53 complex, which induce IL-2 production and activate STAT5- IRF4 axis to promote Th9 differentiation (58). In addition, NO was found to increase the surface expression of IL-4R, which potentiates STAT6-mediated IRF4 dependent expression of IL-9 in Th9 cells (58). Other than IL-4-IL-4R signaling, NO was also shown to enhance TGFβR expression, which in turn enhances TGF-β1 dependent binding of PU.1 to Il9 promoter in differentiation of Th9 cells (58). Consistently with these observations, NO was found to exacerbate airway inflammation while iNOS−/<sup>−</sup> mice were shown to have attenuated airway inflammation due to a decreased frequency of Th9 cells (58). Taken together, these observations clearly indicated that NO affect helper T cells differentiation and functions in autoimmune as well as allergic inflammation, and therefore modulating NO could be a potential metabolic checkpoint in T cells differentiation and disease pathophysiology.

# Adenosine Triphosphate (ATP)

During the metabolic process, adenosine triphosphate (ATP) is generated, and used primarily as a source of energy for cells. In addition to energy source, ATP can act as an extracellular signaling molecule, which mediate cell to cell communication in an autocrine and paracrine manner (59, 60). Glucose feeds into glycolysis to generate ATP during T cell activation and proliferation. In this process of glycolysis, one glucose molecule is metabolized to provide two reduced nicotinamide adenine dinucleotide (NADPH), two molecules of ATP and two pyruvate molecules. Pyruvate further feeds into tricarboxylic acid cycle (TCA) in less active cells such as memory cells and Foxp3<sup>+</sup> Tregs. In TCA cycle, pyruvate is metabolized to generate NADH and reduced flavin adenine dinucleotide (FADH2), which further feeds into OXPHOS, an oxygen dependent step, and generate 36 ATP molecules for one molecule of glucose.

Under physiological conditions, the cellular ATP can be released from healthy cells through exocytosis while necrotic or apoptotic cells releases ATP under pathological conditions (61). As a messenger, extracellular ATP binds to purinergic receptors, P2X, and P2Y that are present on the cell surface. Upon T cell activation via TCR, ATP contributes to the activation of MAPK signaling cascade through P2X receptor and contribute to T cell activation (62). The binding of ATP to its receptors initiates a signaling cascade and integrate various cellular signaling events in inflammatory conditions in both human and mouse T cells (63, 64). Studies have shown that extracellular ATP triggers the effector CD4<sup>+</sup> T cells proliferation while inhibiting the functions of Tregs (65). ATP was found to enhance the differentiation of Th17 cells and exacerbated T-cell-mediated colitis in mouse model (66). Mechanistically, ATP increases the number of CD70highCD11clow lamina propria cells in germ-free mice and upregulate the expression of TGF-β1, IL-6, IL-23p19, which together induce Th17 cell differentiation (66). IL-6, which support the generation of Th17 cells, enhances the generation of cellular ATP as well to further enhance the generation of Th17 cells in a feed forward loop (66). It was further demonstrated that IL-6-mediated ATP-P2XY signaling converts Foxp3<sup>+</sup> Tregs into Th17 cells in vivo, suggesting that ATP can control Tregs and Th17 cells reciprocally (62). In addition to Th17 cells, the role of extracellular ATP was also tested in generation and functions of Foxp3<sup>+</sup> Tregs, as it was demonstrated that P2XY receptor is highly enriched on Foxp3<sup>+</sup> Tregs. Consistently, extracellular ATP was found to suppress the stability and functions of Foxp3<sup>+</sup> Tregs (62). Mechanistically, it was shown that Foxp3<sup>+</sup> Tregs have a specialized function that control extracellular ATP by breaking it down to AMP via CD39, a nucleoside triphosphate diphosphohydrolase-1 (NTPDase 1). The surface expression of CD39 was found to be highly enriched on Foxp3<sup>+</sup> Tregs cells, and is driven by Foxp3 (67). These observations thus imply that Tregs have an intrinsic capacity to control effector cell functions in tissue inflammation by limiting ATP concentration in inflammatory milieu.

Since extracellular ATP provides pro-inflammatory signal to effector T cells and also helps in maturation of dendritic cells, therefore controlling the presence of ATP might be a potential strategy for suppressing tissue inflammations.

