# ROLE OF CD1- AND MR1-RESTRICTED T CELLS IN IMMUNITY AND DISEASE

EDITED BY : Kazuya Iwabuchi and Luc Van Kaer PUBLISHED IN : Frontiers in Immunology

#### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88963-122-3 DOI 10.3389/978-2-88963-122-3

#### 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

# ROLE OF CD1- AND MR1-RESTRICTED T CELLS IN IMMUNITY AND DISEASE

Topic Editors:

Kazuya Iwabuchi, Kitasato University School of Medicine, Sagamihara, Japan Luc Van Kaer, Vanderbilt University School of Medicine, United States

CD1 and MR1 are major histocompatibility complex (MHC) class I-related proteins that bind and present non-peptide antigens to subsets of T cells with specialized functions. CD1 proteins typically present lipid antigens to CD1-restricted T cells, whereas MR1 presents vitamin B-based ligands and a variety of drugs and drug-like molecules to MR1-restricted T cells. The CD1 family of antigen presenting molecules has been divided into two groups: Group 1 contains CD1a, CD1b and CD1c, and Group 2 contains CD1d. Additionally, CD1e is expressed intracellularly and is involved in the loading of lipid antigens onto Group 1 CD1 proteins. Humans express both Groups 1 and 2 CD1 proteins, whereas mice only express CD1d. Group 1 CD1 proteins present lipid antigens to T cells that generally express diverse T cell receptors (TCRs) and exhibit adaptive-like functions, whereas CD1d presents lipid antigens to subsets of T cells that express either diverse or highly restricted TCRs and exhibit innate-like functions. CD1d-restricted T cells are called natural killer T (NKT) cells, which includes Type I or invariant NKT (iNKT) cells expressing semi-invariant TCRs, and Type II NKT cells expressing more diverse TCRs. CD1-restricted T cells have been implicated in a wide variety of diseases, including cancer, infections, and autoimmune, inflammatory and metabolic diseases. Additionally, NKT cells have been targeted for immunotherapy of disease with ligands such as α-galactosylceramide for iNKT cells, or sulfatide for Type II NKT cells. Like iNKT cells, MR1-restricted T cells express semi-invariant TCRs and display innate-like functions. MR1-restricted T cells, also called mucosal-associated invariant T (MAIT) cells, have been implicated in immune responses against a variety of pathogens such as Mycobacterium tuberculosis, Pseudomonas aeruginosa, Helicobacter pylori, hepatitis C virus and influenza virus. Moreover, these cells contribute to autoimmune and inflammatory diseases, including colitis, rheumatoid arthritis, psoriasis, lupus, and diabetes.

Citation: Iwabuchi, K., Van Kaer, L., eds. (2019). Role of CD1- and MR1-restricted T cells in Immunity and Disease. Lausanne: Frontiers Media. doi: 10.3389/978-2-88963-122-3

# Table of Contents

*06 Editorial: Role of CD1- and MR1-Restricted T Cells in Immunity and Disease*

Kazuya Iwabuchi and Luc Van Kaer

*12 Recruitment of MAIT Cells to the Intervillous Space of the Placenta by Placenta-Derived Chemokines*

Martin Solders, Laia Gorchs, Eleonor Tiblad, Sebastian Gidlöf, Edwin Leeansyah, Joana Dias, Johan K. Sandberg, Isabelle Magalhaes, Anna-Carin Lundell and Helen Kaipe

*28 T Cell Receptor Expression Timing and Signal Strength in the Functional Differentiation of Invariant Natural Killer T Cells*

Nyambayar Dashtsoodol, Sabrina Bortoluzzi and Marc Schmidt-Supprian


Maroua Haroun Ferhat, Aurélie Robin, Louise Barbier, Antoine Thierry, Jean-Marc Gombert, Alice Barbarin and André Herbelin

*64 The Pathophysiological Relevance of the iNKT Cell/Mononuclear Phagocyte Crosstalk in Tissues*

Filippo Cortesi, Gloria Delfanti, Giulia Casorati and Paolo Dellabona

*75 iNKT Cells Suppress Pathogenic NK1.1+CD8+ T Cells in DSS-Induced Colitis*

Sung Won Lee, Hyun Jung Park, Jae Hee Cheon, Lan Wu, Luc Van Kaer and Seokmann Hong


Dorian Stolk, Hans J. van der Vliet, Tanja D. de Gruijl, Yvette van Kooyk and Mark A. Exley

*119 Complex Network of NKT Cell Subsets Controls Immune Homeostasis in Liver and Gut*

Idania Marrero, Igor Maricic, Ariel E. Feldstein, Rohit Loomba, Bernd Schnabl, Jesus Rivera-Nieves, Lars Eckmann and Vipin Kumar

*126 Clinical Application of iNKT Cell-mediated Anti-tumor Activity Against Lung Cancer and Head and Neck Cancer*

Mariko Takami, Fumie Ihara and Shinichiro Motohashi

*132 Type II NKT Cells: An Elusive Population With Immunoregulatory Properties*

Avadhesh Kumar Singh, Prabhanshu Tripathi and Susanna L. Cardell

*140 Role of CD1d- and MR1-Restricted T Cells in Asthma* Chiaki Iwamura and Toshinori Nakayama *153 Tissue-Specific Roles of NKT Cells in Tumor Immunity* Masaki Terabe and Jay A. Berzofsky *164 Natural Killer T Cells and Mucosal-Associated Invariant T Cells in Lung Infections* François Trottein and Christophe Paget *181 Increased iNKT17 Cell Frequency in the Intestine of Non-Obese Diabetic Mice Correlates With High* Bacterioidales *and Low* Clostridiales *Abundance* Lorena De Giorgi, Chiara Sorini, Ilaria Cosorich, Roberto Ferrarese, Filippo Canducci and Marika Falcone *188 Invariant Natural Killer T and Mucosal-Associated Invariant T Cells in Asthmatic Patients* Guillaume Lezmi and Maria Leite-de-Moraes *197 Linking CD1-Restricted T Cells With Autoimmunity and Dyslipidemia: Lipid Levels Matter* Sreya Bagchi, Samantha Genardi and Chyung-Ru Wang *209 Factors Influencing Functional Heterogeneity in Human Mucosa-Associated Invariant T Cells* Joana Dias, Caroline Boulouis, Michał J. Sobkowiak, Kerri G. Lal, Johanna Emgård, Marcus Buggert, Tiphaine Parrot, Jean-Baptiste Gorin, Edwin Leeansyah and Johan K. Sandberg *216 CD1d-Invariant Natural Killer T Cell-Based Cancer Immunotherapy:* a*-Galactosylceramide and Beyond* Lisa A. King, Roeland Lameris, Tanja D. de Gruijl and Hans J. van der Vliet *223 It Takes "Guts" to Cause Joint Inflammation: Role of Innate-Like T Cells* Céline Mortier, Srinath Govindarajan, Koen Venken and Dirk Elewaut *231 Insights Into Mucosal-Associated Invariant T Cell Biology From Studies of Invariant Natural Killer T Cells* Lucy C. Garner, Paul Klenerman and Nicholas M. Provine *256 How Lipid-Specific T Cells Become Effectors: The Differentiation of iNKT Subsets* Haiguang Wang and Kristin A. Hogquist *269 Invariant Natural Killer T Cell Subsets—More Than Just Developmental Intermediates* S. Harsha Krovi and Laurent Gapin *286 Latent* Mycobacterium tuberculosis *Infection is Associated With a Higher Frequency of Mucosal-Associated Invariant T and Invariant Natural Killer T Cells* Dominic Paquin-Proulx, Priscilla R. Costa, Cassia G. Terrassani Silveira, Mariana P. Marmorato, Natalia B. Cerqueira, Matthew S. Sutton, Shelby L. O'Connor, Karina I. Carvalho, Douglas F. Nixon and Esper G. Kallas *295 Activation and Regulation of B Cell Responses by Invariant Natural Killer T Cells*

Derek G. Doherty, Ashanty M. Melo, Ana Moreno-Olivera and Andreas C. Solomos


Yoon Jeong Park, Jeu Park, Jin Young Huh, Injae Hwang, Sung Sik Choe and Jae Bum Kim


Günther Schönrich and Martin J. Raftery

*398 CD1d-Restricted Type II NKT Cells Reactive With Endogenous Hydrophobic Peptides*

Yusuke Nishioka, Sakiko Masuda, Utano Tomaru and Akihiro Ishizu


# Editorial: Role of CD1- and MR1-Restricted T Cells in Immunity and Disease

Kazuya Iwabuchi <sup>1</sup> \* and Luc Van Kaer <sup>2</sup> \*

*<sup>1</sup> Department of Immunology, Kitasato University School of Medicine, Sagamihara, Japan, <sup>2</sup> Department of Pathology, Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN, United States*

Keywords: CD1, MR1, natural killer T cells, mucosal-associated invariant T cells, non-peptide antigens, immunity, immune-mediated disease

**Editorial on the Research Topic**

#### **Role of CD1- and MR1-Restricted T Cells in Immunity and Disease**

Products encoded by the major histocompatibility complex (MHC) present peptide antigens to T lymphocytes. MHC class I and class II products present peptide antigens to CD8<sup>+</sup> or CD4<sup>+</sup> T lymphocytes, respectively. In addition to these classical MHC class I molecules (also known as MHC class Ia molecules), numerous non-classical MHC class I molecules (also known as MHC class Ib molecules) have been identified (1). Some of these products are encoded by genes within the MHC, whereas others are encoded by genes outside the MHC. Key differences between class Ia and class Ib molecules are that the former are highly polymorphic, are expressed by most cell types, and present pathogen-derived peptides to CD8<sup>+</sup> T cells, whereas the latter are oligomorphic, are less widely expressed, and are diverse in their capacity to present antigens to subsets of T lymphocytes with innate or adaptive immune functions. Some class Ib molecules such as mouse Qa-1 (HLA-E in human) and H-2M3 bind peptide antigens, others such as CD1 (cluster of differentiation-1) and MR1 (MHC-related protein-1) bind non-peptide antigens, and again others such as the mouse TL (thymus leukemia) and the human MIC (MHC class I polypeptide-related sequence) antigens are not known to bind antigen of any kind. This Research Topic focuses on the immunological functions of T lymphocytes restricted by CD1 or MR1 molecules, which present lipids or small metabolites, respectively, to subsets of T lymphocytes with conventional or unconventional functions [see (2–5) for a general review of these topics].

The CD1 family of antigen-presenting molecules has been divided into two groups: Group 1 contains CD1a, CD1b, and CD1c, and Group 2 contains CD1d (6). Additionally, human CD1e is expressed intracellularly and is involved in the loading of lipid antigens onto the other CD1 isotypes, which are all displayed at the cell surface. Humans express both Groups 1 and 2 CD1 proteins, whereas mice only express Group 2 CD1 proteins (i.e., CD1d). Group 1 CD1 proteins present lipid antigens to T cells that generally express diverse T cell receptors (TCRs) and exhibit adaptive-like functions, whereas CD1d presents glycolipid antigens to subsets of T cells that express restricted TCRs and exhibit innate-like functions (7, 8). CD1-restricted T cells have been implicated in a wide variety of diseases, including cancer, infections, and autoimmune, inflammatory and metabolic diseases.

Seminal studies with human autoreactive T cell lines led to the identification of lipid-reactive T cells restricted by CD1 molecules (9). Subsequent studies identified CD1-restricted T cells specific for microbial lipids, especially those derived from the cell wall of mycobacterial species (10, 11). While some of the initial T cell clones isolated expressed γδ TCRs, it is now clear that the majority of CD1-reactive T cells express αβ TCRs. The antigen-binding groove of CD1 is hydrophobic and

#### Edited and reviewed by:

*Hongbo Chi, St. Jude Children's Research Hospital, United States Lewis Zhichang Shi, University of Alabama at Birmingham, United States*

#### \*Correspondence:

*Kazuya Iwabuchi akimari@kitasato-u.ac.jp Luc Van Kaer luc.van.kaer@vanderbilt.edu*

#### Specialty section:

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

Received: *19 June 2019* Accepted: *22 July 2019* Published: *06 August 2019*

#### Citation:

*Iwabuchi K and Van Kaer L (2019) Editorial: Role of CD1- and MR1-Restricted T Cells in Immunity and Disease. Front. Immunol. 10:1837. doi: 10.3389/fimmu.2019.01837*

**6**

differs in size between distinct CD1 isoforms, which permits these molecules to bind diverse lipid or glycolipid antigens (7, 12). Some lipids, such as the self-antigen sulfatide, can bind with all CD1 isoforms. Each of the CD1 isoforms can present mycobacterial lipid antigens, albeit distinct in structure, to CD1 restricted T cells. Tetrameric CD1 molecules loaded with lipid antigens have been generated for all CD1 isotypes and have greatly assisted in understanding the biology of CD1-restricted T cells (5).

Most Group 1 CD1-restricted T cell clones described are autoreactive, and likely represent a substantial part of the T cell repertoire in humans, estimated at 1/300 to 1/10 circulating T cells (13). The majority of these T cell clones express diverse αβ TCRs, although subsets of Group 1 CD1-restricted T cells expressing conserved TCRs have also been identified (14). As argued in the review article by Bagchi et al., emerging evidence indicates that Group 1 CD1-restricted T cells can contribute to autoimmunity, especially in diseases associated with dyslipidemia. For example, CD1c-restricted T cells have been implicated in progression of systemic lupus erythematosus (SLE). Additionally, CD1a-restricted autoreactive T cells have been implicated in psoriasis and the response to insect venoms. Since Group 1 CD1-restricted T cells recognizing mycobacterial lipids are expanded in patients infected with Mycobacterium tuberculosis or vaccinated with Bacillus Calmette-Guérin, it is likely that these cells exhibit adaptive-like properties and play a protective role during infection. A major challenge in studying the functions of Group 1 CD1 molecules is their absence from mice. Some researchers have therefore resorted to analyzing CD1-restricted T cell responses in guinea pigs, which express several CD1 isoforms. Guinea pigs induce Group 1 CD1-restricted T cell responses to mycobacterial lipids (15), suggesting adaptive-like functions of these cells. Other studies have employed a humanized CD1 transgenic mouse model (16) or humanized immunodeficient mice engrafted with human fetal thymic tissue and hematopoietic stem cells (17). These studies similarly revealed that M. tuberculosis-reactive Group 1 CD1-restricted T cells exhibit adaptive-like functions, suggesting promise for the development of lipid-based vaccines for mycobacterial and other infections. Studies with human CD1 transgenic mice also suggested a role of CD1b-restricted T cells in the generation of skin inflammation (18). Thus, as discussed in the review article by Lepore et al., the available evidence indicates that Group 1 CD1-restricted T cells share many functional properties with conventional, MHC class I-restricted T cells. Group 1 CD1-restricted T cells also hold promise for immunotherapy, as illustrated by studies showing that a subset of CD1c-restricted T cells exhibits strong reactivity for methyllysophosphatidic acid, a lipid antigen produced by leukemic cells (19). When adoptively transferred into immunodeficient mice, such T cells were able to limit the spread of CD1cexpressing human leukemia cells (19). Additionally, the CD1apresented skin antigen, squalene (20), is a widely used adjuvant that enhances vaccine efficacy. Other antigens presented by Group 1 CD1 molecules could be similarly employed to modulate CD1-restricted T cell responses during vaccination or immunotherapy.

CD1d-restricted T cells are also called natural killer T (NKT) cells, which includes Type I or invariant NKT (iNKT) cells expressing semi-invariant TCRs, and Type II or diverse NKT (dNKT) cells expressing more diverse or oligoclonal TCRs (21). NKT cells have been targeted for immunotherapy of disease with ligands such as α-galactosylceramide (α-GalCer) for iNKT cells (22), or sulfatide for dNKT cells (23).

iNKT cells were originally identified as a subset of T cells expressing an invariant TCR in mice and humans (11, 24). The reactivity of these cells was subsequently shown to be restricted by CD1d (25). iNKT cells are highly conserved between mice, humans and other mammalian species (26). The iNKT TCR is made of Vα14-Jα18 and Vβ8.2/7/2 chains in mice or homologous Vα24-Jα18 and Vβ11 chains in humans (2, 3, 21). In mice, iNKT cells are most abundant in liver and adipose tissue, where they can represent up to 30–40% of the T cell population, are present at significant numbers in peripheral blood, spleen, thymus, and bone marrow (1–5% of all mature T cells), and are also found in lymph nodes, skin, and mucosal surfaces in the intestine and lungs (<1% of all T cells). However, these cells are quite rare and have variable frequency in humans (e.g., 0.005–0.2% of all T cells in peripheral blood). A variety of endogenous and exogenous lipid antigens for these cells have been identified (7, 8). All iNKT cells react with α-GalCer (27), an optimized version of a natural product originally isolated from a marine sponge, and α-GalCer-loaded CD1d tetramers selectively bind with iNKT cells. As described in the review by Ren et al., genetic tools to study iNKT cells include Jα18- and CD1d-deficient mice, as well as mouse Vα14-Jα18 and human Vα24-Jα18 TCR transgenic animals. iNKT cells express markers that are characteristic of the natural killer (NK) cell lineage, exhibit an activated or memory phenotype, are enriched in liver and mucosal tissues, have a tendency for autoreactivity, and are unable to generate classical memory responses. In addition to TCR-mediated activation signals, these cells are highly responsive to innate immune activation signals, such as innate cytokines produced in response to toll-like receptor (TLR) signaling in antigen-presenting cells (28). iNKT cells often become activated during situations of sterile inflammation involving release of damage-associated molecular patterns, also called alarmins. As discussed in the perspective by Ferhat et al., the alarmin IL-33 might play a key role in recruiting iNKT cells to the inflammatory site and in enhancing their regulatory and effector functions. iNKT cells can produce a variety of pro- and anti-inflammatory cytokines, and subsets of iNKT cells biased for production of type 1, 2, or 3 cytokines, called iNKT1, iNKT2 and iNKT17 cells, respectively, are enriched in different tissues (29). iNKT cells acquire these unusual properties during their development in the thymus, as discussed in the reviews by Krovi and Gapin,Wang and Hogquist, Yang et al., and Dashtsoodol et al.. A key driver in the acquisition of an innate effector program by these cells is the induction of the transcription factor promyelocytic leukemia zinc finger (PLZF) during thymic selection on CD1d-expressing CD4+CD8<sup>+</sup> (double-positive) thymocytes (30, 31). Additionally, iNKT cells undergo unique epigenetic modifications in the thymus that influence their development, maturation and functional differentiation. Following the activation of iNKT cells by lipids and/or innate signals, these cells can activate and modulate a variety of innate and adaptive immune cells and thus impact overall immune responses and disease (32). The review article by Cortesi et al. discusses how reciprocal interactions between iNKT cells and mononuclear phagocytes in tissues impact physiological and pathological conditions. Additionally, the review articles by Lang, Doherty et al., and Fujii et al. discuss how iNKT cells can interact with B cells to influence humoral immune responses, and Fujii et al. further review the capacity of iNKT cells to license dendritic cells to augment humoral and cellular immune responses.

Consistent with their potent immunomodulatory properties, iNKT cells contribute to protective immune responses against a variety of microorganisms, as discussed in the review articles by Kinjo et al., Trottein and Paget, and Huang et al.. For example, the original research article by Paquin-Proulx et al. provides evidence for dynamic regulation of iNKT cell numbers and functions during M. tuberculosis infection, suggesting a role for these cells in immunity against this organism. Interestingly, several microbial organisms, especially viruses that cause persistent infections, can evade CD1d-restricted immune responses (33), a topic discussed in the review article by Schönrich and Raftery. One particularly clever way to evade iNKT cell responses is illustrated by human immunodeficiency virus (HIV), which can infect iNKT cells, resulting in their depletion during disease progression (34). The original research article by Singh et al. provides evidence that patients with non-progressive HIV infection contain functionally competent iNKT cells, whereas anti-retroviral treatment of patients with progressive infection fails to restore iNKT cell functions.

iNKT cells have been implicated in natural immunity against tumors, a topic discussed by Terabe and Berzofsky and Stolk et al.. These cells also contribute to a variety of autoimmune and inflammatory diseases (28). The review by Bagchi et al. focuses on the role of iNKT cells in autoimmune diseases such as SLE, psoriasis, diabetes, and rheumatoid arthritis that are characterized by dyslipidemia. iNKT cells also play a role in the generation of inflammatory bowel disease, as illustrated by the original research article by Lee et al., which provides evidence for a role of iNKT cells in suppressing disease in a model of colitis mediated by pathogenic CD8<sup>+</sup> T cells. The review articles by Lezmi and Leite-de-Moraes, Iwamura and Nakayama, and Ryu et al. focus on the role of lung iNKT cells in asthma and other pulmonary disorders. These cells also contribute to the pathogenesis of alcoholic and non-alcoholic fatty liver disease, which is discussed in the review articles by Huang et al. and Marrero et al.. The review article by Ververs et al. describes the response of iNKT cells to metabolic activation in diseases such as atherosclerosis and obesity. Several review articles, including those by Park et al., Satoh and Iwabuchi, and Ren et al., discuss the role of iNKT cells in the development of obesity and obesity-associated insulin-resistance, which is influenced by many different factors (35, 36). The divergent findings obtained in different laboratories for the role of iNKT cells in obesityassociated conditions might be caused in part by different models of iNKT cell-deficiency employed, a possibility discussed in the review article by Ren et al.. Additionally, differences in the microbiota present in different mouse colonies might play a role, as the functional status of iNKT cells is impacted by microbiota (37–39). The topic of functional alterations in iNKT cells imparted by gut microbiota is discussed by Marrero et al. in the context of liver diseases and by Mortier et al. in the context of rheumatic diseases. Additionally, the original research article by De Giorgi et al. provides evidence that the abundance of certain microbiota in the gut of diabetes-prone non-obese diabetic mice increases the intestinal iNKT17 subset in these animals, possibly contributing to disease pathogenesis.

The immunomodulatory properties of iNKT cells have been exploited for the development of disease prophylaxis and therapy (22). Many of these studies have been performed with the iNKT cell ligand α-GalCer or its structural analogs. The adjuvant activities of iNKT cells have been exploited for the development of vaccines against microbes and tumors, as discussed in the review articles by Lang, Doherty et al., Fujii et al., and Kinjo et al.. As discussed by Kinjo et al., glycolipid activation of iNKT cells during infection can enhance antimicrobial immunity and hasten microbial clearance. Many studies have explored the therapeutic activities of iNKT cells against tumors and, as discussed in the review articles by Takami et al. and King et al., multiple clinical trials have already been performed. The therapeutic potential of iNKT cells in autoimmune diseases is discussed by Van Kaer and Wu.. In addition to these glycolipid antigen-mediated methods to elicit the therapeutic properties of iNKT cells, diseases where iNKT cells play a pathogenic role might benefit from iNKT cell depletion, a possibility that is being explored in sickle cell disease with antibodies directed against the invariant iNKT TCR (40).

dNKT cells express a more diverse TCR repertoire than iNKT cells (21, 23). In mice, dNKT cells are less prevalent than iNKT cells, whereas in humans dNKT cells outnumber iNKT cells. dNKT cells interact with multiple different lipid antigens, and a prominent subset reacts with the self-antigen sulfatide. While the sulfatide-reactive subset can be identified with sulfatide-loaded CD1d tetramers, rigorous identification of the entire dNKT cell population remains challenging. Curiously, as discussed in the review article by Nishioka et al., a subset of dNKT cells can react with hydrophobic peptides, including a peptide derived from type II collagen. As discussed in the review by Ren et al., the functions of dNKT cells are often deduced from studies comparing CD1d-deficient mice lacking both iNKT and dNKT cells with Jα18-deficient mice lacking only iNKT cells. dNKT cells share with iNKT cells expression of NK and memory markers, response to innate activation signals, innatelike effector functions, and absence of immunological memory. As discussed in the review article by Singh et al., dNKT cells can play both protective and pathogenic roles in a variety of diseases. Several review articles included in this Research Topic discuss the roles of dNKT cells in specific diseases. Trottein and Paget focus on lung infections, Terabe and Berzofsky and Stolk et al. on cancer, Iwamura and Nakayama on asthma, Satoh and Iwabuchi on obesity and insulin resistance, and Marrero et al. on inflammatory diseases of the liver and gut. In some conditions, such as cancer and steatohepatitis, dNKT appear

to play opposing roles to iNKT cells (41). Emerging evidence, discussed in the review article by Marrero et al., has revealed substantial cross-regulation between iNKT and dNKT cells.

Sulfatide has been employed to target the sulfatide-reactive subset of dNKT cells for therapeutic purposes (23). As discussed in the review articles by Singh et al. and Marrero et al., sulfatide has beneficial effects in some infections and models of autoimmunity and liver inflammation. dNKT cells reactive with collagen type II peptide have also been targeted therapeutically (using cognate peptide), employing an experimental model of rheumatoid arthritis, as discussed in the review article by Nishioka et al..

The antigen-binding groove of MR1 is lined with aromatic residues, which facilitates binding with microbial vitamin B metabolites and similar compounds (42). The majority of MR1 restricted T cells express semi-invariant TCRs (43). Such cells are called mucosal-associated invariant T (MAIT) cells, were first identified as a clonally enriched T cell population in humans and mice (44, 45), and were subsequently shown to be restricted by MR1 (46) and to react with vitamin B metabolites (47). MAIT cells express Vα19-Jα33 chains in mice and homologous Vα7.2- Jα33 chains in humans, and these cells can be identified by staining with vitamin B metabolite-loaded MR1 tetramers. In humans, they represent ∼3–5% of T cells in peripheral blood and intestine and up to 40% of T cells in liver. However, these cells are very rare in laboratory mice (<1% of all T cells). In addition to vitamin B metabolites, MAIT cells can react with a range of drugs and drug-like molecules, which can either activate or suppress these cells (48). Genetic tools to study MAIT cells include MR1- and Jα33-deficient mice and transgenic mice carrying the invariant Vα19-Jα33 TCR. As discussed in the review article by Garner et al., many of the studies on MAIT cells have been guided by the available biology on iNKT cells. MAIT cells share with iNKT cells an effector-memory phenotype, enrichment in liver and mucosal sites, expression of PLZF that drives an innate effector phenotype, response to innate signals, and the absence of classical immune memory. The antigen-specificity and activation requirements of MAIT cells are discussed in the review article by Xiao and Cai. Emerging evidence, discussed in the review article from Dias et al., has revealed functional heterogeneity within the MAIT population with regard to cytokine production, akin to the iNKT1 and iNKT17 subsets of iNKT cells (49). MAIT cells have been implicated in immune responses against a variety of pathogens, including but not limited to organisms containing vitamin B metabolites, a topic discussed in the review article by Trottein and Paget for lung infections. For example, the original research article by Paquin-Proulx et al. reports increased frequency of MAIT cells in individuals infected latently with M. tuberculosis, an organism that contains an intact riboflavin synthesis pathway, suggesting a potential role in disease pathogenesis. Moreover, MAIT cells can infiltrate tumors, and their potential role in anti-tumor immunity is discussed in the review article by Stolk et al.. The original research article by Solders et al. shows that MAIT cells are recruited to the intervillous space of the placenta, where they might play a role in fetal-maternal interactions. The review articles by Lezmi and Leite-de-Moraes and Iwamura and Nakayama discuss the potential role of MAIT cells in allergic asthma. The review article by Huang et al. discusses the role of these cells in inflammatory diseases of the liver, and the review article by Chiba et al. discusses their functions in animal models of autoimmune and inflammatory diseases. Finally, like iNKT cells, the development and functions of MAIT cells are substantially modulated by the presence of microbiota, which has a significant impact on disease, a topic discussed in the review article by Mortier et al. for rheumatic diseases. Although not yet fully explored, MAIT cells hold significant potential for therapeutic targeting, using vitamin B metabolites or drug-like molecules that could be employed to either potentiate or suppress MAIT cell responses.

Recent studies have identified subsets of MR1-restricted T cells that do not express the invariant MAIT TCR (42, 50). Some of these cells react with riboflavin-based antigens and exhibit innate-like effector functions, whereas others have a distinct antigen-reactivity and exhibit characteristics of adaptive immune effector functions, as discussed in the review article by Lepore et al.. Thus, these findings suggest additional similarities in the biology of MR1- and CD1d-restricted T cells.

The articles included in this Research Topic illustrate the breath of immune responses regulated by CD1- and MR1 restricted T cells, which include subsets with innate- and adaptive-like properties. An attractive property of the CD1 and MR1 antigen presentation systems, as compared with the classical MHC antigen presentation system, is that they can be more easily targeted in vaccines and immunotherapies. CD1 and MR1 exhibit limited polymorphism and are recognized by T cells that include subsets expressing conserved (i.e., public) TCRs (14). As such, it is feasible to target these cells across the genetically diverse human population in an antigen-specific manner. A better understanding of the basic biology of these cells and their role in different disease processes, together with the development of improved tools to target them, should facilitate the realization of this therapeutic promise.

#### AUTHOR CONTRIBUTIONS

KI and LV served as co-editors for the Research Topic and edited the manuscript. LV wrote the first draft.

# FUNDING

Work in the authors' labs was supported by the Japan Society for Promotion of Science (to KI), the Japan Agency for Medical Research and Development (to KI), the Japan Ministry of Education, Culture, Sports, Science and Technology (to KI), the National Institutes of Health (to LV), the National Multiple Sclerosis Society (to LV), and the American Heart Association (to LV).

#### ACKNOWLEDGMENTS

We are grateful to all colleagues who contributed and/or reviewed manuscripts for this Research Topic.

# REFERENCES


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

Copyright © 2019 Iwabuchi and Van Kaer. 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.

# Recruitment of MAIT Cells to the Intervillous Space of the Placenta by Placenta-Derived Chemokines

Martin Solders 1,2 \*, Laia Gorchs <sup>2</sup> , Eleonor Tiblad<sup>3</sup> , Sebastian Gidlöf 3,4, Edwin Leeansyah5,6 , Joana Dias <sup>5</sup> , Johan K. Sandberg<sup>5</sup> , Isabelle Magalhaes <sup>7</sup> , Anna-Carin Lundell <sup>8</sup> and Helen Kaipe1,2 \*

<sup>1</sup> Clinical Immunology and Transfusion Medicine, Karolinska University Hospital, Stockholm, Sweden, <sup>2</sup> Department of Laboratory Medicine, Karolinska Institutet, Stockholm, Sweden, <sup>3</sup> Center for Fetal Medicine, Karolinska University Hospital and Department of CLINTEC, Karolinska Insitutet, Stockholm, Sweden, <sup>4</sup> Department of Women's and Children's Health, Karolinska Institutet, Stockholm, Sweden, <sup>5</sup> Center for Infectious Medicine, Department of Medicine, Huddinge, Karolinska Institutet, Stockholm, Sweden, <sup>6</sup> Program in Emerging Infectious Diseases, Duke-National University of Singapore Medical School, Singapore, Singapore, <sup>7</sup> Department of Oncology/Pathology, Karolinska Institutet, Stockholm, Sweden, <sup>8</sup> Department of Rheumatology and Inflammation Research, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden

#### Edited by:

Kazuya Iwabuchi, Kitasato University School of Medicine, Japan

#### Reviewed by:

Mariolina Salio, University of Oxford, United Kingdom Marco Lepore, Immunocore, United Kingdom

> \*Correspondence: Martin Solders martin.solders@ki.se Helen Kaipe helen.kaipe@ki.se

#### Specialty section:

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

Received: 30 January 2019 Accepted: 22 May 2019 Published: 06 June 2019

#### Citation:

Solders M, Gorchs L, Tiblad E, Gidlöf S, Leeansyah E, Dias J, Sandberg JK, Magalhaes I, Lundell A-C and Kaipe H (2019) Recruitment of MAIT Cells to the Intervillous Space of the Placenta by Placenta-Derived Chemokines. Front. Immunol. 10:1300. doi: 10.3389/fimmu.2019.01300 The intervillous space of the placenta is a part of the fetal-maternal interface, where maternal blood enters to provide nutrients and gas exchange. Little is known about the maternal immune cells at this site, which are in direct contact with fetal tissues. We have characterized the T cell composition and chemokine profile in paired intervillous and peripheral blood samples from healthy mothers giving birth following term pregnancies. Mucosal-associated invariant T (MAIT) cells and effector memory (EM) T cells were enriched in the intervillous blood compared to peripheral blood, suggesting that MAIT cells and other EM T cells home to the placenta during pregnancy. Furthermore, pregnant women had lower proportions of peripheral blood MAIT cells compared to non-pregnant women. The levels of several chemokines were significantly higher in intervillous compared to peripheral blood, including macrophage migration inhibitory factor (MIF), CXCL10, and CCL25, whereas CCL21, CCL27 and CXCL12 were lower. Migration assays showed that MAIT cells and EM T cells migrated toward conditioned medium from placental explants. A multivariate factor analysis indicated that high levels of MIF and CCL25 were associated with high proportions of MAIT cells in intervillous blood. Blocking of MIF or a combination of MIF, CCL25, and CCL20 in migration assays inhibited MAIT cell migration toward placenta conditioned medium. Finally, MAIT cells showed migratory capacities toward recombinant MIF. Together, these findings indicate that term placental tissues attract MAIT cells, and that this effect is at least partly mediated by MIF.

Keywords: MAIT cells, placenta, chemokines, reproductive immunology, intervillous space, T cells, MIF

# INTRODUCTION

During pregnancy, one of the main functions of the placenta is to provide the growing fetus with oxygen and nutrients from the maternal blood circulation. The pregnant woman's arterial blood fills the intervillous space of the placenta, where it comes in direct contact with the fetal villi protruding from the fetal part of the placenta. Fetal blood vessels run inside the villi, and gas and nutrients

**12**

are exchanged over a thin membrane of fetal cytotrophoblasts and syncytiotrophoblasts (1). Another site for fetal-maternal interactions is the decidua, a maternal membrane reformed from the endometrium during pregnancy. The decidua is readily infiltrated by maternal immune cells that can interact with invasive fetal extravillous trophoblasts [reviewed in (2)].

The circulation of maternal blood in the intervillous space from a physiological perspective has been described (3). The intervillous blood (IVB) is exchanged 2–3 times per minute (1), suggesting that the cell composition in IVB may reflect that of peripheral blood (PB). However, our recent findings showed that mucosal-associated invariant T (MAIT) cells and effector memory (EM) T cells are enriched in the IVB of term placentas (4), indicating that certain immune cell subsets are recruited to or retained in the intervillous space. The potential factors involved in the migration of maternal immune cells to the placenta is still unexplored.

MAIT cells are an invariant type of T cell that respond to microbial derived metabolites from riboflavin synthesis presented by the non-polymorphic MHC class I related molecule (MR1) (5). The MAIT cell T cell receptor invariably uses the Vα7.2-segment coupled with either Jα33, 12 or 20, paired with a limited set of β chains (6–8). Besides the MR1- and T cell receptor-dependent activation, MAIT cells can be partially activated or co-stimulated by cytokines, including IL-7, IL-12, IL-15, and IL-18 (8–11). Thus, MAIT cells can respond in a T cell receptor-independent manner, broadening their relevance to viral infections and autoimmune disorders (12–15). Upon activation, MAIT cells secrete IFN-γ and TNF-α, (16, 17) and MAIT cells from the liver and the female genital tract also secrete measurable levels of IL-17 and IL-22 (18, 19). Activated MAIT cells upregulate the expression of the cytotoxic effector molecules granzyme B and perforin, and can kill infected target cells in an MR1-restricted manner (20). The importance of MAIT cells in pregnancy is unknown.

Placental tissue and fetal trophoblasts produce a wide array of chemokines in early pregnancy (21), and chemokines and their receptors play an instrumental role in trophoblast invasiveness, angiogenesis, and recruitment of immune cells to the decidua [reviewed in (22)]. Macrophage migration inhibitory factor (MIF) is a chemokine-like cytokine highly expressed in placental tissues (23, 24). MIF expression has been shown in syncytiotrophoblasts, cytotrophoblasts and extravillous trophoblasts (23, 25), but its function in pregnancy is unclear.

Since particular immune cell subsets are enriched in IVB compared to PB (4), we hypothesized that the placenta produces chemotactic mediators that are involved in attracting certain maternal immune cells to the intervillous space. Therefore, the main aim of this study was to examine if placental-derived factors are involved in recruiting MAIT cells to the placenta.

#### MATERIALS AND METHODS

#### Sample Collection

Healthy individuals (n = 36, median age 34, range 21–42) donated their placentas after informed consent, subsequent to planned cesarean sections following uncomplicated term pregnancies (median gestational week 39, range 38–42). The regional review board of ethics in research at Karolinska Institutet approved the donation of peripheral blood and placentas (entry numbers 2009/418-31/4, 2010/2061-32, and 2015/1848-31/2). Female blood donors were used as a control for non-pregnant women (n = 27, median age 46, range 22–71).

Data on MAIT cell frequencies from pregnant women has been published previously for 21 out of 36 donors (4). Parts of the T- and MAIT cell phenotype data were published previously for 11 out of 36 donors (4).

#### Cell Isolation

Peripheral blood (PB) samples were collected a few hours before the caesarian sections. The description of isolation of IVB and decidua parietalis has been published previously (4). Briefly, placentas were placed in a sterile container in the operation room, and then taken directly to the lab. After removal of the fetal membranes and clamping of the umbilical cord, the placentas were placed with the maternal side facing up. Visible blood clots were removed, and the surface was washed with phosphate buffered saline (PBS) to remove seeping blood. The placenta was turned around and lifted so that the maternal side faced downwards and IVB was collected on a sterile petri dish and immediately transferred to heparin tubes. The samples of PB and IVB were centrifuged at 600 g for 8 min after which plasma samples were collected and stored at −80◦C until use. The blood was diluted 1:2 with PBS, and mononuclear cells were isolated by density gradient centrifugation (Lymphoprep, Axis-Shield, Oslo, Norway). The fetal membranes were placed with the decidua facing up. After extensive washing and removal of visible blood clots, the decidua parietalis was mechanically scraped off from the chorion, pooled and washed with PBS by repeated short centrifugations. Cells were released from the tissue by using a GentleMACS Dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany). After filtering and washing, mononuclear cells were isolated by density gradient centrifugation (Lymphoprep, Axis-Shield, Dundee, Scotland).

Cells were either stained directly for flow cytometry, or resuspended in RPMI (HyClone, GE Health Sciences, South Logan, UT) medium supplemented with 10% fetal calf serum, 100 U/ml penicillin, 100µg/ml streptomycin (complete medium) containing 10% DMSO and frozen in aliquots in liquid nitrogen. In our previous publication, we found the isolated maternal cell product to have a median fetal DNA contamination of 12.1% (range 5.2–19.4, n = 6) (4).

### Flow Cytometry

IVB and PB samples were analyzed in pairs, either fresh or thawed after freezing. Frozen cells were thawed in complete medium, washed in PBS and counted, and fresh cells were used directly after isolation. The cells were plated in a 96-well plate, ≤ 1 × 10<sup>6</sup> cells/well. Staining was carried out in 50 µL of CliniMACS PBS/EDTA buffer (Miltenyi Biotech, Bergish Gladbach, Germany) supplemented with 0.1% bovine serum albumin (FACS-buffer) to which antibodies were added, and incubated for 30 min at 4◦C. After washing, the cells were stained with 7AAD, which was used to distinguish live from dead cells. The MR1 tetramers were produced by the NIH Tetramer Core Facility as permitted to be distributed by the University of Melbourne, and the MR1 tetramer technology was developed jointly by Dr. James McCluskey, Dr. Jamie Rossjohn, and Dr. David Fairlie. The flow cytometry antibodies used in this study are listed in **Supplementary Table S1**. Data were collected using a BD FACSCanto flow cytometer and analyzed with FlowJo software (Tree Star, Ashland, OR, USA). Data on conventional T cells were analyzed after excluding CD161<sup>+</sup> and Vα7.2<sup>+</sup> T cells.

#### Placenta Conditioned Medium

Placental explant conditioned medium were isolated using a slightly modified version of the method described for 1st trimester placentas by Svensson-Arvelund et al. (21). Following extraction of IVB, the placenta was placed with the maternal side facing up. Using scissors, the decidua basalis was removed, and from the underlying tissue, biopsies of approximately 1 cm<sup>3</sup> were cut. Between 5 and 10 biopsies were taken from different places of the placenta. The tissue biopsies were then pooled and dissected into as small pieces as possible using forceps and a scalpel. Macroscopically identifiable blood vessels were removed. The dissected tissue was then washed extensively with PBS by repeated cycles of short centrifugations until the PBS was no longer colored by the tissue, usually 10–20 times. The tissue was then placed in 24- (50–100 mg of tissue) or 6-well plates (250–500 mg of tissue), and 10 µL of complete medium/mg of tissue was added. Multiple plates were set up at the same time, and incubated at 37◦C for 48 and 72 h. At the end of incubation, the content of multiple wells was harvested and pooled, and following centrifugation (600 g, 8 min), supernatants were collected and frozen in aliquots at −80◦C.

#### Multiplex Chemokine Assay

Paired frozen plasma samples from IVB and PB (n = 25) and placental explant supernatants (n = 5) were analyzed for cytokine and chemokine concentrations using a Magpix with Luminex Xponent software (Luminexcorp, Austin, TX, USA) and the Bio-Plex Pro Human Chemokine 40-Plex kit (BIO-RAD, Hercules, CA, USA) according to the manufacturer's instructions. The detection level of each analyte is shown in **Supplementary Table S2**.

#### Migration Assay

T cells were purified from PBMCs from healthy donors by negative selection (Pan T Cell Isolation Kit, Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's instructions. Placenta conditioned medium from 48 or 72 h cultures was thawed, diluted 50% in complete medium, and 600 µL was added to a 24-well plate. Complete medium only was used as a negative control. Blocking reagents against CCL20 (clone 67310, 1µg/ml, R&D Systems, Minneapolis, MN, USA), CCL25 (clone 52513, 3µg/ml, R&D) and/or MIF (ISO-1, 100µM, Tocris Bioscience, Bristol, UK) or the corresponding isotype/diluent was then added to the wells. ISO-1 was diluted in PBS with 0.1% DMSO. For the MIF migration assay, RPMI medium without serum containing 100 or 250 ng/ml recombinant MIF (R&D Systems) were added to the lower well. The medium control well contained only RPMI without serum. Ten of the experiments were performed using purified T cells (isolated as described above). Six samples were performed by using purified T cells that were also depleted of CD4<sup>+</sup> T cells, using CD4<sup>+</sup> T cell isolation beads (Miltenyi), to enrich CD8<sup>+</sup> T cells. Transwell inserts (5µm pore size, Costar, Cambridge, MA, USA) were placed into each well, and 1 × 10<sup>6</sup> T cells resuspended in 100 µL RPMI without serum were then added on top of the transwell insert. Plates were incubated at 37◦C for 3 h. The transwell inserts were removed, and the content of each well was harvested, transferred to a 15 ml tube, centrifugated (600 g, 8 min), re-suspended in FACS-buffer, and stained for flow cytometry. Careful measures were taken to treat all paired samples equally; they were re-suspended in the same volume of FACS-buffer, washed in parallel, and run for the same time in the flow cytometer. Results are presented as cell counts or proportions within the CD3<sup>+</sup> or CD8<sup>+</sup> T cell populations as indicated in the figures.

#### Statistical Analysis

We used multivariate orthogonal projection to latent structures (OPLS) analyses as a tool to screen for differences between PB and IVB and associations between specific immune cell subsets and chemokine levels. Variables that showed the strongest association to classes or variables were further evaluated by univariate analysis. Multivariate orthogonal projection to latent structures by means of partial least squares discriminant analysis (OPLS-DA) was used to obtain a maximum separation of X-variables (T cell phenotypes or chemokine levels) based on class information (peripheral and intervillous blood, i.e., PB and IVB) (SIMCA software, Sartorius Stedim Biotech, Umeå, Sweden). The twoclass (PB and IVB) discrimination OPLS-DA model presented in **Figure 1A** is a default score scatter observation plot with one predictive component t[1] and one orthogonal component to [1]. The scatter plot of t[1] vs. to[1] is a window in the X space in which the separation of the two classes of observations occurs in the horizontal t[1] direction and the within class variability is expressed in the vertical to[1] direction. Linear OPLS analysis was implemented to investigate associations between a selected Y-variable (proportion of MAIT cells) and a set of X-variables (chemokine levels in IVB). The contribution of each X-variable, variable influence of projection (VIP) values, to the OPLS models were calculated. For the analysis of T cell variabilities, X-variables with a VIP value below 0.81 were excluded and a new model was generated based on remaining variables. The scale presented on the x and y-axis in the OPLS-DA plot (**Figure 1A**) and on the y-axis of the OPLS plots (**Figures 1B**, **4A**) is a dimensionless scale in which the loading vector is normalized to unit length. The validation of OPLS analyses is based on the goodness of fit, or R2, which tells how well the data set can be mathematically reproduced. The goodness of predictivity, or Q2, tells how well the model can predict future data of a test set. Cross validation is performed by SIMCA software to estimate Q2. Briefly, cross validation is performed by dividing the data set into a number of groups and then developing a number of parallel models from the reduced data with one of the groups deleted. After developing a model, the deleted data are used as a test set, and differences between actual and predicted values are calculated for the test set, eventually leading forward to the Q2-value. Logarithmic values were used in the multivariate analysis. The factors contributing most to the separation were further analyzed using the twotailed Wilcoxon matched-pairs signed rank test. To investigate correlations between two factors, the Spearman's rank correlation test was used. To detect differences across two groups of unpaired samples, the two-tailed Mann-Whitney test was used (GraphPad Software, La Jolla, CA). A P of < 0.05 was considered significant.

#### RESULTS

### T Cell Subsets and Phenotypes Are Different in Intervillous Compared to Peripheral Blood

By using multivariate discriminant analysis, we first examined the T cell profiles in PB and IVB based on 53 parameters, which are all displayed in **Supplementary Figure S1**. As depicted in the observation plot in **Figure 1A**, a clear separation was found between PB and IVB based on the T cell variables assessed. The X-variables that showed the strongest associations with PB or IVB are shown in the loading plot in **Figure 1B**, a model based on X-variables with VIP values ≥ 0.81. A VIP column plot for all analyzed parameters is shown in **Supplementary Figure S1**. Further investigating the two groups using univariate statistical analysis, we found several significant differences in the T cell compartment between PB and IVB. The proportions of CD3<sup>+</sup> T cells within CD45<sup>+</sup> cells were similar in PB and IVB, but IVB had higher proportions of CD8<sup>+</sup> and CD4 and CD8 double negative (DN) T cells compared to PB (**Figure 1C**). The proportion of EM CD8<sup>+</sup> T cells was a median 2.1-fold higher in IVB compared to PB, with a concomitant lower proportion of naïve T cells (**Figure 1D**). A similar pattern was observed among CD4<sup>+</sup> T cells, with a 1.7-fold higher proportion of EM cells in IVB compared to PB (data not shown). CD8<sup>+</sup> T cells from IVB were more activated compared to PB CD8<sup>+</sup> T cells, as shown by an increased expression of CD69 and HLA-DR (**Supplementary Figure S2A**). IVB CD8<sup>+</sup> T cells also expressed the co-inhibitory markers PD-1 and TIM-3 to a larger degree compared to PB CD8<sup>+</sup> T cells (**Supplementary Figure S2A**). Taken together, this shows that maternal T cells in the intervillous space to a larger degree are activated and of an EM phenotype compared to their peripheral counterparts. Representative flow cytometry plots are shown in **Supplementary Figure S3**.

## Enrichment of MAIT Cells in Intervillous Blood

As previously shown (4), but now with a larger number of donors, MAIT cells were enriched in IVB both within CD3<sup>+</sup> T cells and within CD45<sup>+</sup> leukocytes, with a median 2.1-fold and 2.3-fold higher proportion, respectively, in IVB compared to PB (**Figure 1E**). This suggests that MAIT cells accumulate in the intervillous space of the placenta during pregnancy. The proportion of MAIT cells in PB from the cohort of pregnant women was compared to that of healthy non-pregnant women. Despite the lower median age of the pregnant compared to the non-pregnant women (34 vs. 46 years, respectively), the non-pregnant women had significantly higher levels of MAIT cells in PB compared to pregnant women (**Figure 1F**). Together, this suggests that a proportion of MAIT cells may leave the circulation and enter the placenta during late healthy pregnancy. The proportion of CD4+, CD8<sup>+</sup> or CD4/CD8 DN conventional T cells did not significantly differ between pregnant and non-pregnant women (**Supplementary Figure S4**). Data on the proportion of conventional EM T cells was not available in this cohort of non-pregnant women.

Throughout this study, MAIT cells were defined by the expression of CD161 and Vα7.2. To confirm that these cells were MR1-restricted, we stained PB and IVB cells from three donors with the MR1 5-OP-RU tetramers. It showed that similar proportions of MAIT cells were found when using the MR1 tetramer or the CD161 and Vα7.2 gating strategy both in PB and IVB (**Figure 1G**). Out of the CD161+Vα7.2<sup>+</sup> cells, a median of 93.6% and 93.9% stained positive for the MR1 tetramer in PB and IVB, respectively (**Supplementary Figure S2B**). The negative control tetramer MR1 6-FP showed little or no staining (**Figure 1G**).

The majority of MAIT cells in both IVB and PB in the pregnant women were CD8+, but intervillous MAIT cells were to a larger degree CD4 and CD8 DN and to a lower extent CD4<sup>+</sup> compared to peripheral MAIT cells (**Supplementary Figure S2C**). This pattern was similar when the cells were gated on MR1 tetramer<sup>+</sup> cells in three donors (**Supplementary Figure S2D**). As observed for the conventional T cells, intervillous MAIT cells had more uniform EM characteristics as compared to peripheral MAIT cells (**Supplementary Figure S2E**). In contrast to conventional CD8<sup>+</sup> T cells, IVB MAIT cells had a less activated phenotype compared to PB MAIT cells with lower expression of CD25 and PD-1, but similar expression of CD69 and HLA-DR (**Supplementary Figure S2F**). Together, intervillous and peripheral MAIT cells display different phenotypic patterns, and conventional T cells and MAIT cells show relative differences between IVB and PB in the expression patterns of both early and late activation markers.

### Placental Tissues Produce High Levels of Several Chemokines Which Attracts MAIT Cells and CD8<sup>+</sup> Effector Memory T Cells

We hypothesized that the observed enrichment of certain T cell populations in the intervillous space of the placenta was due to chemotactic factors produced by the placental tissue. Therefore, we examined the chemokine secretion pattern from placental explants. Several small pieces of placental villous tissue were collected and pooled from the same placenta to provide a representative sample. The tissue was cultured for 48 h, and the supernatant was harvested and analyzed by a 40-plex luminex assay. As shown in **Figure 2A**, placental explants secreted high levels of several chemokines and cytokines, including MIF, IL-6, CXCL8, and CCL2. To investigate the potential chemoattractive capacity of placental tissue on MAIT cells and other T cells, an assay of immune cell migration toward placental tissue conditioned medium was set up. We found that the placental

(Continued)

FIGURE 1 | within class variability (PB, n = 25, IVB, n = 24). (B) OPLS-DA loading plot based on the same data as in (A), following a variable influence of projection (VIP) of 0.81 (Supplementary Figure S1), showing associations between IVB or PB and phenotypic markers on T cells. (C) Proportions of CD3<sup>+</sup> cells within CD45<sup>+</sup> cells (left), and CD4+, CD8<sup>+</sup> and CD4−/CD8<sup>−</sup> cells within CD3<sup>+</sup> cells compared between PB and IVB (right) (n = 24). (D) Median proportions of naïve, central memory (CM), effector memory (EM) and terminally differentiated (TD) cells within CD8<sup>+</sup> T cells, based on the co-expression of CCR7 and CD45RA depicted in a pie chart (top) and as individual values (bottom) (n = 24). Representative dot plots are shown to the right. (E) Percentage of MAIT cells out of total CD3<sup>+</sup> cells (left, n = 24) and CD45<sup>+</sup> cells (right, n = 23) compared between PB and IVB. (F) Percentage of MAIT cells out of total CD3<sup>+</sup> cells compared between peripheral blood from pregnant (n = 35) and non-pregnant women (n = 27). (G) Comparison of MAIT cell characterization methods between CD161+Vα7.2<sup>+</sup> and 5-OP-RU tetramers in PB and IVB (left, n = 3), and representative plots of the stainings, including the negative control tetramer loaded with 6-FP (right). Line in graphs represents the median. Comparisons between paired samples were performed by Wilcoxon test (C–E and G) and unpaired samples by Mann-Whitney test (F). ns = not significant, \*p < 0.05, \*\*\*p < 0.001, \*\*\*\*p < 0.0001.

tissue conditioned medium promoted migration of CD3<sup>+</sup> T cells, CD8<sup>+</sup> T cells, MAIT cells, and CD8<sup>+</sup> EM T cells, as compared to control medium (**Figure 2B**). The number of migrating CD4<sup>+</sup> T cell, DN T cells, as well as CD8+, DN and CD4<sup>+</sup> MAIT cells all increased in the presence of conditioned medium (**Supplementary Figures S5A,B**). The proportion of migrating CD8+, CD4<sup>+</sup> and DN T cells out of CD3<sup>+</sup> T cells was not altered when placenta conditioned medium and control medium was added (**Figure 2C**; **Supplementary Figure S5A**). On the other hand, the proportion of both MAIT cells and conventional CD8<sup>+</sup> EM T cells were higher when placenta conditioned medium was added, compared to control medium (**Figure 2C**). The proportion of CD8<sup>+</sup> and DN migrating MAIT cells was not significantly affected by the conditioned medium, but the percentage of CD4<sup>+</sup> MAIT cells were lower in MAIT cells migrating toward the conditioned medium (**Supplementary Figure S5B**). No significant effect was observed for the proportion of CD4<sup>+</sup> EM T cells (data not shown). Thus, this indicates that chemokines produced from fetal placental tissues attracts T cells in general, but have a stronger attracting capacity on MAIT cells and conventional CD8<sup>+</sup> EM T cells compared to other T cell subsets.

# Chemokine Profile in Intervillous Blood Is Different From That of Peripheral Blood

Following our findings regarding the chemoattractive properties of term placental tissue, we investigated the chemokine pattern in the blood adjacent to the fetal villi in the intervillous space. Plasma from paired PB and IVB samples were analyzed to examine the chemokine profile in these compartments. As shown in the OPLS-DA loading plot in **Supplementary Figure S6**, a large proportion of the investigated chemokines were associated to either PB or IVB. The levels of chemokines and cytokines in PB and IVB in all individual samples are displayed in a heatmap, sorted based on the results from the OPLS-DA (**Figure 3A**). Univariate analysis showed significantly higher levels of a majority of the analyzed soluble factors in IVB compared to PB, whereas other were higher in PB, as indicated in **Figure 3A**. MIF displayed the strongest association with IVB and showed a median 182-fold higher level in IVB compared to PB (**Figure 3B**). Other factors that were higher in IVB included CXCL10, CCL25 and CXCL9 (**Figure 3B**). In contrast, levels of CCL27, CXCL12, and CCL21 were significantly lower in IVB compared to PB. Many of the chemokines that were produced in high levels from the placental explants (**Figure 2A**), were higher in IVB than PB plasma (**Figures 3A,B**). A direct comparison between the levels of chemokines in placental conditioned medium and IVB plasma are shown side by side in **Supplementary Figure S7A**. The relative differences between levels of chemokines in IVB and PB plasma, as determined by the ratio of levels in IVB divided by levels in PB, are shown in **Supplementary Figure S7B**. Together, this suggests that the placental villous tissue was responsible for the production of at least some of the chemokines detected at elevated levels in IVB plasma.

## MIF Levels in Intervillous Blood Correlate With the Proportion of MAIT Cells in the Placenta

The data on paired IVB plasma chemokine levels and IVB T cell composition allowed us to study associations between particular T cell subsets and levels of placenta-derived chemokines. In an attempt to identify factors related to MAIT cell homing to the intervillous space, the proportions of MAIT cells in IVB were analyzed for associations with levels of IVB chemokine levels using OPLS analysis. As shown in **Figure 4A**, higher proportions of MAIT cells were positively associated with higher levels of MIF and CCL25 (**Figure 4A**). In univariate analysis, the proportion of MAIT cells in IVB correlated directly to the levels of MIF in IVB, and there was a trend for correlation with CCL25 levels (**Figure 4B**). Paired data on proportions of MAIT cells from decidua parietalis and IVB chemokine levels was available from 18 of the donors, and showed an even stronger correlation with levels of both MIF and CCL25 in IVB (**Figure 4C**). These associations further support the notion that placenta-derived MIF and CCL25 are involved in attracting maternal MAIT cells to both the decidua and the intervillous space. In contrast, when analyzing the association between proportions of MAIT cells in PB with levels of MIF and CCL25 in the same samples, no correlation was found (**Figure 4D**). Thus, MIF and CCL25 did not appear to affect MAIT cell numbers in general, but rather seemed to be involved in directing MAIT cells from the periphery into the placental intervillous space and to the decidua. IVB plasma levels of CCL21 and CCL27 were inversely correlated to IVB MAIT cell proportions (**Supplementary Figure S8A**).

Conventional CD8<sup>+</sup> EM T cells in IVB showed no correlation with IVB MIF or CCL25 levels (**Supplementary Figure S8B**). However, the proportions of conventional CD8<sup>+</sup> EM T cells instead were significantly correlated to the levels of the CXCR3-ligands CXCL9, CXCL10, and CXCL11 (**Supplementary Figure S8C**). This suggests that MAIT cells and

conventional CD8<sup>+</sup> EM T cells are recruited to the intervillous space by different chemokine gradients.

# Placenta-Derived Chemokines and MIF Promote MAIT Cell Migration

To further investigate the relationship between IVB MAIT cells and chemoattractive factors, we next blocked CCL25 and MIF, identified as putatively important for MAIT cell homing to the placenta, in a migration assay. Neutralizing antibodies to CCL25 or the MIF-inhibiting chemical ISO-1 were used to block these factors in the placental tissue conditioned medium. Since the majority of the MAIT cells are highly positive for CCR6, the receptor for CCL20, we also included neutralizing antibodies to this chemokine. For total CD3<sup>+</sup> T cells, we found that the number of migrating T cells were reduced when ISO-1 or the combination of all three factors were added to the conditioned medium (**Figure 5A**). A similar pattern was observed for the number of migrating MAIT cells. Among the three conditions, only ISO-1 significantly decreased the migration of MAIT cells with a median of 24 % relative to the diluent control (**Figure 5A**). When all three factors were blocked, the number of migrating MAIT cells decreased with a median of 41% relative to isotype

indicate individual donors. (B) The levels of MIF, CXCL10, CXCL9, and CCL25 were significantly higher in IVB compared to PB, whereas the levels of CCL27, CXCL12, and CCL21 where significantly lower. Line in graphs represents the median. Comparisons between paired samples were made using the nonparametric Wilcoxon test. \*\*\*\*p < 0.0001.

and diluent control. When analyzing the proportion of MAIT cells within migrating CD3<sup>+</sup> T cells, we found that the MAIT cell proportions were significantly reduced when all three factors were blocked, whereas there was no significant effect on MAIT cell proportion when only anti-CCL25, anti-CCL20, or ISO-1 were added on their own (**Figure 5A**).

For CD8<sup>+</sup> EM T cells, we found that the number of cells that migrated was significantly blocked by ISO-1, but only with a

parietalis and the measured levels of MIF (left) and CCL25 (right) in paired IVB samples (n = 18). (D) Correlation between the frequency of peripheral blood (PB) MAIT cells and the measured levels of MIF (left) and CCL25 (right) in paired PB plasma samples (n = 24). Correlations between paired samples were made using the nonparametric Spearman correlation test.

median of 13% relative to the control (**Figure 5A**). Neutralization of CCL20 also led to a statistically significant, but small decrease in migration. Blocking of all three factors decreased the number of migrating CD8<sup>+</sup> EM T cell with a median of 30 % relative to control. However, when the proportion of CD8<sup>+</sup> EM T cells among total migrating CD8<sup>+</sup> T cells was analyzed, we found no significant reduction in migration by blockade of any of the factor, either alone or in combination. Instead, an increase in CD8<sup>+</sup> EM T cell proportion was observed when ISO-1 was added (**Figure 5A**). Thus, MIF appear to attract T cells in general, since the number of total migrating T cells were reduced when ISO-1 was added. However, the proportion of CD8<sup>+</sup> EM T cells among migrating CD3<sup>+</sup> T cells was increased when MIF was blocked, suggesting that MIF has little specific effect in promoting migration of CD8<sup>+</sup> EM T cells as compared to other T cell subsets. On the other hand, the proportion of MAIT cells among migrating T cells was reduced when blockade of MIF in combination with CCL25 and CCL20 was performed. Raw data on cell counts after addition of isotype controls, diluent controls, and neutralizing agents are shown in **Supplementary Figure S9**.

Since MIF appeared to be an important factor for MAIT cell attraction, we next examined if recombinant MIF could promote migration of T cells and MAIT cells. MIF significantly increased the number of migrating MAIT cells at a concentration of 250 ng/ml (**Figure 5B**). However, MIF did not promote migration of conventional CD8<sup>+</sup> EM T cells as determined by the number of migrating cells, although there was a trend at the higher concentration of 250 ng/ml (**Figure 5B**). When analyzing

FIGURE 5 | proportion of migrating MAIT cells within CD3<sup>+</sup> T cells (bottom left) and CD8<sup>+</sup> EM T cells among CD8<sup>+</sup> T cells (bottom right) toward CM. Results are expressed as blocking relative to isotype or diluent control. (B) The chemoattractive effect of MIF on the migration of number of MAIT cells (top left) and CD8<sup>+</sup> EM T cells (top right), using increasing concentrations of MIF. Proportion of migrated MAIT cells within CD3<sup>+</sup> T cells (bottom left) and CD8<sup>+</sup> EM T cells among CD8<sup>+</sup> T cells (bottom right) in the absence and presence of recombinant MIF. P values refer to comparisons between the medium control and the indicated levels of MIF. (B) The data on the bottom y-axis for MAIT and CD8<sup>+</sup> EM T cell counts were generated when adding purified T cells and the top y-axis when using CD4-depleted T cells. Line in graphs represents the median. Comparisons between paired samples were made using the nonparametric Wilcoxon test. ns, not significant.

the proportion of MAIT cells and CD8<sup>+</sup> EM T cells among migrating T cells in the parent gate, we found that only MAIT cells were significantly increased when MIF was added and no effect was found for CD8<sup>+</sup> EM T cells (**Figure 5B**). Sufficient counts were recorded in seven of the donors to allow sub-gating. For these, we found that the proportion of migrating CD8<sup>+</sup> MAIT cells were reduced when MIF was added, and there was a trend for an increased proportion of migrating DN MAIT cells (**Supplementary Figure S10**). Thus, these findings support the notion that MIF has a T cell attracting capacity and that this effect is stronger on MAIT cells than on conventional CD8<sup>+</sup> EM T cells.

## MAIT Cells Express Higher Levels of CXCR4 Compared to CD8<sup>+</sup> Effector Memory T Cells

MIF has previously been reported to bind to CXCR4 (26), and we next compared the level of CXCR4 expression on MAIT cells and CD8<sup>+</sup> EM T cells in medium control wells. Although a large proportion of both MAIT cells and CD8<sup>+</sup> EM T cells expressed CXCR4, the proportion of MAIT cell expressing CXCR4 was significantly higher (**Figure 6A**). The median intensity expression (MFI) of CXCR4 was also higher in MAIT cells in five out of six samples (**Figure 6A**). When MIF was added to the migration assay, we found that the proportion of CXCR4<sup>+</sup> CD8<sup>+</sup> EM T cells increased compared to medium control and also that the intensity of expression of CXCR4 was higher (**Figures 6B,C**). We found no enrichment of CXCR4 expressing MAIT cells when MIF was added (**Figures 6B,C**), but this could be explained by that the median expression of CXCR4 on MAIT cells was as high as 99 % in the control condition, a finding also supported by Dias et al. (27). In summary, the higher expression of CXCR4 on MAIT cells may increase the capacity to more efficiently migrate toward MIF compared to CD8<sup>+</sup> EM T cells, and our data support that CXCR4 is involved in the migration of T cells toward MIF.

# DISCUSSION

The decidua and the intervillous space are both sites for fetal-maternal immune interactions where maternal immune cells are in contact with fetal extravillous trophoblasts and syncytiotrophoblasts, respectively. Several studies have examined the composition of maternal immune cells in decidual tissues, in both healthy early and term pregnancies (28–30), as well as in miscarriage and pre-eclampsia (31, 32). In contrast, surprisingly little is known about immune cells in the IVB. In the present study, we show that the T cell compartment in IVB markedly differs from that in PB at term pregnancy, and that placentaderived factors may play a role in recruiting or retaining certain T cell subsets in the intervillous space and decidua.

The multivariate discriminant analysis showed that high proportions of conventional EM T cells and MAIT cells, and activated conventional T cells were strongly associated to IVB. Non-pregnant women had a higher proportion of MAIT cells in the circulation compared to our cohort of pregnant women, but it should be noted that the groups were not age-matched and the sample collection from the non-pregnant women were not adjusted to their menstrual cycles, which could affect the results. The non-pregnant women had a higher median age, and since it has been shown that MAIT cell proportions decrease with age (33), the age difference may mask an even bigger decline in circulating MAIT cell levels in pregnancy. Our findings are thus consistent with a possible homing of peripheral MAIT cells to the placenta during pregnancy. We have previously shown that IVB MAIT cells do not express the proliferation marker Ki67 (4), which further indicate that MAIT cells are recruited or retained rather than proliferating at the site. To further study whether the increase in IVB MAIT cell frequency is due to homing from PB, it would be interesting to follow PB MAIT cell frequencies longitudinally, including before, during, and after pregnancy.

The chemokine pattern in IVB was clearly different from that of PB, and we also found that placental villous explants produced high levels of several inflammatory chemokines. Svensson-Arvelund et. al. (21) and others [reviewed in (22)], have previously shown that first trimester placental tissues and trophoblasts produce several chemokines, but very little is known about how placenta-derived chemokines promote immune cell migration. Nancy et al. showed that murine decidual stromal cells undergo epigenetic modifications that decrease the production of the CXCR3-ligands CXCL9, CXCL10, and CXCL11, which prevented effector T cells from entering the decidua in mouse models (34). The number of T cells in human first trimester decidua is low, (28, 29) but the composition of immune cells changes during pregnancy and T cells become as numerous in decidua as in peripheral blood at term (29, 30). It is therefore reasonable to believe that the expression of chemokines changes as pregnancy proceeds. Our data from term placental villous explant medium and intervillous plasma indicate that the term placenta produces a wide array of chemokines and cytokines, which can be involved in recruiting inflammatory T cells to the decidua and intervillous space at term.

By using a multivariate factor analysis, we found MIF to be the factor most strongly associated with high MAIT cell proportions in IVB. Considering the large inter-individual proportions of MAIT cells and the elevated levels of a large number of

chemokines in IVB, we were surprised to find correlations between MAIT cell proportions and particular chemokine levels. However, the migration assays further supported that MIF was important for MAIT cell migration. Interestingly, decidual MAIT cell proportions from paired samples were strongly associated to MIF levels in the IVB, suggesting that MIF may also be involved in directing MAIT cell into adjacent tissues, and not only to retain them in a blood-filled space. However, it is likely that a combination of several chemokines ultimately regulates retention and migration of cells to a particular tissue, and that other chemokines may be of greater importance in settings other than pregnancy.

It was further evident that MAIT cells and conventional CD8<sup>+</sup> EM T cells were associated with different chemokine patterns. Whereas proportions of MAIT cells were associated to high MIF and CCL25 levels, conventional CD8<sup>+</sup> EM T cells correlated to the CXCR3-ligands CXCL9, CXLC10 and CXCL11. These findings support the results by Nancy et al. showing that the CXCR3-binding chemokines are important for attracting effector T cells to the placenta, but may also contradict the findings that these chemokines are silenced at the fetal maternal interface during pregnancy (34). Powell et al. observed an increased frequency of CXCR3<sup>+</sup> CD4<sup>+</sup> T cells in human 3rd trimester decidua compared to PB following uncomplicated pregnancies (35). Together, this points to differences in homing mechanisms and immune cell recruitment between early and term pregnancy, and also between the human and murine placenta. We have here mainly focused on trying to delineate how MAIT cells are recruited or retained in the intervillous space, but for future studies it would be interesting to investigate if migration of CD8<sup>+</sup> EM T cells toward placental conditioned medium could be blocked by neutralizing antibodies to CXCL9, CXCL10, and CXCL11.

MIF has been described to act as a promoter of inflammation by aiding monocyte and macrophage survival by suppressing p53 (36), increasing their inflammatory response (37, 38), and by counteracting the anti-inflammatory effect of glucocorticoid hormones (39). MIF has a predominant role in the innate immune response, where for example bacterial lipopolysaccharide has been shown to induce systemic release of MIF (40) Apart from monocytes and macrophages, MIF also has a potentiating effect on the pro-inflammatory activity of T cells (26). MIF attracts monocytes by interacting with either CXCR2 or the complex CXCR2/CD74, T cells by binding CXCR4 (26) and B cells by interactions with the complex CXCR4/CD74 (41). The functional effect of the interactions between MIF and monocytes and T cells has been shown in mice with advanced atherosclerosis, where MIF-blockade led to decreased plaque infiltration by monocytes and T cells, together with a regression in plaque size (26). The mechanisms for MIF secretion is poorly understood, but it has been shown that ATP-binding cassette transporter (ABCA1) can impair secretion of MIF (42).

MIF expression has been previously observed in first and third trimester decidual cells (25, 43). NK cells isolated from first trimester decidua synthesize and secrete MIF, and MIF negatively regulates their cytolytic capacity (44). In the placenta, MIF expression has been detected in syncytiotrophoblasts, cytotrophoblasts and extravillous trophoblasts from term pregnancies (23, 25). It was recently shown that MIF has antiapoptotic effects on trophoblasts during early pregnancy, (45) and that MIF is involved in trophoblast invasion (46). In line with our findings, the levels of MIF have been found to be higher in IVB compared to paired samples of PB in women following vaginal delivery at term pregnancy. In the same study, it was further noted that the levels of MIF in IVB was significantly lower in multigravidae women (24). Moreover, high levels of IVB MIF have been correlated to placental malaria in two cohorts (24, 47).

MAIT cells have been shown to express high levels of CXCR4 (27) and moderate levels of CCR9 (17), but low or no expression of CXCR2 (27), suggesting that CXCR4 and CCR9 are the receptors involved in the MAIT cell attracting capacity of MIF and CCL25, respectively. We have here confirmed that the majority of MAIT cells express CXCR4, and that a larger proportion of MAIT cells express CXCR4 compared to CD8<sup>+</sup> EM T cells. Furthermore, the proportion of CXCR4<sup>+</sup> migrating EM T cells increased when MIF was added to the migration assay, suggesting that CXCR4 is involved in the migration of T cells towards MIF. Since MAIT cells express high levels of CCR6, we evaluated blocking of the ligand CCL20, a chemokine that was also highly secreted from placental explants. However, blocking CCL20 did not significantly prevent MAIT cell migration, although there was an effect of blocking CCL20 in the majority of the experiments. This could be due to that other chemotactic factors produced by the placenta, including high levels of MIF, are more important for attracting MAIT cells in this context. On the other hand, migration of conventional EM CD8<sup>+</sup> T cells was partly decreased by the addition by anti-CCL20 antibodies. The relative importance of different chemotactic factors for the migration of MAIT cells and conventional EM CD8<sup>+</sup> T cells and the corresponding biological activity of chemokines is still largely unknown. It has been suggested that MAIT cells are recruited to the liver via CXCR6 and CCR6 and their ligands CXCL16 and CCL20 (48). It would be of great interest to study and compare MAIT cell and EM T cell migration in different physiological settings, including the intervillous space and the liver sinusoids, which are both characterized by low pressure blood flow.

Levels of CCL21 and CCL27 were significantly higher in PB compared to IVB, and inversely related to MAIT cell proportions in IVB. CCL21 is the ligand to CCR7, a chemokine receptor that is highly expressed by naïve T cells and is important for homing to lymph nodes. This is in accordance with the higher proportion of naïve T cells in PB compared to IVB, which are constantly circulating in PB to visit lymph nodes. CCL27 is predominantly expressed by keratinocytes and has been shown to attract CCR10<sup>+</sup> T cells to the skin (49), and MAIT cells only express low levels of CCR10 (50). For future studies, it would be interesting to further study the reverse relationship between low MAIT cell proportions and high levels of CCL21 and CCL27 in IVB.

One limitation with this study is that the number of migrating MAIT cells and CD8<sup>+</sup> EM T cells in the migration assay sometimes were low, which can question the relevance of the results. Indeed, in the assays in which the chemoattractive capacity of MIF was investigated, many of the experiments contained low numbers of migrating MAIT cells. The majority of the experiments were performed with purified T cells, but we also included six experiments where CD4-depleted T cells were added to the transwell. In this way, the cells that were put in the migration assay were enriched for MAIT cells and CD8<sup>+</sup> EM T cells, which consequently also led to higher migrating cell counts. Although MIF appeared to have an attracting effect on T cells in general, as determined by cell counts, the proportion of MAIT cells among migrating CD3<sup>+</sup> was increased after addition of MIF. In contrast, proportions of CD8<sup>+</sup> EM T cells were unaltered in the presence of MIF. Thus, although the robustness of the flow cytometry-based migration assay might be questioned, the results are still in line with the positive correlation between levels of MIF and proportions of MAIT cells in IVB.

The physiological role of MAIT cells in the placenta and their importance during pregnancy is not known. One possible function is protection of the fetus against bacterial and fungal infections, but also against viral infections since MAIT cells are also activated by inflammatory signals in an MR1-independent manner (11, 12, 14). Their rapid cytotoxic effects could provide an efficient innate-like defense mechanism to prevent pathogens from crossing the fetal-maternal barrier. The direct anti-bacterial and tissue protective activity of MAIT cells mediated by IL-17 and IL-22 production, (19), respectively, may also be important. Since many commensal and pathogenic microbes synthesize riboflavin, whereas human cells do not, MAIT cells provide an essentially different mechanism of discrimination between self and non-self compared to conventional T cells (5).

We have previously found that MAIT cells in IVB respond more vigorously with production of cytokines and cytotoxic molecules compared to peripheral blood MAIT cells after stimulation with Escherichia coli, (4) further indicating their importance in mediating anti-microbial inflammatory responses in placental tissues. On the other hand, it can be speculated that aggravated MAIT cell activation in intervillous space could be involved in initiation of preterm birth or play a role in other pregnancy complications. It was recently shown that the proportion of MAIT cells in peripheral blood of patients with pre-eclampsia was lower compared to noncomplicated pregnancies (51), indicating that MAIT cells may have accumulated in the placenta and that they possibly could be involved in mediating the pathophysiological traits of preeclampsia. It is therefore interesting that patients who later developed pre-eclampsia had low levels of MIF in PB at gestational week 13 (52), which was followed by a subsequent pathological rise in MIF in patients with pre-eclampsia, and the highest levels were seen in the patients who developed pre-eclampsia before gestational week 34 (53). This warrants further studies of the role of MIF, MAIT cell homing, and MAIT cell function at the fetal maternal interface in the context of pregnancy complications.

Our data suggests that certain immune cell subsets are enriched in the intervillous space in healthy term placentas. These findings challenge the general conception about placental blood circulation, since the entire blood content does not seem to be exchanged. Since the intervillous blood is surrounded by the fetal placenta, this compartment may be optimal for studying immune dysregulation and for identifying immunological factors involved in complicated pregnancies.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the regional review board of ethics in research at Karolinska Institutet with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was

# REFERENCES


approved by the regional review board of ethics in research at Karolinska Institutet.

## AUTHOR CONTRIBUTIONS

MS conceived, designed, performed, and analyzed the experiments, interpreted the results and wrote the paper. LG performed and analyzed the experiments and interpreted the results. ET and SG interpreted the results. EL and JD designed and performed the experiments. JS designed the experiments and interpreted the results. IM designed the experiments and interpreted the results. A-CL performed all multivariate factor analyses and interpreted the results. HK conceived, designed and analyzed the experiments, interpreted the results, wrote the paper and was the principal investigator of the project. All authors participated in final approval of the manuscript.

## FUNDING

HK was supported by grants from the Swedish Research Council, The Swedish Cancer Foundation, the Swedish Childhood Cancer Foundation, the Cancer Foundation in Stockholm, Stockholm County, and the Karolinska Institutet. MS was supported by Karolinska Institutet. A-CL was supported by grants from the Swedish state under the agreement between the Swedish government and the county councils, the ALF-agreement (ALFGBG-773381). JD was supported by Fundação para a Ciência e a Tecnologia Doctoral Fellowship SFRH/BD/85290/2012, cofunded by the Programa Operacional Potencial Humano-Quadro de Referência Estratégico Nacional and the European Social Fund.

#### ACKNOWLEDGMENTS

We thank Boel Niklasson and Catarina Arnelo and the other midwifes at Karolinska University Hospital in Huddinge for their excellent help in recruiting patients and help with collection of placentas.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01300/full#supplementary-material

The Placenta, From Development to Disease. Oxford: Wiley-Blackwell (2011). p. 17–26.


invariant T cells by MR1. Nature. (2003) 422:164–9. doi: 10.1038/nature 01433


lesional skin comprises mucosa-associated invariant T cells and conventional T cells. J Invest Dermatol. (2014) 134:2898–907. doi: 10.1038/jid.20 14.261


**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 Solders, Gorchs, Tiblad, Gidlöf, Leeansyah, Dias, Sandberg, Magalhaes, Lundell and Kaipe. 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.

# T Cell Receptor Expression Timing and Signal Strength in the Functional Differentiation of Invariant Natural Killer T Cells

#### Nyambayar Dashtsoodol 1,2 \*, Sabrina Bortoluzzi <sup>1</sup> and Marc Schmidt-Supprian<sup>1</sup> \*

<sup>1</sup> Department of Hematology and Medical Oncology, Klinikum rechts der Isar and TranslaTUM Cancer Center, Technische Universität München, München, Germany, <sup>2</sup> Department of Microbiology and Immunology, School of Biomedicine, Mongolian National University of Medical Sciences, Ulaanbaatar, Mongolia

#### Edited by:

Kazuya Iwabuchi, Kitasato University School of Medicine, Japan

#### Reviewed by:

Paolo Dellabona, San Raffaele Scientific Institute (IRCCS), Italy Hyun Park, National Cancer Institute (NCI), United States

#### \*Correspondence:

Nyambayar Dashtsoodol nyambayar.dashtsoodol@tum.de Marc Schmidt-Supprian marc.supprian@tum.de

#### Specialty section:

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

Received: 31 January 2019 Accepted: 01 April 2019 Published: 26 April 2019

#### Citation:

Dashtsoodol N, Bortoluzzi S and Schmidt-Supprian M (2019) T Cell Receptor Expression Timing and Signal Strength in the Functional Differentiation of Invariant Natural Killer T Cells. Front. Immunol. 10:841. doi: 10.3389/fimmu.2019.00841 The CD1d-restricted Vα14 invariant NKT (iNKT) cell lineage in mice (Vα24 in humans) represents an evolutionary conserved innate-like immune cell type that recognizes glycolipid antigens. Because of their unique ability to promptly secrete copious amounts of both pro-inflammatory and anti-inflammatory cytokines, typically produced by different T helper cell types, iNKT cells are implicated in the regulation of various pathologic conditions such as infection, allergy, autoimmune disease, maintenance of transplantation tolerance, and cancer. This striking multifaceted role in immune regulation is correlated with the presence of multiple functionally distinct iNKT cell subsets that can be distinguished based on the expression of characteristic surface markers and transcription factors. However, to date it, remains largely unresolved how this puzzling diversity of iNKT cell functional subsets emerges and what factors dictate the type of effector cell differentiation during the thymic differentiation considering the mono-specific nature of their T cell receptor (TCR) and their selecting molecule CD1d. Here, we summarize recent findings focusing on the role of TCR-mediated signaling and discuss possible mechanisms that may influence the sub-lineage choice of iNKT cells.

Keywords: NKT, CD1d, lymphocyte, development, functional subset, thymus, T cell receptor, developmental pathway

#### INTRODUCTION

Innate-like T lymphocytes are a special group of immune cells assumed to play important immunoregulatory functions by linking and orchestrating the functions of multiple cell types of the innate and adaptive arms of the immune system (1, 2).

The invariant NKT (iNKT) cells, often referred to as Vα14 iNKT cells in mice (Vα24 in humans), represent an evolutionary conserved innate-like immune cell type characterized by the expression of a unique semi-invariant T cell receptor (TCR). This TCR is composed of a single invariant α-chain Vα14Jα18 (encoded by Trav11Traj18) usually paired with a limited range of TCRβ-chains, mostly Vβ8.2, Vβ7, or Vβ2 (TRBV13-2, TRBV29, and TRBV1), in mice and a Vα24Jα18 TCRα-chain (encoded by Trav10Traj18) paired exclusively with Vβ11 (TRBV25) in humans. As opposed to the conventional peptide-recognizing CD8 T cells or CD4 T cells, iNKT cells are specialized in recognizing glycolipid antigens including alpha-galactosylceramide (α-GalCer) presented by the monomorphic MHC class I-like CD1d molecule (3–7). Owing to the

early availability of specific analysis tools such as CD1d tetramers (8, 9), as well as various gene-manipulated mouse models that either lack (10–17) or overexpress (18–22) these cells, to date iNKT cells represent the best-studied innate-like T lymphocyte lineage.

The hallmark feature of iNKT cells is their unique ability to secrete very large amounts of both pro-inflammatory and anti-inflammatory cytokines, typically produced by T helper cell type 1 (TH1), TH2, and TH17 cells. Cytokine secretion occurs rapidly upon activation and does not require clonal expansion or antigenic priming. Therefore, in line with their innate-like ability to robustly produce multiple immunoregulatory cytokines, iNKT cells were implicated in the regulation of various pathologic conditions such as infection, allergy, autoimmune disease, maintenance of transplantation tolerance, and cancer (23, 24). Because of this unique feature to elicit protective, regulatory, and pathogenic functions, it was proposed that iNKT cells constitute a heterogenous population. In fact, recent publications have demonstrated the presence of multiple functionally distinct subsets with discrete cytokine polarization within the iNKT cell lineage that can be distinguished based on the expression of characteristic surface markers and transcription factors (**Table 1**) (25, 26). The differentiation of iNKT cells proceeds within highly restricted context dictated by the recognition of self-glycolipids on CD1d by their semi-invariant TCR. To date, it remains largely unresolved how this puzzling diversity of iNKT cell functional subsets emerges and what factors dictate the type of effector cell differentiation during thymic differentiation.

As a number of excellent reviews have extensively covered advances in the iNKT cell developmental field (27, 30–34), in this mini-review, we attempted to summarize recent findings focusing on the role of TCR-mediated signaling and discuss possible mechanisms that may influence the sub-lineage choice of iNKT cells. A better understanding of the mechanisms underlying the differentiation of iNKT cell functional subsets will eventually help in designing new strategies to explore the therapeutic potential of these cells for the benefit of immunocompromised patients.

#### INKT FUNCTIONAL SUBSETS

In contrast to conventional T cells, iNKT cells can acquire functional maturity in the thymus before their egress to peripheral tissues. Historically, a linear differentiation model was proposed, which postulated that CD1d-selected CD24<sup>+</sup> stage 0 (st0) immature iNKT cells mature through sequential stages: via CD44low NK1.1<sup>−</sup> stage 1 (st1) iNKT cells characterized by IL-4-producing capabilities to CD44high NK1.1<sup>−</sup> (st2) iNKT cells with IL-4- and IL-17-biased features, and finally to terminally matured CD44high NK1.1<sup>+</sup> (st3) iNKT cells with IFN-γ-biased polarization (35). However, direct evidence of differentiation of st1 and st2 iNKT cells into terminally matured st3 iNKT cells was not demonstrated (36). The first evidence probing the above question on whether IL-4-producing thymic iNKT cells can give rise to IFN-γ-producing iNKT cells was addressed by Watarai et al., who demonstrated that IL-4-producing IL-17 receptor TABLE 1 | Differential expression of surface markers and transcription factors on iNKT cell functional subsets.


iNKT, invariant NKT.

B (IL-17RB) expressing thymic iNKT cells do not generate IFN-γ-producing T-bet<sup>+</sup> thymic iNKT cells upon intrathymic transfer (25).

More recently, based on intracellular staining patterns of lineage-specific transcription factors such as T-bet, GATA-3, PLZF, and RORγt, Lee et al. demonstrated that there are at least three distinct iNKT functional subsets in the thymus, designated as iNKT1, iNKT2, and iNKT17, akin to the classification of classical CD4 T helper cell types (TH1, TH2, and TH17 cells, respectively) (26). iNKT1 cells express NK cell-related markers, are T-bet<sup>+</sup> PLZFlow, and produce mainly IFN-γ. Moreover, iNKT2 cells are GATA-3high PLZFhigh and produce high levels of IL-4, whereas iNKT17 cells are defined as RORγt + PLZFintermediate and produce IL-17 upon stimulation. Thus, NKT functional subsets can be distinguished on the basis of their expression profile of characteristic surface markers and signature transcription factors (**Table 1**).

iNKT cells also can be subdivided into CD4<sup>+</sup> and CD4<sup>−</sup> subsets, where both murine and human CD4<sup>−</sup> subsets show TH1 biased cytokine polarization and enhanced cytotoxic activity compared with CD4<sup>+</sup> counterparts (37–39). In line with this, granzyme A (Gzma) and granzyme B (Gzmb), key genes involved in NK-triggered killing, are expressed specifically on CD4<sup>−</sup> iNKT cells (40). Of note, hepatic CD4<sup>−</sup> iNKT cells possess superior antitumor activity compared with thymic or splenic CD4<sup>+</sup> and CD4<sup>−</sup> iNKT cells, suggesting that organ-specific mechanisms might dictate the functional capabilities of resident NKT cells (39). Although it remains incompletely understood whether the development of these diverse functional NKT cell subsets is related to the existence of dedicated precursor cells or is due to specific differentiation programs, a revised model, termed "lineage diversification," is proposed, which suggests thymic iNKT functional subsets represent stable distinct lineages rather than functional maturation stages (26, 30).

In addition to these subsets, iNKT follicular helper cells (iNKTFH) and IL-10-producing iNKT10 cell subsets were described (41–44). Although it is unclear whether iNKTFH cells originate from the thymus, iNKT10 cells can arise in the thymus of F108Y mice, which harbor a mutation in the TCRβ-chain that results in altered iNKT TCR conformation (45). In addition to this, iNKT cells convert into IL-10-producing cells upon repeated injection with agonistic glycolipid α-GalCer (46), which suggests iNKT cells might have some degree of sub-lineage plasticity. In regard to this, there is a report on induction of TGF-β-dependent FOXP3<sup>+</sup> iNKT cells in the periphery upon administration of α-GalCer (47).

Moreover, the relative frequencies of functional subsets differ between inbred mouse strains, such as C57BL/6 mice possess mostly iNKT1 cells, while BALB/c mice have a significantly larger iNKT2 cell subset. Additionally, iNKT cells are tissue resident and show unique tissue distribution patterns (27, 48).

#### THYMIC SELECTION AND DIFFERENTIATION

The development of the iNKT cell lineage proceeds in the thymus, where the precursor cells that successfully assembled their antigen receptor through recombination mediated by the recombination-activating gene (RAG) are positively selected on CD1d-expressing cortical CD4+CD8<sup>+</sup> double-positive (DP) thymocytes (49). Although the majority of iNKT cells are generated from DP thymocytes, it was recently demonstrated that a fraction of iNKT cells develops directly from the CD4−CD8<sup>−</sup> double-negative (DN) stage of thymic ontogeny, bypassing the DP stage. Fate-mapping experiments as well as conditional ablation of RAG2 at the DP stage demonstrated that a fraction of CD4<sup>−</sup> iNKT cells were able to develop directly from DN-stage thymocytes without passing through the DP stage (50). This suggested a scenario in which DNstage thymocyte precursors expressing rearranged Vα14Jα18 generated by random rearrangement events are positively selected on CD1d and develop into mature CD4<sup>−</sup> iNKT cells (**Figure 1**).

Although the exact nature of endogenous ligands presented by CD1d responsible for positive selection of iNKT cells remains to be determined, it was recently reported that mammalian αlinked glycosylceramides represent a candidate for an iNKT cell self-ligand in mice (6). In striking contrast to conventional T cells that undergo clonal deletion upon receiving strong signaling via TCR-MHC interaction, iNKT cells are considered to be selected via strong TCR signals in the thymus. The evidence suggesting agonist selection of iNKT cells is based, among others, on the previously activated and/or memory phenotype of iNKT cells as well as the elevated expression levels of markers associated with TCR signaling such as Egr2 (51) and Nur77 (52) on early iNKT cells. In addition to this strong TCR signaling during selection, the development of iNKT cells requires homotypic interactions between signaling lymphocytic activation molecule family (SLAM) receptors Slamf1 (CD150) and Slamf6 (Ly108) expressed on selector DP thymocytes (53). Reporter mice for TCR signal strength revealed that iNKT cells receive TCR signals during a relatively short developmental time window exclusively at the st0 immature iNKT cell stage (52). These selection signals induce Ras- and Ca2+-dependent transcription factors Egr1 and Egr2, where Egr2 is shown to directly regulate the expression of promyelocytic leukemia zinc finger (PLZF) (51). PLZF represents a key regulator of the innate-like effector functions of iNKT cells (54, 55). However, there is currently no clear consensus on how PLZF is induced, as it was shown that TCR stimulation is not sufficient to induce PLZF expression on pre-selection DP thymocytes (56), suggesting other unknown signals are likely to be required for the development of PLZFexpressing innate T cells such as iNKT. One important factor might be the property of the positively selecting cell type, as CD4 T cells selected by MHC class II-expressing thymocytes express PLZF and show innate-like characteristics (57–59). Moreover, as PLZF expression can be detected in a subset of DN2 thymocytes (60), it is possible that those early-stage precursors expressing PLZF could differentiate later into a PLZF-expressing iNKT cells. In addition to this, premature expression of PLZF on DP thymocytes was demonstrated in pTα/Id2/Id3-deficient mice (61), which suggests a heterogeneity within pre-selection developmental intermediates (62, 63).

Much less is known regarding the negative selection of NKT cells. It was reported that intrathymic injection of α-GalCer and forced expression of CD1d on thymocytes or thymic antigenpresenting cells (APCs) such as dendritic cells (DCs) led to reduced numbers of iNKT cells (64–66). Whether negative selection shapes the iNKT cell population in normal development remains unclear.

#### DEVELOPMENTAL TIMING OF T CELL RECEPTOR EXPRESSION

The early thymic development of iNKT cells presumably mirrors that of conventional αβ T lymphocytes, where the early multipotential bone marrow-derived progenitors differentiate through tightly regulated DN stages 1 to 4 defined based on expression of CD117, CD25, and CD44 markers (67, 68). Major commitment to the αβ T cell lineage occurs at the DN3 stage, where rearrangement of the TCRβ-chain genes and subsequent beta-selection take place (69). Those DN3-stage thymocytes expressing functional pre-TCR, composed of pre-TCRα/TCRβ, differentiate further into the DN4 stage to become DP thymocytes. Although it is well-accepted that rearrangement

of the TCRα-chain occurs at the DP stage, there are some reports demonstrating the presence of TCRα transcripts within DN-stage thymocytes prior to differentiation into the DP stage (70–72). In this regard, it was reported that the DN4-stage thymocytes express Vα14Jα18 iTCR and RAG transcripts and possess iNKT cell potential in vivo (72). Furthermore, TCR sequencing experiments revealed the presence of out-of-frame Trav11Traj18 sequences, providing compelling evidence for ongoing stochastic TCRα-chain rearrangements within late DNstage thymocytes (50).

It seems that iNKT TCR expression during the late DN stage of thymic ontogeny plays a role in shaping the iNKT functional subset choice. Although both DN and DP pathways contribute to the generation of CD4<sup>−</sup> iNKT cells, the former pathway "preferentially" gives rise to IFN-γ-producing TH1 type iNKT cells with augmented cytotoxicity, compared to their counterparts of DP cell origin (50). Of note, such "preferential" development of TH1-type cells appears to be a general attribute of unconventional T cells that are generated as a result of early TCR expression at the DN stage of thymic ontogeny (73).

A potential mechanism for the "preferential" development of TH1-biased iNKT cells might be related to the differentiation stage of precursor cells undergoing positive selection. In this context, it was shown that DN-stage thymocytes normally express the IL-7 receptor (IL-7R), downregulate its expression after differentiating into the DP stage, and then reexpress it as post-selection αβ T cells (74). It was reported that IL-7R determines the fate of cytotoxic effector cells via induction of Runx3, which upregulates genes associated with cytotoxic lineage cells (75). In line with this, gene expression-profiling experiments revealed that the iNKT cells of DN cell origin had elevated expression of the IL-7R and its downstream associated genes characteristic of cytotoxic cells, such as Runx3, Gzmb, and Prf1, compared to iNKT cells of DP cell origin (50).

In addition to their functional bias, NKT cells of DN cell origin had a peripheral distribution pattern different from that of NKT cells of DP cell origin. NKT cells of DN cell origin were present mainly in the liver, a result that could be explained in part by their elevated expression of genes encoding liverhoming factors. In contrast, the NKT cells of DP cell origin were present mainly in the spleen but also in mesenteric lymph nodes, lamina propria, and adipose tissues, and they had high expression of genes encoding the homing factors responsible for peripheral localization, such as CCR6, CCR7, and CCR9 (50, 76–78).

Collectively, the above results suggest that the acquisition of diverse functional characteristics by iNKT cells might be dependent on the timing of TCR expression as well as on the differentiation stage of precursor cells undergoing positive selection.

#### T CELL RECEPTOR SIGNAL STRENGTH

Recently, two groups reported that the TCR signal strength directs the differentiation of iNKT functional subsets (79, 80). Both studies made use of the well-characterized SKG mouse strain, which contains a hypomorphic ZAP70 allele due to a spontaneous mutation in an SH2 domain. ZAP70 (ζchain-associated protein kinase of 70 kDa) is a Syk family tyrosine kinase that is activated upon engagement of TCR and phosphorylates the linker of activated T cells (LAT) and the SH2 domain-containing leukocyte protein of 76 kDa (SLP-76), and thus is thought to play an essential role for T cell development (81–84). Strikingly, analyses of iNKT cells from SKG mice demonstrated that decreased TCR signaling strength leads to a predominance of NKT1 cells, whereas iNKT2 and iNKT17 cell subsets are reduced (79, 80). Based on the above results, it was proposed that higher TCR signals are necessary for the development of iNKT2 and iNKT17 cells, while iNKT1 cell development is relatively undisturbed in the context of reduced TCR signaling capacity (79, 80).

In contrast to the above, analyses of iNKT cells from the YYAA mouse strain (85), another mouse model of hypomorphic ZAP70, revealed that while the frequency of the iNKT2 subset is reduced, the proportion of iNKT17 is actually higher than that of wild-type mice and the percentage of iNKT1 cells is unchanged (80). In addition, more recently, it was shown that the iNKT1 subset is reduced in CD2476F/6F mice, in which the TCR signaling capacity is reduced by ∼60% as a result of phenylalanine (F) substitution of tyrosine phosphorylation sites of the six endogenous immunoreceptor tyrosine-based activation motifs (ITAMs) of CD3ζ, an obligate signal transducer of the TCR/CD3 complex (86).

Moreover, deficiency in the TCR signaling-independent transcription factor SOX4 results in specific reduction of the iNKT1 subset (86). In Sox4-deficient thymocytes, the levels of miR181, which regulates the TCR signaling threshold of DP thymocytes (87), are diminished (86). While the iNKT cell development is impaired in miR181-deficient mice (88), residual iNKT cells in these mice show increased proportions of the iNKT2 and iNKT17 subsets at the expense of reduced frequency of the iNKT1 subset (89). The above-mentioned defects in the iNKT cell development and in the proportion of iNKT subsets seen in miR181-deficient mice are normalized upon introduction of a pre-rearranged Vα14 iTCR transgene (89). Furthermore, the differentiation and functional diversification of PLZF-expressing γδNKT cells occur completely unperturbed in the absence of miR181 (90). As agonist selected T cells depend on miR181 expression, this suggests that γδNKT cells are not agonist selected. Nevertheless, these cells acquire PLZF expression and the ability to produce IFN-γ (with and without miR181), and they expand in the liver in the absence of iNKT cells. These findings argue against a sole role of agonist TCR signals to govern later functional differentiation of innate-like T cells. It was also reported that autophagy influences the iNKT functional maturation, whereby the iNKT1 cell subset is mostly affected via regulation of the cell cycle and survival processes (34, 91, 92). Additionally, let-7 and miR-17 contribute to iNKT subset development via post-transcriptional regulation of PLZF or TGF-βR II expression in post-selection iNKT cells, respectively (93, 94).

Collectively, the current literature does not provide a clear consensus interpretation on how differential TCR signaling strength affects iNKT functional maturation into distinct subsets. Further complicating the matter, it is possible that the kinetics of acquiring TCR signals over time might be as important as the avidity or quantity of individual TCR signaling events per se (95). Sub-lineage choices might occur based on whether TCR signaling persists or ceases as the case of conventional CD4 T or CD8 T cell choice proposed by the kinetic signaling model (96). It is also possible that positive selection and sub-lineage choices are sequential but not simultaneous events. Finally, other undefined TCR-independent factors provided by the microenvironment might affect the differentiation of iNKT functional subsets, as it was reported that iNKT1, iNKT2, and iNKT17 subsets develop, albeit with subtle variations, in mouse models with the monoclonal iNKT TCR specificity (22, 97).

# CONCLUDING REMARKS

Despite tremendous progress in the field, a number of important questions regarding the development of iNKT cell subsets remain unanswered. First, it is not completely understood why strong agonist signaling, which normally results with the clonal deletion in conventional T cells, culminates in the positive selection of the iNKT cell lineage. Second, how stable are these functional subsets and can they interconvert? In this context, it remains unknown what iNKT cell subsets are the precursors of iNKTFH and iNKT10 cells. Third, what are the factors that dictate homing and maintenance of iNKT cell subsets to different tissue sites? As currently there is no consensus view on the precise mechanisms driving the development of the functionally distinct iNKT sub-lineages, it is tempting to hypothesize that multiple mutually nonexclusive mechanisms might exist. A better understanding of functional differentiation mechanisms of the iNKT cell lineage could contribute in developing optimized strategies intended to exploit the unique features of iNKT cells for the benefit of patients.

# AUTHOR CONTRIBUTIONS

ND wrote the first draft. ND, SB, and MS-S edited the manuscript.

# FUNDING

This work was supported by the Deutsche Forschungsgemeinschaft through an SFB 1054 A02 to MS and by the Science and Technology Center Research Grant from the Mongolian National University of Medical Sciences to ND.

# REFERENCES


conventional selection in human thymus. Sci Immunol. (2017) 2:eaah4232. doi: 10.1126/sciimmunol.aah4232


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

# The Role of Autophagy in iNKT Cell Development

#### Guan Yang<sup>1</sup> , John P. Driver <sup>2</sup> and Luc Van Kaer <sup>1</sup> \*

*<sup>1</sup> Department of Pathology, Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN, United States, <sup>2</sup> Department of Animal Sciences, University of Florida, Gainesville, FL, United States*

CD1d-restricted invariant natural killer T (iNKT) cells are innate-like T cells that express an invariant T cell receptor (TCR) α-chain and recognize self and foreign glycolipid antigens. They can rapidly respond to agonist activation and stimulate an extensive array of immune responses. Thymic development and function of iNKT cells are regulated by many different cellular processes, including autophagy, a self-degradation mechanism. In this mini review, we discuss the current understanding of how autophagy regulates iNKT cell development and effector lineage differentiation. Importantly, we propose that iNKT cell development is tightly controlled by metabolic reprogramming.

Keywords: invariant natural killer T cells, CD1d, autophagy, thymic development, metabolic switch

#### INTRODUCTION

#### Edited by:

*Remy Bosselut, National Cancer Institute (NCI), United States*

#### Reviewed by:

*Yuan Zhuang, Duke University, United States Mihalis Verykokakis, Alexander Fleming Biomedical Sciences Research Center, Greece*

\*Correspondence:

*Luc Van Kaer luc.van.kaer@vanderbilt.edu*

#### Specialty section:

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

Received: *17 September 2018* Accepted: *29 October 2018* Published: *14 November 2018*

#### Citation:

*Yang G, Driver JP and Van Kaer L (2018) The Role of Autophagy in iNKT Cell Development. Front. Immunol. 9:2653. doi: 10.3389/fimmu.2018.02653* Invariant natural killer T (iNKT) cells are an innate-like T cell subset found in most mammalian species. Unlike conventional T cells, iNKT cells recognize glycolipid antigens presented by the MHC class I-like, CD1d molecule. iNKT cells express a semi-invariant T cell receptor (TCR) composed of Trav11 (Vα14)-Traj18 (Jα18) and Trbv13-2, Trbv29, or Trbv1 (Vβ8.2, −7, or −2) in mice (1–3) or TRAV10 (Vα24)-TRAJ18 (Jα18) and TRBV25-1 (Vβ11) in humans (1, 4, 5). Analogous invariant TCR chain usage was found in other mammals such as rats (6) and pigs. iNKT cells can be universally activated with the prototypical iNKT cell antigen α-galactosylceramide (α-GalCer) that was originally isolated from the marine sponge Agelas mauritianus (7). Once activated, iNKT cells provide a universal source of T cell help primarily through the rapid production of multiple effector cytokines capable of transactivating an array of immune cells (8, 9). In humans and animal models, α-GalCer has been used to therapeutically target iNKT cells to induce multiple profound effects during different pathological conditions, including cancer, autoimmunity, and infectious disease (8, 10–14).

Like the development of conventional T lymphocytes, iNKT cell development depends on somatic DNA recombination and selection in the thymus. CD1d presentation of endogenous ligands is critical for iNKT cell development and animals lacking CD1d have no detectable iNKT cells (15–17). In sharp contrast with conventional T cells, which require MHC expression by thymic epithelial cells for their development, iNKT cells are positively selected by CD1d-expressing CD4+CD8<sup>+</sup> double positive (DP) thymocytes (16, 18) (**Figure 1**). Nevertheless, a recent study provided evidence that a fraction of iNKT cells develop from late CD4−CD8<sup>−</sup> double negative (DN) stage thymocytes, bypassing the DP stage (19). Negative selection of iNKT cells is not yet clearly defined. Evidence showing that overexpression of CD1d on thymic stromal cells, dendritic cells (DCs), or DP thymocytes in transgenic mice resulted in a variable reduction in the number of iNKT cells suggests that iNKT cells are susceptible to negative selection during their development (20, 21). After the initial selection, iNKT cells transit through four maturation stages, each characterized by sequential acquisition of surface markers: stage 0, CD24+CD44−NK1.1−; stage

**36**

1, CD24−CD44−NK1.1−; stage 2, CD24−CD44+NK1.1−; and stage 3, CD24−CD44+NK1.1<sup>+</sup> (22, 23). iNKT cells become functionally competent to respond to TCR engagement during their maturation in the thymus. Functionally, thymic iNKT cells can be subdivided into iNKT1, iNKT2, and iNKT17 subsets according to their expression of particular transcription factors, surface markers, and cytokines that are expressed by conventional CD4<sup>+</sup> T helper (Th) cell subsets (Th1, Th2, and Th17 cells, respectively). Although the relationships between the different stages of iNKT cells and their subsets remain to be fully explored, stage 1 iNKT cells comprise mainly progenitor cells and include cells with the capacity to produce interleukin (IL)-4 that may be related to iNKT2 cells, stage 2 cells likely include all three subsets, and stage 3 cells predominantly include iNKT1 cells (**Figure 1**). Recent studies have provided evidence that TCR signaling strength governs this iNKT cell subset development, with strong signaling favoring iNKT2 and iNKT17 cell development (24, 25). In addition to these subsets, iNKT follicular helper cells and iNKT10 cells have been identified that resemble T follicular helper cells and regulatory T cells, respectively. Recent studies have revealed a critical role of autophagy, a cellular self-degradation mechanism, in iNKT cell development and function. Here, we review these findings in the context of changes in the metabolic status of developing iNKT cells.

#### SIGNALING PATHWAYS THAT CONTROL iNKT CELL DEVELOPMENT

Many signaling proteins and transcription factors are important for iNKT cell development and/or function. Deficiency of the invariant Vα14 TCR or its ligand CD1d results in a failure in iNKT cell generation (7, 17, 26). Runt-related transcription factor 1 is critical for the ontogeny of functional iNKT cells (18). The E protein transcription factor, HEB, is essential for iNKT cells to develop at their earliest developmental stage. This HEB-mediated regulation, in part, is controlled by modulating the expression of retinoic acid receptor-related orphan nuclear receptor gamma t, a possible HEB target, and the anti-apoptotic molecule Bcl-xL (18, 27–29). Once committed to the iNKT cell lineage, multiple other molecules are required for iNKT cell maturation. TCR engagement activates phospholipase Cγ1, which further leads to production of diacylglycerol (DAG) and inositol-1,4,5-trisphosphate (IP3), both of which are critical for iNKT cell development. DAG induces activation of the Ras guanyl nucleotide-releasing protein 1-Ras-extracellular signalregulated kinase 1/2 pathways and is involved in early iNKT cell development, as well as, late iNKT cell maturation (30). DAG kinases, which are important for controlling intracellular DAG concentration and therefore negatively regulate DAG signaling, are also critical for iNKT cell development (31). IP3 activates the Ca2+-calcineurin-NFAT pathway and regulates the generation of stage 1 and 2 iNKT cells via the early growth response protein

**Abbreviations:** α-GalCer, α-galactosylceramide; Atg, autophagy-related gene; DAG, diacylglycerol; DC, dendritic cell; DN, double negative; DP, double positive; Egr2, early growth response protein 2; HDACs, histone deacetylases; IFN-γ, interferon-gamma; IL, interleukin; iNKT, invariant natural killer T; IP3, inositol-1,4,5-trisphosphate; LC3, microtubule-associated protein 1A/1Blight chain 3; LKB1, liver kinase B1; mTOR, mechanistic target of rapamycin; PI3P, phosphatidylinositol 3-phosphate; PLZF, promyelocytic leukemia zinc finger; RACK1, receptor for activated C kinase 1; SLAM, signaling lymphocyte-activation molecule; Tbkbp1, TBK-binding protein 1; TCR, T cell receptor; Th, T helper; Tsc1, tuberous sclerosis 1; Vps, vacuolar protein sorting.

2 (Egr2) (32). Egr2 directly binds to the Zbtb16 promoter and activates transcription factor promyelocytic leukemia zinc finger (PLZF) that supports iNKT cell transition from stage 1 to stage 2 (33, 34). PLZF is thought to control the innate phenotype and is also expressed by MAIT cells, γδ T cells, and innate lymphoid precursor cells in the fetal liver and adult bone marrow (35–37).

Signaling lymphocyte-activation molecule (SLAM) receptors are critical for early iNKT cell maturation (38). Homotypic interactions of SLAM molecules, Slamf1, and Slamf6 and the downstream recruitment of SLAM adaptor protein and the Src kinase Fyn control formation of the stage 0 iNKT cell lineage by activating the NF-κB signaling cascade (39). Other transcription factors including c-Myc, T-bet, Id2, and GATA-3 were shown to regulate different stages of iNKT cell maturation after they migrate to peripheral lymphoid tissues (29). Many additional transcription factors and signaling molecules such as thymocyte selection-associated HMG box protein, Notch signaling, lymphoid enhancer factor, mechanistic target of rapamycin (mTOR), etc., have been reported to impact iNKT cell development and effector function (29, 40, 41). In addition to these pathways, autophagy and autophagy-related pathways also have been reported to be involved in iNKT cell development.

# A BRIEF INTRODUCTION TO AUTOPHAGY

Autophagy is a highly conserved cellular degradation process. It has been defined as an "auto-digestive" process that promotes the degradation of cytoplasmic proteins and damaged organelles by lysosomes (42). The resulting degradation products are then used in cellular remodeling and in regenerating molecular building blocks during conditions of stress. In this review, we focus on macroautophagy, hereafter referred to as autophagy. The autophagy pathway is tightly regulated by various factors, including nutrient starvation, hypoxia, mitochondrial toxins, and oxidative stress. Over 30 autophagy-related gene (Atg) products that were initially identified in yeast but are largely conserved in higher eukaryotes orchestrate this degradative process. The autophagy process consists of four distinct phases: nucleation, elongation, fusion, and degradation. The morphological hallmark of autophagy is formation of a double-membrane vesicle, termed autophagosome, which is generated in a step-wise manner (42). Nutrient starvation initiates autophagy by inducing dissociation of mTOR from the mTOR substrate complex (ULK1/2, Atg13, FIP200, and Atg101). This dissociation triggers autophagosome nucleation and elongation and leads to recruitment of the class III phosphatidylinositol-3-OH kinase complex, encompassing vacuolar protein sorting (Vps) 34, Vps15, Beclin1, and Atg14, which phosphorylates phosphatidylinositol to generate phosphatidylinositol 3-phosphate (PI3P), a phospholipid critical for membrane trafficking processes. The generation of PI3P leads to recruitment of two ubiquitin-like proteins and initiates the formation of autophagosomes. Briefly, the Atg12 (ubiquitinlike protein)-Atg5-Atg16 complex in conjunction with Atg9 mediates formation of pre-autophagosome structures. During this process, microtubule-associated protein 1A/1B-light chain 3 (LC3, another ubiquitin-like protein) is conjugated to phosphatidylethanolamine with the assistance of Atg4, Atg7 (E1 like enzyme), and Atg3 (E2-like enzyme) to form LC3-II, and associates with newly formed autophagosome membranes until they fuse with lysosomes. The generation of LC3-II is frequently used for monitoring autophagy (43, 44). Upon fusion with lysosomal membranes forming an autolysosome, the autophagic body is degraded by lysosomal esterases, lipases, and proteases and recycled to build new cellular components and energy.

Impaired autophagy is linked to many different diseases, including cancer, inflammatory bowel disease, neurodegeneration, and various cardiovascular, pulmonary, and infectious disorders (45). Autophagy is also implicated in multiple cellular processes, including cell development, survival, and differentiation (46). In the immune system, autophagy plays a critical role in regulating the development of innate and adaptive immunity (47).

# AUTOPHAGY IN T LYMPHOCYTE DEVELOPMENT

Autophagy is crucial for normal T cell development, activation, and differentiation (48). Briefly, at the precursor stage of T cell development, autophagy regulates hematopoietic stem cell selfrenewability and quiescence, which is mediated, at least in part, by the effects of autophagy on reactive oxygen species levels (49). During T cell development in the thymus, autophagy influences thymocyte selection by regulating peptide presentation in stromal cells and professional antigen-presenting cells, which affects survival or proliferation of CD4−CD8<sup>−</sup> DN thymocytes and/or their transition to the DP stage (50). Autophagy is also essential for the maintenance of T cells, especially the longterm survival of naïve T cells in the periphery, via regulation of organelle homeostasis (50). In addition, autophagy is increased in activated T cells (51, 52). Impaired autophagy is associated with T cell malfunction and subset redistribution (50, 53).

# AUTOPHAGY INFLUENCES iNKT CELL DEVELOPMENT AND FUNCTION

# Atg5 and Atg7

Elongation of autophagic vacuoles requires two ubiquitin-like conjugation systems, with critical roles for Atg5 and Atg7. Deficiency in either of these factors blocks most autophagic processes (54, 55). The development and function of iNKT cells were investigated in mouse models with hematopoietic or T cell-specific deletion of Atg5 or Atg7 (56, 57). iNKT cells from these animals displayed an immature phenotype in the thymus (56). Both Atg5- (57) and Atg7-deficiency (56) resulted in defective stage 3 iNKT cell development. These defects correlated with reduced secretion of interferon (IFN) γ and IL-17, but no change in IL-4 when the thymocytes were cultured with α-GalCer in vitro (56) or decreased production of IFN-γ and IL-4 when α-GalCer was injected in vivo (57).

Mechanistically, the Atg7-mediated regulation of iNKT cell development was T cell-intrinsic rather than through the

presentation of exogenous or self-lipid antigens by thymocytes or bone marrow-derived DCs to iNKT cells (56, 57). Loss of Atg7 also led to reduced expression of Bcl-2, Egr2, and PLZF in iNKT cells. Consistent with these findings, Atg5−/<sup>−</sup> iNKT cells showed increased cell death coupled with cell cycle arrest and elevated mitochondrial stress (57). Thus, the defect in survival of iNKT cells in these animals might be due to a combination of increased apoptosis (56) and/or autophagy-dependent regulation of cell cycle progression (58).

During development, iNKT cells undergo metabolic switching and require catabolic processes and autophagy for their transition to a quiescent state after cell number expansion from stage 0 to stage 3 (56, 57). More glucose is required for the stage 0 and 1 iNKT cells than for the more mature stage 2 and 3 iNKT cells, which rely more on increased autophagy (56). Overall, these results indicate that Atg5- and Atg7-dependent autophagy is required for iNKT cell development, especially at later stages of the maturation process, with the strongest effects seen on iNKT1 cells, followed by iNKT17 and iNKT2 cells. If and how these effects are related to the differential TCR signaling requirements observed for these distinct subsets remains to be determined.

# Vps34

Vps34 and its binding partner Beclin1 are important for the initiation of autophagy in yeast (59). However, the function of Vps34 in mammalian cells has been controversial, with contributions to autophagy, phagocytosis, endocytosis, and intracellular vesicle trafficking (60–62). We demonstrated that T cells from mice with a T-cell specific deletion of Vps34 showed profound defects in autophagic flux (53). In agreement with the previous findings, Vps34 deletion in T cells significantly impacted T cell homeostasis and function (53). Additionally, Vps34−/<sup>−</sup> iNKT cells exhibited a developmental blockade at stage 0 (53). Similar to the results in Atg5−/<sup>−</sup> and Atg7−/<sup>−</sup> iNKT cells, the restricted iNKT cell development in conditional Vps34−/<sup>−</sup> mice was T cell-intrinsic and independent of CD1drestricted antigen presentation (53). The reduced iNKT cell frequency in these animals also correlated with a reciprocal increase in natural killer cells in the spleen, liver, and lymph nodes. Because deletion of Vps34 in T cells, heart, or liver results in the loss of versatile cellular functions besides canonical autophagy (53, 62), it is likely that the role of Vps34 in iNKT cell development is more complex than its influence on autophagy alone. Therefore, the role of Vps34-mediated functions in iNKT cell development may be distinct from those of Atg5 and Atg7. This difference is also illustrated by the finding that deletion of Atg5 or Atg7 genes caused a blockade in stages 2 and 3 iNKT cell development, whereas Vps34-deficiency caused a blockade at stage 0.

#### Other Autophagy Regulators

Although direct evidence is lacking, many autophagy regulators have been shown to play critical roles in mediating iNKT cell development (**Figure 2**). Ablation of the metabolic homeostasis regulator of early T cell progenitors liver kinase B1 [LKB1, also known as the upstream kinase of AMPK cascade (63)] erased the development of all iNKT cells in the thymus (64). Importantly, CD1d-expressing human B lymphoblastoid cells treated with AMPK agonists were able to induce iNKT cell activation to levels comparable to α-GalCer treatment (65). However, conditional AMPKα deletion in T cells did not affect thymic iNKT cell frequencies, although iNKT cell subsets and function remain to be investigated (64). AMPK-interacting protein Fnip1 is also critical for iNKT cell development to stage 3 by maintaining metabolic homeostasis in response to metabolic stress (66). Similarly, mTOR signaling negative regulator tumor suppressor tuberous sclerosis 1 (Tsc1) is crucial for iNKT cell development (67, 68). Mice lacking Tsc1 showed markedly reduced iNKT cell frequency and absolute cell numbers in spleen and liver (67), and displayed a developmental block of iNKT cell differentiation at stage 2, with decreased IFN-γ- and increased IL-17-secreting iNKT cells (68). This Tsc1-induced regulation resulted from increased mTORC1 activity, as rapamycin treatment partially rescued reverted IL-17-secreting iNKT cell predominance to IFN-γsecreting iNKT cell predominance in Tsc1-deficient mice (68). In other studies, mTOR was selectively required for thymic iNKT cell development and mTOR-deficiency led to accumulation of stage 0 iNKT cells in the thymus (69), whereas Raptor-deficiency led to severe iNKT cell maturation blockade between thymic stages 1 and 2 (69, 70). The differences in the developmental blockade of iNKT cells between mTOR- and Raptor-deficient animals suggest that mTORC1-independent mTOR pathways such as mTORC2 might be involved in regulating iNKT cell development. These results also suggest that adequate level of mTORC1 is required for iNKT cell development. Taken together, these findings indicate possible regulatory roles for the induction of autophagy downstream of the Tsc-mTOR axis in iNKT cell development.

In addition to metabolic regulators, the role of histone deacetylases (HDACs) in iNKT cell development has been investigated. HDACs are histone-modifying enzymes that mediate removal of acetyl groups from proteins and are strongly involved in the regulation of autophagy [reviewed in (71)]. Deletion of Hdac3 in DP thymocytes completely blocked iNKT cell development without influencing conventional T cell development (72). Loss of Hdac3 in iNKT cells led to a severe reduction of iNKT cells, particularly at stage 3 (73). This depletion of iNKT1 cells was associated with reduced autophagy, although independently of Atg7 and p62 expression, and decreased GLUT1, CD71, and CD98 nutrient receptor expression (73).