Other than Foxp3<sup>+</sup> Tregs, extracellular ATP inhibits Tr1 differentiation by triggering the inactivation of AHR, which is crucial for Tr1 cell differentiation and functions (16). Molecularly, it was demonstrated that ATP induces the expression of HIF-1α, which bind and inhibit the activation of AHR and ARNT (AHR nuclear translocator) thereby blocking Tr1 cell differentiation (16). Moreover, ATP interferes in recruitment of AHR to Il10, Il21, and Entpd1 promoters, and therefore repressing their expression in Tr1 cells (16). Altogether it suggests that extracellular ATP provide crucial signal in maintaining the balance between effector and regulatory T cells in physiological and disease conditions, and therefore ATP could be a potentially metabolic checkpoint to modulate effector and regulatory T cells functions in tissue inflammation.

# Nicotinamide Adenine Dinucleotide (NAD)

NAD is another important energy metabolite which is released by cells through lysis during cellular damage or inflammation mediate intracellular signaling. It has been reported that NAD promotes Th1, Th2, and Th17 cells differentiation (68). Genomewide transcriptomics analysis identified the upregulation of tryptophan hydroxylase-1 (Tph1) in Th1, Th2, and iTreg cells in the presence of NAD which facilitates their differentiation (68). Moreover, it was found that NAD triggers IL-10 production in Th1 cells and IL-10 and IL-17 production in Th2 cells via upregulation of Tph1, which provide protection against EAE by promoting remyelination and axonal regeneration. In addition, NAD skews iTreg cells into Th17 cells, which produce TGF-β1 and have immunosuppressive properties (68).

NAD not only induces immunosuppressive T cells, but also inhibits T cell proliferation via activation of P2X7 receptor signaling (69). Mechanistically, NAD leads to ART2.2-mediated ADP-ribosylation of the cytolytic P2X7 receptor expressed on T cells. Tregs express mono–ADP-ribosyltransferase (ART2.2) which catalyzes the covalent transfer of the ADP-ribose group from NAD<sup>+</sup> onto arginine residues of membrane target proteins resulting in NAD induced cell death of Treg cells (70, 71). The deleterious effects of NAD on Treg cells can be protected by an inhibitory ART2.2-specific single domain antibody (72). In tumor mouse models, administration of exogeneous NAD induces antitumor immune response by selective depletion of Treg cells (72). Thus, NAD is one of the major metabolic signals which serve as checkpoint for CD4<sup>+</sup> T cell differentiation with anti-tumor properties and can be used to treat autoimmune diseases such as MS, IBD and chronic and inflammatory diseases.

## TRANSCRIPTION FACTORS AS METABOLIC CHECKPOINTS OF T CELL DIFFERENTIATION

The interplay between different signaling cascades together play a vital role in regulating T cell differentiation. A number of transcription factors have been reported that regulate different metabolic pathways crucial for T cell differentiation and functions. Here we discussed the role of transcription factors that act as metabolic checkpoints in Th cells differentiation and functions.

### Myc

Myc, a basic helix–loop–helix leucine zipper transcription factor, is induced upon TCR signaling and essential in metabolic reprogramming of T cells activation and functions (73). Expression of Myc is essential for glycolysis and glutaminolysis in activated T cells, as Myc induces the upregulation of GLUT1; also, known as SLC2A1, which is required for glucose uptake as well as functions of pyruvate kinase, lactate dehydrogenase A (LDHA) and hexokinase in glycolysis (73). In addition, it was suggested that Myc promotes glutaminolysis by increasing the expression of glutaminase and glutamine transporters (5). Glycolysis and glutaminolysis are essential for proliferating effector Th cell populations like Th1, Th2 and Th17 cells, therefore the role of Myc in generation of effector T cell was proposed (5). Although Myc and HIF-1α are two most critical transcription factors essential for metabolically active effector T cells, it was found that Myc, but not HIF-1α, is responsible for metabolic reprogramming in T cell activation and rapid cell divisions in proliferating T cells (73).

The role of Myc was identified in regulation of effector and regulatory T cells generation and functions, as Myc found to be essential in maintaining the balance between Th17/Foxp3<sup>+</sup> Treg cells differentiation. Gomez-Rodriguez et al have shown that Myc deficiency in the Itk−/<sup>−</sup> leads to impaired Th17 cell differentiation due to increased expression of Pten (74). In addition, Myc activates miR-19b, which is found to repress PTEN and enhance STAT5 activation. Myc is also found to activate IL-2-mediated PI3K-mTOR pathway that results in enhanced Th17 cells differentiation (74). Deletion of Myc in CD4<sup>+</sup> T cells inhibits PI3K-mTOR pathway, which lead to an increased expression of Foxp3 and resulted in enhanced frequency of Tregs cells. Based on the emerging literature, it is clearly evident that metabolically active T cells require Myc activity, which is essential for modulation of effector and regulatory T cells generation and functions.

# HIF-1α

In addition to Myc, metabolically active T cells express increased levels of HIF-1α. Transcription factor HIF-1α forms a heterodimer and composed of two subunits, and is induced in response to low oxygen concentrations in a state of hypoxia. Under hypoxic condition, HIF-1α dimerizes with HIF-1β, a constitutively expressed ARNT, and translocate into the nucleus where it binds to hypoxia responsive elements (HREs) and mediate transcription of its target genes (75). Under the optimal oxygen supply, a state called normoxia, HIF-1α gets hydroxylated at proline residues by prolyl hydroxylases, which make HIF-1α sensitive to ubiquitination-mediated proteasomal degradation by an E3 ubiquitin ligase (76–78). In addition, FIH-1 (Factor inhibiting HIF-1) also hydroxylate HIF-1α at asparagine residue, which further block the recruitment of p300/CBP therefore limiting the transcriptional activity of HIF-1α (79, 80). Under physiological hypoxia, the balance between nitric oxide and reactive oxygen species (ROS) were found to stabilize HIF-1α (81). The role of HIF-1α is quite established in Th cell differentiation and functions (82). HIF-1α promotes differentiation of Th17 cell by forming a tertiary complex with RORγt and p300 to Il17 promoter, thus enhancing the transcription of IL-17 gene and targeting ubiquitination-mediated proteasomal degradation of Foxp3, resulting in reinforcing the development of Th17 while diminishing the generation of Tregs (83). Taken together these observations thus imply that transcription factor HIF-1α act as a metabolic checkpoint between Th17 and Tregs cell differentiation and functions (84). Consistently, these

observations were further supported with the fact that HIF-1α-conditional deficient animals were found to be relatively resistant to the development of EAE associated with a reduced frequency of Th17 cells (84). In addition, the role of HIF-1α has been found to be associated with Th1 and Th2 cells; however, the deficiency of HIF-1α was not found to impair Th1 and Th2 cell differentiation, suggesting a dispensable role of HIF-1α in Th1 and Th2 cell differentiation (85). In addition, it has been recently reported that HIF-1α increases glycolytic activity in development of Th9 cells. Mechanistically, HIF-1α was found to induce IL-9 promoter activity by binding directly to Il9 promoter in Th9 cells (17). Taken together these observations suggest that HIF-1α is critical for T cells activation and proliferation as well as for their differentiation into various effector T cells.

### IRF4

The transcription factor, interferon regulatory factor 4 (IRF4) is required for TCR-mediated metabolic programming in T cells (86). IRF4 expression is induced upon TCR stimulation, which was found to drive transcriptional program required for the differentiation of Th cell lineages. IRF4 translates TCR-affinity signals into the proliferation of appropriate T cell lineages, and is required for the survival of activated T cells (86). IRF4 binds to promoters of the genes and other downstream factors that are required for differentiation and key metabolic functions in T cells. Association of IRF4 with AP-1, c-Myc and HIF-1α regulates metabolic programming in activated T cells (86). Specifically in the Th differentiation, the initial role of IRF4 was found to be associated with Th2 differentiation, as IRF4-deficient T cells were failed to differentiate into Th2 cells (87). The role of IRF4 was further tested in other Th subsets such as Th17 and Th9 cells. Since Th2 and Th9 cells are sister populations as they share common differentiation factors, therefore, similar to Th2, the role of IRF4 was found to be crucial in Th9 development (88). IRF4 deficiency in animals leads to a defective Th9 cells development and associated with less severe allergic inflammation (88). The role of IRF4 was further identified in the development and functions of Th17 cells (89). TCR mediated induction of IRF4 is crucial for the induction of IL-21, which is found to be a crucial cytokine for amplification of Th17 cells (89). In addition to effector cell populations such as Th2, Th9, and Th17, the role of IRF4 was found to be associated with the function and development of Tregs and Tr1 cells. IRF4 deficiency in Foxp3<sup>+</sup> Tregs leads to defect in Tregs function, which lead to Th2 cell-mediated tissue inflammation (90). It was further suggested that IRF4 in Tregs cells control Th2 mediated tissue pathology (90). In addition to Foxp3<sup>+</sup> Tregs, the role of IRF4 has been also identified in Tr1 cells (91). Taken together these observations indicate that IRF4 is a critical metabolic checkpoint that is crucial for maintaining effector and regulatory T cells generation and functions.