TBK-binding protein 1 (Tbkbp1) is a protein with undefined physiological function but physically interacts with the protein kinase TBK1 (74). A recent study identified Tbkbp1 as a crucial regulator of autophagy stimulated by the cytokines IL-15 and IL-2 (75). This Tbkbp1-mediated autophagy regulation was achieved through activation of ULK1 by antagonizing the inhibitory action of mTORC1. Tbkbp1 deficiency caused a T cell-intrinsic selective loss of stage 3 iNKT cells and a relative accumulation of stage 1 and 2 iNKT cells (75). The predominant loss of stage 3 iNKT cells in Tbkbp1-deficient mice was likely contributed by reduced autophagy, as autophagy inhibitor 3 methyladenine caused a high level of apoptosis in both wildtype and Tbkbp1-deficient mice, even in the presence of IL-15. Receptor for activated C kinase 1 (RACK1) is an adaptor involved in multiple intracellular signaling pathways and is important for the assembly of the autophagy-initiation complex (76). RACK1-deficiency blocked thymic iNKT cell development and migration of mature iNKT cells to peripheral lymphoid organs (77).

### REFERENCES


# CONCLUDING REMARKS AND FUTURE OUTLOOK

Autophagy, a process that is regulated by the metabolic status of cells, is critically important for iNKT cell development. Quiescent T cells can use autophagy to break down intracellular components to supply molecules for oxidative phosphorylation (78). iNKT cells acquire a memory phenotype while developing in the thymus. They undergo metabolic switching during development and differentiation to meet their changing energy demands, with stage 2 and 3 iNKT cells staying in a more quiescent state than the more proliferative stage 0 and 1 iNKT cells. We propose that autophagy is required for this metabolic switch (**Figure 1**). Immature proliferative iNKT cells exhibit high glucose uptake, and this high demand for glucose is reduced upon maturation, which is accompanied by diminished proliferation and increased autophagy. Not surprisingly, ablation of autophagy genes Atg5, Atg7, or Vps34 in iNKT cells led to a defective transition to a quiescent state after population expansion of thymic iNKT cells. However, how autophagy might be regulated by upstream metabolic regulators during iNKT cell development remains to be determined (**Figure 2**). More research is required to elucidate the role of other autophagy-related gene products in iNKT cell development and maturation. Whether autophagy can control the development and differentiation of other innatelike T cells marked by PLZF expression also remains an open question.

# AUTHOR CONTRIBUTIONS

GY wrote the first draft, and GY, JD, and LVK edited the manuscript.

# FUNDING

Work in the authors' labs was supported by grants from the NIH (DK104817 to LVK), NIFA (2016-09448 to JD), NICHD (HD092286 to JD), the Department of Defense (W81XWH-15-1- 0543 to LVK), the National Multiple Sclerosis Society (60006625 to LVK), and the Crohn's and Colitis Foundation of America (326979 to LVK).


homeostasis and function. Nat Immunol. (2011) 12:888–97. doi: 10.1038/ ni.2068


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

# The Role of CD1d and MR1 Restricted T Cells in the Liver

Wenyong Huang, Wenjing He, Xiaomin Shi, Xiaoshun He, Lang Dou and Yifang Gao\*

Organ Transplantation Unit, First Affiliated Hospital of Sun Yat-sen University, Guangzhou, China

The liver is one of the most important immunological organs that remains tolerogenic in homeostasis yet promotes rapid responses to pathogens in the presence of a systemic infection. The composition of leucocytes in the liver is highly distinct from that of the blood and other lymphoid organs, particularly with respect to enrichment of innate T cells, i.e., invariant NKT cells (iNKT cells) and Mucosal-Associated Invariant T cells (MAIT cells). In recent years, studies have revealed insights into their biology and potential roles in maintaining the immune-environment in the liver. As the primary liver-resident immune cells, they are emerging as significant players in the human immune system and are associated with an increasing number of clinical diseases. As such, innate T cells are promising targets for modifying host defense and inflammation of various liver diseases, including viral, autoimmune, and those of tumor origin. In this review, we emphasize and discuss some of the recent discoveries and advances in the biology of innate T cells, their recruitment and diversity in the liver, and their role in various liver diseases, postulating on their potential application in immunotherapy.

Keywords: iNKT cells, MAIT cells, liver diseases, innate T cells, CD1d restriction

# INTRODUCTION

The liver is a primary internal organ that plays a unique role in pathogen defense. Approximately 1/3 of total blood passes through the liver every minute (1, 2). Once the blood enters, it circulates at a reduced flow rate through the sinusoids, which comprise a complex vascular network of capillary-like vessels. The reduced flow rate maximizes the opportunity for pathogens to recognize the hepatic immune environment.

The ability to restrict and eliminate invading pathogens is one of the main features of the immune system. Much hepatology literature has focused on the adaptive, antigen-specific classical T-cell populations and their role in the protection and pathogenesis of liver disease. However, human liver is selectively enriched in innate T cells, including natural killer T (NKT) cells (3) and Mucosal-Associated Invariant T (MAIT) cells (4, 5). These innate T cells are unconventional T cells with diverse functions that play an essential role in liver immune surveillance. Similar to other innate cells, recognition of antigens from microbial, endogenous glycolipids or metabolites activates innate T cells, allowing them to produce cytokines and cytolytic proteins (4, 5). One of the most important features of innate T cells is bridging the innate and adaptive immune response, and various types of cytokines produced after innate T cell activation can modulate CD4<sup>+</sup> and CD8<sup>+</sup> T cell immune response (6). During liver injury, innate T cells infiltrating into the inflammatory site after neutrophils and monocytes are proposed to be sensors that control the local immune response (7). Although innate T cells have previously been defined primarily by phenotypic markers, recent emerging evidence has revealed considerable functional

#### Edited by:

Luc Van Kaer, Vanderbilt University, United States

#### Reviewed by:

Paul Klenerman, University of Oxford, United Kingdom Philipp Hackstein, University of Oxford, United Kingdom, in collaboration with reviewer PK Rafael Solana, Universidad de Córdoba, Spain Moriya Tsuji, Aaron Diamond AIDS Research Center, United States Vipin Kumar, University of California, San Diego, United States

> \*Correspondence: Yifang Gao gaoyf26@sysu.edu.cn

#### Specialty section:

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

Received: 25 April 2018 Accepted: 01 October 2018 Published: 30 October 2018

#### Citation:

Huang W, He W, Shi X, He X, Dou L and Gao Y (2018) The Role of CD1d and MR1 Restricted T Cells in the Liver. Front. Immunol. 9:2424. doi: 10.3389/fimmu.2018.02424

**43**

complexity in this population. This review summarizes the current literature regarding iNKT, type II NKT and MAIT cells, which have important roles in a variety of liver diseases, particularly focusing on their role in human liver diseases.

# THE BIOLOGY OF INNATE T CELLS

iNKT cells, also known as type I NKT cells or classic NKT cells, are a subset of natural killer T (NKT) cells, while another subset of NKT cells is type II NKT cells. iNKT cells are characterized by signatures of both T and NK cells, including broad range expression of molecular markers that are typically associated with NK cells, for example, NK1.1 in mouse and CD161/CD56 in human (8, 9). Notably, these markers alone may be insufficient to distinguish iNKT cells, because in mouse, NK1.1 is not expressed in certain strains, including AKR, BALB/c, CBA/J, C3H, DBA/1, DBA/2, NOD, SJL, and 129 (10). There are reasonable amounts of MAIT cells in human peripheral blood and liver that also express CD56 (11); therefore, costaining with CD1d tetramer should precisely identify iNKT populations. The so-called invariant is based on limited TCR arrangement, Vα14Jα18/Vβ2, Vβ7, and Vβ8 in mouse and Vα24Jα18/Vβ11 in human (12, 13). Mouse and human iNKT cells can recognize lipid and glycolipid antigens of self or microbial origin presented on MHC class-I-like CD1d molecules (14, 15). Activation by the agonist α-galactosylceramide (α-Galcer) allows mice and human iNKT cells to readily proliferate, undergoing significant remodeling of their surface expression patterns with regard to several markers, such as NK1.1 and the semi-invariant TCR, resulting in production of abundant Th1, Th2 and Th17 type cytokines, including IFN-gamma, IL-4, IL-13, and IL-17 (16–18). Apart from the TCR dependent pathway, human iNKT cells can recognize and eliminate target cells expressing NKG2D ligands in a TCR-independent manner (19). Cytokines released by stimulated iNKT cells are able to transactivate other innate and innate-like immune cell subsets, thereby amplifying their initial responses (20–24). In addition, iNKT cells provide both antigen-specific cognate and noncognate aid to B cells (25) and in turn, are activated by B cells (26, 27). Interestingly, unlike the non-cognate iNKT cell– B cell interactions **Figure 1**, antigen-specific cognate iNKT cells induce a more innate-biased B cell response characterized by a discontinuous germinal center B cell expansion and rapid initial proliferation of IL-10-producing B cells that fails to induce humoral memory (28).

There are three functional subsets of iNKT cells in mouse and human, which produce a distinct combination of cytokines and lineage-specific transcription factors, namely, NKT1, NKT2, and NKT17. Murine studies have demonstrated that T-box 21 (T-bet), GATA binding protein-3 (GATA3), and retinoic acid receptor-related orphan nuclear receptor gamma (RORγt) are expressed on iNKT1, iNKT2, and iNKT17 cells, respectively, and these transcription factors are correlated with the function of iNKT cells (29). Mirroring T helper cell subtypes, iNKT1, iNKT2, and iNKT17 cells produce IFN-γ, IL-4, and IL-17 (29). These cytokines allow iNKT cells to interact with other immune cells. For example, IL-4 produced by NKT2 cells at steady state through phosphorylation of signal transducer and activator of transcription 6 (STAT6) regulates CD8 T cells developing to a memory-like phenotype in the thymus (30). Additionally, IL-4 promotes antibody production by B cells and induces dendritic cells to secrete T helper (Th) 2-type chemokines, such as chemokine (C-C motif) ligand (CCL) 17 and CCL22 (30). Moreover, in lymphoid organs, each iNKT subset displays different anatomic localization, which determines their responsiveness to intravenous or oral antigenic challenges (31). Recently, RNAseq analysis has suggested that each iNKT subset has a unique genetic signature, and these footprints are more similar to γδ T cells and innate lymphoid cells (ILCs) than to conventional T cells (32).

Compared to iNKT cells, type II NKT cells express relatively diverse TCRs that can recognize antigens derived from microbial, endogenous glycolipids, phospholipids and endogenous hydrophobic peptides presented by CD1d molecules (33). Type II NKT cells are unable to recognize α-linked glycolipids, for example, α-Galcer, but are responded to βlinked glycolipids (34). Igor Maricic et al. found that mice type II NKT cells are activated by self-phospholipids, for example, lysophosphatidylcholine (LPC), lysosphingomyelin (LSM) and lyso-platelet-activating factor (LPAF) (34). According to their function, type II NKT cells can be divided into pro-inflammatory and anti-inflammatory subsets (34). Stimulating type II NKT cells with self-glycolipid sulfatide inhibits inflammatory responses induced by CD4<sup>+</sup> T (35) and iNKT cells (36) in mice. In a mouse bone marrow transplantation model, donor type II NKT cells inhibit graft-vs.-host disease via releasing IL-4 (37). In a rat model, activating type II NKT cells with SCP2 peptide promotes the inflammatory response through production of inflammatory cytokines IL-5 and IL-6 (38). In an iNKT cell-deficient (Jα18−/−) mouse model, Sagami et al. found adoptive transfer of type II NKT cells exacerbated DSS-induced colitis (39).

MAIT cells are a subset of αβ T cells that possess both innate and effector-like qualities (40, 41). They are preferentially located in the gut lamina propria and express an invariant α chain (42). Similar to other αβ T cells, MAIT cells undergoing conventional TCR arrangement express canonical Vα7.2-Jα33 TCRs paired with variable β-chains in human (43, 44) and Vα19- Jα33 TCR in mice (42). In general, MAIT cells are equipped with effector properties before exiting from the thymus (45, 46). These cells were initially discovered as invariant α chain-expressing cells in the double-negative T cell fraction of human peripheral blood by Steven Porcelli et al. in 1993 (47). Later studies found that MAIT cells express high levels of CD161 and are also present in CD4-positive, as well as CD8-positive, lymphocytes (48).

Distribution of MAIT cells differs between humans and mice. The frequency of MAIT cells in C57BL/6 mice accounts for ∼0.1% of the peripheral T cell population (49), whereas 1–10% of them are identified in human peripheral blood. Moreover, the occurrence of MAIT cells varies widely among tissues in healthy adults, ranging from 2% (ileum) to 60% (jejunum), and they can make up ∼20–50% of intrahepatic T cells.

Interestingly, MAIT cells are absent in germ-free mice, indicating that their expansion in the periphery depends on the presence of microbial ligands (40, 42). MAIT cells are MR1, an MHC class I-like protein, restricted-T cells. Recognizing bacterial-produced vitamin B metabolites presented by MR1 allows MAIT cells to secrete a vast amount of pro-inflammatory cytokines, including IFN-γ, TNF-α, IL-2, and IL-17 (11, 41, 50), which lyse bacterially infected cells (51, 52). Additionally, comparative genomic analysis has demonstrated that MR1 is only expressed in marsupial and placental mammals and is exceptionally highly conserved, particularly at the α1 and α2 domains of ligand-binding grooves (53). Both MR1 and MAIT TCR genes are extremely highly conserved, implying that evolutionary pressure is involved in maintaining conservation of these genes (42).

MAIT cells can defend against microbial activity and infections caused by bacteria or yeast through activating the vitamin B2/riboflavin pathway in an innate manner (42, 45, 54). A study by Michael S. Bennett et al. provided evidence that supernatants from stimulated human MAIT cells promote B cell plasmablast differentiation and IgA, IgG, and IgM production (55). Additionally, human MAIT cells respond to mycobacterium tuberculosis infection and provide an early source of IFN-γ required for activation of the Th1 response (56). Together, these results indicate the potentially important role of MAIT cells in the defense against microbial invasion.

#### CD1D RESTRICTED T CELLS IN LIVER HEALTH AND DISEASE

iNKT cells contribute to a significant subset of lymphocytes in the liver. In mice, iNKT cells are most abundant in the liver (10–30%), with lower frequencies found in the thymus, blood, bone marrow and lymph nodes (0.1–0.2%). In humans, high iNKT cell numbers are detected in the liver (1%) (57, 58), compared to 0.01–0.5% in their peripheral counterparts (59, 60). The distribution of type II NKT cells is difficult to investigate due to their lack of specific surface markers (61). The literature suggests that type II NKT cells may outnumber iNKT cells in humans (61).

iNKT cells mediate various functions in the liver, including hepatic injury, fibrogenesis, and carcinogenesis. As one of the important immune subsets in the liver, iNKT cells demonstrate a pathogenic role in IRI, primary biliary cirrhosis, nonalcoholic fatty liver, and hepatitis. Interestingly, a protective role was identified for these cells in an acute liver injury model (**Figure 2**). To date, literature appears to suggest that iNKT cells exert a protective role during the acute phase of liver injury and a pathogenic role in chronic conditions (62). Some attention has also been focused on the implication of NKT cells in liver transplantation, including their role in ischaemia/reperfusion injury and transplantation rejection (63, 64). The study of type II NKT cells in liver disease progression is limited, focusing on hepatitis viral infection, where type II NKT cells appear to play a controversial role in controlling liver injury. A new subset of type II NKT cells, II NKT-Tfh cells, have been found to regulate metabolic lipid disorders (65).

#### Hepatocellular Carcinoma

HCC is often linked to chronic inflammatory liver diseases, such as NASH and viral hepatitis (66). A murine model suggested that TLR4 and canonical nuclear factor-κB signaling in the liver facilitate NASH-to-HCC conversion. Fundamentally, iNKT cells have a dual role in cancer that either promotes an anti-tumor response or elevates tumor growth via activation of effector T cells promoting Th1 responses or recruitment of regulatory T cells to induce Th2 responses (67, 68). Although hepatic iNKT

cells are rich in number, relatively few studies have attempted to clarify their role in HCC, and results from these studies appear contradictory.

HCC patients exhibit increased iNKT cell numbers in the tumor site compared to the peripheral blood. More importantly, hepatic iNKT cells are found to secrete Th2 cytokines, thus inhibiting tumor-specific CD8<sup>+</sup> T-cell responses (66, 69). In contrast, murine studies identified that CD4<sup>+</sup> iNKT cells could mediate anti-tumor responses through inhibition of the inflammatory response triggered by activation of the oncogenic β-catenin pathway (70). Additionally, iNKT cells are able to suppress tumor growth after adoptive transfer of HCC tumor lines in mice (71, 72).

Very recently, a study by Ma et al. found that commensal bacteria are important regulators of anti-tumor immunity that alter hepatic natural killer T cells. This regulation strengthens IFN-gamma production by hepatic natural killer T cells and promotes anti-tumor effects (73).

Current studies have suggested that type II NKT cells may play an immune regulatory role in cancer settings. In CD1d knockout and Jα18 knockout mice, Terabe et al. found that activation of CD1d-restricted type II NKT cells is sufficient for downregulation of tumor immunosurveillance in mouse fibrosarcoma, mammary carcinoma, colon carcinoma, and lung metastases of the CT26 colon carcinoma models (74). Using the same knockout method, Renukaradhya et al. demonstrated that type II NKT cells supress anti-tumor immunity against B-cell lymphoma (75). In addition, there are higher frequencies of IL-13 releasing type II NKT cells in myeloma patients than in healthy donors (76).

#### Hepatitis Viral Infection (HCV and HBV)

Overall, 57% of liver cirrhosis cases and 78% of liver cancers are caused by chronic HBV and HCV infections, accounting for almost a million deaths every year. A few studies have attempted to identify the role of iNKT cells in controlling HCV infections, in particular, during the initial phase of infection. In human hepatic CD3+CD56<sup>+</sup> cells, including iNKT cells, HCV replication is inhibited in hepatocytes by IFN-γ secretion, and this activity is positively correlated with disease progression (77). Furthermore, it modulates the effectiveness of IFN-alpha in late HCV infection. However, studies have shown that iNKT cells are considerably depleted in chronic HCV infection (78– 80). This finding may suggest that iNKT cells contribute to the early phase of HCV infection but not as much to disease progression.

Similarly, high numbers of activated type I NKT cells have been identified in the early stages of HBV infection in humans (78–80). In agreement with those results, CD1d expression is elevated in HBV+ve liver tissue compared to HBV–ve counterparts (81). Similar to their action in HCV infections, the inhibitory effect of iNKT cells on HBV occurs through secretion of IFN-γ as well, activating the adaptive immune response and inhibiting viral replication (82). Recently, Xu and colleagues have shown that exhaustion marker Tim-3 is upregulated on hepatic iNKT cells from HBV-transgenic mice (83). Blockade of Tim-3 by anti-Tim-3 antibody strongly enhances expression of IL-4, IFN-γ, TNFα, and CD107a in iNKT cells and augments α-Galcer-induced inhibition of HBV replication (83). Interestingly, researchers have found that on the one hand, iNKT cells control the replication of hepatic viruses, while on the other hand, they contribute to virally induced liver injury through production of pro-inflammatory cytokines that induce hepatocyte apoptosis and inhibit proliferation (84–86).

The function of type II NKT cells in hepatitis viral infection is debatable. In a ConA-induced mouse hepatitis model, activation of type II NKT cells with sulfatide or LPC evoke anergy in iNKT cells that suppresses inflammation-triggered liver damage (34). In contrast, hepatic type II NKT cells promote development of liver injury in a transgenic mouse model of acute hepatitis B virus infection (87). Stimulation of type II NKT cells triggers conventional T-cell activation and pro-inflammatory cytokine production, resulting in augmentation of hepatic injury in murine autoimmune hepatitis models (88).

#### Non-alcoholic Fatty Liver Disease

In the modern era, NAFLD is considered the most frequent chronic liver disease in developed countries, affecting ∼10– 20% of the population. NAFLD is characterized by abnormal accumulation of fat in the liver, leading to infiltration of inflammatory cells accompanied by fibrosis or necrosis progressing to liver cirrhosis or hepatocellular carcinoma (HCC) (89, 90). Current studies on systemic analysis of iNKT cell subsets in non-alcoholic fatty liver disease are very limited, with a few studies confirming their importance. In human hepatic CD1d cells, the number of CD3+CD56<sup>+</sup> cells are elevated in NASH patients (91). Reduced iNKT cell counts were found in mice fed high-fat diets and in obese mice (92, 93). In addition, mice with NAFLD lacking iNKT cells showed increased pro-inflammatory mediator factor and increased levels of TLR4 and PDGF2 mRNA (94). Activation of Kupffer cells (KCs) could cause apoptosis in these cells and further contribute to steatosis and insulin resistance (92, 95). Depletion of KCs could reduce hepatic IL-12 expression and rescue iNKT cells from apoptosis, preventing further pathological changes in the disease. Tim-3/galectin-9 is known to regulate the homeostasis of liver iNKT cells in the murine system (96).

Indeed, depletion of KCs via treatment with gadolinium chloride reduces hepatic IL-12 expression and does not lead to iNKT apoptosis, thereby preventing diet-induced hepatic steatosis and insulin resistance. Consistently, activation of the Hedgehog pathway and HSCs have been revealed to be associated with iNKT cells in mice fed an MCD diet or a combination of a CD-HFD (97–99). Using a diet-induced mouse obesity model, Satoh and colleague show that type II NKT cells trigger inflammation in the liver and exacerbate obesity (100).

#### Alcoholic Liver Disease

ALD is caused by chronic alcohol abuse resulting in alcoholic fibrosis or cirrhosis. The disease currently one of the most frequent causes of death. Activation of KCs via LPS/TLR signaling-dependent mechanisms following alcohol consumption result in increased secretion of a variety of pro-inflammatory cytokines and chemokines, in addition to eicosanoids and reactive oxygen species (101, 102). Mechanisms underlying ALD include a complex network of hepatocytes, KCs, DCs and innate T cells (103). Studies found activation of KCs via the LPS/TLR pathway following alcohol intake, which increases secretion of a variety of pro-inflammatory substances (101, 102, 104). In a murine model, increasing TNFα and IL-1β production were observed in alcohol-fed mice that neutralize IL-1β in KCs to allowed iNKT cell accumulation and steatosis. The study also demonstrated that NLRP3 inflammasome and IL-1β secretion are essential factors for hepatic iNKT cells to accumulate and activate in ALD (105). Consistently, gut microbes can also trigger KC NLRP3 activation, resulting in iNKT cell activation (105). A study by Mariric et al. examined the role of both type I and type II NKT cells in alcoholic liver disease, demonstrating that only iNKT cells became activated following heavy alcohol consumption, resulting in inflammation and liver tissue damage. This study suggests that type I and II NKT cells are functionally distinct in liver inflammation and tissue injury (106). A more recent study identified the interplay between IL-10-producing iNKT cells silencing the productive roles of NK cells in alcoholic liver disease (107).

Despite these findings, the role of iNKT cells in human ALD has not been well-examined. In agreement with the murine data, pro-inflammatory cytokine levels were increased in alcoholic hepatitis human subjects, suggesting a correlation with disease severity (57). In addition, in patients with alcoholic hepatitis, NKG2D expression in NK and iNKT cells has been found to correlate with disease severity, suggesting these cells are involved in promoting liver damage (108).

## MAIT CELLS IN LIVER HEALTH AND DISEASE

MAIT cells are significantly enriched in the liver, where they comprise up to 50% of liver-resident lymphocytes. These cells are located primarily in the biliary tract, and in the context of liver infection, MAIT cells can be activated by MR1-presenting bacterial ligands or indirectly via IL-12 and IL-18 produced by antigen-presenting cells in response to Toll-like receptor 8 signaling triggered by viral RNA (109, 110). The importance of MAIT cells in liver immunosurveillance is highlighted by three findings. First, liver MAIT cells are highly activated and express the activation marker CD69, as well as HLA-DR and CD38 (11, 110). This activation status suggests that liver MAIT cells are in a highly activated state, poised to respond to incoming antigens from the gut. Second, intra-hepatic MAIT cells, along with CD56bright NK cells, are the main source of IFN-γ post-TLR8 stimulation by liver-derived mononuclear cells through IL-12 and IL-18 activation (80). Finally, MAIT cells are the predominant IL-17 producers among intrahepatic T cells (∼65% of IL-17<sup>+</sup> T cells) in response to phorbol 12-myristate 13 acetate/ionomycin stimulation (11). As IL-17 targets multiple cell types in the liver, including Kupffer cells and BECs, to produce pro-inflammatory cytokines and chemokines (111), MAIT cells may be important regulators of hepatic inflammation and fibrosis **Figure 3**.

#### Hepatocellular Carcinoma

Recent studies have found that MAIT cells are recruited from peripheral blood to solid tumors in several cancers (112– 114). Infiltration and accumulation of MAIT cells into tumor sites suggest that MAIT cells play an essential role in tumor development. MAIT cells are highly enriched in human liver (109) but are resisted to skew to an IL-17-producing phenotype, as they fail to release to IL-17 upon TCR stimulation (11). On the other hand, the function of tumor-infiltrating MAIT cells is proposed to be impaired in response to a panel of TCR ligands and cytokines (115). In a colorectal liver metastasis setting, IFN-γ produced by hepatic tumor-infiltrating MAIT cells is significantly

suppressed (115). Taken together, the role of MAIT cells in hepatocellular carcinoma is still obscure, and further studies investigating the phenotype of tumor-infiltrating MAIT cells and their interactions with liver-resident cells will help to understand the role of MAIT cells in HCC.

# Hepatitis B Virus (HBV)

To date, the role of hepatic MAIT cells in HBV is still poorly understood. Two studies have compared peripheral MAIT cells in healthy controls and chronic HBV patients, showing opposing results. The first study found that MAIT cells were not deleted nor functionally impaired in HBV patients (116). In contrast, there was a higher frequency of MAIT cells expressing CD38 and releasing granzyme B in HBV patients, suggesting that MAIT cells were more activated in the HBV setting (116). The second study demonstrates that in HBV patients, MAIT cells are in an exhausted phenotype, where the frequency of cells in circulation is reduced, the expression of the early activation marker CD69 is inhibited, and the production of IFN-γ and granzyme B are significantly suppressed (43). Why there are discrepancies between these two studies is not clear. The opposite observations on granzyme B production may be explained by different activation methods, as Boeijen et al. activated MAIT cells with IL-12/IL-18/CD28<sup>+</sup> Escherichia coli (116), while Yong et al. stimulated MAIT cells with PMA/ionomycin (43). It should be taken into consideration that the size of both studies is relatively small. Therefore, the patients could have been in various clinical phases and undergoing different treatments. Indeed, MAIT cells are abundant in the peripheral blood but account for only a small percent of T cells (1–10%) (117). MAIT cells are further enriched in the liver (20% to 50% of T cells), which is also the primary site of infection (117). Therefore, further research with larger cohorts that focus on intrahepatic MAIT cells is required to solve the mystery of MAIT cells in HBV.

# Hepatitis C Virus (HCV)

Several studies have shown that CD8+, rather than CD4+, MAIT cells in the peripheral blood were significantly reduced in the setting of chronic HCV (118, 119). These results may be due to CD8<sup>+</sup> MAIT cells belonging to a newly defined pro-apoptotic phenotype expressing high levels of caspase 3 and 7 (120). Further phenotypic and functional studies reveal that the remaining CD8<sup>+</sup> MAIT cells represent a chronic activation phenotype with signs of immune exhaustion, which is characterized by elevated levels of CD38, HLA-DR, CD69, PD-1, TIM-3, CTLA-4, and Granzyme B (118, 119). Notably, the function of these MAIT cells is also impaired, as reflected by the production of IFNγ and TNFα being actively suppressed upon stimulation with TCR-dependent E. coli but not TCR-independent IL-12+IL-18 (118, 121). This result suggests that the loss and functional impairment of MAIT cells is a non-reversible process in chronic HCV patients, as antiviral treatment cannot reinvigorate these MAIT cells (118, 121, 122). Arguably, Ben Youssef et al found that adult MAIT cells in peripheral blood expand from cord blood Vα7.2<sup>+</sup> CD161high T cells, and this process lasts ∼5 years before filling up the adult MAIT pool (123). Therefore, the dysfunction and loss of MAIT cells after antiviral therapy may be due to the slow kinetics of differentiation and proliferation in MAIT cells.

There is an inverse correlation between the frequency of hepatic MAIT cells with liver inflammation and liver fibrosis in the setting of chronic HCV, demonstrating that MAIT cells are crucial mediators against HCV infection in the liver (121). Similarly, the percentage of hepatic MAIT cells is also reduced in chronic HCV patients (121). Importantly, the expression of HLA-DR and CD69 on MAIT cells is higher in the liver, suggesting that intrahepatic MAIT cells are more activated than are peripheral MAIT cells (121). This difference may because there is a higher frequency of activated monocytes in the liver, as they are an important source of IL-18 (121). MAIT cells are deleted in both blood and liver in the setting of HCV, and it is hypothesized that blood MAIT cells migrate to the organ, where they are further stimulated by inflammatory cytokines, resulting in activation-induced death, a mechanism that has been observed and well-characterized in HIV-induced MAIT cell depletion (121, 124).

#### Non-alcoholic Fatty Liver Disease

The major cause of NASH/ NAFLD is chronic liver inflammation induced by tissue damage or pathogen infection (125). Hegde et al. finds that the number of hepatic MAIT cells is decreased in patients with non-alcoholic fatty liver disease-related cirrhosis (126). Compared with controls, cirrhotic liver MAIT cells exhibit an activated phenotype characterized by increasing IL-17 production with no differences in the percentage of MAIT cells producing granzyme B, IFN-γ, or TNF (126). Another study demonstrated that MAIT cells in NASH patients also display an activated phenotype defined by enhanced cytotoxicity but reduced cytokine production (127). These experiments suggest that MAIT cells are activated and contribute to pathogenesis in NAFLD/NASH.

#### Alcoholic Liver Disease

One of the most frequent complications of ALD is bacterial infection. One study has shown that over 50% of severe alcoholic hepatitis patients suffer from bacterial infection (128). As potent antibacterial lymphocytes in the liver, the number, cytokine production (IL-17) and cytotoxic response (Granzyme B, CD107a) of MAIT cell are impaired in peripheral blood of severe alcoholic hepatitis and alcoholic cirrhosis patients (129). Dysfunction of MAIT cells in ALD patients occurs from exposure to bacterial antigens and metabolites, but not ethanol (129). Importantly, in the liver, microarray data show that expression of transcription factors RORC/RORγt, ZBTB16/PLZF, and Eomes that mediate the function of MAIT cells is lower in ALD patients than in controls (129). Together, these results suggest that MAIT cells in ALD display a defective phenotype, which may explain why there is a high rate of bacterial infection complications in ALD patients.

#### REFERENCES


#### CONCLUSIONS

Herein, we have discussed several key aspects of innate T cells and their potential role in liver diseases. Their enrichment in the liver suggests their unique role in liver disease progression and protection. The distinctive features and functions of innate T cells impart both pathogenic and protective abilities to the host. Thus, modulation of these cells represents a very attractive therapeutic strategy in liver diseases.

Our current knowledge of these cell subsets in the liver and their potential role in liver disease mainly comes from studies in animal models. Data from human and clinical studies are insufficient and are primarily complicated by the opposing effect these cell types have, both pathogenic and protective. In addition, most liver diseases are chronic disorders, and this further complicates analysis of these cells. The dynamic effect of these cells at different time points during the progression of disease could be significantly different regarding both number and function. Another critical factor is that a vast number of immuneregulatory cells resides in the liver, where they all modulate the activity level of iNKT and MAIT cells. In turn, these two innate T-cell subtypes also act as key modulators of other immune cell activity, including KCs, classical innate cells (macrophages and DCs) and conventional T cells. These factors form an involved local liver environment; thus, a molecular understanding of these cross-regulatory effects is key to understanding the liver immune system. Understanding liver immunity and function is also key to maintaining a proper balance between immune tolerance and immunity in the liver. To further understand the mechanism of these cells, it will be essential to develop more specific and reliable reagents to characterize and analyse these cells.

#### AUTHOR CONTRIBUTIONS

WHu, WHe, XS, LD, XH and YG contributed to the writing of the manuscript.

#### FUNDING

YG is supported by Natural Science Foundation of Guangdong Province (Grant number: 2018A030313019).

different from circulating lymphocytes. J Hepatol. (1998) 28:84–90. doi: 10.1016/S0168-8278(98)80206-7


negative mice, with the capacity to rapidly secrete large amounts of IL-4 and IFN-γ upon primary TCR stimulation. J Immunol. (1999) 162: 5156.


T cells demonstrates preferential use of several V beta genes and an invariant TCR alpha chain. J Exp Med. (1993) 178:1. doi: 10.1084/jem.178.1.1


model of hepatitis B virus infection. Immunity (2002) 16:583–94. doi: 10.1016/S1074-7613(02)00305-9


steroids: early response to therapy is the key factor. Gastroenterology (2009) 137:541–8. doi: 10.1053/j.gastro.2009.04.062

129. Riva A, Patel V, Kurioka A, Jeffery HC, Wright G, Tarff S, et al. Mucosaassociated invariant T cells link intestinal immunity with antibacterial immune defects in alcoholic liver disease. Gut (2018) 67:918–930. doi: 10.1136/gutjnl-2017-314458

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

# The Impact of Invariant NKT Cells in Sterile Inflammation: The Possible Contribution of the Alarmin/Cytokine IL-33

Maroua Haroun Ferhat <sup>1</sup> , Aurélie Robin<sup>1</sup> , Louise Barbier <sup>2</sup> , Antoine Thierry 1,3 , Jean-Marc Gombert 1,4, Alice Barbarin<sup>1</sup> and André Herbelin<sup>1</sup> \*

1 INSERM U1082 – IRATI Group, Poitiers, France, <sup>2</sup> Service de Chirurgie Digestive, Oncologique, Endocrinienne et Transplantation Hépatique, CHU Trousseau, Université de Tours, Tours, France, <sup>3</sup> Service de Néphrologie, Hémodialyse et Transplantation Rénale, CHU de Poitiers, Poitiers, France, <sup>4</sup> Service d'Immunologie et d'Inflammation, CHU de Poitiers, Poitiers, France

Although the contribution of iNKT cells to induction of sterile inflammation is now well-established, the nature of the endogenous compounds released early after

cellular stress or damage that drive their activation and recruitment remains poorly understood. More precisely, iNKT cells have not been described as being reactive to endogenous non-protein damage-associated molecular-pattern molecules (DAMPs). A second subset of DAMPs, called alarmins, are tissue-derived nuclear proteins, constitutively expressed at high levels in epithelial barrier tissues and endothelial barriers. These potent immunostimulants, immediately released after tissue damage, include the alarmin IL-33. This factor has aroused interest due to its singular action as an alarmin during infectious, allergic responses and acute tissue injury, and as a cytokine, contributing to the latter resolutive/repair phase of sterile inflammation. IL-33 targets iNKT cells, inducing their recruitment in an inflammatory state, and amplifying their regulatory and effector functions. In the present review, we introduce the new concept of a biological axis of iNKT cells and IL-33, involved in alerting and controlling the immune cells in experimental models of sterile inflammation. This review will focus on acute organ injury models, especially ischemia-reperfusion injury, in the kidneys, liver and lungs, where iNKT cells and IL-33 have been presumed to mediate and/or control the injury mechanisms, and their potential relevance in human pathophysiology.

Keywords: iNKT, sterile inflammatory response, alarmin IL-33, ischemia reperfusion, CD1d-restricted T cells, tissue repair

#### STERILE INFLAMMATION

Inflammation is an important biological process that represents a coordinated response of the innate immune system against specific molecular patterns present in pathogens or in the damaged tissues of the host. From protozoans to metazoans, the innate immune system, arising about ∼1,000 million years ago (1), mediates inflammation as a physiological response to insult, infection, and biological stress in order to restore cellular/tissue integrity, maintain homeostasis and ensure host survival. In recent decades, significant advances have been made on endogenous non-pathogenic activators that trigger inflammation. Endogenous initiators of inflammation can act through the same receptors as pathogens and are referred to as damage-associated molecular patterns (DAMPs).

#### Edited by:

Luc Van Kaer, Vanderbilt University, United States

#### Reviewed by:

Jenny E. Gumperz, University of Wisconsin School of Medicine and Public Health, United States Jae B. Kim, Seoul National University, South Korea

> \*Correspondence: André Herbelin andre.herbelin@inserm.fr

#### Specialty section:

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

Received: 05 July 2018 Accepted: 17 September 2018 Published: 15 October 2018

#### Citation:

Ferhat MH, Robin A, Barbier L, Thierry A, Gombert J-M, Barbarin A and Herbelin A (2018) The Impact of Invariant NKT Cells in Sterile Inflammation: The Possible Contribution of the Alarmin/Cytokine IL-33. Front. Immunol. 9:2308. doi: 10.3389/fimmu.2018.02308

**54**

During tissue injury, necrotic cells in response to stress signals (unfolded protein response, oxidative stress response or autophagy) release sterile stimuli, such as DAMPs and alarmins. Hence, in the absence of infection or any pathogenic trigger, sterile stimuli induce the recruitment of inflammatory cells, production of cytokines and chemokines, and induction of T cellmediated adaptive immune responses (2). This phenomenon is the so-called sterile inflammation response, which uses specific or common pathways to recognize pathogens. Indeed, some DAMPs and alarmins activate pathogen recognition receptors (PRRs) such as Toll-like receptors (TLRs) and the NLRP-3 inflammasome (3, 4), while other endogenous alarmins, such as interleukin (IL)-33, high-mobility group box 1 (HMGB1), and IL-1-α, signal directly via specific receptors that are not PRRs (4). Sterile inflammatory response is the initial step toward wound repair mechanisms mediated by macrophages that clear apoptotic neutrophils and produce factors enhancing the resolution of inflammation and the restoration of homeostasis. However, if not resolved, sterile inflammatory responses become pathological (3, 5, 6).

Sterile inflammation is initiated by mechanical, chemical, or metabolic stimuli. It occurs in acute conditions, such as ischemia reperfusion injury (IRI), crystal-induced arthritis, trauma, toxin exposure, labor, and with chronic illnesses, such as particleinduced lung diseases and atherosclerosis (3). The identification of the cellular factors and mechanisms of sterile inflammation represents a major issue in the elaboration of efficient therapeutic strategies in human diseases.

## INKT CELLS IN STERILE INFLAMMATION: FROM CONCEPT TO IN VIVO COMPLETION

#### Concept

Invariant NKT (iNKT) cells, generally recognized as the archetypal cell subset of innate T-cell receptor (TCR)-αβ lymphocytes, are activated during an early stage of inflammation and subsequently contribute to the development and regulation of innate and adaptive immune responses during infection. However, a major feature of iNKT cells is that their activation does not require the recognition of foreign antigens. Indeed, CD1d-restricted presentation of self-antigens to iNKT cells is induced by endogenous stress and may be stimulated by cytokines that are produced by activated dendritic cells (DCs). Depending on the mode of stimulation, activated iNKT cells rapidly secrete either T helper (Th)1 and Th17 cytokines, interferon (IFN)-γ and IL-17A, respectively, to promote inflammatory responses, or Th2 cytokines, IL-4 and IL-10, to enable repair. iNKT cells therefore represent a unique cell population that is able to sense, trigger and resolve sterile inflammation.

# iNKT Cells in the Initiation of Sterile Inflammation: The IRI Model

IRI represents a complex inflammatory immune response that generally occurs in a sterile environment and results in tissue damage. IRI has been well-documented in different animal models and in different organs, including kidneys, liver, lungs, heart, and brain.

Furthermore, iNKT cells contribute to early events induced by IRI in different organs including the kidneys (7, 8), liver (9– 12), and lungs (13). In brain and heart, iNKT cell recruitment corroborates the severity of IRI, suggesting their implication in the inflammatory response (14, 15).

As a common feature, in all of these organs, IRI induces early iNKT cell activation and pro-inflammatory cytokine production, thereby sensing and relaying sterile danger. In the first 24 h following reperfusion, IFN-γ-, Tumor Necrosis Factor (TNF)-αand IL-17A- producing iNKT cells are closely associated with polymorphonuclear leukocyte (PMN) infiltration and tissue damage. Results have suggested that, once activated, iNKT lymphocytes play a key role in the early development and initiation of sterile inflammation, mainly by rapidly producing large amounts of cytokines contributing to PMN recruitment. Indeed, the use of NK1.1-depleting antibodies, iNKT celldeficient mice (Jα18 KO or CD1d KO) or reconstitution of iNKT cells by transfer experiments have definitively confirmed the role of iNKT cells in the initiation of IRI responses in kidney (7, 8) (**Table 1**, **Figure 1A**), liver (9, 11, 12, 16, 17) and lung (13) (**Table 1**, **Figure 1B**, upper panel). Taken together, these studies lead to the conclusion that activation of iNKT cells is a general mechanism for the initiation of IRI. However, the possible involvement of other cell types such as TCR-γδ cells (34– 36) and NK cells (37), and their possible interactions with iNKT cells during IRI remain to be explored.

# INKT CELLS IN THE INITIATION/PROPAGATION OF STERILE INFLAMMATION: MECHANISMS OF ACTION

#### Involvement of Stress-Induced-Endogenous, Self-Antigen Presentation and Cytokine-Driven Signals

The modes of activation and recruitment of iNKT cells in induction of sterile inflammation remain poorly understood. As innate immune system components, iNKT cells are expected to rapidly and efficiently respond to cell-stress and represent an obvious candidate to participate in endogenous pathways of inflammation, especially in sterile inflammation. However, iNKT cells have not been described as reactive to endogenous nonprotein DAMPs that include ATP, uric acid, heparin sulfate and DNA. Moreover, although iNKT cells express the adenosine 2A receptor (A2AR), their reactivity to ligands of this receptor has been described only under pharmacological conditions and has resulted in a downregulation rather than an induction of iNKT cell functions (11, 25) (**Table 1**).

Numerous studies have documented the involvement of TCR/CD1d interactions in iNKT cell activation during sterile inflammation induced by IRI. In most cases, the blockade of CD1d presentation with anti-CD1d antibodies has prevented iNKT cell-mediated renal and hepatic IRI (8, 11, 16, 17). However, further investigations are needed to characterize the endogenous lipids presented by CD1d and to identify the effective


This table recapitulates the main models of acute sterile inflammation, especially IRI models, that involve iNKT cells in different target organs (kidney, liver, and lung). In parallel, when documented, the involvement of IL-33 is indicated; otherwise it is noted as unknown. The implication(s) of iNKT cells and/or IL-33 has (have) been classified according to their contribution to the initiation and/or resolution phase(s) of sterile inflammation, and accompanied by the relevant bibliography. The mode(s) of action of iNKT cells is (are) specified. In the last column, using the cited bibliography, the existence of an iNKT cell/IL-33 axis is shown, surmised (noted ≪ presumably yes ≫) not surmised (noted ≪ presumably not ≫) or difficult to predict (noted ≪ unpredictable ≫). IRI, Ischemia-Reperfusion injury; A2AR, adenosine A2A receptor; α-GalCer, α-galactosylceramide; iNKT, invariant Natural Killer T; KO, transgenic knockout; IFN, interferon; IL, interleukin; NOX-2, NADPH oxidase 2; ↓, decrease.

presenters of CD1d-relevant lipids. It also remains to determine the inducible role of non-protein DAMPs in their expression by CD1d-presenting APCs.

A specific feature of iNKT cells is their ability to be activated by cytokine-driven signals independently of TCR-engagement and CD1d recognition. During sterile inflammation, their innate IFN-γ producing capacity does not require continuous autoantigen recognition in mice (38, 39), corroborating the in vitro demonstration that the pro-inflammatory cytokine IL-12 (alone or in combination with IL-18) can activate iNKT cells to produce IFN-γ. Indeed, IL-12 and IL-18 amplify both Th1- and Th2-like iNKT cell responses upon TCR engagement (40–43). Accordingly, during renal IRI, we have documented an increase of plasma IL-12, while Marques et al. (44) have reported protection of IL-12-deficient mice. Moreover, in a model of sterile liver injury, Liew et al. (24) highlighted a biphasic mechanism of iNKT cell activation through selfantigen presentation and IL-12/IL-18-driven signals. Lastly, during experimental cerebral ischemia, where iNKT cells have been reported to accelerate brain infarction (14), early detrimental T-cell effects have not been associated with adaptive immunity (36). Taken together, these results from the literature demonstrate that iNKT cells mediate acute sterile inflammation, including IRI, through TCR-engagement and cytokine-driven signals (**Table 1**; **Figure 1A**; **Figure 1B**, upper panel).

FIGURE 1 | The paradigm of the iNKT cell/IL-33 biological axis in orchestration of acute sterile inflammation. A schematic overview of the potential involvement of iNKT cells in concert with IL-33 in sterile tissue damage (A,B, upper panel)) and repair (B, lower panel). (A,B, upper panel): Acute sterile organ injury leads to early iNKT cell activation through the passive release of the alarmin IL-33 that binds to the ST2 receptor constitutively expressed by iNKT cells. IL-33 acts as a requisite co-player leading to complete activation and recruitment of these cells. This mechanism could also involve IL-12 and CD1d-dependent presentation of self-ligands. During kidney IRI, IL-33 released by injured cells promotes iNKT cell activation, recruitment and pro-inflammatory cytokine (IFN-γ, IL-17A) production. IFN-γ- and IL-17A- expressing iNKT cells then contribute to the initiation of inflammation by amplifying neutrophil recruitment, and promoting their pro-inflammatory cytokine and reactive oxygen species (ROS) production, thereby resulting in tissue damage (A). This scenario can be applied to other organs such as liver and lung where both of iNKT cells and IL-33 have been shown separately to contribute to IRI or drug-induced organ injury (for details, see Table 1) (B, upper panel). (B, lower panel): Following acute sterile injury, the innate immune system initiates resolution of inflammation and tissue repair by inducing a shift from M1 to M2 macrophages. By their ability to express Th2-type cytokines, iNKT cells likely contribute to this shift. IL-33 might promote this phenomenon both by recruiting monocytes/macrophages expressing CD1d that in turn activate iNKT cells through their TCR engagement and by amplifying subsequent iNKT cell cytokine production. In this context, activated iNKT cells produce large amounts IL-4, but no IFN-γ. This critical step is a requisite for the transition of monocytes/macrophages from a pro-inflammatory to an anti-inflammatory phenotype with IL-10 production. Th2-type cytokines produced by iNKT cells and monocytes/macrophages suppress inflammation and promote PMN apoptosis (B, lower left panel). Furthermore, IL-33 released shortly after injury could target dNKT (type II NKT) cells and regulatory T cells (Treg), that express ST2 receptor, to counteract IFN-γ-expressing iNKT cells and promote immuno-regulatory cytokine production contributing to the resolution of inflammation (B, lower left panel). In the particular case of lung injury, IL-33 promotes the recruitment of iNKT cells and their IFN-γ production. IFN-γ-expressing iNKT could in turn help to resolve inflammation by counteracting ILC2, eonisophils and neutrophils (B, lower right panel). In the liver, IL-33 can also act as an amplifying factor for ligand-activated iNKT cells, thereby contributing to a shift from the initial pro-inflammatory (pro-Th1) profile of iNKT cells into their (pro-Th2) resolutive profile. As a complementary mechanism, IL-33-driven iNKT cells may in turn sustain protective functions mediated by IL-33 in lung and liver due to their capacity to induce subsequent continuous neosynthesis and secretion of IL-33 by alveolar macrophages and hepatocytes, respectively (B, lower right panel). More precisely, in the liver, neosynthetized IL-33 may promote IL-4-producing iNKT cells that are implicated in the resolution of several sterile inflammatory responses, by suppressing PMN infiltration and enhancing hepatocyte proliferation, thereby preserving tissue function. In this scenario, IL-33 can also directly act on hepatocytes to elicit regeneration after tissue damage (B, lower right panel).

# A Key Role for the Alarmin IL-33 in iNKT Cell Activation and Recruitment in Sterile Inflammation?

#### The Archetypal Alarmin/Cytokine IL-33

Alarmins, a second subset of DAMPs, are tissue-derived nuclear proteins, constitutively expressed at high levels in epithelial barrier tissues and endothelial barriers. These potent immunostimulants include defensins, cathelicidin, eosinophilderived neurotoxin, HMGB1, IL-1-α, IL-18, and IL-33. Once released by necrotic cells after tissue damage, they can activate TLRs or cytokine receptors, and serve as early warning signals to alert adjacent cells/tissues and to mobilize innate and adaptive immune systems.

Among alarmins, the newest member of the IL-1 super-family is IL-33, also called IL-1F11. IL-33 has aroused interest due to its singular action during infectious and allergic responses (45, 46), and acute tissue injury (31, 47). Indeed, IL-33 acts as an alarmin released by necrotic cells after tissue damage, and as a cytokine, due to its inducible expression and subsequent continuous secretion by hematopoietic cells like mastocytes and macrophages (31, 45). By interacting with the ST2/IL-1RAcP receptor complex, IL-33 targets adaptive immune cell subsets, namely conventional CD4 and CD8 T cells and regulatory T cells (Treg). IL-33 also targets several innate immune cells including iNKT cells, NK cells, mastocytes, group 2 innate lymphoid cells (ILC2) and myeloid-derived suppressor cells, thereby influencing their functions and homeostasis (42, 48–50). IL-33 is therefore a crucial alarmin and an ubiquitous immune modulator, with a pivotal role in sterile inflammation.

#### The Alarmin IL-33 in Sterile Inflammation: Relevance to IRI Models

It is generally understood that mechanical, chemical, trauma injury and IRI to several organs, including kidney, liver, lung, and brain, lead to rapid release of IL-33, presumably by damaged cells, further supporting the role of IL-33 as an alarmin (7, 18, 29, 31, 47, 51, 52). However, except during brain trauma (47) and kidney IRI (7) (and section First Evidence of an iNKT Cell/IL-33 Biological Axis Mediating Inflammation During Renal IR: Relevance to Liver and Lung?), it remains to be demonstrated that full-length active IL-33 disappears from the nucleus of damaged cells and immediately increases in circulation. The immune effectors of inflammation and repair targeted by IL-33 also need to be identified. In most of these situations, it is well-recognized that IL-33 release precedes iNKT cell recruitment and/or local activation, opening the question of the role of IL-33 as an early warning signal to alert iNKT cells in sterile inflammation responses.

#### IL-33 Targets iNKT Cells

The existence of an iNKT cell/IL-33 biological axis is wellestablished. Indeed, even though IL-33 alone is not able to fully and completely activate iNKT cells, the alarmin can directly target them as an essential amplifying factor in both primary innate and adaptive immune responses. iNKT cells have an immediate biological reactivity to IL-33 because, like NK cells, they constitutively express on their surface the ST2 chain specific of the IL-33 receptor (42, 49). As a result, IL-33 contributes as a co-stimulatory factor to type 1 (IFN-γ), type 2 (IL-4, IL-10), and type 17 (IL-17A) iNKT cell cytokine production profiles upon TCR engagement (7, 42). Moreover, in combination with IL-12, IL-33 enhances IFN-γ production by iNKT cells. Along with recruitment and local activation of iNKT cells, these functions depend on endogenous IL-12 (28, 42). The same demonstration has been documented in other mammals such as humans (49, 53) and pigs (54). Taken together, these data support the conclusion that IL-33 can recruit iNKT cells and contribute as a co-stimulatory factor to pro-Th1-, pro-Th2-, and pro-Th17- iNKT cell responses in an IL-12-dependent manner.

#### First Evidence of an iNKT Cell/IL-33 Biological Axis Mediating Inflammation During Renal IR: Relevance to Liver and Lung?

Given that both iNKT cells and IL-33 have been shown to contribute to acute organ injury in both kidneys and liver, IRI seems to represent a suitable model of sterile inflammation to test the physiological significance of the iNKT cell/IL-33 biological axis. Therefore, we recently provided the first demonstration that endogenous IL-33 contributes as an alarmin to IRI in the kidneys. Indeed, we highlighted IL-33 rapid release (≤1 h) from its constitutively full-length form expressed at high levels in the nuclei of kidney cells and its transient presence in the extracellular space. Moreover, we further characterize a previously undefined mechanism where IL-33/ST2 engagement promotes iNKT cell recruitment, IFN-γ and IL-17A cytokine production, in the presence of IL-12 as a cofactor for IL-33, resulting in PMN infiltration and activation (7) (**Figure 1A**).

We presume that the coordinated action of IL-33 and iNKT cells will also apply to the liver, where they have been separately shown to contribute to IRI, and where, as in the kidneys, IRI severity depends on IFN-γ and IL-17A (11, 18) (**Figure 1B**, upper panel). Regarding lung IRI, an exacerbating role has been attributed to IL-17A-producing iNKT cells, but in this model, IL-33 release as an alarmin to amplify IL-17A expression by iNKT cells has not yet been addressed (**Figure 1B**, upper panel).

#### The iNKT Cell/IL-33 Biological Axis: A Paradigm Extended to Ozone- and Drug- Induced Organ Injury

The iNKT cell/IL-33 biological axis deserves particular attention when considering organ injury caused by chemical components such as ozone or drugs (**Table 1**, **Figure 1B**, upper panel). Indeed, ozone-driven lung inflammation that requires the presence of iNKT cells (32) is associated with IL-33 release by epithelial cells (33). Furthermore, iNKT cells have been shown to be involved in halothane-induced liver injury (22), while blockade of the IL-33/ST2 axis reduces acetaminophenmediated organ injury (23). It remains to determine whether iNKT cells and IL-33 act in a concerted manner to initiate sterile inflammation in response to liver-targeted drugs.

## THE HYPOTHESIS OF A FUNCTIONAL AXIS BETWEEN INKT CELLS AND IL-33 IN THE RESOLUTION OF STERILE INFLAMMATION

## Impact of iNKT Cells on the Resolution of Sterile Inflammation

Inflammation is a physiopathological and protective response of the organism (host) to infection or sterile tissue damage aimed at neutralizing and eliminating the causing agent/insult. Shortly after the beginning of the inflammatory response, a coordinated and active resolution program involving PMN apoptosis and clearance initiates in order to restore tissue integrity and organ function (6). The resolution of acute inflammation is crucial to ensure proper return to homeostasis and to avoid persistent chronic inflammation, including metabolic diseases and autoimmune syndromes.

Innate immune cells are key actors that orchestrate the switch from acute inflammation to resolution. As regards iNKT cells, one may presume that their functional malleability renders them capable of intervening not only in initiation but also in the resolution of sterile inflammation. However, few studies have described the role of iNKT cells in the resolution of sterile inflammation (24, 26, 27, 55) (**Table 1**). For example, in experimental models of acute liver injury (IRI (9, 10) and ConcanavalinA (ConA)-induced hepatitis (56, 57), iNKT cells display a pro-inflammatory deleterious phenotype. A resolutive role attributed to iNKT cells has only been documented in a drug-induced injury model (58), where iNKT cells have been shown to orchestrate a switch from inflammation to resolution of sterile injury. In response to thermal trauma in the liver, another underlying mechanism of resolution is iNKT cell-derived IL-4, which drives the shift from M1 to protective M2 macrophages (**Figure 1B**, lower left panel). Interestingly, this mechanism is similar to a reported model of sterile inflammation in the peritoneum (55). However, IFN-γ-producing iNKT cells, rather than their IL-4-producing counterparts, are resolving in several models of sterile inflammation in the lung (26–28) (**Table 1**). Taken together, these data reveal that iNKT cell functions required to resolve sterile inflammation and to promote tissue repair strongly depend on the organ microenvironment.

In addition to their natural resolutive function, iNKT cells can orchestrate sterile inflammation in the liver when targeted by their pharmalogical ligand α-Galactosyl-Ceramide (α-GalCer). This pharmacological ligand appears to act by shifting the initial pro-inflammatory (pro-Th1) profile of iNKT cells into their resolutive (pro-Th2) profile (21) (**Table 1**).

Other evidence suggests that iNKT cells and type II NKT cells (also called dNKT cells due to their expression of oligoclonal TCRs and recognition of self-antigens, including sulfatide, in a CD1d-dependent manner) have opposing roles at an early stage of liver inflammation (59). Further studies are needed to determinate if the interactions between the two NKT cellsubsets constitute a general mechanism coordinating initiation and resolution of sterile inflammation. Indeed, it was recently reported that pharmacological activation of dNKT cells by the self-glycolipid antigen sulfatide led to reduced IFN-γ secretion by iNKT cells and prevented hepatic and renal IRI (12, 60) (**Figure 1B**, lower left panel).

Up until now, without pharmacological intervention, there has been no reported experimental model in which iNKT cells exercise their initiating and resolution functions in a coordinated manner. Indeed, the highlighting of two opposite functions of the same cell type in a given experimental model is a challenge. Furthermore, in most IRI models, iNKT cell-dependent tissue lesions are severe and render the resolution phase difficult to analyze.

## Are iNKT Cells Involved in Concert With IL-33 to Promote Resolution and Repair in Sterile Inflammation?

Due to its ability to induce epithelial and endothelial cell proliferation (48, 61), the involvement of IL-33 as a factor promoting repair and tissue regeneration following stress has been proposed. Another element supporting the role of IL-33 in tissue repair is the targeting of Treg and ILC2, contributing to resolution of inflammation of the intestine (62), skeletal muscles (63) and skin/wound repair (64). Lastly, IL-33 counteracts pro-Th1 inflammatory responses by targeting the shift from M1 to M2 macrophages (51, 65, 66). Whether these protective IL-33-driven functions are influenced by iNKT cells is unknown because interactions between Treg and iNKT cells, or ILC2 and iNKT cells, have been described only under pharmacological intervention (67, 68).

However, the iNKT cell/IL-33 biological axis can function as an additional mechanism in several situations, like in the peritoneum and liver, where an IL-4-dependent macrophageiNKT cell circuit suppresses the sterile inflammation response (24, 55). As IL-33 can dramatically increase both IFN-γ and IL-4 productions by iNKT cells, it is unclear whether in this context it favors iNKT cell-mediated protection. We presume that this will occur at least during thermal-induced liver injury where iNKT cells produce IL-4 rather than IFN-γ (24) (**Figure 1B**, lower left panel).

In the lungs, IL-33 itself may be crucial for resolution of tissue injury by amplifying IFN-γ expression by iNKT cells (**Figure 1B**, lower right panel). Indeed, when administered at pharmacological doses, IL-33 induces ILC2-mediated airway inflammation (69) and controls this response via a mechanism that involves IFN-γ-expressing iNKT cells (28). From these data, we propose that IL-33 not only targets ILC2 to promote pro-Th2 inflammatory response, but also contributes concomitantly to recruit iNKT cells and activate their production of IFN-γ. In this way, IL-33 acts as a negative feedback loop to resolve lung inflammation, possibly by counteracting neutrophils (70) and/or ILC2 (71). Interestingly, in response to bleomycine, IL-33 potentiates lung injury (29, 30) whereas iNKT cells attenuate the deleterious response by down modulating the Th2 milieu (26) and producing IFN-γ (27). Taken together, these findings favor the hypothesis that IL-33-driven IFN-γ-expressing iNKT cells may represent a natural mechanism of resolution during sterile inflammation in lung (**Figure 1B**, lower right panel).

It should be emphasized that iNKT cells may also contribute in an indirect manner to protective responses driven by IL-33, by amplifying the inducible synthesis of the alarmin/cytokine (see **Figure 1B**, lower right panel). Indeed, iNKT cells themselves contribute to the subsequent transcriptional and protein synthesis of IL-33 by alveolar macrophages and hepatocytes during acute sterile inflammation responses in lung and liver, respectively (19, 69). Thus, in the above-mentioned ozoneinduced lung inflammation model where iNKT cells are requisite (32), a biphasic injury and inflammation controlled by IL-33 has been proposed (31), as previously shown in an antiparasitic response model (45). IL-33 is first released as an alarmin by epithelial cells and then its synthesis is relayed by alveolar macrophages (vs. mast cells in the anti-parasitic model), thereby triggering tissue protection/repair (vs. boosting a pro-Th2 protective antiparasitic response). It would be important to determine whether iNKT cells contribute to this biphasic function of IL-33. Lastly, during acute hepatic injury induced by ConA, iNKT cell-dependent IL-33 synthesis occurs precisely in hepatocytes (19). As IL-33 is recognized as hepatoprotective (20), one may hypothesize that iNKT cells, by increasing the synthesis of IL-33 in hepatocytes, contribute to the repair phase of hepatocyte damage (**Figure 1B**, lower right panel). This hypothesis can also be applied to hepatic IRI, where IL-33 synthesis is induced in hepatocytes from 4 h after reperfusion (18).

#### THERAPEUTIC STRATEGIES TARGETING THE INKT CELL/IL-33 BIOLOGICAL AXIS

While recent studies have highlighted the pathogenic mechanisms of iNKT cell responses, the beneficial functions of iNKT cells are just beginning to be explored. Pharmacological and cell-based therapies influencing iNKT cell responses in experimental acute organ injury suggest that this is a promising approach for the preservation of organ function during sterile inflammation. We surmised that this approach should take into account the potential implication of IL-33.

Activation of iNKT cells by their pharmacological ligands during sterile inflammation, especially IR, can be protective or exacerbating, depending on the state of APCs (72) and the organ involved (14, 15, 21). IL-33 works as an amplifying factor for ligand-activated iNKT cells (7, 42) and as a protective factor by targeting Treg and ILC2 (50) or parenchymal cells like hepatocytes (73) and myocardiocytes (74, 75). As a result, it can be argued that protection driven by α-GalCer would be more robust if the iNKT cell ligand was co-administered with IL-33. Indeed, in liver and heart IR models, the delivery of either α-GalCer (74, 75) or exogenous IL-33 alone (15, 21) has shown protective effects (**Table 1**; **Figure 1B**, lower right panel). In this context, the contribution of IL-33 to the α-GalCer-driven iNKT cell/Treg cross-talk (67, 68) deserves particular attention.

Treatment with IL-33 alone would also be beneficial when sterile inflammation is accompanied by the natural protective effects of iNKT cells (24, 55). Conversely, in kidney IR models where initiation of inflammation depends on the iNKT cell/IL-33 biological axis, IRI would be counteracted by early blockade of IL-33 together with agonist A2AR treatment, known to attenuate IFN-γ-expressing iNKT cell activation (11, 25).

Taken as a whole, even though therapeutic strategies targeting the iNKT cell/IL-33 biological axis are promising, they might be complex to develop, given that the contribution of iNKT cells and IL-33 in inflammation to exacerbated illness vs. improved recovery, largely depends on the model.

### CONCLUDING REMARKS

Focusing on recent advances in the understanding of the biology of iNKT cells and the alarmin IL-33 obtained from animal models, we propose that iNKT cells and IL-33 form a biological axis in alerting and controlling the immune cells involved in sterile inflammation associated with tissue damage.

The review shows that the involvement and the underlying mechanisms of the iNKT cell/IL-33 biological axis in sterile inflammation depend on the model, organ microenvironment, and the initiating vs. resolutive/repair phase of the inflammation response. Even though more work is required in this area, this review brings new evidence that the iNKT cell/IL-33 biological axis acts during sterile inflammation induced by IR in kidneys and liver and after chemical- or drug- induced acute tissue injury in the kidneys, liver and lungs.

At this stage, an important challenge is to determine how much of the current information in animal models can accurately be translated to human patients. Given the widely recognized deleterious action of the iNKT cell/IL-33 biological axis in renal IR in mice, an immediate challenge would be to improve understanding of the physiopathological impact of this biological axis during IR sequences in organ transplantation. As far as kidney transplantation is concerned, a pilot clinical study indicated a prompt release of lL-33 into the circulation early after organ reperfusion that could be responsible for the activation of iNKT cells (53). Moreover, IL-33 levels and IRI duration are correlated, supporting a close connection between kidney cell injury, IL-33 release and iNKT cell activation. That iNKT cells and IL-33 may also function in a coordinated manner after liver transplantation in humans is an attractive hypothesis and deserves further investigation in our opinion.

In fine, new therapeutic strategies targeting the iNKT cell/IL-33 biological axis could prove beneficial for the long-term survival of organs after acute organ injury. If so, the secondary "repair" functions of the iNKT/IL-33 biological axis, once identified, would need to be protected from treatment focused on the initial, deleterious phase.

More generally, exploration of the crucial and diverse roles of the iNKT cell/IL-33 biological axis during acute sterile inflammation settings may contribute to understanding the mechanisms that control the switch between healthy and pathological inflammation.

# AUTHOR CONTRIBUTIONS

LB, AT and J-MG contributed to literature search for the review. AB and AR contributed to literature search and writing for the review. MF and AH contributed to the literature search for the review, provided writing, and editing of the review.

#### ACKNOWLEDGMENTS

The authors are especially indebted to Jeffrey Arsham for editing the English of their manuscript. This study was

#### REFERENCES


supported by INSERM, CHU de Poitiers, Université de Poitiers, Association pour la Recherche en Immunologie-Poitou Charentes (ARIM-PC), FHU support and Ministère de la Recherche. MF and AB were supported by fellowships provided by INSERM, Région Poitou-Charentes and Octapharma, and Région Nouvelle Aquitaine, respectively.


requires the presence of natural killer T cells and IL-17. J Exp Med. (2008) 205:385–93. doi: 10.1084/jem.20071507


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

# The Pathophysiological Relevance of the iNKT Cell/Mononuclear Phagocyte Crosstalk in Tissues

#### Filippo Cortesi <sup>1</sup> \*, Gloria Delfanti 1,2, Giulia Casorati <sup>1</sup> \* and Paolo Dellabona<sup>1</sup> \*

<sup>1</sup> Experimental Immunology Unit, Division of Immunology, Transplantation and Infectious Diseases, IRCCS San Raffaele Scientific Institute, Milan, Italy, <sup>2</sup> Università Vita-Salute San Raffaele, Milan, Italy

CD1d-restricted Natural Killer T (NKT) cells are regarded as sentinels of tissue integrity by sensing local cell stress and damage. This occurs via recognition of CD1d-restricted lipid antigens, generated by stress-related metabolic changes, and stimulation by inflammatory cytokines, such as IL-12 and IL-18. Increasing evidence suggest that this occurs mainly upon NKT cell interaction with CD1d-expressing cells of the Mononuclear Phagocytic System, i.e., monocytes, macrophages and DCs, which patrol parenchymatous organs and mucosae to maintain tissue homeostasis and immune surveillance. In this review, we discuss critical examples of this crosstalk, presenting the known underlying mechanisms and their effects on both cell types and the environment, and suggest that the interaction with CD1d-expressing mononuclear phagocytes in tissues is the fundamental job of NKT cells.

Keywords: NKT cells, CD1d, monocytes, macrophages, DC, microenvironment

#### INTRODUCTION

Natural Killer T (NKT) cells are a subset of T lymphocytes with innate-like functions characterized by the ability of recognizing lipid antigens presented by the major histocompatibility complex (MHC)-related molecule CD1d (1). NKT cells can be divided into two groups according on their TCR usage. Type I or invariant (i)NKT cells express a TCR made by the invariant rearrangement Vα14-Jα18 (TRAV11–TRAJ18) in mice, and the orthologous Vα24-Jα18 (TRAV10–TRAJ18) in humans, paired with diverse β-chains that utilize a restricted set of Vβ genes (2, 3). Type II NKT cells express different, yet poorly diverse, TCRs other than the semi-invariant Vα14/Vα24 one (4, 5).This review will focus on iNKT cells, the most represented and best characterized subset.

iNKT cells are endowed with a constitutive (i.e., innate) effector-memory phenotype: unlike mainstream MHC-restricted T cells, they rapidly produce large amounts of inflammatory and regulatory cytokines and chemokines upon activation without prior antigen sensitization (6, 7). This innate reactivity, together with their primary localization in tissues, makes iNKT cells effective sentinels of tissue integrity. Mouse and human iNKT cells have been found in lung, intestinal and urogenital mucosae, skin, fat, parenchymatous organs, as well as secondary lymphoid organs. There, they respond to two main types of stimuli, resulting from cell damage and inflammation induced upon pathological processes, namely: (i). signaling from pro-inflammatory cytokines, particularly IL-12 and IL-18 (8, 9); (ii). recognition of microbial or autologous (self) agonist lipids presented by CD1d, which derive from infecting pathogens and from biosynthetic pathways upregulated by stress in immune cells, respectively (10–15). A critical aspect of this function, supported by increasing body of evidence, seems to be represented by the highly regulated crosstalk

#### Edited by:

Michael Loran Dustin, University of Oxford, United Kingdom

#### Reviewed by:

Mauro Nicolas Gaya, INSERM U1104 Centre d'immunologie de Marseille-Luminy, France Jose Alberola-ila, Oklahoma Medical Research Foundation, United States

#### \*Correspondence:

Filippo Cortesi cortesi.filippo@hsr.it Giulia Casorati casorati.giulia@hsr.it Paolo Dellabona dellabona.paolo@hsr.it

#### Specialty section:

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

Received: 26 June 2018 Accepted: 24 September 2018 Published: 12 October 2018

#### Citation:

Cortesi F, Delfanti G, Casorati G and Dellabona P (2018) The Pathophysiological Relevance of the iNKT Cell/Mononuclear Phagocyte Crosstalk in Tissues. Front. Immunol. 9:2375. doi: 10.3389/fimmu.2018.02375

**64**

between iNKT cells and a broad range of CD1d-expressing cell populations of the mononuclear phagocyte system (MPS), represented by monocytes, macrophages, and DCs (10, 14–19). Owed to its extensive diversity and plasticity, the MPS plays essential functions in the organism, including tissue maintenance and healing, innate immune responses and pathogen clearance, and the induction of adaptive immune responses (20–22). Importantly, the cells of the MPS express CD1d in mice and humans and are strategically positioned in tissues to sense stress and convey it to iNKT cells to coordinate a rapid reaction against it. Through these bidirectional interactions with MPS, iNKT cells rapidly modulate the local microenvironment for an immediate tissue reaction, concurrently helping the induction of subsequent adaptive immune responses. In this review, we propose that the interaction with CD1d-expressing MPS in tissues is the fundamental job of iNKT cells, and we will provide examples of the pathophysiological relevance of such interplay.

#### MECHANISTIC ASPECTS OF THE INKT-MYELOID CELL CROSSTALK

The relevance of the interplay between iNKT cells and MPS populations can be defined as not univocal and linear (19), but dependent on several factors that can impact the reciprocal cell regulation in vivo such as: (i) the strength of cognate antigen/iTCR signal, co-stimulation and the maturation state of the mononuclear phagocytic cell; (ii) the iNKT cell subset involved in the interaction; (iii) the physiological vs. pathological status of the host. In this review, we add the tissue context as a fourth factor that has acquired relevance in recent years, as accumulating evidences are highlighting the importance of a fineregulated crosstalk between iNKT cells and CD1d-expressing MPS in tissues for the biology of these cells.

The iNKT cell subsets involved in the interaction with MPS cells and the tissue context are strongly interconnected. Different tissues contain distinct composition of resident iNKT cell subsets, at least in mice (23–26). Based on the differential expression of three key transcription factors (PLZF, Tbet, RORγt) involved in the determination of specific effector phenotypes, mouse iNKT cells acquire TH1- (NKT1, PLZFlow, Tbet+, RORγt <sup>−</sup>), TH2- (NKT2, PLZFhigh, Tbet−, RORγt <sup>−</sup>), and TH17-like (NKT17, PLZFint, Tbetlow, RORγt high) cytokine profiles already upon thymic development. Recent reports suggest that this subsets definition for iNKT cells may not entirely represent the whole spectrum of effector functions displayed by these cells, as their effective cytokine production can sometimes deviate from the one expected from their transcription factor profile (27, 28). This suggests both that iNKT cells may undergo some sort of post-selection functional tuning, and the need for a more comprehensive phenotypical and functional analysis to define their effector profiles. Nevertheless, each known iNKT cell subset egresses from the thymus to survey different peripheral compartments. In C57BL/6 mice, NKT1 cells comprise the >95% of all hepatic iNKT cells, and are also predominant in the prostate, while NKT2 and NKT17 (29) are highly enriched in the intestine and lung mucosae, respectively. In secondary lymphoid organs, NKT1 and some NKT2 cells are contained in the spleen, while LNs harbor NKT1, low NKT2, and expanded NKT17 cells, with the notable exception of mesenteric LNs and Peyer's Patches, in which iNKT2 represent up to 40% of iNKT cells (24, 30). The adipose tissue contains a distinct IL-10 producing regulatory iNKT cell subset (NKT10) (25), which lacks PLZF but express the transcription factor E4BP4, and whose thymic vs. peripheral differentiation is currently unknown (31, 32). The relative frequency and tissue distribution of the iNKT cell subsets varies substantially between different mouse strains, likely correlating with the different dominant types of effector responses classically observed in each strain (24). iNKT cells are sessile cells that exhibit remarkable tissue-residency and limited recirculation, with the notable exception of those cells found in the peripheral blood (23, 25). Together, these characteristics confer iNKT cells a fundamental role in the tissue homeostasis and immune architecture: based on their main cytokine profiles they display in different tissues, iNKT cells modulate in different directions the effector response of the mononuclear phagocytic cells they interact with (33).

The pathophysiological status of the host can also influence iNKT cell distribution and subset balance, which may directly reflect on their communication with the MPS. For instance the relative composition of NKT1, NKT2, and NKT17 cells in a given tissue may be altered from physiology to pathology, as observed in prostate cancer progression (26), or in adipose tissue in lean and obese subjects (34, 35), impacting the quality of the resulting effector functions. This is an intriguing observation, which points to unanticipated effector plasticity and/or ability to migrate into different tissues of iNKT cells that would be relevant to understand.

A parallel aspect impinging substantially on the iNKTmyeloid cell crosstalk is represented by the functional plasticity characterizing the cells of the MPS, particularly monocytes/macrophages, which directly impact the pathophysiological status of the host. Indeed, monocytes are able to differentiate throughout a broad spectrum of effector phenotypes ranging from strongly pro-inflammatory and tissue damaging, to anti-inflammatory and tissue repairing profiles. For macrophages, this complex functional spectrum has been (over)simplified in the widely recognized paradigm of pro-inflammatory M1 and anti-inflammatory M2 populations, mirroring the TH1 and TH2 states of T cells (36), which represent the two functional extremes of the spectrum (37, 38). In vivo, however, macrophages appear often to exhibit mixed phenotypes, with a variable M1/M2 balance, which are modulate by the combination of molecular and cellular signals contained in the local microenvironment, implying a remarkable functional plasticity of this cell population (39).

The interplay between iNKT cells and MPS cells is mutual and embraces different aspects. iNKT cells depend for their functional education on CD1d<sup>+</sup> mononuclear phagocytes (40, 41). At the same time, the maturation and polarization of DCs and monocytes is promoted by iNKT cells (42, 43). Several mechanisms could underlie this interplay, including CD1d engagement (44), cytokine production (45), CD40 ligation (46, 47), purinergic signaling (48, 49). iNKT cell-dependent signaling cues indeed direct the acquisition of either pro-inflammatory or anti-inflammatory effector phenotypes of myeloid cells (50– 53). Based the above considerations, the outcome of the interconnections between iNKT cells and MPS cells in specific anatomical sites can thus be quite different.

#### SECONDARY LYMPHOID ORGANS

iNKT cell distribution in secondary lymphoid organs allows them to exert their "adjuvant" functions for both innate and adaptive immune response, culminating in the non-cognate or cognate help to B cell responses (54–58). In popliteal LNs at steady state, endogenous iNKT cells localize in the interfollicular region and medulla, but not in the T-cell-rich paracortex (59), whereas adoptively transferred iNKT cells are found in the paracortex (60), possibly reflecting the different methods used to detect the cells in situ. In the steady state spleen, both autochthonous and adoptively transferred iNKT cells are found widely distributed throughout the parenchyma, including B and T follicles in the white pulp, the marginal zone (MZ) and the red pulp (56, 61). This iNKT cell distribution at is substantially modified upon antigen-dependent activation. In the popliteal LNs, upon immunization of mice with particulated antigens formulated with the strong lipid agonist αGalCer, the adoptively transferred iNKT cells rapidly move from the paracortex to contact CD169<sup>+</sup> macrophage lining the subcapsular sinus, which express CD1d and can present lymph-borne soluble antigens, resulting in a strong iNKT cell activation and secreting copious amounts of helper cytokines (60). In the spleen, the injection of soluble antigen formulated with αGalCer, or of pathogenic bacteria containing stimulatory glycolipids, results in the massive accumulation, within 8 h from administration, of splenic iNKT cells in the MZ, where the cells are activated upon contacting CD1d+DCs, and possibly also macrophages (61, 62). This iNKT cell re-distribution in secondary lymphoid organs as several functionally relevant consequences for the immune response: (1). It leads to the contact-dependent maturation of macrophages, which can limit potential pathogen spreading in secondary lymphoid organs, and of DCs, which relocate to T cell zones and promote downstream adaptive T and B cell responses, resulting in the so-called non-cognate iNKT cell help (42, 55, 59–62); (2). It elicits the secretion of copious amount of different helper cytokines by iNKT cells, which can stimulate innate and adaptive immune effectors throughout the LN and splenic parenchyma (60, 61); (3). It results in the acquisition of a follicular helper effector phenotype by iNKT cells (iNKTFH: Bcl6highCXCR5highPD-1high) (57, 63, 64), which can ultimately enter into the B cell follicles and help CD1d-expressing B cells presenting the same lipid antigen, providing the cognate iNKT cell help (56, 57, 65). In fact, although the interaction with CD1d-expressing B cells is fundamental to sustain the full iNKTFH cell differentiation and functions (56), the upregulation of the follicular helper molecules by iNKT cells requires the recognition, in first place, of CD1d-expressing myelomonocytic APCs (56), most likely DCs (61), but possibly also CD169<sup>+</sup> macrophages (60, 62). Interestingly, a recent study has gained new mechanistic insight into the critical role of the interaction between LN-resident CD1d+CD169<sup>+</sup> macrophages and endogenous iNKT cells for the delivery of non-cognate help to B cells, activated in the course of influenza virus infection (28). Indeed, as early as 3 days upon influenza virus infection, iNKT cells are found in contact with "stressed" CD1d+CD169<sup>+</sup> macrophages of the subcapsular sinus, in analogy with the results obtained by injecting particulated Ags containing αGalCer. There, iNKT cells become activated by CD169<sup>+</sup> macrophages via CD1dcognate Ag stimulation and secretion of IL-18, without acquiring iNKTFH phenotype. This activation, in turn, induces rapid iNKT cell migration at the B cell follicular border and the secretion of copious IL-4, which is critical for the early phase of germinal center formation and anti-viral antibody responses. Expression of CD1d on macrophages, but not on B cells, is required to elicit IL-4 production by iNKT cells, and mice lacking macrophages or IL-4 develop fewer germinal centers and less influenza specific IgG1 than wild-type mice (28). It is intriguing that IL-18 release by sinus-lining LN macrophages, induced upon inflammasome-depending pathways activated by pathogen-related innate signals, can also elicit the rapid secretion of protective IFNγ by a network of innate and innatelike effectors that include iNKT cells, which are strategically prepositioned for pathogen sensing in secondary lymphoid organ (59).

Collectively, these evidences support a critical role for the iNKT/MPS cell axis in the lymphoid system to rapidly sense infections and damage, and immediately react by promoting local and systemic innate and adaptive immune responses.

# THE LIVER

iNKT cells are the prominent T cell subset in the mouse liver, accounting for up to 30% of T lymphocytes. They are also present in the human liver, though at a 30 times lower frequency; nevertheless, both mouse and human hepatic iNKT cells undergo quantitative and qualitative dynamic changes in chronic inflammation/infections or cancer, suggesting active involvement in the pathological processes affecting the organ (66–68). iNKT cells crawl under basal conditions in liver sinusoids and arrest upon stimulation by cognate antigen recognition, or exposure to inflammatory cytokines IL-12 and/or IL-18 (69–71). The liver contains a rich monocytic/macrophage component, comprising Kupffer cells, which are self-maintaining, tissue-resident phagocytes originating from embryonic yolk sac, and monocytederived macrophages. Kupffer cells and macrophages adjust their phenotypes in response to local signals, which determine their ability to worsen or end liver injury. Both mononuclear phagocyte types express CD1d and can interact with liver iNKT cells, resulting in such functional reprogramming. A paradigm of this function has been highlighted by a recent study using a model of focal hepatic sterile thermal injury assessed by intravital microscopy, revealing that iNKT cells stop and are activated by IL-12, IL-18, and the recognition of self-stress lipid(s) presented by CD1d-expressing CCR2highLy6Chigh

inflammatory monocytes migrating into the injured area. Interestingly, the self-lipid(s)+cytokine stimulation results in iNKT cell production of IL-4, but not IFN-γ, which promotes the transition from inflammatory to reparative (CCR2lowLy6Clow CX3CR1high) monocytes, ultimately leading to the healing of the injury by collagen deposition, wound revascularization and hepatocyte proliferation (53). Interestingly, human iNKT cells extracted from chronic HBV or HCV infected cirrhotic livers exhibit an IL-4high/IFNγ low effector profile skewing, compared to iNKT cells from non-chronic viral infections (72). This is consistent with a pro-fibrotic and tissue repair activity that, in the context of a sustained liver injury, can lead to a pathological form of tissue regeneration. However, in mice, there are also examples of potent IFNγ production by iNKT cells elicited by Kupffer cells during Borrelia burgdorferi liver infections (71), or upon provoked inflammation and autoimmunity, which promotes M1 polarization of the attracted peritoneal macrophages and, in these cases, sustains tissue damage (73, 74). It is possible that the opposite effector responses dominated by IL-4 or IFNγ observed in sterile vs. infectious inflammation may be related also to the different antigenic potency of self vs. bacterial lipid antigens that activate hepatic iNKT cells. Hence, the iNKT cell/MPS crosstalk in the liver is multifaceted depending on the underlying pathological situation, the inflammatory cell type involved, and the weak vs. strong antigen stimulation. All these parameters, collectively, can lead to either tissue repair or damage through the reciprocal modulation of both iNKT cell and macrophage effector functions, even though liver resident iNKT cells are essentially all NKT1 at start. This observation suggests the possibility that the effector profile of liver iNKT cells may change in different pathological situation. As already discussed above, because iNKT cell are reported to be sessile and functionally rigid, an interesting question is whether, under pathological stimuli, liver iNKT cells may either be replaced by newly recruited ones that are endowed with different effector profiles, or undergo functional reprogramming in the organ, implying an unexpected functional plasticity that may apply also to other organs.

### THE PERITONEUM AND OMENTUM

The peritoneum forms a unique microenvironment, which is formed by a thin mesenchymal membrane that lines the abdominal cavity and surrounds the visceral organs. The omentum is a large apron-like peritoneal fold that connects the spleen, pancreas, stomach and transverse colon (75), which encloses adipocytes and specialized compact structures ("milky spots") containing macrophages, DCs, B cells, T cells and mast cells (76). The omental adipocytes expand in obesity, linking the omentum to the adipose tissue and the metabolic control (see below). The peritoneum is an active immune site, in which both branches of the immune system contribute to maintain homeostasis (77). In the murine peritoneum, iNKT cells are present in sizable quantity (78), while the human omentum is highly enriched in iNKT cells, at least 10 time more than any other human organ analyzed (34). Evidences suggest a close interplay between iNKT cells and the abundant population of CD1d<sup>+</sup> macrophages found within the peritoneal membrane. iNKT cells negatively correlated with mouse survival in a model of abdominal sepsis (79, 80), while induction of abdominal sepsis in the peritoneum of iNKT cell-deficient (Jα18−/−) mice results in the reduction of Ly6Clow anti-inflammatory macrophages and decreased mortality compared to WT. The critical interplay between peritoneal iNKT cells and macrophages is further illustrated by a model of acute sterile inflammation, in which peritoneal macrophages phagocyte neutrophils (efferocytosis) leading to CD1d upregulation and IL-4 secretion. This process activates iNKT cells to produce large amounts of IL-4 that, in concert with the macrophage cytokine, sustains the M2 like polarization and the resolution of the inflammation (81). In vivo, peritoneal CD4<sup>+</sup> iNKT cells are the major producers of IL-4 (81), suggesting the possibility that peritoneal iNKT cells are either NKT2, or acquire NKT2 phenotype upon stimulation.

# THE ADIPOSE TISSUE

The immune system contained in adipose tissue (AT) is unique. Sizable quantities of innate-like T cells reside in the omentum and visceral AT of mice and humans (34, 82). Here, iNKT cells primarily interact with CD1d-expressing macrophages (83) and adipocytes (84) to maintain non-inflammatory conditions. In fact, the AT is a sophisticated sensor of metabolic alterations induced by dietary stimuli, and the status of AT-resident macrophages is of great importance for the physiological metabolic control at this site: pathological metabolic alterations associated with obesity results in profound modification of ATmacrophages, inducing pro-inflammatory (M1-like) functions and a consequent increase in local inflammation and insulin resistance (35, 85). Regulatory iNKT10 cells are selectively enriched in the AT and rapidly respond to stimulatory lipids presented by CD1d+ macrophages, or adipocytes, by secreting IL-4 that restrains M1 and promotes M2 polarization (25, 35, 83, 85, 86). However, a prolonged dysmetabolic state provokes down-regulation of CD1d on AT-M2 cells and their switch to an M1-like phenotype that, in turn, leads also to a proinflammatory shift of local iNKT cells (83, 87, 88), again suggesting a plasticity due to the migration of iNKT1 cells from other sites or a functional differentiation of local cells. The presence of iNKT cells in the AT, which is conserved between mouse and human, is crucial for the formation of fat-associated lymphoid clusters (FALC). FALC are noncapsulated structures in the adipose tissue that collect TH2 skewed immune cells, most notably ILC2 (89), which direct the polarization of B1 cells, eosinophils and M2 macrophages (90) in order to maintain the homeostasis in the tissue. FALC are absent in CD1d−/<sup>−</sup> mice, while they can be induced following iNKT cell adoptive transfer in Rag2−/−, suggesting the critical dependency of these structures on iNKT cells (91). Under peritoneal inflammation, the activation of iNKT cells increases the formation of FALC, indirectly inducing the recruitment of beneficial anti-inflammatory myeloid cells and the resolution of inflammation.

#### THE GASTRO-INTESTINAL SYSTEM

In mice, under homeostatic conditions, gut infiltrating iNKT cells (small intestine and lamina propria) are NKT1 (>90%) or NKT17 (<10%) (92). NKT2 are barely detectable in the intestinal epithelium, although they represent up to 40% of iNKT cells of the mesenteric lymph node (LN) and of those infiltrating Peyer's Patches (24). The accumulation of iNKT cells in the small intestine and mesenteric LN has been confirmed also in humans (30). Intestinal macrophages maintain gut homeostasis through the clearance of enteric pathogens and the enforcement of the tolerance to food and microbiota antigens via the production of IL-10 (93, 94). Recent evidences point out that a heterogeneous CD11c<sup>+</sup> myeloid population, which includes both DCs and macrophages, stimulate iNKT cells in the gastro-intestinal system (92), resulting in the control of the intestinal bacteria composition and compartmentalization, regulation of the IgA repertoire and induction of regulatory T cells within the gut. The recognition of microbial lipid products is pivotal for the physiology of intestinal iNKT cells (95–97). In this context, α-glycolipids that are recognized from iNKT cells can originate from the commensal flora (98, 99), or from the diet (100). Upon CD11c<sup>+</sup> myeloid cell-activation, NKT17 and NKT2 cells in the mesenteric LN undergo rapid activation and expansion, suggesting a pathogenic role for these cells in ulcerative colitis (101).

#### THE LUNGS

In the steady state, the mouse lung contains iNKT cells that distribute predominantly in the vasculature, with a minority residing in the interstitium, which are belong to clearly distinct functional subsets. Whereas the majority of the lungassociated vasculature cells are NKT1, the lung interstitium contains the highest frequency of NKT17 in C57BL6 mice (>50%) (24, 102), which is consistent with their involvement in pathogen surveillance. Barrier epithelia (e.g., lungs, colon, skin, LN) produce elevated quantitates of IL-7 (103) which drives NKT17 survival and maintenance (104), thus creating a microenvironment favorable for the accumulation of these effector cells. The clearance of inhaled pathogen is the main feature of lung-resident (alveolar) macrophages (51). The iNKT cell-macrophages axis is once again critical in this context. In a model of viral-induced chronic airway inflammation, iNKT cells are directly recruiting and activating macrophages toward an anti-inflammatory, tissue remodeling M2-state (105). Increased amounts of iNKT cells and of IL-13 producing macrophages have been detected not only in mice, but also in patients with chronic obstructive pulmonary disease (COPD) (105, 106), supporting the involvement of the iNKT cell/macrophage crosstalk in the lung pathophysiology.

iNKT cells react also to a number of pathogens involved in airway infections, including Sphingomonas capsulata (107), Mycobacterium tuberculosis (108), Pseudomonas aeruginosa (109, 110), Streptococcus pneumoniae (111) and Influenza A virus (112, 113), via involvement of local mononuclear phagocytic cells, particularly macrophages. During M. tuberculosis infection in mice, iNKT cells are activated upon interaction with macrophages presenting mycobacterium-specific lipids (108) and help controlling the bacterial load via GM-CSF production (114), which may promote an inflammatory response that ultimately leads to bacterial clearance. A similar mechanism has been identified for P. aeruginosa, where iNKT cells stimulate increased phagocytic clearance of the bacteria in the lung by alveolar macrophages (109). Interestingly, in this context, iNKT cells have a stronger effect in controlling P. aeruginosa in BALB/c compared to C57BL6 mice (110). This difference can be explained by the different iNKT cell subsets that infiltrate the lungs of the two strains, as BALB/c mice contain an higher frequency of NKT1 subset compared to C57BL6 (24, 115). In the case of S. Pneumonia infection, intravital microscopy reveals that interstitial DCs present microbial glycolipids to the few adjacent iNKT cells, resulting in the neutrophil recruitment and CCL17 production. This promotes further iNKT cell migration from vasculature into acutely inflamed lung interstitium, where they assist DC activation and clearance of infection (116). This mechanism for acute inflammation seems conserved also in humans, as suggested by the human iNKT cell ability to drive in vitro the release inflammatory lipid mediators by monocytederived DCs, which can promote neutrophil recruitment and activation (48).

In addition to controlling bacterial infections, iNKT cells were also active in containing pulmonary infection influenza A virus. In this context, iNKT cells orchestrate anti-viral NK and CD8<sup>+</sup> T cell responses (113, 117–119). iNKT cells promote virus control also by promoting differentiation into functional APC of lung-infiltrating immature myeloid derived cells, through CD40 engagement and CD1d cognate recognition (17), or by reducing pathogenic inflammatory monocytes (Ly6ChighLy6G−) via direct lysis (112), which correlates with better influenza outcome in iNKT cell-sufficient compared to insufficient mice.

# THE TUMOR MICROENVIRONMENT

Cancer cells are embedded in the tumor microenvironment (TME), a complex and active milieu in which transformed and non-transformed cells dynamically interact in evolving networks that are continuously rearranged (120). The composition of the TME impinges heavily on the success of cancer therapy, and many studies underline the importance of targeting both the tumor and the supporting stroma for an effective and complete clearance of the malignancy (121). A substantial fraction of immune infiltrate of the TME is composed by tumor associated macrophages (TAMs) (122), which can encompass a spectrum of activation states largely affecting tumor progression, dissemination and response to therapy (36). Different stimuli present in the microenvironment can also rapidly trigger a number of diverse functions in macrophages, which range from the activation of potent pro-inflammatory M1 like responses, to the coordination of M2-like tissue remodeling and immunosuppression.

Despite their low numbers, iNKT cells are also components of the immune infiltrate present in both mouse and human tumors (26, 78, 123–125). Indeed, a growing body of evidences lends support to a critical role for these cells in modulating myelomonocytic cells in the tumor microenvironment. M1 oriented TAMs are generally beneficial for the control of tumors because by exerting critical functions such as antigen presentation, production of inflammatory cytokines and inhibition of angiogenesis (126, 127). By contrast, M2-like TAMs are detrimental, because they exert tumor-supporting, pro-angiogenic, pro-metastatic, and immunosuppressive activities (128). The first hints of iNKT cells interplay with TAMs come from the observation that these cells can kill in a CD1d-dependent manner transferred human macrophages infiltrating a xenograft model of human neuroblastoma in NOD/ SCID/IL-2Rγ-null (NSG) mice (129). The importance of the iNKT cell-TAM crosstalk is further strengthen in the same model, by showing that iNKT cells are recruited into tumor in a CCL20-dependent manner, but inhibited in their anti-tumor activity by macrophage-induced hypoxia (125). In the recent years, this dual relationship has been investigated more in detail. By using a mouse model of oncogene-induced pancreatic cancer, iNKT cells have been shown to have a preferential activity on M2-like macrophages, which are increased in CD1d−/<sup>−</sup> pancreatic cancers (130). iNKT cells delay also the onset and organ infiltration of a mouse model of chronic lymphocytic leukemia (CLL), and their counts in blood independently predicts disease stability in CLL patients (78). iNKT cells remodel the supporting niche of CLL by controlling CD1dexpressing, patient-derived M2-like macrophage population, termed nurse-like cells (NLCs), which sustain leukemia cell survival (78, 131). The unique mechanism by which iNKT cells selectively modulate different subset of TAMs has been recently elucidated in a model of autochthonous prostate cancer (26). In this model, the presence of iNKT cells causally associates with the selective reduction of M2-like TAMs in the tumor microenvironment, leading to the control of tumor progression. Human prostate cancer aggressiveness correlates with reduced intra-tumoral iNKT cells and increased M2 macrophages, underscoring the clinical significance of this crosstalk (26). This selective restriction of M2 TAMs depends on the combinatorial engagement of CD40 and Fas on the surface of macrophages by tumor-infiltrating iNKT cells. Although both molecules are expressed to similar levels on either M1 or M2 TAM populations, the CD40L-CD40 pathway supported the survival only of the M1 population, likely by antagonizing the apoptotic death driven by Fas signaling. By contrast, CD40 expression does not protect M2 TAMs form FAS-dependent killing, suggesting a differential CD40 signaling between M1 and M2 macrophages. Remarkably, the ability to selectively eliminating pro-tumor M2 macrophages seems, thus far, unique to iNKT cells. Interestingly, however, a mouse transgenic model of colon adenocarcinoma represents an exception to this general mechanism. Here, iNKT cells support pre-malignant progression by suppressing TH1 responses, and promoting suppressive Treg and M2-polarization of TAMs, leading to increased intestinal adenomatous polyps formation (132). The dichotomous iNKT cell response in the two mouse tumor models may be related to changes undergoing in the different TMEs. In both healthy prostate and intestine tissues, the NKT1 and NKT17 are mostly represented. However, as tumor progresses, iNKT cells infiltrating intestinal polyps start to produce IL-10, while those in the prostate cancer setting remained TH1-oriented.

On the basis of the described evidences, it is tempting to speculate that tumor-infiltrating iNKT cells lead the immune reprogramming of the local TME by acting primarily on the MPS. This remodeling activity in the tumor context appears critically determined by the specific effector profile exhibited by iNKT cells in the target tissue in physiological conditions, before the development of the malignancy. It will be important to investigate such relationship in different cancers, particularly human ones, given also the interest to define possible different tissue resident iNKT cell subsets, as well as to harness these cells for cancer immunotherapy that exploits their unique potential to reprogram the tumor microenvironment.

#### CHALLENGES AND FUTURE DIRECTIONS

Increasing evidence underscore the relevance of the iNKT cell/mononuclear phagocyte crosstalk in many different tissues, which may contribute to the induction, or the resolution, of tissue damage depending on the local effector phenotype exhibited by the two cell types interacting in the specific tissue. To this respect, iNKT cells are widely located in, non-lymphoid tissues in homeostatic conditions, at least in mice, which include (but are not limited to) the central nervous system (133), kidney (134, 135), eye (136), placenta (137), pancreas (138), and prostate (26). In all these sites, iNKT cells have the possibility of interacting, or have been suggested to interact, with tissue-resident MPS cells. However, the result of this crosstalk has not yet fully elucidated. In some cases, iNKT cells and resident mononuclear phagocytic cells show complementary functions. During acute kidney injury, iNKT cells alleviate the induced damages in different models (135, 139–141). Interestingly, these pathological conditions are highly dependent on renal macrophages, that switch between M1 and M2 phenotype during the acute or the tissue-repair phase (142), or on immature DCs (135), suggesting a link with NKT1 or NKT2 cells. In the eye, iNKT cells contribute to the natural tolerance occurring at this site (136) by cross-talking with T cells, neutrophils and macrophages (143). During reproduction, iNKT cells are present in the placenta, the interface between the mother and the fetus (144), and play role in orchestrating the immune response during infections occurring during pregnancy (145). Considering that iNKT cells consistently infiltrate the placenta also in healthy pregnancies (137), it is reasonable to hypothesize that their role is not limited to pathological conditions but they constantly support the reproduction process, for instance maintaining the status of tolerance induced by IL-10 producing macrophages (146). In the pancreas, iNKT cells promote an innate response against LCMV, by enhancing the local recruitment of pDCs and stimulating their production of anti-viral type I IFNs via OX40-OX40L interaction (138). However, CD1d expression by pDCs is not required for this interaction, suggesting a different, yet undefined, mechanism from those described in other tissues. Given the long standing implication of iNKT cells in the control of Type 1 Diabetes, it would be interesting to assess whether and by what mechanisms these cells may modulate MPS cells in the pancreas (147).

Some of the signals controlling the meeting between the two cell types have been defined, however this remains an open area to explore. Mouse and human iNKT cells express chemokine receptor pattern typical of trafficking toward inflammation sites (148–151) which overlaps at least in part with that of monocytes, supporting a cooperative engagement at inflammatory sites. Indeed, it has been shown that following B. burgdorferi infection Kupffer cells induce CXCR3-dependent clustering of iNKT cells (71), while CXCR6 drives homeostatic iNKT localization in liver sinusoids (69). A final big gap in knowledge concerns details on the presence of different iNKT cell subsets in different human tissues and their possible interaction with MPS cells. Correlative studies suggest undergoing crosstalk between the two cell types also in human tissues, although direct evidence is substantially lacking. A more precise definition of these mechanisms, focusing in particular on the human

#### REFERENCES


system in physiology and pathology, should drive future studies.

#### CONCLUDING REMARKS

The crosstalk between iNKT cells and cells of the MPS has a critical role in both physiological and pathological conditions. The outcome of this interaction is highly dependent on the tissue where it occurs and can be either beneficial or detrimental for the host. Harnessing this crosstalk has potential therapeutic relevance in different pathologies, from cancer to infections, chronic inflammatory diseases or metabolic disorders, as well as to improve vaccine formulation.

#### AUTHOR CONTRIBUTIONS

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

#### ACKNOWLEDGMENTS

We thank the reviewers for their insightful comments and helpful suggestions. This study is supported by a personal fellowship from the Italian Association for Cancer Research (AIRC) to FC (2015-18316), and grants AIRC IG-20081 to GC and LYRA Fondazione Regionale per la Ricerca Biomedica-FRRB to PD.


transcription factor Foxp3 and suppressive function in mice with colitis. Nat Immunol. (2009) 10:1178–84. doi: 10.1038/ni.1791


**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 Cortesi, Delfanti, Casorati and Dellabona. 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.

# iNKT Cells Suppress Pathogenic NK1.1+CD8<sup>+</sup> T Cells in DSS-Induced Colitis

Sung Won Lee1†, Hyun Jung Park 1†, Jae Hee Cheon2†, Lan Wu<sup>3</sup> , Luc Van Kaer <sup>3</sup> \* and Seokmann Hong<sup>1</sup> \*

*<sup>1</sup> Department of Integrative Bioscience and Biotechnology, Institute of Anticancer Medicine Development, Sejong University, Seoul, South Korea, <sup>2</sup> Department of Internal Medicine and Institute of Gastroenterology, Yonsei University College of Medicine, Seoul, South Korea, <sup>3</sup> Department of Pathology, Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN, United States*

#### Edited by:

*Yun-Cai Liu, Tsinghua University, China*

#### Reviewed by:

*Hiroshi Watarai, University of Tokyo, Japan Koji Yasutomo, Tokushima University, Japan*

#### \*Correspondence:

*Luc Van Kaer luc.van.kaer@vanderbilt.edu Seokmann Hong shong@sejong.ac.kr*

*†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: *14 June 2018* Accepted: *03 September 2018* Published: *02 October 2018*

#### Citation:

*Lee SW, Park HJ, Cheon JH, Wu L, Van Kaer L and Hong S (2018) iNKT Cells Suppress Pathogenic NK1.1*+*CD8*<sup>+</sup> *T Cells in DSS-Induced Colitis. Front. Immunol. 9:2168. doi: 10.3389/fimmu.2018.02168* T cells producing IFNγ play a pathogenic role in the development of inflammatory bowel disease (IBD). To investigate the functions of CD1d-dependent invariant natural killer T (iNKT) cells in experimental colitis induced in Yeti mice with dysregulated expression of IFNγ, we generated iNKT cell-deficient Yeti/CD1d KO mice and compared colitis among WT, CD1d KO, Yeti, and Yeti/CD1d KO mice following DSS treatment. We found that deficiency of iNKT cells exacerbated colitis and disease pathogenesis was mainly mediated by NK1.1+CD8<sup>+</sup> T cells. Furthermore, the protective effects of iNKT cells correlated with up-regulation of regulatory T cells. Taken together, our results have demonstrated that CD1d-dependent iNKT cells and CD1d-independent NK1.1+CD8<sup>+</sup> T cells reciprocally regulate the development of intestinal inflammatory responses mediated by IFNγ-dysregulation. These findings also identify NK1.1+CD8<sup>+</sup> T cells as novel target cells for the development of therapeutics for human IBD.

Keywords: CD1d-dependent NKT cells, NK1.1+CD8<sup>+</sup> T cells, Treg cells, IFNγ, DSS-induced colitis

#### INTRODUCTION

Inflammatory bowel disease (IBD) is a chronic inflammatory disorder of the gastrointestinal tract that is caused by dysregulated immune responses to host intestinal bacterial flora (1). Human IBD can be divided into two major types, Crohn's disease and ulcerative colitis (UC), which share symptoms and treatment options but differ in their pathogenic features and expression profiles of inflammatory mediators. Crohn's disease is characterized by excessive T helper 1 (Th1) responses and involves pathogenic lesions in any part of the gastrointestinal tract, whereas UC is characterized by Th2-dominant conditions that result in lesions restricted to the colon and rectum (1). Moreover, it has been reported that increased Th17 immune responses are also associated with the progression of both Crohn's disease and UC (2). In mice, IBD can be acutely induced by oral administration of dextran sulfate sodium (DSS) and this colitis is characterized by elevated Th1 and Th17 cytokine responses.

The development of colitis involves a variety of immune cells, including dendritic cells (DCs), macrophages, monocytes, neutrophils, conventional αβ T cells, γδ T cells, natural killer (NK) cells, natural killer T (NKT) cells, and innate lymphoid cells (ILC) that have been suggested to play either protective or pathogenic roles (3–5). In particular, NKT cells, defined by expression of both NK

cell markers (NK1.1 or DX5) and T cell markers (TCRβ or CD3) and restriction by the MHC class I-related protein CD1d, display several subsets with distinct cytokine profiles in gastrointestinal lymphoid tissues, including Peyer's patches (PP), lamina propria (LP), and mesenteric lymph nodes (MLN) of the small and large intestine (6). A role of CD1d-restricted type I NKT cells, also called invariant NKT (iNKT) cells because of the expression of an invariant TCR α chain (Vα14-Jα18 in mice and Vα24- Jα18 in humans), in regulating the development of IBD is well-established. For example, iNKT cells secreting IL13 play pathogenic roles in the oxazolone-induced UC mouse model (7). In contrast, adoptive transfer of IL9-producing iNKT cells (8) or administration of iNKT cell ligands (9, 10) exerted protective effects against DSS-induced colitis. In addition, roles for CD1drestricted type II and DX5<sup>+</sup> NKT cells in preventing the pathogenesis of IBD have been reported (11, 12). Additionally, CD1d-independent NK1.1<sup>+</sup> T cells have been identified in the intestinal tissues of CD1d knockout (KO) mice (13, 14), but their potential contribution to the pathogenesis of colitis has not been explored.

Previously, Yeti mice were generated to monitor IFNγexpressing immune cells in vivo by targeting an IRES/yellow fluorescent protein (YFP) reporter cassette downstream of the endogenous ifng gene (15). However, these Yeti mice have recently been reported to display autoinflammatory syndromes mediated by chronically elevated levels of IFNγ due to enhanced stability of IFNγ mRNA transcripts by using a polyA bovine growth hormone sequence (16). Thus, Yeti mice can be used to evaluate the role of IFNγ in chronic inflammatory conditions such as IBD.

Here, we have investigated the role of iNKT cells in colitis induced by DSS in Yeti mice with dysregulated IFNγmediated intestinal inflammation. We found that CD1ddeficiency exacerbated intestinal inflammation in these animals. Moreover, we found that disease in these animals was predominantly mediated by NK1.1+CD8<sup>+</sup> T cells. Furthermore, we found that disease suppression mediated by iNKT cells was linked with the expansion of Foxp3<sup>+</sup> regulatory T (Treg) cells.

## MATERIALS AND METHODS

#### Mice

Wild-type (WT) C57BL/6 (B6) mice were purchased from Jung Ang Lab Animal Inc. (Seoul, Korea). IFNγ/YFP (Yeti) cytokine reporter mice were kindly provided by Dr. R. Locksley (University of California at San Francisco, CA, USA). CD1d KO mice were provided by Dr. A. Bendelac (University of Chicago, IL, USA). Jα18 KO mice were provided by Dr. M. Taniguchi (RIKEN, Yokohama, Japan). Yeti mice were further crossed with either CD1d KO or Jα18 KO mice to obtain Yeti/CD1d KO and Yeti/Jα18 KO mice, respectively. All mice in this study were on a B6 genetic background, were maintained at Sejong University, and were used for experiments at 6–12 weeks of age. They were maintained on a 12-h light/12-h dark cycle in a temperature-controlled barrier facility with free access to food and water. Mice were fed a γirradiated sterile diet and provided with autoclaved tap water. Age- and sex-matched mice were used for all experiments. The animal experiments were approved by the Institutional Animal Care and Use Committee at Sejong University (SJ-20160704).

#### Induction of Colonic Inflammation

Mice were provided with 1.5% (w/v) DSS in the drinking water for 5 days. Subsequently, groups of mice were given normal control water for 5 days until sacrifice for experiments. To evaluate the clinical symptoms of DSS-induced colitis, the mice were monitored for a change in the percentage of body weight (0, none; 1, 1–10%; 2, 11–20%; 3, >20%), stool consistency (0, normal; 1, loose stool; 2, diarrhea), and bleeding (0, normal; 1, hemoccult positive; 2, gross bleeding) on a daily basis during colitis induction for 10 days. The body weight was expressed as a percentage of weight change for each individual mouse and was calculated relative to the starting body weight on day 0. These data were used to calculate a disease activity index (DAI).

# Cell Culture and Cell Enrichment by Magnetically Activated Cell Sorting (MACS)

A single-cell suspension of splenocytes was prepared and resuspended in RPMI complete medium consisting of RPMI 1640 (Gibco BRL, USA) medium supplemented with 10% FBS, 10 mM HEPES, 2 mM L-glutamine, 100 units/mL penicillinstreptomycin, and 5 mM 2-mercaptoethanol. Naive CD4<sup>+</sup> T cells from Jα18 KO B6 mice were enriched with the CD4+CD62L<sup>+</sup> T cell isolation kit II (Miltenyi Biotech, Bergisch Gladbach, Germany), following the manufacturer's instructions. The naive CD4<sup>+</sup> T cells were >94% pure among all MACS-purified populations. iNKT cells were enriched using NK1.1<sup>+</sup> iNKT cell isolation kit (Miltenyi Biotech) following the manufacturer's instructions. The NKT cell population was >89% pure among all MACS-purified populations. CD8<sup>+</sup> T cells that include NK1.1+CD8<sup>+</sup> T cells but lack CD1d-dependent NKT cells were enriched from MLN cells isolated from Yeti/CD1d KO mice by negative selection of CD11c<sup>+</sup> cells using anti-CD11c MACS and LD column, followed by positive selection with the CD8<sup>+</sup> T cell MACS system. NK1.1−CD8<sup>+</sup> T cells were enriched from MLN cells isolated from Yeti/CD1d KO mice by first removing NK1.1<sup>+</sup> cells and CD11c<sup>+</sup> cells using anti-CD11c MACS and anti-PE MACS after staining with PE-conjugated anti-NK1.1 (clone PK-136) mAb and LD column, followed by positive selection with the CD8<sup>+</sup> T cell MACS system. Cell populations included >95% CD8<sup>+</sup> cells among all MACSpurified populations. IL15-cultured NK1.1+CD8<sup>+</sup> T cells from CD1d KO MLN were separated using Lympholyte-M (Cedar Lane Laboratories Ltd., Hornby, Ontario, Canada) by density gradient centrifugation and further positively selected for the NK1.1<sup>+</sup> population using anti-PE MACS after staining with PEconjugated anti-NK1.1 (clone PK-136) mAb. The NK1.1+CD8<sup>+</sup> T cell population was >91% pure among all MACS-purified populations.

# In vitro CD4<sup>+</sup> and CD8<sup>+</sup> T Cell Differentiation

Recombinant murine IL15 and human TGFβ were purchased from R&D systems (Minneapolis, MN, USA) and recombinant murine IL2 was purchased from Peprotech (Hamburg, Germany). For in vitro stimulation, rIL15, rTGFβ, and rIL2 were used at a concentration of 10 ng/ml. Lipopolysaccharide (LPS) derived from E. coli (serotype 0111:B4) was purchased from Sigma-Aldrich (St. Louis, MO, USA). For CD4<sup>+</sup> T cell differentiation, anti-IFNγ mAbs (5µg/ml; clone XMG1.2, BD Biosciences) were added at the concentrations indicated in figure legends. Isolated naive CD4+CD62L<sup>+</sup> T cells (1 × 10<sup>5</sup> cells/well) were cultured with 96-well plate-bound anti-CD3ε (10µg/ml) + anti-CD28 (1µg/ml) mAbs in the presence of rTGFβ (10 ng/ml) and rIL2 (10 ng/ml) for 5 days. For CD8<sup>+</sup> T cell differentiation, MLN NK1.1+CD8<sup>+</sup> T cell-depleted CD8<sup>+</sup> T cells (5 × 10<sup>5</sup> cells/ml) isolated from either CD1d KO or Yeti/CD1d KO mice were cultured with rIL15 (10 ng/ml) in 24-well plates for 5 or 10 days.

### Flow Cytometry

The following monoclonal antibodies (mAbs) were obtained from BD Biosciences (San Jose, USA): phycoerythrin (PE)-, or allophycocyanin (APC)-conjugated anti-NK1.1 (clone PK-136); PE-Cy7-, or APC-conjugated anti-CD4 (clone RM4-5); PE-Cy7 conjugated anti-TCRβ (clone H57-597); PE-Cy7-conjugated anti-CD8α (clone 53-6.7); PE-Cy7-, or APC-conjugated anti-CD3ε (clone 145-2C11); PE-Cy7-conjugated anti-CD11b (clone M1/70); APC-conjugated anti-CD25 (clone PC61); APC-conjugated anti-CD44 (clone IM7); APC-conjugated anti-CD314 (NKG2D) (clone CX5); PE-conjugated anti-IL12p40 (clone C15.6); PE-conjugated anti-TNFα (clone XP6-XT22); and PE-conjugated anti-IgG1 (κ isotype control).

The following mAbs from eBioscience were used: PEconjugated anti-Ly49A (clone A1); PE-conjugated anti-FasL (clone MFL3); APC-conjugated anti-F4/80 (clone BM8); PEconjugated anti-Perforin (clone eBioOMAK-D); PE-conjugated anti-IFNγ (clone XMG1.2); PE-conjugated anti-IL17A (clone eBio17B7); PE-conjugated anti-Eomes (clone Dan11mag); and PE-conjugated anti-Foxp3 (clone NRRF-30). The following mAb from R&D Systems was used: Biotin-conjugated anti-IL15. Flow cytometric data were acquired with a FACSCalibur system (Becton Dickinson, USA) and analyzed with FlowJo software (Tree Star, USA).

For surface antibody staining, cells were harvested and washed twice with cold 0.5% BSA-containing PBS (FACS buffer). For blocking non-specific binding to Fc receptors, the cells were incubated with anti-CD16/CD32 mAbs on ice for 10 min and subsequently stained with fluorescence-labeled mAbs. Flow cytometric data were acquired using a FACSCalibur flow cytometer (Becton Dickson, San Jose, CA, USA) and analyzed using FlowJo software (Tree Star Inc., Ashland, OR, USA).

#### Intracellular Cytokine Staining

For intracellular staining, splenocytes were incubated with brefeldin A, an intracellular protein transport inhibitor (10µg/ml), in RPMI medium for 2 h at 37◦C. The cells were stained for cell surface markers, fixed with 1% paraformaldehyde, washed once with cold FACS buffer, and permeabilized with 0.5% saponin. The permeabilized cells were then stained for an additional 30 min at room temperature with the indicated mAbs (PE-conjugated anti-IL12p40, anti-IFNγ, anti-IL17A, and antiperforin; PE-conjugated isotype control rat IgG mAbs). Fixation and permeabilization were performed using a Foxp3 staining kit (eBioscience) with the indicated mAbs (FITC-conjugated anti-Foxp3; PE-conjugated anti-Foxp3; FITC- or PE-conjugated isotype control rat IgG mAbs). More than 5,000 cells per sample were acquired using a FACSCalibur, and the data were analyzed using the FlowJo software package (Tree Star, Ashland, OR, USA).

### Isolation of Colon MLN, IEL, and LP Leukocytes

The MLN were aseptically removed, and single-cell suspensions of the MLN were obtained by homogenization and passing through a 70µm nylon cell strainer. The large intestines were removed and flushed with 20 ml of cold CMF solution (Ca2+- Mg2+-free PBS containing 10 mM HEPES, 25 mM sodium bicarbonate, and 2% FBS). After exclusion of PP, fat, and mucus, the large intestine was cut longitudinally into 5 mm pieces and the pieces of tissue were washed twice with CMF solution. The tissues were transferred into 15 ml pre-warmed EDTA/DTT/FBS/CMF solution (CMF containing 1 mM EDTA, 1 mM DTT, 10% FBS, and 100 units/mL penicillin-streptomycin) and stirred in 37◦C shaking/orbital incubator for 30 min to obtain intraepithelial lymphocytes (IEL). The cell-containing suspension was passed through a 70µm nylon cell strainer (BD Falcon), and put on ice. IELs were purified from the interface of a 40/70% Percoll (GE Healthcare) gradient after centrifugation for 20 min at 2,400 rpm at RT. To isolate the lamina propria lymphocytes (LPLs), the remaining intestinal tissues were cut into small pieces using a scalpel and transferred to conical tubes. Tissues were resuspended in 20 ml of complete RPMI containing 2.5 mg/ml collagenase type IV (Sigma, St. Louis, MO, USA) and 1 mg/ml DNase I (Promega, Madison, USA) and shaken at 200 rpm for 40 min at 37◦C. At the end of the incubation, the digested tissues were dissociated into single-cell suspensions using gentle MACS Dissociator (Miltenyi, Germany) in combination with C Tubes. The cell-containing suspension was passed through a 70µm nylon cell strainer (BD Falcon), and put on ice. LPLs were purified from the interface of a 40/70% Percoll (GE Healthcare) gradient after centrifugation for 20 min at 2,400 rpm at RT.

#### Histology

Distal colonic sections were fixed in 4% paraformaldehyde, embedded in paraffin, and cut into 6µm sections using a microtome (RM 2235, Leica, Germany). The sections were then stained with H&E for the analysis of histological changes. The histological score of each individual mouse was measured as follows: epithelial damage (E), 0 = none; 1 = minimal loss of goblet cells; 2 = extensive loss of goblet cells; 3 = minimal loss of crypts and extensive loss of goblet cells; 4 = extensive loss of crypts; and infiltration (I), 0 = no infiltrate; 1 = infiltrate around

the crypt basis; 2 = infiltrate reaching the muscularis mucosa; 3 = extensive infiltration reaching the muscularis mucosa and thickening of the mucosa with abundant edema; 4 = infiltration of the submucosa. The total histological score was calculated as E+I.

# Cytotoxicity Assay

The flow cytometric CFSE/7-AAD cytotoxicity assay was performed as previously described (17) with minor modifications. NK1.1−CD8<sup>+</sup> T cells and NK1.1+CD8<sup>+</sup> T cells were isolated as described above and suspended in complete RPMI medium. B16 melanoma cells (3 × 10<sup>6</sup> ) were labeled with 500 nM CFSE in Hanks' Balanced Salt Solution for 10 min at 37◦C in a volume of 2 ml. The cells were washed twice in RPMI medium and used immediately. The CFSE-labeled target cells (20,000 cells) were incubated with either NK1.1−CD8<sup>+</sup> T cells or NK1.1+CD8<sup>+</sup> T cells at different effector (E): target (T) ratios (0:1, 3:1, 9:1, and 27:1). After 10 h of incubation, cells were stained with 0.25µg/ml of 7-AAD and were incubated for 10 min at 37◦C in a CO<sup>2</sup> incubator. Cells were washed twice with 1× PBS containing 1% FBS (FACS buffer) and resuspended in FACS buffer. Cytotoxicity was assessed by flow cytometry.

# Statistical Analysis

Statistical significance was determined using Excel (Microsoft, USA). Student's t-test was performed for the comparison of two groups. <sup>∗</sup>P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001 were considered to be significant in the Student's t-test. Two-way ANOVA analysis was carried out using the VassarStats (http:// vassarstats.net/anova2u.html). #P < 0.05, ##P < 0.01, and ###P < 0.001 were considered to be significant in the two-way ANOVA.

### RESULTS

#### Yeti Mice Have Increased Numbers of CD1d-Independent NK1.1+CD8<sup>+</sup> T Cells

Since it has been reported that IFNγ KO mice are protected from DSS-induced colitis (18), we examined whether dysregulated IFNγ expression in heterozygous Yeti mice, in which abnormal IFN<sup>γ</sup> secretion occurs due to modification of the 3′ -untranslated region (UTR) of the IFNγ gene (16), influences colitis severity. We confirmed that heterozygous Yeti mice display splenomegaly and elevated levels of IL12-producing DCs and Th1 cells compared with WT mice (**Figures 1A,B**).