### BCL-6

The transcriptional repressor B-cell lymphoma 6 (Bcl-6) play a role in metabolic regulation of T cells, and is essential in promoting the generation of memory T cells in both CD4<sup>+</sup> and CD8<sup>+</sup> compartments (92). In addition, the role of Bcl-6 has been demonstrated in shaping the development of T follicular helper (Tfh) cells (93), as the balance between Bcl-6 and Blimp1 was found to be critical in generation of Tfh cells, which are crucial for the formation of germinal centers (93).

As a repressor, Bcl-6 is found to downregulate genes of the glycolytic pathways by directly binding to their respective promoters (94). The repression of glycolysis by Bcl-6 is directly linked to IL-2 concentration, as higher concentration of IL-2 leads to downregulation of Bcl-6 and thereby upregulate glycolysis in proliferating T cells. Once the concentration of IL-2 limits, Bcl-6 gets upregulated, and therefore controls T cell proliferation by limiting the glycolysis and generation of metabolic precursors that are crucial for T cells growth (94).

In addition to controlling T cells proliferation, Bcl-6 was also found to influence Th cell differentiation. Bcl-6 was found to inhibit Th1 and CD8<sup>+</sup> Tc1 cell differentiation by repressing glycolysis which is the major metabolic pathway in these T cell subsets (94). Interestingly, T-bet, a master transcription factor of Th1 lineage, is reported to inhibit Bcl-6, and overcome the repression of glycolysis by Bcl-6 in Th1 cells (94). The Bcl-6 mediated repression of glycolytic gene program in effector T cells is so dominant that it cannot be restrained by HIF-1α and c-Myc, which are also known to upregulate glycolysis (94). Bcl-6 by inhibiting glycolysis creates a switch from effector to memory T cell and thus promotes memory cell formation. The role of Bcl-6 was also found to be associated in Th9 cells, as Bcl-6 is found to impair Th9 differentiation by competing with STAT5/STAT6 binding sites on Il9 promoter, and thus represses the transcription of Il9 gene (95). IL-2 induces Th9 differentiation while IL-21 inhibits Th9 differentiation by differential regulation of Bcl-6 expression (95, 96). These observations suggest that Bcl-6 act as an important metabolic regulator of Th differentiation pathways.

### Foxo

Foxo1 is an also essential metabolic checkpoint transcription factor in Th differentiation. Forkhead box O (Foxo) family of transcription factors consists of four Foxo family members: Foxo1, Foxo3, Foxo4, and Foxo6. Foxo1 and Foxo3a were found to be most abundant in T cells, therefore the role of Foxo1 and Foxo3a was extensively studied in T cell development, differentiation and functions (97). Foxo1 and Foxo3 plays a critical role in cell metabolism, apoptosis, cell cycle progression, and detoxification of reactive oxygen species (98, 99). The DNA binding motif of Foxo has been characterized and found to be conserved. The role of Foxo1 and Foxo3 was widely studied in both iTregs and nTreg cell development and functions, as mice deficient in both Foxo1 and Foxo3 have decreased number of nTregs with loss of their suppressive functions (100). Similarly, Foxo1 and Foxo3 deficient mice showed impaired generation of induced Treg cells (Foxp3+−iTregs), which is induced in the presence of TGF-β1 (100). Consistent with these observations, Foxo1, and/or Foxo3a deficient mice develop severe colitis due to non-functional Tregs (100). In addition to Tregs, the role of Foxo1 was found to be associated with Th17 cells

differentiation and function, as Foxo1 deficient mice were found to be susceptible to EAE associated with an enhanced frequency of Th17 cells (101). In addition, it was shown that Foxo1 suppressed the functions of RORγt by directly binding to it, and therefore leads to decreased Th17 cells generation (101). It was further identified that IL-23-IL-23R pathways regulated Foxo1 functions via SGK1, which induces the generation of pathogenic Th17 cells (102).

It has been demonstrated that AKT phosphorylates Foxo at Ser<sup>253</sup> residue causing its nuclear exclusion and ubiquitin mediated degradation. The inhibition of PI3K-AKT pathway promotes Treg cell differentiation through the activation of Foxo (100). Naïve Foxo1-deficient CD4<sup>+</sup> T cells become Tbet+IFNγ <sup>+</sup>Th1 cells and fail to differentiate into Foxp3<sup>+</sup> Treg cells (103). TGF-β-induced Foxo1 inhibits the expression of T-bet while the activation of S1P1 signaling interferes with TGFβ-SMAD3 pathway by activating S6 kinases, downstream of S1P1-induced activation of mTORC1 pathway. This inactivates Foxo1 and thus potentiates Th1 cell differentiation (103). Recent data from our group has identified the crucial role of Foxo1 in promoting Th9 cell differentiation, through inhibition of PI3K/AKT pathway, which activates Foxo1 and increases IL-9 production in Th2, Th9, and Th17 cells. Loss of Foxo1 attenuates IL-9 in Th9 cells and ameliorates allergic inflammation in asthma (104). Moreover, the activation of STAT5 and PI3K-AKT-mTOR signaling pathway by IL-7 increases the abundance of the histone acetyl transferase p300 which promotes the acetylation of histones at the Il9 promoter resulting in dephosphorylation of Foxo1 thus inducing the production of IL-9 and potentiating Th9 cell differentiation (105). Deficiency of Foxo1 inhibits IL-7 mediated Th9 cell differentiation and antitumor activity of Th9 cells (105). These observations clearly indicate that Foxo proteins are critical metabolic checkpoints that affect T cells differentiation and functions in immunity and autoimmunity.

# PPARγ

Peroxisome proliferator-activated receptor gamma (PPARγ) is a ligand-dependent transcription factor which is known to regulate glucose and lipid metabolism, cell growth, differentiation and apoptosis. PPARγ-deficient T cells display enhanced proliferation with an increased activation of ERK and AKT, which lead to increased cytokine production under Th1, Th2, Th9, and Th17 differentiation conditions (106). PPARγ increases the stability of IκBα, Foxo1, and Sirt1, which are the negative regulators of NFκB, thus inhibiting T cell proliferation (106). In addition to T cell proliferation, the role of PPARγ was also found to be associated with T cells activation and differentiation. The role of PPARγ was shown in the differentiation of Th17 cells, as PPARγ was found to interfere with TGF-β/IL-6-induced transcriptional activation of RORγt by preventing the removal of corepressor from the RORγt promoter and thus regulate Th17 cells differentiation. Consistently, PPARγ deficient mice were found to be severely susceptible for tissue inflammation in EAE due to an increased frequency of Th17 cell (107). However, PPARγ activation in CD4<sup>+</sup> T cells didn't show any inhibitory effect on Th1, Th2, or Treg cell differentiation (107).

### SIGNALING TRANSDUCERS AND SENSORS AS METABOLIC CHECKPOINTS OF T CELL DIFFERENTIATION

### mTOR

The mTOR mammalian target of rapamycin (mTOR) is an evolutionary conserved serine/threonine kinase that is a part of PI3K-AKT pathway. mTOR comprises of two functionally distinct complexes: mTORC1 and mTORC2, which play distinct role in Th differentiation. mTORC1 is composed of regulatory associated protein of mTOR (Raptor), Rheb (a small GTPase), G protein β-subunit-like protein (GβL), mammalian lethal with Sec13 protein 8 (mLST8), the proline-rich Protein Kinase B (PKB)/Akt substrate of 40 kDa (PRAS40), and DEP-domaincontaining mTOR-interacting protein (DEPTOR). On the other hand, mTORC2 comprises of mLST8 and DEPTOR, rapamycininsensitive companion of TOR (RICTOR), mSIN1 (mammalian stress-activated protein kinase interacting protein-1) and the protein observed with RICTOR (PROTOR) (108–110). Both mTORC1 and mTORC2 was found to have differential sensitivity for rapamycin, as mTORC1 is rapamycin sensitive and plays a role in autophagy, protein translation, and ribosome biogenesis while mTORC2 is rapamycin insensitive. mTORC1 activation leads to phosphorylation and activation of the ribosomal S6 kinase 1 (S6K1) while mTORC2 phosphorylates Akt at serine 473 (111). mTOR is crucial for regulating glucose metabolism and upregulates glycolysis in T cells, as mTOR–/– cells were shown to have decreased glycolytic activity. The mTOR pathways were found to play an essential role in T cells differentiation and functions. mTOR is required for effector T cell lineage commitment, as activated T cells lacking mTOR fails to differentiate into effector T cell and rather they differentiate into Foxp3<sup>+</sup> Tregs (111). Mechanistically, lack of mTOR leads to Smad3 activation even in absence of TGFβ1 which potentiate the generation of Foxp3<sup>+</sup> Tregs. The role of mTORC1 and mTORC2 was found to be associated with Th1 and Th2 differentiation, respectively. CD4<sup>+</sup> T cells lacking mTORC1 activity were found to have reduced phosphorylation of STAT4 in response to IL-12, which leads to reduced generation of Th1 cells. Although mTORC1 activity was found to be indispensable for Th1 cell differentiation, Th2 cells differentiate even in absence of mTORC1 but they require mTORC2 activity for their differentiation (111). Like Th1 cells, mTORC1 was found to be essential for the differentiation of Th17 cells while deficiency of mTORC2 was not found to be associated with Th17 cells differentiation (111). Recently it has been reported that mTOR induces Th9 cell differentiation by upregulating glycolytic pathway in HIF-1α dependent manner (17). mTORC1 regulates glucose metabolism and function in CD8<sup>+</sup> T cell independent of PI3K and PKB (108).