While studying NKT cells in Yeti mice, we noticed that the proportion and absolute number of NK1.1+CD8<sup>+</sup> T cells among total splenocytes were increased five- to eight-fold in Yeti mice compared with WT mice. Moreover, a similar increase in NK1.1+CD8<sup>+</sup> T cells in Yeti mice was observed in NKT celldeficient Yeti/CD1d KO mice, consistent with the notion that these cells are not CD1d-restricted (**Figures 1C,D**). Collectively, these results suggest that the excessive inflammation in Yeti mice might be associated with an increase in the numbers of NK1.1+CD8<sup>+</sup> T cells.

# Yeti Mice Accelerate Intestinal Inflammation in the Absence of iNKT Cells

Since iNKT cells can provide protective effects against DSSinduced colitis (8, 9), we decided to examine whether these cells play a role in regulating intestinal inflammation in Yeti mice. To address this possibility, we generated iNKT celldeficient Yeti/CD1d KO mice by crossing Yeti mice with CD1d KO mice. We compared the colitic symptoms among WT, CD1d KO, Yeti, and Yeti/CD1d KO mice following 1.5% DSS treatment. Yeti/CD1d KO mice showed more severe weight loss, diarrhea, and bleeding in feces, resulting in a steep increase in the DAI score, compared with the different groups of control mice (**Figure 2A**). Furthermore, the colon length in DSS-treated Yeti/CD1d KO mice was remarkably shortened (**Figure 2B**) and displayed increased signs of colonic inflammation, with loss of epithelial crypts, edema, and infiltration of inflammatory cells (**Figure 2C**). In addition, we found that Th1 and Th17 differentiation of CD4<sup>+</sup> T cells in the spleen, MLN, and LP from Yeti and Yeti/CD1d KO mice was significantly increased as compared with WT and CD1d KO mice after DSS administration. The frequencies of Th1 and Th17 cells in Yeti/CD1d KO mice were two- to three-fold greater in the MLN and LP, but not in the spleen, than those from Yeti mice, indicating that the protective role of iNKT cells in colitis was largely restricted to the MLN and LP (**Figure 2D**). On the other hand, colitis in CD1d KO mice was only slightly increased compared with WT mice, suggesting that lack of iNKT cells by itself is not sufficient to induce exacerbated colitis.

Since CD1d KO mice lack iNKT (Type I) cells as well as type II NKT cells, we generated Yeti/Jα18 KO mice by crossing Yeti mice with Jα18 KO mice that selectively lack iNKT cells. Upon administration of DSS, Yeti/Jα18 KO mice showed significantly higher disease activity than the control mice (**Supplementary Figures 1A,B**). Taken together, these results indicate that CD1d-dependent iNKT cells play a major role in controlling DSS-induced colitis in Yeti mice.

IL10 KO and IL2 KO mice spontaneously develop colitis when raised under conventional conditions by dysregulated immune responses against commensal bacteria (19, 20). To investigate whether WT, CD1d KO, Yeti, and Yeti/CD1d KO mice develop intestinal inflammation spontaneously when raised under conventional conditions, we measured body weights once a week for 12 weeks. We observed that Yeti/CD1d KO mice gained body weight more slowly compared with the other groups of mice (**Figure 2E**). Furthermore, unexpectedly, the incidence of rectal prolapse was significantly increased in Yeti/CD1d KO mice (**Figure 2F**). This was confirmed by histological examination of H&E-stained colon sections (**Figure 2G**). Taken together, these results suggest that colitic pathogenesis in Yeti/CD1d KO mice is mainly attributed to both deficiency of iNKT cells and increase of NK1.1+CD8<sup>+</sup> T cells.

# NK1.1+CD8<sup>+</sup> T Cells With Effector/Memory Phenotypes Strongly Correlate With Severe Intestinal Inflammation in Yeti Mice

Since our results showed that the numbers of NK1.1+CD8<sup>+</sup> T cells were increased in Yeti mice at steady state, we investigated

FIGURE 1 | Altered NKT cell subsets in Yeti mice. (A) Left, a representative picture of the spleens from 8-week-old WT B6 and heterozygous Yeti B6 mice. Middle and Right, Spleen weight and splenocyte number in Yeti B6 and Yeti Balb/c mice, as compared with WT B6 and Balb/c mice. (B) Intracellular IL12 production by isolated DCs (CD11c+) from WT B6 and Yeti B6 mice was assessed by flow cytometry. Intracellular IFNγ production was assessed in splenic CD4<sup>+</sup> T cells from WT B6 or Yeti B6 mice by flow cytometry. (C) The percentage of NK1.1+TCRβ <sup>+</sup> cells among splenocytes and the percentage of either CD4<sup>+</sup> or CD8α <sup>+</sup> populations among NK1.1<sup>+</sup> T cells from 8-week-old WT and Yeti mice are plotted. The proportion and absolute cell numbers of CD4+, CD8+, and DN NK1.1<sup>+</sup> T cells were assessed in WT and Yeti mice at the age of 8 weeks. (D) The percentage of NK1.1<sup>+</sup> populations among CD8α <sup>+</sup> T cells from 8-week-old WT, Yeti, CD1d KO, and Yeti/CD1d KO mice are plotted. The mean values ± SD (*n* = 4 per group in the experiment; Student's *t*-test; \*\**P* < 0.01, \*\*\**P* < 0.001) are shown. Two-way ANOVA (Yeti × iNKT) showed an interaction between these two factors (ns, not significant).

whether NK1.1+CD8<sup>+</sup> T cells are closely related with the intestinal inflammatory processes in Yeti mice. In DSS-induced colitis, the number of NK1.1+CD8<sup>+</sup> T cells in the spleen and MLN from Yeti mice was expanded approximately 6-fold compared with WT mice, and the number of NK1.1+CD8<sup>+</sup> T cells from Yeti/CD1d KO mice was increased eight-fold compared with CD1d KO mice (**Figure 3A**). Moreover, when colitis was spontaneously induced, the number of NK1.1+CD8<sup>+</sup> T cells in the spleen and MLN from Yeti/CD1d KO mice with a prolapse were three- to four-fold higher compared to those from Yeti/CD1d KO mice without prolapse (**Figure 3B**), which strongly indicates that NK1.1+CD8<sup>+</sup> T cells are pathogenic effector cells. Moreover, there was a strong positive correlation between NK1.1+CD8<sup>+</sup> T cell number and the histological score of colon tissue samples in both the spleen (R <sup>2</sup> = 0.91) and MLN (R <sup>2</sup> = 0.94) (**Figure 3C**). Taken together, these findings suggest that NK1.1+CD8<sup>+</sup> T cells are pathogenic cells contributing to severe and spontaneous colitis in Yeti/CD1d KO mice.

In previous reports, NK1.1-expressing CD8<sup>+</sup> T cells were shown to display significant cytotoxicity against tumor cells and to possess a memory phenotype with a pro-inflammatory cytokine production profile (21, 22). Thus, we examined the expression of memory markers (CD44 and Eomes), NK cell receptors (NKG2D and Ly49a), cytolytic molecules (FasL and Perforin), and pro-inflammatory cytokines (IFNγ and TNFα) in MLN-derived NK1.1+CD8<sup>+</sup> T cells from Yeti/CD1d KO mice during DSS-induced colitis. We found that NK1.1+CD8<sup>+</sup> T cells from DSS-treated mice, unlike unstimulated NK1.1−CD8<sup>+</sup>

FIGURE 2 | Lack of iNKT cells accelerates intestinal inflammation in Yeti mice. Daily body weight changes, DAI score (A) and colon length (B) of WT, Yeti, CD1d KO, and Yeti/CD1d KO mice were evaluated after 1.5% DSS treatment. Data are representative of three independent experiments with similar results. (C) On day 10, distal colons from each group were sectioned and stained with H&E. (D) Intracellular IFNγ and IL17 production were assessed in splenic, MLN, and LP CD4<sup>+</sup> T cells from these mice by flow cytometry on day 10. The mean values ± SD (n=5 per group in the experiment; Student's *t*-test; \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001) are shown. Two-way ANOVA (Yeti × iNKT and genotype × tissue) showed an interaction between these two factors (#*P* < 0.05, ##*P* < 0.01, ###*P* < 0.001). Daily body weight changes (E) and prolapse rate (F) of WT, Yeti, CD1d KO, and Yeti/CD1d KO mice housed for 12 weeks under conventional conditions. (*n* = 7 for WT, Yeti, and CD1d KO mice; *n* = 10 for Yeti/CD1d KO mice; Student's *t*-test; \**P* < 0.05, \*\**P* < 0.01). (G) Left, distal colons from these mice were sectioned and stained with H&E at week 12. Right, histologic damages were scored from H&E-stained sections at week 12. (*n* = 5 for WT, Yeti, and CD1d KO mice; *n* = 5 for Yeti/CD1d KO mice (no signs); *n* = 3 for Yeti/CD1d KO (prolapse); Student's *t*-test; \*\**P* < 0.01, \*\*\**P* < 0.001).

FIGURE 3 | mice were assessed by flow cytometry at day 10. The mean values ± SD (*n* = 5 per group in the experiment; Student's *t*-test; \*\**P* < 0.01, \*\*\**P* < 0.001) are shown. Two-way ANOVA (Yeti × iNKT) showed an interaction between these two factors (#*P* < 0.05, ##*P* < 0.01). (B,C) The spleen and MLN were isolated from WT, Yeti, CD1d KO, and Yeti/CD1d KO mice at 12 weeks after initiation of conventional housing conditions. (B) Upper left, the percentage of NK1.1+CD3<sup>+</sup> cells among splenocytes and MLN cells of these mice was determined at week 12. Lower left, the frequency of the CD8α <sup>+</sup> population among NK1.1+CD3<sup>+</sup> cells from the spleen and MLN of these mice was determined at week 12. Right, the absolute numbers of NK1.1+CD8<sup>+</sup> T cells in the spleen and MLN were assessed by flow cytometry at week 12. (*n* = 5 for WT, Yeti, and CD1d KO mice; *n* = 5 for Yeti/CD1d KO mice (no signs); *n* = 3 for Yeti/CD1d KO (prolapse); Student's *t*-test; \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001). (C) Scatter graphs and linear regression analysis of the relationship between the frequency of NK1.1+CD8<sup>+</sup> T cells among total splenocytes or MLN cells and histological score of colon tissue sections in Yeti/CD1d KO mice. The Pearson's correlation coefficient (*R* 2 ) for each plot is indicated. (*n* = 5 for Yeti/CD1d KO mice (no signs); *n* = 3 for Yeti/CD1d KO (prolapse)). (D) The expression of CD44, Eomes, NKG2D, Ly49a, FasL, perforin, IFNγ, and TNFα among NK1.1−CD8<sup>+</sup> T cells, NK1.1+CD8<sup>+</sup> T cells, and NK cells of the MLN from 1.5% DSS-treated Yeti/CD1d KO mice was evaluated by flow cytometry on day 10. The mean values ± SD (*n* = 4 in A–C; *n* = 5 in D; per group in the experiment; Student's *t*-test; \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001) are shown.

T cells, exhibit an effector CD8<sup>+</sup> T cell phenotype with NK cell-like functions, suggesting that NK1.1+CD8<sup>+</sup> T cells are pathogenic cells causing severe intestinal inflammation in Yeti mice (**Figure 3D**).

# The NK1.1+CD8<sup>+</sup> T Cell Population Is Primarily Responsible for Induction of DSS Colitis in Yeti Mice

It has been reported that colitogenic CD8<sup>+</sup> T cells have an effector phenotype with increased expression of IFNγ and Granzyme B (23). Since the increased NK1.1<sup>+</sup> population among Yeti CD8<sup>+</sup> T cells produced high levels of pro-inflammatory cytokines and cytolytic molecules, we decided to examine whether the pathogenic capacity of Yeti CD8<sup>+</sup> T cells was mediated by the NK1.1<sup>+</sup> subpopulation. To test this possibility, we adoptively transferred MLN total CD8<sup>+</sup> T cells or MLN NK1.1−CD8<sup>+</sup> T cells from Yeti/CD1d KO into CD1d KO mice and treated these animals with 1.5% DSS (**Figure 4A**). Mice that received total CD8<sup>+</sup> T cells showed progressive symptoms of colitis compared with uninjected mice, whereas disease in mice that received NK1.1−CD8<sup>+</sup> T cells was very similar to uninjected mice (**Figures 4B,C**). Moreover, mice that received NK1.1−CD8<sup>+</sup> T cells displayed a significant decrease in the frequencies of Th1 and Th17 cells in the MLN and LP compared to mice that received total CD8<sup>+</sup> T cells including NK1.1+CD8<sup>+</sup> T cells (**Figure 4D**). Overall, these results provide strong evidence that NK1.1+CD8<sup>+</sup> T cells play a pivotal role in mediating the pathogenesis of DSS-induced colitis in Yeti mice.

## Lack of CD1d-Restricted iNKT Cells Increases Susceptibility to Colitis Induced by NK1.1+CD8<sup>+</sup> T Cells

It has been reported that the common gamma chain (γc) cytokine IL15 directly induces NK1.1 expression on sorted CD8<sup>+</sup> T cells (24). Consistent with this previous study, we found that NK1.1−CD8<sup>+</sup> T cells cultured with IL15 significantly expressed higher levels of NK1.1 compared to cells cultured without cytokines (**Figure 5A**). Furthermore, we found that IFNγ-YFP reporter knockin increases NK1.1 expression on NK1.1−CD8<sup>+</sup> T cells in response to IL15 stimulation (**Figure 5B**). We also found that IL15-induced NK1.1+CD8<sup>+</sup> T cells expressed high levels of cytolytic molecules such as NKG2D, perforin, and FasL, indicating that IL15-induced NK1.1+CD8<sup>+</sup> T cells display similar phenotypic profiles as Yeti NK1.1+CD8<sup>+</sup> T cells (**Figure 5C**). Furthermore, we demonstrated that IL15-induced NK1.1+CD8<sup>+</sup> T cells, unlike unstimulated NK1.1−CD8<sup>+</sup> T cells, are effector cells with cytotoxic function against B16 melanoma cells (**Figure 5D**).

Previously, iNKT cells and NK1.1+CD8<sup>+</sup> T cells have been implicated in the development of colitis (8, 14), but their interactions remain largely unknown. Importantly, adoptive transfer of IL15-induced NK1.1+CD8<sup>+</sup> T cells resulted in a significant increase in DSS-induced weight loss and DAI score in the absence of iNKT cells, whereas DSS-induced colitis exacerbated by injection of IL15-induced NK1.1+CD8<sup>+</sup> T cells was dramatically halted by the presence of iNKT cells (**Figure 5E**). Upon DSS-induced colitis, when compared with CD1d KO mice, WT mice that received NK1.1+CD8<sup>+</sup> T cells showed significantly decreased frequency of Th1 and Th17 cells in the LP (**Figure 5F**). Collectively, these data demonstrate that NK1.1+CD8<sup>+</sup> T cells are pathogenic in DSS colitis in the absence of iNKT cells.

# NK1.1+CD8<sup>+</sup> T Cells Mediate Inhibitory Effects on Treg Cell Differentiation in Intestinal Inflammation

The loss of colonic Treg cells induces excessive infiltration of inflammatory immune cells that results in colitis (25). A previous study demonstrated that IFNγ inhibits the generation of inducible Treg (iTreg) (26). In addition, activated NKT cells modulate Treg function in an IL2-dependent manner (27). Thus, we examined whether the lack of iNKT cells could affect the distribution of Treg cells in DSS-treated Yeti mice. Intriguingly, DSS-treated Yeti/CD1d KO mice contained fewer Treg cells in the MLN and LP, but not in the spleen when compared with WT, Yeti, and CD1d KO mice (**Figures 6A,B**). Taken together, these data suggest that the combined loss of iNKT cells and dysregulated IFNγ production can induce a decrease in protective Foxp3+CD25+CD4<sup>+</sup> T cells during DSS-induced colitis. As expected, IL15-induced NK1.1+CD8<sup>+</sup> T cells, unlike unstimulated NK1.1−CD8<sup>+</sup> T cells, produced high levels of IFNγ (**Figure 6C**). Moreover, consistent with a previous study (28), our results showed that NK1.1+CD8<sup>+</sup> T cell-derived IFNγ acts as a potent inhibitor of Treg cell differentiation (**Figure 6D**). Upon DSS-induced colitis, when compared with CD1d KO mice, WT mice that received NK1.1+CD8<sup>+</sup> T cells displayed significantly increased frequency of Treg cells in the LP (**Figure 6E**). Collectively, these data strongly suggest that

) from Yeti/CD1d KO mice were i.v. transferred to CD1d KO mice. Daily body weight changes, DAI score (B), and colon length (C) of each group were evaluated after 1.5% DSS treatment. (D) Intracellular IFNγ and IL17 production and the frequencies of Foxp3+CD25<sup>+</sup> cells were assessed in the MLN and LP CD4<sup>+</sup> T cells from these mice by flow cytometry on day 10. The mean values ± SD (*n* = 5 per group in the experiment; Student's *t*-test; \**P* < 0.05, \*\**P* < 0.01) are shown.

T cells from the MLN of CD1d KO mice were cultured with rIL15 for 5 or 10 days. The percentage of the NK1.1<sup>+</sup> population among all CD8<sup>+</sup> T cells was evaluated by flow cytometry at the indicated time points. (B) Purified NK1.1−CD8<sup>+</sup> T cells from the MLN of CD1d KO and Yeti/CD1d KO mice were cultured with rIL15 for 5 or 10 days. The percentage of the NK1.1<sup>+</sup> population among total CD8<sup>+</sup> T cells was evaluated by flow cytometry at the indicated time points. (C) The expression of NKG2D, perforin, and FasL among unstimulated NK1.1−CD8<sup>+</sup> T cells or IL15-stimulated NK1.1+CD8<sup>+</sup> T cells were assessed by flow cytometry on day 10 after cytokine stimulation. (D) Cytotoxicity of either unstimulated NK1.1−CD8<sup>+</sup> T cells or IL15-stimulated NK1.1+CD8<sup>+</sup> T cells from the CD1d KO MLN was evaluated using 7-AAD/CFSE assay. (E,F) Either NK1.1−CD8<sup>+</sup> T cells (1 × 10<sup>6</sup> ) or IL15-stimulated NK1.1+CD8<sup>+</sup> T cells (1 × 10<sup>6</sup> ) from the CD1d KO MLN were i.v. transferred to CD1d KO mice. (E) Daily body weight changes and DAI score of each group were evaluated 10 days after 1.5% DSS treatment. (F) Intracellular IFNγ and IL17 production in LP CD4<sup>+</sup> T cells from these mice were determined by flow cytometry on day 10. The mean values ± SD (*n* = 3 in D; *n* = 4 in A–C; *n* = 5 in E and F; per group in the experiment; Student's *t*-test; \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001) are shown. Two-way ANOVA (genotype × treatment) showed an interaction between these two factors (#*P* < 0.05).

FIGURE 6 | iNKT cells prevent the reduction in the Treg population induced by NK1.1+CD8<sup>+</sup> T cells. (A,B) The spleen, MLN, and LP were obtained from WT, Yeti, CD1d KO, and Yeti/CD1d KO mice at 10 days after 1.5% DSS treatment. The percentage of Foxp3+CD25<sup>+</sup> cells among CD4<sup>+</sup> T cells from the spleen, MLN, and LP of each group was evaluated by flow cytometry on day 10. Data are representative of three independent experiments with similar results. The mean values ± SD (*n* = 5 per group in the experiment; Student's *t*-test; \**P* < 0.05, \*\**P* < 0.01) are shown. Two-way ANOVA (Yeti × iNKT) showed an interaction between these two factors (###*P* < 0.001). (C) The expression of IFNγ among unstimulated NK1.1−CD8<sup>+</sup> T cells or IL15-stimulated NK1.1+CD8<sup>+</sup> T cells were assessed by flow cytometry on day 10 *(Continued)*

NK1.1+CD8<sup>+</sup> T cells become pathogenic by suppressing Treg differentiation whereas iNKT cells can control the effects of NK1.1+CD8<sup>+</sup> T cells on Treg cells during the development of colitis.

#### DISCUSSION

We have demonstrated here that the severity of low dose DSS (1.5%)-induced colitis is exacerbated in heterozygous Yeti mice on an NKT cell-deficient background, indicating that iNKT cells are essential for maintaining intestinal homeostasis in the context of IFNγ-mediated inflammation. We further showed that iNKT cells and Treg cells contribute to colitis protection, whereas expansion of CD1d-independent NK1.1+CD8<sup>+</sup> T cells contributes to colitis pathogenesis.

Previously, Lee et al. suggested that an IL4-dominant environment in the MLN induced by oral administration of α-GalCer suppresses Th1-driven Crohn's disease but promotes Th2-driven ulcerative colitis (29). Based on these previous findings and our results reported here, the MLN might be a critical site for iNKT2 cell-mediated protection against DSSinduced colitis in Yeti mice. In addition to protein antigens that can activate conventional T cells, commensal bacteria contain glycolipid Ags, which are capable of activating iNKT cells through TCR ligation by Ag/CD1d complexes, but not TLR triggering (30). Sphingomonas/Sphingobium species, which contain glycolipid Ags, activate intestinal iNKT cells to produce cytokines in a CD1d-dependent manner (30), whereas intestinal microbe Bacteroides fragilis-derived glycolipid Ags inhibit iNKT cell activation through competitive binding with CD1d, resulting in protection against oxazolone-induced colitis (31). Thus, it is tempting to speculate that reduced survival of Treg cells in a Th1 cytokine-dominant environment triggered in Yeti mice might be inhibited by iNKT cells in response to glycolipid Ags derived from intestinal commensal bacteria.

Next, we investigated the cell types that contribute to disease pathogenesis in DSS-induced colitis under IFNγ dysregulated conditions. Since prior reports that CD1d-independent NK1.1+CD8<sup>+</sup> T cells producing high levels of IFNγ and TNFα are involved in diverse immune responses such as tumor immune surveillance, allogeneic hematopoietic cell transplantation, and viral infection (21, 22, 32, 33), we considered the possibility that such cells might be involved. We found an expansion of NK1.1+CD8<sup>+</sup> T cells during the development of DSS and spontaneous colitis in Yeti mice, and adoptive transfer experiments provided direct evidence for a pathogenic role of these cells in colitis.

Based on the previous report that IFNγ and IL12 do not directly up-regulate NK1.1 expression on the surface of CD8<sup>+</sup> T cells (24), an increase in NK1.1-expressing CD8<sup>+</sup> T cells might be attributed to enhanced IL15 secretion by mononuclear phagocytes (34–36), which is consistent with our observations that IL15 production in both DCs and macrophages from Yeti mice was significantly increased in both the spleen and MLN compared to WT mice, indicating that DCs and macrophages might be critical sources of IL15 to increase NK1.1+CD8<sup>+</sup> T cells in Yeti mice (**Supplementary Figure 2**). In muscle damage mediated by idiopathic inflammatory myopathies, myoblast-derived IL15 induced the differentiation of naive CD8<sup>+</sup> T cells into highly activated and cytotoxic NKG2DhighCD8<sup>+</sup> T cells, which promoted myoblast damage through NKG2D-dependent lysis induced by MHC class I chain-related molecules (MICA and MICB) (37). Moreover, in celiac disease, NKG2DhighCD8<sup>+</sup> T cells induced by IL15 displayed cytotoxicity against intestinal epithelial cells expressing high MIC levels in a TCR-independent and NKG2D-mediated cytolysis pathway (38). Recently, it has been reported that in acute hepatitis A patients, NKG2DhighCD8<sup>+</sup> T cells induced by IL15 promote damage of liver tissue expressing NKG2D ligands such as MICA and MICB (39). Because stress induces increased expression of MICA and MICB on human intestinal epithelial cell lines (40), enhanced expression of murine NKG2D ligands on intestinal epithelial cells of DSS-treated Yeti/CD1d KO mice might trigger NK1.1+CD8<sup>+</sup> T cell activation, which in turn might exacerbate DSS-induced colitis. Moreover, Crohn's disease patients exhibit markedly increased expression of IL15 and IFNγ in the inflamed colon compared with healthy people (41), indicating that NK1.1+CD8<sup>+</sup> T cells involved in intestinal inflammation in Yeti mice might be physiologically relevant to the development of Crohn's disease. It will be of interest to further investigate what types of cells might represent the human counterpart for murine NK1.1+CD8<sup>+</sup> T cells.

It has been reported that repression of IFNγ mRNA decay leads to accumulation of self-reactive CD8<sup>+</sup> T cells, causing pancreas-specific damage (42). In addition, a recent study showed that IFNγ production by CD4<sup>+</sup> T cells from homozygous Yeti mice is regulated by lactate dehydrogenase A (LDHA) via a 3′ UTR-independent mechanism, whereas IFNγ production by NK cells was not affected by these pathways (43). Consistent with the previous report that IFNγ mRNA transcripts are maintained constitutively in NK receptorexpressing innate immune cells such as NK cells and NKT cells (44), we demonstrated that NK1.1+CD8<sup>+</sup> T cells expressing NK receptors constitutively produced large amounts of IFNγ. Thus, our results suggest that NK1.1+CD8<sup>+</sup> T cells might be one of the main cellular sources of increased IFNγ in Yeti mice, in which the stability of IFNγ mRNA transcripts is enhanced.

A previous report revealed that IL15-induced NK1.1+CD8<sup>+</sup> T cells display downregulated levels of TGFβ receptor, implying that NK1.1+CD8<sup>+</sup> T cells are inherently resistant to immune suppression mediated by Treg-derived TGFβ (45). In addition, our data showed that NK1.1+CD8<sup>+</sup> T cells could strongly attenuate Treg development via IFNγ signaling. These findings suggest that IFNγ-producing NK1.1+CD8<sup>+</sup> T cells cannot respond to Treg cells and instead might inhibit Treg functions in the intestine.

In conclusion, the overall severity of DSS-induced colitis is determined by the extent of imbalance between protective and pathogenic cells and factors. Since our results clearly showed that iNKT cells are required for controlling NK1.1+CD8<sup>+</sup> T cellmediated pathogenesis during DSS-induced colitis, glycolipid antigens might be employed for designing more effective and safer therapeutics for IBD. Since iNKT cells are known to have subsets (iNKT1, iNKT2, and iNKT17), in future studies, it will be of interest to determine which subsets of iNKT cells contribute to the protective effect against NK1.1+CD8<sup>+</sup> T cell-mediated intestinal inflammation. It will also be important to confirm such immunoregulatory roles of iNKT cells by employing more specific iNKT cell-deficient mouse model, such as the strain recently generated by CRISPR/Cas9 technology (46).

# AUTHOR CONTRIBUTIONS

SL: study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, statistical analysis. HP: study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, statistical analysis, obtained funding. JC: interpretation of data and review of the manuscript, obtained funding. LW: interpretation of data and review of the manuscript. LVK: interpretation of data and drafting of the manuscript, review of the manuscript. SH: study concept and design, acquisition of data, analysis and interpretation of data, drafting manuscript, statistical analysis, obtained funding, administrative, and material study supervision.

# FUNDING

This research was supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (HI18C0094010018) and Basic Science Research Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Education (2018R1D1A1B07049495).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02168/full#supplementary-material

# REFERENCES


pathology in inflammatory myopathies. Oncotarget (2015) 6:43230–43. doi: 10.18632/oncotarget.6462


**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 Lee, Park, Cheon, Wu, Van Kaer and Hong. 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.

# Lipid-Reactive T Cells in Immunological Disorders of the Lung

Seungwon Ryu1,2, Joon Seok Park <sup>3</sup> , Hye Young Kim1,2 \* and Ji Hyung Kim<sup>4</sup> \*

<sup>1</sup> Department of Biomedical Sciences, Seoul National University College of Medicine, Seoul, South Korea, <sup>2</sup> Institute of Allergy and Clinical Immunology, Seoul National University Medical Research Center, Seoul, South Korea, <sup>3</sup> Department of Microbiology and Immunobiology, Harvard Medical School, Boston, MA, United States, <sup>4</sup> College of Life Sciences and Biotechnology, Korea University, Seoul, South Korea

Regulation of T cell-mediated immunity in the lungs is critical for prevention of immune-related lung disorders and for host protection from pathogens. While the prevalent view of pulmonary T cell responses is based on peptide recognition by antigen receptors, called T cell receptors (TCR), on the T cell surface in the context of classical major histocompatibility complex (MHC) molecules, novel pathways involving the presentation of lipid antigens by cluster of differentiation 1 (CD1) molecules to lipid-reactive T cells are emerging as key players in pulmonary immune system. Whereas, genetic conservation of group II CD1 (CD1d) in mouse and human genomes facilitated numerous in vivo studies of CD1d-restricted invariant natural killer T (iNKT) cells in lung diseases, the recent development of human CD1-transgenic mice has made it possible to examine the physiological roles of group I CD1 (CD1a-c) molecules in lung immunity. Here, we discuss current understanding of the biology of CD1-reactive T cells with a specific focus on their roles in several pulmonary disorders.

Keywords: pulmonary disorders, lipid antigens, CD1 molecules, CD1-restricted T cells, natural killer T cells

### INTRODUCTION

The respiratory system comprises the airways and lungs; as such, it is exposed continuously to foreign materials. Consequently, the immune system in the respiratory system experiences both a huge load and a great variety of foreign antigens. This is significant because this exposure can induce immune reactions that damage lung tissue, thereby transiently or permanently impairing vital respiratory functions. The unique exposure profile of the lung has led to the evolution of immune responses that are not observed in lymphoid organs (1). For example, the surface layer of the lung is composed of airway epithelium, which is a specialized tissue that bears cilia. The epithelial cells secrete mucus that traps inhaled foreign materials; the cilia then move the mucus out of the respiratory system. The lung also harbors a considerable number of innate immune cells, including alveolar macrophages, interstitial macrophages, dendritic cells (DCs), and innate lymphoid cells, all of which act as first line defenders and activate the adaptive immune system, namely, B and T cells. T cells play a particularly important role in controlling immune responses associated with various lung diseases.

Classical T cells recognize foreign or self-antigens complexed with major histocompatibility complex (MHC) I and II molecules on the surface of antigen-presenting cells. When these complexes are recognized by T cell receptors (TCRs), the T cells differentiate into effector cells capable of secreting inflammatory cytokines and cytotoxic proteins that orchestrate downstream immune responses. The role of classical T cells in respiratory health and disease has been well studied [Chen and Kolls published a review on these cells (2)]. However, recent evidence suggests that unconventional subsets of T cells that react to lipid antigens presented

#### Edited by:

Luc Van Kaer, Vanderbilt University, United States

#### Reviewed by:

François Trottein, Centre National de la Recherche Scientifique (CNRS), France Mark L. Lang, University of Oklahoma Health Sciences Center, United States Seddon Y. Thomas, National Institute of Environmental Health Sciences (NIEHS), United States

#### \*Correspondence:

Hye Young Kim hykim11@snu.ac.kr Ji Hyung Kim jay\_kim@korea.ac.kr

#### Specialty section:

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

Received: 03 June 2018 Accepted: 05 September 2018 Published: 26 September 2018

#### Citation:

Ryu S, Park JS, Kim HY and Kim JH (2018) Lipid-Reactive T Cells in Immunological Disorders of the Lung. Front. Immunol. 9:2205. doi: 10.3389/fimmu.2018.02205

**89**

on cluster of differentiation 1 (CD1) molecules also play critical roles in pulmonary immune responses. While the percentage of CD1-reactive T cells in the lung is unclear, a previous study reports a significant percentage of these cells among circulating TCRαβ<sup>+</sup> cells (7% CD4<sup>+</sup> and 0.2% CD4−CD8−) within human PBMC population (3).

While CD1 proteins are expressed in every mammalian species that was examined, humans express five isoforms of CD1, categorized as group 1 molecules (CD1a, CD1b, and CD1c), group 2 molecules (CD1d), and group 3 molecules (CD1e) (4– 7). All CD1 molecules, apart from CD1e, display lipid antigens within their hydrophobic antigen-binding groove. The T cells that recognize lipids bound to group 2 CD1 (CD1d) molecules are known as natural killer T (NKT) cells; these are the best studied among the CD1-restricted T cells. This reflects the fact that, of all the CD1 molecules, only the CD1d gene is conserved in mice and humans: mice lack the other four CD1 isoforms. The presence of CD1d in mice greatly facilitated research on NKT cells, which has shown that these cells play pivotal roles in development of asthma, pulmonary infection, fibrosis, and other pulmonary disorders. NKT cells represent 5–10% of all T cells in the lungs of adult mice (8). Of the iNKT cells in the lung, 0.08% of CD45<sup>+</sup> cells are present in the lung parenchyma and 0.2% are present in the lung vasculature (9). By contrast, T cells in the lungs that are restricted by the group 1 CD1 molecules (CD1a–c) are less well studied. However, a number of recent studies suggest that these T cells also participate actively in immune responses in barrier organs such as the skin and the lungs.

T cells are particularly promising targets in terms of treatments for various lung diseases (2). To facilitate development of these therapies, a better understanding of lipid-reactive T cells is required. To address this, this review will describe the general features of lipid-reactive T cells and the roles they play in development of various pulmonary disorders, including asthma, chronic obstructive pulmonary disorder (COPD), fibrosis, infection, and cancer.

## THE CD1 FAMILY OF LIPID ANTIGEN-PRESENTING MOLECULES

Classical MHC class I and II molecules present peptide antigens to T cells. By contrast, CD1 molecules present hydrophobic antigens such as lipids to T cells (9). Structurally, CD1 molecules resemble MHC class I molecules in that they comprise an α-chain bearing three domains (α-1, α-2, and α-3), which is bound noncovalently to β2-microglobulin (β2m). However, CD1 molecules have some unique characteristics that distinguish them from MHC class I molecules. First, human CD1 genes are located on chromosome 1, whereas human MHC genes are located on chromosome 6 (in mice, CD1 genes are on chromosome 3 and MHC genes are on chromosome 17) (7). Second, CD1 molecules are less polymorphic. Third, MHC I and II molecules have six pockets in their antigen-binding groove (denoted A–F) whereas the binding groove of CD1 molecules harbors at least two antigen-binding pockets, named A′ and F′ ; however, these pockets are narrower and deeper than the A–F pockets in MHC molecules. In addition, these pockets are enriched in hydrophobic residues, which aid the stable binding of lipids to the CD1 groove. The subfamilies of CD1 molecules differ in terms of the size (volume and shape) and properties of these antigen-binding pockets. As a result, the CD1 molecules as a group can present a variety of hydrophobic antigens to T cells (10). Various foreign- and self- antigens that react with CD1 reactive T cells have been identified. These antigens include lipids, phospholipids, glycolipids, and lipopeptides with a large spectrum of size and polarity (10). In general, the hydrocarbon tails, usually alkyl chains, of lipids are buried in the pocket of CD1 molecules and the polar portions protrude, thereby providing a template for TCR engagement (11). Recent studies suggest that many lipid ligands of CD1 molecules, especially CD1c, do not interact directly with TCR; rather they affect the interaction between the TCR and CD1 molecules, thereby allowing or blocking activation of autoreactive T cells (11, 12).

All CD1 isoforms, except CD1e, present antigen. Mice express only CD1d (13), while other mammals (ranging from alpacas to sloths) harbor different combinations of the five CD1 isoforms [these are summarized in the Table 1 from (7)]. CD1e participates in presentation of lipid antigens only indirectly: it trims and transfers lipid antigens prior to presentation to other CD1 molecules (14, 15). CD1a-c molecules are expressed by professional antigen-presenting cells and thymocytes. In particular, Langerhans cells prominently express CD1a while DCs express CD1b and marginal zone B cells express CD1c. The group 2 CD1 molecule, CD1d, is expressed by both hematopoietic and non-hematopoietic cells in various organs, including skin, liver, and colon (16, 17). These differential expression patterns of CD1 molecules suggest that the individual CD1 isoforms may shape local T cell responses by presenting tissue-specific lipids. Moreover, when blood monocytes and hematopoietic CD34<sup>+</sup> progenitor cells are cultured with granulocyte-macrophage colony stimulating factor (GM-CSF) and IL-4, CD1a expression is induced (18–21). This suggests that CD1 expression can be controlled as necessary in vivo.

CD1 molecules load and present lipid antigens in a manner distinct from that of MHC molecules (22, 23). Initially, they are loaded with self-antigen immediately after synthesis in the endoplasmic reticulum (ER); this process is aided by a lipidtransfer protein (LTP) called microsomal triglyceride transfer protein (24, 25). The CD1 molecules then undergo intracellular trafficking (like other proteins synthesized in the ER). Thus, they

**Abbreviations:** MHC, major histocompatibility complex; TCR, T cell receptor; IFN, interferon; IL, interleukin; CD1, cluster of differentiation 1; COPD, chronic obstructive pulmonary disease; ER, endoplasmic reticulum; PBMC, peripheral blood mononuclear cell; hCD1Tg, human CD1 transgenic; Mtb, Mycobacterium tuberculosis; NKT, natural killer T cells; DP, double-positive; α-GalCer, α-galactosylceramide; iGb3, isoglobotrihexosylceramide; AHR, airway hyper-responsiveness; OVA, ovalbumin; BALF, bronchoalveolar lavage fluid; β2M, β2-microglobulin; MDSC, myeloid-derived suppressor cell; IAV, influenza A virus; IPF, idiopathic pulmonary fibrosis; TB, tuberculosis; BCG, Bacillus Calmette-Guérin; DC, dendritic cell; NSCLC, non-small cell lung cancer; SCLC, small cell lung cancer; ILC, innate lymphoid cell; FEV1, forced expiratory volume in one second; MIP, macrophage inflammatory protein; CTGF, connective tissue growth factor; TNF, tumor necrosis factor.

enter the secretory pathway, which traffics them from the ER to the Golgi apparatus and then finally to the plasma membrane. Sometimes, CD1 molecules exchange their self-antigen for an exogenous lipid antigen at either the plasma membrane or in various endosomal compartments, including early and late endosomes and lysosomes (26). Different CD1 molecules preferentially enter different endosomal compartments due to their disparate sorting motifs. For example, CD1b bears the AP3 sorting motif, which promotes entry into lysosomes (17). In addition, CD1a molecules lack sorting motifs and are therefore recycled to early endocytic compartments such as Birbeck granules, a specialized organelle found in Langerhans cells. Trafficking of CD1a molecules to this compartment is important for lipid antigen presentation to CD1a-restricted T cells (27). These disparate trafficking pathways suggest that different CD1 molecules capture distinct subsets of foreign antigens. In the endosomal compartments, capture of exogenous lipids is mediated by LTPs, including the CD1e molecule. For example, the self-lipid antigens expressed on CD1d molecules are removed and replaced by foreign lipid antigens with the help of a lysosomal LTP called saposin (28, 29). It should be noted that while some CD1 molecules retain the self-antigen that they received at the ER, they can also be loaded with another selfantigen. These CD1: self-antigen complexes can be recognized by autoreactive T cells. The original self-antigen can also function as a chaperone that promotes CD1 stability and trafficking to the plasma membrane, or as a determinant of activation of autoreactive T cell activation (12, 17).

#### Lipid-Reactive T Cells Group 1 CD1-Restricted T Cells

Research on group 1 CD1 molecules (CD1a-c) is limited by the fact that mice lack homologs of these proteins. Therefore, most studies have been conducted in vitro using human cells. T cells that recognize CD1a-c are more common in human peripheral blood than T cells that recognize CD1d: ∼2%, ∼1%, and ∼7% of αβTCR<sup>+</sup> cells in human peripheral blood recognize CD1a, CD1b, and CD1c, respectively, whereas only ∼0.1% of αβTCR<sup>+</sup> cells recognize CD1d (3, 30). This indicates the need for further studies on the functions of T cells that are restricted by CD1ac, even though there are some discrepancies regarding their percentages in different models/individuals.

Difficulties associated with studying CD1a-c-restricted T cells can be overcome by using a humanized CD1 transgenic mouse model (hCD1Tg) (31, 32) or humanized SCID mice that have been engrafted with human thymus, liver, and CD34<sup>+</sup> hematopoietic cells (33). Felio et al. used hCD1Tg mice to examine responses of CD1a-c-restricted T cells to Mycobacterium tuberculosis (Mtb) infection. They showed that Mtb-responsive CD1a-c-restricted T cells did not respond quickly to the infection; rather, they became activated later. Moreover, upon second stimulation, they showed boosted responses. Thus, they do not have the innate immune cell-like activities of NKT cells, which are restricted by CD1d and exhibit strong early responses; rather, they more closely resemble classical adaptive lymphocytes (31). This finding was validated by de Lalla et al., who showed that CD1a-c-restricted and self-reactive T cells within adult PBMCs are more likely to be memory T cells than are the same cell populations in umbilical cord blood (3). However, CD1a-c-restricted T cells do not differ from CD1d-restricted T cells in terms of their ability to secrete Th1/Th17 (31, 34) and Th2 (35, 36) cytokines. The percentage and number of CD1drestricted T cells in the mouse spleen increase with age (37). However, in human blood, age is inversely related with the frequency of CD1d-restricted T cells (38), suggesting that agingassociated accumulation of NKT cells may depend on the species or environment. Classical T cells also exhibit age-related changes in memory phenotype (3).

Although our understanding of development of CD1a-crestricted T cells remains poor, Li et al. followed the development of HJ1, a human T cell clone restricted by CD1b; they did this by creating a double transgenic mouse that expressed the TCR of the clone and human CD1b (HJ1Tg/hCD1Tg mice) (34). They showed that to develop properly, HJ1 T cells must encounter CD1b-expressing hematopoietic cells in the thymus. In the absence of CD1b-mediated thymic selection, most HJ1 T cells remained at the CD4+CD8<sup>+</sup> double-positive (DP) stage and did not undergo positive selection. However, if CD1b-expressing hematopoietic cells were present, HJ1 T cells underwent positive and negative selection. During this process, the cells changed from DP cells to DPdull and then to doublenegative cells. These changes were accompanied by expression of activation markers CD69, CD122, and CD44. They also expressed PLZF (promyelocytic leukemia zinc finger, which is encoded by Zbtb16), as did CD1d-restricted invariant NKT (iNKT) cells. These observations suggest that CD1a-c-restricted T cells may develop via the same developmental pathway used by iNKT cells (34). However, further research into the mechanism by which CD1a-c-restricted T cells develop, and into the endogenous lipids involved in this process, is needed.

#### Group 2 CD1-Restricted T Cells

Group 2 CD1 (CD1d)-restricted T cells are called NKT cells. They are a small subset of CD1-restricted αβ T cells that express both NK cell markers (NK1.1 in mice and CD161 in humans) and TCRs. They respond quickly during the early phase of immune reactions by secreting cytokines. They also play a critical role in bridging the innate and adaptive immune responses (39).

There are two subsets of NKT cells. Type I NKT cells are referred to as invariant NKT (iNKT) cells because they have a semi-invariant TCR that is encoded by the Vα14-Jα18 or Vα24-Jα18 TCRα gene segments in mice and humans, respectively. iNKT cells respond strongly to α-galactosylceramide (α-GalCer), a glycolipid ligand first isolated from a marine sponge, when it is presented on CD1d molecules (40). Several other endogenous ligands, including α-linked glycosylceramides (41) and β-linked glycosylceramides (42) stimulate iNKT cells. However, the ligands that are responsible for the selection of iNKT cells in the thymus remain unclear at present. However, the fact that the β-linked glycosylceramides used for some research turned out to be contaminated with αlinked anomers, which are very potent stimulators of iNKT cells, negates some of the findings (41, 43, 44). In addition to α-linked monoglycosylceramides, isoglobotrihexosylceramide (iGb3), a lysosomal glycosphingolipid, is an endogenous ligand for iNKT cells (45). Mice deficient in hexosaminidase b (Hexb), which converts iGb4 to iGb3, fail to develop iNKT cells, supporting the in vivo significance of iGb3 as an endogenous ligand (45). However, iGb3 synthase-deficient mice exhibit normal development of iNKT cells (46). Still, given that iGb3 stimulates iNKT cells in a TCR-dependent manner, it might be a bona fide endogenous ligand. Several exogenous ligands derived from bacteria and fungi have been identified. Glycosylceramides derived from the cell wall of Sphingomonas, and glycolipids from Streptococcus pneumoniae, activate iNKT cells in a CD1d-dependent manner (47–49). In addition to bacterial products, a fungal glycolipid called asperamide B can be presented on CD1d, thereby stimulating both human and mouse iNKT cells (50). Interestingly, microbial products can activate iNKT cells via indirect mechanisms by inducing endogenous glycolipid ligands (51, 52). TLR pathways facilitate biosynthesis of glycosphingolipids, which are endogenous ligands for iNKT cells. Of note, cytokines take part in the TLR-mediated iNKT cell activation. TLR4 and TLR9 mediate IFN-γ production by iNKT cells, not IL-4 production, in IL-12- and type I IFN dependent manners, respectively (51). Such cytokine contributions require TCR engagement for the optimal iNKT activation. However, it is also shown that cytokines themselves could activate iNKT cells independently to TCR triggers at certain levels (51–53). Moreover, cytokines and antigens are differentially required for cytokine production by iNKT cells, depending on the stimulants (54). TLR-induced, not ligand-driven, IFN-γ production by iNKT cells is impaired in the absence of IL-12. On the other hand, ligand- or TLR- mediated IL-4 production by iNKT cells does not require IL-12. During infections, iNKT cells produce IFN-γ in a manner dependent on IL-12 and IL-18; However, expression of IL-4 is independent of IL-12 but dependent on CD1 ligands and MyD88 (54, 55). Interestingly, IL-4 production in iNKT cells is prominent under Th2 inflammatory settings found in airway inflammation (56).

Despite expressing an invariant TCR, iNKT cells are quite heterogenous. When Watarai et al. examined expression of IL17RB (a receptor for IL-25) and CD4 in response to IL-2, they identified three subpopulations of thymic iNKT cells; namely, a CD4+IL17RB<sup>+</sup> population that produces Th2 and Th17 cytokines (IL-13, IL-9, IL-10, IL-17A, and IL-22), a CD4−IL17RB<sup>+</sup> population that produces Th17 cytokines (IL-17 and IL-22), and a CD4+IL17RB<sup>−</sup> subset that produces IFNγ (57). These findings were validated by the study of Lee et al., who showed that NKT cells can be classified as Tbet-expressing NKT1, GATA-3-expressing NKT2, and RORγtexpressing NKT17 lineages (58). Recently, two studies identified new NKT subsets; namely, Bcl-6-expressing follicular helper NKTFH (59) and Nfil3-expressing NKT10 (60, 61) cells. In terms of thymic development (**Figure 1**), iNKT cells develop from a common lymphoid progenitor. Interestingly, however, positive selection of iNKT cells is mediated by DP thymocytes (CD4+CD8+), which present the lipid-CD1d complex, rather than by thymic epithelial cells (62). After positive selection, iNKT cells progress from stage 0 to stage 3 (63). Thus, CD24highCD44lowNK1.1<sup>−</sup> stage 0 NKT precursors first develop into immature stage 1 cells that express CD24low; they also express GATA-3 and therefore produce IL-4 (64). Stage 1 cells then become CD24lowCD44highNK1.1<sup>−</sup> stage 2 NKT cells, which eventually develop into mature CD24lowCD44highNK1.1<sup>+</sup> stage 3 NKT cells that produce IFN-γ (65–67). PLZF is thought to be the key transcription factor that drives NKT cell differentiation (68, 69). While all NKT cells at all stages express Zbtb16 (the gene that encodes PLZF), conditional knockout of Zbtb16 in NKT cells showed that NKT cells cannot develop into immature stage 2 NKT cells without PLZF; mutant mice accumulated stage 1 NKT precursors in the thymus (68). However, considering that NK1.1 is not expressed by all mouse strains, another model of NKT development has been proposed; namely, that the NKT1, NKT2, and NKT17 lineages develop separately from a common NKT precursor rather than by passing sequentially from stage 0 to stage 3 (58, 70).

The second subset of NKT cells, namely, type II NKT cells, express a variable TCR that is restricted by CD1d but is not recognize by α-GalCer (71). Instead, type II NKT cells can be stimulated by sulfatide, β-GalCer, phosphatidylinositol, and phosphatidylglycerol (72). Studies of type II NKT cells are limited by a lack of specific markers for these cells and by a paucity of techniques that permit their analysis (by contrast, iNKT cells can be readily and specifically detected using CD1d/α-GalCer tetramers) (73). However, several experimental strategies have enabled study of the functions of type II NKT cells; these include the use of Jα18-deficient IL-4 reporter mice (74) and 24αβ transgenic mice that express the TCR of a type II NKT cell clone (75). These studies showed that, like iNKT cells, type II NKT cells develop via a linear maturation process marked by changes in expression of CD44 and NK1.1; moreover, like iNKT cells, they differentiate into NKT1, NKT2, and NKT17 cells, as shown by their expression of key transcription factors. While these cells remain relatively poorly characterized, it is possible that type II NKT cells may contribute to lung pathology in humans; indeed, evidence suggests that their close cousins, type I NKT cells, play a role in lung disorders (see below for more details). Moreover, humans harbor more type II NKT cells than type I NKT cells (76). Although the CD1-reactive T cell population in the lung has not been analyzed thoroughly, the spatial distribution of different subsets of iNKT cells in the mouse lung was investigated using a novel approach aimed at distinguishing iNKT cells in lung parenchyma from those in lung vasculature (9). The dominant iNKT subset in lung parenchyma was iNKT17, whereas that in lung vasculature was iNKT1. In both compartments, iNKT2 was the least populating subset. Accumulation of pulmonary iNKT cells relies on the CXCR6-CXCL16 axis (77). By contrast, intratracheal injection of an α-GalCer analog PBS57 promotes extravasation of CD1d-restricted T cells to the lung parenchyma via interaction between CCR4 and CCL17 (8).

#### LIPID-REACTIVE T CELLS IN PULMONARY DISORDERS

#### Asthma

Asthma is a well-known allergic disease of the lung that affects over 7% of adults in the United States of America (78). Symptoms include shortness of breath, coughing, wheezing, and

airway hyper-responsiveness (AHR). Asthma can be classified broadly into two phenotypes, allergic and non-allergic, based on production of critical immune mediators. Allergic asthma accounts for more than 50% of asthma cases. It is known as a "Th2 disorder" because type 2 cytokines are pivotal for progression of the disease (79). While conventional Th2 cells contribute to development of allergic asthma, many studies suggest that lipid-reactive NKT cells and innate lymphoid cells may also play a role in this pathology (80, 81).

A variety of studies suggested that CD1d-restricted NKT cells could be critical inducers of asthma in murine models. An ovalbumin (OVA)-induced model of asthma first showed that mice lacking iNKT cells (i.e., CD1d- or Vα14-Jα18-deficient mice) do not develop AHR, which is a critical feature of asthma (82–85). Mechanistically, IL-4 and IL-13 which are produced by iNKT cells, are essential for development of OVA-induced AHR (83). Other asthma-related features, including airway eosinophilia, the elevation of both type 2 cytokine levels in bronchoalveolar lavage fluid (BALF), and allergen-specific IgE levels in serum, are also dependent on iNKT cells. Moreover, iNKT cells induce asthma even when CD4<sup>+</sup> T cells are absent; when MHC II-deficient mice (which lack conventional CD4<sup>+</sup> T cells but have iNKT cells) were challenged with α-GalCer, the mice developed AHR equally as well as the wild-type mice (86). The contribution of iNKT cells to development of asthma has also been tested with respect to real-world allergens (rather than OVA or α-GalCer, which are experimental triggers). iNKTdeficient mice develop less allergic asthma compared to wild-type mice when exposed to ragweed (87), house dust extracts (88), or a fungus (Aspergillus fumigatus) (50). In addition, Pichavant et al. showed that iNKT cells expressing IL-17A are essential for development of non-allergic asthma induced by exposure to ozone, another real-world cause of asthma (89).

However, studies by other groups suggest that iNKT cells are not required for allergic airway inflammation characterized by eosinophilia and type 2 cytokine production (90–93). These studies show that airway inflammation is dependent on MHC class II molecules, not on CD1d molecules (90, 92). Also, β2mdeficient mice, which are unable to express CD1d molecules, develop asthma despite lacking iNKT cells (91, 93, 94). However, Koh et al. suggested that some NKT cells in β2m-deficient mice are restricted to a β2m-independent form of CD1d, and these cells can compensate for deficiency of conventional iNKT cells. Despite this possibility, iNKT cells might become more dispensable than T cells with respect to inducing AHR when mice are challenged repeatedly with antigens such as OVA (90, 92).

It should be noted that Jα18 (or Traj18) knockout mice, which are often used to study the roles of iNKT cells and type II NKT cells, show a significant (>60%) reduction in TCRα chain diversity (95). A recent study attempted to resolve this issue by developing a new type of Traj18 knockout mouse strain lacking iNKT cells but with normal TCRα diversity. When OVA- and cockroach allergen-induced asthma were assessed in these mice, knockout mice showed significantly less severe AHR than wildtype mice (96). Taken together, the results from murine models of asthma are conflicting; however, iNKT cells may contribute to the pathogenesis of asthma under certain conditions.

Akbari et al. (97) conducted flow cytometry experiments and found that BALF from patients with moderate to severe asthma contained a very high percentage of iNKT cells (about 60% of all CD3<sup>+</sup> cells); BALF from healthy subjects contained very few iNKT cells (97). It should be noted, however, that several other groups found low percentages (<2%) of iNKT cells in BALF from adult asthmatics; these percentages did not differ significantly from those in control subjects detected using CD1d tetramers, an anti-Vα24 antibody, or an anti-Vβ11 antibody (98–101). It was proposed that differences in the percentage of iNKT cells in BALF, sputum, or bronchial biopsies may be due to nonspecific binding of CD1d-tetramers, or to inappropriate gating on iNKT cells (102). Other human studies noted increased percentages of iNKT cells in BALF from childhood asthma patients (0.435% of αβ T cells in asthma vs. 0.116% of αβ T cells in control subjects) (103), and in the lungs of patients with severe asthma compared with patients with well controlled asthma (104). Thus, the link between the percentage of iNKT cells in human lung or bronchial specimens from asthma patients and disease pathogenesis remains still unclear.

A functional study from the United Kingdom, which was based on a human allergen challenge model of asthma, revealed that the percentage of iNKT cells in bronchial biopsies taken 24 h after allergen challenge increased by 15% along with the increase in AHR (105). Moreover, Agea et al. showed that the PBMCs from subjects who were allergic to cypress pollen responded to two specific lipid antigens in the pollen, namely, phosphatidylcholine and phosphatidylethanolamine. These lipid antigens, which are loaded onto CD1 molecules, stimulated T cells within the PBMCs in CD1a- or CD1d-dependent fashion (106). Thus, both CD1aand CD1d-restricted T cells may participate in allergic reactions in humans.

Although most studies to date have focused on the pathogenic roles of iNKT cells in asthma, iNKT cells also have protective functions under specific preconditions. For example, when young mice are infected with influenza A virus, they are protected against allergen-induced AHR when they become adults (107). CD4−CD8<sup>−</sup> DN NKT cellsshowing high expression of CD38 produce IFN-γ and induce CD4 T cell proliferation in a contact dependent manner (108). Glycolipids from Helicobacter pylori activate suppressive NKT cells (107). Notably, Sharma et al. showed that benzo[a]pyrene, a classical environmental air pollutant that causes respiratory disease, suppressed IFN-γ production by CD1a- and CD1d-restricted T cells in vitro by downregulating expression of CD1a and CD1d by human DCs (109). Thus, it appears that certain subsets of NKT cells, such as IFN-γ-producing NKT cells, may suppress AHR.

Because research into the roles of CD1d-restricted iNKT cells in asthma has produced conflicting results, comprehensive studies aimed at characterizing the cause and identifying the contribution made by these cells are needed. Also, we should not overlook the fact that iNKT cells alter the cytokine milieu and affect asthmatic responses under certain conditions. Further studies should determine the role of CD1a-c restricted T cells in the pathogenesis of asthma.

# Chronic Obstructive Pulmonary Disease (COPD)

COPD is another chronic pulmonary disease that causes difficulty in breathing; however, it differs from asthma in several aspects. First, it is not an allergic disease. Second, the airway obstruction is largely irreversible. Third, smoking is a well-known risk factor for COPD. Fourth, it is characterized pathologically by either chronic obstructive bronchiolitis or emphysema (110). Although one study shows that the percentage of iNKT cells in induced sputum from COPD patients are not different compared to that in control subjects (101), others suggest that iNKT cells contribute to the pathogenesis of COPD. For example, mice with chronic lung disease caused by Sendai virus infection harbor large numbers of alternatively-activated IL-13-secreting macrophages in the lung. Experiments with CD1d or Jα18 knockout mice revealed that iNKT cells are required to generate virus-induced IL-13-producing macrophages. Thus, iNKT cells facilitate chronic airway inflammation in mice (111). The same authors showed that a similar mechanism may also participate in COPD. Immunofluorescence staining of lung tissues from patients with COPD and from normal controls revealed that the former harbored significantly higher numbers of Vα24<sup>+</sup> iNKT cells and IL-13+CD68<sup>+</sup> macrophages (111). Wang et al. also showed that patients with COPD harbor higher percentages of CD56+CD3<sup>+</sup> NKT cells in the blood than control subjects (although these data should be interpreted carefully due to the possibility of confounders, i.e., smoking, with respect to NKT cells and COPD) (112). In addition, Pichavant et al. found that oxidative stress generated by cigarette smoke activates DCs and airway epithelial cells, which in turn cause iNKT cells to accumulate in the lung and release IL-17 (113). Furthermore, when mice are challenged repeatedly with intranasal α-GalCer, iNKT cells are activated and the mice develop COPD-like symptoms, including mucus hypersecretion and pulmonary emphysema (114). Taken together, these results suggest that iNKT cells might play a pathogenic role in COPD. To date, no study has investigated the role of CD1a-c-restricted T cells in COPD. However, limited number of studies in humans evaluated CD1a expression in DCs from COPD patients (115–117). These studies suggested either no difference or an increase in the number of alveolar langerin-positive and CD1a<sup>+</sup> immature DCs in response to chronic cigarette smoking, which contribute to the COPD pathology. The role of CD1a-c-restricted T cells in COPD remains unclear.

#### Pulmonary Fibrosis

Pulmonary fibrosis is a collection of chronic progressive lung diseases characterized by abnormal wound healing, which in turn generates thickened and stiff lung tissue. The causes of lung fibrosis are numerous; however, most studies on the role of iNKT cells focused on idiopathic pulmonary fibrosis (IPF). IPF is a chronic, progressive fibrotic lung disease that is diagnosed (when no other cause of fibrosis can be found) according to the results of histology or high-resolution CT. With the exception of the conditionally recommended use of pirfenidone and nintedanib (118), no drugs improve the survival of patients with IPF. Therefore, better understanding of the disease is required (119). A common animal model used to study the pathophysiology of IPF is bleomycin-treated mouse; bleomycin is a chemotherapeutic agent that causes lung fibrosis as a side effect. Jα18 knockout and CD1d knockout bleomycin-treated mice exhibited exacerbated symptoms of lung fibrosis (120). Moreover, treatment with α-GalCer ameliorated lung fibrosis in an iNKT cell-dependent manner (121). Of note IFN-γ produced by iNKT cells was responsible for iNKT-mediated protection against fibrosis (120). However, other studies suggest either that iNKT cells are not necessary for lung pathogenesis, or that they play a pathogenic role, when mice are not treated with exogenous ligands (121, 122). Despite this discrepancy with respect to the impact of endogenous ligands or cytokine-driven activation of iNKT cells on disease pathogenesis, exogenous ligand-induced activation of iNKT cells consistently protects against lung fibrosis (120–122). Thus, iNKT cells likely protect against lung fibrosis. In contrast to CD1d-restricted iNKT cells, the function of CD1ac-restricted T cells has not been studied either human or mouse models of pulmonary fibrosis.

### Lung Infection (Pneumonia)

Pneumonia is an inflammatory lung disease caused by viral, bacterial, or fungal infections (123). Unlike the other pulmonary diseases discussed above, many lines of evidence suggest that group 1 CD1-restricted T cells respond to respiratory infections.

Pulmonary infection with Mtb results in an atypical pneumonia called pulmonary tuberculosis (TB). It is a significant public health concern because TB is a highly contagious airborne disease. Moreover, despite development of the M. bovis-derived Bacillus Calmette-Guérin (BCG) vaccine long time ago, it is still difficult to prevent TB in adults. These factors, along with the high prevalence of TB, its disease course, and practical issues regarding long-term treatment regimens, have led to great interest in understanding Mtb infection and in developing new vaccines that will curb TB (124). In 1994, Beckman et al. showed that mycolic acid derived from the Mtb cell wall binds to CD1b (125). Later, Moody et al. found that mannosylbeta1-phosphodolichol and didehydroxymycobactins, another Mtb lipid and a lipopeptide antigens, activate CD1c- and CD1arestricted T cells, respectively, as well as peripheral blood cells, from patients with TB (126, 127). The percentages of the group 1 CD1-responding T cells in individuals exposed to Mtb are higher than those in control individuals. Furthermore, they produce IFN-γ responses to lipid antigens derived from Mtb, suggesting protective role against the infection (3, 128). However, group 1 CD1-restrictred T cells produce not only Th1 cytokines, but also Th17 and Th2 cytokines (35, 129, 130). Therefore, further studies should investigate physiological role of Th17 and Th2 cytokineproducing group 1 CD1-restricted T cells in infection. Once the human group 1 CD1 transgenic (hCD1Tg) mice became available, more specific functional studies of CD1-restricted T cells in TB infection have been performed. As mentioned above, Felio et al. showed that, when hCD1Tg mice are infected with Mtb, group 1 CD1-restricted T cells exhibit adaptive cell-type responses, and release IFN-γ at a late time-point after primary exposure to Mtb; however, release is much faster after subsequent exposure (31). Contradictory results were reported by Zhao et al.; When they infected hCD1Tg mice (that were also transgenic for a mycolic acid-specific CD1b-restricted TCR) with TB, they found that CD1b-restricted T cells responded more quickly than Mtbspecific CD4<sup>+</sup> T cells to Mtb challenge (131). These disparate reaction patterns may reflect differences in the methods used by Felio et al. and Zhao et al. with respect to antigen challenge; the former used intraperitoneal injection of Mtb (31), whereas the latter used intranasal challenge (131). In any case, these studies show that group 1 CD1-restricted T cells participate in immune responses to Mtb. Moreover, peripheral blood from BCG-immunized subjects contains a high percentage of group 1 CD1-restricted CD8 T cells, which react with live BCG-infected DCs (132). Interestingly, when Parlato et al. compared the gene expression patterns of Mtb antigen-challenged DCs from patients with active and latent TB infection, they found that DCs from patients with active TB lacked CD1a and CD1c expression. By contrast, DCs from patients with latent TB exhibited upregulated expression of CD1a. This suggests that, even though group 1 CD1-restricted T cells may help to protect the host from TB by producing IFN-γ, specific CD1a–c-restricted subsets of these cells may respond to the same infection in different ways (133).

Group 2 CD1-restricted T cells (specifically, iNKT cells) may be involved in immune responses to Mtb. A mouse study showed that injection of an anti-CD1d antibody exacerbated the symptoms of TB, which suggests that iNKT cells may have anti-microbial functions (134). In addition, Sada-Ovalle et al. found that adoptive transfer of iNKT cells reduced the bacterial burden in Mtb-infected mice by producing IFN-γ and killing intracellular bacteria (55). This protective function of iNKT cells suggests that vaccination with common lipid antigens may boost the beneficial anti-TB functions of CD1-restricted T cells (135).

Since human immune responses differ according to the pathogen encountered, it is possible that lipid-reactive T cells respond quite differently to infectious agent other than Mtb. Nevertheless, it is noteworthy that various studies show that iNKT cells also largely play protective roles against other pathogens such as Streptococcus pneumoniae and Saccharopolyspora rectivirgula, the most common causes of community-acquired pneumonia (48, 136) and Farmer's lung (137) respectively. Exceptions include a few cases of noncommon pneumonia occurring in adults. For example, the causative agent of atypical pneumonia is Chlamydia trachomatis. Patients become infected with C. trachomatis either after sexual activity or by vertical transmission from an infected mother. Bilenki et al. showed that, when a mouse model of chlamydial pneumonia is injected with α-GalCer and then infected with C. muridarum, the α-GalCer-activated NKT cells increase the chlamydial burden and worsen the symptoms (138).

Contrary to microbial infections in which CD1-reactive T cells are mainly activated through microbial lipid-CD1-dependent mechanisms, viral infections are a good model in which

#### TABLE 1 | Studies of lipid-reactive T cells in pulmonary disorders.


#### (Continued)

#### TABLE 1 | Continued


(Continued)


to demonstrate the CD1- or TCR-independent mechanisms underlying CD1-restricted T cell activation. Unfortunately, no study has examined the role of group 1 CD1-restricted T cells in pulmonary viral infection, except during systemic viral infections such as HIV (139), which might cause pulmonary infection. Therefore, we focus here on the biology of iNKT cells in viral infection models.

Influenza virus is a representative respiratory virus that causes seasonal flu and pneumonia with high morbidity and mortality (140). Several studies showed that NKT cells control lung inflammation during influenza A virus (IAV) infection through dampening suppression of the virus-specific T cell responses or controlling the IAV-associated lethal inflammation. During IAV (PR8, H1N1) infection, iNKT cells suppress myeloid-derived suppressor cells (MDSCs), which contribute to lung pathology and inhibit the activity of the virus-specific immune cells (141). The MDSC suppressive activity was inhibited by iNKT cells, which were activated in endogenous lipids:CD1d-dependnet manner. MDSCs derived from Hexβ <sup>−</sup>/<sup>−</sup> mice (unable to produce endogenous lipid ligands) failed to activate iNKT cells, thereby sustained their suppressive capability. In another study showing a protective role of iNKT cells in IAV (Scotland, H3N2) infection, however, the underlying mechanism was not mediated by MDSCs, but by a positive effect of iNKT cells on the migration of respiratory dendritic cells to the draining lymph nodes, a phenomenon that favors the expansion of virus-specific CD8<sup>+</sup> T cells (142). Also, iNKT cells suppressed the fatal inflammation by lysing infected inflammatory monocytes (the dominant immune cell in IAV-infected lungs) in a CD1d-dependent fashion (143). In contrast to the CD1d-dependent actions of iNKT cells, a cytokine-mediated, CD1d-independent iNKT cell activation was shown to limit pulmonary inflammation and a secondary bacterial infection during IAV infection. RORγt <sup>+</sup> iNKT cells (iNKT17) secrete a lung-protective cytokine, IL-22, in response to stimulation by IL-1β and IL-23 secreted by IAV-infected DCs (144, 145). Interestingly, the reduction of the burden of

secondary bacterial infection (Streptococcus pneumoniae) after IAV infection was observed when recombinant IL-22 was treated intranasally (146). This new strategy draws an attention considering the severe consequences from the co-infections (147, 148). Although we could speculate that there is no lipid antigen directly from viral genome in cells with virus infection, it should be verified with the in vivo system which enabled to distinguish the origin of effector functions (either TCR-dependent or independent manner) in iNKT cells. A Nur77gfp reporter system, which expresses GFP in response to TCR stimulation only, was used to show that IFN-γ-producing iNKT cells are induced during infection with murine cytomegalovirus (MCMV) without any change in GFP expression (149). It is now anticipated that the previously reported (i.e., not restricted to infection models) effector functions of CD1-restricted T cells will be re-evaluated using novel approaches to better understand the contribution of the TCR to iNKT cell activation.

#### Lung Cancer

The role of lipid-reactive T cells (especially α-GalCer-reactive iNKT cells) in tumors began attracting attention when Kobayashi et al. (150) showed that iNKT cells play protective roles in a B16 melanoma mouse model (150). Although CD1-reactive T cells are best known for their ability to produce cytokines, iNKT cells (like NK cells and cytotoxic T cells) are also cytotoxic; thus, they may be potential targets for therapies aimed at inducing anti-tumor immunity (151). The first in vivo iNKT cell-based immunotherapy regimen tested in patients with a variety of cancers, including breast, colon, liver, skin, kidney, prostate, peritoneum, and lung cancer. The study showed that, when iNKT cells were activated by treatment with α-GalCer-loaded DCs, they contributed to the tumor suppressive environment and reduced the size of the tumors (152). It is well recognized that iNKT cells protect against lung cancers, particularly non-small cell lung cancers (NSCLC), by producing IFN-γ and cytotoxic proteins (153–155). The percentage of NKT cells (CD56<sup>+</sup> CD3<sup>+</sup> T cells) in the blood of NSCLC and SCLC (small cell lung cancer) patients is increased compared to healthy controls (156). In agreement with the beneficial role of iNKT cells in lung cancer, the numbers of iNKT cells were compared following the cancers of differing histologic type and stage. It was found that the number was: (1) lower in adenocarcinoma, which has the worse prognosis than in squamous cell carcinoma (157); (2) lower in stage 1 than in stage IV cancers (156). Another human study showed that the percentage of Vα24 NKT cells in the PBMC of lung cancer patients was 0.01–0.03%; however, patients with recurrent cancers harbored a significantly lower percentage of NKT cells than healthy donors (158). Collectively, the data suggest that more immunosuppressive environments encountered at the later stages of lung cancer reduce the percentage of iNKT cells; this may allow the cancer to escape immune surveillance. In line with this idea, one study shows that injecting in vitro-expanded autologous iNKT cells may be an effective form of iNKT cellbased immunotherapy (159).

Several studies have assessed the role of type II NKT cells in tumor immunity. One such study was conducted by Terabe et al., who showed that when CD1d knockout and Jα18 knockout mice were used as CT26 colon carcinoma metastasis to the lungs, both knockouts developed significantly lower numbers of lung tumor nodules, compared to wild-type mice. However, this effect was more pronounced in CD1d knockout mice (which lack both type I and II CD1d-restricted NKT cells) than in Jα18 knockout mice (which lack only type I CD1d-restricted NKT cells). This suggests that type II NKT may suppress tumor metastasis or growth (160). However, evidence suggests that type I NKT cells promote immunity to CT26 cancer cells while type II NKT cells suppress it. Moreover, an absence of type I NKT cells is associated with stronger type II NKT cell-mediated suppression, whereas type II NKT cells appear to downregulate protective tumor immunity mediated by type I NKT cells. These findings suggest that type I and II NKT cells may counter the effects of each other in a balanced manner (161).

# CONCLUSION AND PERSPECTIVES

Here, we discussed the role of CD1-restricted T cells in lung diseases, including asthma, COPD, pulmonary fibrosis, infection, and cancer. Studies on group 2 CD1-restricted T cells are prolific. Less is known about group 1 CD1-reactive T cells because mice lack group 1 CD1 genes. Nevertheless, the current state of knowledge suggests that, like conventional T cells, group 1 and 2 CD1-restricted T cells can play either a protective role or a pathogenic role in lung disease (**Table 1**).

Despite the large body of work on CD1-restricted T-cell biology, many questions remain in this field, as follows. First, our understanding of how group 1 CD1-restricted T cells participate in lung diseases is limited. The recent development of transgenic mouse models expressing human CD1a, CD1b, or CD1c will improve this situation. Indeed, two studies based on these mice showed that they generate group 1 CD1-restricted T cell responses that are consistent with the responses of human T cell lines (129, 130). While these studies mainly showed that CD1-restricted T cells participate in skin inflammation, these group 1 CD1 transgenic mice may be useful for examining the in vivo roles of CD1-restricted T cells in other diseases. Second, no study has confirmed that autoreactive CD1-restricted T cells contribute to development of lung diseases. By contrast, several reports suggest that self-lipid-reactive T cells may participate in the pathology of psoriasis, which is an autoimmune skin disease (129, 165). Autoimmune pulmonary diseases are either primary autoimmune diseases or systemic autoimmune diseases that show pulmonary manifestations: the latter include systemic sclerosis, systemic lupus erythematosus, rheumatoid arthritis, and Sjogren's syndrome (166). Given that self-lipidreactive T cells are associated with development of psoriatic skin, it is highly likely that there are self-lipid antigens that could promote the lung manifestations observed in systemic autoimmune diseases. In addition, lipid antigens in the lungs still need to be defined. An allergic reaction in the lungs can be induced by house dust mite and ozone, which activate CD1 restricted T cells (35, 89). These substances have the potential to generate a neoantigen for CD1-restricted T cells. For example, house dust mite contains phospholipase A2 (PLA2), which

converts self-lipids into neoantigens. Similarly, it is speculated that ozone may chemically modify self-lipids to render antigenic properties against CD1. Notably, many lipid ligands of CD1c do not make direct contact with the TCR to trigger CD1 mediated autoreactivity (12). Further investigation is required to understand the circumstances in which CD1-autoreactive T cells are activated and how they are involved in lung disorders (**Figure 2**). Finally, the role of NKT cells in asthma should be re-evaluated. There is a large body of work on the role of CD1-restricted T cells in asthma, which is the most studied disease in this field. This work has provided ample evidence that CD1-restricted T cells, especially NKT cells, are involved in the pathology of asthma in mouse models. However, controversy about the role of these cells in asthmatic humans arose in the late 2000s when some studies reported that asthmatics have low percentages of NKT cells (98, 99, 101). It seems quite possible, for example, that NKT cells play an important role in a subset of patients who are allergic to lipids. This possibility is supported by the fact that the triggering antigen in patients with allergic asthma is very heterogeneous; these patients show distinct individual patterns of reactivity to diverse allergens (167). While the most common allergens in patients with asthma do not contain lipid components, it seems possible that at least some patients would be allergic to lipid components that can activate NKT cells.

This review summarizes current knowledge of CD1-restricted T cells, their specific lipid antigens, and their role in lung disease. Further research is likely to reveal hitherto unsuspected contributions to the pathogenesis of various lung diseases.

# AUTHOR CONTRIBUTIONS

SR, JP, JK, and HK wrote and revised the manuscript.

# FUNDING

HK was supported by a grant from the Korea Healthcare Technology R&D Project, Ministry for Health and Welfare Affairs, Republic of Korea (HI15C1736). JK was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2018R1D1A1B07048813).

#### REFERENCES


Am J Respir Crit Care Med. (2015) 192:e3–19. doi: 10.1164/rccm.201506- 1063ST


antigen presenting cells. J Clin Immunol. (2012) 32:1071–81. doi: 10.1007/s10875-012-9697-9


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

Copyright © 2018 Ryu, Park, Kim and Kim. 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.

# Positive & Negative Roles of Innate Effector Cells in Controlling Cancer Progression

Dorian Stolk <sup>1</sup> , Hans J. van der Vliet <sup>2</sup> , Tanja D. de Gruijl <sup>2</sup> , Yvette van Kooyk <sup>1</sup> and Mark A. Exley 3,4,5 \*

*<sup>1</sup> Department of Molecular Cell Biology and Immunology, VU University Medical Center, Amsterdam, Netherlands, <sup>2</sup> Department of Medical Oncology, VU University Medical Center, Amsterdam, Netherlands, <sup>3</sup> Faculty of Biology, Medicine and Health, University of Manchester, Manchester, United Kingdom, <sup>4</sup> Harvard Medical School, Brigham and Women's Hospital, Boston, MA, United States, <sup>5</sup> Agenus, Inc., Lexington, MA, United States*

Edited by:

*Luc Van Kaer, Vanderbilt University, United States*

#### Reviewed by:

*Shinichiro Motohashi, Chiba University, Japan Tonya J. Webb, University of Maryland, Baltimore, United States Randy Brutkiewicz, Indiana University Bloomington, United States*

\*Correspondence:

*Mark A. Exley mexley@partners.org; mark.exley@manchester.ac.uk*

#### Specialty section:

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

Received: *06 June 2018* Accepted: *13 August 2018* Published: *21 September 2018*

#### Citation:

*Stolk D, van der Vliet HJ, de Gruijl TD, van Kooyk Y and Exley MA (2018) Positive & Negative Roles of Innate Effector Cells in Controlling Cancer Progression. Front. Immunol. 9:1990. doi: 10.3389/fimmu.2018.01990* Innate immune cells are active at the front line of host defense against pathogens and now appear to play a range of roles under non-infectious conditions as well, most notably in cancer. Establishing the balance of innate immune responses is critical for the "flavor" of these responses and subsequent adaptive immunity and can be either "good or bad" in controlling cancer progression. The importance of innate NK cells in tumor immune responses has already been extensively studied over the last few decades, but more recently several relatively mono- or oligo-clonal [i.e., (semi-) invariant] innate T cell subsets received substantial interest in tumor immunology including invariant natural killer T (iNKT), γδ-T and mucosal associated invariant T (MAIT) cells. These subsets produce high levels of various pro- and/or anti-inflammatory cytokines/chemokines reflecting their capacity to suppress or stimulate immune responses. Survival of patients with cancer has been linked to the frequencies and activation status of NK, iNKT, and γδ-T cells. It has become clear that NK, iNKT, γδ-T as well as MAIT cells all have physiological roles in anti-tumor responses, which emphasize their possible relevance for tumor immunotherapy. A variety of clinical trials has focused on manipulating NK, iNKT, and γδ-T cell functions as a cancer immunotherapeutic approach demonstrating their safety and potential for achieving beneficial therapeutic effects, while the exploration of MAIT cell related therapies is still in its infancy. Current issues limiting the full therapeutic potential of these innate cell subsets appear to be related to defects and suppressive properties of these subsets that, with the right stimulus, might be reversed. In general, how innate lymphocytes are activated appears to control their subsequent abilities and consequent impact on adaptive immunity. Controlling these potent regulators and mediators of the immune system should enable their protective roles to dominate and their deleterious potential (in the specific context of cancer) to be mitigated.

Keywords: NKT, iNKT cells, CD1d, MAIT cells, gamma-delta T cells, NK cells, cancer immunotherapy

# INTRODUCTION

The importance of the immune system in tumor control and development has been extensively studied and it has been shown that different elements of the innate and adaptive immune system can exhibit anti-tumor activity. Adaptive immune cells are antigen-specific and have enhanced responses to subsequent antigen exposure. Innate-like or semi-invariant T cell subsets can recruit adaptive responses and thereby support eradication of tumor cells. It has become more and more apparent that besides conventional B and T cells and classical NK lymphocytes, other conserved innate T cells, such as natural killer T cells (NKT), γδ T cells and mucosa associated invariant (MAIT) cells, are of great importance in controlling tumor growth. Compared with conventional T cells, these innate T cell subsets are characterized by a limited (γδ T cells) or even (semi)-invariant (iNKT cell populations and MAIT cells) T cell receptor (TCR) repertoire and can have a dual role in tumor immunity. On one hand, they can stimulate or even directly mediate antitumor responses, but on the other hand their regulatory functions may hamper tumor eradication. A deeper understanding of the roles of classical NK cells and these innate T cell subsets in tumor immune biology, has led to new therapeutic options for cancer, whereby manipulation of these invariant subsets has already shown early signs of promising anti-tumor efficacy.

In this Review, we will briefly introduce and then outline our current understanding of the functions and potential of the classical innate NK cells and several semi-invariant subsets of innate immune T cells, and highlight their role in controlling anti-tumor immune responses as well as their therapeutic potential.

#### INNATE LYMPHOCYTE SUBSETS IN NATURAL AND THERAPEUTIC ANTI-TUMOR IMMUNITY

#### NK Cells

Natural killer cells (NK) comprise a classical innate lymphoid cell subset that plays an important role in the defense against infections and cancer (1). NK possess potent cytolytic activity to rapidly kill targeted cells (e.g., virally infected or malignant) and secrete various effector cytokines and chemokines like IFNγ, TNFα, GM-CSF, MIP-1α (CCL3), and RANTES (CCL5) (2, 3) (**Figure 1**). Because of this variety in secreted cytokines, NK activity is also important for proper function of other innate immune subsets such as DCs and macrophages (4, 5). But also in adaptive immune responses, such as cytokine secretion of T and B cells, NK cells seem to have an important contribution (6–9) and therefore NK activation can support tumor specific immune responses.

NK activation is based on the balance between inhibitory and activating signals from various receptors based on "missingself " and "induced-self " ligand interactions. Most important activating receptors are the natural cytotoxicity receptors (NCRs) NKp46, NKp30, NKp44, the C-type lectin NKG2D, the FcR CD16 and some killer cell immunoglobulin-like receptors (KIRs), while the inhibitory receptors include CD94/NKG2A/B and KIR-2DL and KIR-3DL (10). The feature of "missing-self " recognition is based on the situation in which expression of NK-inhibitory MHC-I molecules in the steady state dominates over the expression of NK activating ligands, thereby leaving NK inactive. In contrast, the increased expression of "induced-self " ligands on malignant cells in combination with reduced levels of MHC-I, leads to strong NK triggering and the induction of potent cytolytic activity.

The finding that MHC-I deficient syngeneic tumors were selectively rejected by NK and that the detection of the absence of MHC-I was mediated via inhibitory receptors on NK (10–13) has led to the discovery of multiple indications in which NK are involved in tumor eradication. Different mouse studies using transplanted syngeneic tumor cells showed that either genetic or antibody mediated NK depletion led to increased tumor growth and higher metastasis rates (14–17). Tumor outgrowth could be inhibited through addition of various cytokines that enhance NK activity. Models using methylcholanthrene (MCA) for chemical induction of tumors in combination with NK depletion demonstrated a role for NK, much like iNKT, in immune surveillance at early stages of tumor development (18). Mice deficient for important NK effector molecules such as perforin, IFNγ and the downstream signaling molecule of the IFNγ receptor, STAT1, developed tumors in higher frequencies than WT mice (1, 19). More sophisticated models using RAG2/γc deficient mice, which lack all lymphocytes including NK, iNKT, γδ T, classical CD4<sup>+</sup> and CD8<sup>+</sup> αβ T cells and B cells, showed a higher incidence of tumor growth compared to RAG2 deficient mice alone (which lack αβ T cells and B cells) demonstrating that indeed NK cells are in part responsible for inhibiting tumor growth (20). However, caution is called for the interpretation of these data since these models did not exclusively eliminate NK.

Also in human tumors correlative analyses have indicated a role for NK in tumor elimination. In cancer patients different NK deficiencies and dysfunctionalities have been observed (1, 2, 21–24) and an 11 year follow-up study highlighted that NK function can be a good indicator for cancer development and progression (25). In addition, multiple groups have reported that high levels of tumor infiltrating NK cells (TINKs) represents a favorable outcome for patients with different types of carcinomas and therefore NK cell infiltration appears to be a positive prognostic marker which may also respond to IL-12 (26–28).

Tumors can develop different strategies to evade NK mediated lysis (29) (**Figure 1B**). For example in acute myeloid leukemia, leukemic cells could induce loss or decrease of NCR expression on NK (30, 31) and this phenotype was correlated with a decreased killing capacity (31). Another mechanism by which tumors evade immune surveillance by NK is upregulation of classical and non-classical MHC-I molecules that reduce NK activity by delivery of inhibitory signals (32–34) (**Figure 1B**). Also specific manipulation of NKG2D signaling has been observed in tumors and can explain why in some cases the presence of NKG2D ligands is not sufficient for tumor clearance, but rather promotes tumor growth through NK cell immune

FIGURE 1 | The good and bad of (semi invariant) innate cells in cancer. (A) Overview of anti-tumor responses of NK, iNKT, γδ, and MAIT cells. Activated iNKT can directly kill tumor cells and promote DC triggering which is marked by up regulation of co-stimulatory molecules and enhanced cross-presentation capacities of DCs. iNKT can also directly promote effector T cell activation and differentiation and stimulate γδ mediated anti-tumor responses by secretion of different cytokines. Indirectly, iNKT also support activation of NK cells via IL-12 release of DCs, thereby enhancing anti-tumor effector functions. Expression of MICA/B and ULBP proteins on tumor cells induces activation of both γδ T and NK cells. As a result γδ T cells and NK release different pro-inflammatory cytokines for immune support and are also capable of directly killing malignant cells. Loss of expression of MHC-I molecules serves as another NK activating trigger, leading to perforin release and tumor cell eradication. As well as NK, iNKT, and γδ T cells, tumor infiltrating MAIT cells could also secrete different pro-inflammatory cytokines and potentially kill cancerous cells. (B) Potential tumor-promoting functions of NK, iNKT, γδ, and MAIT cells. Tumor cells possess different mechanisms to escape/manipulate NK cells, leaving NK unable to lyse malignant cells. NK also secrete immune suppressive and angiogenesis stimulating cytokines, which promote tumor growth. As well as functional defects of iNKT, which are marked by decreased IFNγ/IL-4 ratio, cancer cells can also skew iNKT function via secretion of lysophosphatidylcholine (LPC), resulting in IL-13 production by iNKT. This induces production of immuno-suppressive cytokines by MDSC. Also release of IL-17 can promote tumor growth. γδ and MAIT cells and a minor population of NKT cells can also release IL-17, which can inhibit Th1-type responses.

subversion. Normally expression of NKG2D ligands such as the MHC class I chain-related molecules (MIC) A/B and members of the UL-16 binding protein (ULBP) family leads to activation on NK and in patients with colorectal carcinoma expression of MICA even correlates with good prognosis (35). However, tumor cells can release soluble forms of NKG2D ligands and elevated levels of MICA/B and ULBP2 have been detected in sera of patients with various epithelial and hematopoietic malignancies (36–41) (**Figure 1B**). Soluble NKG2D ligands can abrogate NK activation and down regulate and block NKG2D on tumor infiltrating lymphocytes (37, 38, 42). Recently it has become clear that also aberrant glycosylation on tumor cells affects NK activity. Jandus et al. demonstrated that sialic acid containing carbohydrates on tumor cells serve as ligands for the siglec 7/9 receptors on NK and interfere with NK mediated anti-tumor responses (43). This hypothesis is supported by findings that enzymatic induction of high sialylation on tumors dampens activity of NK (44) (**Figure 1B**).

Different studies therefore indicate that NK functions can be turned off in the presence of a tumor, but the coin doesn't always flip from "good" to "inactive" but rather flips to "bad", leaving NK pro-tumorigenic. This has e.g., been described by Bruno et al. who identified NK cells in patients with non-small cell lung cancer that produced substantial levels of vascular endothelial growth factor (VEGF), placental growth factor (PIGF) and IL-8 and therefore might stimulate angiogenesis to enhance tumor growth (45, 46) and actively suppress immune responses (47) (**Figure 1B**).

Therapeutically, it is a challenge to overcome the escape mechanisms that tumors have developed to prevent NK killing and to reverse NK paralysis. Many different strategies are currently tested and some show promising results in preclinical and increasingly clinical studies (10). Most studies have focused on adoptive transfer of autologous, allogeneic or NK cell lines to enhance NK cytotoxicity against tumors. All three approaches show anti-tumor activity but with various efficacy (48–53). Indeed, allogeneic hemopoietic stem cell transplant antileukemia effects are partly mediated by NK cells (49–53). Other promising approaches include blockade of inhibitory receptors on NK using mAb which could recover effective NK mediated killing activity (54–56) and chimeric antigen receptor (CAR) technology that after extensive testing in T cells, have also been applied to NK cells and might constitute a promising approach to enhance NK cell mediated anti-tumor responses (51–53, 57–59). Indeed, autologous CAR-NK might be one way to avoid issues of contaminating allogeneic T cells whilst augmenting the NK activity specifically against the tumor, where appropriate CAR targets are available.

To conclude, the relevance of NK in tumor immune responses has been revealed in many studies. However, immune editing of the tumor and immune suppression perpetrated by the tumor can abrogate NK function limiting NK mediated lysis of tumor cells. More insight in the exact contribution of NK cells in tumor progression and ways to overcome NK paralysis is warranted to optimize NK activating therapies.

#### NKT Cell Populations

There are 2 major populations of CD1d-restricted "NKT" cells (T cells sharing some NK phenotypic and functional properties): The better-known "Invariant natural killer T cells" (iNKT) and polyclonal diverse "non-invariant" NKT cells (60, 61). iNKT are a subset of lymphocytes with a significant role in regulating immune responses, including immune surveillance against tumors. iNKT recognize lipid antigens presented by the monomorphic MHC-like molecule CD1d, predominantly expressed by dendritic cells (DC) and other antigen presenting cells (APC). iNKT were initially identified by their restricted TCR repertoire (Vα14Jα18 in mice and Vα24Jα18 in humans), but subsets expressing variable TCRs do also exist. The basis of the regulatory function of iNKT appears to lie in their capacity for rapid secretion of multiple cytokines upon TCR triggering which is accompanied by an increased CD1d-restricted cytotoxic capacity (60). Cytokines released by iNKT include both regulatory cytokines (e.g., IL-4, IL-10, IL-13) as well as proinflammatory cytokines such as IL-2, IL-17, and IFNγ, reflecting their capacity to suppress or stimulate immune responses (61, 62). iNKT cell cytotoxic activity can be mediated by classical granule-mediated mechanisms, although Fas / FasL dependent killing has also be reported (60, 61) (**Figure 1A**).

iNKT recognize different microbial and endogenous antigens such as gangliosides and glycolipids and therefore play a substantial role during infection (63). However the compound most efficient for activating iNKT is the marine sponge derived glycolipid α-galactosylceramide (αGalCer). Ever since the identification of αGalCer as prototypic high-affinity CD1d binding lipid and potent iNKT stimulant, studies have shown that iNKT activation with αGalCer promotes tumor rejection and protects from the development of metastases in multiple murine tumor models (64–67). This anti-tumor effect could be further improved by injection of αGalCer-pulsed DCs and antimetastatic effects were shown to be driven by IFNγ (68–70). Furthermore, IL-12, a master regulator of Th1 responses, like αGalCer, drives the anti-metastatic activity of T cells, including iNKT, as well as NK cells and effects of low dose IL-12 treatment in murine tumor models can be predominantly mediated by the activity of iNKT (66, 71–74).

Whereas multiple studies have shown the critical role of iNKT in the induction of potent anti-tumor responses in response to stimulation by the above-mentioned exogenous factors such as αGalCer and IL-12, the physiological role of these cells in tumor immunity remains more elusive. However, Smyth et al indicated that, at least in a model of MCA-induced fibrosarcomas, iNKT fulfill an essential role in tumor immune surveillance. Adoptive transfer of iNKT from wild type mice into iNKT cell deficient mice (Jα18 –/–) clearly showed a protective effect on tumor outgrowth without a requirement of additional exogenous stimuli (75). The contribution of iNKT cells to immune surveillance has also been highlighted by findings on their capacity to mature DC and subsequently activate NK and cytotoxic CD8<sup>+</sup> T cells, the latter two of which then become potent cytotoxic cells. Upon recognition of CD1d:lipid complexes and the costimulatory molecules CD80/86 on the surface of DCs, iNKT cells up-regulate the IL-12R (66, 71–74) and CD40L molecule. Subsequently, and mediated by CD40L, iNKT induce DC maturation and release of IL-12. This IL-12 release in turn potently increases IFNγ production by iNKT which then, together with enhanced cross-presentation of DCs after iNKT induced maturation, boosts activation of anti-tumorigenic cytotoxic T lymphocytes (CTL) (76, 77) (**Figure 1A**). In other words, iNKT have the capacity to jump-start immune responses and together with DCs to bridge the innate and adaptive immune systems.

Besides providing a pro-inflammatory status by interaction with DCs, NK and CTL, iNKT have also been found to be able to control tumor growth by killing tumor supportive IL-6-producing CD1d<sup>+</sup> CD68<sup>+</sup> tumor associated macrophages (TAM) (78). Moreover, iNKT could potentially also control myeloid derived suppressor cells (MDSC) in the tumor microenvironment (TME) (79, 80). Absence of iNKT in mice infected with influenza virus resulted in strong expansion of MDSC, but interestingly adoptive transfer of iNKT could abolish suppressive activity of MDSCs. So, by targeting TAM and MDSC, iNKT may skew the TME to a pro-immune milieu.

While the function of iNKT as regulators of immune responses has been widely acknowledged (81, 82), the exact mechanisms polarizing iNKT effector functions remain elusive, thusfar in part limiting their therapeutic potential in clinical trials. Studies in multiple human cancers have revealed selective numerical and/or functional defects in the iNKT cell population. Decreased numbers of circulating iNKT have been found in multiple tumor types such as advanced prostate cancer and are accompanied by decreased IFNγ production and increased IL-4 production by iNKT (83–85). Functional defects of iNKT have been found in human multiple myeloma where development from non-progressive or premalignant gammopathy to progressive disease was marked by a strong decrease in IFNγ producing iNKT in patient blood. However, this functional defect could be reversed by using αGalCer-pulsed matured dendritic cells (DCs) (86).

More functional iNKT defects have been described in the TRAMP prostate cancer model (TRransgenic Adenoma carcinoma of the Mouse Prostate), similar to the functional iNKT defects found in some human malignancies (87). In this model, iNKT were attracted by tumor cells to migrate into prostate tumors mediated through the CCL2-CCR5 axis. Interestingly, these primary prostate tumors as well as mouse and human prostate cancer cell lines and human prostate epithelium can express CD1d, permitting direct interaction with iNKT. Indeed, prostate tumor cells induce selective production of Th2 cytokines by iNKT and thereby bias iNKT effector functions. Interestingly, this aberrant iNKT activation was reversible by the simultaneous addition of αGalCer and IL-12, which allowed iNKT cells to produce IFNγ in response to these CD1d-expressing prostate cancer cells. Restoration of iNKT cell functions by addition of IL-12 with an agonistic CD1d ligand provides the first of several complementary novel approaches for overcoming iNKT defects in malignancy. Besides active skewing of iNKT function via interaction with CD1d on tumor cells, some tumors escape from iNKT cell lysis by loss of CD1d expression and shedding of glycolipids, such as gangliotriaosylceramide which can inhibit iNKT stimulation (88). In addition, it has been shown that iNKT can acquire suppressive functions of regulatory T cells, which is marked by nuclear expression of FoxP3 (89). Finally, an interesting example of the potential complexity of the interactions involving iNKT cells as with other immune components has been recently reported (90). Gut microbiome produced bile acids metabolites positively influenced iNKT cell accumulation and anti-tumor activity in the liver via activating liver sinusoidal endothelial cells to express iNKT chemoattractant CXCL16 (90).

As well as the protective roles for iNKT in cancer and the above-mentioned studies documenting tumor-induced alterations of iNKT cell functions, other studies have also found that some CD1d-restricted NKT cells can suppress anti-tumor responses through regulatory cytokine(s) (91, 92). These "noninvariant" NKT subsets, which are characterized by a diverse TCR repertoire are mostly referred to as type II NKT and can produce high levels of IL-13 through the IL4R/STAT6 pathway, thus promoting tumor recurrence (92) (**Figure 1B**). Based on these findings, Terabe et al. proposed a model in which type II NKT are responsible for downregulating tumor immunity, while type I NKT, as described above, are responsible for tumor protection (93). Additional studies reported that myelomaderived lysophosphatidylcholine (LPC) could induce secretion of IL-13 by a small Vα24−Vβ11<sup>−</sup> subset of NKT (94). Moreover, as IL-13 can induce production of the immunosuppressive cytokine TGF-β by MDSC (93), these data support the hypothesis of NKT driven immune suppression (**Figure 1B**). Together, these findings suggest a suppressive role of non-invariant NKT that could be driven by IL-13. However, our knowledge about type II NKT cells is still limited, in part due to a lack of specific markers for this subset (only CD1d tetramers with sub-optimal ligands are available) and their application is hampered by limited knowledge of glycolipid antigens specific for type II NKT. Therefore future studies are warranted to further specify the complete roles of type II NKT.

Thus far, clinical studies have mainly focused on adoptive transfer of autologous iNKT cell enriched in vitro-expanded populations from peripheral blood mononuclear cells (PBMCs), αGalCer-pulsed monocyte-derived DCs or a combination of activated iNKT and αGalCer pulsed DCs (95–99). Also administration of soluble αGalCer has been tested in clinical trials (98, 99). Currently, clinical benefits are still relatively limited and combinations as well as optimized strategies are being considered (96, 99). Since ex vivo expansion of circulating iNKT has to overcome their low frequencies in blood, induced pluripotent stem cells (iPSCs) for the generation of large numbers of iNKT might provide an alternative (100). Furthermore, a general problem with current approaches might be that although iNKT are systemically activated, their accumulation to the tumor site is not guaranteed. Targeting iNKT to the tumor microenvironment using bi-specific targeting could enhance trafficking to tumor sites and therefore increase the total antitumor response (101). The use of chimeric antigen receptors (CARs), which combine the targeting effect of antibodies to decrease off-target effects with the potent anti-tumor effector functions of iNKT, has been shown to be promising in preclinical studies and has already shown protection targeting GD2 for metastatic neuroblastoma in mice (102, 103).

As detailed above, "Type" I (invariant) NKT possess potent cytotoxic activity against cancer cells, but numerical and functional defects are limiting their full potential. Altogether, it seems that type I NKT dysfunction in cancer may be caused by acquired capacities of tumor cells to immobilize the iNKT arm of anti-tumor defense. Thereby the putative role of iNKT in immune surveillance seems to be extended toward a more controlling role in behavior of cancer cells. On the other hand, non-invariant/diverse NKT subsets ("Type II NKT") can actively downregulate tumor immunity through different mechanisms (91–94). In the future, a more complete and evolving understanding of reversible type I NKT defects together with more insight in the mechanism behind type II NKT cell mediated suppression of antitumor immune responses (or other activities of these less understood and more diverse populations), should help the development and evaluation of novel and successful cancer therapies involving NKT populations (99, 103).

# Gamma-Delta (γδ-) T Cells

γδ-T cells belong to the family of unconventional T cells and differ from conventional αβ T cells, in that most γδ T cells lack expression of the CD4 and CD8 co-receptors. Intriguingly antigen recognition by the γδ TCR is not restricted to MHCclass I and II molecules (104). In humans, 0.5–16% of all CD3<sup>+</sup> cells in peripheral blood and lymphoid tissues is represented by γδ T cells (105, 106). In mice, this percentage varies between 1 and 4% (107). Human γδ-T cells can be divided into two major subsets based on expression of the variable regions of TCR-δ; Vδ1, or Vδ2 (108, 109). Vδ2 cells constitute the most prominent subset in human peripheral blood and are almost always paired with Vγ9 <sup>+</sup> (Vγ9Vδ2) while Vδ1 are more prominent at mucosal areas (110–112). γδ T cells recognize multiple self and non-self-antigens like phospholipids, small proteins and also non-peptidic antigens, so-called pyro-phospho-antigens (pAg), either in complex with butyrophilin 3A1 (BTN3A1, CD277) or effecting a conformational change in BTN3A1/CD277 which in turn leads to Vγ9Vδ2-T cell recognition (113–116). pAgs such as (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMBPP) are not only produced by bacteria, but can also be produced by tumor cells with a relatively high metabolic activity of the mevalonate metabolic pathway resulting in the accumulation of pAg intermediates such as isopentenyl pyrophosphate (IPP)) (114, 117). Vγ9Vδ2 TCR mediated recognition of accumulated pAgs in tumor cells is mediated by BTN3A1/CD277 and results in strong activation and expansion of Vγ9Vδ2-T cells which is marked by the release of multiple pro-inflammatory cytokines including IFNγ, TNF-α and/or interleukin-17 (IL-17 seems to be mostly produced by Vδ1 <sup>+</sup> cells and it is not just a proinflammatory antitumor cytokine) and a strong anti-tumor response (109, 118–120) (**Figure 1**).

Besides activation of Vγ9Vδ2-T cells via TCR ligation, engagement of the natural killer cell receptor NKG2D contributes to the anti-tumor reactivity of Vγ9Vδ2 T and Vδ1 <sup>+</sup> T cells. This is especially interesting since NKG2D can bind stress- or infection induced ligands of the non-classical MHC-I related molecules H60, RAE1, and MULT in mice or MIC-A/B and ULBP1-ULBP6 in humans and while these molecules are absent on healthy cells, they are often expressed by tumor cells (121–124) (**Figure 1A**). Expression of ULBP molecules has been found in multiple types of cancer (leukemia, lymphoma, ovarian and colon carcinomas and hematological malignancies) and can therefore determine susceptibility to Vγ9Vδ2-T cell mediated cytolysis (125–127). Vδ1 <sup>+</sup> T cells not only recognize the stress induced self-antigens MICA/B via NKG2D but can also directly bind MIC molecules via their TCR (128, 129). Interestingly, enhanced expression of MICA/B by oxidative stress on tumor cells has been correlated to an increased frequency of Vδ1 <sup>+</sup> T cells among tumor infiltrating lymphocytes (TIL) (130).

The involvement of γδ-T cells in the elimination of tumors is at least partly based on their ability to interact with different cell types. Besides offering B cell help and triggering of DC maturation (131), Vγ9Vδ2 T cells show characteristics of antigen presenting cells, including the processing and presentation of antigens which allow the induction of naïve αβ T cell proliferation and differentiation (132, 133). This hypothesis has been further expanded by findings that Vγ9Vδ2-T cells, via trogocytosis of CD1d, can function as platform to activate iNKT in a CD1d-restricted manner (134). Since Vγ9Vδ2-T cells have the capacity to interact with different immune cells, they are important for both innate and adaptive anti-tumor responses.

The ability of γδ-T cells to generate huge amounts of pro-inflammatory cytokines, to recognize cell stress via an MHC independent mechanism, to potentiate other immune cell components, both innate and adaptive, and directly mediate cytolysis of multiple tumor types, potentially make γδ-T cells key players in anti-tumor immune responses and as such attractive therapeutic targets.

The potential impact of γδ-T cells on cancer immunotherapy has been reported in multiple studies showing γδ-T cells to be able to recognize and kill multiple different tumor types in vitro including leukemia, numerous carcinomas and neuroblastoma (125, 135–137). Several clinical trials have been conducted using aminobisphosphonates such as zoledronic acid (Zol) to manipulate intracellular levels of IPP (138–140). Administration of a combination of Zol with low dose IL-2 to patients with metastatic breast cancer or prostate cancer was well tolerated and increased peripheral blood Vγ9Vδ2-T cell numbers, which correlated with clinical outcome (141). In addition, synthetic pAgs, such as BrHPP have been tested in clinical trials and been shown to increase recognition of different tumor cells by Vγ9Vδ2-T cells (108). Interestingly, treatment with common chemotherapeutic compounds (e.g., temozolomide) has been shown to increase expression of stress associated NKG2D ligands on tumor cells, thereby possibly sensitizing tumor cells for Vγ9Vδ2-T recognition and opening windows for Vγ9Vδ2-T based immunotherapies (142).

While multiple studies have shown that γδ-T cells exhibit anti-tumor activity, the potential involvement of γδ-T cells in tumor progression remains rather elusive. Recently mouse and human studies emphasized a pro-tumorigenic activity of IL-17 producing and regulatory γδ-T cells (γδ T17/γδ1 Tregs) (**Figure 1B**). Whereas Ma et al. reported on the contribution of IL-17 producing γδ-T cells to the efficacy of anticancer chemotherapies (143), other reports showed an inverse correlation between γδ-T17 cells and overall survival, suggesting immune suppressive and tumor promoting properties of γδ-T cells by promoting accumulation of MDSCs and angiogenesis respectively (144, 145). In a transplantable model of peritoneal and ovarian cancer, γδ T17 (Vγ6 <sup>+</sup>) cells were shown to preferentially produce IL-17 instead of IFNγ and to promote tumor growth (146). Interestingly, a reduction in tumor size was observed in TCRδ and IL-17 deficient mice compared to wild type mice, further suggesting a pivotal role of γδ T17 cells in cancer progression in this model.

Although Vγ6 <sup>+</sup> cells do not exist in humans, enriched amounts of Vδ1 <sup>+</sup> T cells with regulatory properties (γδ1-Tregs) have been identified in TIL of patients with breast cancer (128, 147). These γδ1-Tregs can suppress naïve and effector T cell responses and concordantly block maturation of dendritic cells. A more detailed in-depth study on the correlation of breast cancer TIL phenotypes with clinical outcome revealed that infiltration of γδ1-Tregs was correlated to poor prognosis (148). Together these findings imply a critical role of some γδ-T cell subsets as immune suppressors and emphasize the need for more detailed studies to better understand their regulatory functions in order to ultimately design effective innate γδ-T cell based therapeutic strategies.

Although a lot of effort has been put in understanding γδ-T cell function in tumor immunity, it is becoming clear that the overall impact of γδ-T cells in cancer treatment may depend on the fine balance between anti- and pro-tumorigenic subsets. Current challenges to optimize anti-cancer therapies lie in the quest to determine how γδ-T cell mediated anti-tumor properties can be selectively boosted, while at the same time their suppressive activity is inhibited.

### MAIT Cells

Mucosal associated invariant T (MAIT) cells belong to another discrete subpopulation of T cells that is characterized by a limited TCR repertoire. Most human MAIT cells express Vα7.2- Jα33 and like in iNKT, the invariant α chain is paired with a limited diversity of TCRβ chains (Vß2 and Vß13) (149). CD161 is abundantly expressed on MAIT cells and they are highly sensitive to IL-12 and IL-18 stimulation due to their expression of IL-12R and IL-18R (150). MAIT cells recognize a variety of antigens, including bacterial and fungal derivates and metabolites of vitamin B2 (riboflavin) and B9 (folate), presented by the invariant MHC related 1 molecule (MR1) and they appear to represent important players in antimicrobial immunity (151). Their preferential location of Vα7.2-Jα33 cells is in mucosal tissue such as the gut lamina propria, but MAIT cells are also relatively abundant in peripheral blood and liver (152–154). Compared to iNKT they represent a relatively abundant cell population, with 1–4% of total TCR-αβ<sup>+</sup> T cells (154). MAIT cells can secrete multiple cytokines such as IFNγ, IL-17, and TNF-α and possess lytic activity through the release of granzyme B upon activation (155, 156) (**Figure 1**). Since both the Th1 skewing cytokine IFNγ and the Th17 characterizing cytokine IL-17 are secreted by MAIT cells, these cells might be of great importance in the induction of either advantageous or deleterious immune responses in terms of cancer control (**Figure 1**).

In the past decade most publications on the roles of MAIT cells have focused on protection against infectious pathogens and in some auto immune related disorders, whereas information about their involvement in cancer immunity is relatively scarce. However, findings that accumulated MAIT cells appear to have a protective role in inflammatory bowel diseases (IBD) in humans (152) and the fact that TIL-induced intestinal inflammation present in colorectal cancer (CRC) can alter the prognosis of patients with CRC (157), suggests that intestinal MAIT cells can infiltrate into CRC tumor sites and fulfill a protective function, like in IBD. Indeed, multiple studies have recently reported active accumulation of MAIT cells in CRC while circulating activated and memory MAIT cell numbers were decreased, suggesting active homing to tumor sites (158, 159). Infiltration of MAIT cells in patients with glioblastoma and renal carcinoma in previous reports support this homing and tumor infiltrating capacity (160). Circulating MAIT cells in patients with progressive disease were significantly lower than in early stage CRC patients (158). Although it was shown that tumor infiltrating MAIT cells produced lower levels of IFNγ (and relatively high amounts of IL-17) compared to unaffected colon tissue and that this decrease was independent of lowered expression of MR1 on tumor cells (158, 161), the exact factors in the tumor microenvironment hampering antitumor effector cytokine secretion still remain elusive. Similar effects seem to apply to suppression of function of MAIT cells in CRC metastases to the liver (162). Finally, such defects may be common to a wide variety of cancers (163), since their numbers and activity are also reduced in myeloma patients, although which came first: the defects or the cancer, was a question raised by the finding that carefully age matched (generally older) people have reduced MAIT cells (164).

Until now, little is known about MR1 distribution and it has yet to be elucidated whether MR1 expression on tumor cells could be important in MAIT cell activation. Furthermore, in order to better understand MAIT cell interactions in neoplasms and to exploit MAIT cells for immune therapies, more detailed studies on new ligands and ligand driven expansion are urgently needed.

# CONCLUSIONS: COMBINING INNATE IMMUNE THERAPIES WITH CHEMO- AND OTHER THERAPIES

Innate immune cells may represent the first line of defense against malignancies, e.g., through the MHC-independent recognition of their metabolically stressed state, and their potency as regulators and mediators of tumor immune responses, both innate and adaptive, has been widely acknowledged as discussed above (**Figure 1**). Therefore classical NK cells and the different innate (semi)-invariant T cell subsets have garnered interest in the field of anti-cancer immunotherapies and multiple NK, iNKT, and γδ-T based immunotherapeutic approaches are currently clinically tested. Although these therapies show some promising results, overall clinical benefits are still limited and the explored strategies need to be optimized. Identification of mechanisms underlying NK, iNKT, γδ-T, and MAIT cells defects, which have been observed in patients with cancer, could fuel the development of alternative approaches to current treatments. For example, in the case of immune editing to evade NK effector function, blockade of inhibitory receptors on NK cells to overcome NK paralysis has made its first steps in clinical trials. Overall, defined molecular mechanisms and interactions of tumor cells with immune cells in the tumor microenvironment need to be further investigated in order to understand how local immune suppression of effector cells can be overcome.

There is a general consensus that in order for immunotherapy to be fully effective combinatorial therapies need to be developed and clinically tested. Multiple studies have indicated that low-dose chemotherapeutics can reduce local immune suppression by, for example elimination of MDSC in the tumor microenvironment (165–167), and therefore increase efficacy of already applied immune therapies. Since IL-13 producing type II NKT have been associated with immune suppression mediated via MDSC-derived TGF-β, combining NKT based therapies with chemotherapeutics might reverse the flavor of NKT in this case from "bad" to "good" in terms of cancer control. Therefore, combining immune therapy with chemo-therapy could perhaps also benefit innate immunity based anti-tumor therapies in certain circumstances. Moreover, the immune system comprises many elements which are tightly regulated and connected and as such cross-talk of innate effector cells with each other or with other immune cells could be exploited to enhance the efficacy of current therapies. For example, close crosstalk between iNKTs and Vγ9Vδ2-T cells has been described, in such a way that αGalCer activated iNKT to enhance CD25 expression and IFNγ production of γδ-T cells via secretion of TNF-α (168). The findings that iNKT can potentiate antitumor effector functions of Vγ9Vδ2-T cells and that Vγ9Vδ2-T cells possess unique features to activate iNKT, opens up new avenues to strengthen future iNKT and Vγ9Vδ2 T cell based immunotherapeutic approaches. An alternative approach for combinatorial therapy could be to enhance the interaction of iNKT with DCs using vehicles/vaccines to target both types of cells in order to maximize anti-tumor effects. Indeed, an OVA peptide/CpG vaccine combined with recombinant α-galactosylceramide (αGC)-loaded CD1d-anti-HER2 fusion protein showed increased expansion of OVA-specific CTLs and was likely mediated via maturation of DCs (169).

Nowadays, great advances have been made in the available detection methods for monitoring immune cells in tumors using mass cytometry. Also next generation (single cell) sequencing of (invariant) T cells has proven itself to be helpful for the identification of new invariant T cells (170). These big data approaches facilitate identification and detailed analysis of immune cells and their plasticity in malignancies

#### REFERENCES


and will hopefully contribute to a better understanding of dualistic roles of innate cells in cancer control and progression.

In conclusion, innate immune effector (NK/T) lymphocyte subsets are key in regulating cancer control versus progression. If present hurdles can be overcome and the fine line between their suppression or progression of tumor growth has been further elucidated, NK cells, iNKT, γδ T, and MAIT cells hold great promise for the induction of long lasting anti-tumor immunity.

## AUTHOR CONTRIBUTIONS

ME received and accepted the commission, recruited the coauthors, and edited the manuscript. TG, YK and HV edited the manuscript. DS drafted and edited the manuscript and drew the figure.

# FUNDING

ME was funded by AgenTus which is a company developing cancer immunotherapies. HV is CSO/CMO of Lava Therapeutics, a company developing cancer immunotherapies.


and correlation with soluble MIC. Cancer Immunol Immunother. (2006) 55:1584–9. doi: 10.1007/s00262-006-0167-1


melanoma: a Phase I clinical trial. Clin Cancer Res. (2017) 23:3510–19. doi: 10.1158/1078-0432.CCR-16-0600


molecules on cell surface. Biochem Biophys Res Commun. (2007) 354:613–8. doi: 10.1016/j.bbrc.2007.01.031


bowel diseases. Clin Exp Immunol. (2014) 176:266–74. doi: 10.1111/cei. 12277


functions through the production of TNF-α. Clin Immunol. (2012) 142:194– 200. doi: 10.1016/j.clim.2011.10.006


**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 Stolk, van der Vliet, de Gruijl, van Kooyk and Exley. 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.

# Complex Network of NKT Cell Subsets Controls Immune Homeostasis in Liver and Gut

Idania Marrero1,2, Igor Maricic1,2, Ariel E. Feldstein<sup>3</sup> , Rohit Loomba<sup>2</sup> , Bernd Schnabl <sup>2</sup> , Jesus Rivera-Nieves <sup>2</sup> , Lars Eckmann<sup>2</sup> and Vipin Kumar 1,2 \*

*<sup>1</sup> Laboratory of Immune Regulation, University of California, San Diego, La Jolla, CA, United States, <sup>2</sup> Division of Gastroenterology and Hepatology, Department of Medicine, University of California, San Diego, La Jolla, CA, United States, <sup>3</sup> Department of Pediatrics, University of California, San Diego, La Jolla, CA, United States*

#### Edited by:

*Luc Van Kaer, Vanderbilt University, United States*

#### Reviewed by:

*Marika Falcone, San Raffaele Hospital (IRCCS), Italy Yifang Gao, First Affiliated Hospital of Sun Yat-sen University, China Shintaro Sagami, Kitasato University Kitasato Institute Hospital, Japan*

> \*Correspondence: *Vipin Kumar vckumar@ucsd.edu*

#### Specialty section:

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

Received: *15 May 2018* Accepted: *22 August 2018* Published: *11 September 2018*

#### Citation:

*Marrero I, Maricic I, Feldstein AE, Loomba R, Schnabl B, Rivera-Nieves J, Eckmann L and Kumar V (2018) Complex Network of NKT Cell Subsets Controls Immune Homeostasis in Liver and Gut. Front. Immunol. 9:2082. doi: 10.3389/fimmu.2018.02082* The liver-gut immune axis is enriched in several innate immune cells, including innate-like unconventional and adaptive T cells that are thought to be involved in the maintenance of tolerance to gut-derived antigens and, at the same time, enable effective immunity against microbes. Two subsets of lipid-reactive CD1d-restricted natural killer T (NKT) cells, invariant NKT (iNKT) and type II NKT cells present in both mice and humans. NKT cells play an important role in regulation of inflammation in the liver and gut due to their innate-like properties of rapid secretion of a myriad of pro-inflammatory and anti-inflammatory cytokines and their ability to influence other innate cells as well as adaptive T and B cells. Notably, a bi-directional interactive network between NKT cells and gut commensal microbiota plays a crucial role in this process. Here, we briefly review recent studies related to the cross-regulation of both NKT cell subsets and how their interactions with other immune cells and parenchymal cells, including hepatocytes and enterocytes, control inflammatory diseases in the liver, such as alcoholic and non-alcoholic steatohepatitis, as well as inflammation in the gut. Overwhelming experimental data suggest that while iNKT cells are pathogenic, type II NKT cells are protective in the liver. Since CD1d-dependent pathways are highly conserved from mice to humans, a detailed cellular and molecular understanding of these immune regulatory pathways will have major implications for the development of novel therapeutics against inflammatory diseases of liver and gut.

Keywords: CD1d, lipids, hepatitis, microbiota, epithelium

# INTRODUCTION

The liver is at the center of the interactions between the gut and the rest of the body and little is known about how cellular and molecular interactions in the gut-liver immune axis maintain homeostasis. On the one hand, through the portal circulation, the liver is the primary recipient of gut-derived metabolites and microbial products, and, on the other, the liver secretes products through the biliary system into the gut. In fact, there is a strong association between primary sclerosing cholangitis and inflammatory bowel disease (1, 2). Several factors, including dietary components, particularly fat and alcohol, mucosal damage, infections, medications and toxins, can disturb the intestinal barrier, leading to increased permeability and translocation of bacterial products or metabolites across the epithelial barrier into the portal circulation (3). Under inflammatory conditions, the gut-associated lymphatic tissue is stimulated by the increased influx of pathogen/microbeassociated molecular patterns to secrete pro-inflammatory cytokines (TNFα, IL-1, and IL-6), chemokines, and eicosanoids, all of which can reach the liver and stimulate local responses (4). In this pro-inflammatory environment, both liver parenchymal (hepatocytes) and non-parenchymal cells (intrahepatic lymphocytes, Kupffer cells, sinusoidal endothelial cells and hepatic stellate cells) secrete reactive oxygen species that can contribute to liver injury, inflammation and fibrosis. Thus, in the gut-liver microenvironment, multiple immune and non-immune cells form an interacting network to maintain immune tolerance. In this review, we mainly focus on the interactions between natural killer T (NKT) cell subsets and other innate and adaptive T cells in the gut-liver axis in controlling homeostasis and how activation of different subsets of NKT cells is involved in chronic inflammatory diseases.

#### iNKT AND TYPE II NKT CELL SUBSETS

Both liver and gut are enriched in innate immune cells, including resident macrophages, Kupffer cells, dendritic cells (DC), natural killer cells, and unconventional T cells (5, 6). Unconventional T cells are a diverse population, comprising NKT cells, γδ T cells, mucosal associated invariant T (MAIT) cells, and MHC class I<sup>b</sup> -restricted CD8 T cells. NKT cells are innate-like T cells that express antigen receptors and recognize both exogenous and endogenous lipid antigens presented by a class I MHC-like molecule, CD1d. Following antigenic activation, NKT cells are characterized by their ability to rapidly secrete large amounts of chemokines and cytokines, including IFNγ, TNFα, IL-4, IL-13, IL-17, IL-21, IL-22, and granulocyte-macrophage colonystimulating factor. These factors modulate immune responses triggered by other innate cells and adaptive T and B cells (7–11).

CD1d-restricted NKT cells exist as two main types based on their TCR usage and lipid recognition. Invariant NKT (iNKT) cells express a semi-invariant TCR consisting of TRAV11 TRAJ18 TCR-alpha chains paired with a limited number of TCR-β chains (TRBV13, TRBV29, or TRBV1) in mice or the orthologous TRAV10 TRAJ18 paired with TRBV25 in humans. Most iNKT cells are strongly reactive to the glycosphingolipid αgalactosylceramide (αGalCer) and are abundant in mice, but less frequent in humans (12). Similar to Th cell subsets, iNKT can be divided into subsets that are defined by their transcription factors and/or cytokines secreted, including iNKT1 (T-bet/IFNγ), iNKT2 (Gata-3/IL-4), iNKT10 (IL-10), and iNKT17 (Rorγt/IL-17) (13–15). Recent studies have indicated that iNKT cells can play a protective or a suppressive role in different diseases, such as microbial infections, chronic inflammation, autoimmunity, allergy, and cancer (16–20).

In contrast, type II NKT cells are not reactive to αGalCer, are more abundant than iNKT cells in humans and consist of CD1d-restricted T cells that express a diverse TCR repertoire but not the semi invariant TCR α-chain expressed by iNKT cells (12). Type II NKT cells can also recognize a variety of lipids antigens, including microbial and endogenous glycolipids and phospholipids as well as endogenous hydrophobic peptides (21). Usually, in comparison to the αGalCer/CD1d/TCR interactions, lipid antigens recognized by type II NKT cells, for example, sulfatides or lysophosphatidylcholine (LPC), binds with lower affinity to CD1d molecules and, accordingly, form relatively less stable tetrameric complexes (22–24), whereas αGalCer-loaded CD1d tetramers form stable complexes (25, 26). Consequently, characterization of type II NKT subsets and exploration of their functions have not progressed as rapidly as for iNKT cells. Nevertheless, most studies have so far suggested an immunosuppressive role of antigen-activated type II NKT cells in several experimental models [recently reviewed in (27)].

## CROSS-REGULATION BETWEEN iNKT AND TYPE II NKT CELLS

We have identified a major subset of murine CD1d-restricted type II NKT cells that can recognize sulfatides as well as LPC and express oligoclonal TCRs with a limited number of Vα- and Vβ-chains (Vα3/Vα1 and Vβ8.1/Vβ3.1) (23, 24, 28, 29). Crossregulation between iNKT and type II NKT cells has been shown by our laboratory and others in several models of autoimmunity and cancer (8, 11, 30–32). Thus, sulfatide- or LPC-mediated activation of type II NKT cells leads to activation of hepatic plasmacytoid DCs (pDCs) but not conventional DC (cDCs) through a mechanism that depends on IL-12 and macrophage inflammatory protein 2 (MIP2). This results in induction of anergic hepatic iNKT cells unable to secrete cytokines, which is accompanied by tolerization of cDCs and CNSresident microglia followed by inhibition of the conventional MHC-restricted Th1/Th17 CD4<sup>+</sup> T cells and protection from autoimmune diseases, such as EAE and diabetes as well as suppression of tumor surveillance (11, 29, 33, 34). Consistently, type II NKT cells mediated induction of anergy in iNKT cells and inhibition of conventional CD4+/CD8<sup>+</sup> T cells prevents immune-mediated liver diseases (24, 27, 30, 33, 35–37) (see **Figure 1**). In animal models of inflammatory bowel disease, NKT cells can be either protective or pathogenic (38). Interestingly, colonic type II NKT cells appear to play a pathogenic role in the context of increased CD1d expression or iNKT cell deficiency (38–42). Additionally, sulfatide-reactive IL-13Rα2 <sup>+</sup> type II NKT cells are abundant in the lamina propria of ulcerative colitis patients (43) and have recently been shown to be present in human liver (44). How cross-regulation between iNKT and type II NKT cells influences immunity in these compartments during health or disease is an important area of investigation. It was also shown that activation of type II NKT cells by IL-25 prevented high fat diet-induced obesity and transfer of these cells in obese mice improved weight loss and glucose tolerance (11, 31, 45). On the other hand, type II NKT cells can also promote liver inflammation and obesity in animals fed a highfat diet (46).These different findings indicate that the role of type II NKT cells in the regulation of adipose tissue inflammation and diet-induced obesity remains to be fully understood.

In contrast to a protective role of type II NKT cells, iNKT cells have been shown to promote liver injury contributing to

chronic liver diseases, including ischemic reperfusion injury, Concanavalin A-induced hepatitis, Lieber-DeCarli liquid alcohol diet-induced steatohepatitis and diet-induced nonalcoholic fatty liver disease (27, 30, 35, 36, 47–52). Furthermore, we have shown in murine models of chronic alcoholic liver disease (30, 47) as well as in a choline deficient amino acid defined (CDAA) diet-induced nonalcoholic steatohepatitis (NASH) (Maricic et al. submitted manuscript) that direct inhibition of iNKT cells by blocking the RARγ-signaling pathway or by anti-CD1d blocking antibody suppresses disease.

alcoholic steatohepatitis; NAFLD, nonalcoholic fatty liver disease; IBD, inflammatory bowel disease.

# CHOLINE METABOLISM AND NKT CELLS INVOLVEMENT IN INFLAMMATION

Choline deficiency has been implicated in exacerbation of hepatic steatosis and fibrosis in murine models (53, 54), and in nonalcoholic fatty liver disease in humans (55–57). Recent studies have indicated that alterations in the gut microbiota can lead to choline deficiency and contribute to NASH (58, 59). It is notable that the majority of the US population does not meet the daily recommended intake of choline, a situation that is associated with oxidative damage caused by mitochondrial dysfunction and ER stress (55). Furthermore, inflammation in patients with Crohn's disease and ulcerative colitis is also characterized by decreased levels of choline, phosphatidylcholine, and glycerophosphorylcholine due to increased use of choline (60). The presence of trimethylamine-producing bacteria in the gut significantly reduces choline bioavailability and perturbs choline metabolism, which can contribute to NASH (58, 59, 61). In addition, choline and phosphatidylcholine deficiency results in impaired secretion of very low-density lipoproteins and, consequently, accumulation of fat in the liver (steatosis). Interestingly, choline deficiency results in a decrease of colonic type II NKT cells and attenuates dextran sulfate sodiuminduced colitis (41). In addition, long-term feeding of a cholinedeficient high-fat diet mediate NASH and NASH-induced HCC by promoting liver infiltration by activated iNKT cells and CD8<sup>+</sup> T cells and induction of inflammatory cytokines (52). These observations suggest an important link between choline metabolism, microbiota, and NKT cells that can control liver inflammation. How activation and cross-regulation of iNKT cells and type II NKT cells are able to control immune responses in NASH and inflammatory bowel disease are currently being investigated in our laboratory.

#### NKT CELLS ACTIVATION IN THE LIVER AND GUT

In several models of acute and chronic liver inflammation, we have found that hepatic iNKT cells (as determined using αGalCer/CD1d-tetramers) become activated, but not type II NKT cells [reviewed in (31)] (see **Table 1**). Multiple mechanisms have been shown to be involved in the activation of iNKT cells. Thus, iNKT cells can recognize both microbial and self-derived lipid antigens presented by CD1d-expressing professional APCs, such as DC, resulting in direct activation of iNKT cells. Also, IL-12 and type I IFN produced by activated DC in response to ligands of toll-like receptor (TLR)-4 and TLR-9, respectively, can strengthen NKT activation (62, 63). Furthermore, it has been shown that CD1d-dependent activation of iNKT cells by hepatocytes control liver inflammation (64). Additionally, iNKT cells also become activated by a combination of cytokines, such as IL-12 and IL-18, in the absence of a CD1d-bound agonist (27, 31).

Since NKT cells are also localized in the intestine and CD1d is also expressed by intestinal B cells, DC, macrophages as well as intestinal epithelial cells, they can also mediate lipid-activation of NKT cells thus participating in the intestinal homeostasis and regulation of intestinal immunity. In fact, CD1d-dependent colonization with specific commensal gut bacteria has been shown to be crucial in the maintenance of mucosal homeostasis (65, 66). More recently, the importance of CD1d-dependent activation of iNKT cells in the gut by intestinal lipids presented by CD11c<sup>+</sup> DC has been shown for control of intestinal homeostasis (64, 67). Additionally, it has been also shown that altered commensal gut bacteria induce hepatic accumulation of CXCR6<sup>+</sup> iNKT cells that are able to inhibit liver cancer (68). Furthermore, it is well-known that migratory CD103<sup>+</sup> DC in the gut, which also express high levels of CD1d, play an important role in the activation of iNKT cells (69, 70).

Based upon data from other laboratories and ours, we propose the following model of how diet can lead to activation of iNKT cells in the gut and liver (**Figure 1**). Also, how sulfatide- or LPC-mediated activation of type II NKT cells inhibits iNKTmediated inflammatory cascade followed by protection from chronic inflammatory liver diseases. Alcoholic, choline-deficient, or high-fat diets can initially damage hepatocytes, leading to release of mitochondrial DNA and accumulation and activation of pDCs in the liver. The pDCs activate IL-17-secreting iNKT cells (iNKT17), which induces IL-17 signaling leading eventually to the expansion of conventional Th17 T cells. Chronic feeding of these diets leads to further changes in the gut microbiota, resulting in enterocytes or cDC-mediated presentation of lipids to iNKT cells. Additionally, migration of the CD103+CD11b<sup>−</sup> cDC and activated iNKT cells into liver results in activation or expansion of both MHC-restricted CD4+/CD8<sup>+</sup> T cells and IFNγ/IL-13-secreting iNKT cells as well as hepatic stellate cell activation and fibrosis. Collectively, these data suggest that DC subsets as well as CD1d-expressing parenchymal cells are involved in activation of iNKT cells in both liver and gut.

#### BIMODAL INTERACTIONS BETWEEN NKT CELLS AND GUT MICROBIOTA

Mucosal barrier in the gut plays a crucial role in keeping gut commensal microbes away from the host immune system. Accordingly, in the DSS-induced colitis model, live bacteria have been found in the liver at increased numbers (71). Similarly, high levels of bacterial colonization have been shown in portal blood, liver and peritoneum of patients with Crohn's disease

TABLE 1 | Differential activation, accumulation and influence of NKT cell subsets on other innate cells in liver inflammation.


*a iNKT cells were analyzed in liver mononuclear cells using* α*GalCer-loaded CD1d tetramers and identified as double positive cells for* α*GalCer/CD1d tetramer and TCR*β *(*α*GalCer/CD1d tetramer*+*TCR*β <sup>+</sup>*) in the NK1.1*<sup>+</sup> *gate.*

*<sup>b</sup>Type II NKT cells were also analyzed in liver mononuclear cells and defined as double positive cells for TCR*β *and NK1.1 (TCR*β <sup>+</sup>*NK1.1*+*) in the* α*GalCer/CD1d tetramer*<sup>−</sup> *gate or sulfatide/CD1d-tetramer*<sup>+</sup> *cells.*

*<sup>c</sup>Myeloid cells included both monocytes, neutrophils and macrophages.*

*<sup>d</sup>NK cells were defined as NK1.1*+*TCR*β <sup>−</sup> *cells.*

*<sup>e</sup>Cytokines secretion was determined in culture supernatants after stimulation with PMA & Ionomycin by FACS using BD Cytometric bead array.*

*<sup>f</sup>Apoptotic cells were identified as Annexin V*<sup>+</sup> *cells.*

*IRI, Ischemia Reperfusion Injury; Con A, Concanavalin A-induced hepatitis; Alcohol C*+*B, chronic feeding of Lieber-DeCarli liquid diet for 10 days plus binge on day 11; Alcohol Chronic, chronic feeding of Liber deCarli liquid alcohol diet for 6–8 weeks; CDAA diet, chronic feeding of Choline deficient amino acid defined diet for 20 weeks. NC, not changed; ND, not determined.*

(72). In different settings, the interactions in the gut-liver axis have been shown to be critical in the pathogenesis of several inflammatory liver diseases (4, 73, 74). NKT cells contribute to immunosurveillance in the small and large intestine. Thus, the frequency as well as the functional maturation of NKT cells in the lamina propria and gut epithelium is controlled by lipids derived from the commensal microbiota. At the same time, CD1d activation of NKT cells in the gut can modulate neonatal colonization with commensal bacteria (75, 76). Interestingly, bacterial lipids related to αGalCer can be synthesized by Bacteriodes species and can activate or inhibit iNKT cells (77, 78). Therefore, the frequency, maturation and phenotype of iNKT cells is greatly impacted by the gut microbiota in mice. It is not clear yet whether, in humans, commensal microbes can have a similar effect on iNKT cells. Similarly, it has recently been shown that NKT cells can control microbial communities in the gut. In the absence of NKT cells, mice develop much more severe colitis (66). In addition, lipid presentation within the gut by CD1d-expressing CD11c<sup>+</sup> DCs and macrophages induces activation of intestinal iNKT cells controlling the intestinal bacteria composition (67). Similarly, CD1d-mediated lipid presentation by hepatocytes controls homeostasis of hepatic iNKT cells. Altered CD1d-dependent presentation results in increased hepatic iNKT cell numbers and hepatic inflammation (64). Therefore, experimental data in murine models suggest reciprocal interactions between NKT cells and gut bacteria, and that these interactions have implications in targeting NKT cells or the microbiota in chronic liver diseases or inflammatory bowel disease. Whether similar reciprocal interactions may occur in human need to be determined. It has been shown that gut bacteria derived from chronic alcoholic hepatitis patients can induce liver disease in mice upon colonization (79), and that intestinal microbiota manipulation prevents alcohol-induced liver injury (80). Since murine alcoholic liver disease is mediated by iNKT

#### REFERENCES


cells, it is likely that products derived from critical microbes or diets can activate iNKT cells. Collectively, these studies indicate an important multimodal interaction among gut microbes, NKT cell subsets, and both immune and parenchymal cells in the maintenance of gut and liver homeostasis.

#### FUTURE PERSPECTIVE

The studies reviewed here have revealed a central role of NKT cell subsets in the maintenance of immune homeostasis in the liver and gut. Detailed knowledge of the cellular and molecular interactions among both NKT cell subsets, other innate and adaptive lymphocytes, and the microbiota will be key for developing novel intervention strategies for chronic inflammatory diseases, such as alcoholic liver disease, nonalcoholic steatohepatitis and inflammatory bowel disease. It is becoming increasingly clear that iNKT cells contribute to inflammation while type II NKT cells are protective (30), a clear understanding of the role of these unconventional T cells in large cohorts of patients with chronic inflammatory diseases of liver and gut is needed to develop inhibiting and activating agonists of iNKT cells and type II NKT cells, respectively, as novel therapeutic strategies for these diseases.

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

This work was supported by grants from the National Institutes of Health, USA (R01 CA100660 and R01 AA020864 to VK) and from Lupus Research Alliance, USA (VK).


antitumor immunity in a granulocyte-macrophage colony-stimulating factor-dependent fashion. Proc Natl Acad Sci USA. (2003) 100:8874–9. doi: 10.1073/pnas.1033098100


**Conflict of Interest Statement:** VK is a Scientific Co-founder and Consultant for the GRI-Bio, La Jolla, California.

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

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

# Clinical Application of iNKT Cell-mediated Anti-tumor Activity Against Lung Cancer and Head and Neck Cancer

Mariko Takami <sup>1</sup> , Fumie Ihara1,2 and Shinichiro Motohashi <sup>1</sup> \*

<sup>1</sup> Department of Medical Immunology, Graduate School of Medicine, Chiba University, Chiba, Japan, <sup>2</sup> Department of Otorhinolaryngology, Head and Neck Surgery, Graduate School of Medicine, Chiba University, Chiba, Japan

Invariant natural killer T (iNKT) cells produce copious amounts of cytokines in response to T-cell receptor (TCR) stimulation by recognizing antigens such as α-galactosylceramide (α-GalCer) presented on CD1d; thus, orchestrating other immune cells to fight against pathogen infection and tumors. Because of their ability to induce strong anti-tumor responses and the convenience of their invariant TCR activated by a synthetic ligand, α-GalCer, iNKT cells have been intensively studied for application in immunotherapeutic approaches to treat cancer patients in the clinic. Here, we summarize the clinical trials of iNKT cell based immunotherapy for non-small cell lung cancer, and head and neck cancer. Although solid tumors are thought to be refractory to immunotherapeutic approaches, our clinical trials showed that the intravenous injection of α-GalCer-pulsed antigen presenting cells (APCs) activated endogenous iNKT cells and iNKT cell dependent responses. Moreover, an increase in the number of IFN-γ producing cells in PBMCs was associated with prolonged survival. The marked infiltration of iNKT cells and the accumulation of conventional T cells in the tumor microenvironment were also observed after the administration of α-GalCer-pulsed APCs and/or ex vivo activated iNKT cells. In cases of advanced head and neck squamous cell carcinoma, the increased accumulation of iNKT cells in the tumor microenvironment was correlated with objective clinical responses. We will also discuss potential combination therapies of iNKT cell based immunotherapy to achieve enhanced anti-tumor activity and provide better treatment options for these patients.

Keywords: antigen presenting cells (APCs), invariant natural killer T cells (iNKT cells), cancer immunotherapy, non-small cell lung cancer (NSCLC), head and neck cancer (HNC)

#### INTRODUCTION

Invariant natural killer T (iNKT) cells comprise 1–2% of the mouse spleen and less than 0.1% in human peripheral tissues (1). Although a small population, iNKT cells elicit strong immune responses by producing large amounts of cytokines, which lead to other immune cell responses as well as inducing cytotoxicity; thus, they have an important role in both innate and adaptive immunity.

#### Edited by:

Kazuya Iwabuchi, Kitasato University School of Medicine, Japan

#### Reviewed by:

Yoshihiro Hayakawa, University of Toyama, Japan Christopher E. Rudd, Université de Montréal, Canada

\*Correspondence: Shinichiro Motohashi motohashi@faculty.chiba-u.jp

#### Specialty section:

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

Received: 02 May 2018 Accepted: 16 August 2018 Published: 07 September 2018

#### Citation:

Takami M, Ihara F and Motohashi S (2018) Clinical Application of iNKT Cell-mediated Anti-tumor Activity Against Lung Cancer and Head and Neck Cancer. Front. Immunol. 9:2021. doi: 10.3389/fimmu.2018.02021

**126**

Whereas conventional T cells express diverse T cell receptors (TCR) after TCR rearrangement and recognize their cognate peptide presented on MHC, iNKT cells express monoclonal TCRs composed of a Vα24-Jα18 chain and Vβ11 chain in humans, and a Vα14-Jα18 chain and Vβ8.2 chain in mice that recognize glycolipid antigens presented on CD1d, a MHC class I like molecule (2, 3). Another unique characteristic of iNKT cells that distinguishes them from conventional T cells is the expression of NK cell receptors including CD56, CD16, NKG2D, and CD161 (4). Therefore, iNKT cells recognize non-self in a TCR dependent manner as well as by NK cell receptors. In 1997, Kawano et al. identified αgalactosylceramide (α-GalCer), derived from the marine sponge, as a glycolipid ligand that activates iNKT cells and this discovery has enhanced our understanding of the function and role of iNKT cells in immunity over the past 20 years (5). α-GalCer is a potential therapeutic tool to control immune responses in a variety of disease conditions such as cancer and autoimmunity.

In tumor immunity, the role of iNKT cells was initially demonstrated by using mouse models. CD1d knockout mice or Vα14 knockout mice, which lack iNKT cells, are more susceptible to tumors compared with wild type mice, suggesting that iNKT cells play a crucial role in anti-tumor immunity. Because iNKT cells recognize α-GalCer presented on CD1d expressed on antigen presenting cells (APCs), Toura et al. demonstrated that the administration of α-GalCer loaded dendritic cells (DCs) expanded NKT cells and rejected tumors in a mouse liver metastasis and lung metastasis model (6). Furthermore, even established metastatic tumors were rejected by the administration of α-GalCer loaded DCs. iNKT cells produce large amounts of cytokines including IFN-γ and IL-4 upon TCR stimulation by recognizing α-GalCer presented on CD1d. These events consequently lead to the activation of NK cells and CD8<sup>+</sup> T cells and the conversion of immature DCs to mature DCs; thus, enhancing other immune cell responses to indirectly eradicate tumors (7). iNKT cells also produce granzyme A and granzyme B and express Fas ligand and TRAIL to exert a direct tumor killing effect (8). These pleiotropic functions of iNKT cells are thought to be involved in tumor eradication. Based on these findings, we applied α-GalCer loaded DCs as a tool for iNKT cell based immunotherapy to treat cancer patients.

Immunotherapy is widely recognized as a powerful cancer treatment tool because immune checkpoint blockade with programmed death-1 (PD-1) antibody eradicated tumors dramatically in some patients with lung cancer (9). Since then, a variety of immunotherapeutic approaches has been studied to determine better treatment options for cancer patients. One of these approaches is iNKT cell based therapy and clinical trials of iNKT cell-targeted immunotherapy for lung cancer, and head and neck cancer patients have been conducted at Chiba University. Herein, we will review our clinical trials of iNKT cell based immunotherapy (**Table 1**). We will also discuss the next step of iNKT cell based immunotherapy including combination therapies and induced pluripotent stem (iPS) cell-derived iNKT cells.

# iNKT Cell Based Immunotherapy for Lung Cancer

Lung cancer is a leading cause of cancer death worldwide. Non-small cell lung cancer (NSCLC) is the most common type of lung cancer and is classified as one of three representative subtypes: squamous cell carcinoma, adenocarcinoma, and large cell carcinoma. Although cytotoxic chemotherapeutic agents have been used as the first line treatment for advanced or unresectable NSCLC patients for many years, some PD-1 signaling blockade antibodies have been clinically approved and are used as a first line or second line treatment in recent years (17). These treatments have had impressive outcomes for some, but not all, NSCLC patients; thus, alternative therapies are still highly desired to improve the prognosis of NSCLC patients.

# Establishment of α-GalCer-Loaded APCs

In general, DCs are generated from peripheral blood CD14<sup>+</sup> monocytes cultured in the presence of DC inducing cytokines, GM-CSF and IL-4. Some clinical studies using these cells to target cytotoxic T lymphocytes have been performed (18, 19). However, these culture methods might not be suitable for iNKT cell based clinical trials because a large number of DCs are required to encounter and activate the small population of iNKT cells in the periphery of patients. Therefore, we developed a new method to obtain a large number of APCs from peripheral blood mononuclear cells (PBMCs). In addition to DCs, other immune cells such as activated T cells express CD1d on their cell surface; therefore, the potential use of cultured whole PBMCs instead of DCs was suggested (20). It was reported that PBMCs cultured in the presence of GM-CSF and IL-2 presented α-GalCer on CD1d and induced the expansion and activation of human iNKT cells better than DCs generated in the presence of GM-CSF and IL-4.

# Early Phase Clinical Trials: Administration of α-GalCer-Loaded APCs

Because a large number of APCs can present α-GalCer and activate iNKT cells more effectively compared with monocyte derived DCs, and can be generated from PBMCs by culturing with GM-CSF and IL-2, we designed a phase I clinical trial for NSCLC patients using α-GalCer-pulsed APCs (10). Eleven patients with stage IV or recurrent NSCLC were enrolled in this study and nine completed the treatment. In this phase I study, safety profiles of α-GalCer-pulsed APCs were examined at three different doses at level 1: 5 × 10<sup>7</sup> , level 2: 2.5 × 10<sup>8</sup> and level 3: 1 × 10<sup>9</sup> cells/m<sup>2</sup> . Patients received four intravenous injections of α-GalCer-pulsed APCs over 3 months. While objective clinical responses were not observed in all cases, patients who received the level 3 dose of α-GalCer-pulsed APCs exhibited iNKT cell expansion in the periphery and showed long term survival for over one year. These iNKT cells had high ifn-γ mRNA expression. No severe adverse events over grade II were observed in all cases including the highest dose at level 3, suggesting α-GalCerpulsed APC administration was safe and a feasible treatment. Moreover, the level 3 dose (1 × 10<sup>9</sup> cells/m<sup>2</sup> ) was considered the most effective and safe dose of α-GalCer-pulsed APCs to treat advanced NSCLC patients.


We then designed the next step, a phase I-II clinical trial for NSCLC patients, to extend the study of iNKT cell based immunotherapy (11). Twenty-three patients were enrolled in this study from February 2004 to August 2006, and 17 completed the protocol treatment. Patients were either stage IIIB, stage IV or recurrent NSCLC patients who received standard therapy. The safety of α-GalCer-pulsed APCs and immunological responses were examined at a dose of 1 × 10<sup>9</sup> cells. All patients received two courses with four injections of 1 × 10<sup>9</sup> α-GalCer-pulsed APCs. Regarding adverse events, one patient experienced the recurrence of deep vein thrombosis (DVT) (estimated as a grade III adverse event) and required hospitalization for the continuous injection of heparin. The Chiba University Quality Assurance Committee on Cell Therapy determined no clear relationship between the cell therapy and DVT. Because no severe adverse effects were observed in other patients, the safety of the administration of α-GalCer was confirmed. Regarding immune monitoring, 10 patients had a greater than two-fold increase of IFN-γ producing cells in the periphery after the administration of α-GalCer-pulsed APCs (good responders) whereas seven patients showed mild or no increase of IFN-γ producing cells (poor responders). Moreover, the increase of IFN-γ producing cells in the periphery of patients correlated with the median survival time (MST) and good responders showed a longer MST (29.3 months) compared with poor responders (9.7 months). Overall, the MST of all cases was 18.6 months. Because this clinical trial was not a randomized controlled study, we could not conclude the superiority of α-GalCer-pulsed APCs; however, these data encouraged us to perform a comparative study to compare the α-GalCer-pulsed APCs and standard cytotoxic drugs.

To further elucidate the mechanisms by which the administration of α-GalCer-pulsed APCs trigger iNKT cell immune responses in the tumor microenvironment of NSCLC patients, we performed a clinical study targeting four patients diagnosed as stage IIB and IIIA NSCLC who underwent surgical treatment compared with 6 patient controls (12). Patients received a single intravenous injection of 1 × 10<sup>9</sup> α-GalCerpulsed APCs seven days prior to surgery and characteristic tumor infiltration was analyzed from surgically removed tumor tissue specimens. A higher percentage of iNKT cells was observed in tumor infiltrating lymphocytes (TIL) compared with mononuclear cells isolated from normal lungs and draining lymph nodes. The percentage of iNKT cells in TIL was increased with α-GalCer-pulsed APC administration compared with control groups. Moreover, increased numbers of IFN-γ producing cells in TIL were observed after α-GalCer-pulsed APC administration, indicating that the systemic intravenous administration of α-GalCer-pulsed APCs led to local iNKT cell accumulation in the tumor microenvironment and induced immune responses by producing IFN-γ.

#### Phase I Clinical Trial: Administration of iNKT Cells

To increase iNKT cell numbers in the periphery, the administration of in vitro expanded NKT cells was performed as a phase I clinical trial in six patients with recurrent lung cancer (13). iNKT cells were prepared in vitro from PBMCs cultured in the presence of α-GalCer and IL-2. In vitro expanded iNKT cells (level 1: 1 × 10<sup>7</sup> cells, level 2: 5 × 10<sup>7</sup> cells per injection) were intravenously transferred to patients. Whereas it was previously reported that iNKT cells in cancer bearing patients had a lower frequency and impaired proliferation capability, iNKT cells derived from patients in this study expanded and produced Th1 dominant cytokines including IFN-γ along with tumoricidal activity ex vivo. No patients had severe adverse events during the study. Two of three patients who received a level 2 dose showed increased IFN-γ production while no patients met the criteria for an objective clinical response. These results suggested that the administration of α-GalCer-pulsed APCs was a more effective treatment compared with the administration of ex vivo expanded iNKT cells.

The intravenous injection of α-GalCer-pulsed APCs in patients with advanced or recurrent NSCLC after first line treatment was accepted as an advanced medicine by the Japanese Ministry of Health, Labour and Welfare in 2011. Since then, 35 patients were enrolled in this study and 32 received all courses of treatment. The follow-up was completed in 2017. We are currently analyzing the clinical efficacy and immune responses.

### iNKT Cell Based Immunotherapy for Head and Neck Cancer

Head and neck cancer (HNC) accounts for about 5% of all cancers. Despite the development of multidisciplinary treatment involving surgery, radiotherapy, and chemotherapy for advanced cases, the recurrence rate is still high; thus, the survival rate remains relatively low. Moreover, the quality of life (QOL) of patients who receive these combination therapies is often severely impaired. To improve the prognosis and QOL of patients with head and neck cancer, the development of new therapies is highly desired. Because iNKT cell based immunotherapy for NSCLC patients showed promising results in the treatment of solid tumors, we designed clinical studies of iNKT cell based immunotherapy for HNC patients. While the intravenous administration of α-GalCer-pulsed APCs was used in our clinical trials for NSCLC patients, we found that nasal submucosa injection induced APCs to migrate to the neck lymph node area (21). Furthermore, the nasal submucosa injection of α-GalCer-pulsed APCs increased the number of iNKT cells and IFN-γ producing cells in the peripheral tissues of patients (22). In contrast, the injection of α-GalCer-pulsed APCs into the oral floor submucosa induced tolerance with increased numbers of CD45RA−Foxp3high Tregs instead of anti-tumor activity. These results indicated that the administration of α-GalCerpulsed APCs via the nasal submucosa was a better option for HNC patients. We also confirmed that the number and function of iNKT cells were not affected by radiation therapy, suggesting that iNKT cell based immunotherapy might be an adjuvant treatment of radiation therapy for advanced HNC patients (23).

# Clinical Trials of iNKT Cell Based Immunotherapy for Patients With Advanced and Recurrent HNC

We conducted a phase I clinical trial study of iNKT cell based immunotherapy for patients with recurrent or unresectable HNC using α-GalCer-pulsed APCs (14). Nine patients were enrolled in this study and α-GalCer-pulsed APCs (1 × 10<sup>8</sup> cells/injection) were administrated into the nasal submucosa. During the study period, no serious adverse events over grade 3 were observed. Moreover, the number of peripheral iNKT cells increased in four patients and an increase in IFN-γ producing cells was observed in eight patients. These results suggested that the administration of α-GalCer-pulsed APCs into the nasal submucosa was a safe and effective approach to induce iNKT cell anti-tumor responses. However, the clinical efficacy was not satisfactory in this study.

To improve clinical efficacy and confirm the safety of iNKT cell based immunotherapy in patients with recurrent or unresectable HNC, we conducted a combination therapy with ex vivo expanded iNKT cells and α-GalCer-pulsed APCs. We performed a phase I study of eight patients with recurrent HNC that were refractory to standard therapy (15). Because most HNCs receive blood supply through terminal arteries, the intra-arterial infusion of anti-cancer agents is commonly used (24). Patients received two nasal submucosa injections of α-GalCer-pulsed APCs (1 × 10<sup>8</sup> cells/injection) and one intraarterial infusion of activated iNKT cells (5 × 10<sup>7</sup> cells/injection). No patients showed severe adverse events except one patient who had partial response (PR) with pharynx-skin fistula (grade III). PR was observed in three patients while stable disease (SD) or progressive disease (PD) was observed in four patients. The iNKT cell number in the periphery was increased over 3 fold after the treatment in all patients and increase in IFNγ producing cells in the periphery was over 4-fold in seven patients (15).

In a phase II study, 10 patients with locally recurrent HNC and were indicated for salvage surgery were enrolled. Therefore, ex vivo expanded iNKT cells (5 × 10<sup>7</sup> cells) were administrated via the terminal artery and α-GalCerpulsed APCs (1 × 10<sup>8</sup> cells) were administrated into the nasal submucosa. No severe adverse events over grade 2 were observed in any patients (16). The number of iNKT cells in the periphery was increased in nine patients. Furthermore, IFN-γ producing cells in the periphery were increased in all patients. Additionally, five patients achieved PR and five patients achieved SD. Interestingly, the percentage of infiltrated iNKT cells was higher in PR cases compared with SD cases. Based on these results, we confirmed that the co-administration of ex vivo expanded iNKT cells and α-GalCer-pulsed APCs augmented both anti-tumor immune responses and clinical efficacy.

## Ongoing Advanced Medicine for HNC Patients

A double-blinded randomized study is essential to provide compelling evidence to develop novel therapies for cancers; thus, we conducted a double-blinded randomized study of α-GalCer-pulsed or non-pulsed APC administration to the nasal submucosa in patients with HNC who archived complete response (CR) after standard therapy. The aim of the study was to clarify the role of iNKT cells to prevent the recurrence of HNC after standard treatment in those thought to have minimal residual disease. This study was accepted as advanced medical care by the Japanese Ministry of Health, Labour and Welfare in 2012. Patients diagnosed with stage IV HNC and who achieved CR after standard therapy were enrolled in this study to investigate the prevention of the recurrence of HNC. Patients received two α-GalCer-pulsed APC injections (1 × 10<sup>8</sup> cells/injection) into the nasal submucosa on days seven and ten. Patient recruitment, follow-up and immunological analysis are in progress.

# FUTURE DIRECTION

While iNKT cell based immunotherapy is considered a promising treatment for patients with lung cancer, and head and neck cancer based on the results of previous clinical trials mentioned above, some factors can be improved to provide better treatment options for these patients. Because iNKT cells are such a small population of lymphocytes, it is critical to determine how to efficiently increase the iNKT cell number and enhance their function including IFN-γ production and tumoricidal activity. We are currently investigating two approaches to make our current iNKT cell based immunotherapy more effective. One is to generate iPS cell derived NKT cells (iPS-NKT cells), and the other is combination therapy.

iPS cells are generated from somatic cells by introducing genes to express four transcription factors (Oct4, Sox2, Klf4, and c-Myc) named Yamanaka factors (25). They are reprogrammed to an embryonic stem like cell, which can self-renew and differentiate into multiple cell types. Yamada et al. successfully generated human iPS-NKT cells from iNKT cells isolated from peripheral blood mononuclear cells or cord blood mononuclear cells by introducing Yamanaka factors (26). iPS-NKT cells showed an iPS cell phenotype that was capable of proliferating and that retained NKT cell functions such as cytokine production and cytotoxic activity. The generation of iPS-NKT cells will help acquire sufficient numbers of iNKT cells from patients with low numbers of iNKT cells for next generation iNKT cell based immunotherapy. We are currently performing nonclinical tests of iPS-NKT cells to confirm their safety and efficacy.

Furthermore, we are also investigating potential combination therapies to enhance iNKT cell based immunotherapy. We found that the blockade of PD-1/PD-ligand 1(PD-L1) signaling, an immune checkpoint pathway, enhanced iNKT cell function including cytokine production and tumoricidal activity, suggesting that combined immune checkpoint inhibitor and iNKT cell based immunotherapy might have synergistic effects and exert powerful anti-tumor immunity (27).

# AUTHOR CONTRIBUTIONS

All authors listed have made substantial contributions to the manuscript and have approved it for submission.

#### REFERENCES


# FUNDING

This work was supported by the Japan Agency for Medical Research and Development (No. 18m0304003h0106 to SM).


**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 Takami, Ihara and Motohashi. 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.

# Type II NKT Cells: An Elusive Population With Immunoregulatory Properties

#### Avadhesh Kumar Singh, Prabhanshu Tripathi † and Susanna L. Cardell\*

*Department of Microbiology and Immunology, Institute of Biomedicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden*

Natural killer T (NKT) cells are unique unconventional T cells that are reactive to lipid antigens presented on the non-polymorphic major histocompatibility class (MHC) I-like molecule CD1d. They have characteristics of both innate and adaptive immune cells, and have potent immunoregulatory roles in tumor immunity, autoimmunity, and infectious diseases. Based on their T cell receptor (TCR) expression, NKT cells are divided into two subsets, type I NKT cells with an invariant TCRα-chain (Vα24 in humans, Vα14 in mice) and type II NKT cells with diverse TCRs. While type I NKT cells are well-studied, knowledge about type II NKT cells is still limited, and it is to date only possible to identify subsets of this population. However, recent advances have shown that both type I and type II NKT cells play important roles in many inflammatory situations, and can sometimes regulate the functions of each other. Type II NKT cells can be both protective and pathogenic. Here, we review current knowledge on type II NKT cells and their functions in different disease settings and how these cells can influence immunological outcomes.

Keywords: type II NKT cells, CD1d, sulfatide, T cell receptor, health and disease

# INTRODUCTION

Although conventional MHC-restricted T cells have been the main attraction for immunologists, unconventional T cells are continuously gaining increased attention. Despite of their low frequencies compared to conventional T cells, the subsets of unconventional T cells play an important role in various autoimmune diseases, cancers and infections, and are present both in human and mice. These cells include CD1d-restricted NKT cells, γδ TCR expressing T cells, and MR1-restricted mucosal associated invariant T cells (1). NKT cells are reactive to lipid antigens presented on the non-polymorphic MHC class I-like molecule CD1d (2). These cells are activated early in immune responses and can express immunoregulatory activities that determine different immune outcomes.

In contrast to conventional T cells that express highly diverse TCR, NKT cells appear comparatively limited in its TCR repertoire and ligand reactivities. Based on the nature of their TCR expression, NKT cells comprise two main categories, invariant or type I NKT cells, and diverse or type II NKT cells. Type I NKT cells, the most extensively studied subgroup, express a semi-invariant Vα14-Jα18 TCR in mice, and Vα24-Jα18 in humans paired with a limited repertoire of Vβ-chains (Vβ8.2, Vβ7, and Vβ2 in mice and Vβ11 in humans) (1, 3), while the less explored type II NKT cells utilize a more diverse TCR repertoire (4–7). In mice, type I NKT cells outnumber type II NKT cells,

#### Edited by:

*Luc Van Kaer, Vanderbilt University, United States*

#### Reviewed by:

*Marika Falcone, San Raffaele Hospital (IRCCS), Italy Akihiro Ishizu, Hokkaido University, Japan*

\*Correspondence: *Susanna L. Cardell susanna.cardell@microbio.gu.se*

#### †Present Address:

*Prabhanshu Tripathi, Centre for Human Microbial Ecology, Translational Health Science and Technology Institute, NCR Biotech Science Cluster, Faridabad, India*

#### Specialty section:

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

Received: *03 July 2018* Accepted: *10 August 2018* Published: *28 August 2018*

#### Citation:

*Singh AK, Tripathi P and Cardell SL (2018) Type II NKT Cells: An Elusive Population With Immunoregulatory Properties. Front. Immunol. 9:1969. doi: 10.3389/fimmu.2018.01969*

**132**

while in humans the type II NKT cells are more frequent. Type I NKT cells are reactive to the marine sponge-derived glycolipid α-galactosylceramide (α-GalCer) (8), which forms a stable α-GalCer/CD1d tetramer reagent that can be used for detection of type I NKT cells (9, 10). Availability of α-GalCer/CD1d tetramers opened up new avenues of NKT cell research and laid the foundation for important findings in the field of immunoregulation. In contrast, type II NKT cells are not reactive to α-GalCer (11, 12), instead they are thought to make up an oligoclonal population that recognizes a diverse repertoire of lipid antigens (5, 7, 13). As ligand/CD1d tetramer reagents only recognize a fraction of type II NKT cells, and due to the lack of specific surface markers, it is currently not possible to identify the entire type II NKT cell population using single or combined reagents. Instead, type II NKT cells are often define by indirect approaches. One way of identifying this population is by their CD1d-reactivity and absence of the Vα14 (mouse) or Vα24 (human) TCR α-chains. In a few cases, ligands recognized by type II NKT cells can be loaded on CD1d-tetramers and used to identify subsets of the population (14). Type II NKT cells are abundant in liver of both mice (15) and human (16), although the frequency of the entire type II NKT cell population in different organs is still unknown. However, recent advances in the field have highlighted the importance of type II NKT cells in different diseases. In this review, we discuss recent progress in the study of type II NKT cells and their critical role in different disease settings.

#### ANTIGENS FOR TYPE II NKT CELLS

MHC-II-deficient mice facilitated the discovery of type II NKT cells, when it was noted that while conventional CD4<sup>+</sup> cells were absent, there was still a significant population of peripheral CD4<sup>+</sup> cells (4). Among CD4<sup>+</sup> T cell hybridomas developed from these mice, several were CD1d-autoreactive and they expressed different TCR. These and other type II NKT cell hybridomas were instrumental for the identification of different lipid antigens (5, 11, 15, 17–19). With the identification of more CD1d-restricted ligands for type II NKT cells, the ligands could be categorized into different subtypes - sphingolipids and glycerolipids or phospholipids (7, 20). As α-GalCer is not recognized by type II NKT cells, taken together, this suggests that type I NKT cells and type II NKT cells react to different sets of CD1d-presented antigens, consistent with their distinct TCR, although there are some ligands that are recognized by both NKT subsets.

Consistent with the CD1d-autoreactivity of many type II NKT cell hybridomas, several self lipids have been defined as ligands. Sulfatide, a well studied ligand for type II NKT cells, was first identified as a ligand for the type II NKT cell hybridoma XV19 expressing a Vα1/Vβ16 TCR (4, 15, 19). Subsequently, type II NKT cell reactivity to sulfatide has been shown by independently derived hybridomas (15, 19, 21). Structural analysis of the XV19 TCR in complex with CD1d and sulfatide revealed similarities to both TCR-MHC interactions and type I NKT TCR-CD1d (22, 23). A significant population of sulfatide/CD1d tetramerreactive cells, around 0.2–4% in different organs, were identified in naive mice (15). Sulfatide reactive cells displayed diverse TCR with oligoclonal expansions, and preferentially used Vα3/Vα1- Jα7/Jα9 and Vβ8.1/Vβ3.1-Jβ2.7 TCR segments (24). While type I NKT cells possess germline-encoded CDR3 region of the TCRα chain, type II NKT cells demonstrated predominantly non-germline but also germline encoded CDR3 of TCRα and TCRβ chains (24). The substantial fraction of sulfatide/CD1dreactive cells, and the different TCR used for CD1-restricted sulfatide reactivity emphasized the diverse but oligoclonal nature of the TCR of these cells. Using XV19 hybridoma reactivity, we demonstrated that lysosulfatide, which lacks the fatty acid chain, was the most stimulatory sulfatide isoform, followed by C24:1 and C24:0 (19). Surprisingly, mice genetically deficient in cerebroside sulfotransferase (CST−/−) or UDP-galactose ceramide galactosyltransferase (CGT−/−), key enzymes required in the synthesis of sulfatides, showed the presence of sulfatide reactive type II NKT cells (15, 24). These findings suggest that sulfatide is dispensable for the development of sulfatide-reactive type II NKT cells. Further, the sulfatide reactive XV19 hybridoma showed high autoreactivity to splenocytes from CST−/<sup>−</sup> mice (19). These findings may be explained by reactivity to several ligands by one TCR, shown by the fact that the self-lipids β-glucosylceramide and β-galactosylceramide as well as selfphospholipids can activate the sulfatide reactive XV19 type II NKT cells (17, 25).

Using lipid loaded tetramers, significant populations of human and mouse type II NKT cells were identified reactive to β-glucosylceramide and glucosylsphingosine, lipids that accumulate in Gaucher's disease (26). Sequencing of TCRβchains of the human tetramer positive cells revealed diverse TCR. Lysophosphatidylcholine (LPC) is another ligand recognized by both murine and human type II NKT cells (25, 27, 28). LPC is also recognized by a few human type I NKT cells, but not by murine type I NKT cells (29–31).

Type II NKT cells have also been shown to react to ligands of microbial origin. Phosphatidylglycerol, diphosphatidylglycerol (or cardiolipin), and phosphatidylinositol from Corynebacterium glutamicum or Mycobacterium tuberculosis can activate various type II NKT, but not type I NKT, hybridomas, in a CD1ddependent manner. The specificities of these hybridomas for different lipid antigens were distinct, albeit partially overlapping (18). In a subsequent study, a more potent phosphatidylglycerol antigen from Listeria monocytogenes was identified that had a structure distinct from previously identified mammalian or microbial variants, as it contains short, fully saturated anteiso fatty acid lipid tails (32). Type II NKT cells may not only recognize lipid antigens but a recent study suggests that they also can recognize hydrophobic peptides presented on CD1d (33, 34). A report by Nishioka et al. found that a rat type II NKT cell clone, reactive to vascular endothelial cells, could recognize a CD1d-presented hydrophobic peptide derived from sterol carrier protein 2, a protein implicated in intracellular lipid transfer.

Considering the relatively large populations of primary type II NKT cells that recognize identified lipid ligands, one may speculate that the number of antigens recognized by type II NKT cells is limited. On the other hand, a single type II NKT cell TCR can bind several different antigens. Further, the type II NKT cells that bind a given lipid/CD1d-tetramer have diverse but oligoclonal TCR. Therefore, type II NKT TCR show degeneracy for antigen recognition. Importantly, the TCR repertoire of type II NKT cells appears to largely overlap between mice and human.

# MODES OF TYPE II NKT CELL ACTIVATION

Having characteristics of both NK and T cells, NKT cells can respond to either innate (TCR-independent) or adaptive (TCRdependent) stimulation. Type I NKT cells not only respond very rapidly to stimulation through TCR by secreting diverse cytokines but also to cytokines (IL-12, IL-18, and type I IFN) alone or produced by Toll-like receptors (TLR)-activated DCs, in the absence of TCR-engagement (35). However, information regarding the activation of type II NKT cells is still scarce. In contrast to type I NKT cells, lysophospholipid-reactive type II NKT cell activation was independent of IL-12 during hepatitis B virus (HBV) infection (28). A subset of type II NKT cells were shown to respond with partially CD1d-independent IFNγ production when co-cultured with CpG stimulated DC (36). This implies TCR-independent activation of type II NKT cells on the one hand, but also suggests that TLR activation of DC may have upregulated type II NKT cell ligands on CD1d. It therefore seems feasible that like type I NKT cells, type II NKT cells are not limited to activation by TCR-engagement but can also be activated independently of TCR in a proinflammatory cytokine milieu. However, this needs to be directly addressed. Thus, type II NKT cells can be activated through the TCR by exogenous antigens, such as microbial lipids, or self-lipids that may be upregulated on CD1d in activated DC. Moreover, they can likely be activated indirectly by pathogen derived or endogenous TLR-ligands acting on DC, or by inflammatory cytokines independently of the TCR. It is likely that under most circumstances, both TCR-engagement and TCR-independent stimulatory signals contribute to type II NKT cell activation.

# FEATURES OF TYPE II NKT CELLS

Type I and type II NKT cells share several features that makes them different from conventional T cells, but the two subsets often have distinct functions in specific immune reactions (5, 7, 13). Transcription factors play an important role in the development of MHC-restricted conventional T cells, and their combinations guide the functional profile of these cells. The transcription factor promyelocytic leukemia zinc finger (PLZF), induced by TCR signaling by agonist self-ligands after positive selection, is crucial for the development of type I NKT cells (37, 38). Type I NKT cells can be sub-grouped into distinct functional subsets based on the combinations of transcription factors such as PLZF, T-bet and RORγt: NKT1 (PLZFlowT-bet+), NKT2 (PLZFhiT-bet−) and NKT17 (RORγt <sup>+</sup>) cells that secrete TH1-, TH2-, and TH17-like cytokine patterns, respectively, upon activation (39, 40). Whether type II NKT cells follow similar developmental pathways and can be sub-grouped in similar functional subsets is not yet clearly understood. Studies have found that at least a subset of type II NKT cells have a constitutive production of IL-4 like type I NKT cells, and these type II NKT cells could be identified as IL-4-reporter<sup>+</sup> cells that did not bind the α-GalCer/CD1d-tetramer, or were present in mice lacking type I NKT cells (28, 36). These type II NKT cells exhibited similar developmental requirements as type I NKT cells - PLZF along with signaling lymphocyte activation molecule-associated protein (SAP) played a crucial role (36). The cells were PLZFhi and had an activated phenotype (CD44+, CD62L−, CD69hi) comparable to type I NKT cells. In other studies, mouse and human type II NKT cells were described as PLZFint and having a more resting phenotype (24, 26, 41, 42). In addition, several reports suggest that type I and type II NKT cells have different cytokine-producing capacities, and interestingly, that IL-13 may be more produced by type II NKT cells (15, 26, 27, 36, 41, 42). Taken together, studies using TCR transgenic mice, lipid/CD1dtetramers, and IL-4-reporter mice show that type II NKT cells are diverse also in their phenotype, cytokine secretion and activation state. While they frequently express NKT cell characteristic markers such as PLZF, NK receptors and CD122, some type II NKT cells seem more similar to type I NKT cells while others have a resting phenotype or different functional characteristics. This is in line with the diverse activities that have been associated with type II NKT cells.

# TYPE II NKT CELLS IN DIFFERENT DISEASES

## Multiple Sclerosis (MS) and Experimental Autoimmune Encephalomyelitis (EAE)

MS is a demyelinating autoimmune disease of the central nervous system (CNS) in which myelin-derived protein/lipid antigens are targets for autoreactive T cells. The myelin sheath is a rich source of sulfatide. Sulfatide-reactive T cells are more frequent in peripheral blood of MS patients than in healthy individuals (43), and CD1d-sulfatide reactive type II NKT cells, but not type I NKT cells, accumulate in the CNS in the experimental autoimmune encephalomyelitis (EAE) mouse model for MS (15). Thus, sulfatide-reactive type II NKT cells are thought to play a role in EAE, and in vivo administration of sulfatide at the time of EAE induction prevented the disease in a CD1d-dependent manner (15). Sulfatide-reactive type II NKT cells induced anergy of type I NKT cells, tolerized DCs and CNS microglia and inhibited the effector function of the pathogenic autoantigenreactive CD4 T cells (15, 44).

# Type 1 Diabetes (T1D)

T1D is an autoimmune disease in which autoreactive T cells target pancreatic β-cells of the Langerhans' islets. We have shown that type II NKT cells from 24αβ transgenic mice can protect NOD mice from disease, dependent on costimulators inducible costimulator (ICOS) and PD1 interactions (45, 46). Besides being a major lipid in the myelin sheath, different species of sulfatide are also present in pancreatic β-cells and could serve as ligands for type II NKT cells during T1D immunopathogenesis (47). Sulfatide/CD1d-tetramer positive cells were indeed enriched in pancreas-draining lymph nodes in non-obese diabetic (NOD) mice. Sulfatide administration to NOD mice reduced T1D incidence and islet-specific T cell responses by inducing secretion of the anti-inflammatory cytokine IL-10 from DCs (48, 49), however, protection from T1D in NOD mice by sulfatide was not found in a second study (50).

#### Tumor Immunity

A role for NKT cells in tumor immunity is well established; particularly for type I NKT cells (51). In many human cancers, low levels of circulating type I NKT cells correlate with a poor prognosis; therefore, these cells have been targeted in a series of clinical trials (52). In some cancer forms, type I and type II NKT cells play opposing roles; while type I NKT cells promote, type II NKT cells suppress tumor immunity. Immunosuppression by type II NKT cells was shown to be mediated by IL-13 production resulting in the activation of TGF-β-secreting Gr1+CD11b<sup>+</sup> myeloid derived suppressor cells (MDSCs), which in turn suppressed tumor-specific CD8<sup>+</sup> T cells or type I NKT cells (51). A similar scenario may be present in multiple myeloma patients, in which IL-13 secreting LPC/CD1dtetramer positive type II NKT cells were several fold increased in peripheral blood (27). It may therefore be a way forward to target tumor immunosuppression by type II NKT cells in cancer treatment, as discussed in a recent review (53). However, the roles of NKT cells in tumor immunity are more complex, as both type I and II NKT cells can promote tumor immunity, and there are other cancers in which both type I or type II NKT cells can suppress tumor immunity, as recently discussed in more detail (53, 54). The factors that determine whether NKT cells promote or suppress tumor immunity are not well understood, but it is possible that NKT cell activation by CD1d-expressing tumor cells will favor immunosuppressive functions.

#### Ulcerative Colitis (UC)

UC is a one of two forms of inflammatory bowel disease (IBD) characterized by Th-2-driven mucosal inflammation and tissue destruction in the colon. In UC, sulfatide-reactive type II NKT cells producing IL-13 were increased and suggested to have a colitogenic role (55, 56). Interestingly, in a mouse model for UC, a similar function was attributed to IL-13 producing type I NKT cells that were required to induce the disease (57). However, in different mouse models for IBD, type I NKT cells were either pathogenic or protective (56). Microbial components are abundant in the intestine, which are likely to activate the NKT cells during IBD and compromised intestinal barrier function. 24αβ TCR transgenic mice with elevated CD1d expression spontaneously developed colitis but here IFN-γ and IL-17 were the main players, not IL-13 (58). Moreover, a recent study has suggested the involvement of type II NKT cells in dextran sulfate sodium-induced colitis in mice provided choline-deficient diet (59).

#### Obesity

Obesity is a disease of low-grade adipose-tissue inflammation with a potential cancer risk (60). Both protective and pathogenic roles of type I NKT cells in obesity has been reported by different studies as these cells can either produce anti-inflammatory cytokines IL-4 and IL-10 or pro-inflammatory cytokines IFNγ, in adipose tissues (61–65). However, type II NKT cells exacerbated diet-induced obesity, as deduced from mice that lack type I NKT cells compared with CD1d−/<sup>−</sup> mice (lacking both cell types) (66). Similarly, another recent study indicated that type II NKT cells in ldlr−/<sup>−</sup> mice promote spontaneous obesity, as ldlr−/<sup>−</sup> or ldlr−/−CD1d−/<sup>−</sup> mice are less obese and have less adipose tissue inflammation than ldlr−/−Jα18−/<sup>−</sup> mice (67). By contrast, it was recently shown that sulfatide-induced type II NKT cells prevented high fat diet-induced obesity in mice by regulating adipose tissue inflammation, and their transfer into obese mice resulted in improved weight loss and glucose tolerance (68).

#### Liver Inflammation and Hepatitis

Type I and type II NKT cells can play opposing roles in liver inflammation (69). Type II NKT cells are more frequent in human liver than type I NKT cells (16). In liver ischemicreperfusion injury or a conanavalin A-induced hepatitis model, type I NKT cells were rapidly activated and elicited liver inflammation. This was inhibited by sulfatide-activated type II NKT cells through the activation of plasmacytoid DCs and the production of IL-12 and MIP-2 that induced anergy in type I NKT cells (70, 71). In another study, type II NKT cells inhibited alcohol-induced liver disease in a similar manner (72). Interestingly, sulfatide-reactive type II NKT cells that express IL-13Rα2 were detected in human liver and suggested to play a role in the protection from liver fibrosis (73). Thus, sulfatide activated type II NKT cells regulate pro-inflammatory type I NKT cells in liver inflammation.

#### Infectious Diseases

A role for NKT cells in infectious diseases is well established (74). In fact, as described above, bacteria and viruses can stimulate type I NKT cells without TCR engagement in an innate-like manner. Data suggest that also type II NKT cells can be activated by infections in an innate-like manner, without TCR stimulation. The two NKT subsets exert an opposing role in response to certain infectious agents. In Trypanosoma cruzi infected mice, contrary to type I NKT cells, type II NKT cells showed a proinflammatory effect by reducing the titers of pathogen-specific antibodies (75). By contrast, an opposing role was shown in Schistosoma mansoni infection where type II NKT cells skewed the profile to Th2 cytokine secretion with decreased IFN-γ, here type I NKT cells supported IFN-γ secretion (76). We have demonstrated a protective role of sulfatide-activated type II NKT cells in a murine model for Staphylococcus aureus sepsis accompanied with decreased TNF-α and IL-6 (77). Further, as mentioned earlier, type II NKT cells could be activated by the glycolipid components from Corynebacterium glutamicum or Mycobacterium tuberculosis (18) and phosphatidylglycerol from Listeria monocytogenes (32). Studies have also established a role for NKT cells in antiviral immune responses. Earlier reports have shown that CD1d- and NKG2D-mediated activation of type II NKT cells in HBV infection caused liver damage in mice (78, 79). However, it was more recently found that modified self-lipids such as phosphatidylethanolamine and lysophosphatidylethanolamine, produced during HBV infection, Singh et al. Type II NKT Cells

induced activation of liver type II NKT cells that enhanced antiviral immune responses, demonstrating the dual role of type II NKT cells in this infection (28). Evidence for a protective role of type II NKT cells has also been shown in human immunodeficient virus-1 (HIV-1) infection, where sulfatideinduced type II NKT cells reduced viral replication in humanized severe combined immunodeficiency (SCID-Hu) mice (80), and induced type I NKT cell anergy (81).

#### Graft vs. Host Disease (GVHD)

GVHD is a serious complication that can follow bone marrow transplantation as a treatment for hematological malignancies. Type II NKT cells have been implicated in the regulation of GVHD. Using a mouse GVHD model, it was shown that type II NKT cells in donor bone marrow protected recipient mice from GVHD by producing IFN-γ, which induced apoptosis of donor T cells, and IL-4, that deviated the immune response toward a protective Th2 type (82). In contrast, in this model, type I NKT cells did not have an effect. In this context it is interesting to note that type II NKT cells are more frequent than type I NKT cells in human bone marrow. Human bone marrow derived type II NKT cells displayed a TH2 cytokine profile and suppressed mixed lymphocyte reactions (83).

#### Gaucher's Disease (GD)

GD is an inherited metabolic disease caused by a deficiency of lysosomal glucocerebrosidase, characterized by progressive lysosomal storage of β-glucosylceramide and glucosylsphingosine (84) with an increased cancer risk (85). A functionally unique subset of type II NKT cells, reactive to βglucosylceramide and glucosylsphingosine, was identified in wild type mice and healthy humans (26). These cells constitutively expressed a T-follicular helper phenotype and provided efficient B cell help. Strikingly, these cells were activated and expanded in human GD and its murine disease model, and it was speculated that they might contribute to the increased risk of B cell malignancy observed in GD (26).

#### CONCLUSIONS AND FUTURE PERSPECTIVES

Studies so far have well documented the diverse roles of type II NKT cells in different immunological contexts. The

#### REFERENCES


picture that emerges is of a population with diverse functions and phenotypes, substantially more heterogeneous than type I NKT cells. Recent publications have further established that type II NKT cells are activated and potentially involved in several human diseases, while their targeting in mouse disease models have provided promising results. Type II NKT cells are activated by a range of lipid antigens, and although different type II NKT TCR recognize the same lipids, the TCR repertoire is oligoclonal in nature. One of the greatest challenges is to further define the entire population of type II NKT cells. The identification of additional ligands that can be used to identify type II NKT cells with CD1dtetramers will be instrumental in this quest. Considering the recognition of several lipids by the same type II NKT cell TCR, it will be of importance to elucidate the role of ligand specificity vs. a more promiscuous CD1d-reactivity for type II NKT cell development and activation. The relative role of TCR mediated/adaptive activation and innate, TCRindependent/cytokine mediated stimulation will also be critical to determine. So far type I NKT cells have been targeted in several clinical trials. Considering that type II NKT cells are more abundant in humans, and have been shown to play an immunoregulatory role in several diseases, the type II NKT cells have a largely unexplored immunotherapeutic potential. We have great expectations that the coming years will see rapid progress in this field.

## AUTHOR CONTRIBUTIONS

AKS, PT, and SC made intellectual contribution to the work. AKS and SC wrote the text. PT critically read the manuscript.

#### FUNDING

This work was supported by grants from the Swedish Cancer Foundation, the Swedish Research Council, the Swedish Brain Foundation, the Inga-Britt and Arne Lundberg Research Foundation (SC), and NEURO Sweden, the Edith Jacobson Foundation, the Swedish Foundation for MS Research, Wilhelm and Martina Lundgren's Science Foundation, Petrus och Augusta Hedlunds Stiftelse and The Royal Swedish Academy of Sciences (AKS).


tissue inflammation, insulin resistance, and hepatic steatosis in obese mice. Proc Natl Acad Sci USA. (2012) 109:E1143–52. doi: 10.1073/pnas.12004 98109


**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 Singh, Tripathi and Cardell. 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.

# Role of CD1d- and MR1-Restricted T Cells in Asthma

Chiaki Iwamura<sup>1</sup> and Toshinori Nakayama<sup>2</sup> \*

<sup>1</sup> Division of Immunology, Boston Children's Hospital, Boston, MA, United States, <sup>2</sup> Department of Immunology, Graduate School of Medicine, Chiba University, Chiba, Japan

Innate T lymphocytes are a group of relatively recently identified T cells that are not involved in either innate or adaptive immunity. Unlike conventional T cells, most innate T lymphocytes express invariant T cell receptor to recognize exogenous non-peptide antigens presented by a family of non-polymorphic MHC class I-related molecules, such as CD1d and MHC-related molecule-1 (MR1). Invariant natural killer T (iNKT) cells and mucosal-associated invariant T (MAIT) cells quickly respond to the antigens bound to CD1d and MR1 molecules, respectively, and immediately exert effector functions by secreting various cytokines and granules. This review describes the detrimental and beneficial roles of iNKT cells in animal models of asthma and in human asthmatic patients and also addresses the mechanisms through which iNKT cells are activated by environmental or extracellular factors. We also discuss the potential for therapeutic interventions of asthma by specific antibodies against NKT cells. Furthermore, we summarize the recent reports on the role of MAIT cells in allergic diseases.

Keywords: CD1d, MR1, asthma, invariant NKT (iNKT) cells, mucosal-associated invariant T (MAIT) cells

# INTRODUCTION

Innate-like T cells (CD1-restricted T cells or MHC-related molecule-1 (MR1)-restricted T cells) are classified as innate lymphoid cells that have features similar to those of the cells involved in acquired immunity, such as T cell receptor (TCR) expression (1). However, their TCR repertoire is very limited, and they recognize self or exogenous non-peptide antigens presented by a family of non-polymorphic and MHC class I-related molecules (1).

NKT cells are characterized by the expression of TCRs with a limited repertoire, consisting of Vα14 and Jα18 (in mice) or Vα24 and Jα18 (in humans) (2). In addition, their sets of Vβs are also skewed toward mainly Vβ8.2 (in mice) and Vβ11 (in humans). Since NKT cells have limited TCRs, they are called invariant natural killer T (iNKT) cells. α-galactosylceramide (α-GalCer) presented by CD1d is the most potent and well-analyzed ligand that activates iNKT cells (2). Activated iNKT cells regulate various immune responses to protect us from tumors or infectious diseases (3, 4). However, these cells can also contribute to chronic inflammatory disease, such as allergic inflammation and autoimmune responses (5, 6).

Like iNKT cells, mucosal-associated invariant T (MAIT) cells express a semi-invariant TCR with a unique TCRα chain (Vα19-Jα33 in mice, Vα7.2-Jα33 in humans) and a restricted set of TCRβ chains (7). MAIT cells are activated by a bacterial riboflavin derivative presented by MR1 (8). Although MAIT cells have been suggested to play a role in antibacterial immunity through sensing MR-1-bound microbial products, it has been speculated that these cells may also be involved in regulating beneficial host commensal interactions in the intestine and potentially in the lung (9, 10). As well as participating in antimicrobial immunity, MAIT cells may be involved in the control of chronic inflammation (11).

#### Edited by:

Luc Van Kaer, Vanderbilt University, United States

#### Reviewed by:

Maria Leite-de-Moraes, INSERM U1151 Institut Necker Enfants Malades Centre de Médecine Moléculaire (INEM), France Seddon Y. Thomas, National Institute of Environmental Health Sciences (NIEHS), United States Shin-ichiro Fujii, RIKEN Center for Integrative Medical Sciences (IMS), Japan Rosemarie DeKruyff, Stanford University, United States

#### \*Correspondence:

Toshinori Nakayama tnakayama@faculty.chiba-u.jp

#### Specialty section:

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

Received: 01 April 2018 Accepted: 06 August 2018 Published: 28 August 2018

#### Citation:

Iwamura C and Nakayama T (2018) Role of CD1d- and MR1-Restricted T Cells in Asthma. Front. Immunol. 9:1942. doi: 10.3389/fimmu.2018.01942

**140**

Another interesting feature of these MAIT and iNKT cell populations is their memory-like phenotype (12). They are able to produce effector cytokines, cytolytic molecules, and growth factors at early time points of immune responses. Therefore, they are considered to play an essential role in the host defense immune responses. Both populations have distinct and characteristic tissue localization, as NKT cells reside in the thymus, spleen, lung and liver, while MAIT cells are preferentially found in the gut lamina propria, lung and liver (13, 14). NKT cells are relatively abundant in mice and show a lower frequency in humans, whereas the opposite situation is true for MAIT cells (15).

Upon activation, iNKT cells produce a large amount of both Th1 and Th2 cytokines in addition to inflammatory cytokines, such as interleukin (IL)-17 and tumor necrosis factor α (TNFα). iNKT cells show heterogeneity in their transcriptional factors. Three different subsets of iNKT cells have been shown to produce distinct cytokines, defined as NKT1, NKT2 and NKT17 (16, 17). iNKT cells are also distinguished by their surface molecules, such as CD4 and IL-17RB. CD4+IL-17RB<sup>+</sup> iNKT cells in particular produce IL-13, IL-9, IL-10, IL-17A, and IL-22 (18). Since these cytokines exert different immunoregulatory functions, certain populations of iNKT cells might contribute to the development of chronic disorders, such as allergic diseases (18). MAIT cells produce IFN-γ, IL-4, IL-17, and TNFα (12, 19). Lepore et al. suggested that IL-5 and IL-13 can be produced by MAIT clones (20). It would be possible to identify distinct subsets of MAIT cells, that can produce specific cytokines such as is observed in the case of iNKT cells.

Asthma is a chronic inflammatory disease in the lung that causes recurring periods of wheezing, chest tightness, shortness of breath and coughing. It is well established that allergic asthma is induced by Th2 cell-mediated immune responses (21). Studies by our group and other authors have revealed that chronic airway inflammation in asthma patients is caused by pathogenic memory Th2 cells, which express high levels of IL-33 receptor ST2 and have a CD161highCRTH2high phenotype in human (22– 25).

Memory T cells are considered to play a beneficial role by responding immediately and strongly to the secondary invasion by the same antigen of a microorganism (26). However, memory T cells can induce adverse effects in cases of chronic inflammatory disease if they respond to allergens or self-antigens repeatedly for a long duration (27). Thus, allergen-specific memory Th2 cells, particularly the pathogenic subpopulation of ST2<sup>+</sup> Th2 cells paly important roles in the pathogenesis of IL-5-induced eosinophilic inflammation and fibrotic responses (24, 28). However, it is also recognized that asthmatic patients show heterogeneous phenotypes, including so-called type-1 and type-2 mixed inflammation with neutrophilic infiltration. Thus, other cell types, such as NKT cells likely contribute to the development or exacerbation of asthma (29).

This review describes and discusses the immunoregulatory roles of innate-like T cells in asthma in animal models and human patients.

#### BENEFICIAL AND DETRIMENTAL EFFECTS OF iNKT CELLS FOR ALLERGIC ASTHMA

Many investigators have tried to determine the roles of iNKT cells in asthma over the past 20 years. To this end, Akbari et al. assessed OVA-induced airway hyperactivity (AHR) and allergic airway inflammation in iNKT cell-deficient Jα281 knockout (KO) and CD1d KO mice (30). They noted a significant defect in the development of AHR and inflammation in these NKT cell deficient mice. The defects were corrected by the adoptive transfer of iNKT cells in an IL-4- and IL-13-dependent manner. Therefore, iNKT cells were considered to contribute to the development of AHR and airway inflammation independent of Th2 cells. In addition, the same group showed that nonclassical NKT cells, which are restricted to a β2-microgloblinindependent form of CD1d, also contribute to the development of AHR (31). Woo et al. further suggested that iNKT cells are also required for the generation of Th2 cells by recruiting CD103<sup>+</sup> dendritic cells (DCs) to the lung via the XCL1-XCR1 axis (32). Furthermore, another group suggested that iNKT cells act as an adjuvant to enhance allergic asthma, as systemic iNKT cell activation by α-GalCer administration or adoptive transfer of iNKT cells before OVA challenge significantly augmented the Th2 inflammatory responses (33). These results indicate that iNKT cells have detrimental effects in allergic asthma.

Simultaneously, other groups reported experimental results indicating that iNKT cells are not involved in the development of allergic asthma. OVA-induced allergic inflammation was not reduced in CD1d-deficient mice or β2-microgloblin KO mice lacking iNKT cells (34, 35). Moreover, a protective role of iNKT cells in allergic asthma was suggested. Subsequent AHR in these models can be suppressed by the systemic activation of iNKT cells by α-GalCer treatment or the transfer of α-GalCer-loaded bone marrow-derived DCs before OVA challenge in an IFN-γdependent manner (36, 37). In addition, Grela et al. reported that IFN-γ-producing iNKT cells stimulated with toll like receptor (TLR) 7 agonist (R848) attenuated allergic asthma, which is consistent with the finding that TLR7 stimulation not only enhances viral responses but also alleviates experimental asthma (38).

Thus, iNKT cells display either beneficial or detrimental effects in allergic asthma. These conflicting effects may be due to the various cytokine production patterns of iNKT cells under different conditions. IL-4 or IL-13 production from iNKT cells is required for the development of allergic asthma in mouse models, while iNKT cells can produce IFN-γ, which can suppress the Th2 response and thereby prevent allergic asthma. However, even when employing similar protocols, different institutes obtained completely different findings (33, 36, 37). Since iNKT cells can detect bacterial components through their invariant TCRs or

**Abbreviations:** MR1, MHC-related molecule-1; TCR, T cell receptor; iNKT, invariant natural killer T; α-GalCer, α-galactosylceramide; MAIT, mucosalassociated invariant T; IL, interleukin; TNFα, tumor necrosis factor α; AHR, airway hyperactivity; KO, knockout; TLR, toll like receptor; Myl, myosin light chain; HDE, house dust extract; TIM-1, T cell immunogloblin and mucin domain-1; PtdSer, phosphatidylserine; TSLP, thymic stromal lymphoprotein; DCs, dendritic cells; BALF, bronchoalveolar lavage fluid.

Toll-like receptors, the difference in the lung microbiota may affect the function of distinct iNKT cell subsets, such as NKT1 and NKT2.

Although inducing Th1 bias by iNKT cell activation may result in the inhibition of AHR and eosinophilic infiltration, our recent study shed light on how NKT cell activation can suppress Th2 type inflammation. While immunological memory plays a central role in providing protection against infection or cancer, antigen-specific memory CD4 T cells contribute to the pathogenesis of allergic and autoimmune disorders by recognizing allergens or self-antigens (24, 39). Our data showed that the activation of iNKT cells with α-GalCer during the memory phase resulted in the downregulation of IL-4, IL-5, and IL-13 and up-regulation of IFN-γ in memory Th2 cells (40). These functionally altered memory Th2 cells display a decreased capability to induce Th2 cytokines and eosinophilic airway inflammation. We therefore concluded that activated iNKT cells directly regulate memory Th2 cell function in vivo. Chang et al. showed another inhibitory mechanism for allergic disorder by iNKT cells. They found that influenza infection in neonates helped prevent allergic asthma by inducing CD4negCD8neg iNKT cell activation, which is associated with the expansion of regulatory T cells (41). The inhibitory effect required T-bet and TLR7 expression in iNKT cells. Furthermore, the administration of α-GalCer or glycolipid derived from Helicobacter pylori to neonates recapitulated the result (41), suggesting that infection with certain microorganisms can prevent the subsequent development of allergic asthma by expanding a specific subset of iNKT cells. Therefore, the authors proposed that treatment of children or allergic patients with compounds such as α-GalCer or other glycolipids derived from microorganisms might be effective in preventing or improving the development or symptoms of allergic asthma.

## LUNG iNKT CELL-DEPENDENT ALLERGIC OR NON-ALLERGIC ASTHMA

Lung iNKT cells are relatively abundant compared to iNKT cells in the peripheral blood (14). The activation of pulmonary iNKT cells by the intranasal α-GalCer administration rapidly induced AHR and eosinophilic inflammation in naïve mice, and this effect was independent of conventional CD4 T cells (42). Michel et al. showed that NK1.1neg iNKT cells produced high levels of IL-17 and induced neutrophilic infiltration following the intranasal administration of α-GalCer in a murine model (43). In addition, the development of AHR was observed in non-human primates by the direct activation of pulmonary iNKT cells with α-GalCer, indicating that pulmonary iNKT cells are critical effector cells in these animal models (44). Our previous study showed that α-GalCer induced AHR and neutrophilic infiltration, and the neutrophilic infiltration was significantly attenuated in CD69 deficient mice, indicating that activated iNKT cells-mediated asthmatic responses were dependent on CD69 expression (5). We recently identified myosin light chain (Myl) 9 and Myl12 as functional ligands for CD69 (45). We also showed that the interaction between CD69 on Th2 cells and Myl9 expressed on the luminal side of endothelial cells in the blood vessels recruits activated Th2 cells to the inflammatory site, resulting in airway inflammation (45, 46). CD69 on iNKT cells might therefore induce the migration of iNKT cells to the lung by binding to Myl9 or Myl12 and also play a critical role in the development of AHR and airway inflammation (**Figure 1**).

Even if iNKT cell activation in the lung does contribute to asthma, we are unlikely to be exposed to α-GalCer, a component of marine sponge, in our daily lives. Several studies have indicated that substances naturally existing in our environment, such as allergens, pathogens and air pollution, might activate iNKT cells and cause or exacerbate airway inflammation. Glycolipids from bacteria, such as Sphingomonas, Borrelia, and Leishmania species, are recognized by invariant TCR of iNKT cells (47). In particular, glycolipids purified from Sphigomona cell walls were shown to induce rapid AHR after respiratory administration in wild-type mice but not iNKT-deficient mice (42). Although a glycolipid that can induce iNKT cell activation has not been identified in viruses, Kim et al. suggested that viruses may facilitate CD1d antigen presentation and induce iNKT cell activation in an indirect manner (48). The authors also showed that IL-13 production from macrophages stimulated by iNKT cells during respiratory virus infection induces the development of AHR and mucus production independent of the adaptive immune response. Aspergillus fumigatus is a saprophytic fungus that is ubiquitous in the environment and is commonly associated with allergic asthma (49). Albacker et al. reported that the Aspergillus funmigatus-derived glycosphingolipid asperamide B directly activates iNKT cells in a CD1d-restricted, Myd88 independent, and dectin-1-independent manner (50). The intranasal administration of asperamide B rapidly induced AHR and neutrophil infiltration into the lung, suggesting that fungi can contribute to the induction of asthmatic symptoms by iNKT cells. Therefore, iNKT cells activated by glycolipids from microorganisms may contribute to the development and exacerbation of asthma symptoms in humans.

It was recently revealed that non-glycolipid stimulation could also activate iNKT cells, resulting in the induction of AHR. House dust extract (HDE) contains antigens and is capable of inducing airway inflammation by activating mouse Vα14 or human Vα24 NKT cells (51). The stimulation of mouse Vα14 iNKT cells was shown to be CD1d-dependent and not dependent on TLR agonist present in HDE. Although the antigen in HDE remains incompletely characterized, the authors suggested that the immunostimulatory material in HDE was of neither bacteria nor glycolipid origin (51). Ozone is an air pollutant that has also been reported to be associated with asthma (52, 53). The development of AHR was found to be inducible even in healthy individuals following exposure to ozone, which causes airway epithelial damage, and increased numbers of neutrophils (54, 55). Furthermore, asthmatic patients are more susceptible to the detrimental effects of this pollutant. A murine model of ozone induced-asthma revealed the indispensable role of IL-17 producing iNKT cells for the induction of AHR (56). Although how ozone activates iNKT cells is unclear at present, NKT cells activated by ozone can induce a form of asthma that is characterized by cellular infiltration and AHR.

or both in the airway by producing cytokines.

In addition to the naturally existing molecules in the environment, extracellular factors are also known to activate iNKT cells. T cell immunoglobulin and mucin domain-1 (TIM-1) is an important asthma susceptibility gene and also a receptor for phosphatidylserine (PtdSer) (57), an important marker of cells undergoing programed cell death or apoptosis (58). NKT cells can activate, proliferate, and produce cytokines through recognition of PtdSer by TIM-1 (59). Furthermore, the apoptosis of airway epithelial cells activates pulmonary NKT cells, resulting in AHR and suggesting that TIM-1 serves as a pattern recognition receptor on NKT cells that senses PtdSer on apoptotic cells as a damage-associated molecular pattern (60). Previous studies have shown that apoptosis induced by virus infection or ozone exposure can trigger NKT activation (48, 56, 60), as infection with some viruses triggers apoptosis and externalization of PtdSer. In addition, it has been reported that TLR signaling enhances the activation of iNKT cells. Vultaggio et al. showed that systemic dsRNA (poly (I:C)) selectively upregulates the IL-17 production from iNKT cells activated by α-GalCer. The authors therefore expected that the exacerbation of airway inflammation might be induced by certain virus infections (61). Furthermore, several cytokines involved in the initiation and amplification of Th2 responses have been reported (62). IL-25 is capable of enhancing AHR and is produced by activated Th2 cells, epithelial cells, basophils, and mast cells (63). The administration of recombinant IL-25 induced Th2-type responses, including increased serum IgE levels, eosinophilia, pathological changes in the lung, and AHR. These symptoms induced by IL-25 were not observed in iNKT cell-deficient mice (64, 65). Moreover, iNKT cells expressing IL-17 receptor B were shown to be essential for IL-25-induced AHR using an adoptive transfer model (65). Thymic stromal lymphoprotein (TSLP) is also considered to play an important role in the iNKT cell-dependent asthma model (66). While the targets of TSLP are T cells, mast cells, basophils, and DCs, Nagata et al. demonstrated that TSLP also acts on iNKT cells to enhance AHR by up-regulating their production of IL-13 (67). IL-33 enhanced the production of Th1 and Th2 cytokines in activated NKT cells (68, 69). These results indicate that natural ligands in the environments act as antigens for iNKT cells to induce allergic asthma, and TCR-independent stimuli to iNKT cells may exacerbate the asthmatic symptoms such as AHR (**Figure 1**).

Although it is obvious that the direct activation of lung iNKT cells causes lung inflammation, which types of inflammation are induced is still controversial. Two groups claimed that the intranasal administration of α-GalCer induced allergic airway inflammation because eosinophil infiltration into the lung, a feature of type 2-mediated responses, was observed in IL-4- and IL-13-dependent manners (42, 70, 71). However, neutrophil infiltration, which represents non-allergic airway inflammation, is frequently observed in severe or Th17-mediated asthma (72, 73). The activation of iNKT cells by the intranasal administration of α-GalCer, asperamide B or PtdSer induces pulmonary neutrophil infiltration, suggesting that iNKT cell may contribute to non-allergic airway inflammation (5, 43, 50, 59). In contrast, equivalent numbers of eosinophils and neutrophils have been noted with ozone or poly (I:C) stimulation (56, 61). This discrepancy in outcomes may be due to the activation of distinct subsets of iNKT cells: one produces IL-13 and IL-5, which activate and recruit eosinophils; the other produces IL-17, thereby inducing the recruitment of neutrophils. Additional flow cytometry single cell analyses addressing the precise production profiles of cytokines in iNKT cells are needed in order to discriminate the infiltrated subsets.

With many clinical and experimental examinations, it has been revealed that asthma is more heterogeneous and complex than previously thought. While allergic asthma is induced by allergens and mediated by Th2 cells, a non-allergic form of asthma is caused independent of Th2 responses (29). Nonallergic asthma is induced by multiple environmental factors, such as air pollution (smoke, ozone, and diesel particles) and virus infection. Although the immunological pathways of nonallergy asthma are still unclear, the activation of iNKT cells with their specific ligands or cytokines may contribute to the development of non-allergy asthma.

Taken together, these findings suggest that different types of iNKT cell ligands may activate distinct subsets of iNKT cells, thereby resulting in distinct patterns of airway inflammation. Therefore, lung iNKT cell activation may contribute to the development of various types of asthmatic inflammation (**Figure 1**).

#### THERAPEUTIC INTERVENTION FOR iNKT CELL-DEPENDENT ALLERGIC ASTHMA

As we pointed out above, iNKT cells may play have a critical role in the development or exacerbation of asthma. Although further investigations are needed, Dimaprit (H2 histamine receptor agonist) or intravenous immunoglobulin treatment does appear to suppress iNKT cell-dependent allergic asthma (74, 75). The administration of anti-mouse CD1d monoclonal antibodies (20H2) or CD1d-dependent antagonist has also been shown to suppress OVA-induced AHR and inflammation in murine models (76, 77). Indeed, McKnight et al. reported that anti-mouse CD1d monoclonal antibody (20H2) treatment before the intranasal administration of α-GalCer impaired iNKT cell-induced AHR in an experimental mouse model of asthma, while this antibody did not suppressed OVAinduced allergic asthma. These results suggest that this antibody may attenuate non-allergic asthma (35). Anti-human CD1d antibody (NIB.2) possesses a high affinity for human and cynomolgus macaque CD1d and inhibits NKT cell activation by inhibiting the interactions of the TCRβ chain of iNKT cells with CD1d (78). NIB.2 treatment significantly reduced the cytokine levels and numbers of lymphocytes and macrophages in the bronchoalveolar lavage fluid (BALF) in a primate model of asthma (78). However, this antibody may affect other CD1drestricted T cells that are not involved in airway inflammation (79). Therefore, the development of a more specific method will pave the way for therapeutic interventions to alleviate symptoms.

Mouse invariant monoclonal antibody, NKT14 was found to specifically bind to invariant TCR of mouse iNKT cells and deplete iNKT cells in mice via antibody-dependent cellular cytotoxicity and complement-dependent cytotoxicity for 3 weeks (80). The elimination of iNKT cells was sufficient to prevent murine AHR and pulmonary eosinophilic inflammation elicited by the oropharyngeal inhalation with α-GalCer (71). In addition, NKT14 administration prior to sensitization abrogated either antigen-mediated AHR alone or both AHR and pulmonary inflammation (71, 80).

#### ROLE OF iNKT CELLS IN ASTHMA PATIENTS

In order to determine the role of iNKT cells in human asthma, many investigators have examined iNKT cells in asthma patients with regard to their numbers and the production of cytokines (**Table 1**). An initial report published in 2006 by Akbari et al. stated that more than 60% of CD4 T cells in the BALF from severe asthmatic patients were iNKT cells, while the infiltration of iNKT cells was not observed in patients with other pulmonary diseases, such as sarcoidosis, or in healthy controls (81). Three other supportive reports showed that asthmatic patients display higher frequency of iNKT cells in BALF as compared to healthy control donors do (82, 83, 91). However, the very high numbers of iNKT cells (∼60% among CD4 T cells) reported by Akbari et al. have not been replicated by other investigators.

In contrast, a similar study by other group found that the number of iNKT cells was not increased in patients with asthma (91). Another group reported that iNKT cells were found in low numbers in the sputum or BALF of patients with asthma, chronic obstructive pulmonary disease and healthy controls, with no significant differences among the three groups (84). Mutalithas et al. also reported similar results in the BALF (85).


TABLE

1


iNKT

cells

in

patients

with

asthma.


TABLE 1 |

Continued


Frontiers in Immunology | www.frontiersin.org

TABLE

1


Continued

**147**

BALF,

bronchoalveolar

 lavage fluid; PBMCs, peripheral blood mononuclear

 cells; HCs, healthy controls.

Furthermore, the influx of iNKT cells in the airways was not observed after segmental allergen challenge (92, 93). To explain this discrepancy, Thomas et al. (93) and Vijayanand (84) pointed out that 6B11 antibody was able to stain alveolar macrophages nonspecifically. They suggested that the higher frequency of iNKT cells was due to the non-specific binding to the cells, and that the lymphocyte population should be gated for the analysis of iNKT cells (83, 91). However, Akbari et al. argued that they had already gated the lymphocyte population and used a CD1d-tetramer instead of 6B11 antibody to stain iNKT cells. In addition, those authors readdressed the issue regarding the number of iNKT cells in BALF from patients with severe asthma the next year (86). They confirmed that patients with severe asthma had a significantly increased number of iNKT cells compared to healthy controls. In this report, however, CD1drestricted iNKT cells accounted for 2–7% of total CD3<sup>+</sup> cells in the BALF of asthmatic patients, and only 1 patient with severe asthma had an iNKT cell proportion of 64.5% (93). The findings of Reynolds et al. supported the increase in the number of iNKT cells in the lung using biopsies with allergen challenge (94). Nevertheless Brooks et al. subsequently suggested that the high frequency of iNKT cells detected in BALF was due to the nonspecific staining of dead cells (87). In addition, they also indicated that there was no marked difference in the frequency of 6B11<sup>+</sup> iNKT cells in sputum even when including dead cells in the samples.

After 2010, it was suggested that a reduced iNKT cell frequency in the PBMCs of asthmatic patients did not imply that iNKT cells were irrelevant to the development of asthma. Koh et al. showed that the numbers of NKT cells in peripheral blood did not differ markedly between patients and control groups (88). However, in sputum, the numbers of iNKT cells were significantly increased in patients with asthma. Their subsequent study demonstrated the negative correlation between blood iNKT cell number and eosinophils, cytokines, or chemokines in sputum (95). These results suggested that iNKT cell might be mobilized to the lung during the exacerbation. Two other groups also demonstrated the profound reduction or no increase in iNKT cells in the blood of asthma patients compared to the normal control group (89, 90). However, they also showed an increased IL-4 production in iNKT cells of asthma patients compared to controls. Pedroza showed that pediatric asthmatic patients undergoing exacerbations of asthma displayed increased numbers of iNKT cells in the blood that also produced less IFN-γ and more IL-4 than children with stable asthma or in healthy control children (96). These results suggest that Th2-like iNKT cells might be involved in the development of asthmatic exacerbations.

At present, studies on iNKT cells in asthma patients have provided conflicting results. The frequency of iNKT cells in the lungs is particularly hotly debated. As such, we conclude that the frequency of iNKT cell does not always reflect the severity of the diseases. Although there are some recent reports that suggest no correlation between the blood iNKT cell number and clinical asthma severity (97), it is becoming more widely recognized that iNKT cells likely play a role in the development and possibly exacerbation of allergic asthma. In addition, the studies of iNKT cells in other asthma etiologies, such as chronic, occupational, steroid-resistant, exercise-induced, and aspirininduced asthma, where Th2 cells may not paly a major role, may provide new insights into these type of diseases. We therefore suggest a few experimental design approaches to adopt when studying the role of iNKT cells in particular diseases. First, in the flow cytometry analysis of iNKT cells in patients, lymphocytes, particularly live cells, should be gated for the analysis, and control staining, including with isotype controls, should be performed, with the results compared. This will prevent the contamination of cells with non-specific staining patterns. Second, more than two staining protocol should be employed. At least three different approaches have been established for identifying iNKT cells, such as CD1d-tetramer, anti-Vα24 antibody and 6B11 antibody recognizing the CDR3 region of Vα24-JαQ TCR. Although these approaches should theoretically provide similar results, using multiple staining protocols may help clear up any confusion if controversial results are obtained. Third, in addition to assessing the frequency of iNKT cells, their cytokine production (IL-2, IL-4, IFN-γ, or IL-17) should also be examined by flow cytometry. As we discussed above, it would be difficult to demonstrate the relevance of iNKT cells to diseases by analyzing only the frequency of such a small population. Examining changes in their function may therefore be useful for elucidating their contribution to the pathology of diseases.

#### MR1-RESTRICTED CELLS

MAIT cells are a subset of innate-like T lymphocytes first described in 1999 (98). These MR-1-restricted cells are abundant in humans and can rapidly express a variety of pro-inflammatory cytokines (12). While iNKT cells are suggested to play critical roles in murine models of allergic airway diseases, they are rare in human airways. MAIT cells, by contrast, are 5- to 10-fold more abundant in humans than in mice (15). Since MAIT cells exist in the lung and may be able to produce Th2 cytokines (19, 20), these cells may contribute to the development of asthma. However, several reports have indicated a different role for these cells. Hinks et al. observed a striking deficiency of Vα7.2<sup>+</sup> CD161<sup>+</sup> T cells in blood, sputum, and bronchial biopsy samples, suggesting that the deficiency correlated with the severity of asthma (11, 99). A similar deficiency in humans was observed in autoimmune diseases (systemic lupus erythematosus, rheumatoid arthritis, Crohn's disease, ulcerative colitis, or chronic inflammatory disease, such as type 2 diabetes) (100–103). In addition, it was reported that an increased MAIT cell frequency at 1 year of age was associated with a decreased risk of asthma by 7 years of age (104). These results suggest that MAIT cells may play a protective role against chronic inflammation.

Given that MAIT cells respond to bacterial metabolites, it is possible that MAIT cell activation by gut or lung microbiota is required to prevent asthma. If MAIT cells can exert a suppressive function against chronic inflammation, this hypothesis would be inconsistent with their ability to produce various inflammatory cytokines. In addition, it was also reported that the numbers of MAIT cells producing IL-17 are increased in asthmatic patients (105). Since MAIT-deficient mice have been generated (106), investigations into the function of MAIT cells infiltrating the inflammatory site in mouse models may help provide answers.

#### CONCLUSION

Studies investigating the roles of iNKT cells in allergic responses have helped to explain the Th2-dependent mechanisms underlying the development of allergic asthma. However, iNKT cells also have been suggested to be associated with the development of non-allergic airway inflammation that is induced and/or exacerbated by non-Th2 factors, such as viruses, air pollution and inflammatory cytokines (IL-17 or TNFα). Furthermore, recent studies have suggested that NKT cells or MAIT cells may play a critical role in the inhibition of asthmatic symptoms. Although a clear conclusion has not been reached due to inconsistent results, innate-like T cells apparently have

#### REFERENCES


critical and varied roles in regulating immune responses. As such, more intensive studies will be required in order to elucidate the mechanisms underlying the induction of various types of asthma by innate-like T cells and establish innovative therapeutic strategies.

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

Grants-in Aid for Scientific Research (S) (TN#26221305), the Practical Research Project for Allergic Diseases and Immunology (Research on Allergic Diseases and Immunology) from Japan Agency for Medical Research and development, AMED (JP18ek0410030) (TN).


hyperreactivity in nonhuman primates. J Allergy Clin Immunol. (2008) 121:1287–9. doi: 10.1016/j.jaci.2008.02.006


**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 Iwamura and Nakayama. 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.

# Tissue-Specific Roles of NKT Cells in Tumor Immunity

#### *Masaki Terabe\* and Jay A. Berzofsky\**

*Vaccine Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, United States*

NKT cells are an unusual population of T cells recognizing lipids presented by CD1d, a non-classical class-I-like molecule, rather than peptides presented by conventional MHC molecules. Type I NKT cells use a semi-invariant T cell receptor and almost all recognize a common prototype lipid, α-galactosylceramide (α-GalCer). Type II NKT cells are any lipidspecific CD1d-restricted T cells that use other receptors and generally don't recognize α-GalCer. They play important regulatory roles in immunity, including tumor immunity. In contrast to type I NKT cells that most have found to promote antitumor immunity, type II NKT cells suppress tumor immunity and the two subsets cross-regulate each other, forming an immunoregulatory axis. They also can promote other regulatory cells including regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs), and can induce MDSCs to secrete TGF-β, one of the most immunosuppressive cytokines known. In some tumors, both Tregs and type II NKT cells can suppress immunosurveillance, and the balance between these is determined by a type I NKT cell. We have also seen that regulation of tumor immunity can depend on the tissue microenvironment, so the same tumor in the same animal in different tissues may be regulated by different cells, such as type II NKT cells in the lung vs Tregs in the skin. Also, the effector T cells that protect those sites when Tregs are removed do not always act between tissues even in the same animal. Thus, metastases may require different immunotherapy from primary tumors. Newly improved sulfatide-CD1d tetramers are starting to allow better characterization of the elusive type II NKT cells to better understand their function and control it to overcome immunosuppression.

Keywords: NKT cell, type II NKT cell, iNKT cells, tumor immunology, immune regulation, immune network, tissuespecific immune response, tissue-resident cells

# INTRODUCTION

NKT cells are a small population of true T cells which are distinct from conventional T cells in that their receptor recognizes lipids rather than peptides and is restricted by a non-classical class I-like (class Ib) molecule CD1d (1–8). The term NKT derives from early work in which NK1.1 was used as a marker and evidence that they were early responders like innate NK cells (so "natural killer T cells"), but as many of them are NK1.1 negative, the definition has changed to require just recognition of lipid or glycolipid antigens presented by CD1d (2). This property gives NKT cells a special place in the immune system. As part of the adaptive immune system, they give the adaptive T cell immune system a way to recognize lipids, to which it is otherwise blind, as conventional MHC molecules are limited in general to presentation of peptide fragments of proteins. They also function as part of the

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by: Moriya Tsuji,*

*Aaron Diamond AIDS Research Center, United States Randy Brutkiewicz, Indiana University Bloomington, United States Shinichiro Motohashi, Chiba University, Japan*

#### *\*Correspondence:*

*Masaki Terabe terabe@mail.nih.gov; Jay A. Berzofsky berzofsj@mail.nih.gov*

#### *Specialty section:*

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

*Received: 31 May 2018 Accepted: 25 July 2018 Published: 15 August 2018*

#### *Citation:*

*Terabe M and Berzofsky JA (2018) Tissue-Specific Roles of NKT Cells in Tumor Immunity. Front. Immunol. 9:1838. doi: 10.3389/fimmu.2018.01838*

innate immune system in that they have pre-formed mRNA for cytokines, allowing rapid cytokine production, and are among the first responders in many infectious and inflammatory processes, similar to innate cells (9, 10). Their early cytokine production can potentially set the tone for other adaptive immune responses to follow.

A key discovery was the use of a semi-invariant T cell receptor (TCR) by NKT cells, characterized by semi-invariant TCRα using Vα14Jα18 gene segments together with Vβ8, 7, or 2 in mice, and Vα24Jα18 with Vβ11 in humans (**Table 1**). This was first detected in T cell hybridomas (11, 12) and subsequently in a subset of primary T cells (12–17). Thus, when the Cardell et al., first identified CD1d-restricted T cells that lacked this semi-invariant TCR as residual CD4<sup>+</sup> T cells present in MHC class II-deficient mice (18, 19), these were called non-classical or type II NKT cells and those with the semi-invariant TCR became known as classical or type I NKT cells, or iNKT cells for invariant NKT cells (2).

Another key discovery was that virtually all type I NKT cells responded to a common prototype lipid, α-galactosylceramide (α-GalCer), originally isolated from a marine sponge but probably derived from microbes symbiotic with the sponge (20) (**Table 1**). This greatly facilitated studying the function of type I NKT cells and also led to the development of a marker for these, α-GalCer-CD1d tetramers by the Kronenberg, Bendelac, and Cerundolo labs (21–23). The tetramers filled a key role since NK1.1 was no longer a meaningful marker to characterize these T cells. However, in contrast, type II NKT cells were defined by what they were not, i.e., they were CD1d-restricted T cells that did not use the semiinvariant TCR and did not recognize α-GalCer. The activity of type II NKT cells was primarily studied by comparing wild-type mice that had both types with Jα18<sup>−</sup>/<sup>−</sup> mice that lacked only type I NKT cells and CD1d<sup>−</sup>/<sup>−</sup> mice that lacked both (24). Subsequently, the Kumar lab discovered recognition of sulfatide from myelin sheaths in the central nervous system by type II NKT cells and developed sulfatide-loaded-CD1d tetramers to detect this subset of type II NKT cells (25–27). Presumably, these do not detect all type II NKT cells because the repertoire is more diverse, but they have become the best studied subset. Although production and use of sulfatide tetramers has not been as straightforward


as for α-GalCer-loaded-CD1d tetramers, they have been useful to characterize this subset of type II NKT cells, and our lab has developed more stable sulfatide-CD1d tetramers (28).

Besides dividing NKT cells into subsets based on their TCRα, whether or not they use Vα14Jα18 TCRα, NKT cells, especially type I NKT cells, can be divided based on their surface marker expression, cytokine production, and transcription factor expression. Historically, much effort has been made to use surface markers to identify functional subsets of type I NKT cells (29–34). However, there is some limitation using surface markers since some markers, such as NK1.1, are expressed in limited strains of mice. Therefore, Hogquist's lab proposed to use transcription factors, T-bet, PLZF, and RORγ-t, to distinguish functional subsets of type I NKT cells, especially in the thymus (35). They showed that analogous to CD4<sup>+</sup> T cells subsets, type I NKT cells can be divided into NKT1, NKT2, and NKT17 functional subsets that correspond to Th1, Th2, and Th17, respectively (35, 36). Now more subsets, IL-10 producing E4BP4<sup>+</sup>NKT10 (37, 38), TFH like NKTFH (39, 40) and regulatory T cell (Treg)-like Foxp3<sup>+</sup>NKTreg (41) have been reported (**Table 1**). Although their roles in tumor immunity are not clear yet, it is of interest to explore which subsets are responsible for the functions of NKT cells in tumor immunity discussed below.

#### TYPE I NKT CELLS IN TUMOR IMMUNITY

Type I NKT cells may have evolved in part to recognize certain bacterial pathogens, especially ones that lack LPS to alert the immune system, as lipids from *Sphingomonas*, *Erlichia*, and *Borrelia*, presented by CD1d, have been found to be recognized by the semi-invariant TCR of type I NKT cells (42–45). They can also be activated by LPS-expressing bacteria through non-antigenspecific signals like IL-12 (43, 46) or a combination of IL-12 and IL-18 (47). They can also ameliorate autoimmune diseases such as diabetes (48–50) or experimental autoimmune encephalitis (51), or exacerbate them, such as hepatitis (27, 52).

The focus of this review is NKT cells in cancer, where they also play a critical role. While the role of type I NKT cells in protection against autoimmunity has been mainly through production of Th2 cytokines like IL-4 and IL-13, their protection against cancer has been found to be largely dependent on production of Th1 cytokines, especially interferon-γ (IFN-γ), even though NKT cells have lytic activity and could potentially directly lyse tumors that express CD1d (53, 54). This has been observed both in studies involving treatment with α-GalCer and its analogs and in studies of spontaneous immunosurveillance without an applied treatment. Indeed, α-GalCer was observed to have potent antitumor activity (55–57) even before it was discovered to be a potent agonist for type I NKT cells (20). Dendritic cells (DCs) pulsed with α-GalCer were also found to be therapeutic against established liver metastases of the B16 melanoma and had the advantage that they were less able to induce anergy of NKT cells (58, 59). Analogs of α-GalCer that were skewed more toward IFN-γ induction, such as C-glycoside, were even more potent (60). Furthermore, the protection afforded by α-GalCer was dependent on the sequential induction of NK cells by the NKT cells, both of which could produce IFNγ and provide lytic antitumor activity (61, 62). Even rejection of tumors induced by low doses of IL-12 was found to be dependent on type I NKT cells by the failure to protect Jα18<sup>−</sup>/<sup>−</sup> mice (17, 63–65).

NKT cells have been implicated in protective tumor immunosurveillance in the absence of an applied variable, by comparing tumor formation in wild-type vs Jα18<sup>−</sup>/<sup>−</sup> mice. For example, immunosurveillance against methylcholanthrene-induced sarcomas in mice was found to be deficient in Jα18<sup>−</sup>/<sup>−</sup> mice (66). This study also found that NKT cells were critical for protection in models of immunosurveillance dependent on exogenous production of IL-12, whereas others were more dependent on NK cells. This finding is consistent with the discovery that another mechanism by which NKT cells can promote tumor immunity is by activating DCs to make IL-12, which is a potent inducer of IFN-γ (67) and to be more effective at inducing CD4<sup>+</sup> and CD8<sup>+</sup> T cells (68). The Jα18−/− mice could be reconstituted with purified wildtype liver NKT cells to protect the knockout mice, proving that the protection was dependent on NKT cells (69). Moreover, in that model, the CD4/8 double-negative subset of liver NKT cells was most protective, whereas those from thymus and spleen were less so, implying different activities of different subsets of type I NKT cells from different tissues (70). Similar findings were made for reconstitution of Jα18<sup>−</sup>/<sup>−</sup> mice with NKT cells to protect against lung metastases of a BALB/c sarcoma (71). In this case, interestingly, CD4<sup>+</sup>CD25<sup>+</sup> Tregs were induced that seemed to inhibit protection primarily by decreasing the number of NKT cells in the lungs.

NKT cells express perforin and granzymes in lytic granules, and human NKT cells have been found to use these to lyse a variety of tumor cells *in vitro* (72). In addition, a major mechanism of killing by NKT cells *in vivo* was found to be through Fas–FasL interaction (73). Nevertheless, other studies have found that a major protective mechanism of NKT cells against cancer involves production of IFN-γ and induction of other effector cells downstream, especially NK cells and CD8<sup>+</sup> T cells. For example, protection against the methylcholanthrene-induced tumors by adoptive transfer of wild-type NKT cells into Jα18<sup>−</sup>/<sup>−</sup> mice required their ability to make IFN-γ but not perforin, and on induction of NK cells that did need to be capable of making perforin (69). Moreover, sequential production of IFN-γ first by NKT cells and then by NK cells was necessary (61, 62). NK cell induction by NKT cells is rapid (74) and depends on IL-2, IFN-γ, and in some situations IL-21 (62, 75).

Thus, the major mechanisms by which type I NKT cells protect involve several pathways, production of IFN-γ, activation of DCs to make IL-12 and also be more effective antigen-presenting cells, and then downstream activation of NK cells and CD8<sup>+</sup> T cells that also make IFN-γ and mediate tumor lysis. This appears to apply to most of the α-GalCer analogs that have been studied.

An exception comes from studies in our lab which identified an unusual analog, β-mannosylceramide (β-ManCer) that differs in both the sugar (mannose instead of galactose) and the linkage (β instead of α), which appears to protect against lung metastases in mice by a different mechanism and is considered the first example of a new class of NKT cell agonists that work by a distinct mechanism (76, 77). We found that β-ManCer was a poor inducer of cytokines *in vivo* and *in vitro*, and protected even in IFN-γ−/<sup>−</sup> mice. Rather, the protection was dependent on TNF-α and nitric oxide synthase (NOS) (76). The mechanism by which TNF-α and NOS interact to protect is under study. We also found that, like certain phenyl-glycolipids (78) that can be more efficacious against cancer than α-GalCer (79), β-ManCer induces little or no long-term anergy of NKT cells, in contrast to α-GalCer that induces anergy, so that the NKT cells cannot be re-activated by re-stimulation with α-GalCer or its analogs even 2 months later (77). This is important for repeated dosing if this is to be translated to the clinic. Also, because the mechanism is different, β-ManCer synergizes with α-GalCer to protect when both are used at sub-therapeutic doses (76).

In human cancer patients, defects have been observed in type I NKT cells. Numbers of these were reduced in cancer patients compared with healthy controls in a number of solid tumors (80). Also, production of IFN-γ by NKT cells was significantly decreased in multiple myeloma patients (81). High numbers of infiltrating type I NKT cells was a predictor of overall survival in colorectal cancer (82) and low circulating levels of type I NKT cells was a predictor of poor survival in head and neck squamous cell carcinoma (83). Attempts to use α-GalCer therapeutically in advanced human cancer have not been as successful as predicted from murine studies. Initial studies of α-GalCer itself (80) or autologous DCs pulsed with α-GalCer (84–86) could increase NKT cell numbers and/or cytokine levels, and appeared safe. Expansion of patient's type I NKT cells *ex vivo* and reinfusion also was safe and increased numbers (87). However, none of these treatments resulted in any complete or partial remissions of the cancer. More recent attempts at treatment with α-GalCer-pulsed DCs have achieved prolongation of median survival in lung cancer and some partial responses in head and neck cancer (88, 89). Studies are underway to use induced pluripotent stem cells to generate large numbers of autologous NKT cells for therapy (89).

# TYPE II NKT CELLS IN TUMOR IMMUNITY

In view of all the evidence above in both mice and humans that NKT cells play primarily a protective role in cancer, it came as a surprise when we discovered that NKT cells could also suppress tumor immunosurveillance (90). A BALB/c fibrosarcoma (15- 12RM) that expressed the HIV envelope protein grew, regressed, and then recurred in almost all the mice, but failed to recur in CD1d<sup>−</sup>/<sup>−</sup> mice lacking NKT cells. We traced this to production of IL-13 by the NKT cells that induced myeloid cells (a CD11b<sup>+</sup> Gr1 intermediate population, probably a form of myeloid-derived suppressor cell or MDSC) to make TGF-β, and it was the TGF-β that suppressed the CD8<sup>+</sup> T cell-mediated protection (90, 91). Blockade of either IL-13 or TGF-β or elimination of either the NKT cell or the myeloid cell could interrupt this immunosuppressive circuit and unmask immunosurveillance, preventing the tumor recurrence. The same was true in a CT26 colon cancer lung metastasis model. A puzzle in this pathway was why IL-13 but not IL-4 was necessary, when the protection depended on both the IL-4Rα and STAT6, which are downstream of both IL-4 and IL-13 (90). The solution to this puzzle was found when it was discovered that the signal from IL-4Rα and STAT6 synergized with a signal from TNF-α to upregulate the IL-13Rα2, a second receptor for IL-13 that does not respond to IL-4, and this latter receptor, when triggered by IL-13, induced the myeloid cell to make TGF-β (92). Another concurrent study also found that immunosurveillance was revealed in STAT6-deficient mice, consistent with our findings (93, 94). Nevertheless, in another mouse tumor model, IL-4Rα, STAT6, and TGF-β were not required for immunosuppression by NKT cells, implying that other suppressive mechanisms also exist (95).

The paradox that NKT cells both promoted and suppressed tumor immunity was resolved when it was found that type II NKT cells, still present in Jα18<sup>−</sup>/<sup>−</sup> mice but absent in CD1d<sup>−</sup>/<sup>−</sup> mice, were sufficient to suppress (24). This was true in several tumor models studied in different labs, in which Tregs did not play a critical role. A similar finding that type I NKT cells promoted tumor immunity and type II NKT cells suppressed it was observed in a B-cell lymphoma model (96).

As noted above, the type II NKT cells have been difficult to study due to lack of a good surface marker, with the sulfatide tetramers being difficult to make and use. We have now made stable sulfatide-CD1d tetramers and have been characterizing the type II NKT cells which seem to be enriched in both lung and liver, two tissues that are major targets for tumor metastases (28) (also Kato, Pasquet et al., submitted). This is especially interesting as we have seen the role of type II NKT cells in lung metastasis models in syngeneic mice (24, 91). The question of tissue specificity and tissue microenvironment will be discussed further below.

We have also focused on the suppressive role of TGF-β that is the downstream immunosuppressive cytokine in this pathway. Because it is also involved in Treg induction and function and is also made by tumors, we cannot attribute all the effects of TGF-β blockade to the NKT cell immunoregulatory pathway, but some of it may be due to that pathway. Conversely, blockade of TGF-β could block multiple regulatory pathways simultaneously. It is also possible that TGF-β induced by NKT cells from myeloid cells mediates a negative feedback loop on the NKT cells themselves, as it has been shown that TGF-β downregulates antigen presentation by CD1d by antigen-presenting cells without reducing the level of surface CD1d (97). Blockade of TGF-β unmasked tumor immunosurveillance in both the 15-12RM subcutaneous fibrosarcoma model and the CT26 lung metastasis model in the absence of any other treatment (91). In the TC1 model of an HPV E6/E7 transformed tumor, and in a CT26 subcutaneous tumor model, TGF-β blockade did not protect by itself, but synergized with a vaccine to reduce tumor growth more completely than the vaccine alone (98, 99). Furthermore, blockade of only TGF-β1 and β2 isoforms, without blockade of TGF-β3, was sufficient to mediate both effects (100). This may reduce the number or types of side effects. Also, importantly, as they work by different pathways, we found that anti-TGF-β enhanced the efficacy of anti-PD1 and *vice versa* for improving vaccine efficacy. Thus, the triple combination of vaccine, anti-PD1, and anti-TGF-β gave better protection than any of the possible pairwise combinations (100). This combinatorial efficacy is also supported by recent findings that one reason for failure of anti-PD1 therapy is TGF-β that inhibits T cell entry into the tumor in urothelial and colon cancers (101, 102). A phase I clinical trial of a human anti-TGF-β antibody in advanced metastatic melanoma patients also showed some preliminary evidence of beneficial activity, including an 89% partial response and several cases of mixed response or stable disease, with surprisingly few side effects (103, 104). All these results support the use of anti-TGF-β as a novel checkpoint inhibitor alone or in combination with vaccines and anti-PD1.

# CROSS-REGULATION BETWEEN TYPE I AND TYPE II NKT CELLS

If type I and type II NKT cells generally play opposite roles in cancer immunosurveillance, and type II NKT cells are often immunosuppressive, is it possible that type II NKT cells also suppress type I NKT cells and *vice versa*? We explored this question in mouse tumor models by taking advantage of the ability to selectively stimulate type I NKT cells with α-GalCer and type II NKT cells with sulfatide (105). In a CT26 colon cancer lung metastasis model, treatment of mice with α-GalCer protected against lung tumor nodules almost completely, whereas treatment with sulfatide actually increased the number of lung nodules relative to the vehicle control, consistent with promotion of tumor immunosurveillance by type I NKT cells and suppression of endogenous surveillance by type II NKT cells (105). By contrast, lack of type I NKT cells in Jα18<sup>−</sup>/<sup>−</sup> mice led to increased tumor nodules, consistent with removal of a check on type II NKT suppression when type I NKT cells were absent, whereas CD1d<sup>−</sup>/<sup>−</sup> mice that lack both were protected. Moreover, sulfatide stimulation of type II NKT cells suppressed proliferation and cytokine secretion of type I NKT cells in BALB/c spleen cells induced by α-GalCer. This was not due to competition for binding to CD1d, because APCs could be pulsed separately with α-GalCer and sulfatide and then mixed to achieve the same result. Also, *in vivo* stimulation of both increased the ratio of IL-13/IFN-γ compared with stimulation with α-GalCer alone. The key finding was that in two mouse tumor models, the subcutaneous 15-12RM fibrosarcoma model and the CT26 lung metastasis model, when animals were treated with both α-GalCer and sulfatide, the protective effect of α-GalCer was abrogated or reduced, implying that type II NKT cell stimulation inhibits protection by type I NKT cells (105). In a concurrent study, Vipin Kumar's lab showed that sulfatide-stimulated type II NKT cells inhibited ConA-mediated hepatitis that was dependent on type I NKT cells (27).

These studies indicate the existence of an immunoregulatory axis between type I and type II NKT cells, in which they not only have polar opposite functions, but counteract each other (7, 54, 105–109). This axis is reminiscent, therefore, of the original Th1–Th2 immunoregulatory axis described by Mosmann and Coffman (110, 111) that had such a profound effect on immunology, because the cross-regulation set up a metastable state in which whichever polarized cells got a head start would suppress the other pole, and thus shift the balance to one pole or the other. Because NKT cells serve as first responders on the scene in many types of immune response, the balance along this type I–type II NKT cell immunoregulatory axis could similarly set the tone for subsequent other adaptive immune responses.

#### NETWORKING OF NKT CELLS WITH OTHER REGULATORY CELLS

#### NKT Cells and Myeloid Cells

As a part of the large immune network, NKT cells also interact with other regulatory cells such as Tregs and MDSCs. In addition to the interaction between type II NKT cells and myeloid cells suppressing tumor immunity, type I NKT cells have been reported to interact with MDSCs. MDSCs are converted from expanded immature myeloid cells generated from normal progenitors of myeloid cells and neutrophils (112). They can be further differentiated into myeloid lineage cells such as DCs and macrophages. MDSCs can be converted into immunostimulatory APCs when the MDSCs present α-GalCer and tumor antigens (113). The interaction of α-GalCer-presenting MDSCs with type I NKT cells increased expression of CD11b, CD11c, CD40, MHC II, and CD86 on MDSCs, which are markers of stimulatory APCs. Inoculation of MDSCs loaded with α-GalCer and tumor antigen induced protective tumor immunity dependent on CD8<sup>+</sup> T cells, NK cells, and type I NKT cells, but not CD4<sup>+</sup> T cells and host DCs. Consistent with changes in the surface marker expression, the treatment did not suppress CD8<sup>+</sup> T cells and did not induce Tregs. Although the detailed mechanism of conversion of MDSCs into stimulatory APCs is not clear, the effect of type I NKT cells on interacting MDSCs to induce maturation of the cells to become more stimulating myeloid cells is similar to what has been reported of interaction of type I NKT cells with α-GalCer-presenting immature DCs, which are induced to mature through CD40–CD40L interaction (114). NKT cells not only manipulate MDSCs but also allow CD8<sup>+</sup> T cells to acquire resistance against MDSCs during *ex vivo* expansion (115). On the other hand, inoculation of cell free α-GalCer can induce accumulation of MDSCs through induction of IL-33 production by Kupffer cells (116). Type I NKT cells also modulate suppressive IL-10-secreting neutrophils induced by serum amyloid-A (SAA-1), an acute phase reactant (117). Signaling by SAA-1 also conversely facilitates the modulation of the neutrophils by type I NKT cells. This interaction downmodulates IL-10 production by neutrophils through a mechanism dependent on CD1d and CD40, which induces IL-12 production. Type I NKT cells can also control myeloid cells by directly lysing them, as they express CD1d. In a primary neuroblastoma without MYC-N amplification, CD1d-expressing CD68<sup>+</sup> tumor-associated macrophages (TAMs) stimulate tumor growth through production of IL-6, and a high TAM gene signature is associated with poor prognosis (118). In a xenograft human neuroblastoma model in NOD/ SCID mice, adoptively transferred type I NKT cells selectively killed TAMs by recognizing CD1d to indirectly control tumor growth (118).

By contrast, there are some studies suggesting myeloid cells suppress type I NKT cells in some situations. High neutrophil concentration was reported to suppress type I NKT cell activation and expression of T-bet both *in vivo* in CD18<sup>−</sup>/<sup>−</sup> spontaneous neutrophilic mice and *in vitro* by co-incubating type I NKT cells with neutrophils (119).

# Treg and NKT Cells

In contrast to the relationship between type I NKT cells and myeloid cells, in which most studies demonstrated that type I NKT cells modulate suppressive tumor-associated myeloid cells, type I NKT cells facilitate and collaborate with Tregs to induce immune suppression in multiple settings including allergic asthma, type I diabetes, tolerance induction, and bone marrow transplantation (109, 120–125). Here, we focus the discussion on studies in the context of cancer.

Polyps are pathological sites of inflammation in intestine which subsequently often develop into cancers. Recently, Wang et al. reported that type I NKT cells drive Treg maintenance and/or activation in both polyps and lamina propria of APCmin/<sup>+</sup> mice, which has genetic disposition to develop spontaneous polyps (126). When APCmin−/<sup>+</sup> mice are made deficient for type I NKT cells, there is a significant reduction in the number of polyps. Type I NKT cell-sufficient APCmin/<sup>+</sup> mice have higher number of Foxp3<sup>+</sup> Tregs with more activated phenotype in polyps compared with type I NKT cell-deficient APCmin/<sup>+</sup> mice. The existence of type I NKT cells also altered a myeloid cell population from iNOS-expressing M1 dominant macrophage phenotype in type I NKT cell-deficient animals to CD206-expressing M2 macrophage phenotype and Ly6GhiLy6cint MDSCs. Recognition of commensal bacterial antigens are critical to mature and maintain type I NKT cells in mucosal tissues (119). Cell-to-cell contact of type I NKT cells, weakly stimulated with bacterial antigens or cytokines, with Tregs induces IL-10 production and upregulation of Foxp3 in Tregs to make Tregs more suppressive in humans (127). Thus, it is possible that weakly stimulated type I NKT cells residing in the intestine tend to induce activated suppressive Tregs in the tissue, and this may serve as a mechanism of gut homeostasis.

As described above, type I NKT cells support Tregs both in mice and humans, while paradoxically Tregs suppress type I NKT cell functions. Human Tregs suppress proliferation and cytokine production of type I NKT cells when type I NKT cells are stimulated with relatively weak antigens such as OCH, an α-GalCer analog, or bacterial-derived diacylglycerol *in vitro* (127). The suppression is dependent on both cell-to-cell contact and IL-10. Presumably, IL-10 production induced by cell-to-cell contact between type I NKT cells and Tregs mediates the suppression of type I NKT cells by Tregs. *In vivo*, in the context of cancer, Tregs can suppress type I NKT cells and type I NKT cell-mediated tumor immunity in a model with a methylcholanthrene-induced tumor cell line (71). In this model, vaccination with tumor antigens induced tumor antigen-specific Tregs, reduction of the number of type I NKT cells and accelerated tumor growth. The suppression of type I NKT cells by Tregs affects NKT cell-targeted immunotherapy of cancer. Blockade of Tregs by anti-CD25 mAb treatment enhances the efficacy of DCs pulsed with α-GalCer and tumor-derived antigens to induce effector CD8<sup>+</sup> T cells in a B16.OVA model (128). Similarly, transient depletion of Tregs in DEpletion of REGulatory cells (DEREG) mice, which express a diphtheria toxin receptor under a control of Foxp3 promoter/ enhancer regions, by diphtheria toxin injection enhances the therapeutic effect of an α-GalCer-loaded tumor cell vaccine in a B16F10 melanoma model (129). In contrast to type I NKT cells, interaction between type II NKT cells and Tregs is not yet understood.

#### Balance Between Type II NKT Cells and Tregs Determined by Type I NKT Cells

As we discussed above, there is cross-talk between type I and type II NKT cells forming an immunoregulatory axis. Also, there is a clear interaction between type I NKT cells and Tregs. What is the relationship between those two interactions?