### AMPK and LKB1

Activated T cells possess AMP-activated protein kinase (AMPK), a glucose-sensitive metabolic checkpoint, found to regulate mRNA translation and glutamine-dependent mitochondrial metabolism. AMPK is a heterotrimeric serine/threonine kinase complex which promotes energy conservation in T cells and primarily involved in maintaining T cell bioenergetics and viability (112, 113). During prolonged starvation and stress conditions, AMPK promotes catabolic processes like fatty acid oxidation for limiting energy expenditure and replenishing ATP production rather than anabolic processes which consume ATP (112, 113). AMPK promotes fatty acid oxidation by increasing the expression and phosphorylation of carnitine palmitoyl transferase 1A (CPT1A), which is the rate-limiting enzyme and inhibits acetyl-CoA carboxylase 2 (ACC2) (114). AMPK also enhances mitochondrial biogenesis by promoting the transcriptional activity of peroxisome proliferator-activated receptor-γ coactivator 1α (PGC1α) and oxidative metabolism (114). Furthermore, AMPK found to inhibit glycolysis, glutaminolysis, glycogen and fatty acid synthesis while it promotes oxidative phosphorylation and autophagy (112–114). AMPK is activated by phosphorylation of α subunit at Thr172 by LKB1 under the conditions of bioenergetic stress (113). AMPK is also activated via Ca2+ calmodulin-dependent protein kinases upon TCR triggering (20). AMPK inhibits mTORC1 activity by phosphorylation of TSC2 and RAPTOR, which are important for its activity, thus inhibiting the T helper cell differentiation. T cells deficient in AMPKα1 found to display reduced mitochondrial bioenergetics and metabolic plasticity in response to glucose limitation (112–114). Glucose limitation affect Th1 differentiation by reducing the mRNA and protein level of IFN-γ (115). AMPK play a major role in limiting IFN-γ production under glucose unavailability. AMPKα1-deficient T cells produced enhanced IFN-γ even in the metabolically unfavorable condition (115). As discussed above that effector Th cells relies on glycolysis while Tregs requires on FAO. AMPK is highly enriched and active in Tregs, as it was found that AMPK drives naïve T cells into Tregs (14, 29, 116). AMPK found to negatively regulate glycolysis, which is required for the generation of effector Th subsets. Consistently, metformin, an activator of AMPK, found to block Th17 cell differentiation in vitro and in vivo and alters the ratio of Th17: Tregs in mouse model of colitis and asthma (117, 118). Consistently, metformin treatment was found to ameliorate tissue inflammation in mouse models of autoimmune diseases. Taken together, it indicates that AMPK is crucial for maintaining the balance between effector and regulatory T cells.

The Liver Kinase B1 (LKB1) is a serine/threonine kinase that links cellular metabolism with cell growth and proliferation (119). LKB1 was initially identified in Peutz-Jeghers syndrome, an autosomal dominant disorder that leads to carcinoma (119). LKB1 was found to be upstream of AMPK and two together regulate metabolically active T cells. In fact, LKB deficiency leads to decreased activation of AMPK. LKB1 is a key regulator of lipid and glucose metabolism in T cells where loss of LKB1 increases glucose metabolism (120). Similar to AMPK, LKB1 regulates the functions of both CD4<sup>+</sup> and CD8<sup>+</sup> T cells during inflammation (121, 121). LKB1 primarily regulates metabolism in T cells via AMPK-dependent and independent pathways. In fact, a subset of LKB1 functions was found to be carried out by AMPK, as deletion of AMPK display similar defect in T cell activation. T cell-specific deletion of the gene that encode LKB1 found to affect the number of thymocytes and peripheral T cells (121). In addition, LKB1 deficient T cells were shown to have enhanced activation and cytokine production, which could be due to altered glycolytic and lipid metabolism (121). Loss of LKB1 promotes Th1 and Th17 cell differentiation. AMPK and LKB1 together negatively regulate the inflammatory cytokine production by T cells by inhibiting mTORC1 signaling, which is critical for inflammatory cytokine production (121). These observations indicate the role of LKB1- AMPK axis in metabolism of T cells affecting their differentiation and functions.