In a subcutaneous CT26 colon carcinoma model, it has been shown that blockade of Tregs by anti-CD25 reveals tumor immunosurveillance that rejects transplanted tumors. The effect of anti-CD25 was similar in CD1d<sup>−</sup>/<sup>−</sup> mice that do not have either type of NKT cells. However, it was not the case in Jα18<sup>−</sup>/<sup>−</sup> mice that lack type I NKT cells but retain type II NKT cells (130), indicating that type II NKT cells are a second cell responsible for the suppression of tumor immunity in Jα18<sup>−</sup>/<sup>−</sup> mice. The inability of anti-CD25 treatment to unmask immunosurveillance in Jα18<sup>−</sup>/<sup>−</sup> mice could be reversed by two different approaches. One is blockade of antigen presentation by CD1d by treating mice with anti-CD1d mAb (130). This treatment blocks activation of type II NKT cells in Jα18<sup>−</sup>/<sup>−</sup> mice *in vivo*, and no other cells since these mice lack type I NKT cells. The second is adoptive transfer of type I NKT cells into Jα18<sup>−</sup>/<sup>−</sup> mice, allowing type I NKT cells to counter-regulate the type II NKT cells (105, 130). Both approaches made anti-CD25 treatment effective to induce tumor rejection in Jα18<sup>−</sup>/<sup>−</sup> mice. In NKT cell-sufficient mice, counter-regulation between the two types of NKT cells cancels out their functions to a large extent, allowing Tregs to be the dominant regulator of tumor immunity. It is the status of type I NKT cells that determines dominance of Treg-mediated suppression. Indeed, tipping the balance between type I and type II NKT cells by skewing toward type II NKT cell dominance without removing type I NKT cells, by specifically stimulating type II NKT cells with sulfatide *in vivo*, also made anti-CD25 ineffective to remove immunosuppression of tumor immunity (130). This is clinically relevant because, as noted above, the type I NKT cell number in peripheral blood has been reported to be suppressed in cancer patients (80, 83, 131) compared with healthy donors, and their IFN-γ-producing function is frequently diminished in patients (81, 132–135), making the balance in cancer patients more like that in Jα18<sup>−</sup>/<sup>−</sup> mice. In addition, it is believed that type II NKT cells are more prevalent in humans compared with mice. Therefore, in cancer patients, it is quite possible that imbalance of the type I–type II NKT cell axis may provide an explanation why Treg-targeted therapy development has had very limited success despite much effort invested to develop this type of immunotherapy (136).

## TISSUE SPECIFICITY OF IMMUNE REGULATION AND EFFECTOR FUNCTION IN CANCER

Tissue-resident immune cells such as innate lymphocytes and resident memory T (TRM) cells, mainly CD8<sup>+</sup> T cells, have recently been implicated in tissue homeostasis and tissue-specific immunity against pathogens and cancer. Similar to the situation with innate lymphocytes, as NKT cells have one foot in the innate immune system, distribution of NKT cell functional subsets is distinct among different tissues (31, 32, 38, 119, 137, 138). Thus, it is possible that NKT cells play a distinct role in tumor immunity in different tissues. Moreover, type II NKT cells are most prevalent in the lungs and liver, two organs that are major sites of cancer metastases (28). Type II NKT cells play a critical role in immune suppression of tumor immunity against CT26 tumors in lungs, as NKT cell-deficient CD1d−/− mice, but not type I NKT celldeficient Jα18<sup>−</sup>/<sup>−</sup> mice are resistant to tumor development (24, 107). Tregs do not play a major role in the immune regulation in this model, as anti-CD25 treatment does not alter tumor burden in the lungs (139). By contrast, Tregs rather than NKT cells play a critical role in suppression of tumor immunity against the same CT26 tumors growing in the skin. Therefore, against the same tumor, different T cells in different tissues regulate tumor immunity.

Recent studies suggest that induction of resident memory CD8<sup>+</sup> T (TRM) cells is critical for the protection against tumors in mucosal tissues. Intranasal delivery of a cancer vaccine that induces TRM is reported to reduce tumor progression of TC1 orthotopic head and neck or lung tumors in both prophylactic and therapeutic settings, while intramuscular systemic immunization does not (140, 141). The protection was not transferred from immunized mice to naïve mice through parabiosis, confirming the protective memory CD8<sup>+</sup> T cells induced at the site of induction do not go into the circulation. Consistently, it has been reported that adoptive transfer of effector CD8+ T cells induced by s.c. injection of a DC vaccine loaded with tumor antigens could protect recipients against s.c. tumors but not against gastric tumors (142). In humans, recent studies report that the number of TRM cells in tumors correlates with prolonged survival in a variety of types of cancers (143–149).

TGF-β plays a critical role in induction and retention of tissueresident immune cells. Indeed, blockade of TGF-β, a cytokine required for the differentiation of TRM cells, significantly reduced the number of TRM induced by as well as protective effect of the mucosal vaccine against lung tumors (140). This is in contrast to the effect of anti-TGF-β treatment in some lung metastasis models in which the authors examined the effect of TGF-β blockade in tumor immunosurveillance (91, 100, 150, 151) and studies showing enhancement of tumor vaccine efficacy (98, 100, 152, 153). In addition, recent studies of cancer patients suggested that TGF-β is a factor negating the effect of PD-1 blocking checkpoint inhibitor treatment (101, 102). Further studies are required to provide potential explanations for the sometimes conflicting effects of TGF-β blockade in tumor immunity.

Since immunity against CT26 tumors in different tissues is regulated by different suppressive T cells, we asked whether the effector T cells against CT26 induced in lungs and skin by removing immune suppression are cross-protective against CT26 in the other tissue (139). Mice that rejected subcutaneous tumors after receiving anti-CD25 treatment rejected tumors not only in the skin on the contralateral side but also in the lungs, suggesting that the effector T cells induced in skin go into the systemic circulation. This was further confirmed by the observation that adoptive transfer of splenic T cells from the mice immune to subcutaneous tumors transferred the protection not only against subcutaneous tumors but also against lung tumors. By contrast, splenic effector T cells in CD1d<sup>−</sup>/<sup>−</sup> mice immune to lung tumors could not transfer any protection against subcutaneous tumors, although they did against lung tumors. Thus, there is a unidirectional migration of protective memory T cells from skin to lungs (139). Protective memory T cells in this system are unlikely TRM, as the protection was transferred by splenic T cells. The observations in CT26 models were strikingly different from other studies suggesting that TRM cell induction is crucial for the protection against mucosal tumors. The difference between this study with CT26 models and others is that all other studies induced memory T cells by using vaccines while this study with CT26 tumors in different tissues did not, but rather depended on natural immunosurveillance. As local APCs at the site of priming can determine tissue specificity of activated T cells, and Tregs can affect functions of APCs, it may be possible that APC's ability to determine tissue specificity of induced T cells is changed in the absence of regulatory cells such as Tregs and type II NKT cells. The mechanism of this observation requires further elucidation.

# CONCLUSION

Here, we have seen that both type I and type II NKT cells play critical roles in tumor immunity. While most often type I NKT cells promote tumor immunity and type II NKT cells suppress it, type I NKT cells can also induce suppressive Tregs as well as protective NK cells. Moreover, both have a myriad of interactions with other immune effector and regulatory cells, forming a complex web of immune regulation. Type I and II NKT cells can cross-regulate

#### REFERENCES


each other, forming an immunoregulatory axis that comes into play early in immune responses. By regulating type II NKT cells, type I NKT cells also determine the balance between the latter and Tregs to determine which will dominate in controlling immunity within a particular tumor, regulating the regulators. This has clinical implications because when type I NKT cells are absent, both Tregs and type II NKT cells can suppress in the same tumor, and this situation mimics the situation often found in cancer patients, in which type I NKT cells are deficient in numbers or function. Thus, blockade of both simultaneously may be necessary. The finding that the immune regulation and antitumor effector mechanisms within the same tumor growing in different tissues, such as skin and lungs, are different is also clinically important. The immunotherapy required for a primary tumor may not be the best for metastases growing in a different tissue environment. Thus, understanding the complex multidimensional interactions among a large network of regulatory and effector cells, including type I and type II NKT cells, will be critical to design the most effective immunotherapies for cancer.

# AUTHOR CONTRIBUTIONS

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

# FUNDING

This work was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research (Z01-C-004020), and the Gui Foundation. We apologize to those whose work we were unable to discuss due to space limitations.


having various sugar moieties. *Biol Pharm Bull* (1995) 18(11):1487–91. doi:10.1248/bpb.18.1487


**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 Terabe and Berzofsky. 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.*

# Natural Killer T Cells and Mucosal-Associated invariant T Cells in Lung infections

*François Trottein1,2,3,4,5\* and Christophe Paget6,7*

*1Univ. Lille, U1019 – UMR 8204 – CIIL – Centre d'Infection et d'Immunité de Lille, Lille, France, 2Centre National de la Recherche Scientifique, UMR 8204, Lille, France, 3 Institut National de la Santé et de la Recherche Médicale U1019, Lille, France, 4Centre Hospitalier Universitaire de Lille, Lille, France, 5 Institut Pasteur de Lille, Lille, France, 6 Institut National de la Santé et de la Recherche Médicale U1100, Centre d'Etude des Pathologies Respiratoires (CEPR), Tours, France, 7Université de Tours, Tours, France*

The immune system has been traditionally divided into two arms called innate and adaptive immunity. Typically, innate immunity refers to rapid defense mechanisms that set in motion within minutes to hours following an insult. Conversely, the adaptive immune response emerges after several days and relies on the innate immune response for its initiation and subsequent outcome. However, the recent discovery of immune cells displaying merged properties indicates that this distinction is not mutually exclusive. These populations that span the innate-adaptive border of immunity comprise, among others, CD1d-restricted natural killer T cells and MR1-restricted mucosal-associated invariant T cells. These cells have the unique ability to swiftly activate in response to non-peptidic antigens through their T cell receptor and/or to activating cytokines in order to modulate many aspects of the immune response. Despite they recirculate all through the body *via* the bloodstream, these cells mainly establish residency at barrier sites including lungs. Here, we discuss the current knowledge into the biology of these cells during lung (viral and bacterial) infections including activation mechanisms and functions. We also discuss future strategies targeting these cell types to optimize immune responses against respiratory pathogens.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Marco Lepore, Immunocore, United Kingdom Nicholas M. Provine, University of Oxford, United Kingdom*

#### *\*Correspondence:*

*François Trottein francois.trottein@pasteur-lille.fr*

#### *Specialty section:*

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

*Received: 28 May 2018 Accepted: 16 July 2018 Published: 02 August 2018*

#### *Citation:*

*Trottein F and Paget C (2018) Natural Killer T Cells and Mucosal-Associated Invariant T Cells in Lung Infections. Front. Immunol. 9:1750. doi: 10.3389/fimmu.2018.01750*

Keywords: natural killer T cells, mucosal-associated invariant T cells, mucosal immunity, infection, lung, bacteria, viruses, immunotherapy

The respiratory tract constitutes one of the major sites of entry for pathogens. According to the vital nature of this organ, immune responses in the lung mucosa need to be finely regulated in a time- and strength-dependent manner. First, microbial-derived signals have to be rapidly integrated by the immune system in order to establish an efficient response that will allow containment and elimination of the pathogens. However, if not properly regulated, this inflammatory response can also turn into immunopathology leading to extensive tissue damages and organ failure with devastating outcomes for the host. Thus, the enrichment within the lung mucosa for cells endowed with potent immunoregulatory functions is paramount to combat infections as well as to maintain

**Abbreviations:** TCR, T cell receptor; MHC, major histocompatibility complex; Ag, antigen; NKT, natural killer T; MAIT, mucosal-associated invariant T; GSL, glycosphingolipids; α-GalCer, α-galactosylceramide; PLZF, pro-myelocytic leukemia zinc finger; Tg, transgenic; DC, dendritic cell; BCG, bacillus Calmette–Guérin; IAV, influenza A virus; Ab, antibody; WT, wild-type; CAR, chimeric Ag receptor.

tissue functions and integrity. Among these cells, the family of unconventional αβ T lymphocytes recently emerged as central player in mucosal immunity.

These unconventional αβ T cells differ from conventional αβ T lymphocytes in many respects. After thymic egress, unconventional αβ T cells are readily capable of cytokine/chemokine secretion and cytolysis. Thus, unconventional αβ T cells occupy a unique niche in the immune system (spanning the innateadaptive border of immunity). In contrast to its huge diversity in conventional T cells, the T cell receptor (TCR) repertoire of unconventional αβ T cells is fairly conserved and presents a limited number of V(D)J rearrangements. In addition, unlike conventional T cells that are restricted to the polymorphic major histocompatibility complex (MHC) class I and MHC class II molecules; unconventional αβ T cells recognize conserved antigens (Ags) [including (glyco)lipids, small metabolites, and modified-peptides] presented by the quasi-monomorphic MHC class Ib (e.g., HLA-E/Qa-1b, H2-M3) and MHC class-I like (e.g., group 1 and 2CD1, MR1) molecules.

Due to their quite low representation in rodents and the absence of specific tools to analyze them, unconventional αβ T cells have been overlooked by the scientific community for a long time. Thanks to the development of biological tools (e.g., gene-targeted mice and tetramers), the interest in understanding the functions and role of unconventional T cells in tissue homeostasis and diseases has tremendously grown over the last decades. As a result, the current literature quickly expands to the extent that unconventional T cells are now highly regarded by clinicians as attractive targets for future innovative cell-based immunotherapies. Here, we focused our attention on two major subsets of "innate-like" unconventional αβ T cells namely CD1d-restricted natural killer T (NKT) and MR1-restricted mucosal-associated invariant T (MAIT) cells. We discuss the recent advances in the functions of these cells during lung infections in mice and humans as well as the potential therapeutic opportunities based on harnessing the biology of these cells.

#### GENERAL FEATURES OF CD1d- AND MR1-RESTRICTED T CELLS

CD1d- and MR1-restricted T cells represent a heterogeneous population of cells (**Table 1**). In this section, we review the general features and biological tools available to analyze these particular subsets.

#### CD1d-Restricted T Cells

Natural killer T cells were the first member of the CD1d-restricted T cell family to be described. Originally, NKT cells were named after pioneer studies that observed the co-expression of—initially thought—NK cell-specific markers on a T cell subset (1, 2).


α*-GalCer,* α*-galactosylceramide; GSL, glycosphingolipid;* α*-GlcDAG,* α*-monoglucosyldiacylglycerol; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; PG, phosphatidylglycerol.*

Nowadays, this definition is obsolete since some NKT subsets do not necessarily express NK cell-related molecules (3, 4). By definition, CD1d-restricted T cells comprise the heterogeneous family of T cells that are restricted to the CD1d molecule [for reviews, see Ref. (5–7)]. Specifically, CD1d-restricted T cells can be further segregated into three groups called type I (or "invariant") NKT, type II (or "variant") NKT, and "others" CD1d-restricted T cells regarding their Ag specificity and TCR usage (8).

#### Type I NKT Cells

Type I NKT cells are the most characterized population of CD1drestricted T cells. Type I NKT cells express a semi-invariant TCR composed of an invariant TCRα-chain (Vα14-Jα18 in mice and Vα24-Jα18 in humans) paired with a limited set of TCRβ-chains (5, 7). Type I NKT cells are also defined by their strong capacity to react through their TCR to the glycosphingolipid (GSL) alpha-galactosylceramide (α-GalCer) once inserted into the CD1d molecule (9). In addition, the type I NKT TCR has been shown to recognize a wide range of microbial-derived lipids (6, 10–12), which allows them to specifically respond in a broad set of infectious diseases. Of importance, type I NKT cells can also activate in response to a wide array of cytokines including IL-12, IL-1β, IL-18, IL-23, and IFN-β (13–15).

Type I NKT cells are relatively abundant in mice in which they populate both lymphoid and non-lymphoid tissues. On average, type I NKT cells represent 1–3% of the T cell pool and can account for up to 20–30% of total T cells in specific niches such as the liver (8). In sharp contrast, type I NKT cells are far less frequent in humans (100-fold lower frequencies at similar locations) at the exception of the omentum where they can represent 10–20% of total CD3<sup>+</sup> cells (16). Of note, type I NKT cell subsets can be defined regarding the pattern of cytokines produced including NKT1 [IFN-γ-producing, pro-myelocytic leukemia zinc finger (PLZF)low/T-bet<sup>+</sup>], NKT2 (IL-4-producing, PLZFhigh/T-bet<sup>−</sup>), and NKT17 (IL-17-producing, RORγt +) subset (17, 18).

Upon activation, type I NKT cells can rapidly produce substantial amounts of multiple chemokines (including CCL2, CCL3, CCL4, CCL5, CCL11, and CXCL8) and cytokines such as TNF-α, IFN-γ, IL-4, IL-5, IL-10, IL-13, IL-17, IL-21, IL-22, and GM-CSF (19). This ability to produce a wide range of chemokines and cytokines (with sometimes opposite functions) depends on the nature of the stimuli (e.g., Ags and/or activating cytokines) as well as the cell subset triggered. Through this property, type I NKT cells have been shown to interact with as well as to influence the functions of numerous immune populations including neutrophils, dendritic cells (DCs), macrophages, NK cells, γδT cells, B and conventional T cells (20–28). In addition, type I NKT cells are capable to directly kill transformed and virally infected cells (28, 29).

#### Type II NKT Cells

Conversely, type II NKT cells represent a much broader family engulfing all other CD1d-restricted αβ T cells that react to lipids, but not to α-GalCer (8, 30, 31). Thus, type II NKT cells express a more diverse TCR repertoire. For instance, in mice, type II NKT cells use Vα1, Vα3.2, Vα4, Vα8, or Vα11 TCRα-chains paired with Vβ3, Vβ5, Vβ8, or Vβ11 TCRβ-chains and also comprise oligoclonal Vα3.2-Jα9/Vβ9 and Vα8/Vβ8 subsets (30). The TCR usage of type II NKT cells in humans is far less understood. However, type II NKT cells seem to be more abundant than type I NKT cells in humans (32). In line with their TCR diversity, type II NKT cells recognize Ags of various nature and origin (mammalian and microbial) including glycolipids (lyso)phospholipids and non-lipidic small molecules (33–38).

Due to the lack of specific tools to investigate their entire pool, the functions of type II NKT cells have mainly been proposed indirectly by comparing the phenotypes observed in Jα18- (which lack type INKT cells) versus CD1d-deficient (which lack both type I and type II NKT cells) mice in various settings. Despite being informative, this strategy is not without concerns for several reasons. First, this way of analysis imposes to speculate that NKT cell subsets do not regulate the functions of each other. A recent report also indicates that *Cd1d*<sup>−</sup>/<sup>−</sup> mice present a significant enrichment for mature MAIT cells in the thymus and spleen that could account for the phenotype observed in these mice (39). In addition, the original *J*α*18*<sup>−</sup>/<sup>−</sup> mouse line presented a severely impaired expression of many TRAJ segments (40). However, the recent generation of new mouse lines with specific deletion of TRAJ18 (41, 42) will allow to validate or not the pioneer experiments. Nevertheless, type II NKT cells appear to share conserved phenotypic and functional similarities with type I NKT cells including an effector memory phenotype, cytotoxicity, and secretion of numerous cytokines/ chemokines (8, 31, 43, 44). Thus, they have been proposed to participate in antimicrobial responses, auto-immunity, and cancer (31, 43, 44).

#### Other CD1d-Restricted T Cells

Type I NKT and type II NKT cells account for the vast majority of the CD1d-restricted T cell pool. However, the recent literature indicates that the diversity of this family is much more complex than initially believed (45). For instance, the identification of a minute population of CD1d/α-GalCer tetramer-positive cells in *J*α*18*<sup>−</sup>/<sup>−</sup> mice brought a new layer of complexity in the definition of NKT cells (46). These cells expressed a semi-invariant Vα10-Jα50 TCR with minimal conservation in the amino acid sequence of the TCRα chain expressed on type I Vα14-Jα18 NKT cells (46). Interestingly, α-GalCer-reactive non-type I Vα24<sup>+</sup>-Jα18<sup>+</sup>/Vβ11<sup>+</sup> NKT cells can also be observed in humans (47). This includes Vα24<sup>−</sup> Jα18<sup>+</sup>/Vβ11<sup>+</sup> subset with public TCR repertoires and diverse populations of Vα24<sup>−</sup> Jα18<sup>−</sup>/Vβ11<sup>−</sup> with private repertoires (47). More surprisingly, some human γδT cell subsets can react to lipids (e.g., α-GalCer, sulfatide, andphosphatidylethanolamine) in the context of CD1d (48–51). Existence of CD1dreactive γδT cells has also been proposed in mice (52). Last, an unusual population of T cells bearing a hybrid δ/αβ TCR was shown to recognize CD1d/α–GalCer complex (53). Further investigations are clearly required to evaluate the biological relevance of these rare subsets and to understand why they have been conserved across species. Altogether, this new literature highlights the huge—probably still expanding—diversity of the CD1d-restricted T cell family.

# MR1-Restricted T Cells

Mucosal-associated invariant T cells probably represent the main subset of T cells restricted to the MHC class I-related molecule, MR1 (54, 55). MAIT cells were initially named after their observation in the gut lamina propria of humans and mice (54). Despite their characteristics of tissue-resident memory T cells at barrier sites, MAIT cells can also populate lymphoid tissues and non-lymphoid organs (8) such as the liver (56, 57). In addition, MAIT cells circulate through the bloodstream where they can represent up to 10% of the human T cell compartment. On the contrary, MAIT cells are relatively rare in standard laboratory mouse strains (55, 58).

Mucosal-associated invariant T cells are characterized by the expression of a semi-invariant TCR composed of a canonical TCRα chain (Vα19-Jα33 in mice and Vα7.2-Jα33 in humans) associated with a restricted set of Vβ segments (54, 55, 58, 59). Several subsets of human and mouse MAIT cells have recently been identified, including MAIT cell precursors (39, 60) and non-MAIT mature human MR1-restricted T cells using diverse αβTCR (61–63) that may have specific functions such as tumor Ag recognition.

Through their TCR, MAIT cells recognize small intermediate metabolites from the riboflavin (vitamin B2) pathway of bacteria (64). The first defined Ags were microbial pterin-like compounds presenting a ribityl moiety (64). In addition, the nonenzymatic reaction between a riboflavin precursor(5-amino-6-d-ribitylaminouracil) and small aldehydes (glyoxal/methylglyoxal) of both microbial and host origin generates two instable forms namely 5-(2-oxoethylideneamino)-6-d-ribitylaminouracil and 5-(2-oxopropylideneamino)-6-d-ribitylaminouracil (65). These two molecules represent potent Ags that are able to bind into MR1 and activate mouse and human MAIT cells (65, 66).

These discoveries subsequently enabled the development of mouse and human MR1 tetramers loaded with MAIT cell Ags that represented a major breakthrough in the investigation of MAIT cell biology including development, phenotype, and functions (39, 58, 67). Reminiscent with NKT cell biology, MAIT cells can respond to TCR signals and/or to various activating cytokines including IL-12, IL-1β, IL-18, and IL-23 (56, 68, 69). Upon activation, MAIT cells produce large amounts of Th1- and Th17 related cytokines such as IFN-γ, TNF-α, IL-17A, and IL-22 (70). Additionally, MAIT cells are capable of efficiently kill bacteriainfected cells (71, 72).

Due to their scarce representation in common laboratory mouse strains, understanding MAIT cell biology is challenging by comparing wild-type (WT) and *Mr1*<sup>−</sup>/<sup>−</sup> mice. To better assess their functions, transgenic mice (*V*α*19i*Tg × Cα−/−) displaying high content of MAIT cells have been developed (73). However, the use of these mice is not without concerns since MR1 tetramer-positive cells can still be detected in Vα19i TCR transgenic (Tg) mice that lack MR1 (58). In addition, a certain abnormality in MAIT cells have been reported in these Tg mice (67). Nevertheless, MAIT cells can be found at high frequencies in unusual inbred mouse strains such as *Mus musculus castaneus* (CAST mice) (74). Interestingly, the reason for this MAIT phenotype in CAST mice relies on a single locus located on chromosome 14. Thus, a congenic mouse presenting a high level of MAIT cells (20× compared to classical C57BL/6) on a C57BL/6 background (named B6-MAITCAST) was generated (74) and will be likely to be helpful to investigate the functions of MAIT cells.

Given their cytokine profile and cytotoxic capacity, MAIT cells intuitively emerged as cell subsets specialized in host defense against bacteria. However, recent evidences indicate that MAIT cells get activated in many pathological situations such as acute and chronic viral infections (68, 75–78), solid cancers and hematological malignancies (79–82), as well as many inflammatory disorders including type I and type II diabetes (83, 84), inflammatory bowel disease (85, 86), graft-versus-host disease (87), chronic obstructive pulmonary disease (88, 89), and multiple sclerosis (90, 91).

## LUNG CD1d-RESTRICTED NKT CELLS AND MR1-RESTRICTED MAIT CELLS IN HEALTH

#### CD1d-Restricted NKT Cells

In mice, type I NKT cells account for around 2–5% of lungresident T lymphocytes. Lung type I NKT cells are mainly resident either as marginated cells within the vasculature or located in the lung interstitial parenchyma (92, 93). The lungs are particularly enriched for NKT17 cells compared to the vast majority of tissues (3). Interestingly, type I NKT cell location in the lung tissue is strongly dependent on the subsets. While NKT1 and NKT2 cells are predominantly found in the vasculature, NKT17 cells are at frontline within the lung parenchyma (93, 94). However, the factors that regulate their homing and homeostasis in the lung tissue are yet to be defined. Of note, microbiota seems to regulate lung type I NKT cell homeostasis since germ-free mice display an increase frequency of type I NKT cells, which is dependent on hypermethylation and increase levels of CXCL16 (95).

# MR1-Restricted MAIT Cells

Mucosal-associated invariant T cells are also present in the lung tissue of mice in which they account for approximately 2 and 0.3% of resident T lymphocytes in C57Bl/6J and BALB/c, respectively (67). Akin to NKT cells, lung MAIT cells predominantly display a phenotype of IL-17-producing cells defined by high expression of IL-7Rα and IL-18R1 and the lack of NK1.1 expression (67). In line, stimulation of purified lung MAIT cells of naive mice induced strong IL-17A production but little IFN-γ (67). In addition, they present a phenotype of effector memory cells (CD44high CD62Llow). The precise pulmonary niches of MAIT cells have not been determined, so far, but should be revealed soon, for instance, using antibody (Ab)-mediated *in vivo* labeling. How lung MAIT cells rely on commensal bacteria is currently unknown; however, there is a severe impairment in MAIT cells in the thymus, spleen, and gut of germ-free mice (39, 54). While NKT cells and MAIT cells appear to patrol the lungs in the steady state, their contribution to lung physiology and tissue integrity remains to be determined.

# CD1d-RESTRICTED NKT CELLS AND MR1-RESTRICTED MAIT CELLS IN LUNG INFECTIONS

A large body of evidence in both preclinical and clinical settings has recently suggested a key role for both NKT cells and MAIT cells in host response against lung pathogens (**Table 2**). Here, we compared the mode of activation and functions of NKT cells and MAIT cells in infectious respiratory diseases. We focused our interest on pathogens that provide data for both populations.

### Mycobacteria

Mycobacteria are the causative agents of tuberculosis (TB) due to a group of closely related species including *Mycobacterium tuberculosis*, *Mycobacterium bovis*, *M. microti*, and *M. africanum*. More than a billion of people have died of TB during the last three centuries. Nowadays, TB remains the major cause of death due to infection, causing an estimation of 10 million new cases and ~2 millions of fatalities per year worldwide (121). In addition, one-quarter to one-third of the global human population is latently infected (122, 123), constituting a large reservoir for future spread of active TB. TB control mainly relies on a 100-year-old vaccine [Bacillus Calmette–Guérin (BCG)] (124) that saved the lives of many children without impacting on TB prevalence. In addition, the emergence of resistances for antituberculosis drugs, developed several decades ago, becomes worrying (125).

#### NKT Cells

The role of NKT cells during *Mycobacterium* infection has been extensively studied in preclinical models (96–98, 126). Early type I NKT cell activation (IFN-γ release and cytotoxicity) has been reported during *M. bovis* BCG and *M. tuberculosis* infections (96, 99, 126). In addition, type I NKT cells can also exert antimycobacterial effector function *in vitro* through the secretion of GM-CSF (127). The mechanisms that drive type I NKT cell activation have been shown to rely on both adaptive and innate cues such as CD1d-mediated Ag recognition and the activating cytokines IL-12 and IL-18 (97, 126). Several mycobacterial-derived lipids such as phosphatidylglycerol, cardiolipin, or phosphatidylinositol activate type II but not type I NKT cells (35) even if an early study has suggested CD1ddependent antigenic properties for phosphatidylinositol mannoside on type I NKT cells (128). Despite their early activation during mycobacterial infections, type I NKT cells do not appear to have a critical role in the control of TB (100–104) despite some controversial results have been reported regarding granuloma formation (96, 105, 126). Akin to *J*α*18*−/− mice, CD1ddeficient mice did not significantly differ in their susceptibility to mycobacterial infection compared to control mice (106). This suggests that neither type I nor type II NKT cells play a cardinal role in TB outcome in mice.

Several clinical reports indicated that patients with active TB present reduced levels of circulating NKT cells compared to patients with latent forms (129–134). However, NKT cells from patients with active TB present an activated phenotype and can secrete high amounts of IFN-γ (129, 135). Of note, NKT cell activation in TB patients leads to PD-1 expression at cell surface and subsequent apoptosis of IFN-γ-producing NKT cells, a phenomenon that can be reversed through PD-1 blockade (136, 137). Human NKT cells have also been shown to exert direct bactericidal effects on *M. tuberculosis*-infected macrophages (126). In addition, some NKT cells from pleural effusion of TB patients present characteristics of TFH-like cells (138). Thus, it is possible that TFH-like NKT cells produce IL-21 to participate in local B cell response against *M. tuberculosis* (138).

Although preclinical models do not point toward a key role for NKT cells in TB, their rapid response and immunoregulatory functions poised them as an interesting target for cell-based therapies. In line, exogenous activation of type I NKT cells with α-GalCer (alone or in combination with classical chemotherapy) protected mice from lethal TB (96, 99, 139, 140). The molecular determinants that drive this protective effect are not fully understood but are likely to rely on enhanced innate response such as IFN-γ production and cytotoxicity. Moreover, α-GalCer (or analogs) incorporation into BCG vaccine enhances immune responses in mice by modulating T cell priming *via* type I NKT cell activation (141). α-GalCer can also exert adjuvant effect on TB vaccines using the sublingual route by enhancing Ag-specific IFN-γ-producing T cells (142). In addition to its primary application as a vaccine for TB prevention, epidemiological studies have suggested that BCG vaccination may prevent IgE-mediated allergic diseases in mice and humans (143, 144). Interestingly, the BCG-mediated IgE response suppression appears to be dependent on IL-21 secretion by type I NKT cells that mediates specific apoptosis of IgE-expressing B cells (145).

#### MAIT Cells

The role of MAIT cells in preclinical model of TB has been explored. Mouse MAIT cells can inhibit intracellular growth of *M. bovis* BCG in macrophages (107). This effect relies on MR1 and IL-12-dependent IFN-γ secretion by MAIT cells (107). Interestingly, upon BCG infection at low dose, *Mr1*<sup>−</sup>/<sup>−</sup> mice displayed a higher bacterial burden in the lung during the early course of infection compared to MAIT cell-proficient mice. However, similar bacterial loads were observed at later stage of infection (107) suggesting an important role for MAIT cells in early control of bacterial growth although inefficient to prevent long-lasting establishment of the bacteria in the lung tissue. The role of MAIT cells in controlling mycobacterial infection has also been evaluated using *V*α*19i* Tg mice (69). Upon intravenous injection of *M. abscessus*, *V*α*19i* Tg × *Mr1*<sup>−</sup>/<sup>−</sup> mice displayed slightly but significant higher bacterial loads than their *Mr1*<sup>+</sup>/<sup>+</sup> littermates (69). Despite having an increased proportion of MAIT cells, it is important to mention that *V*α*19i* Tg × *Mr1*<sup>+</sup>/<sup>+</sup> mice have a comparable bacterial burden compared to WT mice (69). Thus, MAIT cells play a limited protective role in the early course of mycobacterial infections.

Akin to NKT cells, clinical studies in patients with active TB indicated a striking decrease in frequency of circulating MAIT cells compared to healthy controls (69, 71, 146). This strong reduction is paralleled with an enrichment of MAIT cells


in the lung tissue and pleural effusions of TB patients (71) suggesting a recruitment to the inflamed tissue. In addition, one can argue that this relative disappearance can also rely on other mechanisms such as increased apoptosis (147) or TCR downmodulation (rendering them virtually undetectable) as a consequence of overactivation. Indeed, MAIT cells from blood and pleural effusions of TB patients displayed a phenotype of activated/memory cells (CD69<sup>+</sup> CD45RO<sup>+</sup>) associated with a higher capacity to produce antimycobacterial cytokines such as IFN-γ and TNF-α (147). This phenotype can be further exemplified since MAIT cells from these patients had a reduced response to unspecific restimulation (147). Reminiscent with the data generated in studies on NKT cells, MAIT cells from TB patients expressed high levels of PD-1 and *ex vivo* PD-1 blockade restored IFN-γ production by these latter (147). Human MAIT cells can also mediate cytotoxicity against mycobacteria-infected cells. Thus, they can recognize and kill *M. tuberculosis*-infected cells including DCs and lung epithelial cells in a MR1-dependent manner (71, 148, 149). Of note, non-MAIT Vα7.2-Jα12 MR1-restricted T cells can also efficiently kill BCG-infected THP-1 (a monocytic cell line) (72).

#### Other Bacteria

Concurrent literature for NKT cells and MAIT cells in the context of bacterial lung infection is rather limited. Some data are nevertheless available for models of pneumonia induced by the Gram-negative bacterium *Francisella tularensis* or by the Grampositive bacterium *Streptococcus pneumoniae*.

#### *Francisella tularensis*

*Francisella tularensis* is the causative agent of tularemia (also referred to as rabbit fever); a zoonosis that spreads *via* arthropod vectors after contact with infected animals (150). Depending on the site of infection, tularemia can lead to several clinical manifestations including pneumonia, the most deadly form of the disease with a mortality rate up to 60% (151). Incidence rates have drastically dropped over the last century despite some local outbreaks have been noticed over the last decades (152). In addition, *F. tularensis* can be considered as a potential agent for bioterrorism (153). Given the high virulence of *F. tularensis*, preclinical data mainly derived from attenuated strains such as *F. tularensis* Live Vaccine Strain (LVS).

#### NKT Cells

Surprisingly, the role of NKT cells has only recently emerged in experimental pulmonary tularemia (108), a model of immunopathology that can be related to sepsis-like disorder (154). To explore this, authors studied survival and body weight loss in various mouse strains including *J*α18<sup>−</sup>/<sup>−</sup>, *Cd1d*<sup>−</sup>/<sup>−</sup>, and *V*α*14i* Tg mice. Unfortunately, due to the apparent variability in this model and the lack of data using all mouse lines in a single experiment, interpretation of the results is not straightforward (108). Thus, *J*α*18*<sup>−</sup>/<sup>−</sup> and *V*α*14i* Tg mice were more susceptible to infection compared to WT controls while *Cd1d*<sup>−</sup>/<sup>−</sup> mice were more resistant. Altogether, these data implied that NKT cells probably play deleterious role during pulmonary tularemia even if data obtained with *J*α*18*<sup>−</sup>/<sup>−</sup> mice do not support this hypothesis. To explain this, authors claimed that CD1d-deficient mice represent a better model than *J*α*18*<sup>−</sup>/<sup>−</sup> mice to study the role of NKT cells in disease (108). However, several additional hypotheses are worth mentioning to explain these results. First, it is possible that type I NKT and type II NKT cells play opposite functions during pulmonary tularemia. Although, *V*α*14i* Tg mice present a similar phenotype than *J*α*18*<sup>−</sup>/<sup>−</sup> mice, it has to be kept in mind that type I NKT cells generated from *V*α*14i* Tg mice present unique characteristics compared to *bona fide* type I NKT cells and a profound bias in their CD8<sup>+</sup> T cell repertoire (155). Given the role of MAIT cells during pulmonary tularemia (see below), the phenotype observed in *Cd1d*<sup>−</sup>/<sup>−</sup> mice can also at least partially rely on the high level of MAIT cells in these mice (39). Conversely, *J*α*18*<sup>−</sup>/<sup>−</sup> mice are likely to have reduced MAIT cell frequencies since 60% of the TCRα repertoire diversity is actually lacking in these mice (40).

The resistance of *Cd1d*−/− mice to tularemia is mainly associated with enhanced formation of tertiary lymphoid structures in lungs of infected mice associated with mitigated inflammation including reduced pro-inflammatory cytokines and inflammatory cell recruitment (108). Collectively, NKT cells have deleterious functions during tularemia although a more refine analysis of the potential differential roles of type I NKT and type II NKT cells in this model should be considered.

#### MAIT Cells

CD4<sup>−</sup>CD8<sup>−</sup> double-negative (DN) T cells have been shown to represent a major responding T cell subset in the lungs of LVSinfected mice (156). Despite the lack of specific tools, authors unequivocally demonstrated a couple of years later that this strong accumulation of DN T cells during pulmonary LVS infection was mainly attributable to MAIT cells (109). Indeed, most of these cells expressed transcripts for the invariant Vα19-Jα33 α-chain rearrangement associated with TCR β-chains containing Vβ6 or Vβ8 segments (109). In addition, the accumulation of DN T cells was absent in *Mr1*<sup>−</sup>/<sup>−</sup> mice. Interestingly, *Mr1*<sup>−</sup>/<sup>−</sup> mice exhibited higher bacterial burden in the lungs compared to WT mice from day 10 postinfection (109).

During *in vivo* infection, MAIT cells produced antibacterial cytokines including IFN-γ, TNF-α, GM-CSF, and IL-17A and contributed to recruitment of adaptive CD4<sup>+</sup> and CD8<sup>+</sup> T cells (109, 110). Interestingly, MAIT cells controlled the recruitment of activated CD4+ T cells through GM-CSF-mediated reprogramming of monocytes into monocyte-derived DCs (110).

In addition, cultures of MAIT cells with LVS-infected monocytes led to a MAIT cell-dependent reduction in bacterial growth, a mechanism that relied on secretion of IFN-γ, TNF-α, and nitric oxide (109). *In vitro* activation of MAIT cells in this model was dependent on MR1 and the activating cytokines IL-12 and IL-18 according to the strain of bacterium used (109, 157). Altogether, MAIT cells appear as important antibacterial players during tularemia.

#### *Streptococcus pneumoniae*

*Streptococcus pneumoniae* (the pneumococcus) is the major bacterium responsible for community-acquired pneumonia in western countries and accounts for ~2 million of deaths per year worldwide. In normal conditions, *S. pneumoniae* colonizes asymptomatically nasopharynx of healthy individuals. However, when the immune equilibrium is broken, pneumococcus carriage can lead to mild disease such as otitis media or sinusitis and more occasionally turns into severe complications such as pneumonia, sepsis, and meningitis (158). In addition, presence of *S. pneumoniae* is often found in biological fluids of hospitalized patients for severe influenza A infection (159) as well as patients with exacerbated chronic lung inflammation such as chronic obstructive pulmonary disease (160).

Despite vaccination is an efficient strategy to prevent pneumococcus spread and to control infections, the available vaccines have, however, some issues [for reviews, see Ref. (161, 162)]. In addition, the emergence of antibiotic-resistant strains (163) represents an important threat for the management of pneumococcal infections in clinics.

#### NKT Cells

Many studies have highlighted an important role for type I NKT cells in host response against pulmonary pneumococcal infection. Using *S. pneumoniae* serotype 1 or 3 strains, we and others reported that *J*α*18*<sup>−</sup>/<sup>−</sup> mice were more susceptible to infection and displayed higher bacterial burden compared to WT controls (111–113). The underlying mechanisms of this protective activity rely on IFN-γ secretion by type I NKT cells that regulate early recruitment of neutrophils (113). Elaborately, type I NKT cell-derived IFN-γ controls the secretion of CXCL2 and TNF-α (113), two critical mediators of lung neutrophilia. Type I NKT cells migrate to the lung parenchyma only 24 h after *S. pneumoniae* infection in a CCL17-dependent manner, a mechanism, which is crucial for mouse survival (92). The mechanisms involved in type I NKT cell activation during *S. pneumoniae* infection have been extensively explored and depends on activating cytokines, pneumococcal-derived Ag, or both according to the strain studied. First, akin to other bacteria, the cell wall of several *S. pneumoniae* strains is constituted of glycolipids (α-glucosyldiacylglycerol) that can serve as CD1drestricted type I NKT cell Ags (114). The recognition of this Ag *in vivo* is important for host response to *S. pneumoniae* (114). The general mechanism that allows accessibility and loading of such microbial-derived Ags into CD1d are currently unknown and deserve further investigations. In addition, type I NKT cell activation during pneumococcal infection is also dependent on the release of IL-12 especially by CD103<sup>+</sup> DCs (112, 115). No data are currently available regarding type II NKT cells during pneumococcal infection.

Exogenous activation of type I NKT cells protects against lethal *S. pneumoniae* infection independently of the strain used (113, 164). Mechanistically, this protective activity relied on IFN-γ and IL-17A production and subsequent neutrophil recruitment that controlled bacterial elimination (164). While these protective effects have been obtained using α-GalCer, exogenous activation of type I NKT cells with α-mannosylglycolipids has also been shown to partially protect mice against pneumococcal infection (165).

Surprisingly, despite strong evidence for a critical role in host response to pneumococcus in experimental models, no clinical studies have so far addressed the dynamics and functions of NKT cells in patients with severe *S. pneumoniae*-driven pneumonia.

Type I NKT cells have been shown to provide help for B cells in mounting Ab responses (26, 166). In line with this literature, NKT cells have important role in the production of specific antipneumococcal Abs and class switching in response to pneumococcal vaccines in experimental models (167, 168). A recent clinical study enrolling vaccinated subjects with the 23-valent pneumococcal polysaccharide vaccine reported a positive correlation between frequency of circulating type I NKT cells and the serum concentrations of specific IgG directed against serotypes 14 and 19F (167). Interestingly, injection of liposomes containing synthetic NKT cell Ags and pneumococcal capsular polysaccharides led to the generation of long-term memory B cell response (169, 170). Generation of such a response was dependent on the recognition of lipids and capsular polysaccharide antigens by type I NKT cells and B cells, respectively, to elicit cognate and direct NKT–B cell interactions. Thus, harnessing NKT cell functions (adjuvant effect or B cell help) during pneumococcal vaccination might represent an interesting avenue of research in the future.

Of note, pneumococcal infection and vaccination have been proposed to reduce airways allergic diseases as a result of immune modulation in both experimental and clinical studies (171–173). Interestingly, this mechanism relied on the regulatory functions of some components of the pneumococcus including the bacterial toxin pneumolysin and type-3-polysaccharide (174). These molecules induced the generation of regulatory T cells that control type I NKT cell accumulation and limit the development of airway hyperresponsiveness (174).

#### MAIT Cells

Early works suggested that MAIT cells do not react to *Streptococcus* group A (69). However, genomic and transcriptomic analyses of numerous strains indicate that *S. pneumoniae* expresses enzymes involved in the synthesis of riboflavin metabolites (175, 176) as well as a highly conserved riboflavin operon (177). Thus, activation and functions of MAIT cells during pneumococcal infection have recently been investigated. Human MAIT cells can produce IFN-γ in response to accessory cells infected by *S. pneumoniae* clinical isolates (116, 177). Interestingly, the magnitude of MAIT cell response differed from one isolate to another and was proposed to depend on genetic variations in their expression of the riboflavin pathway (116) suggesting a role for MR1-dependent Ags in this model. Conversely, another study suggested that MAIT cell activation, in response to THP-1 cells cultured with fixed pneumococci was unchanged upon MR1 blockade but was fully abrogated following IL-12 and IL-18 neutralization (177). This discrepancy might stem from differences in experimental design as well as in the level of metabolic activity (riboflavin metabolism) of the various strain used. In line with the first hypothesis, the use of macrophages in the culture assay led to IFN-γ release by MAIT cells in a MR1-dependent manner (177). Environmental factors can also determine the generation of MAIT cell Ags by the pneumococci. Thus, modulation in riboflavin availability in the culture assay significantly influenced MAIT cell activation in response to pneumococcus-infected DCs. Specifically, excess of riboflavin blunted MAIT cell response while culture in riboflavin-free assay medium increased MAIT cell activity (116, 177). In addition, heat stress can induce the riboflavin operon in *S. pneumoniae* resulting in higher MAIT cell Ag availability (177).

Using *in vivo* models of pneumococcal infections with virulent or avirulent strains in Vα19i Tg × Cα−/− mice, a small proportion of lung MAIT cells produced IFN-γ and IL-17A (116). Although a more detailed kinetic analysis is required, these levels were relatively low compared to cytokine production observed for other unconventional T cell subsets in this model including γδT cells and NKT cells (112, 178, 179). However, the precise functions of MAIT cells in pneumococcus-induced pneumonia are currently unknown since data from either *Mr1*<sup>−</sup>/<sup>−</sup> or Vα19iTgCα−/<sup>−</sup>*Mr1*<sup>−</sup>/<sup>−</sup> mice are not yet available.

Akin to NKT cells, the phenotype and dynamics of MAIT cells in patients with severe pneumonia associated with presence of pneumococci are rather limited. Of note, the levels of circulating MAIT cells in critically ill patients with severe bacterial infections were markedly decreased compared to age-matched healthy donors (180). Surprisingly, the decrease in MAIT cell frequency was less pronounced in patients with streptococcal infections including pneumonia compared to non-streptococcal infections (180). This might imply a minimal role for MAIT cells during *S. pneumoniae*-induced pneumonia in humans although this observation should be confirmed by the use of MR1 tetramers.

#### Influenza A Virus (IAV)

Influenza A viruses belong to the family of *Orthomyxoviridae* viruses and are characterized by a segmented single-stranded RNA genome of negative polarity. IAV can be further segregated regarding the expression of two important proteins namely hemagglutinin and neuraminidase proteins. IAV are responsible for seasonal highly contagious infections characterized by a severe pulmonary immune pathology leading to human morbidity and mortality (181). About five millions of clinical cases and 250,000 to 500,000 deaths are reported worldwide every year (182, 183). More sporadically, transversal infections of animal strains to humans or co-infection with multiple IAV strains in individuals can result in re-assortment of genes generating highly virulent new strains leading to life-threatening pandemics (184, 185). Recent epidemiological studies also indicated that hospitalized patients for severe pneumonia during IAV epidemics or pandemics are often coinfected with bacteria including *S. pneumoniae* (methicillinresistant) *Staphylococcus aureus*, *Pseudomonas aeruginosa*, and *Haemophilus influenza* (186). These secondary bacterial infections contribute significantly to the excess morbidity and mortality of influenza [for a review, see Ref. (186)]. Mechanistically, infection with IAV dampens innate antibacterial immunity and alters pulmonary barrier functions thus favoring local bacterial outgrowth and dissemination from the lungs [for reviews, see Ref. (187, 188)]. Considering the clinical impact of bacterial superinfection post-influenza, the development of preclinical models should be encouraged for proper translational studies.

#### NKT Cells

We and others have demonstrated the pivotal role of type I NKT cells in host response to IAV infection. Keeping in mind that they represent a heterogeneous population composed of subsets endowed with specialized functions, type I NKT cells have been shown to exert multiple functions during the course of IAV infection. Despite mechanisms differ from one experimental setting to another, type I NKT cell deficiency has been consensually shown to increase susceptibility (survival, enhanced inflammation) in several experimental influenza (H1N1, H3N1, and H3N2) infection models (117–119). For instance, a pioneer study described a new mechanism in which type I NKT cells could inhibit the suppressive functions of myeloid-derived suppressor cells resulting in enhanced IAV-specific CD8<sup>+</sup> T cell response, enhanced viral clearance, and increased survival (119). We also described a role for type I NKT cells in mounting IAV-specific CD8<sup>+</sup> T cell response (118). In our model, this effect relied on NKT cell-mediated control of lung CD103<sup>+</sup> DC emigration to the draining lymph nodes (118). In parallel, type I NKT cells exert anti-inflammatory functions during IAV infection, thus preventing immunopathogenesis. First, type I NKT cells control the recruitment of inflammatory monocytes (117) as well as the production of IFN-γ by NK cells (118), two mechanisms that can blunt IAV-associated pathogenesis (189, 190). In addition, type I NKT cells produced the tissue protective cytokine IL-22 during the early course of IAV infection (120). Although it did not affect viral loads *in vivo* (191), IL-22 protected against epithelial damages due to viral replication (120, 192). Through this mechanism, IL-22 might reduce secondary bacterial infection post-influenza (191, 193).

The activation mechanisms of type I NKT cells during IAV infection have been proposed to rely on either CD1d-dependent or CD1d-independent pathways. First, type I NKT cells have been shown to directly control the suppressive activities of myeloid-derived suppressor cells in a CD1d-dependent manner (119). The nature of the ligand involved is yet to be defined but appears to be a neo-synthesized GSL of host origin (119). On the other hand, we demonstrated that IL-22 production by type I NKT cells during IAV infection was dependent on IL-1β and IL-23 secretion by accessory cells and did not require CD1d (120).

The potential role of type II NKT cells during IAV is currently unknown. Of note, *CD1d*<sup>−</sup>/<sup>−</sup> mice are more susceptible to H1N1 IAV infection than *J*α*18*<sup>−</sup>/<sup>−</sup> mice although no differences between genotypes were observed regarding viral loads (119). This might suggest a role for type II NKT cells in regulating host immune response rather than controlling viral replication.

The role of type I NKT cells in secondary bacterial infection post-influenza has also been examined. Using a model of post-IAV invasive secondary pneumococcal infection, type I NKT cells were shown to limit susceptibility to superinfection and lethal viral/bacterial synergism (112). In this setting, type I NKT cells act early after IAV infection by preventing tissue damages (through IL-22) (118). On the other hand, at the peak of bacterial susceptibility, type I NKT cells are in a state of hyporesponsiveness; an effect that depends on the immunosuppressive cytokine IL-10 (112). Interestingly, IL-10 blockade restored type I NKT cell activity and slightly increased resistance to bacterial superinfection (112). Of interest, α-GalCer instillation tempered secondary bacterial infection with lower bacterial outgrowth and dissemination (194) suggesting that type I NKT cells could be instrumental in clinics to combat IAV infection and its complications.

#### MAIT Cells

Since they cannot generate MAIT cell Ags derived from the riboflavin pathway, viruses were intuitively thought to be unable to activate MAIT cells. Nevertheless, MAIT cells can be activated in the absence of TCR ligation by many cytokines (e.g., IL-1β, IL-2, IL-7, IL-12, IL-15, IL-18, IL-23, and type I IFN) (75, 195, 196) released by accessory cells through engagement of various pattern recognition receptors including toll-like receptors and/or nodlike receptors.

Pioneer experiments demonstrated that MAIT cells isolated from Vα19iTg mice do not respond to DCs infected with a broad set of viruses (69). However, as previously mentioned, MAIT cells from Tg mice differ from conventional ones in that they present a naive profile and are devoid of the key transcription factor PLZF (67). Since PLZF controls expression of many chemokine/cytokine receptors (197) including IL-7Rα and IL-18Rα on MAIT cells (39), this can simply explain the absence of response of MAIT cells from Tg mice to virus-induced activating cytokines. The putative functions of MAIT cells during experimental influenza infection are still ignored.

Human MAIT cells can respond *in vitro* to IAV by CD69 and granzyme B upregulation as well as IFN-γ secretion (68, 75). In these settings, purified or enriched MAIT cells were cultured with peripheral blood mononuclear cells and IAV-infected epithelial cells or IAV-exposed macrophages. Mechanistically, MAIT cell activation was dependent on IL-18 secretion by monocyte/macrophages and did not require functional MR1 molecule (68, 75). Notably, IAV replication was not required since ultraviolet-irradiated IAV were still able to activate MAIT cells (75). These data suggest a potential role for MAIT cells in host response during IAV infection.

Compared to healthy donors, the frequency of circulating MAIT cells was decreased in a small cohort of 16 patients hospitalized for severe pneumonia due to infection with the Asian lineage avian IAV (H7N9) (68). Interestingly, individuals who recovered from pneumonia had higher levels of circulating MAIT cells compared with those who succumbed (68). Another clinical study confirmed the reduced MAIT cell frequencies in patients with acute IAV infection (75). This decrease was even more pronounced in critically ill patients admitted in intensive care unit compared to patients with mild symptoms (75).

#### Future Directions

Natural killer T cells and MAIT cells participate in host response to numerous respiratory pathogens. These cells might have some overlapping functions during lung infections. To better elucidate their differential contribution, integrated analysis of activation dynamics and functions of NKT cells and MAIT cells in experimental models as well as in patients with severe lung infection should be encouraged. In addition to the models discussed above, NKT cells have been reported to play important role *in vivo* in many other areas of lung infections including respiratory syncytial virus, *P. aeruginosa*, *Legionella pneumophilia*, *Chlamydia* spp. as well as the fungi *Cryptococcus neoformans* and *Aspergillus fumigates* [for a review, see Ref. (198)]. Thus, it would be worth to examine the activation status and putative functions of MAIT cells in these models.

Regarding the limitations enumerated previously, the conclusions drawn from the phenotype observed in *J*α*18*<sup>−</sup>/<sup>−</sup>, *Cd1d*<sup>−</sup>/<sup>−</sup>, *Mr1*−/−, *V*α*14i* Tg, and *V*α*19i* Tg mice should be taken with caution. Therefore, reconstitution experiments with organ-specific cells in these knock-out/transgenic mice appear mandatory to confirm previous findings. In addition, selective *in vivo* depletion of these populations by means of monoclonal Ab represents an alternative approach as recently reported for type I NKT cells (199). Additionally, since these cells are likely to regulate homeostasis and functions of each other, crossing NKT cell- and MAIT cell-deficient mice will certainly provide new interesting tools to determine the specific or redundant functions of these cells.

Although the evolutionary conservation of CD1d and MR1 across mammals renders rodent models of pulmonary infections relevant to study NKT cell and MAIT cell biology, the expected widespread of humanized mouse models will surely provide new insights on the development and regulation of the immune response during lung infections. In addition, these models will be likely to highlight the NKT cell and MAIT cell species' differences in data generated from mouse and human studies.

The current therapeutical arsenal for clinicians to combat lung infections is almost exclusively based on the use of antimicrobial drugs and standardized daily management. However, fatal outcomes in patients with severe lung infection are often associated with dysregulated immune response. In this condition, innovative strategies of immune intervention targeting host factors should be proposed. Thus, according to their immunoregulatory functions, NKT cells and MAIT cells are well poised for cell-based therapy to fight lung infections and their complications.

#### THERAPEUTIC OPPORTUNITIES AND CONCLUDING REMARKS

The last century has witnessed incredible progresses in the control of infectious diseases, which significantly reduced the human death toll, especially in western countries. However, the incidence of lung infections has significantly increased over the last decades due to major changes in environment and lifestyle (leading to immune homeostasis dysregulation and increased contact with new pathogens). These events represent key societal and public health threats. In addition, booming in resistance to antimicrobial drugs also constitute a new challenge for clinicians. Thus, the development of new therapeutical options to combat lung infections and their complications is urgently needed.

Harnessing the biology of CD1- and MR1-restricted cells is currently highly regarded to fight cancer (200, 201). By contrast, exploiting NKT cell and MAIT cell in clinical trials for lung infections is still in its infancy. Yet, targeting these cells offers several advantages that may lead to improve current management of lung infections. First, NKT cells and MAIT cells are restricted by quasi-monomorphic Ag-presenting molecules. Thus, they have the potential to be manipulated by "universal" ligands rendering most of patients eligible for NKT/MAIT cell-based therapies. Second, engineering of polarizing Ags (only developed for NKT cells so far) allows considering the generation of tailored-made responses according to the immune profile of the patients. Last, experimental and clinical studies indicated that NKT cell-based therapies are safe and feasible (202). Thus, many strategies might be tested to harness CD1d- and MR1-restricted T cell functions during lung infections.

The adjuvant properties of NKT cells and MAIT cells could be exploited in the design of more efficient vaccines. The replacement of conventional adjuvants by NKT/MAIT cell Ags (α-GalCer and its analogs, ribloflavin metabolites) in classical vaccines could be used to optimize the magnitude and duration of the adaptive immune. Through the ability of NKT cells and MAIT cells to subsequently activate/mature accessory cells including DCs, this strategy is likely to improve the development of the memory response.

In addition, these Ags might also generate short-term and organ-specific immune responses with positive outcomes (8). As clinical studies predominantly reported decreased levels of circulating NKT cells and MAIT cells in patients with severe pneumonia and associated with reduced prognosis, the proliferative activity (*in vivo* or *ex vivo*) of these Ags might help to the replenishment of the NKT/MAIT cell pool. Of note, the highly unstable nature of MAIT cell Ags precludes its use in expansion protocols and, therefore, represents a major limitation in these strategies. However, recent efforts in generating more stable forms of MAIT cell Ags (203, 204) should help to circumvent this issue in the near future. Despite being more laborious, the expansion of MAIT cells can also be achieved through reprogramming to pluripotent stem cells and subsequent redifferentiation using the stromal cell line OP9/DL1 (205).

The effects of combining NKT/MAIT cell Ags with antimicrobial drugs should also be encouraged. By reducing the concentrations of antibiotics or antiviral drugs used in conventional therapy for a similar antimicrobial activity, this strategy could help to lessen the development of drug resistance.

Chimeric Ag receptor (CAR) T cell therapy using conventional T cells has recently emerged as a promising approach in clinics (206). However, the high polymorphism in HLA system limits the development of CAR T cell therapy due to potential appearance of side effects such as graft-versus-host disease. Since many bacteria express glycolipid Ags for NKT cells, the recent

#### REFERENCES


generation of CAR-expressing NKT cells (207, 208) offers an interesting and highly innovative approach in the context of antibacterial therapies. Similar approaches could be developed in the future with CAR-expressing MAIT cells.

Dealing with their hyporesponsiveness during lung infections represents another food for thought for NKT/MAIT cell-based therapy. Increased expression of the checkpoint inhibitor PD-1 on NKT cells and MAIT cells in patients with TB has been associated with reduced capacity to produce antimycobacterial cytokines (136, 137, 147). Blockade of PD-1 expression has recently emerged as a potential strategy for infectious diseases including lung infection and sepsis (209–212). Thus, the possibility that the protective activity of anti-PD-1 treatments might partially rely on the restoration of NKT cell and MAIT cell functions should be considered. Treatment with α-GalCer has also been shown to lead to NKT cell hyporesponsiveness due to TCR internalization (213) and/or induction of PD-1 expression (214). New formulations and a better delivery of α-GalCer to potent Ag-presenting cells (such as DCs) by means of dedicated nanovectors can prevent the development of NKT cell anergy (215–218). These strategies might optimize NKT cell-based therapies to combat respiratory infections.

Natural killer T cells and MAIT cells have become the focus of intense investigation in the last decade. It is now clear that harnessing the biology of these cells has the potential to offer innovative therapeutic approaches in a medical field in which clinicians require a new and efficient arsenal to treat infections and their complications.

# AUTHOR CONTRIBUTIONS

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

#### ACKNOWLEDGMENTS

We apologize to colleagues whose works could not be cited due to space constraints. CP is supported by the INSERM and FT by the CNRS.

#### FUNDING

This work was supported by the Inserm, the CNRS, the University of Lille Nord de France, the Pasteur Institute of Lille, and the Institut National du Cancer (INCa, under reference R08046EE/ RPT08003EEA and R13071EE/RPT13001EEA).

involved in airway neutrophilia. *J Exp Med* (2007) 204:995–1001. doi:10.1084/ jem.20061551


in vivo through direct interaction with dendritic cells. *J Immunol* (2003) 171: 5140–7. doi:10.4049/jimmunol.171.10.5140


*Am J Respir Crit Care Med* (2010) 182:841–7. doi:10.1164/rccm.200912- 1921OC


interferon responses to *Francisella* species in vitro but not in vivo. *Infect Immun* (2018) 86:e00117-18. doi:10.1128/IAI.00117-18


D39 of *Streptococcus pneumoniae* and comparison with that of unencapsulated laboratory strain R6. *J Bacteriol* (2007) 189:38–51. doi:10.1128/JB. 01148-06


**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 Trottein and Paget. 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.*

# Increased iNKT17 Cell Frequency in the Intestine of Non-Obese Diabetic Mice Correlates With High *Bacterioidales* and Low *Clostridiales* Abundance

*Lorena De Giorgi1 , Chiara Sorini2 , Ilaria Cosorich1 , Roberto Ferrarese3 , Filippo Canducci3 and Marika Falcone1 \**

*1Experimental Diabetes Unit, Division of Immunology, Transplantation and Infectious Diseases, IRCCS San Raffaele Scientific Institute, Milan, Italy, 2Department of Medicine, Karolinska Institute, Stockholm, Sweden, 3Department of Biotechnology and Life Sciences, University of Insubria, Varese, Italy*

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Chyung-Ru Wang, Northwestern University, United States Tonya J. Webb, University of Maryland, Baltimore, United States*

> *\*Correspondence: Marika Falcone falcone.marika@hsr.it*

#### *Specialty section:*

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

*Received: 30 April 2018 Accepted: 16 July 2018 Published: 30 July 2018*

#### *Citation:*

*De Giorgi L, Sorini C, Cosorich I, Ferrarese R, Canducci F and Falcone M (2018) Increased iNKT17 Cell Frequency in the Intestine of Non-Obese Diabetic Mice Correlates With High Bacterioidales and Low Clostridiales Abundance. Front. Immunol. 9:1752. doi: 10.3389/fimmu.2018.01752*

iNKT cells play different immune function depending on their cytokine-secretion phenotype. iNKT17 cells predominantly secrete IL-17 and have an effector and pathogenic role in the pathogenesis of autoimmune diseases such as type 1 diabetes (T1D). In line with this notion, non-obese diabetic (NOD) mice that spontaneously develop T1D have an increased percentage of iNKT17 cells compared to non-autoimmune strains of mice. The factors that regulate iNKT cell expansion and acquisition of a specific iNKT17 cell phenotype are unclear. Here, we demonstrate that the percentage of iNKT17 cells is increased in the gut more than peripheral lymphoid organs of NOD mice, thus suggesting that the intestinal environment promotes iNKT17 cell differentiation in these mice. Increased intestinal iNKT17 cell differentiation in NOD mice is associated with the presence of pro-inflammatory IL-6-secreting dendritic cells that could contribute to iNKT cell expansion and iNKT17 cell differentiation. In addition, we found that increased iNKT17 cell differentiation in the large intestine of NOD mice is associated with a specific gut microbiota profile. We demonstrated a positive correlation between percentage of intestinal iNKT17 cells and bacterial strain richness (α-diversity) and relative abundance of *Bacterioidales* strains. On the contrary, the relative abundance of the anti-inflammatory *Clostridiales* strains negatively correlates with the intestinal iNKT17 cell frequency. Considering that iNKT17 cells play a key pathogenic role in T1D, our data support the notion that modulation of iNKT17 cell differentiation through gut microbiota changes could have a beneficial effect in T1D.

Keywords: natural killer T cells, interleukin-17, dendritic cells, microbiota, autoimmune diabetes

#### INTRODUCTION

Vα14iNKT cells play pleiotropic functions in the immune system that either boost or dampen T and B cell immunity in infections, antitumor responses, and autoimmune diseases (1, 2). These opposite immunological functions are mediated by iNKT cells with different cytokine-secretion phenotype. In fact, iNKT cells are classified into iNKT1, iNKT2, and iNKT17 based on their release of Th1, Th2, and Th17 cytokines. iNKT1 cells that predominantly secrete IFN-γ play an effector adjuvant function that enhances T cell responses and is fundamental to clear infections and tumors while iNKT2 cells releasing IL-4 and IL-13 mainly provide B cell help and are involved in allergic reactions. The iNKT17 cells have been recently characterized and they also play effector functions in infections (3), asthma (4), and autoimmune diseases like collagen-induced arthritis (5) and autoimmune type 1 diabetes (T1D) (6, 7). In fact, non-obese diabetic (NOD) mice that spontaneously develop T1D have an increased frequency and absolute number of iNKT17 cells (1) that are directly responsible for triggering autoimmune diabetes (6).

The factors that regulate iNKT cell expansion and acquisition of a specific cytokine-secretion phenotype and function are still largely unknown. Most iNKT cells are already committed toward a specific cytokine-secretion phenotype when they exit the thymus (8). However, it is now clear that the iNKT cell repertoire can be expanded and modulated in the periphery and recent evidence indicates that the intestinal environment and the microbiota composition are instrumental to control iNKT cell expansion and acquisition of a specific cytokine-secretion phenotype (9). In support to this notion, lack of commensal gut microbiota in germ-free mice leads to less mature and hyporesponsive iNKT cells (10, 11). Moreover, housing conditions and the resulting differences in the gut microbiota composition alter iNKT cell functional maturation, release of cytokines, and acquisition of effector functions (10). The relative abundance of some gut commensal microbes plays a direct effect on iNKT cells. For example, *Bacteroides fragilis* limits iNKT cell expansion in the gut mucosa by providing inhibitory sphingolipid antigens that bind the iNKTCR (12). On the other hand, some lipids derived from *Sphingomonas* species are capable to activate iNKT cells and promote their intestinal expansion (13, 14) and functional maturation (10). Commensal microbiota can also influence iNKT cells through antigen-independent mechanism such as epigenetic regulation of CXCL16 expression that promotes iNKT cell recruitment to the gut mucosa (11). Although these findings demonstrate that commensal microbiota influence iNKT cell number and function, the capacity of the gut environment to drive iNKT cells toward a specific cytokine-secretion phenotype is yet to be determined.

Here, we show that NOD mice have increased iNKT17 cell frequency in the intestinal mucosa that is more evident than in peripheral lymphoid organs and liver. Moreover, we found that the augmented iNKT17 cell percentages correlate with a specific gut microbiota profile characterized by high bacterial richness, increased relative abundance of *Bacteroidales*, and reduction of *Clostridiales* strains.

#### MATERIALS AND METHODS

#### Mice

Females NOD mice were purchased from Charles River Laboratories (Calco, Italy). In some experiments mice received antibiotic treatment (ampicillin, 1 g/L; neomycin 1 g/L; metronidazole, 1 g/L; vancomycin, 0.5 g/L) for 1 week in drinking water. All mice were maintained under specific pathogen-free conditions in the animal facility at San Raffaele Scientific Institute, and all experiments were conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) according with the rules of the Italian Ministry of Health.

# Cell Isolation

Mononuclear cells were isolated from intestinal tissues as previously described (15, 16). After removal of the Peyer's patches, small and large intestines were flushed with PBS, opened longitudinally, and predigested with 5 mM EDTA and 1 mM DTT for 20 min at 37°C. After removing epithelial cells and adipose tissue, the intestine was cut into small pieces and incubated in HBSS containing 0.5 mg/mL collagenase D, 1 mg/mL dispase II (Roche Diagnostics GmbH, Mannheim, Germany), and 5 U/mL DNase I (Sigma-Aldrich, St. Louis, MO, USA) for 20 min at 37°C. Digested tissues were washed, suspended in 5 mL of 40% Percoll (Sigma-Aldrich, St. Louis, MO, USA), and overlaid on 2.5 mL of 80% Percoll solution. Percoll gradient separation was performed by centrifugation at 1,000 *g* for 20 min at 20°C, and cells at the interface were collected. For lymphocyte isolation from liver, the total organ was meshed and hepatocytes were removed by Percoll gradient centrifugation. Splenocytes and lymph node cells were isolated by mechanical disruption of the tissues.

#### Flow Cytometry

Single cell suspensions were stained for 20 min at 4°C with the following fluorochrome-conjugated monoclonal antibodies or tetramers in FACS buffer (PBS with 5% FBS, 0.1% sodium azide): PE anti-mouse αGalCer-loaded CD1d tetramers (PBS57/ Dimerix from the NIH Tetramer Facility, Washington, DC, USA), FITC anti-mouse TCRβ, PerCP anti-mouse CD4, Pacific Blue CD8, APC-Cy7 anti-mouse CD3 (BD Biosciences, San Diego, CA, USA). For intracellular cytokine staining, single-cell suspensions were stimulated for 2.5 h with 50 ng/mL phorbol 12-myristate 13-acetate (PMA) and 1 µg/mL ionomycin (both from Sigma-Aldrich, St. Louis, MO, USA) in the presence of 10 µg/mL Brefeldin A (Sandoz, Princeton, NJ, USA). Cells were collected and labeled for iNKT cell surface markers and then fixed and permeabilized with fixation and permeabilization buffer (BD Biosciences, San Diego, CA, USA) and stained with PE-Cy7 anti-mouse IL-17 and FITC anti-mouse IFN-γ mAbs (BD Biosciences, San Diego, CA, USA). Dead cells were stained with AmCyan-conjugated fixable viability dye (eBioscience, San Diego, CA, USA) and excluded from the analysis. Flow cytometry data were acquired on a FACSCanto II and analyzed with FACS Diva software (BD Biosciences, San Diego, CA, USA).

#### DC Cytokine Secretion

To analyze the cytokine secretion profile of the intestinal dendritic cells (DCs), CD11c<sup>+</sup> cells were isolated by magnetic separation from intestinal single cell suspensions and stimulated *in vitro* with 1 µg/mL LPS for 20 h. IL-1β, IL-6, TNF-α, and IL-23 in the cell culture supernatants were quantitated with a BD cytometric bead array (CBA from BD Biosciences, CA, USA). The data were analyzed with FCAP-Array software v1.0.1 (Soft Flow, St. Louis Park, MN, USA).

#### DC-iNKT Cell Coculture Experiments

To assess the capacity of intestinal DCs to induce iNKT cell expansion and iNKT17 cell differentiation *in vitro*, CD11c<sup>+</sup> cells were purified from single cell suspension obtained from intestinal tissues by magnetic separation. iNKT cells were isolated from total splenocytes by staining with PE anti-mouse PBS57/

Dimerix and magnetic separation with anti-PE MicroBeads (Miltenyi Biotec, Bologna, Italy). Intestinal DCs were cocultured with purified iNKT cells for 7 days at 1:1 ratio with the addition of IL-7 and IL-15 (10 ng/mL) every 48 h. After 7 days, the cytokine secretion profile of iNKT cells was determined by FACS analysis.

Figure 1 | Increased frequency of iNKT17 cells in the spleen, liver, and intestine of non-obese diabetic (NOD) mice. (A) Measurements of percentages of total iNKT cells in spleens, lymph nodes, and liver of female 6-week-old Balb/c and NOD mice (8 mice/group). Single cell suspensions obtained from different organs of age- and sex-matched NOD and Balb/c female mice were stained with αGalCer-loaded CD1d tetramers (PBS57-DimersX) in combination with anti-TCR-β monoclonal antibodies and FACS analyzed. Data are presented as mean percentage ± SEM of iNKT cells (PBS57-DimerX+TCRβ+) out of total TCRβ+ T cells. (B) Small and large intestines were digested with collagenase D and DNase I and single cell suspensions were obtained by Percoll gradient. Cells were stained and analyzed as in (A). (C,D) Total single cell suspensions obtained from the different tissues were stimulated with phorbol 12-myristate 13-acetate/ionomycin for 2.5 h, stained with PBS57-CD1d tetramers in combination with anti-TCR-β monoclonal antibody, then fixed and permeabilized and stained with anti-IL-17 (C) or anti-IFN-γ (D) monoclonal antibody. Data are expressed as mean percentage ± SEM of IL-17+PBS57-DimerX+TCRβ+ or IFN-γ+PBS57-DimerX+TCRβ+ out of total iNKT cells (PBS57-DimerX+TCRβ+). (E) Cytokine secretion profiles of intestinal dendritic cells (DCs) from NOD and Balb/c mice. Intestinal DCs were purified from the large intestine of NOD or Balb/c mice by magnetic bead separation with anti-CD11c mAb and stimulated *in vitro* with LPS (1 µg/mL). After 20 h, supernatants were collected and cytokine secretion was quantified by cytometric bead array. (E,F) Intestinal DCs from NOD mice were more efficient in inducing iNKT cell expansion and iNKT17 cell differentiation. iNKT cells were isolated from splenocytes of BALB/c and NOD mice by magnetic separation and stimulated *in vitro* with αGalCer-pulsed DCs obtained from the intestinal tissues of NOD or Balb/c mice as specified. After 7 days, the cells were collected and FACS analyzed for expression of iNKT cell markers (PBS57-DimersX and anti-TCR-β monoclonal antibodies) and intracellular stained for IL-17. Data are presented as mean percentage ± SEM of iNKT cells out of total TCRβ+ T cells (left panel) or IL-17+PBS57-DimerX+TCRβ+ out of total iNKT cells (right panel). The *p* values were calculated using a paired Student's *t* test. \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001.

#### Microbiota Analysis

Total bacterial DNA was isolated from the mucosa and luminal content of the large intestine of NOD mice and BALB/c mice using PowerFecal™ DNA Isolation kit (Qiagen, Hilden, Germany) following the manufacturers' instructions. The V3- V4- V5- region of the 16S rRNA gene was amplified using universal primers. The analysis of microbiota metabolically active was performed by pyrosequencing of rRNA cDNA 16S (GS Junior, Roche, Roche Diagnostics GmbH, Mannheim, Germany). Sequences with a high-quality score were used for the taxonomic analysis with QIIME (Quantitative Insights Into Microbial Ecology version 1.6).

#### Statistical Analysis

Statistical significances of the differences in the percentages of positive cells were calculated by unpaired two-tailed Student's *t*-test using GraphPad Prism software. Statistical analysis of the microbiota profiling data was performed on the proportional representation of the taxa (summarized to phyla, class, order, family, and genus levels) using one-way ANOVA test with Bonferroni's correction. For all tests, a *p* value <0.05 was considered statistically significant.

#### RESULTS AND DISCUSSION

iNKT cell with an IL-17-secreting phenotype (iNKT17 cells) have a crucial pathogenic role in autoimmune T1D (6). In female, NOD mice that spontaneously develop T1D, although the overall iNKT cell number is reduced, there is a selective increase in the iNKT17 cell percentage possibly due to their enhanced thymic differentiation (8). Since recent evidence indicates that the intestine is important for iNKT cell maturation and functional differentiation (10), we used female NOD mice with an enlarged iNKT17 cell repertoire to ask whether the gut mucosa is a preferential site for iNKT17 cell expansion. Total iNKT cells and iNKT17 cells were analyzed in the peripheral lymphoid organs (spleen and peripheral lymph nodes), liver and small and large intestine of NOD and control non-autoimmune Balb/c mice. Secretion of IL-17 and IFN-γ by iNKT cells (iNKT17 and iNKT1 cell phenotypes) was detected after staining with αGalCer-loaded CD1d tetramers and stimulation with PMA and ionomycin. Our analysis confirmed a significantly reduced percentage (**Figure 1A**) and absolute number (Figure S1 in Supplementary Material) of iNKT cells in the spleen of NOD mice. On the contrary, the iNKT cell presence in the intestinal tissues as well as in the liver was augmented in NOD mice compared to control Balb/c mice both in terms of relative percentages (**Figures 1A,B**) and absolute numbers (Figure S1 in Supplementary Material). We also confirmed that, in NOD mice, there is a selective expansion of the iNKT17 cell subset in all organs analyzed (**Figure 1C**). However, we noted that the increase in iNKT17 cell frequency in NOD mice was much more significant in the intestinal tissues (particularly in the large intestine) in comparison with peripheral lymphoid organs and liver. In fact, while there is a twofold increase of iNKT17 cells in the spleen, liver, and small intestine of NOD mice in comparison with control Balb/c mice, we detected a fourfold increase in the large intestine of NOD mice (**Figure 1C**). The frequency of other iNKT cell subsets (iNKT1 cells) was similar between NOD and control Balb/c mice in all organs analyzed (**Figure 1D**). These results suggest that there is a preferential expansion of iNKT17 cells in the large intestine of NOD mice.

Next, we asked whether increased intestinal iNKT17 cell frequency in NOD mice is due to the presence in their gut mucosa of pro-inflammatory dendritic cells (DCs) that preferentially drive iNKT cells toward an IL-17-secreting iNKT17 phenotype. The factors that regulate peripheral iNKT17 cell differentiation have not yet been characterized, but we hypothesized that DC release of Th17-priming cytokines such as IL-6 and IL-23 could also drive iNKT17 cell differentiation (17, 18). In line with this notion, we found that intestinal DCs of NOD mice secrete higher amount of IL-6 in comparison with their counterparts from control Balb/c mice (**Figure 1E**). Moreover, intestinal DCs of NOD mice were more efficient in promoting iNKT cell expansion (**Figure 1F**, left panel) and inducing iNKT17 cell differentiation *in vitro* compared to intestinal DCs of Balb/c mice (**Figure 1F**, right panel). However, intestinal DC are not directly responsible for iNKT17 cell differentiation since stimulation of iNKT cells from Balb/c mice with

Figure 2 | Increased intestinal iNKT17 cell frequency in non-obese diabetic (NOD) mice correlates with the commensal microbiota composition. Microbiota profiling was performed on microbiota samples obtained from the lumen and mucosal tissue of the large intestine of female 6-week-old Balb/c and NOD mice (8 mice/ group). 16S mRNA analysis and QIIME software were used for the analysis of microbiota composition and species distribution between the two groups. (A) Alpha diversity of the commensal microbiota composition isolated of NOD mice and BALB/c mice. (B) Correlation between total number of bacterial species detected in the large intestine and percentage of intestinal iNKT17 cells in NOD mice. (C) Profiles of the commensal microbiota composition at the genus level in Balb/c and NOD mice. (D) Relative abundances of *Bacteriodales* and *Clostridiales* strains in Balb/c and NOD mice. Data are presented as mean percentage ± SEM of the relative abundance of the different bacterial strains in the two murine strains. (E,F) Correlative analyses between the relative abundance of *Bacteroidales* or *Clostridiales* strains and the frequency of iNKT17 cells in the large intestine of NOD mice. Statistical analysis of the microbiota profiling data was performed by using one-way ANOVA test with Bonferroni's correction. (G) Percentages of iNKT17 cells in intestinal the intestine and liver changes upon antibiotic treatment. 5-week-old NOD mice were treated with broad-spectrum antibiotics (ampicillin, 1 g/L; neomycin 1 g/L; metronidazole, 1 g/L; vancomycin, 0.5 g/L) for 1 week in drinking water. Single cell suspensions were obtained from different tissues, stimulated with phorbol 12-myristate 13-acetate + ionomycin for 4 h, stained with PBS57-CD1d tetramers in combination with anti-TCR-β monoclonal antibody, then fixed and permeabilized and stained with anti-IL-17 monoclonal antibody. Data are expressed as mean percentage ± SEM of IL-17+PBS57-DimerX+TCRβ+ out of total iNKT cells (PBS57-DimerX+TCRβ+). \**p* < 0.05, \*\**p* < 0.01.

intestinal NOD DCs was not sufficient to drive them toward an iNKT17 cell phenotype (**Figure 1F**, right panel). We concluded that intestinal iNKT cells of NOD mice were intrinsically more prone to acquire a biased iNKT17 cell phenotype than iNKT cells from Balb/c mice.

Having established that the frequency of iNKT17 cells is increased in the intestine of NOD mice, we wanted to highlight the mechanism responsible for this effect. The composition of the gut commensal microbiota can have a strong impact on iNKT17 cell expansion. To explore this possibility, we analyzed the profiles of the intestinal microbial community by ultra-deep pyrosequencing of barcoded 16S rRNA gene amplicons on samples obtained from the luminal content and mucosa of the large intestines of NOD mice and Balb/c mice. Our taxonomic, functional, and diversity microbiome profiling revealed significant differences in the gut commensal microbiota composition in NOD mice vs control Balb/c mice. Specifically, we found that NOD mice have a higher bacterial richness (**Figure 2A**) that directly correlates with iNKT17 cell percentages in the large intestine (**Figure 2B**). At the genus level, we demonstrated that the microbiota composition in the large intestine of NOD mice is very different compared to non-autoimmune Balb/c mice (**Figure 2C**) with a selective increase in the relative abundance of *Bacteroidales* and reduction of *Clostridiales* strains (**Figure 2D**). Importantly, we found that in the NOD mice the iNKT17 cell frequency positively correlates with the relative abundance of *Bacteroidales* species (*p* > 0.01) (**Figure 2E**) and inversely correlates with the presence of *Clostridiales* strains (*p < 0.05*) (**Figure 2F**). To furtherly highlight the role of the gut microbiota in iNKT17 cell expansion, we treated NOD mice with broad-spectrum antibiotics and measured the percentages of iNKT17 cells in the gut mucosa and systemically (liver tissue). Our data show statistically significant alterations in the percentages of iNKT17 cells in the liver and in the intestine of antibiotic-treated NOD mice compared to untreated NOD controls (**Figure 2G**). However, we observed that iNKT17 cell frequency decreased in the liver but increased in the intestinal tissue. This discrepancy may be related to the uncomplete deletion of endogenous commensal microbiota induced by the antibiotic treatment in NOD mice (data not shown). This could have favored microbial species that negatively regulate iNKT17 cells in the liver (probably by passing through the intestinal barrier into the systemic circulation) and other species that positively affect iNKT17 cell expansion in the gut mucosa.

Our data suggest that the intestinal environment of NOD mice regulates iNKT17 cell differentiation though gut microbiota modification. Previous reports have shown that microbial regulation of iNKT cell expansion and functional maturation extends beyond the intestinal compartments. For example, in germ-free mice, iNKT cells are hyporesponsive not only in the gut mucosa but also in peripheral lymphoid organs (10). Immune cells, including iNKT cells, after being modulated in the gut, travel to secondary lymphoid organs and peripheral tissues (19). Hence, the gut microbiota composition of NOD mice could promote T1D by favoring intestinal expansion of effector iNKT17 cells that from the gut mucosa move to pancreatic lymph nodes and tissues to promote T1D pathogenesis.

#### ETHICS STATEMENT

All mice were maintained under specific pathogen-free conditions in the animal facility at San Raffaele Scientific Institute and all experiments were conducted in accordance with the Institutional Animal Care and Use Committee.

# AUTHOR CONTRIBUTIONS

LDG, CS, and IC performed all *in vivo* and *in vitro* experiments. LDG analyzed data and prepared manuscript's figures. RF and FC performed microbiome analysis. MF served as principal investigator, analyzed data, and wrote the manuscript.

#### ACKNOWLEDGMENTS

We thank the NIH Tetramer Facility (Washington, USA) for the kind donation of αGalCer-loaded CD1d tetramers (PBS57/ Dimerix). This work was supported by a Research Grant from the Juvenile Diabetes Foundation (Grant 2-SRA-2014-28- Q-R) to MF.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01752/ full#supplementary-material.

# REFERENCES


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

*Copyright © 2018 De Giorgi, Sorini, Cosorich, Ferrarese, Canducci and Falcone. 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.*

*Guillaume Lezmi1,2,3 and Maria Leite-de-Moraes2,3\**

*1AP-HP, Hôpital Necker-Enfants Malades, Service de Pneumologie et d'Allergologie Pédiatriques, Paris, France, 2Université Paris Descartes, Paris, France, 3 Laboratory of Immunoregulation and Immunopathology, INEM (Institut Necker-Enfants Malades), CNRS UMR8253 and INSERM UMR1151, Paris, France*

Recent studies have highlighted the heterogeneity of asthma. Distinct patient phenotypes (symptoms, age at onset, atopy, and lung function) may result from different pathogenic mechanisms, including airway inflammation, remodeling, and immune and metabolic pathways in a specific microbial environment. These features, which define the asthma endotype, may have significant consequences for the development and progression of the disease. Asthma is generally associated with Th2 cells, which produce a panel of cytokines (IL-4, IL-5, IL-13) that act in synergy to drive lung inflammatory responses, mucus secretion, IgE production, and fibrosis, causing the characteristic symptoms of asthma. In addition to conventional CD4+ T lymphocytes, other T-cell types can produce Th2 or Th17 cytokines rapidly. Promising candidate cells for studies of the mechanisms underlying the pathophysiology of asthma are unconventional T lymphocytes, such as invariant natural killer T (iNKT) and mucosal-associated invariant T (MAIT) cells. This review provides an overview of our current understanding of the impact of iNKT and MAIT cells on asthmatic inflammation, focusing particularly on pediatric asthma.

Keywords: asthma, NKT cells, mucosal-associated invariant T cells, children, patients, CD1d, MR1, Th2 cells

# INTRODUCTION

Asthma is now considered to encompass different conditions characterized by common symptoms (wheeze, cough, shortness of breath, and chest tightness), variable degrees of airflow limitation, and different pattern of inflammation. Most patients with asthma have an eosinophilic infiltration of the airways, associated with increased production of type 2 cytokines including IL-4, IL-5, IL-13 secreted by Th2 cells, together with allergic comorbidities (1). However, around 50% of adults with asthma do not fall into this description (2). Asthmatic patients with a neutrophil-high signature were described in both adults and children (3–5). This neutrophilic-predominant endotype is less well understood than the Th2 endotype and may be related to the activation of the IL-17 pathway (1, 6). Intriguingly, despite eosinophilic airway inflammation is a key feature of severe asthma in schoolchildren, there is no clear evidence for a Th2 type cytokine signature in bronchial mucosa or bronchoalveolar lavages in that population (7, 8). Alternative mechanisms may, therefore, be involved in the pathogenesis of asthma in this group. Recent studies have suggested the potential role of unconventional T cells, such as invariant natural killer T (iNKT) and mucosal-associated invariant T (MAIT) cells in asthma pathogenesis. These T lymphocytes usually reside in the tissues, including those of the airways and can respond rapidly to stimuli by producing Th2 and Th17 cytokines. Here, we review the field of asthma immunity, focusing on the role of iNKT and MAIT cells in asthmatic patients.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Chiaki Iwamura, National Institutes of Health (NIH), United States Seddon Y. Thomas, National Institute of Environmental Health Sciences (NIEHS), United States Rosemarie DeKruyff, Stanford University, United States*

#### *\*Correspondence:*

*Maria Leite-de-Moraes maria.leite-de-moraes@ parisdescartes.fr*

#### *Specialty section:*

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

*Received: 28 April 2018 Accepted: 17 July 2018 Published: 30 July 2018*

#### *Citation:*

*Lezmi G and Leite-de-Moraes M (2018) Invariant Natural Killer T and Mucosal-Associated Invariant T Cells in Asthmatic Patients. Front. Immunol. 9:1766. doi: 10.3389/fimmu.2018.01766*

# iNKT CELLS

The major characteristic of iNKT lymphocytes is their expression of T cell receptors of limited diversity recognizing lipid antigens presented by the non-polymorphic MHC-like molecules, CD1d (9, 10). iNKT cells express an invariant TCRα chain, Vα14-Jα18 (or TRAV11 TRAJ18) in mouse, and Vα24-Jα18 (or TRAV10 TRAJ18) in humans, together with a limited set of TCRβ chains (10, 11). The invariant TCRα chains of mice and human are very similar, enabling the iNKT cells to recognize the same glycolipids in both species. A classic example is provided by α-galactosylceramide (α-GalCer), an antigen capable of stimulating both mouse and human iNKT cells (12, 13). α-GalCer and its analog, PBS57, are currently used as antigens for the production of a CD1d-tetramer complex capable of specifically identifying iNKT cells (14). iNKT lymphocytes respond rapidly to specific lipid antigens, in a TCR-dependent manner, but they also respond to pro-inflammatory cytokines (12, 15, 16). Indeed, IL-12 and IL-18 induce the production of IFNγ, whereas IL-1β and IL-23 will promote the secretion of IL-17A (or IL-17) and IL-22 (17, 18). Further, iNKT cells have been shown to express IL-25 (or IL-17E), thymic stromal lymphopoietin, and IL-33 receptors that will favor their secretion of IL-4, IL-13, and IFNγ (19–22). Human iNKT cells require TGFβ for the production of IL-17 and IL-22 (23). TCR-dependent and TCR-independent pathways can act in synergy to stimulate iNKT cells more strongly (15, 24).

The thymic differentiation of iNKT cells is tightly controlled. At least three major iNKT cell subsets mature in the thymus, the iNKT1 (IFNγ and IL-4 producers), the iNKT2 (IL-4 and IL-13 producers), and the iNKT17 (IL-17 and IL-22 producers) (25–28). Maturation in the thymus is regulated by the Slam-Associated Protein (29), the transcription factors PLZF, Egr2, ThPOK, Runx1, and RORγt, the microRNA Let-7, and the cytokine IL-7 (30–38). iNKT cells undergo several maturation steps [see Ref. (10, 35) for more information], before migrating to peripheral organs as CD4<sup>+</sup>CD8<sup>−</sup> and double-negative (CD4<sup>−</sup>CD8<sup>−</sup>) cells. In humans, there is also a CD4<sup>−</sup>CD8<sup>+</sup> subset (39). iNKT cells are mostly resident in tissues, where they can "patrol" to identify threats to the body. For instance, these cells have been shown to perform an intravascular immune surveillance function in the liver, spleen, and lung (40–43). The primary function of iNKT cells is to protect the host from infections (24, 44, 45). However, in some conditions, iNKT cell activation favors tissue injury, including lung (**Figure 1**), as discussed below.

#### iNKT Cells and Murine Asthma Models

Mouse models are widely used to help clarify the role of iNKT cells in asthma. Studies initially focused on allergic asthma, with ovalbumin (OVA) as the allergen, associated with aluminum hydroxide as adjuvant, for the systemic immunization followed by intranasal (i.n.) OVA challenge. First analysis showed no major difference in the severity of allergic airway inflammation in β2microglobulin (β2m)<sup>−</sup>/<sup>−</sup> and CD1d<sup>−</sup>/<sup>−</sup> mice, which lack iNKT cells (46–48). However, other studies reported that iNKT cell-deficient (Jα18−/− and CD1d−/−) mice had attenuated asthma symptoms including airway hyperresponsiveness (49), airway eosinophilia, Th2 inflammation, and OVA-specific anti-IgE production (50, 51). The adoptive transfer of IL-4- and IL-13-producing iNKT cells restored the asthma severity, demonstrating that iNKT cells favored allergic asthma symptoms through the production of these cytokines (50, 51). iNKT cells did not recognize OVA as an antigen, but their ability to promote lung inflammation was reduced by the treatment of mice with anti-CD1d antibodies, indicating that endogenous lipidic antigens stimulated the iNKT cells (50). More recently, another study comparing distinct iNKT cell-deficient mice strains (β2m<sup>−</sup>/<sup>−</sup> and CD1d−/−) reported that NKT cells were dispensable for T celldependent allergic airway inflammation (52), even though AHR was not analyzed.

A possible reason to explain the discrepancies between studies concerning the implication of iNKT cells in asthma severity is that, in addition to iNKT cells, type II NKT cells were also absent in β2m<sup>−</sup>/<sup>−</sup> and CD1d<sup>−</sup>/<sup>−</sup> mice (53, 54), while β2m<sup>−</sup>/<sup>−</sup> mice also lack CD8 T cells. Then, it is not excluded that the absence of type II NKT and CD8<sup>+</sup> T cells could influence the effect of iNKT cells in asthma severity. Another point is that asthma symptoms are more severe in 129/Sv mice compared to BALB/c and C57BL/6 animals (55). The iNKT cell-deficient mice cited here (β2m<sup>−</sup>/<sup>−</sup>, CD1d<sup>−</sup>/<sup>−</sup>, and Jα18<sup>−</sup>/<sup>−</sup> mice) were created on a 129/Sv background. Some results showing no significant differences in airway eosinophilia used 129/Sv × C57BL/6 CD1d−/− mice (47), while those describing CD1d<sup>−</sup>/<sup>−</sup> and Jα18<sup>−</sup>/<sup>−</sup> mice as more resistant to asthma used CD1d<sup>−</sup>/<sup>−</sup> and Jα18<sup>−</sup>/<sup>−</sup> backcrossed with BALB/c animals (51). In our hands, Jα18<sup>−</sup>/<sup>−</sup> (backcrossed at least 10 times in C57BL/6) presented lower allergen-induced airway inflammation and AHR than controls (50). Recently, Kronenberg's team created a new mouse strain deficient for iNKT cells. These mice presented no airway eosinophilia and significantly less pulmonary resistance in response to OVA challenge than did their wild-type (WT) counterparts (56). Hence, the discrepancies reported may also result from a possible low number of backcross of the knockout mice used. Finally, the microbiota differences between the animal houses where the studies were performed cannot be excluded. In this context, an elegant study by Blumberg's team showed that iNKT cells accumulated in the lung and in the colonic lamina propria in germ-free (GF) mice, rendering these animals more susceptible to OVA-induced asthma and oxazolone-induced colitis (57). The colonization of neonatal GF mice with a normal flora or *Bacteroides fragilis* decreased the number of iNKT cells and protected the mice against these diseases, clearly establishing a link between iNKT cells, the microbiota, and disease (57, 58).

These studies were highly informative but were designed to analyze a specific allergic asthma model. They, therefore, underestimated the complexity of asthma pathogenesis. It was subsequently shown that α-GalCer, the cognate antigen for iNKT cells, protects sensitized mice against asthma symptoms when administered 1 h before the first challenge (59). The mechanisms involved are dependent on IFNγ production by α-GalCer-stimulated iNKT cells (59). In another context, α-GalCer, administered i.n. at the time of sensitization, was found to act as an adjuvant, enhancing asthma symptoms (42). This study echoed those in non-human primates showing that the administration of α-GalCer alone induces AHR in monkeys (60). The iNKT cells are resident mostly in the intravascular space

Figure 1 | Proposed roles of mucosal-associated invariant T (MAIT) and invariant natural killer T (iNKT) cells in the lung. These unconventional T lymphocytes are present in the lung at steady state and express several chemokine- and interleukin-receptors. Bacteria, fungus, virus, pollutants, and airway allergens will directly or indirectly stimulate MAIT and iNKT cells. Cytokines produced by epithelial cells, namely, IL-25, IL-33, and thymic stromal lymphopoietin, could activate these cells. Antigen-presenting cells (APC) present antigens to MAIT and iNKT cells in the context of MR1 and CD1d molecules, respectively. Activated APC produce IL-12, IL-18, and IL-23 that will stimulate MAIT and/or iNKT cells. Following TCR-dependent or TCR-independent activation, MAIT and iNKT cells secrete IFNγ, IL-17, IL-4, or IL-13. IFNγ contributes to lung protection and promotes potential protective Th1 responses against asthma. IL-17, in turn, could have a dual effect since it is known that this cytokine promotes neutrophils recruitment and activation to protect lung from injury, but IL-17 can also enhance neutrophilic asthma severity. Finally, IL-4 and IL-13 will favor Th2 immune responses and then amplify allergic eosinophilic airway inflammation.

rather than in the pulmonary tissue itself, and they are rapidly mobilized after exposure to airborne lipid antigen, to which they respond by the secretion of cytokines (42). Thus, different lipid antigens in the airways, unrecognized by conventional T cells, may amplify airway inflammation by acting on iNKT cells.

Other asthma models have recently been used to investigate the role of iNKT cells. Intranasal administration of the natural House Dust Mite allergen without adjuvant has been shown to induce iNKT cell recruitment in the lung. The iNKT cells were stimulated *via* OX40–OX40 ligand interactions to generate a pathogenic Th2 cytokine environment (61). In this model, iNKT-deficient mice displayed significantly lower levels of pulmonary inflammation than WT mice (61). iNKT cells were further implicated in the model of asthma induced by *Aspergillus fumigatus* (62). This fungus, which is associated with a severe form of asthma, expresses asperamide-B, a glycolipid specifically recognized by both human and mouse iNKT cells (62). The i.n. administration of *A. fumigatus*- or asperamide-B rapidly induces AHR, by activating pulmonary iNKT cells in an IL-33-ST2- and IL-4/IL-13-dependent manner (62).

Overall, these findings indicate that iNKT cells promote allergic asthma inflammation and AHR principally through the secretion of IL-4 and IL-13. The Th2 paradigm explains many features of asthma, but this disease is not limited to pro-Th2 allergic immune responses and may also include a number of different phenotypes, such as neutrophilic asthma (63–65). In this context, the i.n. administration of α-GalCer activates IL-17-secreting iNKT (iNKT17) cells, which, in turn, recruit neutrophils to the airways (25). iNKT17 cells are also required for the pathogenic mechanism responsible for disease severity in the model of asthma induced by ozone, a major air pollutant (66). These findings indicate that iNKT2 and iNKT17 cell populations may contribute to asthma inflammation in different ways.

#### iNKT Cells and Asthmatic Patients

Several studies have analyzed the possible implication of iNKT cells in the physiopathology of human asthma. Studies analyzing the frequency of iNKT cells in the [bronchoalveolar-lavage fluid (BALF)] or bronchial tissues of asthmatic patients have reported discordant results (67–70). The study of Akbari et al. (68) found that about 60% of the pulmonary CD4<sup>+</sup>CD3<sup>+</sup> T cells in adult patients with moderate-to-severe persistent asthma were iNKT cells. These results were not reproduced by Vijayanand et al. (69), who found that up to 2% of the T cells obtained from airway biopsy, BALF, and sputum induction from subjects with mild or moderately severe asthma were iNKT cells. The study of Thomas et al. (71) also observed less than 2% of iNKT cells among gated T lymphocytes from BALF of asthmatic patients. Of note, further analysis from the initial group have demonstrated that only a small fraction of T cells in the lung of adult asthmatic patients were iNKT cells (72). In our hands (73), iNKT cells accounted for less than 1% of T cells in BALF from severe asthmatic children. The discrepancies with the first study (68) could be due to the limited number of samples, the heterogeneity of the cohort, or to non-specific staining of cells in BALF, as suggested by the study of Thomas et al. (71).

It was showed that the frequency of iNKT cells in the blood of adult asthmatic patients was similar to that in blood from control donors (74). Further, it was suggested that pro-Th2 iNKT cells may be particularly frequent in blood from asthmatic patients, and that these cells was associated with lung function (67). Our previous study indicated that the percentage of peripheral blood iNKT cells did not differ significantly between asthmatic children classified as exacerbators (1 or more severe exacerbations in the last 12 months) and those classified as non-exacerbators (75). Similarly, it has recently been reported that there is no relationship between the frequency of iNKT cells and that of IL-4- or IFNγ-producing iNKT cells in the blood of 1-year-old children and asthma-related clinical outcomes at the age of 7 years (76).

There is now a consensus that a limited number of iNKT cells is present in the BALF of adults and pediatric patients with severe asthma. However, several questions remain unanswered: Is the presence of iNKT cells in the BALF associated with specific asthma endotypes? What role do iNKT cells play in the pathophysiology of asthma? Further studies are therefore required to characterize the mechanisms by which iNKT cells could contribute to asthma.

# MAIT CELLS

Invariant natural killer T and MAIT cells may be considered to be "twins" in several respects. Like iNKT cells, most MAIT cells express an invariant TCRα chain (Vα7.2-Jα33 or TRAV1-2- TRAJ33) and a small number of TCRβ chains (77). MAIT cells are restricted by the MHC class I-related molecule MR1 and recognize microbial-derived vitamin B2 (riboflavin) metabolites, such as the 5-(2-oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU) (78). The endogenous ligands able to either select MAIT cells in the thymus or to potentially stimulate these cells in peripheral lymphoid organs remain to be defined.

Mucosal-associated invariant T cells, similar to iNKT cells, develop in the thymus, where they became functionally competent and able to produce IFNγ, TNFα, and IL-17 in response to stimulation (79). MAIT cells produce low-to-moderate levels of IL-4 and IL-13 when stimulated (80–83). Thymic MAIT development is also directed by the transcription factor PLZF and is dependent on microRNAs (79, 84). Human MAIT cells are CD8<sup>+</sup>, CD4<sup>+</sup>, or double-negative (CD4<sup>−</sup>CD8<sup>−</sup>) and express high levels of CD161 and IL-18Rα (85). MAIT cells can be activated in a TCRdependent and -independent manner. In the latter situation, they can be stimulated by pro-inflammatory cytokines, such as IL-7, IL-12, IL-18, and IL-23 (29, 86, 87). Indeed, MAIT cells, in addition to IL-18Rα, can also express IL-7Rα, IL-23R, IL-12Rβ1, CCR5, CXCR6, CCR6 (83, 84, 88, 89). These receptors will allow the activation of MAIT cells by IL-7, IL-12, IL-18, and IL-23 and their migration to peripheral tissues.

Despite their striking similarities, iNKT and MAIT cells also differ in several important ways. Unlike iNKT cells, MAIT cells are rare in conventional laboratory mouse strains and abundant in humans. In healthy individuals, MAIT cells account for up to 10% of peripheral blood T cells and are numerous in the gut, lung, and liver (49, 85, 89). The expansion of the MAIT cell population in response to commensal flora antigens explains their abundance in mucosal tissues, in which they are involved in antimicrobial responses (49). Their presence in the liver may be explained by the constant exposure of this organ to bacterial products absorbed from the gut. There is a clear causal relationship between the number of MAIT cells and the presence and the diversity of the commensal flora, as shown by the absence of MAIT cells from the peripheral organs of GF mice (49).

#### MAIT Cells and Infections

In addition to the commensal flora, pathogens may also stimulate MAIT cells, which play a crucial role in antimicrobial defenses, through the secretion of IFNγ, TNFα, and IL-17 and the killing Lezmi and Leite-de-Moraes iNKT, MAIT Cells, and Asthma

of target cells through the production of cytotoxic perforin and granzyme B molecules (29, 87, 90). MAIT cell analysis, in both humans and in experimental models, has been greatly facilitated by the use of antigen-loaded MR1 tetramers (83, 84). Experimental studies in non-human primates have reported the activation of circulating MAIT cells in response to Bacillus Calmette–Guerin vaccination and *Mycobacterium tuberculosis* infection (91). MAIT cells from the spleen of these macaques produced IFNγ, TNFα in response to stimulation by *Escherichia coli* in a TCR-dependent manner (91). Intranasal inoculation with *Salmonella typhimurium* in mice induced a striking enrichment in IL-17-producing MAIT cells in the lungs (92). The response of MAIT cells to lung infection with *S. typhimurium* was rapid and dependent on the MR1 presentation of riboflavin biosynthesisderived bacterial ligands (92). These findings are consistent with previous reports indicating that patients infected with mycobacteria have many more MAIT cells in the infected lung and fewer MAIT cells in the blood than uninfected controls (93, 94).

Infections with viruses, such as dengue virus, hepatitis C virus, influenza A virus, and HIV-1 can activate human MAIT cells. MAIT cells do not recognize virus antigens, because no riboflavin metabolites are found in host cells or viruses (78), but they may be activated by cytokines produced during viral infection, such as IL-18 in synergy with IL-12, IL-15, and/or IFNα/β (29, 95). Activated MAIT cells during virus infections robustly secrete IFNγ and granzyme B (29, 95).

Mucosal-associated invariant T cells have also been implicated in non-infectious diseases. Several studies have reported large decreases in MAIT cell number in the peripheral blood of patients with the following diseases: antineutrophil cytoplasm antibody-associated vasculitis, chronic kidney disease, Crohn's disease, ulcerative colitis, newly diagnosed and relapsed multiple myeloma, obesity and type 2 diabetes (96–100). However, the mechanisms by which MAIT cells influence these human diseases remain to be elucidated.

#### MAIT Cells and Adult Asthmatic Patients

Despite the prevalence of MAIT cells in the lung, and their involvement in airway infections, very little is known about the possible role of these cells in asthma. MAIT cells are detected in human fetal lung and are numerous in the lungs of adult rhesus macaques (91, 101), consistent with a protective role against infections in this organ. The frequency of MAIT cells is significantly lower in the peripheral blood, sputum, and bronchial biopsy specimens of asthmatic patients than in control subjects (102). The percentage of MAIT cells in BALF does not differ significantly between these two groups (102). A re-analysis of the results, comparing patients with mild, moderate, or severe asthma to healthy donors, showed that the lower frequency of MAIT cells was significant only in the peripheral blood and sputum of patients with moderate or severe asthma (102). The results of this study suggest that the frequency of MAIT cells is negatively correlated with clinical severity. Furthermore, MAIT cell frequency is associated with serum vitamin D3 concentrations and the use of oral corticosteroids. Proof of concept for the association between corticosteroid use and MAIT frequencies was provided by the demonstration of a decrease in the frequency of circulating MAIT cells in 12 patients with moderate asthma treated with oral corticosteroids for seven days (102). It remains to elucidate whether corticosteroids may modify MR1 expression and MAIT cell activation. Some drugs can influence antigen presentation by MR1 molecules. For instance, doxofylline, a bronchodilator used to treat asthma, is known to upregulate MR1 expression weakly, but does not act as a MAIT cell agonist (103). Thus, some of the drugs currently used in asthma treatment may influence MAIT cell functions.

Mucosal-associated invariant T cells are present and can be activated in the lung (**Figure 1**). However, to date, there is no evidence indicating that MAIT cells could recognize *via* their TCR any airway allergens or pollutants potentially implicated on asthma. Consequently, MAIT cells could be activated either directly by endogenous compounds presented by MR1 molecules or indirectly by pro-inflammatory cytokines present in the lung of asthmatics, namely IL-1β, IL-7, and IL-23 (104, 105). These two possibilities are not mutually exclusive. Of note, asthma exacerbations are frequently associated with virus infections (104), which indirectly activate MAIT cells through the induction of proinflammatory cytokines (29, 95). MAIT cells then activated will secrete IFNγ and/or IL-17. Knowing that MAIT cells secrete low levels of Th2 cytokines, namely IL-4 and IL-13 (80–82, 88–90, 106), we cannot exclude the possibility that these lymphocytes will promote Th2 responses (**Figure 1**). However, as discussed before, asthma is a complex pathology that is not restricted to overproduction of Th2 cytokines. Further, IL-17 and IFNγ production were associated with asthma severity in some steroid-resistant patients (107, 108). Overall, these studies have provided a basis for further analyses of the role of MAIT cells in asthma, potentially on steroid-resistant asthma, and of the mechanisms by which these cells affect asthma severity.

## MAIT Cells and Pediatric Asthmatic Patients

Asthma is frequent in children, but little is known about the possible influence of MAIT cells on the pathophysiology of this disease in childhood. We recently reported a similar frequency of circulating MAIT cells between exacerbators and nonexacerbators, in a population of asthmatic children (75). However, the frequency of IL-17-producing MAIT (MAIT-17) cells was found to be positively correlated with the number of severe exacerbations and negatively correlated with the asthma control test (ACT) score (75). No significant modification of the frequency of IFNγ-producing MAIT cells was observed (109). These findings indicate a possible association of MAIT-17 cells with asthma symptoms. Interestingly, higher levels of IL-17 production by MAIT cells have been observed in a number of non-infections pathologies, such as obesity, type 2 diabetes, and inflammatory bowel disease (99, 100), indicating that mechanisms other than infections may favor IL-17 production by MAIT cells.

Another recent study reported that an association between a high frequency of circulating MAIT cells at 1 year of age and a lower risk of asthma by the age of 7 years (76). Furthermore, this high frequency of MAIT cells was correlated with higher frequency of IFN-γ-producing CD4+ T cells, indicating a possible protective effect of MAIT cells as children grow older (76). IL-17 production by MAIT cells did not correlate with asthma in this study (110). Taken together, the results of these two studies suggest that MAIT-17 cells may be associated with asthma symptoms, whereas pro-Th1 MAIT cells may promote protection (75, 76).

# CONCLUSION

Our understanding of the biology of both iNKT and MAIT cells and their role in asthma has increased considerably in recent years (**Figure 1**). As a result, many new questions have been raised concerning the mechanisms by which iNKT and MAIT cells could promote human severe asthma. For example, time may be an important element, because asthma often begins early in childhood, when the number and functional properties of lung iNKT and MAIT cells may be fixed. Studies

## REFERENCES


conducted in children may, therefore, be crucial. Analyses of circulating iNKT and MAIT cells, as biomarkers, may be informative, but data for BALF and bronchial biopsies are still lacking. Finally, detailed analyses of the frequency and functional subsets of these cells in the context of different asthma endotypes may be crucial for the development of new therapeutic approaches.

# AUTHOR CONTRIBUTIONS

GL and MLM wrote the manuscript.

# FUNDING

This work was supported by funds from INSERM, CNRS, and Paris Descartes University, Paris, France.


an IL-17-producing CD4-NK1.1- NKT cell population. *Proc Natl Acad Sci U S A* (2008) 105(32):11287–92. doi:10.1073/pnas.0801631105


airway inflammation in mice. *J Exp Med* (1999) 189(3):553–62. doi:10.1084/ jem.189.3.553


Lezmi and Leite-de-Moraes iNKT, MAIT Cells, and Asthma

requires the presence of natural killer T cells and IL-17. *J Exp Med* (2008) 205(2):385–93. doi:10.1084/jem.20071507


the home environment". *J Immunol* (2018) 200(10):3317. doi:10.4049/ jimmunol.1800344

110. Chandra S, Wingender G, Greenbaum JA, Kronenberg M. Response to comment on "development of asthma in inner-city children: possible roles of MAIT cells and variation in the home environment". *J Immunol* (2018) 200(10):3317–8. doi:10.4049/jimmunol.1800397

**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 Lezmi and Leite-de-Moraes. 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.*

# Linking CD1-Restricted T Cells With Autoimmunity and Dyslipidemia: Lipid Levels Matter

#### *Sreya Bagchi, Samantha Genardi and Chyung-Ru Wang\**

*Department of Microbiology and Immunology, Northwestern University, Chicago, IL, United States*

Dyslipidemia, or altered blood lipid content, is a risk factor for developing cardiovascular disease. Furthermore, several autoimmune diseases, including systemic lupus erythematosus, psoriasis, diabetes, and rheumatoid arthritis, are correlated highly with dyslipidemia. One common thread between both autoimmune diseases and altered lipid levels is the presence of inflammation, suggesting that the immune system might act as the link between these related pathologies. Deciphering the role of innate and adaptive immune responses in autoimmune diseases and, more recently, obesityrelated inflammation, have been active areas of research. The broad picture suggests that antigen-presenting molecules, which present self-peptides to autoreactive T cells, can result in either aggravation or amelioration of inflammation. However, very little is known about the role of self-lipid reactive T cells in dyslipidemia-associated autoimmune events. Given that a range of autoimmune diseases are linked to aberrant lipid profiles and a majority of lipid-specific T cells are reactive to self-antigens, it is important to examine the role of these T cells in dyslipidemia-related autoimmune ailments and determine if dysregulation of these T cells can be drivers of autoimmune conditions. CD1 molecules present lipids to T cells and are divided into two groups based on sequence homology. To date, most of the information available on lipid-reactive T cells comes from the study of group 2 CD1d-restricted natural killer T (NKT) cells while T cells reactive to group 1 CD1 molecules remain understudied, despite their higher abundance in humans compared to NKT cells. This review evaluates the mechanisms by which CD1-reactive, self-lipid specific T cells contribute to dyslipidemia-associated autoimmune disease progression or amelioration by examining available literature on NKT cells and highlighting recent progress made on the study of group 1 CD1-restricted T cells.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Gennaro De Libero, Universität Basel, Switzerland S. Subramanian, University of Washington, United States*

#### *\*Correspondence:*

*Chyung-Ru Wang chyung-ru-wang@ northwestern.edu*

#### *Specialty section:*

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

*Received: 03 April 2018 Accepted: 29 June 2018 Published: 16 July 2018*

#### *Citation:*

*Bagchi S, Genardi S and Wang C-R (2018) Linking CD1-Restricted T Cells With Autoimmunity and Dyslipidemia: Lipid Levels Matter. Front. Immunol. 9:1616. doi: 10.3389/fimmu.2018.01616*

Keywords: CD1, dyslipidemia, antigen presentation, autoreactive T cells, natural killer T cells, animal models

# INTRODUCTION

Over the years, it has become apparent that a range of rheumatological and dermatological autoimmune diseases like systemic lupus erythematosus (SLE) and psoriasis are associated with dyslipidemia (1, 2). In most cases, dyslipidemia in autoimmune diseases is characterized by altered serum cholesterol, triglyceride (TG), low-density lipoprotein (LDL), and high-density lipoprotein (HDL) levels (3–5). In this review, the words dyslipidemia and lipid abnormalities will be used interchangeably, while hyperlipidemia will be used to refer to cases in which an increase in cholesterol, TGs, and LDL was observed.

Atherogenic lipid profiles, characterized by increased serum cholesterol and TGs, have been observed up to 10 years prior to diagnosis of rheumatoid arthritis (6), while hypercholesterolemia in

**197**

SLE patients has only been reported after onset of disease (7, 8). Overall, these data indicate that dyslipidemia might be linked to autoimmune disease. However, whether dyslipidemia acts as a potential trigger for the initiation of autoimmune diseases has not been investigated in-depth. It is known that inflammation plays a key role in autoimmune diseases (9–12). Several studies have shown that obesity gives rise to low-grade chronic inflammation and the incidence and severity of autoimmune diseases, particularly, rheumatoid arthritis and psoriasis, are increased in obese patients compared to population controls (13). These data suggest that metabolic inputs influence inflammatory outputs (14). In obese individuals, for example, adipocytes release proinflammatory cytokines and adipokines, like TNF-α, IL-1β, and leptin, which activate both innate and adaptive arms of the immune system (14, 15). Therefore, in addition to genetic and environmental stress, dyslipidemia-induced chronic inflammation could be a driver of autoimmune diseases, suggesting an intricate interplay between lipid metabolism, activation of the immune system, and subsequent development of autoimmune diseases.

One of the major players of the adaptive immune system involved in the pathophysiology of autoimmune diseases is selfantigen reactive T cells (16). Even though most self-reactive T cells are eliminated in the thymus by the process of negative selection, some can escape. These T cells can recognize self-antigens in the periphery (16), so the autoantigens present within the affected tissues are thought to activate autoreactive T cells. Antigen-presenting cells like macrophages, dendritic cells, and B cells present the antigens to T cells, which secrete pro-inflammatory cytokines, leading to tissue damage (17). In the case of SLE, activated Th1 and Th2 effector T cells help B cells to produce autoantibodies (18). These antibodies can form immune complexes, which will then damage the kidney and lead to nephritis (19). In psoriasis, Th1-related cytokines like IFN-γ secreted by effector T cells are known to play a pathogenic role (20). Additionally, IL-17A is considered to be pathogenic in several autoimmune diseases and IL-17A blocking antibodies are currently used for treating autoimmune diseases like rheumatoid arthritis and psoriasis (20–22). Thus far, most studies have looked in to the role of peptide-specific autoreactive T cells in autoimmune disease initiation, progression, and maintenance. However, given that inflammation forms the crux of most autoimmune diseases, and that dyslipidemia is a potential trigger of chronic inflammation, it is imperative to uncover the role of lipid-specific autoreactive T cells in dyslipidemia-associated autoimmune diseases. Since lipid antigen-presenting molecules are widely expressed on a range of antigen-presenting cells in different tissues, it is conceivable that they contribute to the activation of cognate T cells when presented with lipids in inflammatory environments (**Figure 1)**. Work by our lab and others have demonstrated a role for lipid-autoreactive T cells in psoriasis (23, 24). Thus, understanding the role of self-lipid reactive T cells in dyslipidemia-associated autoimmune diseases would not only lead to better treatment options for a myriad of these diseases but also allow for development of preventive measures to either delay or eliminate their progression. Thus, the subsequent sections of this review will focus on detailed discussions of lipid reactive T cells and their role in major dyslipidemia-associated autoimmune diseases like SLE, psoriasis, and RA, as well as its associated comorbidities such as atherosclerosis and obesity.

#### CD1 MOLECULES

CD1 are a subset of MHC I-like molecules capable of presenting lipid antigens to T cells (25). Unlike MHC molecules, which are highly polymorphic, CD1 molecules exhibit very limited polymorphism, suggesting that antigens presented by each CD1 molecule are similar from one individual to the other (26). CD1 molecules are found in placental mammals and also birds, which indicates their ancient lineage (27). However, there are varying numbers and types of CD1 isoforms present in different animal species (28–32). CD1 molecules are divided into groups 1, 2, and 3 based on sequence homology and patterns of expression. Group 1 CD1 consists of CD1a, CD1b, and CD1c while CD1d is the sole group 2 CD1 molecule (33). While the abovementioned CD1 isoforms are expressed on the cell surface and present lipid antigens to T cells, the group 3 CD1, CD1e, acts as a chaperone in intracellular compartments to aid with antigen loading onto other CD1 molecules (34–36).

# CD1 Structure

*CD1* genes are paralogs of *MHC* genes and are unlinked from the *MHC* locus; genes encoding all CD1 isoforms are located on the long arm of chromosome 1q22-23 in humans (37–39). Like MHC I molecules, CD1 molecules form heterodimers of heavy α chains with β2 microglobulin (β2m), held together by non-covalent interactions (40–43). The antigen-binding grooves of CD1 molecules are usually narrower, deeper, and more voluminous than MHC I molecules and are lined with hydrophobic/neutral residues that facilitate binding of lipid molecules (44–48). This structural diversity allows CD1 isoforms to bind a range of different lipids, thus suggesting that each isoform may play a non-redundant role in the immune system.

#### Antigens Presented by CD1 Molecules

Several studies have shown that CD1 molecules can present self-lipids to cognate T cells; yet, the physiological implication of self-lipid presentation under homeostatic and disease conditions remains unclear. We have recently shown that under conditions of hyperlipidemia, presentation of phospholipids and cholesterol by CD1b to cognate T cells drove the development of an inflammatory skin disease resembling psoriasis. In line with our findings, other groups have shown that CD1b can present phospholipids and GM1, a prototypic ganglioside, to T cells (49, 50). Apart from CD1b, CD1d is known to bind a range of glycosphingolipids and phospholipids (51–55). Interestingly, even though the antigen-binding groove of each CD1 molecule is unique, sulfatide, a sulfated glycolipid, is presented by all CD1 molecules, suggesting that each CD1 isoform is capable of presenting both shared and unique lipids (56). Additionally, CD1a can present the autoantigens phosphatidylcholine, lysophosphatidylcholine, and skin-derived apolar, headless oils (57, 58). CD1c can present a unique leukemia-associated methylated-lysophosphatidic acid and cholesteryl esters (59, 60). The ability of CD1 molecules to present such a diverse array of self lipids suggests their potential role in eliciting T cell responses under both steady state and pathological conditions.

# CD1 Expression and Tissue Distribution

In humans, CD1 molecules are distributed on a myriad of cell types and tissues. Both group 1 (CD1a, CD1b, and CD1c) and group 2 CD1d molecules are expressed on double-positive (CD4<sup>+</sup>CD8<sup>+</sup>) thymocytes (61). In peripheral tissues, group 1 CD1 molecules are expressed exclusively by professional antigen-presenting cells. Dendritic cell subsets from lymph node and skin can express any of the group 1 CD1 isoforms, while B cells can express CD1c (61–63). In contrast to group 1 CD1, group 2 CD1d expression is more widely distributed, found on both hematopoietic and nonhematopoietic-derived cells. Examples of CD1d-expressing cells include epithelial cells of the small bowel and colon, keratinocytes in skin, and hepatocytes in liver (61). CD1 expression can be altered in various autoimmune and inflammatory conditions, thus dictating the flavor of lipid-specific T cell responses. For example, CD1d was upregulated in the psoriatic plaques from patients with active psoriasis, while patients with active psoriasis and dyslipidemia exhibited increased CD1b expression in skin (23, 64). In contrast, CD1d expression was lower in B cells isolated from SLE patients compared to healthy controls, resulting in reduced ability to stimulate CD1d-restricted T cell cytokine production *ex vivo* (65). CD1 a, b, c, and d were shown to be upregulated in human atherosclerotic plaques compared to non-atherosclerotic arteries, leading to the potential for increased CD1-restricted T cell activation and inflammation (66). Whether CD1 molecules become up or downregulated in the context of autoimmunity and inflammation seems to be dependent on their specific environments and much is still to be learned. Many researchers utilize mice for immunological studies; however, mice only express CD1d and do not express group 1 CD1 molecules, limiting the ability to study CD1-restricted T cells in the context of autoimmunity (67). As a result, little is known about group 1 CD1-restricted T cells under normal and pathologic conditions leading to autoimmunity.

# Dysregulation in Antigen Presentation Under Dyslipidemia

Homeostatic presentation of lipid antigens by CD1 molecules can be disrupted by dyslipidemia. Cholesterol uptake and storage is a tightly regulated process that becomes dysregulated with genetic pre-dispositions and/or chronic overnutrition. Under steady-state conditions, cholesterol metabolism is regulated both intracellularly (SREBP2 pathway) and in circulation (LDL metabolism) (68). Dysregulated cholesterol metabolism within antigen-presenting cells has been linked to autoimmunity: Ito et al. reported that cholesterol accumulation within CD11c<sup>+</sup> antigen-presenting cells drives autoreactive B cell and T cell expansion and promotes a lupus-like syndrome in mice (69). Our lab showed that serum from hyperlipidemic mice enhances IL-6 production by DCs, driving IL-17A production by autoreactive CD1b-restricted T cells in a model of psoriasis (23). Given that both group 1 and group 2 CD1 molecules are expressed by antigen-presenting cells and are upregulated in multiple autoimmune conditions, it will be important to further characterize the role that antigen-presenting cells play in autoimmunity and dyslipidemia in driving CD1-restricted T cell pathology.

# CD1-RESTRICTED T CELLS

## CD1d-Restricted Natural Killer T (NKT) Cells

CD1d-restricted NKT cells are divided into two main subsets, based on T cell receptor (TCR) usage. Type I NKT cells have an invariant TCR α chain (Vα14-Jα18 in mice and Vα24-Jα18 in humans) and thus are also referred to as invariant NKT (iNKT) cells (70–73). In mice, three β chains (Vβ7, Vβ8.2, and Vβ2) predominantly associate with the invariant α chain, while in humans, the invariant α chain pairs with Vβ11, an ortholog of the mouse Vβ8 (70–73). Unlike type I NKT cells, which have a semi-invariant TCR and recognize the marine sponge-derived lipid α-galactosylceramide (α-GalCer), type II NKT cells have diverse TCR usage and do not recognize α-GalCer (74–78). Rare populations of NKT cells (making up <1% of hematopoietic cells) have also been reported: NKT cells that use γδ chains for their TCR (primarily Vγ1.1 and Vδ6.3), and α-GalCer-reactive NKT cell population harboring a Vα10-Jα50 TCR were identified in mice (79, 80). Vα24<sup>−</sup> CD1d-α-GalCer-specific T cells expressing CD4 or CD8αβ and using diverse Vα/Vβ chains have been reported in humans (78). NKT cells are "educated" in the thymus where CD1d-expressing cortical thymocytes mediate their positive selection as opposed to thymic epithelial cells that select conventional T cells (81). The lipid(s) responsible for the selection of NKT cells are largely unknown, though a recent study suggested that ether-linked lysophospahtidylethanolamine and lysophosphatidic acids might play a role in the thymic selection of iNKT cells (82).

In the thymus, NKT cells are characterized by the expression of the transcription factor PLZF (promyelocytic leukemia zinc finger protein), which is thought to impart the "innate-like" features to these T cells (83, 84). As NKT cells exit the thymus, they exhibit a pre-activated phenotype and have the capacity to rapidly produce Th1, Th2, and Th17 related cytokines upon TCR stimulation (85). These cytokines are produced when NKT cells interact with either self or foreign lipid antigens presented by the CD1d molecule. It is known that a small proportion of conventional T cells are self-peptide reactive. In contrast, most NKT cells can recognize self-lipid antigens, although the ability of self-lipids to stimulate cytokine production is dependent on two inputs: (1) the strength of TCR signaling and (2) the presence of cytokine driven co-stimulation (e.g., IL-12/IL-18 secreted by TLR-activated DCs) (86, 87). The nature of self-lipids that activate NKT cells during steady state and/or during a specific pathogenic challenge remain largely unknown. Thus, by virtue of their pre-activated status and their ability to be activated by self-antigens in the presence of the correct cytokine milieu, it is conceivable that NKT cells play an important role, either pathogenic or protective, in a range of infectious and autoimmune diseases as well as tumor immunity.

# Group 1 CD1-Restricted T Cells

In contrast to the copious amounts of information available on NKT cells, progress on group 1 CD1-restricted T cells is limited. Most studies have made use of long-term T cell clones isolated from patients infected with *Mycobacterium tuberculosis* and *Mycobacterium leprae* (88–94), though T cell clones derived from multiple sclerosis patients showing autoreactivity to several self glycosphingolipids have been described in the literature (49, 95). These T cell clones mostly have a diverse αβ TCR repertoire and can be CD4 or CD8 single positive or double negative, and capable of producing Th1, Th2, and/or Th17 related cytokines (88–94). While some of these T cell clones recognize lipid antigens from the mycobacterial cell wall, most CD1-restricted T cell clones described are autoreactive. In fact, the frequency of autoreactive group 1 CD1-restricted T cells makes up between 1 in 10 and 1 in 300 of all circulating T cells in humans, suggesting that they represent a substantial part of the T-cell repertoire in humans (96). Since most of the knowledge about these T cells comes from the study of T cell clones isolated from humans, the developmental program of group 1 CD1-restricted T cells and their physiologic responses during infection and autoimmunity are mostly unknown. Therefore, our lab generated a TCR transgenic mouse, expressing a CD1b-restricted self-lipid reactive TCR (HJ1Tg) and crossed it with mice co-expressing group 1 CD1b and CD1c (hCD1Tg) (97).

Characterization of HJ1Tg/hCD1Tg mice demonstrated that positive selection of autoreactive group 1 CD1-restricted T cells was mediated by thymocytes, similar to iNKT cells (97). CD1bautoreactive HJ1 T cells were enriched in the liver and exhibited an activated/effector phenotype (CD44hi, CD69<sup>+</sup>, CD122<sup>+</sup>) in the naïve setting (97). Additionally, HJ1 T cells were shown to be protective against *Listeria monocytogenes* infection (97) and contribute to antitumor immunity against a CD1b<sup>+</sup> T cell lymphoma (97, 98). In a recent study, we also demonstrated that under conditions of hyperlipidemia, HJ1 T cells contributed to the development of an inflammatory psoriasis-like skin inflammation (23). Although the nature of the lipid(s) recognized by HJ1 T cells in the context of *Listeria* infection remains unknown, HJ1 T cells in the context of hyperlipidemia were most likely activated by excess phospholipid and cholesterol species (23). These data suggest that while transient activation of group 1 CD1-autoreactive T cells may play a protective role in infections, chronic activation of autoreactive CD1-restricted T cells could lead to detrimental effects like initiation of inflammatory conditions and autoimmunity. A comparison of group 1 and group 2 CD1-restricted T cells is in **Table 1**.

## THE ROLE OF CD1-RESTRICTED T CELLS IN AUTOIMMUNE DISEASES CORRELATED WITH DYSLIPIDEMIA

# Systemic Lupus Erythematosus

It has been well documented that atherosclerosis-related cardiovascular complications comprise the leading cause of death in SLE patients (2). Age, arterial hypertension, smoking habits, diabetes mellitus, obesity, and dyslipidemia are all known to be risk factors for the development of atherosclerosis-related cardiovascular diseases (CVD) among these patients (2). Dyslipidemia is specifically defined in SLE patients as increased total cholesterol, TG, LDL, and decreased levels of HDL (2). Studies have reported approximately 30% of SLE patients displaying signs of hyperlipidemia at the time of diagnosis, which increased to about 60% at 3 years post-diagnosis (99). Even though the mechanisms that cause hyperlipidemia in lupus patients is not well understood, some studies have suggested that the activity of enzyme lipoprotein lipase (LPL), which metabolizes lipids, is reduced and autoantibodies to LPL could be a reason for its reduced activity (100, 101). Like all autoimmune diseases, lupus is characterized by inflammation, often systemic. Under such conditions, LDL


Table 1 | Comparison of group 1 and group 2 CD1-restricted autoreactive T cells.

particles are more prone to oxidation (ox-LDL) (102). In fact, the process of atherosclerosis is initiated when macrophages ingest ox-LDL to become foam cells. Therefore, not surprisingly, anti-ox-LDL antibodies are present in lupus patients, contributing to the development of dyslipidemia-related cardiovascular ailments (103).

Since dyslipidemia and related pathologies are common in lupus patients and T cells are important players in lupus, exploring the role of lipid-reactive T cells is important. As iNKT cells are the most studied lipid-responsive T cell subset, copious amounts of information are available on these T cells in lupus, though some of the data are contradictory. When lupus was induced in C57BL/6 mice by injecting apoptotic cells, iNKT cells were activated and produced more IL-10 compared to IFN-γ; the absence of iNKT cells was associated with exacerbated disease as a result of increased autoantibody production and glomerular immune complex deposition (104). NKT cells also played a protective role when pristane, a hydrocarbon oil, was injected into BALB/c mice to stimulate a lupus-like disease. Absence of NKT cells (CD1d<sup>−</sup>/<sup>−</sup> mice) in this model led to increased nephritis and serum autoantibodies (105). iNKT cell-specific cytokine production, IL-4, and TNF-α, were also decreased, with an expansion of marginal zone B cells (105). Interestingly, activation of iNKT cells with α-GalCer, upon induction of lupus using pristane, protected BALB/c mice from disease, but aggravated disease in SJL mice (106). As noted in the SJL mice, several studies have reported that iNKT cells in BWF1 mice induced autoantibody production from B cells, which promotes lupus pathogenesis (107, 108). In humans, CD1d expression on B cells and iNKT cell frequency and proliferative capacity generally decreases in lupus patients (65, 109, 110), suggesting a protective role for iNKT cell during lupus. However, another study showed that iNKT cells from SLE patients could induce CD1d-dependent CD40/CD40L-dependent anti-dsDNA antibody production by B cells, demonstrating a pathogenic role for iNKT cells in SLE (111). Aside from NKT cells, the role of other lipid-reactive T cells in SLE remains largely unexplored. One study showed that CD1c-restricted T cells isolated from SLE patients can promote class-switched IgG autoantibodies, mediated by CD1c, IL-4, and CD40 (112). Given that dyslipidemia is very prevalent in lupus patients, it is not surprising that lipidspecific autoreactive T cells may contribute to disease progression. However, due to the conflicting role of iNKT cells and the dearth of knowledge about other CD1-restricted T cells in lupus, more research needs to be conducted before their potential can be harnessed in the clinic.

#### Psoriasis

Psoriasis, a primarily T cell driven autoimmune disease, which affects about 1–3% of the world's population, is associated with hyperlipidemia (3, 113). Psoriatic patients also have a higher risk of developing cardiovascular disease such as atherosclerotic plaque formation (3). Additionally, psoriatic patients exhibit signs of systemic inflammation, which is largely mediated by neutrophils and T cells (20, 114). Although psoriatic patients are hyperlipidemic and psoriasis is a T cell driven disorder, the role of self-lipid reactive CD1-restricted T cells in psoriasis remains nebulous. Multiple studies have shown an increase in iNKT cells and CD1d expression within psoriatic lesions of mice and humans (64, 115, 116). Human skin engraftment of psoriatic and non-psoriatic skin onto SCID (severe combined immunodeficiency) mice has been the sole mouse model to study NKT cells in psoriasis; this model recapitulates a psoriasis-like disease in engrafted human skin when injected with psoriatic patient-derived lymphocytes (117, 118). An NKT cell line derived from a psoriatic patient was capable of producing IFN-γ when cocultured with CD1dexpressing keratinocytes and induced psoriasis plaque formation when injected into engrafted human skin SCID mice (64, 117). Since CD1d is expressed on keratinocytes and their expression is upregulated in psoriatic skin, it is thought that lipid antigens presented by these cells can activate NKT cells *in vivo* (61, 64).

Given the high frequency of autoreactive CD1-restricted T cells in humans and the presence of excess lipids under conditions of hyperlipidemia, it is surprising that the role of lipid-specific T cells remains understudied. Hyperlipidemia is induced in mice by either feeding them a high-fat diet, using mice that eat excessively due to genetic manipulations (obese mice) or knocking out genes important for lipid clearance from the blood. Recent work by our lab showed that upon induction of hyperlipidemia, CD1b-autoreactive T cells contributed to the development of psoriasis-like skin inflammation, characterized by a Th17 phenotype (23). Furthermore, there was preferential accumulation of phospholipids and cholesterol in the skin of these mice and the aforementioned lipids could be presented by CD1b to activate CD1b-autoreactive T cells. In humans, CD1b expression was increased in psoriatic compared to normal human skin and there were more CD1b-autoreactive T cells in the blood of affected individuals (23). Additionally, circulating CD1a-autoreactive T cell frequency increases in psoriasis patients compared to healthy controls (119). Presentation of self-lipids on CD1a drives the activation of cognate T cells with a Th17 effector phenotype. This results in the development of psoriatic-like skin inflammation (24, 119). Finally, headless apolar skin lipids can be presented by CD1a to CD1a-autoreactive T cells (58). It can be anticipated that some of these T cells might play a role in psoriasis, especially because they produce IL-17A and IL-22 in response to antigenic stimulation (58). A more indepth examination of these T cells in psoriasis is warranted to determine the necessity of autoreactive CD1-restricted T cells in driving disease progression.

## Rheumatoid Arthritis

Rheumatoid arthritis is an autoimmune disease that affects the connective tissue of the synovial joints. It affects about 0.3–1% of the world's population, with women being more susceptible than men (120). Interestingly, it is now clear that affected individuals have a 50% increased risk of developing CVD, even though levels of lipids associated with an increased risk of CVD, like cholesterol and LDL, do not always show an association with the development of RA (121–124). The mechanisms of RA disease development and progression are not fully understood; it is speculated that RA-associated inflammation consists of a proinflammatory milieu with self-reactive T and B cells potentially contributing to disease pathogenesis. Thus, it is conceivable that self-lipid reactive T cells might play a role in disease development. It has been demonstrated that iNKT cell number and function is altered in the peripheral blood of RA patients (125). Along with decreases in iNKT cell frequency, their capacity to secrete Th2 related cytokine secretions is also diminished. These findings suggest that iNKT cells may be involved in the disease process of RA, although further studies need to be completed to link these findings with physiological relevance (126). In collagen-induced arthritis (CIA) mouse models, which are commonly used as a model for RA, the absence of iNKT ameliorates disease severity (127, 128). Similar results were obtained when the interaction of CD1d with NKT cells was blocked by administration of anti-CD1d antibody (128). Interestingly, administration of OCH, a ligand known to skew iNKT cells to a Th2 type response, ameliorated disease in wildtype but not iNKT cell-deficient mice (129). Further, neutralization of IL-4 and IL-10 abrogated the OCH-mediated therapeutic benefit (129). Another mouse model of RA uses serum or immunoglobulins from K/BxN mice transferred to wild-type mice to promote joint inflammation. RA disease pathology score was reduced in NKT cell-deficient mice transferred with serum compared to wild-type mice (130). Additionally, iNKT cells were shown to infiltrate the joints of mice given K/BxN serum, secreting IFN-γ and IL-4, which inhibited anti-inflammatory TGF-β secretion in joint fluid (130). These studies show that CD1d-restricted NKT cells contribute to the pathogenesis in multiple mouse models of RA. However, the role of group 1 CD1-restricted T cells in RA remains unknown. Further, whether the role of self-lipid reactive T cells changes in this disease model under conditions of dyslipidemia needs to be elucidated.

#### Obesity and Atherosclerosis

Due to the ease of tracking iNKT cells *in vivo*, it is no surprise that studies have examined the role of these T cells in high-fat diet fed, obese mice. In general, iNKT cells in adipose tissues have been implicated in maintaining immune homeostasis by producing IL-10, inducing an anti-inflammatory phenotype in macrophages (mediated by IL-4/STAT-6) and controlling the number and function of regulatory T cells (131). Recent studies have shown that CD1d expression and iNKT cell numbers decrease in obese humans and mice, leading to increased recruitment of proinflammatory macrophages in adipose tissues and insulin resistance (132–135). The cause of this decrease in iNKT cells is unclear, but insufficient stimulation of these T cells due to decreased CD1d expression could be a possibility. In contrast to the findings of the aforementioned studies, Wu et al. demonstrated that iNKT cells cause tissue inflammation in both high fat diet fed and obese mice (136). Further, adipocyte-specific CD1d deficiency ameliorated high fat diet-induced obesity and insulin resistance (137). A pathogenic contribution of iNKT cells under conditions of hyperlipidemia was also reported using β2m knockout mice injected with α-GalCer and fed a high fat diet. These mice had decreased inflammatory macrophage infiltration into adipose tissue compared to control C57BL/6 mice injected with α-GalCer and fed with a high fat diet (138). Since β2M knockout mice lack both CD8<sup>+</sup> and NKT cells, the contribution of NKT cells alone to macrophage recruitment was not examined. Lastly, while one study has demonstrated a pathogenic role for type II NKT cells in high fat diet fed mice (139), no studies to date have explored group 1 CD1 expression and the role of group 1 CD1-restricted T cells in obesity.

Aside from diet-induced hyperlipidemia, knocking out genes important for lipid clearance like Apolipoprotein E (*Apoe*) and low density lipoprotein receptor (*Ldlr*) results in accumulation of lipids, characterized by increased serum cholesterol, TG and LDL with decreased HDL (140, 141). This lipid profile is similar to dyslipidemia observed in patients with autoimmune diseases. Traditionally these knockout mouse models have been used to study the formation of atherosclerotic plaques in murine blood vessels; unlike humans, mice do not naturally develop atherosclerotic plaques, even when put on high fat diet. Interestingly, in both *Apoe* and *Ldlr* knockout mice, a pathogenic role for iNKT cells has been reported. For example, CD1d deficiency significantly reduced atherosclerotic burden in atherosclerosisprone mice (142–144). When mice were injected with α-GalCer, the process of atherosclerosis was accelerated and secretion of proatherogenic cytokines increased (142–144). Finally, it has been reported that iNKT cells play a pathogenic role during the initial phases of atherosclerotic plaque formation, but not during the later stages (143). Since atherosclerotic plaques harbor different species of oxidized and modified lipids, it is thought that CD1d-expressing DCs present these lipids to NKT cells, leading to their activation. This causes secretion of proinflammatory


Table 2 | Lipid-specific T cell involvement in autoimmune diseases and dyslipidemia.

*ND, not determined.*

cytokines like IFN-γ, aggravating disease. A recent study showed that a monoclonal CD1b-autoreactive T cell did not significantly contribute to plaque formation in ApoE<sup>−</sup>/<sup>−</sup> mice at later stages of disease (30-week-old mice) (23). Since it is now appreciated that immune cells are instrumental in the maintenance of atherosclerotic plaques, it is of interest to further explore the role of group 1 CD1-autoreactive T cells in this disease, especially because autoreactive group 1 CD1-restricted T cells form a large proportion of CD1-reactive T cells in humans, and group 1 CD1 molecules are expressed in human atherosclerotic plaques but not in normal arterial walls (66).

## The Effect of Statins, a Class of Lipid-Lowering Drugs, on Autoimmune Diseases

Statins, which are prescribed for lowering lipid levels in individuals with atherosclerosis, have favorable effects on autoimmune disease activity. In SLE patients, statin treatment was shown to lower proteinuria (145, 146). Additionally, statins can also have immunomodulatory functions such as increasing the presence of regulatory T cells, which are crucial for keeping autoimmune diseases at bay (147–149). Several studies have also noted a decrease in psoriasis area severity index scores in patients with psoriasis upon statin treatment (150, 151). However, another study did not find any statistically significant differences in psoriasis activity upon statin treatment (152). Therefore, it is too early to determine whether statins indeed ameliorate psoriasis. Finally, statin treatment has been shown to reduce disease activity scores of RA patients (153). These studies suggest a close link between lipid abnormalities and autoimmune diseases.

# CONCLUSION

Several autoimmune diseases have been strongly linked to lipid abnormalities (2). Whether dyslipidemia acts as a trigger for disease development or whether dyslipidemia is an outcome of autoimmune diseases is not well understood. In diseases like atherosclerosis and SLE, hyperlipidemia is known to develop before disease onset (2) and cause chronic inflammation (154). T cells play an important role in autoimmune disease progression; therefore, it is essential to decipher the role of T cells, specifically self-lipid reactive T cells, in dyslipidemia-associated autoimmune diseases. Current information on the role of these T cells in autoimmune diseases is shown in **Table 2**. As mentioned above, most of the knowledge comes from the study of iNKT cells. Even so, a slew of contradictory results has made it difficult to reach definitive conclusions about their role under conditions of dyslipidemia-associated autoimmunity. The conflicting results regarding the role of iNKT cells in various diseases may arise from a range of factors. For example, it is known that different subsets of iNKT cells have different functions (155). A majority of iNKT cells are CD4<sup>+</sup> and can secrete IFN-γ and IL-4 under different stimuli (156). However, NK1.1<sup>−</sup> iNKT cells are known to secrete more IL-17A (157). Additionally, the localization and microenvironment affect the function of iNKT cells. For example, adipose tissue-derived iNKT cells are known to play a regulatory role by not only producing IL-10 but also stimulating the activity of regulatory T cells (131). Furthermore, the genetic background of the mice can also affect iNKT function and thus their role in various disease conditions. It is known that iNKT cells from BALB/c mice secrete more IL-4 compared to IFN-γ than iNKT cells from C57BL/6 mice (155). Finally, microbiome compositions, which differ among animal housing facilities, may lead to discrepancies in results. Therefore, it is important to carefully evaluate these parameters before comparing results about iNKT cells in scientific studies. An additional parameter to note is that mouse models often do not fully mimic human autoimmunity. For example, mouse models of RA share autoreactive T cells, citrullinated autoantibodies, and macrophage/neutrophil infiltrate, but based on the method of disease induction, do not recapitulate the distribution of rheumatoid pannus and necessity of T cells as drivers of disease seen in humans (158). These reasons above highlight the importance of corroborating data from mouse models with human studies.

Aside from CD1d, the study of group 1 CD1-restricted T cells under conditions of dyslipidemia and autoimmunity remains obscure. However, recent studies from our lab demonstrated a pathogenic role of CD1b-autoreactive T cells in hyperlipidemiaassociated psoriasis-like skin inflammation (23). Additionally, very little is known about CD1d-restricted type II NKT cells, owing to the lack of markers to track them *in vivo*. The presence of type II NKT cells in mice makes them more amenable than group 1 CD1-restricted T cells to study in murine mouse models; however, the differences in lipid metabolism between mice and humans makes it critical to corroborate any murine findings of type II NKT cells with human studies. It is important to study these T cell subsets in more depth not only because they form a substantial proportion of the total human T cell population but also because a majority of these T cells are known to be autoreactive.

Though the role of iNKT cells has been deciphered in some autoimmune diseases, the identity of self-lipids recognized by these T cells remain largely unknown. The general thought is that antigen-presenting cells expressing CD1-self-lipid complexes activate CD1-restricted T cells. Activated T cells produce cytokines that are either immunomodulatory or inflammatory.

#### REFERENCES


The nature of the cytokines secreted is influenced by the functional state of the antigen-presenting cells, the type of self-lipid being presented, and the microenvironment. It is crucial to identify these lipids not only for tracking these T cells *in vivo,* but also for harnessing their potential in the clinic. Additionally, very few studies have examined the role of CD1-restricted T cells under dyslipidemia and autoimmune disorders. Thus, the study of self-lipid reactive T cells in this context is still in its infancy; more research investigating the links between autoimmunity, inflammation, and dyslipidemia could inform both diagnosis and treatment of autoimmune diseases in the future.

#### AUTHOR CONTRIBUTIONS

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

#### ACKNOWLEDGMENTS

We sincerely apologize to colleagues whose work we have not cited due to space constraints or oversight. We thank Lavanya Visvabharathy and Eva Morgun for critical reading and helpful discussions. This work was supported by the National Institutes of Health R01 grants AI057460 and AI43407.


autoimmune disease. *Immunity* (2016) 45(6):1311–26. doi:10.1016/j.immuni. 2016.11.008


with collagen-induced arthritis. *Clin Exp Immunol* (2005) 141(1):47–53. doi:10.1111/j.1365-2249.2005.02817.x


diseases: a preliminary study. *Lupus* (2003) 12(8):607–11. doi:10.1191/09612 03303lu429oa


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

# Factors Influencing Functional Heterogeneity in Human Mucosa-Associated Invariant T Cells

*Joana Dias1 , Caroline Boulouis1 , Michał J. Sobkowiak1 , Kerri G. Lal1,2,3, Johanna Emgård1 , Marcus Buggert1 , Tiphaine Parrot1 , Jean-Baptiste Gorin1 , Edwin Leeansyah1,4 and Johan K. Sandberg1 \**

*1Center for Infectious Medicine, Department of Medicine, Karolinska Institutet, Karolinska University Hospital Huddinge, Stockholm, Sweden, 2U.S. Military HIV Research Program, Walter Reed Army Institute of Research, Silver Spring, MD, United States, 3Henry M. Jackson Foundation for the Advancement of Military Medicine, Bethesda, MD, United States, 4Program in Emerging Infectious Diseases, Duke-National University of Singapore Medical School, Singapore, Singapore*

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Derya Unutmaz, Jackson Laboratory for Genomic Medicine, United States Agnes Lehuen, Institut National de la Santé et de la Recherche Médicale (INSERM), France Asako Chiba, Juntendo University, Japan*

> *\*Correspondence: Johan K. Sandberg johan.sandberg@ki.se*

#### *Specialty section:*

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

*Received: 26 April 2018 Accepted: 27 June 2018 Published: 10 July 2018*

#### *Citation:*

*Dias J, Boulouis C, Sobkowiak MJ, Lal KG, Emgård J, Buggert M, Parrot T, Gorin J-B, Leeansyah E and Sandberg JK (2018) Factors Influencing Functional Heterogeneity in Human Mucosa-Associated Invariant T Cells. Front. Immunol. 9:1602. doi: 10.3389/fimmu.2018.01602*

Mucosa-associated invariant T (MAIT) cells are unconventional innate-like T cells that recognize microbial riboflavin metabolites presented by the monomorphic MHC class I-related (MR1) molecule. Despite the high level of evolutionary conservation of MR1 and the limited diversity of known antigens, human MAIT cells and their responses may not be as homogeneous as previously thought. Here, we review recent findings indicating that MAIT cells display microbe-specific response patterns with multiple layers of heterogeneity. The natural killer cell receptor CD56 marks a MAIT cell subset with distinct response profile, and the T cell receptor β-chain diversity influences responsiveness at the single cell level. The MAIT cell tissue localization also influences their response profiles with higher IL-17 in tissue-resident MAIT cells. Furthermore, there is emerging evidence that the type of antigen-presenting cells, and innate cytokines produced by such cells, influence the quality of the ensuing MAIT cell response. On the microbial side, the expression patterns of MR1-presented antigenic and non-antigenic compounds, expression of other bioactive microbial products, and of innate pattern recognition ligands all influence downstream MAIT cell responses. These recent findings deepen our understanding of MAIT cell functional diversity and adaptation to the type and location of microbial challenge.

Keywords: mucosa-associated invariant T cells, MHC class I-related, CD56, cytokines, microbial immunity, antibacterial immunity, mucosal immunology

# INTRODUCTION

Mucosa-associated invariant T cells are unconventional T cells operating on the border between the innate and adaptive immunity and respond promptly in an innate-like manner to antigens presented by the MHC class I-related (MR1) protein (1, 2). Human mucosa-associated invariant T (MAIT) cells express a semi-invariant T cell receptor (TCR) characterized by the uniform use of the Vα7.2 segment paired with Jα12, 20, or 33, whereas the β-chain diversity is broader but still limited (3–7).

The naturally occurring MR1-presented MAIT cell agonists identified to date are metabolites derived from the riboflavin biosynthesis pathway (8, 9). This limited set of antigens coupled with the high evolutionary conservation of MR1 (10) has favored the notion that MAIT cells may be functionally homogeneous and responding in a largely undifferentiated manner to microbes capable of riboflavin production. However, recent studies have demonstrated that MAIT cells are in fact fairly heterogeneous in their phenotype and function, and that their response patterns are influenced by a range of factors.

# MAIT CELL RESPONSE PATTERNS VARY BASED ON MICROBIAL STIMULI AND TISSUE LOCALIZATION

We recently found that peripheral blood MAIT cells respond differently to distinct microbes in both the quality and quantity of cytokines produced (7). The Gram-negative bacterium *Escherichia coli* induced production of interferon (IFN)γ and tumor necrosis factor (TNF), as well as TCR downregulation, at significantly higher levels than the opportunistic fungus *Candida albicans,* and the MAIT cell polyfunctional cytokine profile significantly differed in response to these microbes (7). Moreover, several studies have provided evidence that MAIT cell effector functions vary based on tissue localization. Upon bacterial stimulation, peripheral blood and gastric MAIT cells produce IFNγ and TNF, and degranulate (2, 4, 11–15), whereas MAIT cells from the female genital mucosa display a bias toward production of interleukin (IL)-17 and IL-22 (16). Moreover, in response to *Mycobacterium tuberculosis* stimulation, MAIT cells from tuberculous pleural effusions display an enhanced capacity to produce IFNγ, IL-17F, and granzyme B than circulating MAIT cells (17). Upon phorbol myristate acetate and ionomycin stimulation, MAIT cells from the liver and adipose tissue produce more IL-17 and IL-10, respectively, than their peripheral blood counterparts (18, 19). Data from mouse models further support a role of MAIT cells in the control of type 1 diabetes *via* maintenance of gut integrity and control of anti-islet autoimmune responses (20), as well as of pulmonary infection by *Francisella tularensis* live vaccine strain (LVS) (21, 22). Overall, these findings suggest the existence of MAIT cell response patterns that vary with tissue localization and depend on the microbes encountered.

Antimicrobial immune responses are an outcome of the interplay between effector cells, antigen-presenting cells (APCs), and microbes. Recent findings have indicated that MAIT cells are phenotypically heterogeneous and comprise functionally distinct subsets (7). Thus, the functional compartmentalization of the MAIT cell population, together with distinct characteristics of APCs and microbes, may influence MAIT cell responses upon microbial encounter.

## MAIT CELLS—NOT AS HOMOGENEOUS AS THEY FIRST MAY SEEM

Adult peripheral blood MAIT cells were long considered phenotypically homogeneous in that they express a restricted semiinvariant TCR α-chain and predominantly exhibit a CD45RO<sup>+</sup> CCR7<sup>−</sup>CD62L<sup>−</sup>CD28<sup>+</sup> effector memory phenotype (3, 7, 23, 24), as determined by individual assessment of surface receptors (23, 24) and by screening of their surface immune-proteome (7). However, MAIT cells vary in their expression of TCR Vβ segments (3–7), and of the natural killer (NK) cell-associated receptor CD56 (7). Thus, the discovery of these phenotypically distinct MAIT cell populations suggested the existence of subsets that could potentially exhibit different functional properties.

# THE TCR **β**-CHAIN COMPOSITION INFLUENCES MAIT CELL ANTIMICROBIAL RESPONSES

Although less diverse than that of other T cells (5, 6), the Vβ usage of MAIT cells adds some diversity to their overall TCR β-chain repertoire. We observed that the Vβ segment expression can influence MAIT cell responses, as MAIT cells expressing Vβ8<sup>+</sup>, Vβ13.1<sup>+</sup>, and Vβ13.6<sup>+</sup> were hyporesponsive to *E. coli*, and Vβ13.2<sup>+</sup> MAIT cells were slightly hyperresponsive to *C. albicans* when compared with the total MAIT cell population (7). Lopez-Sagaseta et al. (25, 26) had previously reported different binding affinities between MAIT cell TCRs with different Vβ segments and MR1 in complex with a MAIT cell agonist. Thus, while the semi-invariant α-chain is indispensable for TCR recognition of MR1–ligand complexes (25, 27), the TCR β-chain may influence MAIT cell antimicrobial responses by fine-tuning the overall TCR–ligand–MR1 interaction. In light of the aforementioned findings, one can speculate that localization or accumulation of Vβ13.2<sup>+</sup> MAIT cells, which comprise a significant proportion of the total MAIT cell population (7), at sites of *C. albicans* colonization, such as the genitourinary tract (28), could boost local immune responses against this opportunistic pathogen.

Mucosa-associated invariant T cell subpopulations defined by Vβ expression also have differential proliferative capacity *in vitro*. MAIT cells that express the more abundant Vβ's proliferate more *in vitro* in response to *E. coli* than the less abundant ones (7). This finding raises the possibility that the *in vivo* interactions with microbes believed to drive the expansion of MAIT cells from the low frequencies seen in cord blood also shape the Vβ repertoire by selectively driving the expansion of more responsive MAIT cell subsets in an antigen-dependent manner. If this is the case, the MAIT cell TCR repertoire might be influenced by vaccination strategies that expose individuals to microbial antigens. In agreement with this, Howson et al. (29) recently reported a preferential expansion of certain MAIT cell clonotypes in human volunteers challenged with *Salmonella enterica* serovar Paratyphi A (29). Interestingly, the MAIT cell clonotypes that expanded *in vivo* were more strongly activated *in vitro* in an MR1-dependent manner than those that contracted during infection, potentially due to higher functional avidity between their TCRs and MR1 ligands (29). Thus, the MAIT cell TCR β-chain repertoire may function as a bacterial infection signature of any given individual. Furthermore, factors such as the geographic location, diet, or medication usage [all of them known to affect the microbiota (30, 31)] might shape the MAIT cell TCR β-chain repertoire as well. Hinks et al. (32) reported that the levels of MAIT cells in peripheral blood and bronchial tissues were affected in steroid-treated chronic obstructive pulmonary disease patients when compared with non-steroid-treated patients (32). However, it remains to be investigated if this or any other factor influences the MAIT cell TCR β-chain repertoire through its effect on the microbiota.

### CD56 MARKS A MAIT CELL SUBSET WITH ENHANCED INNATENESS

A proportion of human MAIT cells in peripheral blood express the NK cell marker CD56 (7). Interestingly, the CD56-expressing MAIT cells have a higher capacity to respond to IL-12 and IL-18 than their negative counterparts (7). This can possibly be explained by their higher expression levels of IL-12R and IL-18R, as well as higher levels of the transcription factors PLZF, Eomes, and T-bet (7). The higher responsiveness of CD56<sup>+</sup> MAIT cells to innate cytokines may make them more efficient in mounting MR1-independent responses during viral and bacterial infections, as well as sterile inflammatory conditions. In addition, CD56<sup>+</sup> MAIT cells are reportedly more abundant in the liver than in peripheral blood (19, 33). Whether this MAIT cell subset has protective, pathogenic, or modulatory roles in liver diseases such as viral hepatitis remains to be determined.

In summary, the type and magnitude of effector functions mounted in response to stimuli can be influenced by factors intrinsic to the MAIT cells: the TCR β-chain composition and CD56 expression (**Figure 1**). Thus, the relative amounts of CD56<sup>+</sup> and CD56<sup>−</sup> MAIT cells and of Vβ-defined MAIT cell subsets, the latter potentially already determined by previous microbial encounters *in vivo*, might play an important role in determining how the MAIT cell compartment will respond to a new antigenic challenge.

#### DIFFERENTIAL DEPENDENCE ON MR1 FOR MAIT CELL CYTOKINE PRODUCTION

At steady state, MR1 is mostly retained intracellularly (34–37) and traffics to the cell surface upon ligand availability and APC activation (34, 36, 38). Interestingly, blocking experiments have revealed differential MR1-dependency in MAIT cell responses (7). In responses to both *E. coli* and *C. albicans*, most IFNγ and virtually all TNF produced by MAIT cells was dependent on the TCR–MR1 interaction (7). However, a small proportion of IFNγ was produced in an MR1-independent manner (7). The dependency on MR1-mediated antigen presentation for MAIT cell production of TNF suggests tight regulation of their pro-inflammatory responses. This may be an important regulatory mechanism to prevent MAIT cell activation in response to riboflavin biosynthesis-competent commensal microbes that do not actively produce MAIT cell agonists at steady state, but still reside close to MAIT cells. Moreover, the low levels of MR1 on most cells at steady state may prevent continuous activation of

cytokine production and cytolytic capacity, in different tissues, such as peripheral blood and the female genital mucosa.

MAIT cells by MR1 ligands from commensal microbes. As MR1 can bind extracellular ligands directly on the cell surface, these ligands, even if present at homeostatic levels, could otherwise lead to unnecessary MAIT cell pro-inflammatory responses (36).

# INNATE CYTOKINES ACTIVATE MAIT CELLS IN AN MR1-INDEPENDENT MANNER

Microbes activate APCs to secrete cytokines, such as IL-12 and IL-18, which induce IFNγ production by MAIT cells independently of TCR signaling and MR1 (39). In agreement with these initial findings, Jesteadt et al. (40) recently reported different MAIT cell responses to two microbes that differ in their capacity to activate the inflammasome, a proinflammatory innate immune system collection of receptors and sensors that involve activation of caspase-1 and inflammatory molecules to both microbes and host-derived proteins (40, 41). In contrast to *Francisella tularensis* LVS, *Francisella novicida* is a strong inflammasome activator that induced high levels of IL-18 production by macrophages and subsequent high levels of IFNγ production by murine MAIT cells (40). Moreover, the magnitude of the *in vitro* IFNγ production in response to *F. novicida* was directly influenced by the concentration of IL-18 in the cultures (40). Therefore, one can expect differential MAIT cell responses to microbes with different ability to induce IL-18 production by the APCs.

Other cytokines can have a range of effects on peripheral blood MAIT cells. In the absence of microbial stimulation, IL-7 induces GrzB and upregulates Prf expression without concomitant production of cytokines (14), whereas IL-15 in combination with IL-18 and/or IL-12 induces IFNγ and GrzB production (42–44). Upon stimulation with suboptimal doses of *E. coli*, both IL-7 and IL-15 augment cytokine and cytolytic molecule expression by MAIT cells (14, 42). Moreover, the combination of MR1 antigen presentation with either IL-12, or IL-7 and IL-12, induces GrzB production by MAIT cells in response to *E. coli* (13) or nontypeable *Haemophilus influenzae* (NHTi), respectively (45). Thus, it is plausible that APCs shape MAIT cell antimicrobial responses through the cytokines they produce upon microbial-mediated activation. In fact, the capacity of MAIT cells to be activated by cytokines alone underlies their ability to respond *in vitro* to several viruses, including dengue virus, influenza virus, and hepatitis C virus, in a process dependent on IL-12 and IL-18, IL-18 alone, and IL-18 and IL-15, respectively (43, 46). In addition, both IFNα and IFNβ were shown to activate MAIT cells (43, 47) and further contribute to the MAIT cell response to HCV (43). Moreover, MAIT cells can respond to the superantigen staphylococcal enterotoxin B independently of MR1 and in an IL-12, IL-18, TCR Vβ, and HLA class II-dependent manner (48, 49).

#### DIFFERENT TYPES OF APCS VARY IN KEY FUNCTIONS REQUIRED FOR MAIT CELL ACTIVATION

As the *MR1* gene is ubiquitously expressed (50–52), many different cell types are able to present antigen to MAIT cells. The repertoire of innate cytokines and the extent to which MR1 is upregulated and brought to the cell surface upon activation and ligand availability vary not only with the type of stimulation but also with the type of APC (38, 53). Professional APCs (DCs, macrophages, and B cells) are efficient in microbe internalization and processing, as well as in delivering costimulatory signals to T cells. Kurioka et al. showed that the MR1-dependency of the MAIT cell response to pneumococci varied with the type of APC used (54). While the MAIT cell response in the presence of monocytes was MR1-independent, it was partially MR1-dependent in the presence of monocytederived macrophages (54). We previously observed that the addition of anti-CD28 to *E. coli*-fed monocytes cultured with Vα7.2<sup>+</sup> cells boosted MAIT cell IFNγ production (55), thus indicating that monocytes are not intrinsically efficient in delivering co-stimulatory signals and that the magnitude of the MAIT cell response varies with the degree of co-stimulation provided.

In conclusion, several aspects of the APC shape the magnitude and quality of MAIT cell antimicrobial responses. Such APCintrinsic factors include the surface expression of MR1–antigen complexes, the innate cytokines produced, and the panel of costimulatory and co-inhibitory receptors expressed upon microbial exposure (**Figure 1**). MAIT cells are likely to encounter different APCs *in vivo*, and their responses will ultimately be influenced by the type and representation of APCs.

## MICROBE GROWTH CONDITIONS INFLUENCE THE PRODUCTION OF MR1-PRESENTED LIGANDS

Mucosa-associated invariant T cells sense microbes through antigens presented by MR1 on the surface of APCs. The naturally occurring activating antigens identified thus far belong to the riboflavin biosynthesis pathway (8, 9), expressed in many different species of bacteria and fungi (56, 57). However, the ability of a microbe to activate MAIT cells depends not only on its capacity to produce MR1-presented agonists but also on whether MAIT cell non-stimulatory MR1-binding compounds are also produced and to what extent.

The type and concentration of MR1-presented compounds vary with the microbial growth conditions. 5-(2 oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU), the most potent MAIT cell agonist identified thus far, requires the riboflavin intermediate precursor 5-amino-6-d-ribitylaminouracil (5-A-RU) and either methylglyoxal or glyoxal for its generation (9), whereas natural MAIT cell non-stimulatory compounds derive from folic acid (8, 58, 59). The concentration of these precursor molecules at effector sites will likely dictate the amount of antigenic and non-antigenic MR1-presented compounds. It was recently shown that *Streptococcus pneumoniae* clinical isolates respond to exogenous availability of riboflavin by downregulating the *ribD* gene (which encodes the enzyme pyrimidine deaminase/ reductase essential for the production of 5-A-RU), with a consequent decrease in MAIT cell stimulatory potential (60). On the other hand, heat-stress was shown to induce the riboflavin operon in pneumococci, with upregulation of the riboflavin pathway genes within 2–4 h under such conditions (54).

It is also possible that other, yet unknown MAIT cell antigens with different requisites for their formation exist. Meermeier et al. (61) reported that MR1-restricted non-classical TRAV1-2- (Vα7.2- ) MAIT cells could be activated in an MR1-dependent manner by *Streptococcus pyogenes*, a riboflavin biosynthesis-incompetent microbe. This suggests that MR1 can present MAIT cell agonists other than riboflavin metabolites (61).

Other microbial products that are not presented by MR1 may also influence MAIT cell responses. For instance, lactate was shown to dampen NK and T cell activation in response to *Staphylococcus aureus* (62). Furthermore, short-chain fatty acids derived from bacterial fermentation, such as acetate, butyrate, and propionate, promote differentiation of T cells in a cytokine milieu-dependent manner (63, 64). Further investigation is required to determine if MAIT cells respond similarly to these microbial products.

In conclusion, the type of MR1-presented compounds and other bioactive microbial products will likely shape the functional characteristics of MAIT cell antimicrobial responses, as previously exemplified in *in vitro* competition experiments between MAIT cell activating and non-activating MR1-binding compounds (58, 59, 65).

#### MICROBIAL GENETIC BACKGROUND MAY PLAY A ROLE IN MAIT CELL RESPONSES

Recent studies showed that different *S. pneumoniae* isolates activated MAIT cells to different extents, as assessed by CD69 upregulation and IFNγ production (54, 60). Interestingly, Hartmann et al. found that *S. pneumoniae* isolates with similar MAIT cell activating properties grouped together with regard to their multilocus sequence type, suggesting a link between the MAIT cell response and microbial genetic background (60). The pneumococcus strain groups inducing higher levels of MAIT cell responses expressed significantly higher levels of the *ribD* gene and of MAIT cell ligands (60). Thus, differences in the genetic background between microbes influence their capacity to activate MAIT cells.

#### THE MICROBIAL PAMP SIGNATURE AND PROPENSITY FOR PHAGOCYTOSIS CAN AFFECT MAIT CELL RESPONSES

Different classes of microbes express distinct PAMPs, which can trigger toll-like receptors (TLRs) in APCs. Recently, Ussher et al. showed that the IFNγ production by MAIT cells upon *E. coli* stimulation can be positively or negatively affected by pretreatment of APCs with TLR agonists (38). Therefore, the PAMP–TLR

#### REFERENCES

interaction might be another factor shaping MAIT cell antimicrobial responses.

Given that phagocytosis of particles depends on their size and shape (66), geometrically different microbes may have different propensity to be phagocytosed. Moreover, certain microbes contain a polysaccharide capsule, and variations in this structure are known to influence the rate of phagocytosis (67, 68). Thus, the intracellular microbial load may vary quite extensively with the type of microbe. Interestingly, by using *S. enterica* serovar Typhi and *E. coli* as microbes and a B cell line as APCs, Salerno-Goncalves et al. found that the quality of the MAIT cell response depended on the microbial load (69).

In summary, MAIT cell antimicrobial responses can be influenced by several microbe-intrinsic factors, including their genetic background, physical characteristics, and PAMP repertoire, as well as their ability to produce MAIT cell antigens and other microbial products (**Figure 1**). These factors will not only influence MAIT cell functions but also dictate the amount of microbe that is required for optimal responses. In our study of MAIT cell responses to *E. coli* and *C. albicans*, we found that the optimal dose necessary for maximal MAIT cell activation, as assessed by CD69 upregulation and IFNγ production, was much higher for *E. coli* than for *C. albicans* (7).

#### CONCLUDING REMARKS

In conclusion, numerous factors influence the quality and magnitude of MAIT cell antimicrobial responses, including the MAIT cell TCR β-chain composition, the expression of NK cellassociated receptors, and the TCR–ligand–MR1 interaction. Predominance of MAIT cell subsets with distinct effector functions at sites of microbial invasion and their co-localization with functionally heterogeneous conventional CD4<sup>+</sup> and CD8<sup>+</sup> T cells that recognize distinct antigens (70–72) builds multifaceted immune barriers of immunosurveillance able to efficiently target pathogens with different requirements for eradication.

#### AUTHOR CONTRIBUTIONS

JD, CB, EL, and JS wrote the manuscript; MS provided the figure; MS, KL, JE, MB, TP, and J-BG contributed to manuscript editing and revision; JD and JS revised the final manuscript.

#### FUNDING

This work was supported by grants to JS from the Swedish Research Council (2016-03052), the Swedish Cancer Society (CAN 2017/777), and the US National Institutes of Health (R01DK108350). JD was supported by Fundação para a Ciência e a Tecnologia (FCT) Doctoral Fellowship SFRH/BD/85290/2012, cofunded by the POPH-QREN and the European Social Fund (FSE).

<sup>1.</sup> Treiner E, Duban L, Bahram S, Radosavljevic M, Wanner V, Tilloy F, et al. Selection of evolutionarily conserved mucosal-associated invariant T cells by MR1. *Nature* (2003) 422:164–9. doi:10.1038/nature01433

<sup>2.</sup> Le Bourhis L, Martin E, Peguillet I, Guihot A, Froux N, Core M, et al. Antimicrobial activity of mucosal-associated invariant T cells. *Nat Immunol* (2010) 11:701–8. doi:10.1038/ni.1890

<sup>3.</sup> Tilloy F, Treiner E, Park SH, Garcia C, Lemonnier F, De La Salle H, et al. An invariant T cell receptor alpha chain defines a novel TAP-independent major

histocompatibility complex class Ib-restricted alpha/beta T cell subpopulation in mammals. *J Exp Med* (1999) 189:1907–21. doi:10.1084/jem.189.12.1907


**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 Dias, Boulouis, Sobkowiak, Lal, Emgård, Buggert, Parrot, Gorin, Leeansyah and Sandberg. 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.*

# CD1d-Invariant Natural Killer T Cell-Based Cancer Immunotherapy: **α**-Galactosylceramide and Beyond

*Lisa A. King1 , Roeland Lameris1 , Tanja D. de Gruijl1 and Hans J. van der Vliet1,2\**

*1 Department of Medical Oncology, VU University Medical Center and Cancer Center Amsterdam, Amsterdam, Netherlands, 2 Lava Therapeutics, Den Bosch, Netherlands*

CD1d-restricted invariant natural killer T (iNKT) cells are considered an attractive target for cancer immunotherapy. Upon their activation by glycolipid antigen and/or cytokines, iNKT cells can induce direct lysis of tumor cells but can also induce an antitumor immune response *via* their rapid production of proinflammatory cytokines that trigger the cytotoxic machinery of other components of the innate and adaptive immune system. Here, we provide an overview of various therapeutic approaches that have been evaluated or that are currently being developed and/or explored. These include administration of α-GalCer or alternative (glyco) lipid antigens, glycolipid-loaded antigen-presenting cells and liposomes, strategies that enhance CD1d expression levels or are based on ligation of CD1d, adoptive transfer of iNKT cells or chimeric antigen receptor iNKT cells, and tumor targeting of iNKT cells.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*François Trottein, Centre national de la recherche scientifique (CNRS), France Toshinori Nakayama, Chiba University, Japan*

#### *\*Correspondence:*

*Hans J. van der Vliet jj.vandervliet@vumc.nl*

#### *Specialty section:*

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

*Received: 30 April 2018 Accepted: 19 June 2018 Published: 02 July 2018*

#### *Citation:*

*King LA, Lameris R, de Gruijl TD and van der Vliet HJ (2018) CD1d-Invariant Natural Killer T Cell-Based Cancer Immunotherapy: α-Galactosylceramide and Beyond. Front. Immunol. 9:1519. doi: 10.3389/fimmu.2018.01519*

Keywords: iNKT cells, CD1d, α-GalCer, glycolipids, cancer immunotherapy

# THE INVARIANT NATURAL KILLER T (iNKT) CELL AS TARGET FOR CANCER IMMUNOTHERAPEUTIC APPROACHES

Invariant natural killer T cells belong to a population of T lymphocytes, which harbor distinct characteristics of both natural killer (NK) and T cells (1). These cells express a semi-invariant T cell receptor (TCR), in humans consisting of a Vα24-Jα18 chain paired with Vβ11, and NK cell markers (e.g., CD161 and NKG2D) (2). Two CD1d-restricted NKT cell subtypes exist, the classical (type I) iNKT and non-classical (type II) NKT subsets expressing a diverse TCR repertoire (1). Both subsets are able to secrete immunoregulatory cytokines upon glycolipid recognition presented *via* the human leukocyte antigen class I-related molecule CD1d (2). iNKT cells release, upon their interaction with CD1d, a broad spectrum of cytokines, which in turn activate T cells, NK cells, B cells, and dendritic cells (DCs), thereby initiating a T helper (Th) 1, Th2 or Th17 response (1, 3–6). The role that the type I CD1d-restricted iNKT cell population can play in the antitumor immune response will be the main focus of this review.

CD1d-restricted iNKT cells can play a role in mediating antitumor immunity in various ways: indirectly *via* recognition of glycolipid-loaded CD1d molecules expressed by antigen-presenting cells (APCs), directly *via* recognition of glycolipid loaded CD1d expressed by tumor cells and alternatively *via* a TCR-independent manner through cytokines (6, 7). In case of recognition of glycolipid-loaded CD1d on APCs, the antitumor effect is mediated *via* secretion of inflammatory cytokines. Ligation of glycolipid-loaded CD1d molecules by iNKT cells amplifies IL-12 production and, like CD4<sup>+</sup> T helper cells, can induce maturation of DCs, conversely resulting in enhanced IFN-γ

production by the interacting iNKT cells (8, 9). Secretion of these inflammatory cytokines in turn promotes the cytolytic function of cytotoxic CD8<sup>+</sup> T cells and NK cells. In case of recognition of tumor cells expressing CD1d, iNKT cells can exert a direct antitumor effect *via* secretion of perforin and granzymes and death inducing receptors (e.g., Fas and TRAIL) reviewed by Bassiri et al. (9). Because of the cytotoxic capacity of iNKT cells and their ability to orchestrate pro- and anti-inflammatory immune responses, these cells are very attractive targets to exploit for cancer immunotherapy. Here, we will outline multiple strategies that can be used in order to promote iNKT cell based cancer immunotherapy.

# **α**-GALACTOSYLCERAMIDE

Several glycolipids have been demonstrated to act as activating agents for both murine and human iNKT cells, of which, α-galactosylceramide (α-GalCer) is the best known and most intensely studied. This glycosphingolipid was originally isolated from the marine sponge *Agelas mauritianus* and activates iNKT cells in a very potent way (10). Upon activation with α-GalCer, iNKT cells secrete Th1, Th2, and Th17 cytokines, modulating immune responses against tumors, microbial infections, viral infections, and auto-immune diseases (3, 5, 11, 12). α-GalCer-induced antitumor immune responses in several *in vivo* models using different tumor types (13). Subsequent clinical studies with α-GalCer did not show any adverse events but also did not result in clinically relevant antitumor effects in advanced cancer patients (14). The effect of α-GalCer may be limited by the relatively short-lived and in part antagonizing nature of the mix of Th1 and Th2 cytokines that is produced by activated iNKT cells, followed by long-term anergy of iNKT cells (10, 15, 16).

Changing the route of administration of α-GalCer may enhance efficacy. Direct intravenous administration of α-GalCer in patients with solid tumors led to an increase in serum cytokine levels but also to the disappearance of iNKT cells from the circulation within 24 h (14). Furthermore, upon repeated systemic administration of α-GalCer, increases in serum cytokine levels were no longer observed, which was in line with the induction of iNKT cell anergy observed in mouse studies. The anergy of iNKT cells in these cases may have been related to the fact that also non-professional APCs presented α-GalCer to iNKT cells (16). An attractive alternative route of administration to overcome these problems might be the skin. Here, α-GalCer would be taken up predominantly by skin-residing DCs or DCs in skin-draining lymph nodes. A study performed by Bontkes et al. compared the effect of intradermal versus intravenous injections and indeed showed prevention of iNKT cell anergy by intradermal injection of α-GalCer (17). Furthermore, intradermal α-GalCer triggered an earlier iNKT cell response and an increase in systemic iNKT cell numbers, leading to enhanced protective immunity in response to intradermal vaccination with protection against tumor outgrowth in five out of six mice. To add to this, Tripp et al. showed presentation of α-GalCer directly to iNKT cells in the draining lymph nodes in an *in vivo* mouse model, thereby bypassing migratory DCs and possibly iNKT cell anergy (18). Intranasal injection of α-GalCer was also shown to effectively reduce iNKT cell anergy in an *in vivo* mouse model as repeated dosing of α-GalCer *via* this route boosted iNKT cells and DCs without inducing anergy, as opposed to the intravenous route (19).

# ANALOGS OF **α**-GalCer

As a result of the limited effects of α-GalCer in clinical studies, subsequent research focused on the development of glycolipid analogs with more distinct iNKT cell activating properties. Whereas some of these glycolipids predominantly induce Th2 type cytokine production in iNKT cells and were suggested to be mainly of potential use in auto-immune diseases (e.g., OCH and α-GalCer20:2), other glycolipid activators (e.g., those encompassing an aromatic ring in either the acyl- or shingosine tail) induced a predominant Th1 type immune response (20). Such Th1-biased glycolipids are more effective in triggering TCR activation and iNKT cell expansion compared to α-GalCer (21). These Th1 biased analogs include, e.g., α-C-GalCer and 7DW8-5. α-C-GalCer is a C-glycoside analog of α-GalCer and harbors a methyl group instead of a glycosidic oxygen. In a mouse melanoma metastasis model, α-C-GalCer was found to increase IL-12 and IFN-γ production and to decrease IL-4 production in comparison with α-GalCer and in addition exerted a more potent prophylactic effect against lung metastasis (22). Also in combination with monoclonal antibodies targeting tumor necrosis factor-related apoptosis-inducing ligand receptor (DR5) and 4-1BB, α-C-GalCer outperformed α-GalCer in experimental (established) mouse breast and renal tumors (23). Furthermore, while high concentrations of α-GalCer led to toxicity, this was not observed with α-C-GalCer.

The synthetic α-GalCer analog 7DW8-5 has a shorter fatty acid tail with a fluorinated benzene ring at the end and binds stronger to the CD1d molecule than α-GalCer (24). In vaccination studies, 7DW8-5 induced 100-fold stronger IFN-γ production by iNKT cells as compared with α-GalCer (24). When used as adjuvant for vaccination with tumor-associated antigens (TAAs) in a B cell lymphoma mouse model, IL-12 and IFN-γ production and an enhanced magnitude of the CD8<sup>+</sup> T cell response were observed leading to an enhanced antitumor response (25).

# GLYCOLIPID-LOADED APCs

Professional APCs are well equipped to provide optimal stimulatory signals to T cells that recognize their cognate antigen and are thereby capable of mediating antigen-specific immune responses against various targets. This potential of APCs could be used as a means to further strengthen the antitumor effect of glycolipids. Indeed, APCs that were loaded with α-GalCer *ex vivo* enhanced antitumor immune responses compared to α-GalCer alone in a B16 melanoma mouse model (26). This observation triggered multiple clinical phase I studies using mature or immature DCs pulsed with α-GalCer. Nieda et al. started the first phase I clinical trial where they administered α-GalCer pulsed immature moDCs to 12 patients with metastatic malignancies (27). They found increased serum IL-12 and IFN-γ levels and activated T and NK cells, indicating that NKT cells indeed bridged innate and adaptive immunity. This group performed another phase I clinical trial involving 12 patients with metastatic solid tumors (28). Effective iNKT cell activation was observed using immature moDCs pulsed with α-GalCer. Therapy was well tolerated and the majority of the patients experienced disease stabilization. Of note, intravenously administered α-GalCer pulsed DCs induced greater immunological effects compared to intradermally administered α-GalCer pulsed DCs. Several phase I and II clinical trials have been performed focusing on lung cancer using either α-GalCer pulsed DCs, peripheral blood mononuclear cells (PBMCs), or APCs (29–31). No adverse effects were observed and treatment was found to be safe and well tolerated. Chang et al. performed a phase I study in advanced cancer patients using intravenous administration of mature α-GalCer pulsed moDCs (32). Activation and a persistent expansion of the iNKT cell pool in combination with signs of secondary activation of other immune cell populations (including B cells, NK cells, and T cells) were observed as well as an increase in serum levels of IL-12 and IFN-γ. Interestingly, in a phase I clinical trial with asymptomatic myeloma patients, combination of low-dose lenalidomide with α-GalCer pulsed mature moDCs led to increased activation of innate immune cell subsets including iNKT, NK cells, monocytes, and eosinophils and a reduction in tumor-associated monoclonal immunoglobulin in three of four patients with measurable disease (33). Gasser et al. administered autologous moDCs loaded with α-GalCer, synthetic long peptides spanning immunogenic regions of the cancer-testis antigen NY-ESO-1, and short MHC-Ibinding peptide sequences from the influenza virus intravenously in eighth high-risk stage II–IV melanoma patients (34). In three of these patients, a significant increase of peripheral iNKT cells was observed and four patients showed increased frequencies of IFN-γ positive cells when PBMCs were re-stimulated with α-GalCer. Five melanoma patients showed increases in cytokines related to α-GalCer stimulation found in the serum, and an increase in circulating NY-ESO-1-specific T cells was detected in seven patients.

To further enhance the effect of α-GalCer loaded APCs, combinations with chemotherapeutic agents known to induce immunogenic cell death were investigated. These chemotherapeutics can promote immune responses against the tumor by inducing activation of multiple cell death pathways and by enhancing the subsequent uptake of tumor peptides by APCs in the context of damage-associated molecular patterns (35). For example, gemcitabine and mafosfamide were tested in combination with α-GalCer-loaded bone marrow-derived DCs in a murine metastatic breast cancer model (36). Chemotherapy alone resulted in an increase in tumor cell CD1d expression, facilitating recognition by iNKT cells. Furthermore, α-GalCer-loaded DCs in combination with gemcitabine or mafosfamide led to increased IFN-γ production and a significant increase in survival.

# INCORPORATION OF GLYCOLIPIDS IN NANOVECTORS

Another approach that is being explored with the aim to enhance the effect of α-GalCer entails its incorporation into nanovectors, which can act as vaccine carriers to induce an immune response by delivering their content to endosomes. Presentation of α-GalCer by CD1d-expressing APCs was indeed improved using liposomes and resulted in increased expansion and IFN-γ production by iNKT cells and a potent anti-metastatic effect in a highly malignant metastatic lung murine cancer model (37). Khan et al. used liposomes incorporating glycosphingolipids isolated from *Spingomonas paucimobilis*, which can, like α-GalCer, specifically activate iNKT cells (38). When these liposomes were loaded *ex vivo* onto bone marrow-derived DCs and used as treatment for mice with dimethyl-α-benzanthracene-induced tumors, a more sustained secretion of IFN-γ and a potent antitumor response was induced compared to administration of glycosphingolipids alone. A different approach consists of iNKT cell activation *via* targeted delivery of α-GalCer and OVA or tumor self-antigens (PLGA)-based nanoparticles that target the endocytic pathway of the cross-presenting CD8α+ DC subset *via* DEC205 (39) or Clec9a (40). Delivery of α-GalCer to CD8α+ DCs *via* this route enhanced iNKT cell transactivation of NK and T cells and a cytotoxic T cell response in *in vivo* mouse models and could promote both prophylactic and therapeutic antitumor responses in an advanced solid tumor model in mice. Notably, this approach could also target human CLEC9A-expressing DC to mediate the expansion of tumor self-antigen specific CD8<sup>+</sup> T cells in PBMCs samples of melanoma patients *in vitro*, thereby underscoring the translational potential of this approach*.*

# CD1d-INDUCING AGENTS

As iNKT cells can directly kill CD1d-expressing tumor cells, one can hypothesize that the efficacy of iNKT cell-based antitumor responses can also be improved by increasing CD1d-expression levels on tumor cells as this may facilitate their recognition by iNKT cells. It has been reported that inhibitors of histone deacetylases that regulate expression, cell cycle progression, and cellular proliferation are able to induce CD1d expression levels (41). Next to this, all-trans retinoic acid and certain chemotherapeutics have also been reported to increase CD1d expression levels and are, therefore, of potential interest either alone or in combination with other iNKT cell-based therapeutic approaches (36, 42).

# ADOPTIVE TRANSFER OF iNKT CELLS

It is known that relatively low numbers of iNKT cells are present in peripheral blood of healthy individuals. This number is further reduced in many, but not all, cancer types (43–46). A higher number of circulating iNKT cells predicted improved outcome in head and neck squamous cell carcinoma (HNSCC) patients treated with curative-intent radiotherapy (47). Several studies were designed to increase the size of the iNKT cell population. Mouse *in vivo* studies support this strategy as adoptive transfer of murine iNKT cells, activated with IL-12 *ex vivo*, showed a potent antitumor response in a B16 melanoma and a lung metastasis model (48). Adoptive transfer of iNKT cells has the added advantage of reversing the defective iNKT cell IFN-γ production commonly observed in cancer patients, which is known to be important for promoting antitumor immune responses (45, 49, 50).

Several clinical trials have been performed using adoptive transfer of iNKT cells. In a phase I study of patients with advanced melanoma, Vα24 iNKT cells were isolated from patients PBMCs and *ex vivo* expanded for several weeks (45). After adoptive transfer, an increased number of iNKT cells and an increased activation state of iNKT cells and other immune cell subsets was observed without signs of toxicity. A slightly different approach was used in clinical trials in patients with HNSCC. Here, iNKT cells were isolated from PBMC and expanded *ex vivo* with α-GalCer and IL-2, while APC fractions were generated from PBMC by culturing them in the presence of GM-CSF and IL-2 (51). Expanded iNKT cells were then intra-arterially infused in the tumor-feeding artery while α-GalCer pulsed APCs were injected in the nasal submucosa. Therapy was found to be safe and resulted in an objective response rate of 50%. Increased intratumoral accumulation of transferred iNKT cells was associated with improved clinical outcome. Additional clinical trials were designed combining administration of expanded iNKT cells with α-GalCer pulsed DCs in patients with recurrent HNSCC reviewed by Motohashi et al. (52). Again, combination therapy appeared to exert beneficial clinical effects with disease stabilization and tumor regression associated with increased intratumoral iNKT cell numbers (53, 54). Based on these positive results, additional trials involving the adoptive transfer of iNKT cells in various tumor types were initiated, the results of which are eagerly awaited (NCT03093688; NCT02619058; NCT01801852).

# TUMOR TARGETING OF iNKT CELLS

All strategies described above are based on either the intravenous or intra-arterial administration of iNKT cells or the systemic or intradermal/intranasal activation of iNKT cells. Although these approaches can trigger antitumor immune responses, antitumor activity may be more pronounced and consistent when one can specifically target and activate iNKT cells in the tumor microenvironment. The potential of this approach has been demonstrated using a bispecific molecule generated by genetic fusion of a single chain variable fragment (scFv) targeted to a specific tumor peptide and CD1d, which can be loaded with specific glycolipids to allow iNKT cell activation. The antitumor activity of such a bispecific approach outperformed the activity of α-GalCer as was demonstrated *in vivo* in mice inoculated with Her2 or CEA expressing tumors using Her2- and CEA-targeted constructs, respectively (55–57). iNKT, NK, and T cells were found to accumulate at the tumor site using these targeted approaches and, in addition, treatment was not accompanied by iNKT cell anergy as iNKT cells remained responsive to repeated injections of the CD1d fusion proteins loaded with α-GalCer (55, 56).

# CHIMERIC ANTIGEN RECEPTOR (CAR) iNKT CELLS

Another strategy combining tumor targeting of iNKT cells with an increase in the size of the iNKT cell population consists of adoptive transfer of CAR-expressing iNKT cells. CAR therapies were first applied to conventional T cells resulting in the approval by the Food and Drug Administration of two CAR-T cell therapies for hematological malignancies: one for acute lymphoblastic leukemia and one for advanced lymphoma. The currently used CARs consist of a scFv for antigen binding, the TCR ζ chain for TCR activation, and one or two signaling domains from the co-stimulatory molecules CD28 and/or 4-1BB (58). After the introduction of the CAR, there is still a large diversity in TCR specificity and function among conventional CAR-T cells, whereas CAR iNKT cells (due to the invariant nature of their TCR) constitute a more homogenous population with respect to both their function and specificity to CD1d, and this may translate into a different and perhaps more predictable and manageable toxicity profile (58).

Invariant natural killer T cells were reported by Heczey et al. to be a safe and effective platform for CAR redirected cancer immunotherapy in neuroblastoma (58). This approach showed effective *in vitro* cytotoxicity of Vα24 human iNKT cells with a CAR targeting the ganglioside GD2 antigen expressed by neuroblastoma cells. Also, iNKT cells retained their ability to kill tumorassociated macrophages as a result of TCR-mediated recognition

of CD1d. Using this therapeutic approach in a neuroblastoma mouse model, transfer of GD2-specific CAR iNKT was shown to induce antitumor activity resulting in prolonged survival of mice. Of importance, GD2-specific CAR iNKT cells did not lead to graft versus host disease even after repeated infusions.

In a B cell lymphoma model, Tian et al. demonstrated that CD19-specific CAR-iNKT cells expressing CD62L, a ligand involved in homing of naïve and central memory T cells to secondary lymphoid organs, was the predominant CAR iNKT population that mediated tumor regression (59). This potentially allows for future selection of a more effective CAR-iNKT approach. A phase I clinical trial is currently ongoing wherein children with neuroblastoma are treated with GD2 CAR and IL-15 expressing autologous iNKT cells (NCT03294954).

## CD1d-SPECIFIC ANTIBODIES

Instead of targeting iNKT cells that interact with CD1d, effects of monoclonal antibodies specific for CD1d have also been explored. Yue et al. showed that direct ligation of CD1d by monoclonal antibodies on the DC could, at least in part, mimic iNKT cell help to DCs (60). Ligation of several monoclonal CD1d antibodies led to downstream signaling *via* NF-κB, resulting in IL-12 production and moDC maturation. Recently, CD1d-specific single domain antibodies (sdAb) have been identified with a similar ability to induce

# REFERENCES


DC maturation and IL-12 production *via* CD1d ligation (61). sdAb have several advantages over conventional monoclonal antibodies, including extended stability, low immunogenicity, ease of production and, due to their small size (15 kDa), deep and homogenous tumor penetration (62). As sdAbs can also be easily cloned to other molecules, e.g., TAAs, this can provide a vaccine encompassing a stimulatory signal to DCs, which will promote the efficient initiation and development of a tumor-associated antigen specific cytotoxic T cell response.

# CONCLUDING REMARKS

Within the last two decades, the role of iNKT cells within the antitumor immune response has been intensely studied. It is now recognized that iNKT cells play an important role in orchestrating immune responses, making strategies exploiting these cells potentially valuable for cancer immunotherapy. Several approaches for therapeutic manipulation of iNKT cells are being explored (illustrated in **Figure 1**), which may ultimately translate into a more effective therapy for cancer.

# AUTHOR CONTRIBUTIONS

LK wrote the manuscript. HV co-wrote and reviewed the manuscript. RL and TG reviewed the manuscript.


**Conflict of Interest Statement:** LK and Rl are funded by Lava Therapeutics. HV acts as chief scientific officer of Lava Therapeutics.

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

*Copyright © 2018 King, Lameris, de Gruijl and van der Vliet. 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.*

*Céline Mortier 1,2†, Srinath Govindarajan1,2†, Koen Venken1,2‡ and Dirk Elewaut 1,2\*‡*

*1Department of Rheumatology, Ghent University Hospital, Ghent, Belgium, 2Unit for Molecular Immunology and Inflammation, VIB Center for Inflammation Research, Ghent University, Ghent, Belgium*

Innate-like T cells such as invariant natural killer T (iNKT) cells and mucosal-associated T (MAIT) cells, characterized by a semi-invariant T cell receptor and restriction toward MHC-like molecules (CD1 and MR1 respectively), are a unique unconventional immune subset acting at the interface of innate and adaptive immunity. Highly represented at barrier sites and capable of rapidly producing substantial amounts of cytokines, they serve a pivotal role as first-line responders against microbial infections. In contrast, it was demonstrated that innate-like T cells can be skewed toward a predominant pro-inflammatory state and are consequently involved in a number of autoimmune and inflammatory diseases like inflammatory bowel diseases and rheumatic disorders, such as spondyloarthritis (SpA) and rheumatoid arthritis. Interestingly, there is link between gut and joint disease as they often co-incide and share certain aspects of the pathogenesis such as established genetic risk factors, a critical role for pro-inflammatory cytokines, such as TNF-α, IL-23, and IL-17 and therapeutic susceptibility. In this regard dysregulated IL-23/IL-17 responses appear to be crucial in both debilitating pathologies and innate-like T cells likely act as key player. In this review, we will explore the remarkable features of iNKT cells and MAIT cells, and discuss their contribution to immunity and combined gut–joint disease.

Keywords: innate-like T cells, invariant natural killer T cells, mucosal-associated invariant T cells, CD1, MR1, rheumatic diseases, inflammatory bowel disease, gut–joint axis

# INTRODUCTION

Over the past decades, innate-like T cells have gained increasing attention given their unique biology and potential involvement in multiple immune and inflammatory diseases. Those cells, with overlapping features of both the innate and adaptive immune system, are characterized by an antigen-specific semi-invariant T cell receptor (TCR) with restricted V(D)J rearrangement. Innatelike T cells are able to rapidly produce cytokines, which makes them an ideal first-line defense against microbial infections (1). However, it has become clear that these cells show functional plasticity and can be skewed toward a more pro-inflammatory state (2). Two members of this unconventional T cell population are invariant natural killer T (iNKT) cells and mucosal-associated invariant T (MAIT) cells. Both cell types have the unique feature of recognizing atypical non-peptide antigens presented by highly conserved MHC-related molecules, respectively CD1 and MR1. iNKT cells respond to glycolipid molecules, whereas MAIT cells can be activated by vitamin B2 (riboflavin) metabolites, which are intermediates from bacterial and yeast biosynthetic pathways (3). Gamma

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Luis Graca, Universidade de Lisboa, Portugal Saschiko Miyake, Juntendo University, Japan*

> *\*Correspondence: Dirk Elewaut dirk.elewaut@ugent.be*

*† These authors have contributed equally to this work. ‡ Shared supervision.*

#### *Specialty section:*

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

*Received: 30 April 2018 Accepted: 15 June 2018 Published: 29 June 2018*

#### *Citation:*

*Mortier C, Govindarajan S, Venken K and Elewaut D (2018) It Takes "Guts" to Cause Joint Inflammation: Role of Innate-Like T Cells. Front. Immunol. 9:1489. doi: 10.3389/fimmu.2018.01489*

delta (γδ) T cells are a third innate-like T cell population (4), but the focus here will be on CD1- and MR1-restricted T cells. In this review, we want to highlight the intriguing nature of these cells and discuss what is known about their role in rheumatic diseases like spondyloarthritis (SpA) and rheumatoid arthritis (RA), next to inflammatory bowel diseases (IBD).

Spondyloarthritides are a group of chronic inflammatory disorders that primarily affect the musculoskeletal system and are multifactorial in origin. Some subtypes affect mainly the axial joints (spine and sacroiliac joints), with ankylosing spondylitis (AS) as prototypical disease, while others have a more peripheral manifestation (arthritis of the limbs and enthesitis) (5). RA, another chronic rheumatic disorder, has an autoimmune basis and is characterized by the presence of autoantibodies direct against i.a. citrullinated antigens (6). Most commonly involved are the small joints of hands and feet, often with a symmetrical distribution, whereas in SpA joint inflammation is generally non-symmetrical. IBD is the collective term for a group of inflammatory diseases of the digestive tract, leading to gastrointestinal complaints. The two best known subtypes are Crohn's disease (CD) and ulcerative colitis (UC) (7).

Remarkably, SpA often is accompanied by extra-articular manifestations, such as acute anterior uveitis, psoriasis, or IBD. Histological evaluation showed that about 50% of SpA patients without gastrointestinal symptoms have microscopic intestinal inflammation (8), of which a fraction (5–10%) develops CD over time (8, 9). Furthermore the presence of subclinical gut inflammation is associated with shifts in the composition of the gut microbiome (10–12). A state of intestinal dysbiosis has also observed in RA and IBD patients (6, 11, 13–15). Additionally in RA, there is a significant correlation with periodontitis and the presence of *Porphyromonas gingivalis* in the oral cavity. This bacterium could play an important role in the pathogenesis of RA through citrullination of proteins using a specific enzyme (peptidyl arginine deiminase), potentially leading to the production of anti-cyclic citrullinated peptide (anti-CCP) autoantibodies relevant to RA disease (16, 17). This finding underscores that microorganisms can have a direct pathological role in disease pathogenesis. On the other hand, alterations in microbial composition can also play an indirect role by modulation of specific immune cell functions relevant for these diseases. Hence, the ability of recognizing bacterial antigens (or derived products) combined with their clear presence at barrier sites, makes innate-like T cells an appealing target to study in the context of the gut–joint axis in rheumatic diseases.

A crucial role for pro-inflammatory cytokines in the pathogenesis of SpA, RA, and IBD, is confirmed by current knowledge from genome-wide association studies (GWAS) and anti-cytokine trials. Interestingly, SpA, RA, and IBD share clinical responsiveness to anti-tumor necrosis factor (TNF)-α therapy but significantly differ in their response toward inhibition of other key inflammatory cytokines like IL-17. Over the years, the interleukin (IL)-23/IL-17 immune axis has manifested as a major player in the pathogenesis of SpA (18). GWAS studies have revealed polymorphisms in the *IL23R* gene associated with both SpA and IBD (19). Furthermore, there is extensive evidence from *in vivo* models, translational studies, and clinical trials (2, 20–22). Curiously, anti-IL-17 treatment was not effective in patients with RA or IBD with some reports even suggesting a worsening of IBD, which might be linked to an effect on barrier integrity (23–25). IL-23 is essential for the terminal differentiation and inflammatory functions of T helper-17 (Th17) cells. Interestingly, it has been shown that also innate-like T cells express the key Th17 transcription factor retinoic acid receptor-related orphan receptor-γt (RORγt) and that they can respond toward IL-23 by producing IL-17 and related cytokines like IL-22 (22). The importance of this finding was underscored by a mouse study, in which IL-23 overexpression (an SpA-like model using minicircle DNA technology) could induce enthesitis independent of conventional Th17 cells (26). As disease induction did require the presence of CD4<sup>−</sup>CD8<sup>−</sup> T cells, there could be a role for IL-23 responsive innate-like T cells (27).

#### iNKT CELLS

#### Biology and Localization

Invariant natural killer T cells are CD1d-restricted T cells which express a semi-invariant TCR consisting of an invariant α chain [in particular, the variable (V) and joining (J) segments Vα14–Jα18 in mice and Vα24–Jα18 in humans], combined with a restricted β chain repertoire, usually Vβ2, Vβ7, or Vβ8.2 in mice and Vβ11 in humans (28, 29). Identification of these cells in mice can be performed by the use of CD1d tetramers and in humans by using CD1d tetramers, a specific Vα24Jα18 Ab (clone 6B11) or the combination of anti-Vα24 and anti-Vβ11 antibodies. In contrast to conventional T cells which detect self or foreign peptide antigen–MHC complexes, iNKT cells recognize only glycolipid antigens bound to CD1d, a MHC class I-like glycoprotein (30). Currently, identified antigens are predominantly of nonmammalian nature, with α-galactosylceramide (α-GalCer) as the most potent and best studied example. However, also microbial derived (31) and endogenous ligands have been described (28, 32, 33). Of note, the human genome encodes five CD1 genes (CD1a, b, c, d, and e) whereas only CD1d is expressed in mice, and human CD1a, b, and c restricted T cells have been described too (34).

A hallmark of iNKT cell biology is the ability to secrete large amounts of cytokines and chemokines upon TCR recognition of lipid antigen–CD1 complexes or *via* indirect (TCR independent, mainly cytokine driven) stimulation, hereby acting as a "bridge" between innate and adaptive immune responses (35, 36). In analogy to classification of conventional T cells based on their cytokine production, iNKT cells can be subdivided in NKT1, NKT2, and NKT17 cells (37). Each of these subsets expresses distinct transcription factors which correlate with their capacity to secrete specific cytokines. NKT1 cells are T box transcription factor TBX21 (T-bet) positive and primarily secrete interferon (IFN)-γ, NKT2 cells express high levels of GATA-binding protein 3 (GATA3) and promyelocytic leukemia zinc finger protein (PLZF), and produce IL-4 and IL-13, and NKT17 cells express RORγt next to intermediate levels of PLZF and produce IL-17 as signature cytokine (38–40). All these subsets acquire their functional capacity during the development in the thymus and are distributed to the peripheral organs in a tissue-specific manner (41). However, there are also reports suggesting that peripheral iNKT cells are able to further functionally differentiate under inflammatory conditions (42, 43). In addition, it is also clear that iNKT cells experience further maturation at mucosal surfaces (e.g., lung and gut) as evidenced from experiments with germ-free mice (44, 45).

Finally, next to above-mentioned subsets, also other particular iNKT cells, such as NKTreg (FOXP3<sup>+</sup>) (46), NKTFH (CXCR5<sup>+</sup> PD-1hi) (47), NKT10 cells (48), and adipose tissue residing iNKT (PLZF-E4BP4<sup>+</sup>) cells (49) have been described and warrant further investigation. The frequency of iNKT cells in mice is substantially higher compared to humans. The majority of murine iNKT cells are found in the liver (20–40%), whereas iNKT cells constitute only 1% of cells in the human liver. Moreover, the iNKT cell frequencies in human peripheral blood samples shows significant inter-donor variation (approximately 0.01–0.5% of T cells) which makes the study of human iNKT cell biology more challenging.

#### Contribution to Gut and Joint Disease

Considering the ability of iNKT cells to produce copious amounts of immunomodulatory cytokines, several studies have assessed the capacity of iNKT cells to modulate autoimmune diseases (50–54). Some have shown that activation of iNKT cells can protect from joint inflammation, while others mentioned exacerbation of disease (2, 35, 52–58). In TNF<sup>Δ</sup>ARE/<sup>+</sup> mice, a TNF-driven SpA-like animal model for combined gut and joint inflammation, iNKT cells can dampen arthritis and ileitis by producing immunomodulatory cytokines after activation by TNF-driven CD1dhigh dendritic cells (DCs). Interestingly, the frequency of the latter cell population is increased in synovial fluid from SpA patients (52). This example, next to evidence from an iNKT cell-dependent infectious disease *in vivo* model, suggests that inflammatory DCs can pick up antigens from the microbiota or microbialderived products at the intestinal draining sites and subsequently activate iNKT cells. Furthermore, the crosstalk between DCs and iNKT cells was found to be TNF-mediated (52, 59, 60). Collageninduced arthritis (CIA) and collagen antibody-induced arthritis (CAIA), two mouse models for RA, have revealed contradictive results. While several reports suggested a pathogenic role (55, 56, 61, 62), iNKT cells protected from disease in a number of studies (54, 63, 64). Conflicting outcomes could originate from differences in the stimulating ligand and the time point of iNKT cell activation, since these appeared to be crucial factors (54). Regarding human joint disease, it has been described that RA patients have lower frequencies of both CD4− and CD4+ iNKT cells in peripheral blood compared to healthy controls, and they were skewed toward a Th1 phenotype (65–67) and a more restricted iNKT-TCR repertoire (68). There is no clear information regarding iNKT cell function in SpA disease so it will be of interest to study these, but also other innate-like T cells, in the context of joint–gut pathology in SpA patients.

Similar to joint disease, dichotomous effects of iNKT cells were observed for IBD (69). In dextran sodium sulfate-induced colitis, a model for human UC, activation of iNKT cells by α-GalCer ameliorated disease (70, 71). Adoptive transfer in iNKT deficient mice also had a protective role (70, 72, 73). In contrast, iNKT cells exacerbated inflammation in oxazolone-induced colitis, another UC model, as shown from results in CD1d and iNKT-deficient mice (74). Again it is clear that iNKT cells are involved in the pathogenesis, possibly even serving a dual role depending on the type of IBD (UC versus CD like) and the exact conditions of activation and further research is warranted to elucidate the mechanisms, ideally by using CD1d tetramer stainings. A large cohort of IBD patients showed that iNKT cells were decreased in the blood in both CD and UC compared to healthy individuals (75). The intestinal lamina propria of UC patients was found to have a strong abundance of type 2 iNKT cells that produced high amounts of the cytotoxic cytokine IL-13 (76, 77). Further studies are required to understand whether these disturbances in cell numbers in patients are a result of disease or whether iNKT cells are involved in development or persistence of inflammatory gut and joint disorders.

# MAIT CELLS

#### Biology and Localization

Mucosal-associated invariant T cells are an evolutionarily highly conserved cell population with two defining traits: the expression of a semi-invariant TCR and restriction of recognizing antigens presented by the MHC class I-related molecule MR1. Similar to iNKT cells, their TCR consists of an invariant TCR α chain paired with a limited array of Vβ chains (Vα7.2Jα33 paired with Vβ2 or Vβ13 in humans and Vα19Jα33 paired with Vβ6 and Vβ8 chains in mice) (3). Also, MAIT cells can be stimulated by both TCRactivation and TCR-independent signals, such as IL-18 (36, 78). In both humans and mice, the majority of MAIT cells in peripheral blood and tissues are CD4<sup>−</sup>CD8<sup>−</sup> or CD8<sup>+</sup> (in particular, more CD8αα than CD8αβ), besides very few CD4-expressing cells (79). The development occurs in the thymus, followed by an extrathymically maturation, a process that is regulated by multiple factors, including MR1, commensal gut microbiota, and the transcription factor PLZF (80), as illustrated by their absence in MR1-deficient and germ-free mice (81) and their severely reduced frequency in PLZF-deficient mice (82).

Identifying MAIT cells in human blood and tissue can be based on expression of TCR Vα7.2 (TRAV1-2) combined with the NK cell receptor CD161 and/or IL-18Rα (CD218). However, some of these surrogate markers are not present throughout the whole ontogeny, which has challenged accurately defining the cells. Recently, the production of MR1 tetramers meant a major revolution in this field, enabling the specific detection of MAIT cells in both humans and mice (82, 83). This has led to increased understanding of the development, which in mice occurs in three stages with only stage 3 being functionally competent. This model is largely in parallel with the development in humans (80). Furthermore, MR1 tetramers have allowed to describe different subsets within the MAIT cell population (84).

Mucosal-associated invariant T cells are predominantly found at mucosal and epithelial barrier sites. They are most abundant in the gastrointestinal tract and associated organs, such as mesenteric lymph nodes and the liver (in the latter organ representing 20–45% of all human T cells), but can also be found in the blood (1–8% of all human T cells). However, a lot of variation exists in the frequency of MAIT cells among humans, with age as an important determining factor (84). In mice, MAIT cells have a much lower frequency (at least 10-fold less than in humans) but are also mainly found at mucosal surfaces (82, 85). Because of their localization in close contact with the microbiota, it is believed that MAIT cells serve an essential role in modulating host-microbial interplay (81). They recognize and can be activated by vitamin B2 (riboflavin) metabolites, such as ribityllumazines [for example, 5-OP-RU or 5-(2-oxopropylideamino)-6-D-ribitylaminouracil] and pyrimidines (80). As many vitamin biosynthetic pathways are restricted to bacteria and yeasts, it is believed that MAIT cells detect these antigens to respond toward microbial challenges.

The majority of MAIT cells (>80%) in peripheral blood of healthy humans was found to produce the Th1-related cytokines interferon-gamma (IFN-γ) and TNF (80). Only a small fraction could produce IL-17A, consistent with a low expression of RORγt in healthy subjects. However, it seems that peripheral expansion and maturation is particularly important in human MAIT cells, illustrated by a dominant IL-17A-producing MAIT cell population in the liver (86).

#### Contribution to Gut and Joint Disease

In contrast with their role as first-line responders against microbial infections (87), MAIT cells are also thought to be involved in a number of inflammatory and autoimmune disorders. In many of these diseases, a reduced systemic MAIT cell frequency compared with healthy individuals was observed, together with an increased abundance at sites where inflammation occurred (88, 89). For instance, IBD patients were found to have decreased peripheral blood MAIT cells with an enrichment in inflamed intestinal tissue (90, 91). In both RA (92) and AS (93, 94), there was a systemic decrease in MAIT cells accompanied by elevated cell numbers in the synovial fluid. It should be noted that in some of these diseases, like IBD and RA but not AS, results could be confounded by the use of corticosteroids as this has been associated with lower systemic MAIT cell frequencies (88). Furthermore, the identification in these studies was based on the expression of surrogate markers (TCRVα7.2<sup>+</sup>CD161hi) and not MR1-tetramer stainings. Upon activation, CD161 can be downregulated on MAIT cells, which could also have influenced these results (88).

Next to changes in frequencies, there were also phenotypical alterations in these diseases. In IBD, MAIT cells expressed higher levels of activation markers such as CD69 and they produced more IL-17 (90, 91). UC patients showed increased IL-18 serum levels and interestingly, a correlation was found with CD69 expression, suggesting that induced IL-18 secretion could have a role in activation of MAIT cells in these patients (91). The activation status of MAIT cells was positively correlated with disease activity of AS patients (94). An upregulation of IL-17 in these cells could also be observed in peripheral blood of AS patients compared to

regulatory capacities (e.g., iNKT cells) or through skewing to pro-inflammatory profiles (e.g., IL-17).

healthy controls, together with a lower IFNγ production (93, 94). Curiously, the higher proportion of IL-17<sup>+</sup> MAIT cells in AS was only seen in male patients, while no differences in other clinical parameters existed. Another important finding was that MAIT cells in synovial fluid from AS patients show even higher IL-17 levels than in peripheral blood (93). These results support the idea that MAIT cells can contribute to inflammatory diseases in both the gut and joint. Interestingly, an elevated IL-17 production by MAIT cells was not found in RA neither in peripheral blood nor in synovial fluid (93), suggesting a differential mechanism in RA and AS disease. An important role could be attributed to IL-7 as gut and joint tissues of AS patients contained higher IL-7 levels than healthy controls, next to a higher IL-7R expression in blood-derived MAIT cells from AS patients. Furthermore, IL-7 priming induced IL-17 production by MAIT cells and, even more interesting, this response was substantially higher in AS patients (93). Anti-TNF therapy did not affect the MAIT cell number nor did it decrease production of IL-17 or IFNγ by MAIT cells, further underscoring the IL-23/IL-17 axis in innate-like T cells as a potential therapeutic target (94).

An effector role for MAIT cells in arthritis was demonstrated in MR1-deficient mice, after both CIA and CAIA disease induction. MR1 deficiency significantly reduced arthritis and adoptive transfer of MAIT cells to MR1<sup>−</sup>/<sup>−</sup> mice exacerbated arthritis in CAIA (95). The situation in IBD is less clear, with one report showing that adoptive transfer of MAIT cells in mice with TNBSinduced colitis resulted in milder disease (96). However, caution should be taken in interpreting this result, since the identification of MAIT cells in which study was only based on Jα33 TCR, meaning that also non-MAIT conventional T cells were included (88).

#### CONCLUSION

After believing for a long time that solely MHC–peptide complexes can be recognized by T cells, it is now known that TCRs can also bind (glyco)lipid, vitamin metabolites, and other non-peptidic antigens. These cell types include iNKT and MAIT cells, restricted to MHC class I-related molecules CD1d and MR1, respectively. Their evolutionary highly conserved nature indicates a strong selective pressure to be maintained in immune responses. Showing features of both the innate and adaptive immunity, these innate-like T cells act at the interface of the two systems. In this regard, it is not surprising that these cells, next to their distinct but still unclear resident role in liver tissue, are predominantly found at mucosal barriers, i.e., at sites where there is a close encounter

#### REFERENCES


with microorganisms. Next to direct activation by recognizing microbial-derived ligands *via* their semi-invariant TCR, they can also be activated indirectly (e.g., by cytokine and TLR-mediated signaling), upon which they respond by rapidly producing copious amounts of effector molecules as a first-line defense making them excellent gatekeepers against potential invasive pathogens (36, 97). However, innate-like T cells show a dichotomous phenotype, being not only protective but they are also thought to be involved in a number of immune and inflammatory diseases. Indeed, iNKT and MAIT cells might be skewed toward a predominant pro-inflammatory state in which secretion of key pathogenic cytokines such as IL-17 can cause tissue pathology (**Figure 1**) (26, 27). A pathological role for innate-like T cells is supported by evidence from diverse experimental models, although some conflicting results might reflect an aberrant role depending on the disease phenotype, activation kinetics, and the background of the animals (52, 57, 95). Interestingly, SpA-like gut and joint pathology shown in TNF- and IL-23-dependent animal models, such SKG and TNF<sup>Δ</sup>ARE/<sup>+</sup> mice does not develop under germ-free conditions (98, 99). This underscores the relevance of the host (immune)-microbial interplay in the induction of SpAlike disease features. Additionally, a state of dysbiosis as discovered recently in SpA patients (10–12) might contribute to chronicity of disease by e.g., dysregulating immunomodulatory T cell function and cytokine (TNF and IL-17) mediated responses. Although one has to keep in mind that a causal relationship has not been proven yet, we would postulate that it takes "guts" to cause joint inflammation as observed in SpA pathology. However, many questions still need to be addressed. For example, a specific role for innate-like T cells in these microbiota-mediated pathological events clearly awaits further investigation, especially in light of the complexity of the human disease. Future in depth immunoprofiling of innate-like Tcells, next to other immunomodulatory cells, in gut and joint tissues from SpA patients, combined with further exploration of their function and role in experimental models under different microbial conditions (e.g., conventional versus germ-free housing), will shed additional light on the precise nature of the relationship between these unconventional cell populations and the microbiota, and their contribution to gut and joint diseases.

# AUTHOR CONTRIBUTIONS

CM and SG contributed equally to this work. KV and DE shared supervision.


**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 Mortier, Govindarajan, Venken and Elewaut. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner 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.*

*Lucy C. Garner <sup>1</sup> \*, Paul Klenerman1,2 and Nicholas M. Provine1 \**

*<sup>1</sup> Translational Gastroenterology Unit, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom, 2Peter Medawar Building for Pathogen Research, University of Oxford, Oxford, United Kingdom*

Mucosal-associated invariant T (MAIT) cells and invariant natural killer T (iNKT) cells are innate-like T cells that function at the interface between innate and adaptive immunity. They express semi-invariant T cell receptors (TCRs) and recognize unconventional non-peptide ligands bound to the MHC Class I-like molecules MR1 and CD1d, respectively. MAIT cells and iNKT cells exhibit an effector-memory phenotype and are enriched within the liver and at mucosal sites. In humans, MAIT cell frequencies dwarf those of iNKT cells, while in laboratory mouse strains the opposite is true. Upon activation *via* TCR- or cytokine-dependent pathways, MAIT cells and iNKT cells rapidly produce cytokines and show direct cytotoxic activity. Consequently, they are essential for effective immunity, and alterations in their frequency and function are associated with numerous infectious, inflammatory, and malignant diseases. Due to their abundance in mice and the earlier development of reagents, iNKT cells have been more extensively studied than MAIT cells. This has led to the routine use of iNKT cells as a reference population for the study of MAIT cells, and such an approach has proven very fruitful. However, MAIT cells and iNKT cells show important phenotypic, functional, and developmental differences that are often overlooked. With the recent availability of new tools, most importantly MR1 tetramers, it is now possible to directly study MAIT cells to understand their biology. Therefore, it is timely to compare the phenotype, development, and function of MAIT cells and iNKT cells. In this review, we highlight key areas where MAIT cells show similarity or difference to iNKT cells. In addition, we discuss important avenues for future research within the MAIT cell field, especially where comparison to iNKT cells has proven less informative.

Keywords: mucosal-associated invariant T cells, natural killer T cells, innate-like T cells, phenotype, development, activation, effector function, subsets

#### INTRODUCTION

Mucosal-associated invariant T (MAIT) cells and invariant natural killer T (iNKT) cells are two populations of innate-like T cells that have emerged in recent years as crucial players in the development and maintenance of immunity. This is demonstrated by the array of infectious, inflammatory, and malignant diseases in which they have been implicated and in which they play diverse roles (1–5). Depending on the nature of the infectious or inflammatory setting, these can range from host protective functions, for example, antimicrobial or antitumor responses, to the augmentation of disease (1–5).

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Lucia Mori, Universität Basel, Switzerland Dominic Paquin Proulx, United States Military HIV Research Program, United States*

*\*Correspondence:*

*Lucy C. Garner lucy.garner@merton.ox.ac.uk; Nicholas M. Provine nicholas.provine@ndm.ox.ac.uk*

#### *Specialty section:*

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

*Received: 10 May 2018 Accepted: 14 June 2018 Published: 28 June 2018*

#### *Citation:*

*Garner LC, Klenerman P and Provine NM (2018) Insights Into Mucosal-Associated Invariant T Cell Biology From Studies of Invariant Natural Killer T Cells. Front. Immunol. 9:1478. doi: 10.3389/fimmu.2018.01478*

Mucosal-associated invariant T cells and iNKT cells function at the bridge between innate and adaptive immunity. While they express a T cell receptor (TCR), similar to conventional T cells of the adaptive immune system, their TCRs are semi-invariant and recognize a limited range of non-peptide ligands presented by monomorphic MHC-like molecules (6, 7). Consequently, the TCRs of MAIT cells and iNKT cells may function in a manner more akin to that of the pattern-recognition receptors expressed on innate immune cells, for example, macrophages and dendritic cells (DCs). Furthermore, MAIT cells and iNKT cells display an effector-memory phenotype prior to antigen exposure, can be activated by cytokines independent of their TCR, and can rapidly exert their effector functions upon activation without the requirement for clonal expansion, properties more analogous to innate immune cell types (8, 9). Given these distinctive characteristics, they are likely to play particularly important roles during the early stages of an immune response, prior to the differentiation of conventional effector T cells.

Although MAIT cells and iNKT cells exhibit many similarities, they also show important differences that are often disregarded. For instance, MAIT cells are the largest subset of donor-unrestricted T cells in humans, and their frequency in peripheral blood and certain tissues can be more than 100-fold greater than that of iNKT cells, whereas in mice iNKT cells are the more abundant population in most tissues (10, 11). Moreover, while MAIT cells in humans form a homogeneous population with mixed Th1/Th17 functionality, iNKT cells are highly diverse and can be divided into functionally distinct subsets (5, 11).

Given their much higher frequency in mice and the earlier availability of tetramers for their specific identification, iNKT cells have been more widely studied than MAIT cells. Furthermore, because of the similarities in their phenotypes, findings from the iNKT cell field have often been assumed to also apply to MAIT cells. With the discovery of MAIT cell ligands and the recent generation of tetramers for accurate MAIT cell identification (7, 12, 13), it is timely to consider the phenotype, development, and function of MAIT cells, how this relates to iNKT cells, and where gaps remain in our understanding. This review will focus on key areas of similarity and difference between MAIT cells and iNKT cells and will highlight important remaining questions in the MAIT cell field, many of which should now be feasible to address using the newly available tetramers.

# KEY CHARACTERISTICS

#### Frequency and Localization

Mucosal-associated invariant T cells represent a relatively large population of lymphocytes in humans, comprising up to 10% of peripheral blood T cells (14). iNKT cells are comparatively rare, with an average frequency of around 0.1% of T cells, although both MAIT and iNKT cell frequencies are highly variable between individuals (15–17). Interestingly, iNKT cells are far more abundant than MAIT cells in mice (18, 19).

Mucosal-associated invariant T cells preferentially localize to peripheral tissues (11), analogous to iNKT cells (10). In humans, MAIT cells are particularly enriched in the liver (5–50% of T cells) and are also abundant in adipose tissue, in the lung, in the female genital tract, and to varying degrees in the gut, while their frequency is low in peripheral lymph nodes (12, 14, 20–26). MAIT cells and iNKT cells in mice show a largely similar tissue distribution to human MAIT cells, with enrichment in the liver and lung (18, 19). Due to their low abundance, the tissue distribution of human iNKT cells remains poorly characterized, although they are particularly enriched in adipose tissue (27), comparable to human MAIT cells (20).

Evidence from parabiosis studies in mice suggests that iNKT cells comprise predominantly tissue-resident populations that do not recirculate, in contrast to conventional CD4<sup>+</sup> and CD8<sup>+</sup> T cells (28, 29). The capacity of tissue MAIT cells to recirculate has not yet been examined. In support of a tissue-resident phenotype, MAIT cells lack expression of the lymph node homing receptors CD62L and CCR7 (14) and express tissue-resident T cell markers in mucosal tissue, including CD69, CD103, and CRTAM (25, 30). In addition, human liver MAIT cells express LFA-1 (31), a molecule that is required for retention of liver iNKT cells in mice (28). MAIT cells accumulate in the lungs of mice following intranasal infection with *Salmonella enterica* serovar Typhimurium and remain *in situ* for at least 7 weeks post-infection, implying long-term retention in tissues (32). Finally, MAIT cells express the transcription factor PLZF (33), and conventional CD4<sup>+</sup> T cells in mice acquire a tissue-resident phenotype following ectopic expression of PLZF (28). However, CCR7<sup>−</sup>CD103<sup>−</sup> MAIT cells have recently been identified in human thoracic duct lymph at a similar frequency to that in peripheral blood (34). As CCR7 is required for lymph node entry, the authors suggest that MAIT cells in the lymph must have exited from non-lymphoid tissues. Based on these findings, it is possible that tissue MAIT cells comprise largely resident populations, while MAIT cells in certain tissues and/or particular subsets, are capable of recirculation. Such a model would need to be tested in mouse parabiosis experiments.

In mice, MAIT cell frequency is under considerable genetic control. MAIT cells show differential abundance in different strains of mice (19), and increased MAIT cell numbers in CAST/ EiJ mice can be mapped to a single genetic locus (35). Similarly, iNKT cell frequency is strongly regulated by genetic factors, as indicated by longitudinal and twin studies in humans, and analyses of iNKT cell frequency in different wild-type and congenic mouse strains (36–40). In addition to genetics, MAIT cell frequency is influenced by a number of environmental factors. Their frequency decreases in the blood with age (after ~25 years old) and in numerous diseases, while they expand in certain tissues upon infection or inflammation (3, 32, 41–44), comparable to iNKT cells (10, 45, 46). Moreover, the frequency of Vα7.2<sup>+</sup>CD161hi T cells (a proxy for MAIT cell frequency) shows no correlation in human mothers and neonates, and the correlation in Vα7.2<sup>+</sup>CD161hi T cell frequency at birth is equally high in monozygotic and dizygotic twins (47). This suggests that environmental factors may dominate over genetic factors in regulating MAIT cell frequency in humans. However, these findings need to be confirmed using the MR1/5-OP-RU [5-(2-oxopropylideneamino)-6-dribitylaminouracil] tetramer for MAIT cell identification, as MR1/5-OP-RU tetramer<sup>+</sup> MAIT cells comprise only a small fraction (<20%) of Vα7.2<sup>+</sup>CD161hi T cells at birth, in contrast to adults, where Vα7.2<sup>+</sup>CD161hi T cells are typically >95% MR1/5- OP-RU tetramer<sup>+</sup> (47). Therefore, further research is required to establish the relative role of genetic and environmental factors in regulating MAIT cell frequency in mice and humans.

### TCR Usage

The semi-invariant αβ TCRs of MAIT cells and iNKT cells comprise a largely invariant TCRα chain paired with a biased repertoire of Vβ chains. In humans, MAIT cells express a Vα7.2- Jα33/12/20 (TRAV1-2/TRAJ33/12/20) TCRα chain preferentially paired with Vβ2 or Vβ13 (TRBV20 or TRBV6) (12, 48–50), while the iNKT TCR comprises a Vα24-Jα18 (TRAV10/TRAJ18) TCRα chain paired exclusively with Vβ11 (TRBV25) (**Table 1**) (48, 51, 52). Conventional T cells recognize short peptide antigens presented by highly polymorphic MHC Class I or MHC Class II molecules. By contrast, MAIT cells and iNKT cells recognize non-peptide ligands bound to monomorphic MHC Class I-like molecules, namely riboflavin metabolites bound to MR1 (7, 13, 22), and glycolipid/phospholipid antigens bound to CD1d (6), respectively (**Table 1**).

#### Phenotype

In humans, MAIT cells are predominantly CD8<sup>+</sup> (70–90%), with some CD4<sup>−</sup>CD8<sup>−</sup> (DN) (10–20%), and a minor population of CD4<sup>+</sup> cells (**Table 1**) (12, 16, 43). iNKT cells can also be CD8<sup>+</sup> (absent in mice), DN, or CD4<sup>+</sup> (**Table 1**) (60–62, 98, 99). Within the CD8-expressing subset, both MAIT cells and iNKT cells predominantly express CD8αα homodimers or are CD8α+βlow (**Table 1**) (12, 61–63), in contrast to conventional T cell populations that are mainly CD8αβ+ (>90%). CD8αα homodimers might function to inhibit T cell activation, although their physiological role remains poorly defined (100).

Human MAIT cells display an effector-memory phenotype and characteristic expression of several surface molecules (CD161, CD26), cytokine and chemokine receptors (IL-18Rα, CCR5, CCR6), and transcription factors (PLZF, RORγt, T-bet) (**Table 1**) (11). As their phenotype is largely homogeneous and MR1 tetramers have only recently been developed, human MAIT cells are routinely identified using surrogate markers, most commonly as Vα7.2<sup>+</sup>CD161hi T cells, but also using Vα7.2 combined with IL-18Rα or CD26. In contrast to the homogeneity of MAIT cells, iNKT cells show considerable heterogeneity and thus must be directly identified using CD1d/α-galactosylceramide (αGalCer) tetramers or with an antibody to the invariant Vα24-Jα18 TCRα chain in humans. While CD4<sup>+</sup> and CD4<sup>−</sup> iNKT cell populations show disparate expression of memory markers, chemokine receptors, and natural killer (NK) cell receptors (62, 65, 84), the predominant CD4<sup>−</sup> population shows resemblance to MAIT cells, displaying an effector-memory phenotype and similar expression of surface receptors (**Table 1**) (65, 84, 101). Human MAIT cells coexpress the transcription factors PLZF, T-bet, and RORγt (91, 102), whereas their expression is subset specific for mouse MAIT cells, with cells expressing PLZF and either T-bet or RORγt (**Table 1**) (19, 102). This dichotomous expression of T-bet and RORγt is also seen in mouse iNKT cells (**Table 1**) (86). Therefore, the expression of a mixed Th1/Th17 transcription factor profile appears unique to human MAIT cells.

In summary, MAIT cells and iNKT cells show many overlapping characteristics, including expression of semi-invariant TCRs, recognition of non-peptide ligands, and an innate-like effectormemory phenotype. However, the phenotype of iNKT cells is considerably more heterogeneous than that of MAIT cells. MAIT cells and iNKT cells predominantly localize to peripheral tissues under homeostatic conditions, especially the liver and mucosal tissues, and are therefore optimally positioned to act as a first line of defense at the site of microbial infection.

# MOUSE MODELS

#### TCR Transgenic

Transgenic mouse models are widely used to study the phenotype and function of MAIT cells and iNKT cells, and their role in different disease settings. While use of these models has provided major contributions to our understanding of both cell subsets, it is also important to be aware of their limitations.

Mice that constitutively express the MAIT and iNKT cell invariant TCRα chain, namely, Vα19-Jα33 (termed iVα19 in several studies) (16, 103, 104) and Vα14-Jα18 (Vα14-Jα281 nomenclature used in early studies) (105, 106), respectively, have been generated on a Cα−/<sup>−</sup> background. As intended, these mice have an increased frequency of the target cell population. However, as is commonly observed in TCRα transgenic models, normal T cell development is dysregulated. TCR diversity is greatly reduced, T cell numbers are significantly decreased in the thymus and many peripheral tissues, and the mice harbor an expanded population of DN T cells. In addition, as mice overexpressing the MAIT or iNKT invariant TCRα chains also harbor other T cell populations (16, 103–105), comparison of mice deficient and sufficient in MR1 or CD1d, respectively, is necessary in order to identify features specific to the cell subset of interest.

Along with global changes in T cell development, MAIT cells and iNKT cells from TCRα transgenic mice exhibit certain differences in their phenotype, function, subset distribution, and tissue localization compared with their wild-type counterparts. For example, MAIT cells from iVα19 TCRα transgenic mice display a naïve phenotype, lack expression of PLZF, and secrete considerable amounts of IL-10 and Th2 cytokines (16, 103, 104), in contrast with MAIT cells from wild-type mice (19). Moreover, while wild-type iNKT cells produce both IL-4 and IFN-γ, T cells from Vα14-Jα18 TCRα transgenic mice produce high levels of IL-4, but little IFN-γ following stimulation with αCD3 (105). However, several groups have generated refined Vα14-Jα18 TCRα mouse models using somatic cell nuclear transfer (107) or induced pluripotent stem cells (108), in which iNKT cells can secrete high levels of both IL-4 and IFN-γ.

Vβ transgenic mice, for example, Vβ6 and Vβ8 transgenic mice, can be studied as an alternative to TCRα transgenics or can be crossed with TCRα transgenics to further increase MAIT or iNKT cell frequency, and to decrease unwanted TCR specificities (16, 109). An important limitation of these models is that, as MAIT and iNKT cell populations utilize multiple TCRβ chains, the forced usage of a single Vβ will bias the antigen specificity, and thereby the functionality of the generated population. In addition to the

MAIT Cell Insights From iNKT Cells


Table 1 | Characteristics of human and mouse mucosal-associated invariant T (MAIT) and invariant natural killer T (iNKT) cells.

*DN - double-negative/CD4*−*CD8*− use of double transgenics, MAIT or iNKT cell frequency can be increased by studying transgenic mice on a RAG<sup>−</sup>/<sup>−</sup> or TAP<sup>−</sup>/<sup>−</sup>Ii<sup>−</sup>/<sup>−</sup> background (16, 109). However, in these mice, interactions between MAIT or iNKT cells, and other conventional T cells (and B cells in RAG<sup>−</sup>/<sup>−</sup> mice), which might influence their phenotype and development in a wild-type setting, are completely absent.

#### Non-Transgenic

Given the scarcity of MAIT cells in mice and the limitations of TCR transgenic models, alternative models with increased MAIT cell frequency have been developed. A mouse strain (CAST/EiJ) with 20-fold greater frequency of MAIT cells than C57BL/6 mice was identified, and crossing these strains generated a B6-MAITCAST strain with increased frequencies of MAIT cells (35). These MAITCAST cells display a phenotype more consistent with MAIT cells from wild-type animals, including expression of PLZF, but some phenotypic and functional abnormalities remain. An alternate, non-genetic approach to increase the frequency of MAIT cells in mice is through the intranasal administration of MR1 ligand (5-OP-RU) combined with a toll-like receptor (TLR) agonist, which increases their frequency to approximately 50% of lung αβ T cells (32). Further work will be required to understand how this "boosting" may impact on the phenotype and function of MAIT cells, and thereby to establish the robustness of this experimental approach. Regardless of potential current shortcomings, efforts to develop mouse models with increased MAIT cell frequencies, while avoiding the limitations of TCR transgenic systems, appear promising.

#### MHC/TCR Knockout

Models with reduced, rather than increased, MAIT or iNKT cell frequencies have also been generated either by altering the TCR repertoire or by removing the MHC molecules that are essential for MAIT or iNKT cell selection. iNKT cell-deficient Jα18<sup>−</sup>/<sup>−</sup> mice are widely used; however, a recent study showed that TCRα rearrangement is perturbed in the original Jα18<sup>−</sup>/<sup>−</sup> strain (110, 111). TCRα rearrangements using Jα segments upstream of Jα18 are almost completely absent, and therefore, along with other T cell populations, MAIT cell frequency is reduced. Consequently, lack of MAIT cells may contribute to the phenotype of Jα18<sup>−</sup>/<sup>−</sup> mice. However, newer Jα18<sup>−</sup>/<sup>−</sup> models have now been generated that exhibit a normal TCRα repertoire (except for the lack of Jα18) (112–115), thus addressing this concern.

Mice lacking MR1 or CD1d lack MAIT cells (22) or iNKT cells (116–118), respectively. However, they also lack other MR1- or CD1d-restricted T cells. A population of MR1-restricted non-MAIT T cells was recently identified in humans (119), which if present in mice would be absent in MR1<sup>−</sup>/<sup>−</sup> animals. CD1d<sup>−</sup>/<sup>−</sup> mice lack not only iNKT cells (type I) but also diverse (type II) NKT cells. In addition, CD1d<sup>−</sup>/<sup>−</sup> mice on the BALB/c background, and to a lesser extent the C57BL/6 background, exhibit a marked increase in the frequency of MAIT cells, which might further confound studies using these mice (102). Jα18<sup>−</sup>/<sup>−</sup> mice have the advantage that they lack only iNKT cells. It is important to bear this in mind when using MR1<sup>−</sup>/<sup>−</sup> and CD1d<sup>−</sup>/<sup>−</sup> mice, as any identified phenotypes may not be directly attributable to MAIT or iNKT cells, respectively.

Therefore, while transgenic mouse models enable the role of MAIT cells and iNKT cells to be interrogated *in vivo* in health and disease, caution is necessary when using these models. Newer models continue to be developed that aim to overcome some of the drawbacks of existing models. Nevertheless, results from any mouse model should be validated in other models to avoid findings that result from peculiarities of the chosen experimental system. Moreover, it is important to bear in mind that discoveries from mouse studies may not directly translate to humans, given the vastly different frequencies of MAIT and iNKT cells in these species, in addition to other differences, for example, in functional subsets and tissue distribution.

## DEVELOPMENT

#### Thymic Development Selection

The earliest stages of MAIT cell development in the thymus show many similarities to the thymic development of iNKT cells. As with conventional T cells, the semi-invariant TCR of MAIT cells is generated *via* random recombination (49, 120); however, its formation requires an extended CD4+CD8+ (DP) thymocyte lifespan. Initial rearrangement of the TCRα locus utilizes 3′ Vα and 5′ Jα segments, with later rearrangements using progressively more 5′ Vα segments and more 3′ Jα segments (termed proximal to distal rearrangement) (121, 122). Thus, formation of the MAIT cell semi-invariant TCR that incorporates the 5′ most Vα segment (TRAV1-2) occurs late in the lifespan of DP thymocytes. A long DP thymocyte half-life is likewise necessary for generation of the iNKT TCR and hence for iNKT cell development. iNKT cells are absent in *Rorc*<sup>−</sup>/<sup>−</sup> mice (RORγt-deficient) that show a reduced DP thymocyte lifespan, but their development is rescued upon expression of the rearranged Vα14-Jα18 TCRα chain or the antiapoptotic protein Bcl-xL (123, 124). In peripheral blood T cells from *RORC*<sup>−</sup>/<sup>−</sup> patients, 5′ Vα-3′ Jα TCRα pairings are absent, and hence these patients lack both MAIT cells and iNKT cells (125), presumably due to lack of rearrangement of their characteristic TCRα chains at the DP thymocyte stage.

Following TCR expression, conventional T cells undergo positive selection on cortical thymic epithelial cells that present self-peptides on MHC Class I and MHC Class II. By contrast, MAIT cells are selected by MR1-expressing DP thymocytes (126, 127), comparable to the CD1d-dependent selection of iNKT cells (128, 129) (**Figure 1A**). For iNKT cells, selection is dependent on the presentation of endogenous lipid antigens by CD1d (**Figure 1A**) (130). Based on this paradigm and circumstantial evidence (131, 132), it is highly plausible that MAIT cell selection also involves an endogenous ligand(s), although such ligands have yet to be identified.

Conventional T cells are positively selected in the thymus when their TCR exhibits moderate affinity for MHC/self-peptide, while thymocytes expressing high affinity TCRs are removed from the repertoire. The strength of the TCR-MR1/ligand interaction required for MAIT cell positive selection has not been investigated, but agonist selection is hypothesized based on the following information. First, a number of unconventional T cell

Figure 1 | Comparison of mucosal-associated invariant T (MAIT) cell and invariant natural killer T (iNKT) cell thymic and peripheral development in mice. (A) MAIT cells and iNKT cells are positively selected by MR1- and CD1d-expressing DP (double-positive/CD4+CD8+) thymocytes, respectively. iNKT cell positive selection involves an endogenous ligand(s). A similar role for endogenous ligand(s) in MAIT cell selection is postulated, but such ligands have yet to be identified. Concomitant with T cell receptor (TCR)-MHC-Ib/ligand binding, homotypic interactions between signaling lymphocyte activation molecule (SLAM) family receptors are essential for iNKT cell, but not MAIT cell, development. (B) iNKT cells also undergo negative selection, while negative selection has not been studied for MAIT cells. (C) Following selection, MAIT and iNKT thymocytes differentiate through similar stages defined by the expression of CD24 and CD44. Stage 1 and stage 2 iNKT thymocytes are highly proliferative, whereas the proliferative capacity of thymic MAIT cells is currently unknown. A number of shared factors are required for thymic differentiation, including microRNAs (miRNAs) and PLZF, but the requirement for IL-18 and exogenous ligand (from commensal bacteria) is specific to MAIT cells. Conversely, the transcription factors Egr1, Egr2, and c-Myc have been implicated in iNKT cell development, but not investigated in MAIT cell development. iNKT cells express PLZF and exhibit effector functions at stage 1, while MAIT cells acquire effector capacity at stage 3. (D) MAIT cells and iNKT cells exit the thymus with a CD24−CD44+ memory phenotype. In the periphery, MAIT cells undergo expansion, probably driven by the presentation of exogenous ligands from commensal bacteria, whereas iNKT cell frequency remains relatively constant. (E) iNKT cell homeostasis is predominantly regulated by cytokines, in particular IL-7 and IL-15. By contrast, the role of MR1 and cytokines in MAIT cell homeostasis is currently unknown.

lineages are selected by agonist ligands, including iNKT cells, regulatory T cells, and CD8αα gut intraepithelial T cells (133). Second, compared with conventional thymocytes, the high avidity interaction between the iNKT TCR and selecting glycolipids results in strong TCR signaling (134) and therefore prolonged upregulation of the TCR-induced transcription factors Egr1 and Egr2 (135). The transcription factor PLZF, encoded by *Zbtb16*, is a direct downstream target of Egr2 (135), and both MAIT cells and iNKT cells upregulate PLZF expression during thymic development, contrasting to conventional T cells. Finally, mouse MAIT cells upregulate CD44 expression, and mouse and human MAIT cells can acquire effector function within the thymus (102), properties of antigen-experienced conventional T cells.

MR1 is essential for MAIT cell positive selection (**Figure 1A**) (126). However, whether engagement of other cell surface receptors is required, is currently unknown. By contrast, homotypic interactions between at least two signaling lymphocyte activation molecule (SLAM) family members (SLAMF1 and SLAMF6) are required alongside TCR-CD1d/ligand engagement, for iNKT cell development (**Figure 1A**). In mixed bone marrow chimeras, the frequency of iNKT cells in *Slamf1*/*Slamf6*-deficient populations is significantly reduced compared with wild-type, with a specific defect at the transition from stage 0 (CD24hi) to stage 1 (CD24lo) (136). SLAMF6 costimulation has been shown to augment TCR signaling, resulting in increased Egr2 expression and consequently enhanced expression of PLZF (137). MAIT cell development is independent of SLAM receptors, as patients deficient in SLAM-associated protein, an intracellular adaptor required for SLAM signaling, lack iNKT cells but show normal numbers of MAIT cells (16).