### SIRT1

SIRT1 is a mammalian homolog yeast NAD<sup>+</sup> dependent type III histone deacetylase which plays an important role in a variety of biological and metabolic processes in immune cells. SIRT1 is highly expressed in dendritic cells (DC) and reciprocally regulates Th1 and iTreg cell differentiation (122). SIRT1 signaling in DCs inhibits Th1 differentiation while promotes Tregs generation via modulation of DC derived T-cell polarizing cytokines such as IL-12 and TGF-β1 in a HIF-1α dependent manner (122). SIRT1-HIF1α signaling axis in DC reciprocally regulates Th1 and iTreg cell generation by modulating IL-12- STAT4 and TGFβ1-SMAD3 pathways in a mTOR independent manner (122). However, an unexpected proinflammatory role of SIRT1 has been established in the generation and functions of Th17 cells. SIRT1 activates RORγt by deacetylation enhancing Th17 cell differentiation thereby promoting autoimmunity (123). In addition to Th1, Th17, and Tregs, the role of SIRT1 has also been established in the differentiation and functions of IL-9 producing Th9 cells (17). SIRT1 inhibits Th9 cell differentiation by negatively regulating mTOR-HIF1α dependent glycolytic activity (17). Overexpression of SIRT1 negatively regulates IL-9 producing CD4<sup>+</sup> T cell differentiation while SIRT1 deficiency promotes Th9 cell differentiation by inducing mTOR-HIF1α dependent glycolytic activity (17). Taken together observations clearly indicate that SIRT1 is one of the crucial metabolic checkpoints in regulating T cells differentiation and functions in tissue inflammation.

# METABOLIC CHECKPOINTS DURING TISSUE INFLAMMATION

T cell encounters a variety of micro-environmental cues during differentiation into effector and regulatory T cells. The molecules which are essential for host metabolism affect and modulate T cells functions. In addition to antigen and cytokines, metabolic precursors, play a key role in activation, proliferation, and differentiation of T cells that lead to diverse immunological responses in tissue inflammation. An activated effector T cell undergoes a dramatic metabolic shift to support its growth, proliferation, differentiation, and functions. This metabolic shift occurs during T cells activation in tissue inflammation due to limited availability of oxygen and nutrients. Tumor infiltrating lymphocytes (TILs) are good example of such changes, where TABLE 1 | List of drugs targeting metabolic checkpoints in clinical trials in inflammation and cancer.


(Continued)

### TABLE 1 | Continued


(Continued)

TABLE 1 | Continued


tumor microenvironment support less oxygen and glucose, which leads to metabolic shift in TILs (124–126). Under hypoxic environment in tumor, T cells express HIF-1α, which further modulate the course of immune response in tumors and its associated inflammation (127).

The metabolic shift known as Warburg effect is an essential phenomenon of active and growing T cells in tissue inflammation to support its energy demand in terms of ATP production and synthesis of biosynthetic precursor and intermediate molecules (128). The Warburg effect is particularly evident during tissue inflammation when activated T cell shifts toward aerobic glycolysis providing glucose-6-phosphate for the pentose phosphate pathway (PPP) generating 3-phosphoglycerate. This 3-phosphoglycerate is then utilized for the serine biosynthetic pathway which is essential for the biosynthesis of various cytokines required to induce effector functions of T cells in tissue inflammation. The aerobic glycolysis provides pyruvate to the TCA cycle leading to the synthesis of citrate, which is required for the membrane fatty acid synthesis (129). Metabolic reprogramming of T cells in tissue inflammation toward aerobic glycolysis further allow these cells to overcome stressful microenvironment, such as reduced cellular oxygen level, during tissue inflammation (130, 131). Moreover, lesser amount of ATP is generated through oxidative phosphorylation due to low levels of oxygen in the vicinity of tissues during inflammation (132, 133). Thus, aerobic glycolysis constitutes the major metabolic pathway in activated T cells, activated B cells, activated macrophages, DCs, stimulated natural killer cells and neutrophils during tissue inflammation (2, 9, 84, 134–139).