The role of negative selection in MAIT cell development has not been investigated (**Figure 1B**). By contrast, although not explicitly demonstrated, highly autoreactive iNKT cells likely undergo negative selection on DCs (**Figure 1B**) (138, 139). Addition of the agonist glycolipid αGalCer or CD1d overexpression during iNKT cell development results in decreased iNKT cell frequency *in vitro* and *in vivo* (138, 139). It seems likely that high avidity self-reactive MAIT thymocytes also undergo negative selection. Alternatively, peripheral MAIT cell activation could be controlled by other mechanisms, for example, dampened TCR signaling compared with conventional T cells (85).

#### Differentiation

The differentiation of thymic MAIT cells following selection remains relatively unexplored. However, a recent paper by Koay et al. identified three stages of MAIT cell development in mice and humans (102). The described developmental pathway in mice, with stages defined by the expression of CD24 and CD44 (stage 1 – CD24<sup>+</sup>CD44<sup>−</sup>, stage 2 – CD24<sup>−</sup>CD44<sup>−</sup>, stage 3 – CD24<sup>−</sup>CD44<sup>+</sup>), is remarkably similar to the linear differentiation model of iNKT cell development (**Figure 1C**) (130). In mice, thymic stage 1 (CD24loCD44lo) and stage 2 (CD24loCD44hi) iNKT cells are highly proliferative (**Figure 1C**) (140). Intrathymic iNKT cell proliferation requires expression of the transcription factor c-Myc (141). Mouse MAIT cells accumulate in the thymus with age, and stage 3 MAIT cells are more abundant than stage 1 and stage 2 in fetal thymic organ culture (102). These data suggest that murine MAIT cells proliferate in the thymus, similar to iNKT cells, but direct measures of *in vivo* proliferation have not been performed. By contrast, human thymic MAIT cells are present at low frequency irrespective of age (16, 102, 127), and T cell receptor excision circle (TREC) analysis of MAIT cells in human thymus and cord blood identified no differences in TREC concentration compared with conventional T cells (127). Whether human thymic and cord blood iNKT cells show enhanced proliferative capacity relative to conventional T cells is unclear, as prior studies have reported conflicting findings (73, 74). Thus, additional independent direct assessments of MAIT and iNKT thymocyte proliferation are needed to clarify the extent of their intrathymic proliferative capacity.

Development of MAIT cells along the linear developmental pathway requires a number of different factors, some of which are also necessary for iNKT cell development (**Figure 1C**). MR1 is essential at all stages of MAIT cell development *in vitro*, and peripheral MAIT cells are nearly absent in MR1-deficient mice (16, 19, 22, 102). Likewise, CD1d is essential for the development of iNKT cells (116–118, 142). In the absence of commensal bacteria and IL-18 *in vivo*, stage 3 MAIT cells are reduced, while the frequency (but not number) of stage 1 cells is increased (22, 102). Moreover, MAIT cell development beyond stage 1 requires microRNAs (miRNAs), as the abundance of stage 2 and stage 3 MAIT cells is decreased in *Drosha*-deficient mice (102). By contrast, PLZF is necessary only for MAIT cell maturation from stage 2 to stage 3 and for their acquisition of effector function (19, 102).

Invariant natural killer T cell development similarly requires miRNAs and PLZF (33, 92, 143). Consistent with a shared developmental niche, MAIT cell frequency is markedly increased in CD1d-deficient mice on a BALB/c background, with only minor differences to wild-type on the C57BL/6 background, although the increase in both strains is statistically significant (102). By contrast, the number of iNKT cells in the spleen and thymus of MR1-deficient mice is similar to that of wild-type, perhaps due to the much lower frequency of MAIT cells compared with iNKT cells in these mouse strains (102). In addition, the frequency of MAIT cells and iNKT cells in humans is positively correlated in adult peripheral blood (43). Despite shared development needs, the absolute requirement for commensal bacteria appears unique to MAIT cells (22, 102), as iNKT cell frequency is relatively conserved in germ-free (GF) mice compared with either specific pathogen-free (SPF) mice or mice harboring a conventional microflora (144, 145).

Stage 3 mature MAIT cells in human thymus coexpress RORγt and T-bet (102). By contrast, stage 3 MAIT cells in mice comprise two subsets, namely, RORγt <sup>+</sup>T-bet<sup>−</sup> and T-bet<sup>+</sup>RORγt <sup>−</sup> cells (102). Analogous to MAIT cells, thymic iNKT cell subsets have not been identified in humans, while iNKT cells comprise at least three different subsets in mouse thymus, named NKT1, NKT2, and NKT17 (discussed in more detail in Section "Subsets") (86, 146, 147). It is unclear whether RORγt <sup>+</sup>T-bet<sup>−</sup> and T-bet<sup>+</sup>RORγt − MAIT cells represent different developmental stages or distinct subsets derived from a shared progenitor. Recent studies suggest that iNKT cell subsets arise from a common PLZFhi precursor population and represent stable lineages with distinct transcriptional and epigenetic programs (86, 146, 147). However, whether the classic developmental stages model or the newer lineage segregation model best describes iNKT cell development, remains uncertain. Moreover, the specific signals required for commitment to the different iNKT cell subsets are largely unknown, although a multitude of factors, including cytokines and transcriptional regulators, can differentially regulate NKT1, NKT2, and NKT17 development (148). It would be worth investigating whether similar factors also modulate the differentiation of thymic MAIT cell subsets.

#### Acquisition of Innate-Like Effector Function

Mucosal-associated invariant T cells can acquire innate-like effector function in the thymus and secrete cytokines upon activation (102), comparable to iNKT cells (15, 140, 149, 150). Expression of the transcription factor PLZF is necessary and sufficient to drive innate-like effector differentiation (33, 92, 151–153). In PLZF-deficient mice, MAIT and iNKT cell development is almost completely abrogated and residual cells exhibit a CD44lo phenotype, reduced expression of characteristic phenotypic markers, and impaired cytokine secretion (19, 33, 92, 102). In addition to direct regulation by Egr2 (135), PLZF expression is regulated by the binding of Runx1 to a shared intronic enhancer in several innate lymphoid lineages, including iNKT cells (154). Therefore, Runx1 likely also regulates PLZF expression in MAIT cells. During thymic MAIT cell development, PLZF expression begins at stage 2 (mouse – CD24<sup>−</sup>CD44<sup>−</sup>, human – CD161<sup>−</sup>CD27<sup>+</sup>) and is highest at stage 3 (mouse – CD24<sup>−</sup>CD44<sup>+</sup>, human – CD161<sup>+</sup>CD27pos-lo) (**Figure 1C**) (102). By contrast, PLZF expression is induced immediately following positive selection in iNKT cells and its expression peaks in thymic stage 1 cells (CD24loCD44lo) (**Figure 1C**) (33, 92, 135). Consequently, thymic stage 1 iNKT cells can secrete cytokines upon stimulation (140, 149, 150), while MAIT cells do not acquire this functionality until stage 3 of thymic differentiation (102).

Although they can acquire effector function within the thymus, MAIT cells in humans are typically thought to exit the thymus as naïve cells and acquire their effector-memory phenotype in the periphery (14, 16, 47, 63, 127). This is supported by the naïve phenotype of MAIT cells in thymus and cord blood, and their rapid acquisition of CD45RO in neonates, such that >80% of blood MAIT cells express CD45RO by 1 month of age (14, 16, 47, 63, 127). However, further studies are required to fully define exactly when and where MAIT cells acquire their effector-memory phenotype and function, given that some thymic MAIT cells express PLZF and CD45RO (16, 102, 155). Naïve stage 2 (CD161<sup>−</sup>CD27<sup>+</sup>) MAIT cells were recently shown to predominate in human thymus and were found to a lesser degree in cord blood and young blood (102). Thus, the majority of MAIT cells may exit the thymus at stage 2 and undergo further maturation in the periphery, while a small population matures to stage 3 in the thymus (102). Stage 3 mature MAIT cells (CD24<sup>−</sup>CD44<sup>+</sup>) are the main population in mouse thymus (102). Therefore, contrasting to human MAIT cells, but comparable to mouse iNKT cells, mouse MAIT cells probably exit the thymus as CD44<sup>+</sup> memory cells (140). Similarly, human iNKT cells may leave the thymus as effector-memory cells, as they already display a CD45RO<sup>+</sup> memory phenotype in thymus and cord blood (66, 67, 156).

In conclusion, thymic MAIT cell development shows many similarities to that of iNKT cells, including selection on DP thymocytes, development through similar stages post-selection, a shared requirement for developmental factors, and the possibility to acquire innate-like effector function in the thymus. However, the role of MAIT cell negative selection and the extent of their intrathymic proliferation have yet to be examined. While mouse iNKT cells and likely mouse MAIT cells exit the thymus as CD44<sup>+</sup> effector-memory cells, human MAIT cells appear to leave the thymus as naïve cells and acquire innate-like effector function extrathymically, although the exact timing of their thymic exit needs to be clarified. The reason for such disparity between mouse and human MAIT cells in the location of effector maturation is currently unclear.

# Peripheral Development

#### Changes in Abundance

While their frequency is relatively constant in the thymus, MAIT cells undergo a large population expansion in the periphery, reminiscent of intrathymic iNKT cell expansion in mice, increasing over 100-fold from <0.01% of T cells at birth to 1–10% of T cells by adulthood (14, 41–43, 47, 102). The increase in MAIT cell numbers is gradual and occurs over a number of years, although estimates for the age at which adult frequencies are reached vary between studies (6–25 years of age) (14, 41–43, 47, 102). By contrast, peripheral iNKT cell frequencies remain relatively constant from birth to adulthood (66, 157, 158).

Though MAIT cell thymic selection is independent of B cells (16), B cells are crucial for peripheral MAIT cell expansion in mice (**Figure 1D**). Peripheral MAIT cells are almost entirely absent in B cell-deficient mice, and transfer of B cells is sufficient to induce MAIT cell expansion in iVα19/Vβ6 RAG<sup>−</sup>/<sup>−</sup> mice (16, 22). Whether cognate interactions between B cells and MAIT cells are necessary for such expansion has not been established. In humans, the role of B cells remains uncertain. Treiner et al. observed a reduced frequency of MAIT cells (as measured by the presence of Vα7.2-Jα33 TCRα transcripts) in the blood of patients who lack B cells due to a mutation in Bruton tyrosine kinase (*BTK*) (22). However, only four patients were analyzed, one of which had a normal number of MAIT cells. A study of common variable immunodeficiency (CVID) provides indirect evidence against a role for B cells in regulating human MAIT cell frequency (159). Although the abundance of B cells and MAIT cells was variably decreased in CVID patients, the frequency of MAIT cells showed no correlation with that of B cells. A major confounding factor in these human studies is the increased occurrence of infections in patients with *BTK* deficiency and CVID, which can independently modulate MAIT cell frequency. Thus, whether B cells have a role in MAIT cell expansion or at other stages of their development in humans requires further investigation. B cells are not essential for iNKT cell development, but they do play an important role in human peripheral iNKT cell homeostasis (160), as discussed below.

It is widely hypothesized that peripheral MAIT cell expansion and maturation is driven by the presentation of microbial antigens on MR1, derived from either commensal or pathogenic bacteria (**Figure 1D**). Although this has yet to be formally proven, a variety of evidence supports this hypothesis. In humans, MAIT cells are naïve in thymus, cord blood, and in newborns, but rapidly acquire a memory phenotype in the blood during the first month of life (14, 16, 47, 102), concomitant with exposure to bacteria. MAIT cells are absent in GF mice (22) and expand upon microbial reconstitution with a single strain of bacteria (17). Furthermore, MAIT cells undergo MR1-dependent proliferation *in vitro* and *in vivo* in response to bacteria, for example, in the lungs of mice infected with *Salmonella enterica* serovar Typhimurium (32, 91, 161). The TCR repertoire of MAIT cells also supports microbemediated expansion. While the TCR repertoire is polyclonal in cord blood, it is oligoclonal in adult blood (47, 49, 50, 63), consistent with the hypothesized expansion of specific clones in response to particular bacteria. This is plausible, as MAIT cells with distinct TCRs are activated *in vitro* following stimulation with different bacteria (162), and in a human *in vivo Salmonella enterica* serovar Paratyphi A challenge setting, the relative abundance of different MAIT cell clonotypes changes in response to infection (163).

#### Phenotypic and Functional Maturation

In addition to expansion, MAIT cells undergo maturation in the periphery, as indicated by marked phenotypic changes. Approximately half of MAIT cells in the thymus are either DP or CD4<sup>+</sup>, whereas MAIT cells in adult blood are predominantly DN or CD8<sup>+</sup> (**Figure 1D**) (16, 49, 102). Furthermore, CD8<sup>+</sup> MAIT cells in the thymus and cord blood are CD8αβ+, whereas roughly half of CD8α+ MAIT cells in adult blood express CD8αα homodimers (16, 63, 102). CD8αα+ MAIT cells are thought to arise from CD161hiCD8αβ+ cells in the periphery (47, 63). iNKT cells undergo similar phenotypic changes with age. CD4<sup>+</sup> cells comprise 80–90% of iNKT cells in human thymus and cord blood, and progressively decline in the periphery to comprise on average 40% of iNKT cells in adult blood (15, 73, 158). This may result from the preferential peripheral expansion of CD4<sup>−</sup> iNKT cells, as CD4<sup>−</sup> iNKT cells show reduced TREC content and increased proliferation in response to IL-15 compared with CD4<sup>+</sup> iNKT cells (73). However, alternative explanations, such as CD4 downregulation, remain possible. It is unknown if CD4<sup>+</sup> and CD8<sup>+</sup> MAIT cells show differences in their proliferative capacity. Analogous to MAIT cells, a large proportion of CD8<sup>+</sup> iNKT cells in human blood express CD8αα (61, 62), but whether these arise in the periphery has not been investigated.

Following thymic exit, MAIT cells acquire a memory CD45RO<sup>+</sup> phenotype and upregulate the expression of characteristic phenotypic markers, such as CD161 and IL-18Rα (**Figure 1D**), while downregulating the expression of lymph node homing receptors, including CD62L and CCR7 (14, 47, 63, 102, 127, 155). These changes are gradual, as MAIT cells in cord blood, young blood, and adult blood exhibit an increasingly mature phenotype (47, 63, 102, 127, 155). iNKT cells undergo similar extrathymic phenotypic changes (15, 67, 73, 156, 158, 164), although they already exhibit a memory phenotype in the thymus and cord blood (15, 66, 67, 140). Upregulation of NK cell receptors on mouse iNKT cells is dependent on CD1d (142), while IL-7 can upregulate CD161 on human cord blood iNKT cells *in vitro* (74). The signals required for NK cell receptor upregulation on developing MAIT cells are currently unknown.

As well as phenotypic changes, MAIT cells undergo further functional differentiation following thymic exit. Although stage 3 MAIT cells in human thymus can produce IFN-γ and TNF-α following PMA and ionomycin stimulation, their capacity to do so is significantly decreased compared with peripheral blood MAIT cells (102). Moreover, the majority of human MAIT cells may exit the thymus at stage 2 before they acquire effector capacity. In contrast to adult blood Vα7.2<sup>+</sup>CD161hi T cells, cord blood Vα7.2<sup>+</sup>CD161hi T cells are unable to produce IFN-γ or Granzyme B in response to overnight stimulation with *Escherichia coli*infected THP-1 cells (47). However, as MAIT cells comprise the majority of Vα7.2<sup>+</sup>CD161hi T cells in adult blood, but only a small fraction of Vα7.2<sup>+</sup>CD161hi T cells in cord blood (47), whether this finding also applies to MAIT cells needs to be confirmed using the MR1 tetramer. Similar to the findings for MAIT/Vα7.2<sup>+</sup>CD161hi T cells, human thymic and cord blood iNKT cells appear functionally immature compared with adult blood iNKT cells. Early reports suggested that thymic and cord blood iNKT cells were incapable of cytokine production without prior *in vitro* expansion (66, 73). However, more recently, freshly isolated iNKT cells from the thymus and cord blood were shown to secrete cytokines, including IFN-γ, TNF-α, and IL-4, in response to TCR and/ or PMA and ionomycin stimulation (15, 74). Consequently, the capacity of human thymic and cord blood iNKT cells to produce cytokines needs to be clarified. In contrast to human MAIT cells and iNKT cells, mouse thymic MAIT (102) and iNKT (140, 149, 150) cells strongly produce cytokines, suggesting possible species-specific differences in when cytokine-producing capacity is acquired.

#### Homeostasis

The requirements for MAIT cell proliferation and survival in the periphery are poorly understood (**Figure 1E**). Conventional memory T cells depend predominantly on cytokines for peripheral maintenance (165), suggesting stimulation with cytokines, as opposed to MR1, might be key for MAIT cell homeostasis. iNKT cells exhibit subset-specific requirements for cytokines. While IL-15 is indispensable for the survival and functional maturation of most iNKT cells in mice (75, 166), NKT17 cell homeostasis is exclusively dependent on IL-7 (76) (**Figure 1E**). Moreover, IL-15 and IL-7 preferentially stimulate the proliferation of CD4<sup>−</sup> and CD4<sup>+</sup> human iNKT cells, respectively (73). In contrast to iNKT cells, MAIT cells proliferate only in response to IL-15 (161), and not IL-7 (91), despite their exquisite sensitivity to stimulation with either cytokine (30, 91, 167–169). Cytokines that signal *via* STAT3 are required for human MAIT and iNKT cell development and homeostasis, as indicated by the 4- and 20-fold reduction in their frequency, respectively, in patients with heterozygous loss-of-function mutations in *STAT3* (77). The central role of STAT3 appears to be downstream of the IL-23 receptor (and possibly the IL-21 receptor) in MAIT cells, and the IL-21 receptor in iNKT cells. IL-18 is similarly necessary for MAIT cell development and/or survival, as IL-18-deficient mice exhibit reduced thymic and peripheral MAIT cell frequencies (102). Interestingly, the role of IL-18 appears independent of IL-18 receptor signaling, as MAIT cell development is normal in IL-18Rα-deficient mice (102). Therefore, further work is necessary to determine the specific role of IL-23 and IL-18 in regulating MAIT cell frequency and to establish the requirement for IL-7, IL-15, and additional cytokines in MAIT cell homeostasis. Furthermore, it remains to be investigated whether RORγt <sup>+</sup> MAIT cells and T-bet<sup>+</sup> MAIT cells in mice are differentially regulated by cytokines, as has been demonstrated for the equivalent murine iNKT cell subsets.

It is unknown if tonic TCR signaling is necessary for MAIT cell homeostasis (**Figure 1E**). iNKT cell homeostasis in mice appears independent of CD1d (**Figure 1E**). iNKT cells can survive for weeks in the periphery of mice in the absence of CD1d (142, 170), and the homeostatic expansion of iNKT cells in lymphopenic hosts is CD1d independent (75, 166). By contrast, CD1d may play a role in human iNKT cell homeostasis through lipid antigen presentation on B cells. Compared with iNKT cells from total PBMCs, iNKT cells from B cell-depleted PBMCs (but not from PBMCs depleted of other CD1d<sup>+</sup> populations) display reduced proliferation and cytokine production *in vitro* upon stimulation with αGalCer + IL-2 (160). In addition, iNKT cells exhibit decreased frequency and altered functionality in systemic lupus erythematosus patients, associated with reduced CD1d expression on immature B cells (160, 171, 172). Restoration of CD1d expression is sufficient to reverse these defects both *in vitro* and *in vivo* (160). Thus, it is worth examining whether MR1 has a role in MAIT cell homeostasis, particularly in humans. However, although MR1 is widely expressed by hematopoietic and nonhematopoietic cells, it is largely retained in the endoplasmic reticulum prior to ligand exposure (132, 173–175). Consequently, the ability of MR1 to modulate MAIT cell homeostasis may be limited compared with CD1d, which is frequently present at the cell surface (176).

In conclusion, MAIT cells and iNKT cells undergo further extrathymic maturation. However, while peripheral iNKT cell frequency remains relatively constant with age, MAIT cells undergo a large population expansion from birth to adulthood. B cells have an important, but differing, role in MAIT cell and iNKT cell peripheral development. Compared with MAIT cells, more is known about the role of cytokines in the peripheral maintenance of iNKT cells. Given that MAIT cells express similar cytokine receptors to iNKT cells, including the receptors for IL-7 and IL-15 (14), it is worth investigating the role of these cytokines in MAIT cell homeostasis and peripheral maturation. With the availability of MR1 tetramers (12, 13) and mice with an increased frequency of MAIT cells (35), this can now be examined *in vivo* using cytokine-deficient mice.

#### Fetal Development

Similar to human iNKT cells (64), human MAIT cells develop in fetal thymus and can be identified in both lymphoid and non-lymphoid peripheral tissues in the second trimester of fetal development (155). As the timing of early MAIT cell and iNKT cell development in humans is comparable (47, 64, 155), and iNKT cells develop postnatally in mouse thymus (150, 177), it is likely that MAIT cells also undergo postnatal development in mice.

Before discussing human fetal MAIT cell development, it is important to note that, while fetal iNKT cell development has been studied using the CD1d tetramer, fetal MAIT cell development has so far only been investigated using the surrogate MAIT cell markers Vα7.2 and CD161. As previously mentioned, the MAIT cell populations defined as MR1/5-OP-RU tetramer<sup>+</sup> or Vα7.2<sup>+</sup>CD161hi are essentially the same in adult blood, while MR1/5-OP-RU tetramer<sup>+</sup> MAIT cells comprise <20% of Vα7.2<sup>+</sup>CD161hi T cells at birth (47). Moreover, the majority of MAIT cells in human thymus are CD161<sup>−</sup>CD27<sup>+</sup> stage 2 cells, and stage 2 MAIT cells are also present at lower frequencies in cord blood and young blood (~20% and ~10% of MR1/5-OP-RU tetramer<sup>+</sup> MAIT cells, respectively) (102). Therefore, using Vα7.2 and CD161 for fetal MAIT cell identification will fail to capture these CD161<sup>−</sup> MR1/5-OP-RU tetramer<sup>+</sup> MAIT cells. Overall, findings from the study of Vα7.2<sup>+</sup>CD161hi T cell development in fetal tissues may not accurately reflect the developmental pathway of MAIT cells. This should be taken into consideration when interpreting the findings discussed below, all of which were made using Vα7.2 and CD161 to identify "MAIT" cells.

#### Frequency and Localization

During fetal development, Vα7.2<sup>+</sup>CD161hi T cells comprise ~0.05% of T cells in human thymus, significantly lower than their frequency in adult blood (155). Their frequency in the thymus remains low and relatively constant after birth, at least up until the age of 14 (102). In contrast to MAIT cells, iNKT cell frequency in early fetal thymus is similar to that in adult blood (~0.1% of T cells) (156). However, their frequency decreases with gestational age in the thymus, cord blood, and neonatal peripheral blood, such that they are rare in postnatal thymus, while it increases in fetal peripheral tissues, particularly the small intestine and spleen (47, 64, 73, 156, 158, 178). This suggests a wave of iNKT cell development in the thymus early during fetal life (156), accompanied by the gradual population of peripheral tissues. This might also be true for MAIT cells, as the frequency of Vα7.2<sup>+</sup>CD161hi T cells in the blood of neonates decreases with gestational age (47). However, whether their thymic frequency also decreases, is currently unknown. Contrary to iNKT cells, no correlation was observed between gestational age and Vα7.2<sup>+</sup>CD161hi T cell frequency in fetal tissues (155), although the sample size was relatively low.

As in adults, Vα7.2<sup>+</sup>CD161hi T cells are enriched in fetal peripheral tissues, including the lung, liver, and small intestine, with lower frequencies in the thymus and secondary lymphoid organs (SLOs) (155). iNKT cells are similarly enriched in the small intestine, but relatively depleted in the liver, lung, and SLOs (64). The frequency of Vα7.2<sup>+</sup>CD161hi T cells in fetal tissues is low compared with the corresponding adult tissues, particularly in the liver, where they are ~100-fold less frequent in the fetus (14, 155). The frequency of human iNKT cells in adult peripheral tissues is poorly characterized, due to their low abundance. Nonetheless, at least in liver and spleen (15, 179, 180), their frequency appears largely similar to that in fetal tissues.

#### Phenotypic and Functional Maturation

In all fetal tissues, Vα7.2<sup>+</sup>CD161hi T cells are less differentiated than in adult blood (155), analogous to iNKT cells (64, 74, 156). Nevertheless, Vα7.2<sup>+</sup>CD161hi T cells in fetal peripheral tissues, particularly the small intestine, show a more mature phenotype than those in the thymus and SLOs, with increased expression of IL-18Rα and CD45RO, and reduced expression of CD62L (155). In addition, peripheral tissue Vα7.2+CD161hi T cells are functionally more mature than their counterparts in lymphoid organs, producing increased IFN-γ *in vitro* following *E. coli* stimulation (155). Similarly, iNKT cells are phenotypically and functionally more mature in peripheral tissues compared with lymphoid organs (64). However, while Vα7.2<sup>+</sup>CD161hi T cells are naïve in cord blood (16) and only a fraction express CD45RO in fetal thymus (155), >80% of iNKT cells exhibit a memory CD45RO<sup>+</sup> phenotype in both cord blood (66, 67) and fetal thymus (156). Moreover, the proportion of Vα7.2<sup>+</sup>CD161hi T cells that produce IFN-γ is significantly reduced in fetal peripheral tissues compared with adult blood (155), whereas iNKT cells in fetal small intestine and adult blood show largely comparable IFN-γ production in response to αGalCer (64). As iNKT cells from GF mice exhibit reduced cytokine production compared with their counterparts from standard SPF mice (181), and the fetal environment is typically thought to be sterile (182), it is perhaps surprising that fetal iNKT cells do not display reduced functionality compared with those in adult peripheral blood. However, this could be understood if the fetal environment was not entirely sterile, conflicting with the "sterile womb hypothesis" (182). In support of this suggestion, a number of recent papers provide evidence for the presence of microbes during fetal development, although these findings remain highly controversial [reviewed in Ref. (182)].

The discovery of mature CD45RO<sup>+</sup> Vα7.2<sup>+</sup>CD161hi T cells in fetal peripheral tissues appears at odds with the requirement for commensal bacteria for the development and maturation of MAIT cells in mice (17, 22, 102). Moreover, MAIT cells exhibit a naïve CD45RA<sup>+</sup> phenotype in cord blood (16) and neonates (47), and rapidly upregulate CD45RO following birth, concomitant with their exposure to riboflavin-synthesizing commensal bacteria (14, 16, 47, 102). This suggests microbe-driven maturation of human MAIT cells, akin to mouse MAIT cells. The reason for the discordant findings in fetal tissues and postnatal blood is currently unclear. As mentioned earlier, it is possible that MAIT cells in peripheral tissues comprise largely tissue-resident populations distinct from those in blood, similar to what has been proposed for iNKT cells based on parabiosis experiments in mice (28, 29). However, this has yet to be investigated for MAIT cells. Regardless, this would not explain why MAIT cells undergo maturation in fetal peripheral tissues. Maturation could be understood if the fetal environment was not completely GF, as discussed above. Alternatively, it is possible that other unknown factors can mediate fetal MAIT cell maturation.

In summary, we have a very limited understanding of fetal MAIT cell development. Only one paper has addressed MAIT cell development in human fetal tissues and the MR1 tetramer was not used in this study. Nevertheless, the findings for Vα7.2<sup>+</sup>CD161hi T cells are reminiscent of iNKT cell fetal development, with Vα7.2<sup>+</sup>CD161hi T cells undergoing maturation in fetal peripheral tissues, particularly at mucosal sites. Now that MR1 tetramers are readily available, it will be necessary to establish whether MR1/5- OP-RU tetramer<sup>+</sup> MAIT cells show similar fetal maturation to Vα7.2<sup>+</sup>CD161hi T cells and if so, to explore the factors driving such maturation.

#### ACTIVATION

#### Mechanisms

Analogous to iNKT cells, MAIT cells can be activated by TCR signals, cytokine signals independent of the TCR, or by combined TCR and cytokine signals (**Figure 2**).

The MAIT cell semi-invariant TCR recognizes bacterial and yeast riboflavin metabolite ligands in the context of MR1, with the most potent ligands being 5-OP-RU and 5-OE-RU [5-(2-oxoethylideneamino)-6-d-ribitylaminouracil] (**Table 1**; **Figure 2A**) (7, 13). By contrast, iNKT cells recognize various glycolipid and phospholipid antigens bound to CD1d (**Table 1**; **Figure 2A**), including glycosphingolipids from *Sphingomonas* spp. and *Bacteroides fragilis*, diacylglycerols from *Borrelia burgdorferi* and *Streptococcus pneumoniae*, and endogenous lysophospholipids (57). Although a wide range of lipid antigens have been identified for iNKT cells, compared with only a few for MAIT cells, the list of known antigens for both subsets is likely not exhaustive.

As the riboflavin synthesis pathway is present in diverse pathogenic and commensal bacteria, as well as in yeast, but absent in mammals, recognition of riboflavin metabolites enables MAIT cells to effectively discriminate self from non-self. No endogenous ligands have been identified for MAIT cells, although this is an active area of research. The ability of bacteria to activate MAIT cells *in vitro* strongly correlates with the presence of the riboflavin metabolic pathway (7) and activation of MAIT cells by a number of viruses, including dengue, influenza A, and hepatitis C, is MR1-independent (168, 183–185)*.* Thus, presentation of

by cytokines, iNKT cells can be directly activated *via* certain natural killer (NK) cell receptors, such as NK1.1 in mice. IFN-γ production may predominate following TCR-independent activation, although this requires further investigation. (C) MAIT cells and iNKT cells can be activated through a combination of TCR and cytokine signaling. In this setting, iNKT cell antigens are typically weak self-antigens (bold), but endogenous ligands for MAIT cells have not been identified.

endogenous ligands by MR1 does not appear to be important for MAIT cell activation in bacterial or viral infections. However, endogenous ligands could have a key role *in vivo* in inflammation and cancer, or in specific infectious settings. This is plausible, as MR1 can bind endogenous ligands to activate non-MAIT T cells (119). In contrast with MAIT cells, self-lipid ligands are known to play a key role in iNKT cell biology. Although a number of endogenous ligands can activate iNKT cells, including lysophosphatidylcholine (human iNKT cells only) and ether-bonded mono-alkyl glycerophosphates (186, 187), recent evidence suggests that, at least in mice, α-linked glycosylceramides are the major endogenous ligands (188).

In the absence of riboflavin metabolites, MAIT cells can be activated by cytokines independent of their TCR (**Figure 2B**). Similar to iNKT cells (189–192), they are potently activated by IL-12 + IL-18, as well as by various combinations of IL-12, IL-15, IL-18, and type I interferons (30, 167, 168, 193). In general, a single cytokine is insufficient to induce significant activation. MAIT cells express high levels of IL-18Rα and IL-18 appears dominant for their TCR-independent activation, at least in viral infections (**Figure 2B**) (17, 168, 183). By contrast, IL-12 is key for iNKT cell activation in the absence of TCR stimulation (**Figure 2B**) (78, 191, 192).

It is unknown if MAIT cells are permanently amenable to TCR-independent cytokine stimulation. The capacity of human iNKT cells to respond to cytokine stimulation alone appears to reflect a transitory state that depends on prior TCR stimulation. In response to weak TCR stimulation by CD1d/ self-lipid, histone H4 acetylation at the *IFNG* locus leads to a transient increase in the responsiveness of iNKT cells to innate stimulation with IL-12 + IL-18 independent of additional TCR signaling, which decays over a period of hours to days (79). As iNKT cells adoptively transferred into CD1d<sup>−</sup>/<sup>−</sup> or CD1d<sup>+</sup>/<sup>+</sup> mice show comparable responses to a number of bacteria and viruses (191, 194), cytokine-dependent activation of mouse iNKT cells may be entirely TCR independent. However, it remains possible that iNKT cells undergo TCR signaling in donor mice prior to adoptive transfer.

Mucosal-associated invariant T cells can integrate both TCR- and cytokine-dependent signals to augment functional capacity (30, 91, 167, 169, 193, 195, 196), similar to iNKT cells (197–199) (**Figure 2C**). Many of the cytokines that can drive TCRindependent activation have been shown to costimulate TCR signaling for MAIT cells, including IL-12, IL-15, and/or IL-18. These are typically produced by antigen-presenting cells (APCs) downstream of TLR activation. MAIT cell activation following *E. coli* stimulation of THP-1 cells is mediated by TLR4-induced IL-12 + IL-18, combined with MR1-dependent TCR activation by microbial ligand(s) (193). In this model, early MAIT cell activation depends predominantly on TCR signals, while both TCR and cytokine signals are crucial at later time points. In a similar manner, iNKT cells are activated by self-lipids together with IL-12 following TLR4 or TLR7/8 stimulation of DCs (197, 198, 200). The mechanisms underlying TCR and cytokine synergy in MAIT cells and iNKT cells remain to be established. However, TCR signaling-induced histone acetylation at the *IFNG* locus (79), as discussed above, may play a role in iNKT cells.

As MAIT cells are hyporesponsive to stimulation *via* the TCR alone, synergy between TCR and cytokine signaling likely plays a key role in robust MAIT cell activation *in vivo* (30, 85, 169). This is supported by a recent study showing that both metabolites from the riboflavin biosynthesis pathway and costimulatory signals are required for MAIT cell accumulation *in vivo* following bacterial lung infection (32). For iNKT cells, innate signaling from IL-12 provides the dominant signal for activation in many bacterial infections, even in the presence of cognate microbial lipid antigens (78). Cytokine signaling might also dominate in activating MAIT cells. TCR stimulation is insufficient to induce sustained MAIT cell effector responses *in vitro* (30), and in certain bacterial settings, blocking cytokines, as opposed to MR1, has a greater impact on MAIT cell activation (195, 201). Moreover, a central role for cytokines would potentially explain why MAIT cells are not constitutively activated by TCR-dependent sensing of commensal bacteria. However, the relative role of TCR- and cytokine-mediated activation will be influenced by many factors, including the nature of the APC. MAIT cell activation in response to *Streptococcus pneumoniae in vitro* is driven purely by cytokines when THP-1 cells are used as the APC, whereas in the presence of monocyte-derived macrophages, activation is driven by both MR1 and cytokines (201).

In addition to TCR- and cytokine-dependent activation, MAIT cells could potentially be activated *via* NK cell receptors, some of which can directly activate iNKT cells (**Figure 2B**). For example, NKG2D engagement triggers degranulation of human CD4<sup>−</sup> iNKT cells (202), and mouse iNKT cells produce IFN-γ following crosslinking of NK1.1 (203), although the significance of TCR- and cytokine-independent activation *in vivo* remains unknown. In contrast to iNKT cells, direct NK cell receptor-mediated activation of MAIT cells has yet to be reported. Despite high expression of NKG2D (14), the cytotoxic activity of MAIT cells against *E. coli*-infected HeLa cells (overexpressing MR1) *in vitro* is unaffected by the presence of anti-NKG2D antibody (204). Nevertheless, reports have suggested both costimulatory and coinhibitory roles for the NK cell receptor CD161 on MAIT cells (80, 204). Similarly, CD161 can costimulate the activation of human iNKT cells (205).

#### Regulation

Costimulatory and coinhibitory molecules, including CD28, ICOS, OX40, and PD-1, have an important role in regulating iNKT cell activation and effector function *in vitro* and *in vivo* (**Figure 2A**) (206, 207). In addition to simply augmenting or dampening the magnitude of responses, engagement of specific costimulatory receptors on iNKT cells has been shown to skew the induced effector response (206, 207). For example, blocking the interaction of CD28 with CD86 more strongly inhibits IFN-γ production compared with IL-4 production by murine iNKT cells *in vitro* in response to αGalCer, thus promoting a Th2-biased response (208). While MAIT cells express various costimulatory and coinhibitory molecules (209), we have limited understanding of their functional role, although a few have been shown to modulate MAIT cell effector function *in vitro* (85, 210–212). For example, costimulation with αCD28 augments MAIT cell cytokine production and proliferation upon αCD3 stimulation (85). By contrast, the coinhibitory molecule PD-1 is upregulated on MAIT cells in several bacterial and viral infections, including hepatitis C (213) and tuberculosis (TB) (211), and PD-1 blockade leads to enhanced IFN-γ production by MAIT cells from active TB patients in response to *in vitro* stimulation with live bacillus Calmette–Guérin (211). Nevertheless, the role of costimulatory and coinhibitory molecules in modulating MAIT cell activation *in vivo*, and their capacity to differentially skew the MAIT cell effector response, has yet to be investigated.

In addition to the expression of coinhibitory molecules, two additional mechanisms may function to negatively regulate MAIT cell activation and/or to switch off MAIT cell effector functions upon resolution of infection or inflammation. First, MR1 binding antagonist ligands, including 6-FP (6-formylpterin) and Ac-6-FP (acetyl-6-formylpterin), competitively inhibit MAIT cell activation *in vitro* in response to synthetic agonist ligand (**Table 1**) (58, 59, 214). Furthermore, intranasal administration of Ac-6-FP can inhibit MAIT cell accumulation in a dose-dependent manner in the lungs of mice following intranasal administration of 5-OP-RU and a TLR agonist (riboflavin-deficient *Salmonella enterica* serovar Typhimurium) (59). Second, MAIT cells exhibit a proapoptotic phenotype driven by PLZF, akin to iNKT cells (215). This sensitivity to activation-induced cell death may function to restrain the MAIT cell effector response in order to minimize immunopathology. However, the role of these mechanisms in regulating MAIT cell activation in physiological settings is currently unknown.

#### Therapeutic Modulation

Although microbe-driven TCR-dependent MAIT cell activation requires expression of the riboflavin biosynthesis pathway (7), non-riboflavin activating and inhibitory MR1 ligands have recently been identified (59). Keller et al. used various *in silico* approaches to screen libraries of small organic molecules and drugs for potential MR1 ligands (59). Identified targets were then tested for their ability to modulate MR1 expression and MAIT cell activation. Metabolites of the drug diclofenac were found to activate certain Jurkat MAIT cell lines *in vitro*, depending on their TCRβ chain usage, whereas 3-F-SA (3-formylsalicylic acid) could inhibit 5-OP-RU-dependent MAIT cell activation *in vitro* and *in vivo* (**Table 1**). While these findings imply that common therapeutics might inadvertently affect human MAIT cell activity *in vivo*, they also indicate the potential to design drugs to modulate MAIT cell function. Whether iNKT cell activation and function might also be influenced by common drugs is unknown. However, as lipid-based drug delivery systems are increasingly employed to improve oral bioavailability (216), it will be important to investigate the impact of such lipid-based formulations on iNKT cell function.

To summarize, diverse activation mechanisms are available for MAIT cells that are largely shared with iNKT cells. However, whereas iNKT cells can be activated by self-ligands in combination with cytokines, endogenous ligands for the MAIT cell TCR are yet to be identified. Due to their capacity for cytokine-mediated activation, MAIT cells and iNKT cells can play key roles in diverse infectious, as well as inflammatory and malignant diseases, even in the absence of their cognate microbial antigens. The relative importance of TCR- and cytokine-dependent activation *in vivo* is likely to be context-dependent and influenced by the nature of the pathogen and its TCR/TLR ligands, the type of activated APCs, the availability of costimulatory/coinhibitory molecules, and the stage of infection or inflammation. Nevertheless, the role for cytokines appears more important than for conventional T cell activation and may even dominate in MAIT cell and iNKT cell activation in some settings, despite the presence of microbial TCR ligands.

## EFFECTOR FUNCTIONS

#### Cytokine Production

Upon activation, MAIT cells rapidly produce cytokines such as IFN-γ, TNF-α, and IL-17 (**Table 1**) (14). MAIT cells typically secrete a limited range of pro-inflammatory cytokines. By contrast, iNKT cells secrete a huge variety of both pro- and antiinflammatory cytokines, including IL-4, IFN-γ, IL-10, and IL-17 (**Table 1**) (9).

The factors that govern which cytokines MAIT cells produce under different stimulatory conditions remain poorly characterized. Human MAIT cells secrete IFN-γ following both TCR- and cytokine-dependent activation, whereas TNF-α production is more contingent on TCR signaling (**Figure 2**) (14, 30). Though all human MAIT cells express RORγt, in addition to other type 17-associated molecules, such as CCR6 and the IL-23 receptor (14, 70), IL-17 production *ex vivo* is usually only detected following PMA and ionomycin stimulation (14, 70), and not upon TCR or cytokine stimulation alone (14, 30, 85, 169). However, certain cytokines, such as IL-7 or IL-23 + IL-1β, can induce IL-17 production when combined with a TCR stimulus (169). In addition, MAIT cells may exhibit functional plasticity driven by cytokines, as has been demonstrated *in vitro*. For example, CD161hiCD8α+ T cells (predominantly MAIT cells) develop a more Tc1-like phenotype following culture with αCD3/αCD28 + IL-12 for 14 days (85).

Similar to MAIT cells, the profile of cytokines produced by iNKT cells varies under different stimulation conditions and there is limited knowledge regarding the factors that regulate this. iNKT cells secrete both IFN-γ and IL-4 upon TCR stimulation with microbial antigens (**Figure 2A**) (217). By contrast, cytokine-dependent activation by viruses or TLR ligands stimulates predominantly IFN-γ production and not IL-4 (**Figure 2B**) (217). Chemically modified analogs of the iNKT cell ligand αGalCer have been identified that induce qualitatively different cytokine responses *in vitro* and *in vivo*, specifically Th1-biased, Th2-biased, or mixed Th1/Th2 responses (218–220). Although the exact mechanisms for this remain unknown, the stability of ternary TCR-CD1d/glycolipid complexes appears to have an important role, with prolonged TCR stimulation favoring Th1 biased responses (218, 220). As discussed, activation of different costimulatory pathways can also skew the iNKT cell response to antigen stimulation (206, 207). Therefore, the type of lipid antigens and costimulatory molecules available to activate iNKT cells *in vivo* will alter the nature of the cytokine response. Whether different MAIT cell ligands/chemical modifications of MAIT cell ligands, or costimulatory pathways, can skew MAIT cell cytokine production, remains to be investigated. As the range of cytokines produced by MAIT cells is less functionally diverse than that of iNKT cells, the capacity to drastically alter the overall immune response by skewing MAIT cell cytokine production may be more limited than with iNKT cells.

Human MAIT cells in different tissues exhibit differential cytokine production. MAIT cells in the female genital mucosa appear skewed toward type 17 functions, secreting increased IL-17 and IL-22, and decreased IFN-γ and TNF-α, compared with blood MAIT cells (26). IL-22-secreting MAIT cells are also enriched in fetal small intestine (155), while MAIT cells in adipose tissue exhibit the unique capacity to secrete IL-10 (20). Different iNKT cell subsets preferentially localize to certain tissues in mice (18, 87). As a result, challenge with αGalCer induces distinct cytokine responses depending on the route of antigen delivery and thus the nature of the iNKT cell subsets activated (18). Whether variation in MAIT cell cytokine production across tissues can similarly be explained by the tissue-specific enrichment of different MAIT cell subsets is currently unknown. Of interest, IL-10-producing NKT10 cells preferentially localize to adipose tissue (87, 221), suggesting that the adipose tissueenriched IL-10-producing MAIT cells in humans could comprise a distinct subset (20).

#### Cytotoxic Activity

In addition to cytokine secretion, MAIT cells and iNKT cells display direct cytotoxic activity. MAIT cell killing is mediated *via* the Perforin/Granzyme pathway and is independent of Fas/FasL and NKG2D (91, 161, 204). While their cytotoxic capacity (i.e., Perforin and Granzyme expression) is enhanced upon activation *via* the TCR and/or with cytokines, target cell killing is MR1 dependent (**Figure 2**) (91, 161, 204). In contrast to MAIT cells, iNKT cell killing can be mediated *via* both Perforin/Granzymeand Fas/FasL-dependent pathways (**Figure 2**) (62, 98, 222). The cytotoxic capacity of iNKT cells varies between subsets. Human CD4− iNKT cells show increased expression of cytotoxic molecules and superior cytotoxic activity compared with CD4<sup>+</sup> cells (62, 223, 224). Whether CD4<sup>+</sup>, CD8<sup>+</sup>, and DN MAIT cells show differences in cytotoxic activity is currently unknown. Akin to MAIT cells, iNKT cell cytotoxicity is largely dependent on CD1d and antigen, although alternative CD1d-independent mechanisms have been described (202, 222, 224). In particular, human CD4<sup>−</sup> iNKT cells can kill targets through a CD1d-independent NKG2D-dependent pathway (202). MR1-independent pathways for MAIT cell killing have yet to be identified.

#### Immune Interactions

While knowledge of MAIT cell crosstalk with other immune cell subsets remains relatively limited, recent studies have identified important interactions with a number of immune cell types, including DCs, B cells, and NK cells. Through contact with myeloid cells, MAIT cells appear to have important immune regulatory functions. For example, upon antigen-specific activation, MAIT cells upregulate CD40L and induce CD40-dependent DC maturation, and in synergy with TLR ligands, promote the secretion of IL-12 (225). DC-derived IL-12 can subsequently enhance MAIT cell activation. MAIT cells can also influence monocyte differentiation *in vivo*. MR1<sup>−</sup>/<sup>−</sup> mice show enhanced susceptibility to pulmonary infection with *Francisella tularensis* live vaccine strain and delayed bacterial clearance (44). In this setting, MAIT cells promote early GM-CSF production in the lungs, resulting in the differentiation of inflammatory monocytes into DCs and the recruitment of activated CD4<sup>+</sup> T cells into the lungs (44, 226). In addition to their effects on myeloid populations, MAIT cells can provide noncognate B cell help *in vitro* (227). In response to TCR-dependent or TCR- and cytokine-dependent activation, MAIT cells secrete factors that act on B cells to promote the differentiation of memory cells into plasmablasts and to increase antibody production (227). In these experiments, TCR stimulation was essential for the capacity of MAIT cells to provide B cell help (227). Finally, in the context of whole blood, activated MAIT cells promote NK cell transactivation in an MR1- and IL-18-dependent manner (225). Although MAIT cells can facilitate monocyte differentiation *in vivo*, whether the described crosstalk with DCs, B cells, and NK cells, occurs *in vivo*, and exactly where such interactions would take place, remains to be determined.

More is known about the immune interactions of iNKT cells. Through crosstalk with an array of immune cell types, iNKT cells can profoundly influence the nature and quality of both innate and adaptive immunity. iNKT cells engage in similar interactions to those described for MAIT cells; however, differences can be identified in the requirements for MAIT/iNKT cell activation in these settings and in the downstream effects on the immune response. Human iNKT cell clones drive monocyte differentiation in a CD1d-dependent manner (228). By contrast, MAIT cellmediated monocyte differentiation is MR1-independent, at least for transgenic MAIT cells *in vitro* (226). Analogous to MAIT cells, bidirectional interaction between iNKT cells and DCs leads to DC maturation and NK cell transactivation, but also results in increased peptide-specific CD4<sup>+</sup> and CD8<sup>+</sup> T cell responses (229– 232). A similar function has yet to be described for MAIT cells. In addition to non-cognate B cell help, iNKT cells can provide cognate B cell help (233). In some settings, help is provided by a dedicated subset of iNKT cells, known as follicular helper NKT (NKTFH) cells (93, 94, 234). Other specialized iNKT cell subsets also engage in key immune interactions. For example, mouse and human Foxp3<sup>+</sup> invariant regulatory NKT (iNKTreg) cells have been shown to suppress naïve T cell proliferation *in vitro* (95, 96). Whether comparable functions can be performed by specialized MAIT cell subsets is currently unknown.

In summary, iNKT cells and MAIT cells rapidly produce cytokines, exhibit cytotoxic activity, and can influence the function of both innate and adaptive immune cell populations. MAIT cells typically produce pro-inflammatory cytokines, whereas iNKT cells secrete vast amounts of both pro- and antiinflammatory cytokines. While immunoregulation *via* cytokine secretion is the dominant function of iNKT cells, the relative importance of cytokine secretion versus cytotoxic activity for MAIT cells, is currently unknown. iNKT cells profoundly influence the immune response through their crosstalk with other immune cell subsets, and limited studies reveal similar interactions for MAIT cells. Given their abundance in humans and their rapid effector function, MAIT cells are likely key orchestrators of innate and adaptive immunity in humans.

# SUBSETS

Although human MAIT cells are thought to be largely homogeneous in phenotype and function, recent findings suggest that they may be more diverse than currently appreciated. Lepore et al. identified subpopulations of human MAIT cells with distinct cytokine secretion profiles (50), although whether these represent separate lineages is currently unknown. Human MAIT cells also show heterogeneous expression of certain NK cell-associated receptors, including CD56 and CD84, the expression of which correlates with their functional response to cytokine stimulation (209). Finally, MAIT cells in certain tissues exhibit altered cytokine production. For example, MAIT cells are skewed toward a Th17-like phenotype in female genital tract and express lower levels of the transcription factors PLZF and Eomes compared with peripheral blood MAIT cells (26), suggesting that they may comprise a distinct MAIT cell subset.

Mucosal-associated invariant T cells in humans can be CD4<sup>+</sup>, CD8<sup>+</sup>, or DN (**Table 1**) (16). Coreceptor expression appears to have little practical significance, particularly for CD8<sup>+</sup> and DN MAIT cells (43, 235). However, limited studies have characterized the phenotype and function of the minor CD4<sup>+</sup> population. Moreover, surrogate methods for MAIT cell identification were recently shown to poorly define CD4<sup>+</sup> MAIT cells (43, 235), and thus findings from previous studies that used such approaches, require validation. Initial studies using the MR1/5-OP-RU tetramer indicate some disparity between CD4<sup>+</sup> and CD8<sup>+</sup>/DN subsets, including differential expression of certain transcription factors (PLZF, Eomes), chemokine receptors (CCR4, CXCR6), adhesion molecules (CD56), and NK cell receptors (NKG2A, NKG2D) (43, 235). Nevertheless, the role of the CD4 coreceptor and whether CD4<sup>+</sup> MAIT cells comprise a distinct subset with specific functionality is currently unknown. Unlike MAIT cells, human CD4<sup>+</sup>, CD8<sup>+</sup>, and DN iNKT cells show clear phenotypic, functional, and transcriptional differences (15, 61, 62, 84, 202, 224, 236–240). However, given the considerable heterogeneity within each of these populations (15), it is unlikely that they represent genuine iNKT cell subsets.

The evidence for MAIT cell subsets in mice is more convincing. Two subsets of MAIT cells, a major RORγt <sup>+</sup>IL-17<sup>+</sup> population, and a smaller T-bet<sup>+</sup>IFN-γ+ subset have been identified in the thymus, spleen, and lung (**Figure 3**) (19, 102). These subsets can show plasticity *in vivo*. Following intranasal infection with *Salmonella. enterica* serovar Typhimurium, RORγt <sup>+</sup> MAIT cells in the lung upregulate T-bet expression to become RORγt <sup>+</sup>T-bet<sup>+</sup> cells that can secrete either IL-17 or IFN-γ, though the majority produce IL-17 (32). In contrast to mice, human thymic and peripheral blood MAIT cells coexpress RORγt and T-bet, although T-bet expression can similarly increase upon activation (85, 91, 102, 167). Furthermore, the majority of human MAIT cells express IFN-γ and a smaller fraction produce IL-17, while some secrete both cytokines (**Figure 3**), highlighting important differences between mouse and human MAIT cells (14).

Similar to MAIT cells, iNKT cells in mice differentiate into distinct subsets within the thymus. These mirror conventional CD4<sup>+</sup> T helper cell subsets in their expression of master transcription factors and cytokines, namely PLZFloT-bet<sup>+</sup>RORγt <sup>−</sup> NKT1 cells that secrete IFN-γ, PLZFhiT-bet<sup>−</sup>RORγt <sup>−</sup> NKT2 cells that secrete IL-4 and other Th2 cytokines, and PLZFintT-bet<sup>−</sup>RORγt <sup>+</sup> NKT17 cells that secrete IL-17A and IL-22 (**Figure 3**) (86, 241). NKT1, NKT2, and NKT17 cells show highly divergent transcriptional programs and differ in their expression of chemokine receptors, NK cell receptors, cytotoxic molecules, and cell cycle-related genes (146, 147, 242). T-bet<sup>+</sup>IFN-γ+ and RORγt <sup>+</sup>IL-17<sup>+</sup> MAIT cells in mice could represent "MAIT1" and "MAIT17" subsets (19, 102), akin to NKT1 and NKT17 cells, respectively. By contrast, no "MAIT2" subset comparable to NKT2 cells has been identified, and MAIT cells in mice and humans show little to no production of Th2 cytokines (14, 19).

In addition to the major iNKT cell subsets that develop in the thymus, a number of minor, highly specialized subsets have been identified in mouse peripheral tissues and/or lymphoid organs, but not in the thymus. These include NKTFH cells (93, 234), iNKTreg cells (95), and IL-10-secreting NKT10 cells (87, 221). However, it is important to note that IL-10 can also be produced by activated NKT1, NKT2, and NKT17 cells following stimulation with αGalCer (243). Human iNKT cells with a similar phenotype and/ or function to NKTFH cells (93), iNKTreg cells (95, 96), and NKT10 cells (87) have been reported (**Figure 3**). By contrast, comparable MAIT cell subsets have not been defined in mice or humans, although a high proportion of MAIT cells secrete IL-10 in human adipose tissue (**Figure 3**) (20). As NKT10 cells are enriched in mouse adipose tissue (87, 221), IL-10-secreting MAIT cells could represent a dedicated "MAIT10" subset. Whether MAIT cells can have regulatory or follicular helper-type functions is currently unknown.

In brief, two distinct MAIT cell subsets, analogous to NKT1 and NKT17 cells, are present in mice. Whereas the MAIT cell population is biased toward RORγt/IL-17 expression in C57BL/6 mice, iNKT cells favor the expression of T-bet/IFN-γ. This could suggest functional segregation between innate-like T cell populations in mice, although depending on the tissue and mouse strain, NKT17 cells might still be more abundant than MAIT cells. In comparison with mouse MAIT cells, human MAIT cells appear more homogeneous and exhibit a mixed Th1/Th17 phenotype, although there is evidence for some phenotypic and functional diversity. Moreover, human MAIT cells preferentially secrete IFN-γ as opposed to IL-17. Multiple iNKT cell subsets are present in both mice and humans, although they are better defined in mice. Compared with MAIT cells, iNKT cells appear more functionally diverse, although it is possible that additional MAIT cell subsets remain to be identified.

## AVENUES FOR FUTURE MAIT CELL RESEARCH

With the recent generation of MR1 tetramers (12, 13), it is now possible to detect MAIT cells in wild-type mice (19), as well as in human settings where MAIT cell frequency is low, for example in the thymus (102). Consequently, MAIT cells are now being studied in an increasingly wide variety of settings, including in numerous human diseases and animal disease models. Undoubtedly, this will lead to a greatly improved understanding of the role of MAIT cells in both health and disease.

Compared with iNKT cells, our knowledge of MAIT cell phenotype, development, regulation, and function remains limited and there are many important questions that need to be

Figure 3 | Functional capacity of mucosal-associated invariant T (MAIT) cells and invariant natural killer T (iNKT) cells. Where subsets of MAIT or iNKT cells have been defined, characteristic cytokines, transcription factors, and/or surface markers, are illustrated. MAIT cells and iNKT cells exhibit overlapping functions, although a wider range of functions have been described for iNKT cells. In mice, distinct type 1 and type 17 MAIT and iNKT cell subsets have been identified. By contrast, human MAIT cells exhibit a mixed type 1/type 17 phenotype. Human iNKT cells secrete IFN-γ and IL-17 (only *in vitro* under pro-inflammatory conditions), but whether these cytokines are produced by distinct subsets, remains to be established. Unlike MAIT cells, iNKT cells also show type 2 functions, such as IL-4 secretion. In mice, IL-10-producing iNKT cells comprise a distinct subset with altered transcription factor expression. Human MAIT cells and iNKT cells can produce IL-10, and human IL-10-producing MAIT cells are enriched in adipose tissue, similar to mouse NKT10 cells. However, whether these IL-10-producing populations comprise distinct subsets, is currently unknown. Finally, multiple specialized subsets of iNKT cells have been identified in mice, including NKTFH cells and iNKTreg cells. Human iNKT cells with similar phenotypes and/or functions have also been identified (NKTreg only *in vitro*), but analogous populations have yet to be described for MAIT cells.

addressed. The search for novel MAIT cell ligands, both for their selection in the thymus, and their activation in the periphery, is a particularly active area of investigation. It is currently unknown whether MAIT cells can recognize endogenous antigens, analogous to iNKT cells, and if so, how these might influence MAIT cell development and function *in vivo*.

There are many gaps in our understanding of MAIT cell development, including the signals required for positive and potentially negative selection in the thymus; the transcriptional and epigenetic regulation of MAIT cell differentiation; and the timing and location of MAIT cell maturation. As MAIT cells appear to be mature in fetal tissues (155), but naïve in cord and neonatal blood (16, 47), they may comprise predominantly tissue-resident populations, similar to iNKT cells (28, 29, 221). This could be addressed in parabiosis studies. Moreover, it will be important to determine the signals governing peripheral MAIT cell maturation and the maintenance of homeostasis. In addition to IL-18 and IL-23 (77, 102), it is worth investigating the role of IL-7 and IL-15, given MAIT cell responsiveness to these cytokines (30, 91, 167–169) and their function in iNKT cell homeostasis (73, 75, 76, 166, 244). Whether MR1 is required for peripheral MAIT cell survival and maintenance *in vivo* is currently unknown, although CD1d does not appear to be essential for the homeostasis of iNKT cells (75, 142, 166, 170).

While MAIT cells can be activated through the TCR or with cytokines, or by a combination of both (11), which of these activation mechanisms predominates *in vivo* remains to be determined. iNKT cell activation appears to be dominantly driven by cytokines, even in settings where microbial ligands are present (78), although their sensitivity to cytokines alone may require prior TCR priming (79). It is hypothesized that cytokines might also be key for MAIT cell activation, especially given the indiscriminate expression of their ligands in both commensal and pathogenic bacteria (7, 17), and their relative hyporesponsiveness to TCR stimulation (30, 85, 169). Despite expression of various NK cell receptors, costimulatory molecules, and immune checkpoint molecules (14, 209), how stimulation of these receptors modulates MAIT cell activation, is essentially unknown.

Similar to iNKT cells (86), transcriptionally and functionally distinct MAIT cell subsets have been identified in mice (19), although these require further characterization. Human MAIT cells exhibit a relatively uniform phenotype, and coexpress key transcription factors, suggesting a single MAIT cell subset (14, 91). However, recent studies indicate previously unappreciated phenotypic and functional heterogeneity in blood MAIT cells (14, 50, 209), as well as skewed or unique functions in certain tissues (20, 26, 155). Thus, the existence of specialized MAIT cell subsets remains an open question. In addition, it is unclear to what extent MAIT cells can show functional plasticity driven by environmental factors. It is likely that the presence of subsets and the ability of MAIT cells to display functional plasticity both contribute to MAIT cell diversity.

Recently, a number of studies have increased our knowledge of MAIT cell interactions with other immune cell populations (225–227), although our understanding of MAIT cell crosstalk remains much more limited than for iNKT cells (9). Furthermore, the role of MAIT cell interactions *in vivo* and their influence on the overall innate and adaptive immune response is largely unknown. As the interactions identified for MAIT cells resemble those of iNKT cells, it may be possible to use known iNKT cell interactions as a framework for the further investigation of MAIT cell crosstalk.

Though outside the scope of this review, we have little understanding of the role of MAIT cells in disease, despite their altered frequency and function in numerous infectious, inflammatory, and malignant diseases (1, 4, 245). Given that MR1 tetramers are now widely available, it is possible to address the role of MAIT cells *in vivo* in models of disease. Recently, MR1 ligands have been discovered among drugs and drug-like molecules (59). This indicates the potential for the future development of therapeutics to modulate MAIT cell function (59), akin to the development of αGalCer as a stand-alone therapy or vaccine adjuvant (246).

#### CONCLUSION

Invariant natural killer T cells have been more extensively studied than MAIT cells due to their higher frequency in mice, the earlier

#### REFERENCES


development of tetramers for their specific identification, and the earlier generation of relevant transgenic mouse models. However, the MAIT cell field is growing rapidly, due to the recent development of the MR1 tetramer and hence the possibility to examine MAIT cells in wild-type mice. Comparison of MAIT cells to iNKT cells has in many cases identified shared biology. However, there are also instances in which comparison of MAIT cells to iNKT cells has revealed distinct biology. Therefore, while iNKT cell research provides a useful framework for the study of MAIT cells, it is important that both populations are studied individually, and findings from the iNKT cell field are not presumed to also apply to MAIT cells. Given their abundance in humans, their capacity for rapid cytokine production in response to TCR and/or cytokine stimulation, and their interactions with other immune cell populations, MAIT cells are likely key players in the immune system, both in health and disease. In support of this, they show altered phenotype and function in numerous human diseases, and exhibit protective or deleterious roles in mouse models of infection or inflammation. Consequently, MAIT cells represent an attractive target for therapeutic manipulation, especially considering their high frequency in humans and their recognition of a monomorphic MHC-like molecule. To realize this goal, further research is necessary to develop a greater understanding of MAIT cell development, function, and regulation, and their specific roles in disease. We should continue to leverage our accumulated knowledge of iNKT cell biology as a platform to more completely understand the unique, and shared, biology of MAIT cells. In doing so, such investigations will likewise enhance our understanding of iNKT cell biology.

# AUTHOR CONTRIBUTIONS

LG wrote the bulk of the manuscript and contributed to the figures. PK contributed to the planning, editing, and scope of the review. NP edited and revised the manuscript, and created the figures.

#### ACKNOWLEDGMENTS

We would like to thank Dr. Mariolina Salio and Dr. Chris Willberg for critical reading of the manuscript.

#### FUNDING

LG is supported by a Wellcome Trust PhD Studentship [109028/Z/15/Z]. PK is funded by the Wellcome Trust [WT109965MA]; the Medical Research Council (STOP-HCV); and an NIHR Senior Fellowship; and the NIHR Biomedical Research Centre (Oxford). NP is supported by an Oxford-UCB Postdoctoral Fellowship.


in the small intestine. *Mucosal Immunol* (2014) 7:1233–43. doi:10.1038/mi. 2014.13


and rapidly produce IL-17 upon receptor ligation in an IL-6-independent fashion. *J Immunol* (2008) 180:5167–71. doi:10.4049/jimmunol.180.8.5167


*Eur J Immunol* (2000) 30:620–5. doi:10.1002/1521-4141(200002)30:2<620:: AID-IMMU620>3.0.CO;2-4


antibody production, and T cell stimulation by B cells in vitro. *J Immunol* (2013) 191:1666–76. doi:10.4049/jimmunol.1202223


**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 Garner, Klenerman and Provine. 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.*

# How Lipid-Specific T Cells Become effectors: The Differentiation of iNKT Subsets

*Haiguang Wang and Kristin A. Hogquist\**

*Department of Laboratory Medicine and Pathology, Center for Immunology, University of Minnesota, Minneapolis, MN, United States*

In contrast to peptide-recognizing T cells, invariant natural killer T (iNKT) cells express a semi-invariant T cell receptor that specifically recognizes self- or foreign-lipids presented by CD1d molecules. There are three major functionally distinct effector states for iNKT cells. Owning to these innate-like effector states, iNKT cells have been implicated in early protective immunity against pathogens. Yet, growing evidence suggests that iNKT cells play a role in tissue homeostasis as well. In this review, we discuss current knowledge about the underlying mechanisms that regulate the effector states of iNKT subsets, with a highlight on the roles of a variety of transcription factors and describe how each subset influences different facets of thymus homeostasis.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Paolo Dellabona, San Raffaele Scientific Institute (IRCCS), Italy Sebastian Joyce, Vanderbilt University, United States*

> *\*Correspondence: Kristin A. Hogquist hogqu001@umn.edu*

#### *Specialty section:*

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

*Received: 01 May 2018 Accepted: 12 June 2018 Published: 26 June 2018*

#### *Citation:*

*Wang H and Hogquist KA (2018) How Lipid-Specific T Cells Become Effectors: The Differentiation of iNKT Subsets. Front. Immunol. 9:1450. doi: 10.3389/fimmu.2018.01450*

Keywords: invariant natural killer T cell, subsets, development, lipid, thymus, mucosal-associated invariant T cell

#### INTRODUCTION

Natural killer T cells (NKT) were named because they express T cell receptor (TCR)–CD3 complexes as well as the natural killer (NK) cell receptor NK1.1 (CD161) (1, 2). Later, research discovered that NKT cells express a semi-invariant TCR, characterized by a Vα14-Jα18 TCRα chain coupled with a limited Vβ repertoire (Vβ2, Vβ7, or Vβ8.2) in mice, and an invariant Vα24-Jα18 paired with Vβ11 in humans (3, 4). Owning to this semi-invariant TCR, invariant natural killer T (iNKT) cells recognize self- and foreign-lipid antigens presented by the CD1d molecule and could be specifically detected using CD1d tetramer loaded with a cognate lipid antigen.

iNKT cells originate in the thymus, but in contrast to the conventional peptide specific CD4<sup>+</sup> or CD8<sup>+</sup> T cells, which are positively selected by cortical thymic epithelial cells, the positive selection of iNKT cells solely relies on the interactions among cortical double-positive (DP) thymocytes (5–7). DP thymocytes expressing the rearranged Vα14-Jα18 TCR recognize high-affinity lipid antigens presented by CD1d molecules on neighboring DP thymocytes (4). iNKT cells highly express the transcription factor promyelocytic leukemia zinc finger protein PLZF (*zbtb46*), which is essential for their effector program (8, 9), for specifying the tissueresident properties of iNKT cells, and for their ability to produce cytokines early after stimulation (8–10).

It has been realized that iNKT cells are a heterogenous population, and recent evidence from various groups suggest that there are three major functional iNKT subsets at steady state according to their expression of lineage-specific transcription factors and cytokine-producing potential. The three iNKT subsets are designated NKT1, NKT2, and NKT17, in analogy to the classical CD4 T helper lineages. NKT1 cells are PLZFlow T-bet<sup>+</sup> and produce both IFN-γ and low amounts of IL-4 after stimulation. They express NK1.1 and other NK receptors and represent the subset that "NKT" cells were named after. NKT2 and NKT17 cells, in contrast, do not express NK1.1. NKT2 cells are PLZFhigh and produce high amounts of IL-4 at steady state and after stimulation. NKT17 cells are PLZFintermediate ROR-γt + and produce IL-17 after stimulation (11). Through intra-thymic transfer and fetal thymic organ culture (FTOC), previous studies demonstrated that each iNKT subset (NKT1, NKT2, and NKT17) is terminally differentiated; i.e., do not give rise to other cell subsets (11–13). iNKT cells play diverse roles in immunity due in part to the existence of these three functional subsets. The subsets produce distinct cytokines and reside in distinct tissues. With accumulating knowledge regarding the biology of iNKT cells, in this review, we summarize recent advances in the development and differentiation of iNKT subsets, as well as their role in maintaining the immune homeostasis.

## THE DEVELOPMENT AND DIFFERENTIATION OF iNKT CELLS

#### Initial Positive Selection

Like the conventional CD4+ and CD8+ T cells, iNKT cells originate from precursors undergoing TCR gene rearrangement in the thymus. A lineage tracing study using transgenic mice (RORγt–Cre × ROSA26lsl-EGFP) in which ROR-γt triggers permanent expression of green fluorescent protein (GFP) confirmed that iNKT cells were derived from ROR-γt + DP thymocytes in the thymus (5) (**Figure 1** "Selection"), while a minor population could arise from DN thymocytes bypassing DP stage (14). Moreover, ROR-γt itself is essential for iNKT cell generation, in that, it supports DP survival through regulating Bcl-xL expression, allowing for optimal rearrangement of Vα14-Jα18 TCR chains

maintain potential to produce IFN-γ or IL-17, respectively, after stimulation.

(15). Similarly, an E protein transcription factor, HEB, promotes survival of DP thymocytes through regulating both ROR-γt and Bcl-xL expression, which opens the window of time to allow distal Jα rearrangement (16). Downstream of the initial selection of DP thymocytes, c-Myc has been shown to control the maturation of iNKT cells (17, 18). Moreover, c-Myb has also been shown to play a central role in this process, as it supports long half-life of DP thymocytes to allow Vα14 to Jα rearrangement (19).

Immediate post-selection precursor iNKT cells are characterized as CD1d tetramer<sup>+</sup> CD44<sup>−</sup> CD24<sup>+</sup> CD69<sup>+</sup>, termed as "stage 0" iNKT cells. The strong TCR signal during iNKT selection was directly demonstrated using a reporter mouse in which a GFP cassette was inserted in the Nur77 locus (an immediate-early gene upregulated by TCR stimulation), wherein the GFP level indicates the TCR signal strength (20). In these mice, stage 0 iNKT cells express a high level of GFP indicating they received strong TCR signal during selection (20). Beside this strong TCR signal, the development of iNKT cells also relies on a "second signal" generated through homotypic interactions between signaling lymphocyte activation molecule family (SLAMF) receptors, SLAMF1 and SLAMF2, expressed on the DP thymocytes (21). In addition to supporting a long half-life in DP thymocytes, c-Myb also promotes the expression of CD1d and SLAMFs, which are essential for positive selection of iNKT cells (19). Deficiency of c-Myb completely abrogates the generation of iNKT cells, as CD24<sup>+</sup> stage 0 iNKT cells were NOT detected (19).

Historically, in B6 mice, the maturation of iNKT cells beyond stage 0 was described as a stepwise linear model from stage 1 to 3 based on expression of CD44 and NK1.1. In this model, the stage 0 iNKT cells develop into CD24<sup>−</sup> CD44<sup>−</sup> NK1.1<sup>−</sup> stage 1 cells, then upregulate CD44 to become stage 2 cells, and finally acquire NK1.1 expression to become stage 3 cells in a linear fashion (22). This model fits some but not all the available data. For example, NKT17 cells were known to express CD44 but not NK1.1 (stage 2), never become NK1.1<sup>+</sup> (stage 3) (12). Alternatively, based on the expression of transcription factors, PLZF, Gata3, T-bet, and ROR-γt, CD24<sup>−</sup> iNKT cells could be very well categorized into three distinct subsets, NKT1, NKT2, and NKT17 as described above. Similar to NKT17 cells, intrathymic transfer of "stage 2" IL-4 producing NKT2 cells (IL-4<sup>+</sup> IL-17RB<sup>+</sup> CD4<sup>+</sup>) showed that they do not give rise to T-bet<sup>+</sup> NK1.1+ "stage 3" cells either (11). Therefore, a revised lineagediversification model for iNKT cell development, in which a common progenitor gives rise to the distinct lineages of NKT1, NKT2, and NKT17 cells (**Figure 1**) was suggested. We herein discuss the promoting and inhibitory factors for selection, specification, and differentiation of iNKT cells, which are summarized in **Table 1**.

#### Specification

Stage 0 iNKT cells arise from DP thymocytes in the thymic cortex (6). However, in CD1d tetramer-based immunofluorescence and histocytometric analysis, thymic iNKT subsets were found to be predominantly localized in the thymic medulla (23) (**Figure 1**). Consistent with this, the thymic medullary environment was reported to impact the functional maturation of iNKT cells (24). Therefore, the nature and localization of the common progenitor that directly gives rise to distinct subsets is unclear. Furthermore, the signals that drive their migration from cortex to medulla, as well as the medullary factors that control the differentiation of iNKT subsets has not yet been reported. A previous study demonstrated that the chemokine receptor CCR7, which responds to the chemokines CCL21, is important for thymocytes trafficking from the cortex to the medulla (25). Additionally, the number of iNKT cells was significantly reduced in CCR7<sup>−</sup>/<sup>−</sup> mice (26). Interestingly, single cell RNA-seq analysis of thymic iNKT cells suggested that PLZFhigh iNKT cells might comprise a progenitor population (27). Previous work showed IL-4<sup>−</sup> PLZFhigh iNKT cells could further differentiate into T-bet<sup>+</sup> NKT1 cells when sorted and intra-thymically transferred into thymus (11), suggesting they maintain precursor potential. Further analysis of this IL-4<sup>−</sup> PLZFhi iNKT cell population by RNA-seq and PCA analysis confirmed they have the least similarity to the three effector subset (28). Taken together, it could be inferred that CCR7<sup>+</sup> cells within


PLZFhigh iNKT cells might serve as the common progenitor for iNKT subsets (**Figure 1** "Specification").

#### Factors Involved in Specification and/or Effector Differentiation Cytokines

## *IL-15, TGF-β, and IL-25*

Numerous studies have demonstrated cytokines produced in the local environment play central roles in determining the differentiation of CD4<sup>+</sup> T helper subsets (Th1, Th2, and Th17) (29). Similarly, the differentiation of iNKT subsets is heavily influenced by different cytokine signals (**Figure 1** "Effector differentiation"). For instance, it's been shown that NKT1 cells highly express CD122 (IL2Rβ), and CD122-mediated IL-15 signaling is essential for the differentiation of NKT1 cells (30). Likewise, the absence of TGF-β signaling (CD4-Cre × TGF-βRIIflox/flox and CD4-Cre × Smad4flox/flox) led to complete loss of ROR-t<sup>+</sup> NKT17 cells (31). Both NKT2 and NKT17 cells express IL-17RB (IL-25 receptor), which was essential for the production of IL-13, IL-9, IL-10, and IL-17 after stimulation with αGalCer (13), demonstrating that the cytokine production of activated iNKT cells is influenced by a signal through this receptor. It was further shown that such effect was dependent on E4BP4, a transcription factor that regulates IL-10 and IL-13 production in CD4<sup>+</sup> T and iNKT cells (32, 33). Interestingly, E4BP4 seems to be upregulated in iNKT cells only after stimulation with IL-25 or αGalCer (13, 33), but not expressed by thymic or most peripheral iNKT cells in the steady state (except the adipose iNKT cells) (13, 33, 34). Though inferred by the data, iNKT subsets defined by transcription factor expression as NKT1, 2, and 17 were not directly evaluated in the study (13). Thus, whether the development of NKT2 and/or NKT17 cells is controlled by the IL-17RB/IL-25 axis remains or be defined. In a scenario where IL-25 signaling controls differentiation of NKT cells, it would be important to define the source of IL-25 in the thymus (**Figure 1**; **Table 1**). A recent study demonstrated that a type of specialized epithelial cells, called tuft cells, are the solely source of IL-25 in the gut (35). It will be interesting to check the thymus for this lineage of epithelial cells as well.

#### Transcription Factors

#### *Egr2*

Strong TCR signaling in stage 0 iNKT cells commits their fate to iNKT lineage, as it leads to elevated expression of the transcription factors Egr1 and Egr2, which influence further development of iNKT cells (36). In agreement with Egr2 directly binding the PLZF promoter, Egr1 and Egr2 together are critical for PLZF induction, which indicates that Egr1 and Egr2 may be upstream of PLZF in determining iNKT lineage fate (36). In addition, Egr2-deficient iNKT cells failed to express CD122, indicating that elevated Egr2 expression not only specifies iNKT lineage at an early stage but its sustained expression may also further influence differentiation of iNKT subsets (36). In addition, a cytoskeletal remodeling protein, P21-activated kinase 2 (Pak2) also influences the development of iNKT cells, especially NKT1 and NKT2 cells, possibly through regulation of the two critical transcription factors, Egr2 and PLZF (37).

#### *KLF Family Factors*

The transcription factor Kruppel-like factor 2 (KLF2) is essential for T cells egress from thymus and lymph node, because it's required for the expression of sphingosine 1 phosphate receptor type 1 (S1P1) in T cells (38). Unexpectedly, thymocytes in KLF2 deficient mice (CD4-Cre × KLF2flox/flox) displayed a memory phenotype (CD44high CXCR3<sup>+</sup> CD122<sup>+</sup>) that was shown to be an IL-4-dependent cell-nonautonomous effect (39). Furthermore, this effect was due to the expansion of IL-4-producing PLZFhigh T cells (mostly NKT2 cells), showing that KLF2 negatively regulates the differentiation of NKT2 cells (40). Another member of the Kruppel-like family, KLF13, plays the opposite role. KLF13 deficiency (KLF13<sup>−</sup>/<sup>−</sup>) led to a diminished population of IL-4 producing PLZFhigh iNKT cells (41).

#### *Hobit*

Though serving as an important factor that instructs the tissue retention program in iNKT cells and resident memory T cells (Trm) (42), the transcription factor Hobit was also shown to regulate the differentiation of iNKT cells (43). Hobit expression is high in CD44high NK1.1<sup>+</sup> iNKT cells (mostly NKT1 cells), but low in CD44low NK1.1− and CD44high NK1.1<sup>−</sup> iNKT cells (mostly NKT17 and NKT2 cells) (43). Accordingly, the number of CD44high NK1.1+ iNKT cells was significantly reduced in Hobitdeficient mice, while the abundance of CD44low NK1.1<sup>−</sup> and CD44high NK1.1<sup>−</sup> iNKT cells remained intact (43). Though the iNKT subsets were not distinguished in the study, it could be inferred from the data that Hobit promotes the differentiation and/or thymic retention of NKT1 cells.

*Lymphoid Enhancer Factor 1 (LEF1) and T Cell Factor 1 (TCF1)* The transcription factors LEF1 and TCF1 are essential for T cell development including early commitment to the T cell fate, transition from DN to the DP thymocytes, as well as following CD4/CD8 choice (44). The critical role of LEF1 and TCF1 in the differentiation of iNKT subsets has also been shown. Deletion of TCF1 at DP stage (CD4-Cre × Tcf7flox/flox) led to a severe defect in all three iNKT subsets (45). In addition, iNKT cell development was similarly impaired in absence of LEF1 (Vav-Cre × Lef1flox/flox) (46). LEF1 was required for the proliferation and survival of iNKT cells, especially the massive expansion after stage 0 (46). Interestingly, though it influenced the development of all three iNKT subsets, LEF1 showed a preference in promoting the differentiation of NKT2 cells (46).

#### Chromatin Modifiers

Epigenetic modifications also regulate development and differentiation of iNKT cells. The TET-family dioxygenases, TET1, TET2, and TET3, oxidize 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC), which is an important DNA modification critical for various biological processes (47–49). Simultaneous deletion of *Tet2* and *Tet3* resulted in uncontrolled TCR-mediated expansion of NKT17 cells through suppression of T-bet and ThPOK (50). Jarid2, a component of polycomb repressive complex 2 that methylates histone 3 lysine 27 (H3K27), is also involved in iNKT cells development. Upregulated after TCR stimulation, Jarid2 directly binds to the PLZF promotor as a transcriptional Wang and Hogquist iNKT Subset Differentiation

repressor. Therefore, deficiency of Jarid2 led to significant expansion of PLZFhigh NKT2 cells (51). In addition, the transcriptional repressor NKAP was shown to be required for the development of iNKT cells, as the iNKT development was completely abrogated at stage 0 in mice deficient of NKAP (CD4-Cre × NKAPflox/flox) (52). How NKAP regulates iNKT cell development is not clear, but its interaction with the histone deacetylase 3 (Hdac3) may be important, as NKAP is known to associate with Hdac3 and a similar defect of iNKT cells was observed in Hdac 3 conditional knockout mice (CD4-Cre × Hdac3flox/flox) (53). A recent report demonstrated that the H3K27me3 histone demethylase UTX is essential for iNKT cell development, especially the differentiation of NKT1 cells, as there was considerably fewer T-bet<sup>+</sup> NKT1 cells in UTX-deficient mice while NKT2 and NKT17 cells were not affected (54). UTX not only directly binds to the promoters of T-bet and CD122 genes but also influences the epigenetic landscape and transcription of PLZF-activated genes (54).

#### MicroRNAs (miRNAs)

MicroRNAs are small noncoding single-strand RNAs (~22 nt) that modulate the stability and transcriptional activities of messenger RNAs (mRNAs) and *via* this mechanism influence the transcriptomes of various cells, leading to further effects on cellular proliferation, apoptosis, lineage commitment, and differentiation (55). Perhaps not surprisingly, complete loss of mature iNKT cells was observed in mice lacking Dicer (CD4- Cre × Dicerflox/flox), which are incapable of generating functional miRNAs in T cells, thus demonstrating that miRNAs are essential for the development of iNKT cells (56). miR-181a is abundant in DP thymocytes and could augment TCR signaling strength *via* enhancing the basal activation of TCR signaling molecules, such as increased basal phosphorylation level of Lck and ERK (57). Deletion of miR-181a (miR-181a/b-1<sup>−</sup>/<sup>−</sup> mice) completely blocked iNKT cell development at the DP/Stage 0, which was presumably due to reduced responsiveness to TCR signals as exogenous agonistic ligand (αGalCer) could rescue iNKT cell generation (58). The miR-17–92 family cluster is also critical for the development of iNKT cells, in that absence of miRNAs of the miR-17–92 family cluster (triple knockout of three paralogs miR-17–92, miR-106a–363, and miR-106b–25 clusters) resulted in almost complete ablation of the three iNKT effector subsets (59). Excessive TGF-β signaling was seen in the remaining triple knockout iNKT cells, but it did not solely account for the impaired iNKT cell development, because deletion of TGF-βRII did not fully restore the hemostasis of iNKT cells (59). It was further found that the Let-7 family miRNAs, the most abundant family of miRNAs in mammals, tightly controls the differentiation of iNKT subsets (60, 61). Let-7 miRNAs are abundant in NKT1 cells while low in NKT2 and NKT17 cells, targeting *Zbtb46* mRNAs and inhibiting PLZF expression, therefore, directing iNKT cell differentiation into PLZFlow NKT1 lineage (61). Moreover, Lin28 inversely regulates Let-7 miRNAs, and Lin28 transgenic mice, which are practically deficient in Let-7 miRNAs, showed significantly increased NKT2 and NKT17 cells (61). miR-150 is expressed in lymphocytes (B, T, and NK cells) and has been implicated in their maturation. Correspondingly, miR-150 expression is expressed in iNKT cells after stage 0 (62, 63). In a mixed bone marrow chimera system, cell-intrinsic deficiency of miR-150 mildly affected iNKT cell development (62, 63), while overexpression of miR-150 substantially blocked maturation of iNKT cells beyond stage 0 (62). This suggests that fine-tuning of miR-150 level might be critical for iNKT cell development. Though the molecular pathway underlying this miR-150-dependent iNKT cell development is unclear, regulation of c-Myb by miR-150 could be involved (62, 63).

#### Cellular Protein Degradation System

While playing a central role in iNKT cell development, PLZF is initially induced in the stage 0 iNKT cells, and its expression can be regulated by the transcription factor Runx1 through direct binding to a critical enhancer of PLZF gene (64). Using Chip-Seq analysis, PLZF was shown to bind and regulate multiple genes, especially a broad set of immune effector genes expressed in iNKT cells (65). Beside directly regulating the expression of various genes, PLZF was also shown to transport an E3 ubiquitin ligase, cullin 3 (CUL3), from cytosol to nucleus, which would induce unique and essential ubiquitination patterns in iNKT cells (66). The number of iNKT cells was dramatically decreased in mice lacking CUL3 (CD4-Cre × CUL3flox/flox), further substantiating the importance of PLZF–CUL3 interaction in the development of iNKT cells (66). In line with its association with CUL3, PLZF has also been reported to interact with enhancer of zeste homolog 2 (Ezh2) methyltransferase (67). Moreover, Ezh2 directly methylates PLZF, causing its ubiquitination and subsequent degradation. Deletion of Ezh2 leads to sustained expression of PLZF and substantial expansion of PLZFhigh NK1.1<sup>−</sup> iNKT cells (mostly IL-4-producing NKT2 cells) (67).

#### Endogenous Selecting Lipid-Ligand and TCR Specificity

The generation of iNKT cells depends on recognition of lipid antigen presented by CD1d molecules on DP thymocytes. This antigen is most likely to be a self-lipid(s), because iNKT cells emerge early in life (6, 68), before stable colonization of commensal bacterial. Moreover, the phenotype and function of thymic and most peripheral iNKT cells (except pulmonary and intestinal iNKT cells) are normal in germ-free mice (69, 70). Regulated lipid metabolism in DP thymocytes is critical for thymic selection of iNKT cells, and the transcription factor Bcl11b plays a vital role in this process (71). Bcl11b-deficient (CD4-Cre × Bcl11bflox/flox) thymocytes showed deficient presentation of endogenous lipid antigens, dysregulated endo-lysosomal compartment, and alterations in genes involved in lipid metabolism (71). Moreover, in a mixed bone marrow chimera system, Bcl11b-deficient DP thymocytes (TCR-α−/<sup>−</sup>/CD4-Cre × Bcl11bflox/flox) failed to support selection of iNKT precursors in Bcl11b-sufficient DP thymocytes (β2m<sup>−</sup>/<sup>−</sup>/Bcl11b-Wt) (71). CD1d molecules can traffic between cell membrane and cytosolic organelles, surveying the endolysosomal compartment (72). A mouse model that expresses CD1d with a truncated cytoplasmic tail showed a severe defect in intracellular trafficking, and the number of iNKT cells was significantly reduced, suggesting the selection of iNKT cells relies on endosomal trafficking of CD1d molecule (73).

Though a great effort has been made to understand the stimulatory thymic self lipid(s), controversy remains, as reviewed elsewhere (22, 74, 75). Briefly, iGb3, an endogenous lysosomal glycosphingolipid, though thought to be presented by LPS-activated dendritic cells that activate iNKT cells (76), is unlikely to be a major selecting ligand for iNKT cells given that the development and function of iNKT cells are normal in isoglobotrihexosylceramide (iGb3)-deficient mice (77). Instead, glycosphingolipids (GSL) have been implicated in the development of iNKT cells as mice deficient of GSL-synthesizing enzyme glucosylceramide (GlcCer) synthase (GSC) in hematopoietic cells (Vav-Cre × GCSflox/flox) showed mild reduction of iNKT cells in both thymus and periphery (78). While stage 0 iNKT cells were not examined in the study, it remains unclear whether GSL are involved in the positive selection of iNKT cells. A recent report demonstrated the selecting ligands likely to be α-linked glycosylceramides (79). Since all glycosylceramides in mammals were believed to be β-anomers due to that mammalian glycosylceramide synthases are β-transferases (80), this finding is somewhat surprising. Earlier studies pioneered by the Brenner group showed, though initially thought to be a potent lipid self-antigen for iNKT cells, that β-glucopyranosylceramide (β-GlcCer) actually does NOT possess antigenic activity to iNKT cells (81, 82). The observed activity of β-GlcCer is likely due to inclusion of an α-GlcCer species (82). These observations suggested the possibility that α-glycolipids are endogenous antigenic lipids for iNKT cells (82). However, nuclear magnetic resonance spectroscopy analysis at the time did not render a definitive identity (82). It is possible that an unknown alternative enzymatic pathway, unfaithful enzymatic activities, or unique stressed cellular environments could confer production of small amounts of α-linked glycolipids, though the exact mechanism remain to be discovered (79, 83). The peroxisome-derived ether lipids seem to be partially involved in the iNKT cell development, as mice deficient in the peroxisomal enzyme glyceronephosphate O-acyltransferase (GNPAT) harbor moderately reduced iNKT cells and GNPAT<sup>−</sup>/<sup>−</sup> thymocytes are unable to support maturation of iNKT cells (84). However, the number of stage 0 iNKT cells are NOT changed in GNPAT<sup>−</sup>/<sup>−</sup> mice (84), suggesting that peroxisome-derived lipids may not be the predominant selecting ligands for iNKT cells, but rather influence later developmental events of iNKT cells. The lysosomal phospholipase A2 (LPLA2), which modifies lysophospholipids in the lysosome, has been shown to play a role in thymic selection of iNKT cells, as both CD1d endogenous antigen presentation and iNKT cell numbers were negatively affected in the absence of LPLA2 (85). Taken together, considering that maturation of iNKT cells after positive selection of stage 0 iNKT cells requires the presence of CD1d in the thymus (86), it is possible that the endogenous lipid ligands for iNKT cells are presented in both thymic cortex and medulla and are displayed by different antigenpresenting cells (APCs). In this fashion, they may influence both selection (in the cortex) and effector differentiation (in the medulla) of iNKT cells (**Figure 2**).

Consistent with a potential role of specific self-lipids in effector differentiation, it was noted that the three iNKT subsets express distinct but stable Vβ repertoires (11, 87, 88). For example, NKT2 cells show a higher usage of Vβ7 (11). Thus, a few studies have raised the hypothesis that differential TCR signaling events due to biased TCR Vβ gene usage could impact the differentiation of iNKT subsets (87, 88). Through generation of retrogenic mice expressing different CDR3β to manipulate iNKT TCR β chain *in vivo*, a recent study clearly demonstrated the half-life of TCR-Ag-CD1d interaction governs the frequency of different iNKT subsets in a cell-intrinsic manner. The number of NKT2 cells strongly correlated with the *t*1/2 of tetramer binding (89). As mentioned above, a high level of Nur77GFP was seen in NKT2 cells in the steady state, suggesting continuous TCR signaling in NKT2 cells (11). However, it is less clear whether such continuous TCR stimulation is required for the steady-state production of IL-4 in NKT2 cells and/or the development of PLZFhigh NKT2 cells (**Figure 2**). Since NKT2 cells reside in the thymic medulla, further efforts are required to elucidate where and how TCR binding kinetics of NKT2 cells might control their differentiation (**Figure 2**).

#### iNKT CELLS MODULATE TISSUE HOMEOSTASIS

#### Major Role of iNKT-Derived IL-4 Thymus

Using KN2 mice, in which a human CD2 cassette was knocked into the IL-4 gene locus, and human CD2 expression on cell surface indicates active secretion of IL-4, a previous study demonstrated that thymic NKT2 cells produce abundant IL-4 at steady state (11). Another group showed thymic iNKT cells may also produce IL-13 at steady state in IL-13GFP mice (90). Because iNKT cells are predominantly localized in the medulla, IL-4 produced by NKT2 cells could influence a variety of immune events in that environment (**Figure 2**). Indeed, steady-state production of IL-4 selectively activates STAT6 in medullary CD8<sup>+</sup> single-positive thymocytes, which drives them to become memory phenotype (CXCR3<sup>+</sup> CD122<sup>+</sup> Eomes<sup>+</sup>) (11, 40). This population of IL-4 induced memory T cells has been categorized as innate memory T cells (91), and they maintain greater function compared to naïve CD8<sup>+</sup> T cells. They are well equipped to produce IFN-γ in response to TCR stimulation and showed much better expansion after infection with *listeria monocytes* (LM) (40). Moreover, the developmental exposure to IL-4 is critical for CD8<sup>+</sup> T cells to mount robust Th1 responses to acute or chronic lymphocytic choriomeningitis virus infection (92, 93). Therefore, innate memory T cells are beneficial to the host for their functional superiority (94). Nevertheless, we do not yet understand, in the bigger picture, why iNKT cells recognition of medullary selflipids should control this process.

IL-4 impacts other immune cells beyond CD8 T cells in thymic medulla. A recent study demonstrated that the type 2 cytokines (IL-4/13) produced by iNKT cells could influence in the thymic emigration of mature thymocytes (90). IL-4Rα−/<sup>−</sup> mice showed accumulation of mature T cell in the thymus and reduced recent thymic emigrants in the periphery (90). Medullary thymic epithelial (mTEC) cells express IL-4Rα and can respond to the type 2 cytokines, as pSTAT6 level went up in mTEC of FTOC when IL-4 and IL-13 were added in the culture (90). Moreover,

disorganization of the thymic medulla was observed in mice deficient of IL-4Rα, that the medulla contained some epithelial-free areas revealed by the ERTR5 staining (90). It was speculated that IL-4/13 signaling in mTEC might promote the egress of mature T cells from thymus, though the specific mechanism remains to be uncovered (90). While the S1p–S1p1 axis remains intact in the IL-4Rα−/<sup>−</sup> mice (90), it is possible that the IL-4/13 produced by NKT2 cells serve as a novel regulator of thymic emigration of T cells.

#### Periphery

In the periphery, iNKT cells are critical for restoring homeostasis under stressed conditions. The regulatory role of iNKT cells has been implicated in the type 1 diabetes, where iNKT cells are less frequent and biased toward Th1 cytokine production in diabetic siblings than in their non-diabetic siblings (95). The protective role of iNKT cells has been shown in the mouse model for type 1 diabetes [non-obese diabetes (NOD) mouse], as CD1d<sup>−</sup>/<sup>−</sup> NOD mice, lacking iNKT cells, have a higher risk and earlier onsets of diabetes compared to CD1d+/+ counterparts (96). Such protection is dependent on the IL-4 production by iNKT cells (97, 98), and activation of iNKT cells to produce IL-4 by cognate lipid antigen α-Galcer prevents diabetes in NOD mice (99, 100).

Recent studies highlighted the key role of iNKT cells in regulating the pathogenesis of graft-versus-host disease (GvHD), a severe immunological dysregulation that frequently occurs after allogeneic hematopoietic stem cell transplantation (101, 102). Higher frequency of iNKT cells in patients correlated with lower risk of GvHD (102). In murine studies, stimulation with α-Galcer or adoptive transfer of iNKT cells confer substantial protection against GvHD (103, 104). Furthermore, the iNKT cell-derived IL-4 and following regulatory T cells expansion seem to be critical for optimal suppression of GvHD (102–104). These data point iNKT cells as promising therapeutic regimen for GvHD patients.

Invariant natural killer T cells are rare in most peripheral sites (0.1–1% of lymphocytes), but highly enriched in the liver, representing nearly 30% of hepatic lymphocytes (23). They are actively involved in restoring tissue homeostasis after sterile liver injury as demonstrated in a recent report (105). Shown by intravital microscopy, iNKT cells randomly patrol the sinusoids within liver in the steady state, while they rapidly move toward the injury site after injury (105, 106). Arrested at the injury site due to TCR stimulation and IL-12/18 signals, iNKT cells produce IL-4 to promote a series of events that are vital for optimal tissue repair, including increased proliferation of hepatocytes, the switch of monocyte subtypes from CCR2high CX3CR1low to CCR2low CX3CR1high, as well as reduced collagen deposition (105).

Altogether, these studies demonstrated that iNKT cells are potent regulator in immunity, largely due to their ability to produce abundant cytokines. Most of the studies implicated iNKT cell-derived IL-4 as the critical factor in restoring tissue homeostasis. Therefore, to unleash the therapeutic potential of iNKT cells, it will be important to have better understanding of the underlying mechanisms, especially the relevant APCs and the precise stimulatory lipid antigens that activate iNKT cells to produce IL-4.

#### Role of Other iNKT Subsets

Invariant natural killer T cells also have strong potential to produce other cytokines (IFNγ by NKT1 and IL-17 by NKT17). However, the role of these subsets and cytokines on tissue homeostasis has not been deeply explored, although it should be noted that NKT17 are abundant in the lung. iNKT cells also express a variety of other stimulatory or inhibitory molecules; therefore, they might influence immune homeostasis through direct cell contact. One of the molecules expressed by iNKT cells is RANK ligand (RANKL) (24). Signals through tumor necrosis factor family receptors (TNFRSF) RANK promotes Aire expression in mTEC (107). While iNKT cells express RANKL, and Aire<sup>+</sup> mTECs were significantly reduced in CD1d<sup>−</sup>/<sup>−</sup> mice (24). It strongly suggests that iNKT cells could regulate the development of mTEC through direct cross-talk to induce RANK signals. Further RNA-Seq analysis demonstrates that only NKT2 and NKT17 cells highly express RANKL (28), suggesting that iNKT subsets may have unique effects in modulating tissue homeostasis in the thymus (**Figure 2**).

#### THE PARALLELS IN DEVELOPMENT OF iNKT CELLS AND MUCOSAL-ASSOCIATED INVARIANT T (MAIT) CELLS

The MAIT cells are another specialized lineage of innate-like T cells, expressing a semi-invariant TCR, that Vα7.2-Jα33 chain predominantly paired with Vβ2 or Vβ13 in human and Vα19-Jα33 chain predominantly paired with a Vβ6 or Vβ8 in mice (108). They are remarkably abundant in human tissues, making of 1–10% of T cells in peripheral blood, nearly 10% of T cells in intestine and up to 40% of T cells in liver (109, 110). Therefore, MAIT cells have attracted great interest in terms of elucidating their development and function. Recently, with the discovery of the vitamin B metabolites as cognate antigens and successful manufacturing of MR-1 tetramer to accurately detect MAIT cells in mice and human (111, 112), we have gained a more clear understanding of their development and homeostasis. Surprisingly, the thymic development of MAIT cells parallels many aspects of iNKT cells (**Figure 3**).

Mucosal-associated invariant T cells originate in the thymus where their selection depends on the interaction with the MR-1 expressing DP thymocytes (113). Positively selected immature MAIT cells are CD24<sup>+</sup> CD44<sup>−</sup>, which give rise to the CD24<sup>−</sup> CD44<sup>+</sup> mature MAIT cells (70). These CD44<sup>+</sup> MAIT cells are comprised of at least two distinct subsets, T-bet<sup>+</sup> MAIT cells and ROR-γt <sup>+</sup> MAIT cells reminiscences the NKT1 and NKT17 cells. Moreover, like iNKT cells, MAIT cells express PLZF and depends on PLZF for their differentiation, as CD44<sup>+</sup> MAIT cells were absent in PLZF-null mice (70). Furthermore, microRNA plays indispensable role in the development of both MAIT cells and iNKT cells, the expansion and differentiation of MAIT cells beyond the CD24+ stage were severely impaired in Droshadeficient mice (70). With the notion that MAIT cells development might parallel the development of iNKT cells, it is reasonable to reference what we learned from iNKT cells to facilitate and advance our understanding of MAIT cells. Many tools designed and hypotheses raised for research of iNKT cells could be applied to that of MAIT cells. Using CD24 and CD44 to distinguish immature and mature MAIT cells, as well as examine expression and dependency of PLZF in MAIT cells are both good examples of that. Taken one step further, more questions could be asked: (1) whether MAIT cells receive strong TCR signal like iNKT cells during selection; (2) whether the two MAIT cell effector subsets require differentiation cues similar to those for NKT1 and NKT17 cells; (3) whether thymic MAIT effector cells predominantly reside in medulla; (4) and PLZF induce tissue residency program in iNKT cells—is it the same in MAIT cells?

#### MORE iNKT SUBSETS: NKT10, NKTFH, AND ADIPOSE iNKT

Beside the three effector subsets in the thymus, additional functional subpopulations of iNKT cells have been described. Follicular helper iNKT cells (NKTFH) were detected after immunization with α-GalCer-conjugated proteins or haptens (114, 115). NKTFH adopt the phenotype of MHC-II restricted T follicular helper cells (TFH), expressing a variety of classical TFH surface markers and transcription factor, including PD-1, CXCR5, ICOS, and Bcl6 (114, 115). NKTFH initiate and localize in germinal centers, provide both cognate and noncognate help to lipid and protein-specific B cells, respectively (114, 115). However, NKTFH-dependent germinal center reactions failed to generate long-lived plasma cells (114). Another specialized subpopulation of iNKT cells emerges after stimulation with αGalCer is the regulatory NKT10 cells, characterized by predominant IL-10 production (33). Unlike T regulatory (Treg) cells, NKT10 don't express Foxp3, rather, they highly express E4BP4, a

transcription factor regulates IL-10 and IL-13 production in CD4 T and iNKT cells (32, 33).

Adipose iNKT cells have gained focus for their crucial role in modulating Treg cells and macrophages, which are correlated with the onset of obesity (116). However, the generation or selection of adipose iNKT cells has been a puzzle. Interestingly, adipose iNKT cells have been found to share phenotype with NKT10 cells, in that they both produce abundant IL-10 and rely on E4BP4 for their regulatory function (34). A recent discovery showed that recognition of CD1d by iNKT TCR controls the development of iNKT cells in the adipose tissue (117). TCRα–TCRβ pairing of iNKT TCR creates a hydrophobic patch, which is critical for maintaining TCR conformation as well as its recognition of CD1d molecule (117). Partial disruption of this patch by substitution of a single amino acid in TCR Vβ8.2 chain (F108Y), while recognition of CD1d preserved, significantly alters the development of iNKT cells, results in an enrichment of iNKT cells in the adipose tissue (117). It is unclear whether this is due to altered selection in the thymus or enhanced proliferation/competitive advantage of adipose iNKT cells on site.

# CONCLUDING REMARKS

T cells play a central role in protecting the body from infectious agents and cancer, but at the same time can cause autoimmune diseases when dysregulated. iNKT cells are a specialized lineage of T cells that recognize foreign and self-lipids in a manner quite distinct from conventional T cells. Though iNKT cells are a numerically small population, their striking ability to rapidly produce large amounts of cytokines renders them potent regulators of immunity—implicated in antimicrobial responses, antitumor immunity, and autoimmune and allergic diseases. Despite past progress, a number of questions regarding the development of iNKT cells remain unanswered. First, what is the nature of endogenous lipids recognized by iNKT cells, especially the lipids presented by cortical DP thymocytes that induce positive selection of iNKT cells? Second, evidence suggests that NKT2 cells produce large amounts of IL-4 at steady state in the thymus. It is of great interest to understand how this process is regulated. Are antigenic lipids and TCR stimulation required, and if so what is the identity of APCs?

Finally, though iNKT cells are found in most tissues, the frequency of iNKT subsets varies greatly in different organs. For instance, NKT17 are enriched in lung and skin draining LN, while liver iNKT cells are predominantly NKT1. What dictates this striking bias in the distribution of iNKT subsets? How does this differential distribution influence immune responses and/or modulate tissue homeostasis? What is the phenotype of iNKT cells that recently emigrated from thymus to seed in periphery? What are the environmental and cell-intrinsic factors that regulate differentiation or homing of iNKT subsets in various peripheral sites? iNKT stimulatory lipids are well-tolerated in human trials. Through selective activation of different iNKT effector subsets, iNKT cells can modulate immune responses and tissue homeostasis in different fashions. This can only be possible

#### REFERENCES


with a better understanding of the developmental steps that drive iNKT cells into functionally distinct subsets.

### AUTHOR CONTRIBUTIONS

HW drafted the manuscript. KH supervised the writing and edited the manuscript.

#### ACKNOWLEDGMENTS

We thank Dr. Hristo Georgiev for reading the manuscript, and all present and past members of the Hogquist and Jameson labs for productive discussions and assistance.

## FUNDING

This work was supported by NIH grant R37 AI39560 (KH) and UMN doctoral dissertation fellowship (HW).


recognition. *Proc Natl Acad Sci U S A* (2013) 110(13):5097–102. doi:10.1073/ pnas.1302923110


**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 SJ and handling Editor declared their shared affiliation.

*Copyright © 2018 Wang and Hogquist. 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.*

*S. Harsha Krovi1 and Laurent Gapin1,2\**

*1Department of Immunology and Microbiology, University of Colorado Anschutz Medical Campus, Aurora, CO, United States, 2Department of Biomedical Research, National Jewish Health, Denver, CO, United States*

Invariant natural killer T (iNKT) cells are a CD1d-restricted T cell population that can respond to lipid antigenic stimulation within minutes by secreting a wide variety of cytokines. This broad functional scope has placed iNKT cells at the frontlines of many kinds of immune responses. Although the diverse functional capacities of iNKT cells have long been acknowledged, only recently have distinct iNKT cell subsets, each with a marked functional predisposition, been appreciated. Furthermore, the subsets can frequently occupy distinct niches in different tissues and sometimes establish long-term tissue residency where they can impact homeostasis and respond quickly when they sense perturbations. In this review, we discuss the developmental origins of the iNKT cell subsets, their localization patterns, and detail what is known about how different subsets specifically influence their surroundings in conditions of steady and diseased states.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Sebastian Joyce, Vanderbilt University, United States Kristin Hogquist, University of Minnesota Twin Cities, United States*

*\*Correspondence: Laurent Gapin laurent.gapin@ucdenver.edu*

#### *Specialty section:*

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

*Received: 02 May 2018 Accepted: 05 June 2018 Published: 20 June 2018*

#### *Citation:*

*Krovi SH and Gapin L (2018) Invariant Natural Killer T Cell Subsets—More Than Just Developmental Intermediates. Front. Immunol. 9:1393. doi: 10.3389/fimmu.2018.01393*

Keywords: invariant natural killer T cells, subsets, development, homeostasis, cytokine secretion

# INTRODUCTION

Adaptive immunity has long been appreciated as a chief means through which various jawed vertebrates stave off infectious pathogens. One of the main features differentiating the adaptive immune system from its innate counterpart is the generation and expression of diverse antigen receptors. Antigen sensors of the innate immune system are proteins with fixed sequences and pattern-recognition motifs encoded within the germline (1). By contrast, antigen receptor generation by adaptive immune cells involves complex somatic DNA rearrangements that juxtapose genes otherwise separated by thousands to millions of base pairs (2). Each cell randomly rearranges its antigen receptor locus to ensure that few cells express identical receptors. In so doing, the cells express clonal receptors and develop an antigen receptor repertoire diverse enough and with the potential to recognize the plethora of existing antigens that the host is likely to encounter. Because the antigen receptors in the adaptive immune system are heterodimers, with both subunits undergoing independent rearrangements, the combinatorial diversity has been estimated to exceed 1015 unique sequences (3).

The two types of cells belonging to the adaptive immune system, B and T lymphocytes, have each evolved in different ways to efficiently respond to infections. In particular, B cells produce and secrete antibodies that target and bind to different conformational epitopes on pathogens (4). T cells, on the other hand, express cell membrane-tethered antigen receptors (TCRs, composed of α-chains paired to β-chains), thereby necessitating their proximity to their targets in order to initiate an immune response. These TCRs primarily recognize their target ligands by interacting with major histocompatibility complex (MHC) proteins, which present linearized peptides, expressed on adjacent cells (5). The presented peptides are processed fragments from full length proteins and can be presented by MHC-I molecules interacting with TCRs expressed on CD8<sup>+</sup> T cells or MHC-II proteins interacting with TCRs expressed on CD4<sup>+</sup> T cells (6). Correspondingly, these T cells not only interact with different MHC molecules but also produce distinct responses.

To ensure that T cells are capable of mounting immune responses against all kinds of invading pathogens, T cells have further evolved to differentiate into functionally distinct subsets. Indeed, CD4<sup>+</sup> T cells can differentiate into TH1, TH2, TH17, Treg, among others, upon exposure to their cognate antigens (7). Each of the subsets produces a distinct set of cytokines with the capacity to skew the immune response in a specific direction. For example, TH1 cells produce IFNγ, a pro-inflammatory cytokine that promotes increased antigen presentation by MHC molecules and increased phagocytosis by macrophages, to name a few of its effects. CD4<sup>+</sup> T cell differentiation into TH1 cells is thought to be primarily due to intracellular pathogens (8). TH2 cells, though, produce a different set of cytokines, including IL-4, IL-5, and IL-13, and assist in combating extracellular pathogens such as parasites and helminths (9). Through this division of labor, functionally different T cell subsets can resolve infections by producing responses catered to the pathogen.

The substantial TCR repertoire diversity in the adaptive immune system serves as a double-edged sword. Many different TCRs are expressed by the T cell population *en masse*, but only a few T cells expressing TCRs specific for a given antigen exist within the population (10). Thus, T cells undergo extensive proliferation when they first encounter their antigen to generate enough cells with the proper antigen-specific TCRs. The activated cells can then travel to the site of infection and execute their appropriate functions upon antigen re-challenge. As this process usually takes several days, the adaptive immune system is considered a slow and deliberate yet specific form of response. After the immune response has been resolved, some of the cells differentiate into memory cells, which exhibit faster response times in the event that the pathogen reinfects the host (11).

Due to the delayed kinetics of this "conventional" arm of the adaptive immune system, other "innate-like" adaptive lymphocytes play a crucial role early during an infection (12). These cells are unique because they express markers associated with memory cells despite not having encountered their antigens previously. Additionally, they are functionally poised and capable of responding within hours as opposed to days, in line with their innate-like capabilities. Though they make up only a fraction of the overall T cell population, they still exert crucial and sometimes nonredundant functions. One such lymphocyte population, which will be the focus of this review, is the invariant natural killer T (iNKT) cell lineage. These cells straddle the innate-adaptive boundary because they respond quickly upon stimulation (within hours), yet, express a TCR that underwent somatic rearrangement (13). Indeed, the vast majority of iNKT cells express an identical TCRα chain (TRAV11-TRAJ18 in mice, TRAV10-TRAJ18 in humans) paired to a restricted set of TCRβ chains (TRBV1, TRBV13, and TRBV29 in mice, TRBV25 in humans), with some notable exceptions (14–16). Furthermore, instead of interacting with peptides presented by MHC molecules, the TCRs expressed by iNKT cells recognize (glyco)lipids presented by CD1d, a non-polymorphic MHC-I-like molecule (17).

Analogous to the aforementioned functionally different conventional T cell subsets, iNKT cells also come in different flavors, each of which exhibits a different functional profile (18–20). Such division of labor between functionally different iNKT cell subsets perhaps could explain why iNKT cells have been implicated in ameliorating or exacerbating a variety of diseases and illnesses ranging from autoimmunity to cancer. Historically, iNKT cells have been lumped into one category despite their varied roles in responses. However, with the recent identification of functionally distinct iNKT cell subsets, how and which iNKT cell subsets might affect the development of the immune system and its response need to be updated. In this review, we will focus on how the different iNKT cell subsets develop and consequently, to what extent each of these subsets actively participates in immune responses.

#### iNKT SUBSETS

Initially, an intriguing population of mature T cells was identified in the thymus by their lack of expression of CD4 or CD8 coreceptors (double negative, DN) but with surface expression of a TCR, thereby distinguishing them from other immature thymocytes (21). These DN cells were functionally competent since they could produce IL-4 and IFNγ readily after stimulation and also expressed the natural killer (NK) cell marker NK1.1 (22–24). This was particularly novel because T cells were not traditionally considered to be cytokine secretion-competent in the thymus and suggested that functional competence by these cells might be acquired during their development. Sequencing the TCRs from these cells repeatedly provided investigators with the same TCRα chain sequence (25, 26), and it was eventually determined that the cells required CD1d expression for their development, suggesting that they recognized lipids instead of peptides (27). Due to the expression of a TCR as well as NK markers by these cells, the name iNKT took preferential hold as a label for these cells.

With the discovery of the marine sponge-derived lipid α-galactosylceramide (αGC) that when bound to CD1d strongly stimulated these cells and the advent of MHC-loaded tetramer technology, iNKT cells could now be tracked with profound resolution (28–30). Interestingly, it became readily apparent that not all the cells that were identified by αGC-loaded CD1d tetramers were NK1.1<sup>+</sup>, suggesting phenotypic heterogeneity within the iNKT compartment. Because the NK1.1<sup>+</sup> cells composed the overwhelming majority of the total tetramer<sup>+</sup> population in the thymus in C57BL/6 (B6) mice, the NK1.1<sup>−</sup> cells were thought to perhaps represent developmental intermediates. Indeed, support for this idea came from experiments in which intrathymic transfers of NK1.1<sup>−</sup> cells could generate NK1.1<sup>+</sup> cells (31, 32). Interestingly, stimulating the NK1.1<sup>−</sup> cells led to the production of larger amounts of IL-4 compared to IFNγ, in stark contrast to what the NK1.1<sup>+</sup> cells produced, which was primarily IFNγ and little IL-4 (31–33). Additionally, the iNKT cells that were primarily exported from the thymus were NK1.1<sup>−</sup> cells while the NK1.1<sup>+</sup> cells were retained in the thymus (34, 35). Thus, it was unclear why the intermediates had a different cytokine secretion profile compared to the terminally matured population and furthermore, why/how the immature cells emigrated from the thymus if they were truly meant to give rise to the mature iNKT cells (36).

Only recently has this conundrum been resolved due in large part to the work by the Hogquist group. Instead of identifying iNKT cells simply by the tetramer and NK1.1, they also stained the cells with transcription factors known to endow specific fates (19). Because of years of work in understanding conventional CD4<sup>+</sup> T cell differentiation, it was known that the master transcription factors engendering the TH1, TH2, and TH17 fates were T-bet, GATA-3, and RORγt, respectively (8). In addition, large scale screens by two groups had recently identified that all iNKT cells require the expression of the zinc finger transcription factor promyelotic leukemia zinc finger (PLZF) (37, 38). In fact, the few remaining iNKT cells (as identified by the tetramer) found in PLZF-deficient mice resembled naïve T cells that secreted IL-2 upon stimulation but not significant levels of IL-4 or IFNγ, highlighting the transcription factor's importance (37, 38). By using antibodies targeting T-bet, GATA-3, RORγt, and PLZF, Hogquist and colleagues had the surprising finding that not all thymic iNKT cells expressed each of the transcription factors. Instead, three distinct thymic subpopulations were identified based on their staining: PLZFhi GATA-3hi (iNKT2), PLZFint RORγt <sup>+</sup> (iNKT17), and PLZFlo, T-bet<sup>+</sup> (iNKT1) (19). Moreover, NK1.1 primarily stained the iNKT1 cells. Thus, although a few iNKT1 precursors were present in this pool, many of the NK1.1<sup>−</sup> cells were terminally differentiated cells themselves. This was further confirmed by stimulating the different subsets *in vitro*, with iNKT1 cells producing large amounts of IFNγ and a little IL-4, iNKT2 cells producing large amounts of IL-4, and iNKT17 cells secreting IL-17, putting to rest the functional discrepancy between NK1.1<sup>+</sup> and NK1.1<sup>−</sup> cells (19).

Interestingly, not all thymic iNKT subsets are equally represented in different strains of mice. Certain strains, such as the BALB/c, have large proportions of iNKT2 and iNKT17 cells with a correspondingly reduced proportion of iNKT1 cells. On the other extreme, B6 mice instead possess largely iNKT1 cells and few iNKT2/iNKT17 cells (19). This is particularly important because previous work in these mice had led to the prevalent idea that, generally speaking, B6 mice tended to be predisposed to a "TH1 phenotype" while BALB/c mice displayed a "TH2 phenotype" (39, 40). Whether the iNKT cell compositions in either of these strains are a cause or a consequence of these phenotypes is unknown. Another point of note is that the reason for the overrepresentation of NK1.1<sup>+</sup> iNKT cells in previous experiments was the predominant use of the B6 mouse for the study of iNKT cell development. In fact, the antibody targeting NK1.1 in B6 mice does not recognize the epitope present in BALB/c mice due to an allelic variance (41). The use of this antibody necessitated experiments in B6 mice since the cells were initially characterized by their expression of NK1.1 (22). Now, however, iNKT cells in all strains are identified by their ability to interact with the αGC (or its analog PBS57) loaded CD1d tetramer and their transcription factor profile (or surface proteins known to be specifically upregulated by these transcription factors) serves as a readout of the subset proportions.

The terminal differentiation status of these subsets has also been challenged due to the discovery of new iNKT cell subsets in the periphery. Although only the three aforementioned subsets are largely represented in the thymus, analysis of other tissues in both steady-state and immunization conditions has revealed the presence of novel iNKT subsets. In the adipose tissue, a special iNKT cell population, named iNKT10, has been identified that depends on expression of the transcription factor E4BP4 for its role in maintaining adipose tissue homeostasis (42). Similarly, an iNKTFH population expressing the transcription factor Bcl6 has been observed in the peripheral lymphoid organs of immunized mice (43). This population secreted IL-21 and provided cognate help for B cells undergoing affinity maturation, much like conventional TFH cells in germinal centers (43–45). Ongoing work should help determine whether these additional subsets are indeed generated at low frequencies in the thymus or if they differentiate into their observed subsets within other tissues.

#### iNKT CELL SUBSET DEVELOPMENT

CD4<sup>+</sup> CD8<sup>+</sup> [double positive (DP)] thymocytes serve as the progenitors for all cells belonging to the αβ T cell lineage (46, 47). iNKT cells are no different as they also principally originate from DP precursors (48, 49), with a minor proportion utilizing an alternative pathway (**Figure 1**) (50). DP cells randomly rearrange their TCRα loci to generate the invariant TCRα chains that pair with suitable TCRβ chains (49). While DP precursors of conventional T cells are selected by MHC I/II on thymic epithelial cells (TECs), iNKT cell DP precursors are instead positively selected by self-lipids presented by CD1d expressed on fellow DP thymocytes (51, 52). The DP–DP interaction provides the iNKT precursor with the obligate lipid/CD1d ligand along with a distinctive homotypic co-stimulation through members of the signaling lymphocytic activated molecules (SLAM) family of receptors. Signals derived from SLAM family receptor interactions are required to produce mature iNKT cells because iNKT cells are notably absent in mice in which an adapter downstream of SLAM receptors (SAP) has been deleted (38, 53). Interestingly, different mouse strains also express different alleles of the SLAM receptors. For example, BALB/c mice possess an allele of SLAMF6 (Ly108) that is hyperphosphorylated upon engagement compared to the B6 Ly108 allele (54, 55). This hyperphosphorylation has a functional effect because a stronger signal is consequently transduced in BALB/c DP thymocytes (56), suggesting that signals received by iNKT cell precursors during development in the thymus might not be equivalent across mouse strains.

Co-stimulations of DP cells *via* the TCR and the SLAM receptors elicit a strong signal in the iNKT precursors leading to a high expression of the transcription factor Egr2 (56). Without Egr2, thymocytes are arrested early during iNKT cell development (57–59). High expression of Egr2 is dispensable for conventional T cell development (57), suggesting that iNKT cells are unique in their requirement for stronger-than-normal agonistic signals to properly mature. Indeed, post-positive selection iNKT cells, commonly referred to as stage 0 iNKT cells, expressed the highest levels of Nur77 (encoded by *Nr4a1*), an immediate early protein induced upon TCR signaling, compared to all other thymocytes (60). Furthermore, Egr2 has been demonstrated to bind directly

tissues. RANKL expressed by medullary iNKT2 (and iNKT17) cells also promotes maturation of medullary thymic epithelial cells (mTECs) into Aire+ MHC-IIhi mTECs, which mediate negative selection of medullary SP thymocytes and Treg maturation.

to the promoter and positively regulate the transcription of *Zbtb16*, the gene encoding PLZF. In the absence of Egr2 and its related protein Egr1, PLZF levels in the few remaining iNKT cells are significantly lower, further corroborating the link between strong TCR/SLAM signaling and proper iNKT development (59). iNKT cells are imbued with an activated/memory phenotype relatively early during their ontogeny, primarily due to their expression of PLZF (37, 38). In fact, ectopic expression of PLZF is sufficient to promote a memory phenotype even in conventional CD4<sup>+</sup> T cells at steady state (38, 61).