The metabolic reprogramming of T cells is well controlled at various metabolic checkpoints during homeostasis and inflammation. Metabolic shift toward aerobic glycolysis is supported by an enhanced expression of GLUT1 for increased glucose transport inside the cell for enhancing the rate of aerobic glycolysis (12). Interestingly, effector T cells efficiently adapts to hypoglycemia at the site of inflammation where glucose levels are low. They do so by internalizing glutamine from their surrounding environment and catabolizes it through glutaminolysis for the continuity of TCA cycle (136). The glutamine supply to the differentiating effector T cells is crucial for maintaining Th1 cell differentiation, as it is observed that in the absence of glutamine during tissue inflammation, T cell skew toward Treg phenotype (140–142). Recently, similar role of amino acids such as leucine and arginine has also been established for differentiation and functions of effector T cells (143, 144).

The key players that bring about these metabolic reprogramming are mTOR and transcription factors such as Myc and HIF-1α as discussed above. Activity of these regulators in turn modulates the AMPK activity (145). mTOR activation is responsible for a number of dynamic changes within the proliferating activated T cells such as enhanced mRNA translation and fatty acid synthesis, maintenance of compartment Myc levels which in turn, is crucial for the induction of glycolytic gene expression (73, 143, 146, 147). Remarkably, different mTOR complexes triggers distinct metabolic programming that lead to effector vs. memory T cell generation (134). For example, mTORC1 signaling shifts the metabolism toward aerobic glycolysis during the proliferation of effector T cells; whereas mTORC2 is required for the metabolic reprogramming in memory T cells (134). Reduction of glycolysis together with an increased oxidative catabolism downregulate mTOR signaling, which skew T cells to become Tregs. Taken together it clearly indicates that mTOR serves as a key metabolic checkpoint for the development of effector and regulatory T cells that influences the outcome of tissue inflammation (111). The role of AMPK was also suggested in modulating effector and regulatory T cell response, as AMPK were shown to promote the development of regulatory and anti-inflammatory T cells while limiting the generation of effector T cells (145). Low cellular energy and insufficient nutrient supply trigger AMPK activation leading to inhibition of mTOR and upregulation of fatty acid catabolism, which support Tregs development and functions. In addition, at transcriptional level, upregulation of HIF-1α and Myc gene expression along with suppression of Bcl-6 is essential for the metabolic reprogramming in T cells during tissue inflammation (73, 84, 148, 149), which is initiated by Myc in CD4<sup>+</sup> T cells while it is maintained further in CD8<sup>+</sup> T cells by AP-4 and IRF4 transcription factors as soon as Myc activity declines (86, 150). Thus, these metabolic checkpoints play a crucial role in tissue inflammation and have enormous potential for immunotherapy.

### CONCLUSION

The metabolic checkpoints are essential to maintain balance between pro- and anti-inflammatory T cells during tissue inflammation. These metabolic checkpoints include various cellular components ranging from cellular metabolites, cell signaling molecules and transcription factors. The dynamic interactions between these checkpoints determine the outcome of T cell response in tissue inflammation. Emerging data clearly indicate that metabolic checkpoints are different for effector, memory, and regulatory T cells, and so does the factors that regulate these metabolic reprogramming. As discussed above in this review, a dysregulated metabolic checkpoint is indicative of imbalance between effector, memory, and regulatory T cells. Many of these metabolic checkpoints are in an area

### REFERENCES


of active research and being tested as potential therapeutic targets for inflammatory diseases. However, the regulatory metabolic network of these checkpoints in tissue inflammation, autoimmune diseases as well as in infection is not completely understood. Nonetheless, several therapeutic drugs are available or undergoing clinical trials for ameliorating tissue inflammation targeting metabolic checkpoints as summarized in **Table 1**. More broader understanding is required for different metabolic checkpoints to formulate new immunotherapies as well as immunomodulation for various inflammatory conditions and tumor microenvironment.

# AUTHOR CONTRIBUTIONS

SR, ZAR and AA contributed to writing the review. AA and SR corrected and edited the review.

### ACKNOWLEDGMENTS

This work was supported by Wellcome Trust/DBT India alliance intermediate fellowship (IA/I/12/1/500524), Department of Biotechnology, Government of India and Core grant of Translational Health Science and Technology Institute. SR was supported by a Ph.D. fellowship from Council of Scientific and Industrial Research (CSIR). ZAR was supported by a Postdoctoral fellowship (N-PDF) from Department of Science and Technology (DST), Government of India.


receptor and Ca2<sup>+</sup> in T lymphocytes. J Exp Med. (2006) 203:1665–70. doi: 10.1084/jem.20052469


**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 Roy, Rizvi and Awasthi. 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.