One of the outstanding questions in iNKT cell biology currently is how iNKT cell subset differentiation occurs in the thymus. Conventional T cells in the periphery require antigeninduced priming as well as the appropriate cytokine milieu to be properly polarized into different functional subsets. For example, generating TH1 cells from naïve CD4<sup>+</sup> T cells requires TCR-mediated signals in addition to the cytokine IL-12 that promotes commitment to the TH1 lineage through the actions of the transcription factor signal transducer and activator of transcription 4 (7). TH2 cells, though, require TCR-mediated signals in conjunction with IL-4 to be polarized into their lineage (9). Polarization into different lineages in this manner has been demonstrated to be dependent on specific factors both *in vitro* and *in vivo*. However, since iNKT cells differentiate in the thymus unlike conventional T cells that differentiate in the periphery, it is unclear if they follow a similar differentiation course.

Several factors, besides the canonical transcription factors, have been revealed to be relevant for specific iNKT cell subset development and maintenance, primarily through the use of knockout mice. The transcription factor lymphoid enhancer factor 1 (Lef-1) was shown to influence differentiation into the iNKT2 lineage (62). Lef-1 expression strongly correlates with PLZF expression and Lef-1<sup>−</sup>/<sup>−</sup> mice have significantly fewer iNKT2 cells capable of producing IL-4 (62). Mice lacking the serine protease Serpinb1 generate more iNKT17 cells in the thymus and more iNKT17 cells are found in peripheral tissues in KO mice as well (63). When the transcription factor Bcl11b was specifically knocked out of PLZF<sup>+</sup> cells using a PLZF-Cre mouse, iNKT17 cells were preferentially maintained at the expense of iNKT1 and iNKT2 cells, suggesting that Bcl11b is required for iNKT1 and iNKT2 cell survival but is dispensable for iNKT17 cell survival (64). The transcription factor Th-POK appears to restrict the generation of iNKT17 cells because in its absence, iNKT17 cell numbers are increased and more iNKT cells are capable of producing IL-17 (65). Th-POK itself is epigenetically regulated by the micro-RNA miR133b and more iNKT17 cells are observed in mouse strains where iNKT cells express higher levels of this miRNA (66). In this manner, many other factors have similarly been described to be preferentially important for the development and/or maintenance of one subset over others. Despite the abundance of studies reporting proteins required for different subsets, what is unknown is the stage of development during which these proteins are first expressed. In other words, it is unclear if these factors themselves are the force behind the development of specific subsets or are instead a consequence of commitment to a given subset. If the former, then, different precursors should exist, each with a specific phenotype, which predispose cells into a given subset. Support for this hypothesis is currently lacking. Instead, it has been shown that DP precursors exist in a poised state on a population level. Approximately 1,000 genes in bulk DP cells are transcriptionally silent yet possess both permissive (H3K4me3) and repressive (H3K27me3) histone modifications at their transcriptional start sites (67, 68). About 14% of these genes code for transcription factors, some of which are implicated in influencing lineage diversification. These results suggest that a given DP precursor is capable of adopting various fates but only commits to one upon receiving specific cues (68). Thus, differential signals received by DP thymocytes consequently could drive the commitment of individual precursors into distinct lineages that are then preferentially dependent on specific proteins.

Several hypotheses, not all of which are mutually exclusive, can be formulated to explain how iNKT cell subsets might arise from differential signaling during development. First, the cells might enter distinct pathways, as observed in the periphery, due to stimulation by different cytokines that impose commitment to a specific lineage. In support of this idea, each of the thymic subsets expresses a unique composition of cytokine receptors (19, 69). Indeed, iNKT1 cells display a dependence on the cytokine IL-15 for survival while iNKT17 cells instead require IL-7 to maintain their numbers (70, 71). In addition, iNKT2 and iNKT17 cells also seem to require IL-25 for homeostasis and function (72). However, this idea pre-supposes that commitment only occurs after upregulation of cytokine receptors that can enforce lineage specification. Instead, cytokine receptor expression frequently occurs as a consequence of expression of particular transcription factors themselves (73–75). Because expression of the transcription factors would already signify commitment, this implies that an even earlier event drove the cells to differentially upregulate these proteins. Furthermore, stage 0 iNKT cells do not express any appreciable levels of transcripts coding for lineage-determining transcription factors, even at the single cell level (69). Instead, their transcriptomes are reminiscent of uncommitted DP cells having recently undergone positive selection, suggesting that although commitment to a subset can be reinforced by cytokine receptor signaling, it is unlikely to be the original signal driving diversification.

Another hypothesis is that recognition of specific self-ligands during selection has the potential to shape the subset ratio. Accordingly, in a recent study, antigen-specific iNKT cells were readily identified in the thymus when using CD1d tetramers loaded with a variety of lipids (76). Even though all iNKT cells could be identified using tetramers loaded with the potent antigen αGC, different subpopulations reacted exclusively to specific other lipids. Although the authors did not further categorize the responding cells based on their functional subsets, it remains an appealing idea that differential recognition of CD1d-presented lipids might have dramatic consequences for iNKT precursors. In this scenario, lineage commitment would be expected to occur early after positive selection for each cell. Different microenvironments within the thymus could present high levels of specific lipids, thereby specifically promoting selection of certain iNKT cell subsets. This seems an unlikely proposition because selection itself has been demonstrated to occur on cortical DP thymocytes while the majority of the mature subsets (approximately 70%) take up residence in the medulla of the thymus, implying that migration to different thymic niches occurs well after positive selection (77). It is instead plausible that the precursors encounter different antigens merely by chance in a homogenous cortical environment, although this possibility remains to be formally demonstrated.

Different ligands, though, are only distinguished by iNKT DP precursors through the use of a diverse TCR repertoire (19, 78, 79), raising a third non-mutually exclusive possibility that iNKT cell subsets might arise due to differential signaling transduced by their TCR during positive selection. As the cells undergo selection, the strength of the signal perceived by each cell due to the nature of the TCR as well as the specific ligand being recognized could instruct each cell to adopt and commit to a specific lineage. This is an intriguing idea since TCR signal strength influencing fate decisions has been demonstrated in a variety of contexts (80–82). In addition, at the population level, it is reasonable to postulate that the precursor cells express an even distribution for a variety of markers, precluding the predisposition of any one cell to enter a given pathway. However, the nature of the αβ TCR (the TCRβ chain in particular in the case of iNKT cells) does vary from cell to cell, making it likely that the signals transduced by the TCRs could similarly vary. The signals thus generated might have a small range of strengths but through co-stimulation by SLAM receptors (and perhaps other coreceptors), this range could be amplified and engender distinct fates to cells that land on either end of the spectrum. In agreement with this, mature iNKT subsets express different levels of Nur77, with iNKT2 cells expressing the highest, followed by iNKT17 cells, and finally with iNKT1 cells expressing the lowest levels (19). Using Egr2 as a marker for strength of TCR-mediated signals during positive selection, data generated in our lab also confirm this hierarchy (83). Thus, the cells within the subsets seem to retain a memory of the signals they received as precursors, with iNKT2 cells having received the strongest signals followed by iNKT17 and iNKT1 cells.

One way that cells might retain the signaling information could be through the transcription factor GATA-3. TCR signaling has been previously demonstrated to upregulate GATA-3 protein levels (84). Many of the genes coding for the components of the TCR complex, namely the *Tcra, Cd3d,* and *Cd3g* loci, are direct targets of GATA-3 (75, 85–87). In mice lacking GATA-3, expression of these different genes is significantly reduced. Furthermore, GATA-3 has also been previously shown to autoregulate its own expression in a positive feedback loop (88). Therefore, stronger signaling during positive selection could potentially lead to higher and sustained GATA-3 levels and consequently, higher TCR levels. In support of this, the TCR levels (and GATA-3 levels to some extent) on the different subsets follow the same pattern as Nur77 and Egr2 do, perhaps suggesting that signals received during selection could be maintained in this manner (19, 63).

Pairing the invariant TCRα chain with different TCRβ chains can also affect the affinity with which the TCR heterodimer interacts with antigen/CD1d and consequently, how efficiently the TCR can initiate and propagate a signal intracellularly (89). Interestingly, in retrogenic mice generated with distinct TCRβ chains, the proportions of each of the subsets could be linked to the avidity of the TCR for its ligand (90). Similarly, when clonal mice were generated using nuclei from iNKT cells expressing different TCRs, the proportion of PLZFhi iNKT cells in the thymus directly correlated with the avidity of the TCR for lipid/CD1d (91). Finally, different studies have revealed that TCR signaling regulates the expression levels of several proteins involved in chromatin remodeling and in whose absence, the subset ratios are vastly altered (68, 92, 93). With the advent of myriad technologies allowing immunologists to assess transcriptomic and epigenomic signatures at the resolution of a single cell, it will become paramount in the future to pursue single cell analyses on the stage 0 iNKT cells immediately following positive selection and determine if TCR signaling-mediated differences can already be identified within these cells. Although a recent study did conduct single-cell RNA-sequencing analysis on stage 0 iNKT cells, the study concluded that these cells were similar to other positively selected conventional cells (69). As this study only analyzed 45 stage 0 iNKT cells, obtaining greater depth by sequencing more stage 0 iNKT cells could potentially provide more information on otherwise non-sampled low-abundance transcripts and/or accessible loci in different cells. With this information, perhaps an early signature can be identified that correlates with eventual iNKT cell subset.

#### iNKT SUBSET TISSUE HOMEOSTASIS

After developing in the thymus, iNKT cells have been observed in various tissues throughout the body (13). Unfortunately, due to an incomplete understanding of iNKT cell subsets, only their presence or absence in various tissues could be ascertained until recently. Some studies had identified iNKT cells in different tissues by αGC-CD1d tetramer staining, which remains the gold standard (30, 94, 95). This staining, however, was rarely done in conjunction with staining for the master transcription factors associated with the subsets, precluding their identification. In other studies, cells were frequently identified by their co-expression of NK1.1 and TCRβ (78, 96, 97). This strategy is perhaps problematic for multiple reasons. First, since staining for NK1.1 is not successful in all strains (41), it is entirely possible that observations made using the B6 mouse model are not generalizable to all mouse strains, as demonstrated in BALB/c and non-obese diabetic (NOD) NK1.1-congenic mice (98). Second, NK1.1 does not exclusively mark iNKT cells as conventional CD8<sup>+</sup> T cells can also co-express NK1.1, potentially obfuscating the real iNKT population (99, 100). Indeed, cytokine stimulation can lead to upregulation of NK1.1 and other NK cell-related markers in CD8<sup>+</sup> T cells, perhaps suggesting that iNKT1 cells acquire NK1.1 expression in a similar manner (101). And finally, since iNKT1 cells are primarily the only cells expressing NK1.1, learning about iNKT cell tissue localization through the use of this marker is by necessity restricted to this subset. Despite these drawbacks, some aspects of the tissue distribution patterns of iNKT cell subsets could be gleaned from early studies.

Of the subsets, iNKT1 cells have been indirectly demonstrated to remain long-term thymic residents and accumulate over time. When congenically marked thymic lobes were transplanted in recipient mice, while different kinds of iNKT cells were observed early after transplantation, only NK1.1<sup>+</sup> iNKT cells persisted in the thymus as time progressed (34). In fact, over 50% of the mature αβ TCR<sup>+</sup> cells remaining of donor origin were these likely iNKT1 cells that were maintained for a long period of time. This finding is in stark contrast to the conventional T cell population that is rapidly turned over in the thymus (102, 103). One possible explanation for thymic retention of iNKT1 cells is that T-bet drives expression of the chemokine receptor CXCR3, allowing them to be maintained in the thymus due to high levels of the cognate CXCR3 ligand CXCL10 (35). Another explanation for this phenomenon could be that T-bet in iNKT1 cells induces the expression of the gene *Il2rb* coding for the protein CD122 (73), thereby supporting the response to the trans-presented IL-15 cytokine (104). This cytokine is produced by cells in the thymic medulla and not only serves as a survival cytokine for iNKT1 cells but also help stabilize T-bet itself in those cells (70, 105). In Krovi and Gapin iNKT Cell Subsets

addition, IL-15 has been previously shown to upregulate CD69 in cells sensitive to this cytokine and indeed, iNKT1 cells do express high levels of CD69 (106, 107). Because ectopic overexpression of CD69 prevents thymic egress of conventional T cells (108), this IL-15-induced CD69 could potentially also play a role in iNKT1 cell thymic retention.

Despite the thymic retention, NK1.1<sup>+</sup> iNKT cells are also found in the periphery. Interestingly, large numbers of NK1.1<sup>+</sup> iNKT cells are found in the liver (109). This could be linked back to iNKT1 cells expressing T-bet and their sensitivity to IL-15. As previously mentioned, T-bet expressing cells also concomitantly express CXCR3 while IL-15 has been shown to condition cells to express CXCR6 in humans (110). The ligands for both these chemokine receptors (CXCL9/CXCL10 for CXCR3 and CXCL16 for CXCR6) are present in abundant quantities in the liver (111– 113). Thus, by following their chemotactic gradients, it is not surprising that iNKT cells compose 20–30% of T-lymphocytes in the liver (30, 94). Moreover, liver iNKT cells also establish strong residency upon arrival, as evidenced by their reduced circulation in parabiotic mice (94). Long-term residency by these lymphocytes has been proposed to be due to the high expression of the transcription factor Hobit (114). Induced by both T-bet and IL-15, Hobit has been shown to be preferentially expressed in liver iNKT cells, preventing their egress from the liver (114), although a recent study disputes this finding (115). Nevertheless, thymic iNKT1 cells also express high levels of Hobit perhaps suggesting they might maintain their residency in a similar manner (116).

With B6 mice remaining a popular mouse model to study iNKT cells, iNKT2 and iNKT17 cell localization has been largely understudied. While in some cases, there has been some direct evidence of a specific subset, iNKT2/iNKT17 presence in peripheral tissues has instead been frequently inferred, either by chemokine/ cytokine receptor expression or by their cytokine secretion profile. iNKT17 cells, in particular, were initially identified as IL-17 producing iNKT cells within the NK1.1<sup>−</sup> population by several groups (117, 118). Thereafter, using RORγt-GFP reporter mice, a unique population of iNKT cells was identified in the thymus that was dependent on RORγt for secretion of IL-17 (119). Since RORγt expression is strongly correlated with expression of the chemokine receptor CCR6, iNKT17 cells are specifically directed to the skin (120, 121). Additionally, expression of this chemokine receptor also endows some iNKT17 cells to enter lymph nodes as they are enriched in peripheral lymph nodes compared to the other sub-lineages. Similar to CCR6, expression of CD103 is also high on iNKT17 cells, leading to preferential retention of these cells in the skin, where epithelial cells express the CD103 ligand, E-cadherin (121, 122). iNKT17 cells also uniformly express high levels of the protein Syndecan-1 (CD138), although the reason for why they express this is unknown (123).

Insight into iNKT2 cell localization, however, has been further hindered by the lack of unique markers defining this subset. Unlike iNKT1 and iNKT17 cells, cytokine secretion is insufficient to specifically identify iNKT2 cells since IL-4 is also secreted by iNKT1 cells. Additionally, while iNKT2 cells express high levels of GATA-3, iNKT1, and iNKT17 cells also express this transcription factor, albeit at slightly lower levels (19). And finally, although expression of the cytokine receptor IL-17RB (specific for the cytokine IL-25) on iNKT cells has been demonstrated to enrich for IL-4/IL-13-secreting cells, iNKT17 cells also express this receptor, thereby preventing the use of this marker to specifically distinguish iNKT2 cells in tissues (72).

Through the use of reporter mice and transcription factor staining, a recent study has resolved these ambiguities by shedding substantial light on iNKT cell tissue distribution as well as location within tissues (77). In this seminal study, iNKT cell subsets were identified by their transcription factor expression and analyzed in many different tissues. Additionally, by developing a technique called histocytometry, the authors were able to identify the intra-tissue localization of the iNKT cell subsets. For example, it can now be appreciated that approximately 70% of the thymic iNKT cells, irrespective of subset, reside in the medullary space. This could be due to greater accessibility to homeostatic/survival cytokines (IL-15 for iNKT1 and IL-25 for iNKT2/iNKT17) in the medulla as well as chemokine-mediated trafficking. Remarkably, the relative iNKT subset distribution within tissues is not equivalent across different strains of mice as evidenced by strain-specific iNKT cell subset distribution patterns (77). For example, skindraining lymph nodes were largely enriched for iNKT17 cells in the NOD background and to a lesser extent in B6 mice. However, iNKT2 cells were the principal subset present in these lymph nodes in BALB/c mice. Other tissues also similarly contained different ratios of the subsets across the strains. It is currently unclear if this corresponds merely to the proportion of each subset generated in the thymus in different strains, since BALB/c mice generate significantly more iNKT2 cells, or if strain-specific tissue-homing biases also exist.

New subsets of iNKT cells besides the three described here have also been identified. iNKT cells producing IL-10 are abundant in adipose tissues, where they make up approximately 30% of all T cells (124). Acquiring the moniker iNKT10 due to their ability to produce IL-10, these cells express low levels of PLZF and are dependent on the transcription factor E4BP4 for their functional competence (42). Although these cells are thymically derived as they are absent in adipose tissues of athymic nude mice, they could not be identified in detectable numbers in the thymus of a WT mouse (42). However, in mice expressing a transgene with a modified TCRβ chain that results in fewer iNKT cells due to improper signaling, more iNKT10-like cells were observed in the thymus that preferentially homed to adipose tissue (125). Currently, though, how they arise in a WT mouse is unknown. Therefore, it is possible that one of the three thymic subsets gives rise to this new subset that differentiates in the periphery. What and how specific cells home and differentiate within adipose tissue is uncertain. It is conceivable that due to their expression of T-bet and ability to produce IFNγ after stimulation with PMA/ ionomycin, they are cells that deviate from the iNKT1 lineage due to the adipose tissue microenvironment (42).

Another subset that has also received attention of late is the iNKTFH subset, which, analogous to the conventional TFH population, expresses Bcl6 and helps in antibody class-switching and somatic hypermutation (43, 45). This population was initially described in secondary lymphoid organs upon immunization with antigen in conjunction with αGC, prompting these cells to form stable contacts with B cells and induce germinal centers through the secretion of IL-21. Since this subset has been found in the spleen and iNKT1 cells are also found in higher numbers in the B cell zone (77), perhaps iNKTFH cells represent another branch-off subset from the iNKT1 lineage. This would suggest that iNKT1 cells are somewhat plastic in the periphery and can adopt other fates based on the inflammatory cues they receive.

#### iNKT SUBSET FUNCTIONS AT STEADY STATE

Although T cells commonly circulate throughout the host body, so, they can properly survey all sites for any perturbations, many cells also establish long-term residency in various tissues (126). After a primary immune response has been cleared, a proportion of the antigen-specific cells are retained in the tissue to guarantee a faster response in the future. In the absence of any immune response, however, these cells are not quiescent and sessile but rather dynamically interact with other cells in the tissues to shape their microenvironment in crucial ways. Perhaps owing in part to their memory phenotype, iNKT cells similarly establish longterm residency in several different tissues (42, 94, 127). Beyond that, even without establishing residency, they play important roles in maintaining homeostasis even in steady-state conditions. For instance, their effector status enables them to readily secrete cytokines upon stimulation, which can have dramatic consequences for their surroundings (19, 77). Thus, they serve as a rheostat for how nearby cells acquire phenotypes that correspondingly influence tissue equilibrium.

Ample evidence exists that iNKT cells in the thymus skew the thymic microenvironment in substantial ways (**Figure 1**). Importantly, the ratio of the subsets affects the phenotypes of other conventional cells. For example, iNKT2 cells in the thymus affect the phenotype and functionality of CD8<sup>+</sup> T cells. Usually, thymic CD8<sup>+</sup> T cells exhibit naïve characteristics and display antigen-response kinetics that are delayed compared to memory CD8+ T cells. Through the use of IL-4 reporter mice, it was discovered that thymic iNKT2 cells constitutively produce IL-4 (19). This IL-4 conditions the surrounding CD8<sup>+</sup> T cells to upregulate CXCR3 and Eomes and exhibit memory traits (128, 129). These "innate" memory CD8<sup>+</sup> T cells display antigen-response kinetics reminiscent of memory cells despite never having encountered antigen previously (130). In so doing, they can play a major role in combating chronic viral infections by mounting rapid and robust responses (131, 132). Mutant mice with larger numbers of iNKT2 cells compared to wild-type mice or different strains of mice that endogenously produce large numbers of iNKT2 cells consequently have larger numbers of innate memory CD8<sup>+</sup> T cells. For example, only ~15% of the total iNKT cells in 8-weekold B6 mice thymi are iNKT2 cells while similarly aged CBA and BALB/c mice have ~40 and ~50%, respectively (19). The IL-4 produced by these iNKT2 cells has been directly demonstrated to affect the numbers of innate memory CD8<sup>+</sup> T cells, with B6 mice thymi possessing <4% while CBA and BALB/c mice thymi contain ~30 and ~60% innate memory CD8<sup>+</sup> T cells, respectively.

In addition to affecting the CD8<sup>+</sup> T cells, the IL-4 produced at steady state by iNKT2 cells also conditions SIRPα+ thymic dendritic cells (DCs) to upregulate and produce the chemokines CCL17 and CCL22 (19). These chemokines interact with CCR4, also expressed by iNKT2 cells, perhaps implying a positive feedback loop whereby iNKT2 cells are drawn to the medulla by these chemokines where they enact their effects and further ensure their continued presence due to their sustained production of IL-4. Regulatory T cells (Tregs) appear to also be increased in number and proportion by the iNKT2-produced IL-4 (133). These Tregs exhibit more of an activated phenotype and, in fact, have a greater suppressive capacity in an immune response. Recent data have also identified IL-4 as an inhibitory cytokine for early thymic progenitors (ETPs) to commit to the T cell lineage (134). ETPs stimulated through the IL-4 receptor upregulated the myeloidspecific transcription factor C/EBPα, presumably halting their development into T cells. It would be curious to see if mice with a higher frequency of iNKT2 cells had correspondingly fewer ETPs seeding the T cell pool in the thymus. Finally, IL-4 promotes thymic egress of SP4 thymocytes in a S1P/S1PR1-independent manner (135). Although how IL-4 leads to an accumulation of SP thymocytes is currently unknown, it is clear that the pleiotropic effects of IL-4 by iNKT2 cells markedly change the thymic landscape, reinforcing their importance in tissue maintenance.

Significantly fewer MHC-IIhi Aire+ medullary thymic epithelial cells (mTECs) exist within the thymus of CD1d<sup>−</sup>/<sup>−</sup> mice compared to the B6 control mice (105). Aire is a transcription factor exclusively expressed in mTECs that promotes the expression of peripheral tissue antigens and tolerance of developing SP thymocytes. Both central tolerance of SP4 thymocytes and generation of Tregs depends on MHC-II and Aire expression by mTECs. Reduction of the number of cells capable of carrying out these tasks compromises both of these functions (136). mTECs in a CD1d<sup>−</sup>/<sup>−</sup> mouse are enriched for an immature phenotype (MHC-IIlo Aire<sup>−</sup>). Interestingly, this mTEC developmental arrest is critically dependent on RANKL expression by NK1.1<sup>−</sup> cells iNKT cells, suggesting a potential other role for iNKT2 (and possibly iNKT17) cells in the thymus beyond their production of IL-4. It would be of further relevance to identify if BALB/c mice, which have much higher numbers of iNKT2 and iNKT17 cells in the thymus, have an even more profound defect in mTEC maturation in the absence of CD1d than was described in B6 mice.

It remains unclear why iNKT2 cells play such key roles in influencing different thymic compartments when iNKT1 cells have been identified as long-term thymic residents. What role(s), if any, iNKT1 and iNKT17 might have in maintaining thymic homeostasis is currently unknown.

Substantially less evidence exists for iNKT subsets impacting steady-state functions of other tissues. Production of IL-4 by iNKT2 cells continues to condition the peripheral tissues by contributing to the high IgE levels found in the sera of BALB/c mice as well as promoting a proportion of CD4<sup>+</sup> T cells in the mesenteric lymph nodes (mLNs) to constitutively express the activated form of the transcription factor STAT6 (phospho-STAT6) (19, 77). Activated STAT6 translocates to the nucleus from the cytosol and promotes expression of GATA-3, implicating iNKT2 cells in potentially influencing the "TH2-bias" observed in BALB/c mice (137). Beyond this IL-4-mediated role of iNKT2 cells, our understanding of iNKT subset functions at steady-state in peripheral tissues is limited. Parabiosis experiments have determined that iNKT cells establish long-term residency in hosts in the liver and the lung (94, 127). In the liver, based on their expression of NK1.1, the resident cells are largely iNKT1 while expression of IL-17RB suggests an enrichment of iNKT2/ iNKT17 cells in the lung. Although it is possible that the reason for their tissue residency is simply to act as sentinels that kickstart the overall immune response during an infection, tissue-resident lymphocytes quite frequently have roles beyond that. Thus, future experiments where iNKT cells are prevented from accumulating in those tissues, perhaps by conditional deletion of chemokine ligands in those tissues, should help illuminate how iNKT subsets are affecting tissues in non-infectious settings.

Recently, though, the increased attention paid to iNKT10 cells has uncovered some interesting functions of these cells in maintaining adipose tissue homeostasis. Experiments conducted using parabiotic mice demonstrated that iNKT10 cells establish long-term residency in adipose tissue where they support an immunosuppressive environment (42). Upon stimulation, over half of them secrete IL-10, which helps induce an anti-inflammatory "M2" macrophage phenotype. Moreover, in contrast to other peripheral iNKT cell subsets, these cells produce high amounts of IL-2 upon stimulation. This, in conjunction with the IL-10, also promotes Treg expansion with a highly suppressive phenotype. This supports the idea that iNKT cells in the adipose tissue might also be producing these two cytokines at steady state, but this remains to be formally demonstrated. Although many iNKT10 functional features have primarily been uncovered by stimulating these cells with αGC, the cells also express high levels of PD-1 and *Nr4a1* even at steady state. This could indicate that the iNKT cells perhaps receive continuous TCR-mediated signals in the adipose tissue (42). Indeed, adipocytes themselves display high levels of CD1d molecules. Yet, the nature of the lipids that might be presented to iNKT10 cells by adipocytes remains to be discovered.

# iNKT SUBSETS IN IMMUNE RESPONSES

Because of their varied responses, iNKT cells have been demonstrated to be involved in myriad immune responses in which they can be either protective or pathogenic (138, 139). In mice infected with *Streptococcus pneumoniae*, iNKT cells produce IFNγ within hours of infection (140, 141). Preventing iNKT cells from getting activated by using an antibody that blocks CD1d recognition by iNKT TCRs significantly increased bacterial loads, suggesting that iNKT cell activation contributed to bacterial clearance. Similar findings have been observed in other models in which mice have been infected with *Pseudomonas aeruginosa* or *Mycobacterium tuberculosis*, where iNKT cell deficiency also correlated strongly with increased bacterial burdens, hinting that iNKT cells are perhaps involved in helping clear different kinds of pathogenic bacteria (142–144). When mice deficient in iNKT cells were injected with fibrosarcoma cells, tumor progression was inhibited significantly only upon transfer of iNKT cells (145). Yet again, this protective effect was evident only when the recipient mice expressed CD1d, perhaps implying that the fibrosarcoma cells expressed lipids capable of activating iNKT cells when presented by CD1d. On the other hand, in a model of implanted osteosarcoma, 88% of CD1d<sup>−</sup>/<sup>−</sup> mice rejected the tumors compared to only 24% of WT mice (146). The reasons for the contradicting roles of iNKT cells in tumor models remain unclear. Finally, transferring iNKT cells into the diabetes-prone NOD mouse conferred resistance to diabetes and, in one study, reduced the incidence of diabetes from 90 to 10% (147). In contrast, anti-CD1d treatment of (NZBxNZW)F1 mice led to increased protection from lupus (148, 149). Indeed, transferring iNKT cells from (NZBxNZW) F1 mice into healthy recipients was sufficient to transfer disease (150). Thus, iNKT cells can modulate the course of the immune response in a variety of manners, depending on the models being studied.

Although iNKT cell responses have been characterized in different diseased-state conditions, the specific iNKT cell subsets contributing to the response are largely unknown. Usually, the contribution of iNKT cells to an immune response is determined through the use of a CD1d<sup>−</sup>/<sup>−</sup> mouse model and/or a TRAJ18<sup>−</sup>/<sup>−</sup> mouse model, both of which lack iNKT cells. However, both of these mouse models have drawbacks. The original TRAJ18<sup>−</sup>/<sup>−</sup> mouse was generated by introducing a neomycin resistance gene into the *Traj18* locus (151). Interestingly, these mice lack approximately 60% of their overall TCR repertoire due to an inability to express TCR rearrangements involving TRAJ genes upstream of *Traj18* (152), potentially due to the presence of the neomycin resistance gene (153). Thus, these animals not only lack iNKT cells but also a substantial proportion of their conventional TCR repertoire, potentially obfuscating some of the findings discovered in studies using these mice. Repeating these experiments in mice where *Traj18* was deleted without the presence of a neomycin resistance gene should help clarify the original results (154–156). In the case of the CD1d<sup>−</sup>/<sup>−</sup> mouse model, new data have revealed that one of the four widely distributed knockout strains (157) continues to possess a small number of iNKT cells (158). Therefore, this specific knockout strain cannot be considered iNKT cell-deficient mice and conclusions obtained using these mice should be reassessed and instead be reevaluated using mice in which iNKT cells are completely absent (159, 160). Besides the use of these mouse models, the iNKT cell contribution to an immune response is further characterized only by the cytokines that affect the progression of the disease, frequently IFNγ and IL-4, but not the phenotype of the iNKT cells secreting those factors. Although this cytokine-secretion profile is more indicative of the specific subsets involved, it is often insufficient since iNKT1 cells are also capable of producing IL-4. Thus, a greater effort needs to be put forth to identify the subsets involved in any disease based on not just their cytokines produced but also by the transcription factors expressed.

A few studies have shed some light on the roles of specific iNKT cell subsets in diseases, albeit indirectly. One study used a transplantable tumor model to determine that CD4<sup>+</sup> T cells negatively regulated tumor rejection. Upon further examination, it was discovered that CD4<sup>+</sup> iNKT cells were the primary source of IL-13, creating an immunosuppressive environment that prevented tumor rejection (161). When these iNKT cells were depleted, either through the use of depleting antibodies targeting CD4 or CD1d<sup>−</sup>/<sup>−</sup> mice, tumors were rejected at a significantly higher frequency. It is possible that the responding iNKT cells were iNKT2 cells due to their production of IL-13 and the fact that iNKT2 cells would have been abundant in the BALB/c mice in which these experiments were conducted (19, 72, 162). In addition, iNKT2 cells are enriched in the CD4<sup>+</sup> population (19, 63), further lending credence to this idea, although it is perhaps worth revisiting these experiments using currently available tools.

Interestingly, there appears to be a tissue-specific bias associated with iNKT cells capable of mediating tumor rejection. When bulk iNKT cells from the liver were transferred into TRAJ18<sup>−</sup>/<sup>−</sup> mice harboring a sarcoma, they were capable of halting tumor progression (163). However, bulk splenic and thymic iNKT cells were not similarly capable of rejecting the tumor growth. It was further determined that the DN hepatic iNKT cells were significantly better at tumor rejection compared to the CD4<sup>+</sup> hepatic iNKT cells. This latter finding could perhaps be because upon stimulation, CD4<sup>+</sup> iNKT cells tend to produce more TH2 cytokines that would be immunosuppressive compared to the TH1 cytokines that DN iNKT cells are more biased to produce (79). Indeed, when IL-4−/− iNKT cells, irrespective of their tissue origin, were transferred into mice with tumors, they more potently rejected tumors compared to WT iNKT cells (163). Curiously, however, CD4<sup>+</sup> and DN iNKT cells from different tissues all produced similar levels of IFNγ and IL-4, suggesting that although IL-4 does have an impact on tumor rejection, other differences between the subsets and their tissue-origin also likely affect the functions of the iNKT cells *in vivo*. The specific functional subsets associated with these findings, though, remain unknown because iNKT1 cells, for example, can be found in both the CD4<sup>+</sup> and DN compartments (63). Thus, in light of what is known today, these experiments bear repeating with the use of transcription factor staining to identify any differences between the various sources of iNKT cells. It would be especially interesting to identify gene signature differences between the same iNKT subsets but from different tissues to understand if certain tissues impose functional variations.

In other studies, iNKT cells were also discovered to be relevant in airway hyperreactivity (AHR). One of the hallmarks of asthma, AHR features eosinophilia in the airways, enhanced mast cell growth, and increased levels of serum IgE (164). Conventional TH2 cells play a major role in exacerbating antigen-induced AHR. However, it appears that iNKT cells can prime the immune system initially to bias the response toward a TH2 phenotype. In fact, in several models of AHR, iNKT-deficient mice do not develop AHR (165, 166). Further, intratracheal administration of a lipid agonist and a protein antigen strongly activated the pulmonary iNKT cells to prime CD4<sup>+</sup> cells specific for the protein antigen to polarize into TH2 cells (127). Indeed, IL-4/IL-13 produced by iNKT cells was discovered to be fundamental for mice to succumb to AHR in these studies, although this remains controversial since a later study found that iNKT cells were dispensable for airway inflammation (167). Despite this, the iNKT cells secreting the type 2 cytokines were subsequently identified as expressing IL-17RB, providing some evidence to suggest that the cells potentially promoting AHR were possibly iNKT2 cells (168, 169). Only IL-17RB<sup>+</sup> cells, usually expressed by iNKT2 and iNKT17 cells (72), were capable of recapitulating the symptoms upon transfer into a iNKT cell-deficient mouse by secreting IL-4 and IL-13 (168, 169). Interestingly, in a different model of AHR in which mice were exposed to ozone instead of an allergen/antigen, IL-17 production by iNKT cells was also required in addition to IL-4/IL-13, perhaps implicating that iNKT17 cells also can contribute to AHR in certain contexts (170). The IL-17 produced by iNKT cells led to increased neutrophilia instead of eosinophilia in the airways and has been demonstrated in a separate study to be dependent on c-Maf, a transcription factor also involved in promoting the proper function of TH17 cells (171–173).

Administration of αGC intravenously in mice can activate the vascular-localized iNKT cells. In this fashion, the hepatic and the red-pulp splenic iNKT cells, which are primarily iNKT1 cells, respond within minutes by producing IFNγ and IL-4. Serum increases of both these cytokines can easily be detected in these conditions (77) and the IL-4 secreted under these conditions appears to have long-range effects as demonstrated by the increased phosphorylation of STAT6 in CD4<sup>+</sup> T cells in other tissues, such as LNs, despite the iNKT cells in those tissues remaining unstimulated. Thus, blood-borne pathogens that are capable of activating iNKT cells could possibly activate iNKT1 cells due to their localization that could then condition T cells in distal tissues. Analogously, since iNKT2 cells are present in high numbers in the mLNs of certain mouse strains, oral administration of αGC largely activated these cells and caused them to secrete IL-4 in large quantities (77). However, perhaps due to a lack of proximity to the circulation, the IL-4 produced in this setting had primarily local effects, with only the CD4<sup>+</sup> T cells in mLNs increasing their phospho-STAT6 levels while T cells in other tissues were unaffected.

Bacteria express their own lipids, some of which might serve as stimulatory antigens to iNKT cells. Viruses, however, hijack host machinery for their own purposes and thus, are devoid of any lipids themselves and thought to not activate iNKT cells directly. However, viral infections can lead to upregulation of CD1d by triggering toll-like receptors (TLRs) (174). Additionally, infection can also lead to activation of hypoxia-inducible factor, which in turn could alter the lipid metabolism and allow antigenic selflipids to be presented to iNKT cells (175). Thus, viral infections could lead to activation of iNKT cells in a CD1d-dependent manner (176). Alternatively, activation of innate cells such as DCs through TLRs could prompt them to secrete pro-inflammatory cytokines such as IL-12 and IL-18 that consequently activate iNKT cells (177, 178). These activated iNKT cells can secrete IFNγ that promotes an antiviral response (179, 180). Thus, iNKT cells can participate in viral infections, potentially in a protective role. However, the specific subsets involved in viral clearance are unknown. Although the production of IFNγ by iNKT cells strongly suggests that the subset involved is the iNKT1 subset, this remains to be formally demonstrated.

Interestingly, a new study has highlighted that iNKT cells influence humoral immunity during Influenza A virus infection (181). A previous study had identified iNKT cells as important in curbing myeloid-derived suppressor cell (MDSCs) function in influenza infection (182). The MDSCs in influenza-infected mice suppressed influenza-specific immune responses, leading to high titers of the virus. In a CD1d-dependent manner, iNKT cells were able to restrict the activity of MDSCs and consequently boost the immune responses directed against influenza. Thus, a role for iNKT cells in combating influenza virus infection had already been established. In the recent study, the authors focused on how iNKT cells affect B cell responses in influenza infection. These cells influence B cell germinal center formation and antibody class switching despite not being iNKTFH cells. The iNKT cells are the primary secretors of IL-4 early during the infection and express CXCR3, suggesting that they are possibly iNKT1 cells. CD1d-mediated interactions with CD169<sup>+</sup> macrophages were critical for the production of IL-4 by the iNKT cells. This response underscores a novel role that iNKT1 cells potentially play in mounting an immune response against viral pathogens. Although why iNKT1 (if the cells are indeed iNKT1) cells are producing IL-4 and the iNKT2 cells are not is unknown. It could be due to a possible abundance of iNKT1 cells in the mediastinal LNs or it could be that the macrophages present lipids on CD1d only capable of activating iNKT1 cells. Confirming the specific subset involved in this immune response and why these cells are preferentially activated is paramount.

More recently, with iNKT10 cells entering the fold, a new role has been added to the growing list of iNKT cell roles (183). In a model of diet-induced obesity in which mice were fed with high-fat diets (HFD), iNKT cells were depleted from adipose tissues, although this was reversible once their diets were switched to standard fat diets (124). Mice lacking iNKT cells fed HFD weighed more, had large adipocytes, elevated fasting blood glucose levels, and increased insulin resistance. Furthermore, there was an increased infiltration of proinflammatory macrophages into the adipose tissues, an important intermediary step in the inflammation and pathogenesis associated with obesity. As mentioned previously, it is unknown what cytokines the adipose tissue-resident iNKT cells secrete at steady state to maintain healthy adipocytes. However, these iNKT10 cells are different from other tissue localized iNKT cells because of their expression of E4BP4 (42). When iNKT cells transfected with E4BP4 were stimulated, they secreted more IL-10 than E4BP4<sup>−</sup> stimulated iNKT cells, directly correlating E4BP4 to IL-10 production in iNKT cells. In addition, upon stimulation, the adipose-resident iNKT cells were capable of expanding Tregs in the adipose tissue in an IL-2-dependent manner and the adipose Treg population is substantially reduced in iNKT cell-deficient mice (42). Thus, the immunoregulatory role that the iNKT10 subset plays in adipose tissue to prevent obesity-related illnesses could be due to direct secretion of IL-10 (and possibly IL-2) at steady state.

#### HUMAN iNKT CELL SUBSETS

In humans, iNKT cell numbers are substantially more variable compared to the inbred mouse strains routinely used. Usually, their frequencies are lower in human blood compared to mouse blood (around 0.01–0.1% compared to 0.2–0.5% in mice), and their frequencies are more variable in other tissues when compared to the analogous mouse tissues (184, 185). Humans instead have larger proportions of other innate-like T lymphocyte populations, such as mucosal-associated invariant T cells and the group I CD1-restricted T cells (12). Despite the reduced frequencies, human iNKT cells can be isolated from healthy individuals and patients and analyzed for function and phenotype. Unfortunately, a rigorous manner of identifying human iNKT cells from clinical samples has not always been consistently employed. Many times, the cells were identified by staining for human NK-cell markers such as CD56 and CD161 (the human counterpart of NK1.1) but as in mice, these markers are also expressed on other T cell populations (186, 187). In other studies, iNKT cells were identified by using antibodies targeting the TCRβ chain used by these cells (TRBV25), but this is also problematic since other T cell populations also express this TCRβ chain (188). More recently, iNKT cells have been identified either through the use of αGCloaded human CD1d tetramers or by using an antibody targeting the invariant TCRα rearrangement unique to iNKT cells (189, 190). Both tools have provided greater resolution into understanding iNKT cell function in humans.

Invariant natural killer T cells in humans can be broadly categorized as DN, CD4<sup>+</sup>, and a small percentage of CD8<sup>+</sup> cells (185). There appears to be some functional conservation of these subpopulations between species since the DN cells tend to have a TH1 bias while the CD4<sup>+</sup> cells have a TH2 bias, although the CD4<sup>+</sup> cells are also capable of secreting TH1 cytokines (184, 191). This suggests that perhaps human iNKT1 cells are present in both the CD4<sup>+</sup> and DN fractions while iNKT2 cells are primarily present within the CD4<sup>+</sup> fraction. Further evidence to support this hypothesis stems from the fact that the DN cells express higher levels of several NK receptors compared to the CD4<sup>+</sup> cells, similar to how murine iNKT1 cells primarily express the NK receptors (185, 191).

Interestingly, it has been shown that human iNKT cells also express high levels of PLZF compared to other T cell populations (37, 38). Additionally, the CD4<sup>+</sup> iNKT cell population appears to express higher levels of PLZF as identified by mRNA levels, perhaps because more iNKT2 cells are present within this population. Indeed, iNKT cell numbers and phenotype appeared to be significantly altered when a patient with biallelic PLZF deficiency was analyzed (192). Human iNKT cells also require SLAM receptor-mediated signals for proper development because humans lacking the adapter SAP lack any observable iNKT cells (193). Despite these studies, whether or not functional iNKT cell subsets follow a similar developmental path in humans and mice has not been formally addressed.

The identification of functionally distinct human iNKT cell subsets with differential expression of master transcription factors, similar to what is observed in mice, is currently limited. Instead, the cells are usually sub-divided based on their cytokinesecretion profile and/or their expression of the CD4 coreceptor. For example, iNKT cells found in the cord blood of humans appear to have an intrinsic bias to secrete IL-17 and cannot produce IFNγ (194). Furthermore, a RORγt inhibitor selectively impaired IL-17 production by iNKT cells in different tissues, suggesting that some iNKT cells could indeed constitutively express RORγt (although this was not formally tested) that endows them with the ability to secrete IL-17 upon stimulation (195). Different iNKT subsets identified by their differential expression of CD4 could induce secretion of different isotypes of antibodies by B cells. In particular, CD4<sup>+</sup> iNKT cells were unique in their ability to induce expansion of a CD1dhi CD5hi Breg population (196). These cells, however, did not express CXCR5 or PD-1 at high levels when placed in co-culture with the B cells, suggesting that they are not likely to be iNKTFH cells. As in mice, immunosuppressive functions were associated with CD4<sup>+</sup> iNKT cells in various tumor settings, further implicating a functional distinction between CD4<sup>+</sup> and CD4<sup>−</sup> iNKT populations (197). In other models of autoimmunity, iNKT cell numbers were found to be reduced and functionally impaired in their ability to secrete IL-4, although whether this

reflects the loss of a specific subset is unknown (138). Reduction in iNKT cell numbers was also observed in obese patients in the peripheral blood, and these numbers appeared to increase once the patients underwent bariatric surgery (124). Whether these cells are the human equivalents of the murine iNKT10 population remains to be explored. Thus, overall, there are primarily tidbits of information regarding functional diversity in human iNKT cells without a cohesive paradigm comparable to the one established in mice. Future work should focus on understanding iNKT cell function in diseased states with increased granularity, with special attention paid to linking function to transcription factors expressed by different cells.

# CONCLUDING REMARKS

Subset differentiation of iNKT cells is a complex and multifaceted process. Despite this complexity, the final subsets are surprisingly similar phenotypically to other cell types belonging to their corresponding functional group. For example, there is a striking

# REFERENCES


similarity between iNKT1 cells, NK cells, TH1 cells, and ILC1 cells (198). Similarly, iNKT2 and iNKT17 cells share similarities to their γδ and ILC counterparts. What then makes iNKT cell subsets special when other cells occupy similar niches and respond similarly? Two different aspects provide iNKT cells with a unique ability to influence the immune response. First, their ability to recognize lipids in an antigen-specific manner allows these cells to sample an antigen space that would otherwise be unmonitored by conventional T cells. Second, the kinetics of their responses to antigenic stimulation allow the iNKT subsets to rapidly skew the course of the immune response in directed ways. By establishing an initial path for the immune response, iNKT cells have the potential to dictate how downstream adaptive cells are polarized and, consequently, how they respond. Thus, understanding the functional diversity within iNKT cells is essential to be able to manipulate the immune system. By gaining a greater understanding about iNKT cell subsets and their functions, one can hope to target specific subsets in an effort to influence various immune responses in the future.

## AUTHOR CONTRIBUTIONS

Both authors wrote and edited the manuscript.

## FUNDING

This work was supported by National Institutes of Health Grants AI121761 and AI124076 (to LG).


acid receptor-related orphan receptor (gamma)t+ and respond preferentially under inflammatory conditions. *J Immunol* (2009) 183(3):2142–9. doi:10.4049/jimmunol.0901059


and NK cells in a PLZF deficient patient. *PLoS One* (2011) 6(9):e24441. doi:10.1371/journal.pone.0024441


**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 SJ and handling Editor declared their shared affiliation.

*Copyright © 2018 Krovi and Gapin. 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.*

*Dominic Paquin-Proulx1\*, Priscilla R. Costa2, Cassia G. Terrassani Silveira2, Mariana P. Marmorato2 , Natalia B. Cerqueira2 , Matthew S. Sutton3 , Shelby L. O'Connor3 , Karina I. Carvalho4, Douglas F. Nixon1† and Esper G. Kallas2†*

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Sebastian Joyce, Vanderbilt University, United States Salah Mansour, University of Southampton, United Kingdom*

#### *\*Correspondence:*

*Dominic Paquin-Proulx dpaquin\_proulx@gwu.edu*

> *† Co-senior authors.*

#### *Specialty section:*

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

*Received: 25 March 2018 Accepted: 05 June 2018 Published: 19 June 2018*

#### *Citation:*

*Paquin-Proulx D, Costa PR, Terrassani Silveira CG, Marmorato MP, Cerqueira NB, Sutton MS, O'Connor SL, Carvalho KI, Nixon DF and Kallas EG (2018) Latent Mycobacterium tuberculosis Infection Is Associated With a Higher Frequency of Mucosal-Associated Invariant T and Invariant Natural Killer T Cells. Front. Immunol. 9:1394. doi: 10.3389/fimmu.2018.01394*

*1Department of Microbiology, Immunology and Tropical Medicine, The George Washington University, Washington, DC, United States, 2School of Medicine, University of São Paulo, São Paulo, Brazil, 3Department of Pathology and Laboratory Medicine, University of Wisconsin-Madison, Madison, WI, United States, 4Hospital Israelita Albert Einstein, Instituto Israelita de Ensino e Pesquisa, São Paulo, Brazil*

Increasing drug resistance and the lack of an effective vaccine are the main factors contributing to *Mycobacterium tuberculosis* (Mtb) being a major cause of death globally. Despite intensive research efforts, it is not well understood why some individuals control Mtb infection and some others develop active disease. HIV-1 infection is associated with an increased incidence of active tuberculosis, even in virally suppressed individuals. Mucosal-associated invariant T (MAIT) and invariant natural killer T (iNKT) cells are innate T cells that can recognize Mtb-infected cells. Contradicting results regarding the frequency of MAIT cells in latent Mtb infection have been reported. In this confirmatory study, we investigated the frequency, phenotype, and IFNγ production of MAIT and iNKT cells in subjects with latent or active Mtb infection. We found that the frequency of both cell types was increased in subjects with latent Mtb infection compared with uninfected individuals or subjects with active infection. We found no change in the expression of HLA-DR, PD-1, and CCR6, as well as the production of IFNγ by MAIT and iNKT cells, among subjects with latent Mtb infection or uninfected controls. The proportion of CD4− CD8+ MAIT cells in individuals with latent Mtb infection was, however, increased. HIV-1 infection was associated with a loss of MAIT and iNKT cells, and the residual cells had elevated expression of the exhaustion marker PD-1. Altogether, the results suggest a role for MAIT and iNKT cells in immunity against Mtb and show a deleterious impact of HIV-1 infection on those cells.

Keywords: mucosal-associated invariant T cells, invariant natural killer T cells, *Mycobacterium tuberculosis*, HIV-1, CCR6

**Abbreviations:** Mtb, *Mycobacterium tuberculosis*; MAIT, mucosal-associated invariant T; iNKT, invariant natural killer T; NHP, non-human primate; PBMC, peripheral blood mononuclear cell; ART, antiretroviral treatment.

# INTRODUCTION

*Mycobacterium tuberculosis* (Mtb) infection is a major cause of death globally. Several factors contribute to this phenomenon, including increased drug resistance (1) and the absence of a highly effective vaccine (2). In the majority of Mtb-infected individuals, there are no clinical signs of tuberculosis (TB), and the infection will be eliminated or remain latent (3). Immunocompromised individuals, including those infected with HIV-1, have a high risk of developing active Mtb infection. CD4+ and CD8+ T cells are believed to be important for immune control against Mtb (4, 5), but there is no known biomarker that can predict the progression from latent to active TB.

Innate-like unconventional T cells can rapidly produce cytokine after antigen exposure, and they have been implicated in the defense against Mtb (6). Mucosal-associated invariant T (MAIT) cells recognize pyrimidine intermediates derived from the riboflavin biosynthesis pathway (7, 8) presented by MR1 (9). Invariant natural killer T (iNKT) cells recognize glycolipids presented by CD1d (10, 11). Both MAIT and iNKT cells have been shown to directly recognize Mtb-infected cells (12, 13), and numerous studies have shown that their frequency is reduced in blood during active Mtb infection (12, 14–20). In humans, there is indirect evidence that MAIT and iNKT cells could play a role in controlling Mtb infection. For example, MAIT cells from tuberculous pleural effusions were shown to produce more IFNγ, IL-17, and granzyme B after stimulation with Mtb antigens (21). Both MAIT and iNKT cells are depleted during infection with HIV-1 (22–24) and HTLV-1 (25, 26). Infection with both of these pathogens is associated with a greater risk of developing active Mtb infection (27–30).

Results from non-human primate (NHP) animal models of Mtb infection suggest that CD8+ iNKT cells can play a protective role in preventing Mtb pathology (31), and MAIT cells were activated following BCG vaccination and Mtb infection (32). In mouse models, both iNKT and MAIT cells reduced bacterial burden following Mtb infection (13, 33). However, limited knowledge is available on the role of MAIT and iNKT cells in controlling Mtb infection in humans. Contradicting results have been reported regarding MAIT cell frequency in blood during latent Mtb infection (12, 17), and only one study has measured the frequency of iNKT cells in latent Mtb infection (16). A more detailed characterization of MAIT and iNKT cells in latent Mtb infection is still needed to better understand their role in immunological control of Mtb.

In the current confirmatory study, we evaluated MAIT and iNKT cell frequency, phenotype, and functionality in uninfected individuals and subjects with latent or active Mtb infection with and without HIV-1 infection. We found that both cell types were increased in subjects with latent Mtb infection compared with uninfected individuals and subjects with active Mtb infection. Latent Mtb infection was further associated with an increase in the proportion of CD4− CD8+ MAIT cells. Active Mtb infection was associated with elevated surface expression of the activation marker HLA-DR on both MAIT and iNKT cells, as well as of the exhaustion marker PD-1 on iNKT cells. No significant difference was observed between the groups in the production of IFNγ following *in vitro* stimulation of MAIT and iNKT cells.

### MATERIALS AND METHODS

#### Ethics Statement

HIV-1-uninfected (*n*= 41, age range 23–70) (**Table 1**) and -infected (*n* = 16, age range 22–68) (**Table 2**) subjects were enrolled in the study. There was no significant difference in age between any of the subgroups. Definition of Mtb infection was based on a positive PPD skin reaction above 10 mm and/or a positive TB-Spot test, in the absence (latent) or presence (active) of clinical signs or symptoms of TB. The study was approved by the University of São Paulo institutional review board (CAPPesq), and written informed consent was provided by all participants according to the Declaration of Helsinki. All samples were anonymized.

Table 1 | Demographics and *Mycobacterium tuberculosis* (Mtb) status of HIV-1-uninfected subjects.


Table 2 | Demographics and *Mycobacterium tuberculosis* (Mtb) status of HIV-1-infected subjects.


#### Sample Collection

Peripheral blood mononuclear cells (PBMCs) were isolated by density-gradient sedimentation using Ficoll-Paque (Lymphoprep, Nycomed Pharma, Oslo, Norway). Isolated PBMCs were washed twice in Hank's balanced salt solution (Gibco, Grand Island, NY, USA), and cryopreserved in RPMI 1640 (Gibco), supplemented with 20% heat inactivated fetal bovine serum (FBS; Hyclone Laboratories, Logan, UT, USA), 50 U/ml of penicillin (Gibco), 50 µg/ml of streptomycin (Gibco), 10 mM glutamine (Gibco), and 7.5% dimethylsulfoxide (Sigma, St. Louis, MO, USA). Cryopreserved cells from all subjects were stored in liquid nitrogen until used in the assays.

#### Flow Cytometry and Antibodies

Cryopreserved specimens were thawed and washed, and counts and viability were assessed using the Countess Automated Cell Counter system (Invitrogen, Carlsbad, CA, USA). Cells were washed and stained in Brilliant Violet Stain Buffer (BD Biosciences, San Jose, CA, USA) at room temperature for 15 min in 96-well V-bottom plates in the dark. Samples were then washed and fixed using Cytofix/Cytoperm (BD Biosciences) before flow cytometry data acquisition. Intracellular staining was performed in Perm/ Wash Buffer (BD Biosciences). Monoclonal antibodies used in flow cytometry: CD3 AF700, CD3 PerCP-Cy5.5 (both clone UCHT1), CD4 BV605 (clone RPA-T4), CD8 BV711 (clone RPA-T8), CD161 BV421 (clone DX12), CCR6 BV786 (clone 11A9), HLA-DR APC (clone L243), IFNγ APC (clone B27), and PD-1 PE-Cy7 (clone EH12.1) were all from BD Biosciences, TCRα24 FITC (clone C15) and TCR Vβ11 PE (clone C21) were from Beckman Coulter (Indianapolis, IN, USA), and TCR Vα7.2 PercP-Cy5.5 (clone 3C10) was from BioLegend (San Diego, CA, USA). Live/dead aqua fixable cell stain was from Life Technologies (Eugene, OR, USA). Data were acquired on a BD LSRFortessa instrument (BD Biosciences) and analyzed using FlowJo Version 9.8.5 software (TreeStar, Ashland, OR, USA).

#### Functional Assays

Mucosal-associated invariant T cell function was determined *in vitro* using paraformaldehyde-fixed *Escherichia coli* stimulation (One Shot Top10, Life Technology, multiplicity of exposure 10) in the presence of 1.25 µg/ml anti-CD28 mAb (clone L293, BD Biosciences) (34). *E. coli* was fixed for 5 min in 1% paraformaldehyde. PBMCs were further cultured for 24 h at 37°C/5% CO2 in RPMI medium supplemented with 10% FBS. Monensin (Golgi Stop, BD Biosciences) was added during the last 6 h of the stimulation. iNKT cell function was determined *in vitro* using α-GalCer (KRN7000, Enzo Life Science, Farmingdale, NY, USA) at 100 ng/ml. PBMCs were further cultured for 8 h at 37°C/5% CO2 in RPMI medium supplemented with 10% FBS. Monensin (Golgi Stop, BD Biosciences) was added during the last 6 h of the stimulation.

# Statistical Analysis

All statistical analyses were performed using Graph Pad Prism version 6.0 h for Mac OSX (GraphPad Software, La Jolla, CA, USA). Results were tested for normal distribution, and appropriate ANOVA or Kruskal–Wallis test was used for comparison between groups. Mann–Whitney *U*-test was used for comparison between HIV-1-uninfected and infected subjects. *p*-Values ≤ 0.05 were considered statistically significant.

#### RESULTS

#### Increased Frequency of MAIT and iNKT Cells in Latent Mtb Infection

First, to confirm previous studies, we evaluated the frequency of MAIT and iNKT cells in a cohort of patients with latent (age range 23–63, median 39) or active Mtb infection (age range 26–58, median 45) compared with uninfected controls (**Table 1**; **Figure 1A**, age range 26–70, median 42). We found no difference in the frequency of both cell types between active Mtb infection and uninfected controls (**Figures 1B,C**). However, there was a significant increase in the frequency of MAIT and iNKT cells in individuals with latent Mtb infection compared with active Mtb infection, or uninfected controls. Next, we investigated if there was a change in the CD4+ and CD8+ subset distribution of MAIT and iNKT cells between the groups. As expected, MAIT cells were mostly CD4− CD8+ and CD4− CD8−, whereas iNKT cells were mostly CD4+ CD8− and CD4− CD8−, in all three groups. Furthermore, we found that there was an increase in the proportion of CD4− CD8+ MAIT cells in the latent Mtb infection group compared with the control group (**Figure 1D**). No change in the distribution of iNKT cell subsets was observed between the groups. Our results suggest that latent Mtb infection is associated with an increased MAIT and iNKT cell frequency.

#### MAIT and iNKT Cells Are Activated in Active, but Not Latent, Mtb Infection

Previous studies have reported that MAIT and iNKT cells have increased expression of activation and exhaustion markers, but reduced expression of CCR6 in active Mtb infection (15, 16, 32, 35). Therefore, we investigated the expression of HLA-DR, PD-1, and CCR6 in latent, and active Mtb infection (Figure S1 in Supplementary Material). We found increased expression of HLA-DR on MAIT and iNKT cells in active, but not in

latent, Mtb infection (**Figure 2A**; Figure S1A in Supplementary Material). Increased immune activation has been associated with a lower frequency of MAIT cells in chronic viral infections and in conditions associated with inflammation (23, 25, 36). In our cohort, there was no significant association between HLA-DR expression and the frequency of MAIT and iNKT cells in active Mtb infection (Figure S2A in Supplementary Material). Furthermore, PD-1 expression was elevated on iNKT cells, but not on MAIT cells, in active Mtb infection (**Figure 2B**; Figure S1B in Supplementary Material). CCR6 expression was reduced only in active Mtb infection for both MAIT and iNKT cells (**Figure 2C**; Figure S1C in Supplementary Material). Reduced

CCR6 surface expression could be indicative of ligand binding and an early sign of tissue recruitment and activation. However, there was no significant association between CCR6 expression and the frequency of MAIT and iNKT cells or their HLA-DR expression in subjects with active Mtb infection (Figures S2B,C in Supplementary Material). Our results show that active Mtb infection is associated with increased activation and exhaustion of MAIT and iNKT cells compared with uninfected controls, in contrast to latent Mtb infection.

## Normal MAIT and iNKT Cell IFN**γ** Production in Latent and Active Mtb Infection

The IFNγ response is essential in the control of Mtb (37). Furthermore, MAIT and iNKT cells have been shown to recognize Mtb-infected cells *in vitro* (12, 13). Thus, a high innate T cell IFNγ production could be associated with control of Mtb infection. We evaluated the IFNγ production by MAIT and iNKT cells *in vitro* following antigen stimulation. PBMCs were stimulated with fixed *E. coli* or α-GalCer to activate MAIT and iNKT cells, respectively, and IFNγ production was evaluated by flow cytometry (**Figures 3A,C**). There was a small, albeit non-significant, increase in the production of IFNγ by MAIT and iNKT cells in individuals with latent Mtb compared with uninfected controls and individuals with active Mtb infection. However, there was no significant change in the production of IFNγ between all groups for both cell types (**Figures 3B,D**). There was also no association between the age of the subjects and the production of IFNγ by the MAIT or iNKT cells (*r* = 0.1567, *p* = 0.4544 and *r* = −0.0930, *p* = 0.6585, respectively).

#### Reduced MAIT and iNKT Cells in HIV-1 Infection Is Independent of Mtb Infection

HIV-1 infection is associated with an increased susceptibility to Mtb infection and more severe TB disease, even in individuals on long-term antiretroviral treatment (ART) (29, 30). We enrolled HIV-1-infected subjects without active Mtb infection (**Table 2**, age range 31–68, median 41.5, for Mtb uninfected subjects and age range 22–52, median 43, for individuals with latent Mtb infection) and investigated their MAIT and iNKT cell frequency and phenotype. All but one of the HIV-1-infected individuals were on ART, and the median duration of treatment was 5.5 years. As previously reported, the frequencies of MAIT and iNKT cells were decreased in HIV-1-infected individuals without active Mtb infection compared with HIV-1-uninfected subjects (**Figure 4A**). Similar to HIV-1-uninfected subjects, there was a trend for increased frequencies of MAIT (median 0.46 vs 1.20%) and iNKT cells (median 0.02 vs 0.05%) in HIV-1-infected subjects with latent Mtb infection, compared with the Mtb-uninfected individuals (**Figure 4B**). This increase did not reach statistical significance likely due to the low number of subjects in each group. There was a positive association between the frequencies of MAIT and iNKT cells in uninfected and HIV-1-infected subjects (Figures S3A,B in Supplementary Material). PD-1 expression was increased on MAIT and iNKT cells in HIV-1-infected individuals (**Figure 4C**). Finally, CCR6 levels were reduced on MAIT cells in HIV-1 infection, but not on iNKT cells (**Figure 4D**). Our results indicate that alterations of MAIT and iNKT cells caused by HIV-1 infection are not restored after long-term ART.

# DISCUSSION

Several populations of innate T cells have been proposed to play a role in the immune response against Mtb (6). In this study, we sought to confirm previous investigations of the frequency

(\**p* < 0.05, \*\**p* < 0.01, and \*\*\**p* < 0.001).

latent Mtb (*n* = 7), and subjects with active Mtb infection (*n* = 8) (D).

of MAIT and iNKT cells in latent and active Mtb infections (12, 16–18). We found an increase in their frequencies in latent Mtb infection. Our observation that there was a higher frequency of MAIT cells in latent Mtb infection is in agreement with the results of Gold et al. (12), but in contrast to those of Wong et al. (17). One difference between those studies was how MAIT cells were defined. Wong et al. identified MAIT cells as CD161++ CD8+ T cells whereas Gold et al. identified MAIT cells as CD8+ Vα7.2+ T cells. In our studies, we used CD161++ Vα7.2+ T cells to identify MAIT cells, which includes the CD8− subset of MAIT cells. Latent Mtb infection could induce proliferation of MAIT and iNKT cells by constant low exposure to stimulating antigens. We would expect this to be associated with higher expression of activation markers, but we found no change in HLA-DR expression on MAIT and iNKT cells in latent Mtb infection. Furthermore, NHP studies have shown no change in peripheral MAIT cell frequency following BCG vaccination or Mtb infection (32). Another possibility is that a higher frequency of innate T cells before infection is associated with control of Mtb. Addressing these would require a longitudinal study.

Interestingly, we found that in latent Mtb infection there was an increase in the proportion of CD4− CD8+ MAIT cells but no change in iNKT cell subsets. There is growing interest in understanding the heterogeneity of MAIT cells (38–40). CD8+

MAIT cells have a higher expression of CCR6 compared with the CD8− CD4− subset, and they have higher levels of Granzyme A, Granzyme K, and Perforin compared with the CD4+ subset (38). Thus, CD8+ MAIT cells could have a higher combined capacity to migrate to and kill Mtb-infected cells compared with the other subsets of MAIT cells.

We found that active Mtb infection was associated with increased expression of HLA-DR and decreased expression of CCR6 on MAIT and iNKT cells. Lower CCR6 levels could impair the capacity of the cells to migrate to the lungs. We also found increased PD-1 expression on iNKT cells, but only in active Mtb infection. Notably, there was no change in the IFNγ production of both cell types after antigen stimulation in active Mtb infection. This is in contrast with a previous study that found less IFNγ production from MAIT cells in active Mtb infection (18). It remains to be determined if elevated HLA-DR and PD-1 expression by MAIT and iNKT cells in active Mtb infection are a cause or a consequence of the disease. In this regard, PD-1 expression by conventional CD4+ T cells is needed to prevent immune mediated pathology in response to Mtb infection (41). Chronic activation of MAIT and iNKT cells by direct recognition of Mtb-infected cells and exposure to elevated levels of inflammatory cytokines could lead to this abnormal phenotype.

Figure 4 | Mucosal-associated invariant T (MAIT) and invariant natural killer T (iNKT) cells are reduced in HIV-1 infection. Frequency of MAIT (left panel) and iNKT (right panel) cells in HIV-1-uninfected (*n* = 27) and infected (*n* = 16) subjects without active *Mycobacterium tuberculosis* (Mtb) infection (A). Frequency of MAIT (left panel) and iNKT (right panel) cells in Mtb uninfected (*n* = 10) and with latent Mtb infection (*n* = 6) HIV-1-infected subjects (B). PD-1 expression (C) and CCR6 expression (D) by MAIT (left panel) and iNKT (right panel) in HIV-1-uninfected (*n* = 27) and -infected (*n* = 16) subjects without active Mtb infection. The lines and whiskers represent the median and interquartile range, respectively (\**p* < 0.05 and \*\**p* < 0.01).

*Mycobacterium tuberculosis* infection is a major comorbidity associated with HIV-1 infection, and HIV-1-infected subjects on ART remain at higher risk of developing active Mtb infection (29, 30). We report here a concomitant decline in MAIT and iNKT cells in a cohort of mostly ART-treated HIV-1-infected individuals. Residual MAIT and iNKT cells had elevated levels of PD-1, a marker associated with exhaustion. Furthermore, residual MAIT cells had lower levels of CCR6, suggesting an impaired capacity to migrate to mucosal tissue. This dysregulation of MAIT and iNKT cells was observed in some HIV-1-infected subjects who were on ART for over 15 years, suggesting that these cells do not recover even with viral suppression. Combination of ART with immunotherapies such as IL-7 and IL-2 has shown some capacity to increase MAIT and iNKT cell frequency (42, 43). Altogether, our results suggest that HIV-1-associated defects in MAIT and iNKT cells could be in part responsible for the increased susceptibility of HIV-1-infected individuals to Mtb infection and more severe disease progression.

One limitation of our study is that we only explored cell population in the peripheral blood. Studying MAIT and iNKT cells in the lung would be highly relevant in the context of Mtb infection. Studies using animal models may be better positioned to address this question.

Overall, our results suggest a role for MAIT and iNKT cells in the successful immune control of Mtb infection. Further research is needed to develop strategies to restore these cells in ART-treated HIV-1-infected individuals.

# ETHICS STATEMENT

The study was approved by the University of São Paulo institutional review board (CAPPesq), and written informed consent was provided by all participants according to the Declaration of Helsinki. All samples were anonymized.

# AUTHOR CONTRIBUTIONS

Performed experiments: DP-P, PC, CS, MM, NC, and MS. Analyzed data: DP-P. Design study: DP-P, SO, KC, DN, and EK. Wrote the manuscript: DP-P, DN, and EK. All the authors reviewed the manuscript.

# ACKNOWLEDGMENTS

The authors would like to thank all patients and healthy controls for their time and efforts toward this study.

# FUNDING

This work was supported in part by NIAID (R01 AI52731) to DN and in part by R21 AI127127 to SO.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01394/ full#supplementary-material.

# REFERENCES


CD161++ Valpha7.2+ T cell subsets. *Front Immunol* (2017) 8:1031. doi:10.3389/fimmu.2017.01031


43. Moll M, Snyder-Cappione J, Spotts G, Hecht FM, Sandberg JK, Nixon DF. Expansion of CD1d-restricted NKT cells in patients with primary HIV-1 infection treated with interleukin-2. *Blood* (2006) 107(8):3081–3. doi:10.1182/ blood-2005-09-3636

**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 SJ and handling Editor declared their shared affiliation.

*Copyright © 2018 Paquin-Proulx, Costa, Terrassani Silveira, Marmorato, Cerqueira, Sutton, O'Connor, Carvalho, Nixon and Kallas. 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.*

# Activation and Regulation of B Cell Responses by invariant Natural Killer T Cells

#### *Derek G. Doherty\*, Ashanty M. Melo, Ana Moreno-Olivera and Andreas C. Solomos*

*Discipline of Immunology, Trinity Translational Medicine Institute, Trinity College Dublin, Dublin, Ireland*

CD1d-restricted invariant natural killer T (iNKT) cells play central roles in the activation and regulation of innate and adaptive immunity. Cytokine-mediated and CD1d-dependent interactions between iNKT cells and myeloid and lymphoid cells enable iNKT cells to contribute to the activation of multiple cell types, with important impacts on host immunity to infection and tumors and on the prevention of autoimmunity. Here, we review the mechanisms by which iNKT cells contribute to B cell maturation, antibody and cytokine production, and antigen presentation. Cognate interactions with B cells contribute to the rapid production of antibodies directed against conserved non-protein antigens resulting in rapid but short-lived innate humoral immunity. iNKT cells can also provide non-cognate help for the generation of antibodies directed against protein antigens, by promoting the activation of follicular helper T cells, resulting in long-lasting adaptive humoral immunity and B cell memory. iNKT cells can also regulate humoral immunity by promoting the development of autoreactive B cells into regulatory B cells. Depletions and functional impairments of iNKT cells are found in patients with infectious, autoimmune and malignant diseases associated with altered B cell function and in murine models of these conditions. The adjuvant and regulatory activities that iNKT cells have for B cells makes them attractive therapeutic targets for these diseases.

*Patricia Barral, King's College London,* 

*\*Correspondence:*

*United Kingdom*

*Edited by: Luc Van Kaer, Vanderbilt University, United States Reviewed by: Mark L. Lang, University of Oklahoma Health Sciences Center, United States* 

*Derek G. Doherty derek.doherty@tcd.ie*

#### *Specialty section:*

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

*Received: 30 April 2018 Accepted: 31 May 2018 Published: 18 June 2018*

#### *Citation:*

*Doherty DG, Melo AM, Moreno-Olivera A and Solomos AC (2018) Activation and Regulation of B Cell Responses by Invariant Natural Killer T Cells. Front. Immunol. 9:1360. doi: 10.3389/fimmu.2018.01360*

Keywords: invariant natural killer T cells, B cells, antibodies, disease, CD1d, glycolipids

#### INVARIANT NATURAL KILLER T (iNKT) CELLS CONTROL INNATE AND ADAPTIVE IMMUNE RESPONSES

Invariant natural killer T cells are frequently considered a "bridge" between the innate and adaptive immune systems. They are classed as innate T cells because their T cell receptors (TCRs) are semiconserved and display specificity for conserved non-peptide antigens. They display effector-memory phenotypes and can respond immediately to infection or inflammation without the need for prior antigen priming. iNKT cells possess multiple effector functions, similar to those of conventional

**Abbreviations:** APC, antigen-presenting cell; APRIL, a proliferation-inducing ligand; BAFF, B cell activation factor; Bcl-6, B cell lymphoma-6; BCR, B cell receptor; CAR, chimeric antigen receptor; CLL, chronic lymphocytic leukemia; CVID, common variable immunodeficiency; Breg cell, regulatory B cell; DC, dendritic cell; α-GalCer, α-galactosylceramide; Ig, immunoglobulin; IL, interleukin; iNKT cell, invariant natural killer T cell; iNKTFH cell, follicular helper invariant natural killer T cell; MDSC, myeloid-derived suppressor cell; MHC, major histocompatibility complex; NK cell, natural killer cell; PD-1, programmed death-1; SAP, signaling lymphocytic activation molecule-associated protein; SLE, systemic lupus erythematosus; TCR, T cell receptor; TFH cell, follicular helper T cell; Th cell, helper T cell; Treg cell, regulatory T cell.

T cells of the adaptive immune system, such as targeted granular release of cytolytic mediators and the release of T helper type 1 (Th1), Th2, Th17, and regulatory (Treg) cytokines, allowing them to activate, polarize, and regulate adaptive immune responses. Ultimately, iNKT cell responses can dictate the outcomes of microbial infections, autoimmune diseases, and cancer, and for this reason, they are attractive potential targets for therapeutic intervention for multiple types of disease. However, iNKT cells are more than simply the conjoining cell type linking innate and adaptive immunity. They can stimulate and regulate multiple cell types at many levels and thereby are central controllers of innate and adaptive immune responses.

Invariant natural killer T cells, also known as type 1 NKT cells, are clonally expanded T cells expressing a TCR composed of an invariant α-chain (Vα24-Jα18 in human and Vα14-Jα18 in mice) paired with a restricted set of β-chains, which displays specificity for glycolipid antigens presented by CD1d (1, 2). This T cell population is the best characterized member of a wider repertoire of CD1d-restricted T cells, mostly with undefined TCR specificities. CD1d-restricted T cells other than iNKT cells are collectively termed type 2 NKT cells (3, 4). The present review will focus mainly on type 1 NKT cells. Type 1 or iNKT cells express a number of stimulatory receptors that are frequently found on natural killer (NK) cells, such as NK1.1 in mice and NKG2C and NKG2D in humans. Their TCRs can recognize a number of self (5, 6) and microbial (7, 8) glycosphingolipids; however, most research on murine and human iNKT cells has utilized the prototypic glycolipid, α-galactosylceramide (α-GalCer), which binds to CD1d and activates murine and human iNKT cells (9). Activation of iNKT cells with α-GalCer *in vitro* results in target cell killing and the rapid release of multiple growth factors and cytokines (1, 2). iNKT cells are of particular interest because of their ability to produce cytokines associated with all of the CD4<sup>+</sup> helper T (Th) cell lineages, including the Th1 cytokines interferon-γ (IFN-γ) and tumor necrosis factor-α (TNF-α), the Th2 cytokines interleukin-4 (IL-4), IL-5, and IL-13, the Th9 cytokine IL-9, the Th17 cytokines IL-17A and IL-22, and the Treg cytokine IL-10 (10, 11). These cytokines contribute to the activation and polarization of CD4<sup>+</sup> and CD8<sup>+</sup> T cells (12) and natural killer (NK) cells (12, 13). Cytokines and CD1d-dependent interactions between iNKT cells and dendritic cells (DCs) (14, 15), macrophages (16), neutrophils (17, 18), and myeloid-derived suppressor cells (MDSC) (19, 20) lead to the activation and regulation of the effector activities of these cells (**Figure 1**).

Invariant natural killer T cells make essential contributions to adaptive immune responses by promoting the maturation of dendritic cells into antigen-presenting cells (APCs). Physical interactions between activated iNKT cells and DC result in the expression of major histocompatibility complex (MHC) class II molecules, co-stimulatory molecules such as CD80 and CD86, and the release of IL-12 by the DC. This adjuvant effect involves CD1d/TCR, CD40/CD40L, and CD80/86/CD28 interactions between the two cells (21–24). iNKT cells also promote DC maturation *in vivo*: a single intravenous dose of α-GalCer can stimulate the maturation of DCs into APCs, capable of activating CD4<sup>+</sup> and CD8<sup>+</sup> T cells specific for a co-administered protein (25). iNKT cells prime DC to cross-prime CD8<sup>+</sup> T cells (15, 26, 27) and to promote CD4<sup>+</sup> Th1 and Th2 cell differentiation and activation (14, 28–30). A subset of iNKT cells can also kill DC (31).

Invariant natural killer T cells can also provide help for B cells. They can boost antibody responses by directly interacting with B cells presenting glycolipid antigens on CD1d (cognate B cell help) and indirectly by activating follicular helper T (TFH) cells specific for protein antigens presented by B cells on MHC class II (non-cognate B cell help). This adjuvant activity makes iNKT cells and glycolipids attractive targets for boosting vaccine responses and preventing antibody-mediated diseases. iNKT cells also can regulate pathogenic B cell responses. Here, we review recent findings on the roles and mechanisms by which iNKT cells influence B cell function and how they may contribute to the pathogenesis of and protection against diseases that involve aberrant B cell proliferation, maturation, or regulation.

#### ACTIVATION OF B CELLS

B cells are uniquely able to recognize antigens that bind their surface immunoglobulin (Ig) receptors, resulting in the release of soluble antibody, which mediates the humoral immune response through pathogen neutralization, opsonization, and complement fixation. In the adaptive immune system, naïve antigen-specific lymphocytes are rarely activated by antigen alone. Naïve T cell activation requires an antigen-specific signal through the TCR and a co-stimulatory signal from a professional APC. Naïve B cell activation requires antigen recognition by the Ig receptor and additional signals that can come either from a CD4<sup>+</sup> T cell (thymus-dependent) or, in some cases, directly from microbial components (thymus-independent). B cells and T cells sample antigens in secondary lymphoid tissues, the lymph nodes, and the spleen, which provide a microenvironment that is ideal for enabling physical interactions between T cells and B cells and APCs, such as macrophages and DC. Lymph nodes receive antigens from the tissues *via* the lymphatics. Lymph nodes are divided into lobules, each consisting of an outer B cell-rich cortical region, a T cell-rich paracortical region, and an inner medulla (32). The B cells cluster together in lymphoid follicles. Upon antigenic stimulation, B cells proliferate and form germinal centers, where their Ig genes undergo somatic hypermutation and class switch recombination (33–37). The spleen receives antigens from the blood and consists of the white pulp embedded in red pulp. T cells and B cells accumulate in the white pulp, whereas erythrocytes dominate the red pulp. Murine spleen has an additional B cell-rich area, the marginal zone, between the white and red pulp, a region that is absent in human spleen (38).

Thymus-dependent B cell responses require the dual recognition of antigen by B cells and T cells. DC internalize protein antigens in the tissues and migrate *via* the lymphatics to the T cell-rich zones of lymph nodes. Here, they present antigenic peptides bound to MHC class II molecules to naïve T cells. T cell activation is associated with their differentiation into TFH cells, characterized by the expression of the B cell lymphoma-6 (Bcl-6) transcription factor, CD40 ligand, inducible T cell costimulator, the chemokine receptors CXCR4 and CXCR5, IL-21, programmed death-1 (PD-1), and signaling lymphocytic activation

these cells.

and dendritic cells (DCs), macrophages, neutrophils, and myeloid-derived suppressor cells (MDSC) lead to the activation and regulation of the effector activities of

molecule-associated protein (SAP) (33, 39). The TFH cells then relocate to the borders between the T and B cell areas, where they interact with antigen-specific B cells. In the lymph node follicles, the same antigen binds to a B cell receptor (BCR), resulting in internalization, processing, and cell-surface presentation on MHC class II molecules. The B cell migrates to the T–B borders, where it presents antigen to a TFH cells (34–36). Upon recognition of the peptide–MHC class II complex, the T cell expresses CD40 ligand, which ligates CD40 on the B cell, leading to B cell proliferation and differentiation. The T cell also secretes the cytokines IFN-γ, IL-4, IL-10, and IL-21, which are required for Ig isotype switching (40, 41). Cognate T cell–B cell interactions result in the proliferation of B cells in germinal centers, and somatic hypermutation, and affinity maturation of their Igs, resulting in the generation of long-lived antibody-secreting plasma cells and memory B cells (36). Cognate T–B cell interactions can also stimulate extrafollicular proliferation of B cells and their maturation into plasmablasts, which do not undergo affinity maturation

and are short-lived. These B cells mediate transient innate-like responses (35).

Thymus-independent B cell responses are elicited by nonprotein antigens that do not stimulate T cells, such as bacterial polysaccharides and microbial toll-like-receptor ligands. These antigens are recognized by subsets of innate B cells, such as marginal zone B cells and B-1 cells, which do not reside in the follicles. Thymus-independent responses generally lead to rapid antibody responses to pathogens mediated by short-lived, lowaffinity extrafollicular plasma cells (42, 43).

## iNKT CELLS CAN PROVIDE NON-COGNATE B CELL HELP

Invariant natural killer T cells can activate, regulate, enhance, and sustain humoral immune responses. In the steady state, iNKT cells are distributed throughout the spleen and lymph nodes. Upon activation, they consolidate in the marginal zones of the spleen and the interfollicular regions and medulla of the lymph nodes, where they can interact with APCs and T cells. Later, they are found in the germinal centers (44–48). Coadministration of α-GalCer with immunizing antigen to mice results in enhanced production of antibodies specific for the antigen (49–51). This help provided by iNKT cells is non-cognate and does not require the expression of CD1d by B cells, but requires the co-expression of CD1d and MHC class II by DC and CD40 ligand expression by iNKT cells (52). Upon administration of α-GalCer, iNKT cells residing in the marginal zones of the spleen are activated by CD8α+ DC, resulting in reciprocal activation of the DC and their relocation to the borders between the T and B zones of the white pulp where they activate helper T cells specific for the co-administered antigen (**Figure 2A**). These T cells acquire TFH functions and provide help to cognate B cells (45, 46). The result is a typical thymus-dependent B cell response, with formation of germinal centers, antibody class switching and affinity maturation, and the induction of long-lived antibody-secreting plasma cells and memory B cells (**Figure 2A**) (46, 51, 53, 54). iNKT cellassociated B cell activation factor (BAFF) and a proliferationinducing ligand (APRIL) are required for long-term maintenance of the B cell responses (55). Non-cognate help from iNKT cells for the generation of alloreactive antibodies following hepatocyte transplantation in the absence of exogenous glycolipid administration has been demonstrated (56), suggesting that endogenous iNKT cell-activating glycolipids are present. In summary, noncognate B cell help by iNKT cells boosts adaptive immunity by promoting the generation of long-lived antibody responses and B cell memory.

#### iNKT CELLS CAN PROVIDE COGNATE B CELL HELP

Invariant natural killer T cells can also provide direct cognate help for B cells reactive against lipid-containing antigens internalized through the BCR (**Figure 2B**) (57, 58). This has been demonstrated using protein or hapten antigens that are physically linked to α-GalCer, which were shown to be internalized by B cells, leading to presentation of α-GalCer on CD1d molecules and resulting in the acquisition of TFH phenotypes by iNKT cells. This results in reciprocal activation of the B cells, leading to the formation of extrafollicular plasmablasts and germinal centers, affinity maturation, and the generation of robust protein- or hapten-specific immunoglobulin M (IgM) and IgG responses, but not long-lived memory cells (59, 60). Cognate B cell help provided by iNKT cells requires CD1d expression by B cells, CD40-CD40 ligand signaling, CD80-CD86 costimulation, and IFN-γ but not IL-4. The iNKT cells acquire TFH phenotypes, including the expression of Bcl-6 and their ability to provide cognate B cell help requires IL-21 (**Figure 2B**).

Cognate B cell help by iNKTFH cells that recognize environmental antigens occurs in nature. Mattner and co-workers (61) demonstrated that NKT cells can provide direct cognate help to B cells during infection of mice with *Sphingomonas*, a bacterium that carries iNKT cell agonist lipid antigens in its cell wall. Cognate recognition by iNKT cells of the tumor antigen, *N*-glycolyl-GM3, is also thought to contribute to the generation of specific antibodies, which are present in some people. This antigen binds to CD1d on human B cells and is presented to iNKT cells leading to iNKT cell activation, which may reciprocally drive antibody production by the B cells (62). Sphingolipids that accumulate in Gaucher's disease are recognized by murine and human type 2 NKT cells that express TFH phenotypes. Injection of mice with these lipids results in NKTFH cell expansion, induction of germinal center B cells, and the production of anti-sphingolipid antibodies. These lipids also activate human NKT cells which provide cognate help to B cells for antibody production *in vitro* (63). Stimulation of iNKT cells isolated from the pleural fluid of humans with tuberculosis with *Mycobacterium tuberculosis* antigens results in their expression of CXCR5, release of IL-21 and the provision of cognate B cell help for the production of IgG and IgA (64). Both TFH and iNKTFH cells contribute to B cell help and the production of antibodies to *Clostridium difficile* toxin B in mice (65). Furthermore, CD1d expression by B cells was required for iNKT cell-mediated B cell help to a protein antigen in mice coimmunized with a mixture of the protein antigen and α-GalCer, indicating that cognate iNKT cell–B cell interactions can play a role in the development of antibody responses to protein antigens (53). With some notable exceptions (66, 67), iNKTFH cells appear to promote weaker antibody responses than conventional TFH cells, with smaller germinal centers and negligible differentiation of long-lived plasma cells and memory B cells (59, 60, 68). Thus, cognate B cell help by iNKT cells boosts innate immunity by promoting the generation of robust, but short-lived, antibody responses to non-protein antigens.

Cognate iNKT cell help for B cell production of antibodies has also been demonstrated unequivocally *in vitro* using co-cultures of primary human B cells and fresh or expanded autologous iNKT cells. Human iNKT cells promote proliferation of autologous naïve and memory B cells and the subsequent production of IgM, IgG, and IgA *in vitro* by a mechanism that requires B cell-iNKT cell contact and CD1d, but not α-GalCer, suggesting that the iNKT cells recognize an autologous ligand (69–71). Surprisingly, antibody blocking experiments suggested that CD40–CD40 ligand interactions were not required for the provision of cognate B cell help by human iNKT cells. Humans have distinct subsets of iNKT cells, including CD4<sup>+</sup>, CD4<sup>−</sup>CD8α−β− (double-negative), and CD8α+ iNKT cells with distinct but overlapping functional activities (10–12). Separate analysis of the B cell helper activities of CD4<sup>+</sup>, CD8α+, and double-negative iNKT cells revealed that all subsets similarly induced B cell proliferation, but CD4<sup>+</sup> iNKT cells induced higher levels of antibody release (69, 71).

## iNKT CELLS INFLUENCE B CELL FUNCTIONS OTHER THAN ANTIBODY PRODUCTION

In addition to their roles in antibody production, B cells influence T cell responses. B cells are potent APCs for T cells. They can prime CD4<sup>+</sup> T cells without the requirement for DC or macrophages (72). Similar to T cells, B cells can produce Th1 and

Figure 2 | Invariant natural killer T (iNKT) cells provide non-cognate and cognate help for antigen-specific B cells. (A) Coadministration of protein antigen and an iNKT cell ligand, such as α-galactosylceramide (lipid antigen) results in internalization by a dendritic cell (DC) and simultaneous presentation of peptide fragments of the protein antigen on major histocompatibility complex (MHC) class II to a naïve CD4+ T cell and of the lipid antigen on CD1d to iNKT cells. IFN-γ production by the iNKT cell reciprocally promotes MHC class II antigen presentation and expression of CD40 by the DC, whereas IL-2, IL-4, B cell activation factor (BAFF), and a proliferation-inducing ligand (APRIL) production promotes maturation of the peptide-specific CD4+ T cell into a follicular helper T (TFH) cell. The TFH cell then provides antigen-specific help for the proliferation of B cells in germinal centers, affinity maturation, antibody class switching, and the generation of long-lived antibodysecreting plasma cells and memory B cells. (B) iNKT cells recognizing lipid antigens presented by DC differentiate into follicular helper invariant natural killer T (iNKTFH) cells capable of activating B cells specific for lipid antigens or proteins or haptens conjugated to the lipid antigens. B cell activation is mediated by CD40 and CD80/86 ligation by the iNKTFH cells and the production of IL-21 and IFN-γ. Cognate B cell help from iNKT cells results in plasmablast expansion, germinal center formation, modest affinity maturation, and primary class switched antibody production.

Th2 cytokines (73). Since iNKT cells are unique in that they can selectively secrete Th1, Th2, Th17, or Treg cell cytokines (10, 11, 16, 74), it is possible that these cells may promote and/or regulate cytokine production, antigen presentation, and conventional T cell activation by B cells.

Our group was the first to show that cognate interactions between CD4+ iNKT cells and B cells results in the differentiation of B cells into cells with phenotypes of regulatory B (Breg) cells *in vitro* (71). Breg cells are immunosuppressive cells that help maintain immunological tolerance *via* the production of IL-10, IL-35, and transforming growth factor-β (75). Murine B cells expressing CD5 and high levels of CD1d (CD1dhiCD5<sup>+</sup> B cells), mainly marginal zone B cells, secrete IL-10 resulting in immunosuppression and protection against autoimmune disease (76, 77). In humans, a B cell population that produces IL-10 and inhibits Th1 cell responses is present in the CD24hiCD38hi B cell compartment, and this subset is functionally impaired in patients with systemic lupus erythematosus (SLE) (78, 79). We found that CD4<sup>+</sup> human iNKT cells, but not CD8α+ or DN iNKT cells, induced the expansion of both CD1dhiCD5+ and CD24hiCD38hi B cells *in vitro* by a mechanism that required cell–cell contact but not activation of the iNKT cells with α-GalCer. Co-culturing CD4<sup>+</sup> iNKT cells with B cells also induced IL-4 and IL-10 production by the B cells and inhibited their ability to stimulate proliferation of alloreactive and antigen-specific conventional T cells (71). Cognate iNKT cell help by murine iNKT cells has also been shown to induce the expansion of IL-10-producing Breg cells *in vivo*, which was associated with a decrease in germinal center B cell and TFH cell expansion (80). These findings suggest that, as well as promoting antibody production by B cells, CD4<sup>+</sup> iNKT cells can induce the differentiation of B cells into immunosuppressive cells with impaired ability to present antigen to conventional T cells.

The roles of iNKT cells as promoters and regulators of B cell maturation and antibody responses have important implications for infectious and autoimmune diseases and cancers. In the following sections, we review recent research on the roles of iNKT cells in these classes of diseases, focusing particularly on infectious diseases where antibodies are required for protection, B cell lymphomas and leukemias, and B cell-mediated autoimmune diseases. Finally, we discuss the possible role and treatment potential of iNKT cells in immunodeficiencies that result in impaired or absent antibody responses.

## iNKT CELLS PROTECT AGAINST INFECTIOUS DISEASE

Invariant natural killer T cells play a central role in the protection against infection. A number of bacterial glycolipids have been shown to bind to CD1d and stimulate iNKT cells (7, 8). Administration of α-GalCer prior to pathogen challenge improved disease outcomes in experimental models of infection, including *Plasmodium falciparum*, *Cryptococcus neoformans*, *Pseudomonas aeruginosa*, and *M. tuberculosis* (81, 82). Mice lacking CD1d or iNKT cells suffered increase burdens of *Borrelia burgdorferi*, *Streptococcus pneumonia*, *P. aeruginosa*, *M. tuberculosis*, and *Chlamydia pneumonia.*

In murine models, CD1d and iNKT cells are required to generate protective antibody responses against several pathogens, including *P. falciparum* (83), *S. pneumoniae* (84), influenza virus (85, 86), herpes simplex virus (87), *Bacillus anthracis* (88), and *Borrelia* species (89). CD1d and iNKT cells are also needed for the production of IgE antibodies specific for allergens in experimental models of airway inflammation (90, 91) and for the generation of antibodies specific for allo- and xenoantigens in grafted mice (92). Thus, iNKT cells provide an adjuvant activity for the induction of antibody responses. These data suggest that exogenous stimulation of iNKT cells could be used to protect against infection, with glycolipid antigens serving as vaccine adjuvants.

#### iNKT CELLS PROTECT AGAINST CANCER

Invariant natural killer T cells are most notable for their roles in antitumor immunity. Mice with deletions in the CD1d or Vα14Jα18 TCR genes, which lack iNKT cells, are predisposed to developing cancer and protection against cancer can be restored by adoptive transfer of iNKT cells (93). Furthermore, glycolipid activation of iNKT cells can both prevent and reverse tumor growth in mice (9, 94). iNKT cells can kill a number of human tumor cell lines *in vitro*, while their activation *in vivo* leads to downstream activation of natural killer (NK) cells and CD8<sup>+</sup> T cells which infiltrate tumors (95, 96). iNKT cells are deficient and functionally impaired in most human cancers studied (97, 98). These observations have led to a number of clinical trials involving the adoptive transfer of α-GalCer-pulsed autologous DC, *ex vivo* expanded iNKT cells, or both, in cancer patients (99, 100). Although these therapies stimulated antitumor immune responses in the patients, clinical efficacy has to date been limited.

Invariant natural killer T cells play important roles in B cell cancers. Chronic lymphocytic leukemia (CLL), the most common leukemia in adults, is characterized by the expansion of mature monoclonal CD5+ B cells (101). These cells can accumulate in the bone marrow and interfere with hematopoiesis, resulting in deficiencies of erythrocytes, platelets, lymphocytes, and antibodies. Since B cells express CD1d, they could potentially prime iNKT cells for cytolysis, one of the cardinal functions of iNKT cells. However, iNKT cells could alternatively provide B cell help for proliferation and antibody production. Recent studies (102–104) have found that circulating iNKT cells are depleted, but functionally active in patients with CLL. Reports are conflicting regarding whether CD1d expression by B cells is higher or lower in CLL patients compared to healthy controls (102, 103, 105–107). Our group provided evidence that downregulation of CD1d expression by CLL cells underlies the functional deficiency of iNKT cells in CLL patients, since the induction of CD1d expression by B cells using retinoic acid restored cytolytic killing of CLL cells by iNKT cells *in vitro* (104). iNKT cells are also depleted from the circulation of patients with multiple myeloma, a malignancy associated with the accumulation of transformed plasma cells in the bone marrow (108, 109). Myeloma cells express CD1d and are sensitive to lysis by NKT cells, but CD1d expression is downregulated during the progression of the disease and eventually lost altogether. iNKT cells are also depleted and functionally impaired in patients with non-Hodgkin's lymphoma (110) and human herpesvirus 8 multicentric Castleman disease, a virus-induced B cell lymphoproliferative disorder (111). Thus, it appears that iNKT cells play roles in B cell tumor immune surveillance but that these cells become suppressed or depleted during the course of the diseases.

### iNKT CELLS PROTECT AGAINST AUTOIMMUNE DISEASE

Invariant natural killer T cells can also protect against autoimmune and metabolic diseases. Numerical and functional deficiencies of iNKT cells are found in patients with type 1 diabetes (112, 113) and in non-obese diabetic mice, a model of type 1 diabetes (114, 115). Adoptive transfer of iNKT cells from healthy mice protected non-obese diabetic mice from developing diabetes (116). iNKT cells are also depleted from the circulation of patients with multiple sclerosis and they expand in patients during remission (117, 118). They are deficient in mice that are predisposed to developing experimental autoimmune encephalomyelitis (119, 120), and these mice can be protected from developing encephalomyelitis by injection of α-GalCer (121) or by overexpressing the Vα14Jα18 TCR. Patients with SLE (122) and lupus-prone mouse strains (119) have reduced numbers of iNKT cells, suggesting that these cells may protect against lupus. B cell-mediated stimulation of iNKT cells is deficient in patients with SLE, suggesting that B cell defects underlie these iNKT cell deficiencies (123). However, iNKT cells may promote the development of lupus in some animal models. Recent studies have also implicated iNKT cells in obesity and metabolic diseases, including type 2 diabetes (124, 125).

Several lines of evidence suggest that iNKT cells and Breg cells can prevent the production of pathogenic autoantibodies, such as anti-double-stranded DNA IgG antibodies and rheumatoid factor, which are pathogenic in patients with SLE and rheumatoid arthritis, respectively, and in murine models of these diseases (126, 127). Inhibition of autoantibody production is mainly mediated by CD5<sup>+</sup> marginal zone B cells, which express high levels of CD1d and produce IL-10 in a contact- and CD1d-dependent manner (128, 129). In contrast, iNKT cells can promote nonautoreactive antibody production by follicular B cells *via* the production of IL-17 and IL-21 (128–130). iNKT cells and CD1d also limit autoreactive B cell activation and symptoms of disease in a model where circulating apoptotic cells trigger autoantibody production, resembling the situation in SLE patients (131). Thus, it appears that iNKT cells prevent autoimmunity in lupus-prone mice by inducing autoreactive B cells to differentiate into Breg cells. However, Shen and co-workers (132) showed that iNKT cells promote the production of anti-double-stranded DNA IgG in SLE patients and that expanded iNKT cells from SLE patients, but not healthy donors, induced the production of these autoantibodies by autologous B cells. Therefore, it appears that iNKT cells can either promote or prevent the production of antibodies by B cells and that this balance of activities may determine whether or not an individual develops SLE.

# iNKT CELLS AND ANTIBODY DEFICIENCIES

Invariant natural killer T cells are also depleted and functionally altered in the circulation of patients with common variable immunodeficiency (CVID), a group of primary antibody deficiencies characterized by recurrent infections and susceptibility to autoimmunity, enteropathy, and lymphoid malignancy (133, 134). Patients with CVID display defects in B cell differentiation, resulting in accumulations of naïve and lessdifferentiated B cell populations and depletions of class switched memory B cells and plasmablasts (135, 136). This led Erazo-Borrás and co-workers (137) to investigate if the iNKT deficiency underlies the defect in B cell differentiation. They found that, although total iNKT cells were depleted, iNKTFH cells were expanded in the patients. However, α-GalCer-pulsed iNKT cells were unable to induce autologous B cell proliferation although they induced proliferation of healthy donor B cells. These findings suggest that iNKT cells are not impaired, but B responsiveness to iNKT cells is impaired in patients with CVID. Indeed, lipid presentation by B cells is required for the maintenance of iNKT cells (123), suggesting that the defects in B cell differentiation in patients with CVID may lead to the depletions in iNKT cells observed.

# CONCLUDING REMARKS

Innate recognition of self and microbial glycolipids enables iNKT cells to drive both cellular and humoral immune responses. Cellular immune responses, as exemplified in antitumor immunity, are mediated in part by the ability of iNKT cells and DC to reciprocally activate each other, resulting in the licensing of DC to activate conventional T cells and NK cells. Similar cognate interactions with B cells contribute to the rapid production of antibodies directed against conserved non-protein antigens resulting in robust, but short-lived, innate humoral immunity. Later, iNKT cells can provide non-cognate help for the generation of antibodies directed against protein antigens, by promoting the maturation and activation of TFH cells, resulting in long-lasting adaptive humoral immunity and the generation of memory B cells. The ability of iNKT cells to promote innate and adaptive humoral immune responses is balanced by their ability to induce maturation of B cells into Breg cells capable of inhibiting the generation of autoantibodies and preventing autoimmune disease. Based on studies in mice, iNKT cells hold great promise as immunomodulators for the treatment of disease. A key challenge with these multifunctional cells will be to harness the protective immunity that they offer for various types of disease while suppressing pathogenic immune stimulation. Several steps have been made to fine-tune iNKT cell stimulation for the treatment of disease. Synthetic glycolipid ligands, including structural analogs of α-GalCer, that bind to CD1d and exhibit partial agonist activity for iNKT cells can increase their therapeutic value while eliminating their pathogenic potential (138–140). Nanoparticle formulations which deliver glycolipid adjuvants to relevant APCs may further enhance their efficacy (141, 142). Tissue-specific modulation of CD1d expression, using epigenetic modifying drugs or retinoic acid, can render cells more susceptible to killing by iNKT cells (104, 143). The ability of iNKT cells to mediate immune protection can also be enhanced using cytokines or by antibody-mediated blocking of immune checkpoint inhibitors, such as PD-1 or cytotoxic T lymphocyte antigen-4 (144). Modification of iNKT cells to express chimeric antigen receptors (CAR-iNKT cells) or homing chemokine receptors may also prove effective for delivering the effector functions of these cells to target organs in the body (145).

#### REFERENCES


# AUTHOR CONTRIBUTIONS

All authors contributed to the conception, writing, and critical revising of this review.

# FUNDING

The authors acknowledge support from Consejo Nacional de Ciencia y Tecnología (CONACyT, Mexico; to AM), the Department of Immunology, Trinity College Dublin (to AM-O) and Imcyse (Sart Tilman, Belgium; to AS).


against blood group A carbohydrates. *Blood* (2013) 122:2582–90. doi:10.1182/ blood-2012-02-407452


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

# The Conventional nature of non-MHC-Restricted T Cells

#### *Marco Lepore\*† , Lucia Mori and Gennaro De Libero\**

*Experimental Immunology, Department of Biomedicine, University of Basel and University Hospital of Basel, Basel, Switzerland*

The definition "unconventional T cells" identifies T lymphocytes that recognize nonpeptide antigens presented by monomorphic antigen-presenting molecules. Two cell populations recognize lipid antigens and small metabolites presented by CD1 and MR1 molecules, respectively. A third cell population expressing the TCR Vγ9Vδ2 is stimulated by small phosphorylated metabolites. In the recent past, we have learnt a lot about the selection, tissue distribution, gene transcription programs, mode of expansion after antigen recognition, and persistence of these cells. These studies depict their functions in immune homeostasis and diseases. Current investigations are revealing that unconventional T cells include distinct sub-populations, which display unexpected similarities to classical MHC-restricted T cells in terms of TCR repertoire diversity, antigen specificity variety, functional heterogeneity, and naïve-to-memory differentiation dynamic. This review discusses the latest findings with a particular emphasis on these T cells, which appear to be more conventional than previously appreciated, and with the perspective of using CD1 and MR1-restricted T cells in vaccination and immunotherapy.

Keywords: CD1, MR1, lipid antigens, immunotherapy, vaccines

#### INTRODUCTION

T lymphocytes recognize complexes made by antigen-presenting molecules and small antigenic molecules. Small peptides generated after digestion of large proteins can form stable complexes with MHC molecules. Their generation may occur in different intracellular compartments, which are the stations where cellular protein degradation occurs. Short peptides usually associate with the MHC molecules that co-localize in the same cellular organelles.

T cells may also recognize non-peptide antigens, which are bound and presented by diverse nonpolymorphic antigen-presenting molecules.

One group of T cells reacts to lipids, which form complexes with CD1 molecules. The four CD1 proteins with antigen presentation capability (CD1a, CD1b, CD1c, and CD1d) have different antigen-binding pockets and traffic alongside diverse intracellular routes. These features confer unique lipid-binding capacities to individual CD1 isoforms and allow them presenting a large variety of lipid molecules derived from microbes and plants or of self-origin.

Invariant natural killer T (iNKT) cells are the best-characterized CD1-restricted T cells. They are stimulated by the prototype lipid α-galactosylceramide (α-GalCer), by a variety of CD1d-presented lipid antigens made by several bacteria and also by peroxisome-derived self-lipids. Because they recognize conserved lipids shared among different microbes, they represent a large fraction of total T cells.

A second group of T cells recognizes small metabolites from the mevalonate pathway synthesized by APCs or of microbial origin. Human T cells expressing the TCR Vγ9/Vδ2 recognize APC accumulating endogenous isopentenyl-diphosphate or in the presence of microbial

*Edited by: Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Patricia Barral, King's College London, United Kingdom Qibin Leng, Institut Pasteur of Shanghai (CAS), China Mark L. Lang, University of Oklahoma Health Sciences Center, United States*

#### *\*Correspondence:*

*Marco Lepore marco.lepore@immunocore.com; Gennaro De Libero gennaro.delibero@unibas.ch*

*†Present address:* 

*Marco Lepore, Immunocore, Abingdon, United Kingdom*

#### *Specialty section:*

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

*Received: 03 May 2018 Accepted: 01 June 2018 Published: 14 June 2018*

#### *Citation:*

*Lepore M, Mori L and De Libero G (2018) The Conventional Nature of Non-MHC-Restricted T Cells. Front. Immunol. 9:1365. doi: 10.3389/fimmu.2018.01365*

**306**

(E)-4-hydroxy-3-methyl-but-2-enyl diphosphate (1) and require the expression of butyrophilin 3A1 (2).

A population named mucosal-associated invariant T (MAIT) cells recognize transient metabolites generated during the microbial synthesis of riboflavin (vitamin B2) when associated with MHC-I-related molecule 1 (MR1) (3). MAIT cells react to a wide range of riboflavin-producing microbes and are abundant in human blood, liver, intestine, skin, and other organs (4). **Table 1** summarizes some of their characteristics.

An important feature of all these T cells is their capacity to promptly mediate diverse effector functions without the need for previous expansion after antigen recognition. Because of this ability, they are also defined as being innate-like T cells, as their immediate response mimics that of cells belonging to the innate immune compartment.

Innate-like T cells recognize almost ubiquitous and evolutionary conserved antigens, which are frequently encountered in the body, such as the aforementioned lipids and small metabolites. In addition, as is the case for iNKT and MAIT cells, they express a semi-invariant and oligoclonal TCR, in which a highly conserved TCR Vα chain is associated with a small number Vβ chains. In some cases, restricted amino acid use also occurs in the CDR3 regions of the α and β or γ and δ chains and it appears to be selected by antigen stimulation (4). Therefore, these cells evolved the capacity to recognize conserved microbial products, in some instances necessary for microbial survival and thus representing permanent microbial signatures.

The discovery of T cells reactive to non-protein antigens presented by non-polymorphic molecules, has prompted their definition as "unconventional," with the aim of differentiating them from "conventional" peptide-specific T cells restricted to MHC molecules. This appellative, although helpful in distinguishing the two populations, may generate some confusion as the adjective "unconventional" is often used as synonym for innate-like T cells. Indeed, it is now clear that additional T cell populations exist in humans, which are unconventional in their capacity to recognize non-peptide antigens presented by non-polymorphic proteins, but nevertheless are conventional in their close similarity to adaptive MHC-restricted T cells. Non-peptidic-specific T cells, which are not iNKT, MAIT, or Vγ9Vδ2 cells, display a polyclonal TCR repertoire and recognize diverse antigens. In addition, they expand from a naïve pool upon antigen encounter, acquire diverse specialized functions after a few days of maturation, and give rise to memory responses, thus displaying a naïve-to-effector/ memory differentiation dynamic. All of these characteristics are common to conventional peptide-specific T cells.

Heterogeneous non-innate-like T cells restricted to CD1 and recognizing diverse microbial and self-lipid antigens have been reported in many studies. TCR γδ cells reacting to lipids presented by group 1 CD1 molecules have been also identified. Finally, a new population of MR1-restricted T cells, which do not recognize riboflavin-related metabolites, has recently been isolated. These non-peptide-specific T cells, which are adaptivelike and MHC-unrestricted, remain poorly characterized and represent large populations of T cells not previously appreciated. This review will focus on the current knowledge of their features, role in immunity and diseases, and their potential applications in immunotherapy.

#### ADAPTIVE-LIKE T CELLS RESTRICTED TO CD1

CD1 proteins display some unique structural characteristics, which make them specialized in presenting lipids rather than peptides to T cells (5, 6). The CD1 antigen-binding groove is bulky, with a volume ranging between 1,280 and 2,200 Å3 across individual CD1 isoforms (CD1b > CD1c > CD1d > CD1a) (7–10). The grove commonly contains two pockets, called A′ and F′ in analogy to the A and F pockets found in MHC-I. These pockets extend deeply within the molecules and allocate the acyl chains of lipid antigens (11). Indeed, they are almost entirely unexposed to the solvent surface and are lined with non-polar amino acids, which mediate hydrophobic interactions with the aliphatic tails of the antigens. The connection between this highly hydrophobic antigen-binding groove and the hydrophilic external environment is generally provided by a single surface portal (F′ portal) that allocates the polar head-group of the bound lipids, thereby making it available for TCR interaction, and allows lipid loading/exchange (4, 12–14).

Individual CD1 isoforms display substantial differences in the dimension and shape of their antigen-binding clefts (11). CD1a has the smallest groove, with the F′ pocket partially exposed to the external surface, probably to allow rapid lipid exchange (10, 15–17). CD1b shows an additional pocket (C′) and a tunnel (T′), which connects A′ and F′ clefts (7, 18–22). This conformation permits CD1b to bind lipids with very long acyl chains (23). In CD1c, two more portals are present (D′ and C′ portals), which allow additional access points for lipid molecules to the antigenbinding cavity (9, 24–26) and are likely to be responsible for the high flexibility and versatility of this CD1 isoform in presenting a wide range of lipid structures (27).

Table 1 | MR1-restricted human T cells.


*a Preferential usage.*

*bAgs not identified.*

*5-OP-RU, 5-(2-oxopropylideneamino)-6-D-ribitylaminouracil; 6-FP, 6-formylpterin; Ac-6-FP, acetyl 6-formylpterin.*

A large variety of structurally different self-, microbe-, and plant-derived lipids stimulate specific adaptive-like T cells when presented by CD1 proteins (4, 12, 14, 17, 24, 26, 28, 29). Generally, each of them preferentially binds one of the four CD1 isoforms equipped with antigen presentation capacity (CD1a, b, c, and d), although promiscuity is also observed (4, 12, 14, 17, 24, 26, 28–30). Several factors such as the unique variations in the architecture of individual CD1 isoforms, their diverse intracellular trafficking routes, their differential pH requirements for optimal loading and their co-localization with distinct lipid-remodeling enzymes and chaperon proteins, dictate the type and the size of lipid antigens they present to T cells (13, 31–34).

More than two decades ago, pioneer studies described human T cells reacting to CD1-expressing cell lines in the absence of foreign antigens (35, 36). At the same time, the novel T cell reactivity mediated by CD1 was attributed to the recognition of lipid rather than peptide antigens (37, 38). These early discoveries triggered incredibly intense and prolific research activity aimed at enumerating, characterizing, and classifying CD1-restricted lipid-specific T cells as well as understanding their immunological roles. The exclusive expression in mice and rats of the CD1d isoform, due to the lack of group 1 CD1 genes in these laboratory animals, focalized these studies on CD1d-restricted T cells, whose most abundant subset is innate-like iNKT cells (39). Invariant NKT cell antigen specificities, developmental programs, functional capacities, and impact in immunity have been extensively characterized in *in vitro* studies and in animal models and these findings currently feed clinical research aiming to assess their therapeutic potential [reviewed in Ref. (40–42)].

Additional T cells restricted to group 1 CD1 isoforms have been identified (28, 43–46), and they resemble conventional MHC-restricted T cells specific for peptide antigens in several aspects. For this reason, we define them here as adaptive-like.

CD1-restricted adaptive-like T cells can be divided into two groups, based on the source of their antigens. The first group includes T cells restricted to group 1 CD1 (CD1a, CD1b, and CD1c) and recognizing exogenous lipids derived from the cell wall of *M. tuberculosis* (43, 46). These T cells comprise diverse subsets that might be classified according to their TCR usage. The expression of a germline-encoded TRAV1-2/TRAJ9 TCR chain, conserved among individuals and preferentially paired with TRBV6-2, defines a population of mycolate-specific CD1b-restricted T cells called germline-encoded mycolyl-reactive (GEM), which is contained in the CD4<sup>+</sup> T cell compartment (20, 47, 48). A second subset recognize glucose-monomycolates (GMM), also presented by CD1b, and has been named LDN5-TCR like, because the TCR Vα/Vβ pair found in the prototypic cell clone LDN5 (49) is frequent in this subset (48, 50). These cells display TCRs repertoire biased toward TRAV17 and TRBV4-1 chains, and diverse expression of the CD4 and CD8 co-receptors (48, 50). Additional *Mycobacterium*-reactive T cells include other CD1b-restricted T cells specific for mycolic acid (MA) (48) glycerol monomycolates (51), diacylated sulfoglycolipids (52, 53) and lipoglycans (54–56), CD1c-restricted T cells recognizing mycoketides (57, 58), and T cells stimulated by the lipopeptide dideoxymycobactin presented by CD1a (59). These T cells preferentially express the CD4 co-receptor and display a polyclonal TCR repertoire. Interestingly, they also include a small population of TCR γδ cells displaying the Vδ1 chain (46).

CD1-mediated T cell recognition of mycobacterial antigens occurs *via* direct and specific interaction of the TCR with the polar head of CD1-bound lipids (**Figure 1A**). Importantly, small

Figure 1 | Modes of CD1-restricted TCR binding to CD1–lipid antigen complexes. (A) The TCR directly interacts with both the CD1 α1 and α2 domains and the bound lipid antigens. Key residues of the CDR3α and CDR3β loops directly contact the lipid antigens, allowing discrimination of small structural variations of their polar heads exposed to the solvent. (B) The TCR directly interacts with CD1 only and does not contact the lipid antigens. The antigens are often, but not always, headless lipids, which do not protrude out of the CD1 portals and probably induce small conformational changes favoring TCR binding. Lipid antigens that do not directly contact the TCR have been defined as "permissive." (C) TCR binding is prevented by CD1 ligands that display large polar heads or contain solvent-exposed chemical groups that mediate repulsion with key residues of the TCR CDR3α and/or CDR3β loops. Ligands in this category have been defined as "non permissive." (D) TCR binding occurs despite the presence of large and complex ligand polar heads, consisting of multiple sugar subunits. The TCR interacts with both CD1 and only a portion of the exposed lipid antigen head, which probably remains partially excluded from the binding surface area. This mode has not been supported by crystallographic studies, yet.

variations in the structure or the stereochemistry of the lipid head-groups abrogate T cell recognition, thus supporting the fine antigen specificity of these T cells. For example, structural studies have demonstrated that a GEM TCR grasps the glucose ring of the GMM, acting like molecular tweezers (20). Interestingly, this TCR did not react to the same scaffold lipids displaying a mannose or a galactose instead of the glucose, suggesting that even small variations in the orientation of hydroxyl groups on the antigen head moiety, can strongly impact T cell reactivity (20). Similarly, CD1b-restricted T cells specific for the sulfoglycolipid Ac2SGL failed to recognize a version of this molecule devoid of the sulfate-group linked to sugar head-group, indicating an important role of this small moiety in mediating a direct interaction with the TCR (52). The size of the hydrophilic head is also important. A T cell clone specific for ganglioside GM1, which is made of four linear sugars and a branched sialic acid, did not recognize GM2 or GM3, which lack the terminal galactose of GM1 and the lateral sialic acid, respectively (**Figure 1D**) (60). Diverse mycoketide-specific T cells restricted to CD1c were also able to discriminate stereochemistry and structure alterations of their cognate antigens bound to CD1c (57, 58), thus further highlighting a remarkable fine specificity of these T cells.

A second group of adaptive-like CD1-restricted T cells recognizes target cells expressing CD1 isoforms in the absence of foreign antigens (28, 61). The autoreactivity of these T cells is due to the recognition of complexes formed by CD1 proteins and lipid molecules synthesized within APCs (28, 61). Diverse endogenous lipids including phospholipids, sphyngolipids, terpenes, and oils, stimulate group 1 CD1- and CD1d-restricted cells distinct from iNKT cells (17, 24, 28, 61).

The expression of CD4 and CD8 co-receptors and TCR repertoire are heterogeneous among these T cells (28, 61), although recent studies suggested that they might preferentially use recurrent TCR β chains. An example is provided by CD1c-autoreatcive T cells, which have recently been shown to be enriched in T cells expressing the TRBV4-1 or TRBV5-1 genes (26, 62). It is noteworthy that self-reactive CD1-restricted T cells include not only TCRαβ cells but also a small population of TCR γδ cells displaying the Vδ1 or Vδ3 chains (63–65).

Two general mechanisms of antigen recognition have been described for these T cells. The first one relies on cognate interaction of TCR with the polar head of lipid antigens (**Figures 1A,D**). This has been observed for CD1a, CD1b, and CD1c-restricted T cell clones reacting to sulfatide (66) and CD1b-restricted clones responding to gangliosides (60). A second mechanism has been recently elucidated by crystallographic studies, which showed direct interaction between the TCR and the CD1 molecule but not with bound lipid antigens (**Figures 1B,C**). This mode of recognition was described for CD1a and CD1c-autoreactive T cells (16, 26), and implies that CD1 proteins bind "permissive" lipids. The presence in the CD1-binding pockets of such lipids might explain the observed frequent cross-reactivity toward diverse self-lipids, including headless molecules such as triglycerides, squalene, and cholesteryl esters. These important structural studies are uncovering key aspects of the interaction between TCRs and CD1-lipid complexes. However, a crystal structure is a snapshot of this interaction and is one of the many events required for T cell activation. Two aspects remain to be investigated: (i) whether recognition of permissive lipids requires additional signals provided by non-TCR molecules and (ii) whether different permissive lipids show a hierarchy of T cell-stimulatory potencies when APCs expressing physiological levels of CD1 molecules are tested. In addition, it will be very interesting to compare different autoreactive T cells and investigate whether unique endogenous lipids play major roles in the stimulation of these T cells in physiological and pathological settings. An example is provided by CD1c-restricted T cells recognizing methyl lysophosphatidic acids (mLPA), a newly defined lipid species accumulating in leukemia cells (67) (**Figure 2**). mLPA induced potent activation of specific CD1c-restricted T cells when exogenously added to CD1c<sup>+</sup> B cells, which are not recognized in the absence of mLPA due to the scarce endogenous amounts of this lipid (67). Importantly, the same CD1c-restricted T cells strongly recognized and killed CD1c<sup>+</sup> leukemia cells, which already have high mLPA quantities (67). These findings suggested that, at least in this case, mLPA, and not other permissive lipids, is the physiological antigen responsible for this reactivity (**Figure 2**).

A third type of CD1-restricted T cells shows characteristics of both groups described above. Indeed, they display dual reactivity toward self- and exogenous lipids derived from bacteria or plants (22, 30, 68). The basis of this cross-reactivity might rely on molecular mimicry or structural similarity between exogenous and endogenous stimulatory antigens.

The lack of group 1 CD1 genes in rodents limited the study of adaptive-like CD1-restricted T cell physiological functions and roles in diseases. *Ex vivo* data obtained with antigen-loaded CD1 tetramers and *in vivo* experiments in guinea pigs (that express several group 1 CD1 molecules) or in humanized mice, indicate that lipid-specific adaptive-like T cells participate in immunity against bacterial infections and that they might also be involved in autoimmunity and cancer (4) (**Figure 2**).

CD1-retricted T cells recognizing mycobacterial antigens were found expanded in *M. tuberculosis*-infected patients and in BCGvaccinated individuals, supporting their adaptive-like properties and their role in protection (47, 51, 52, 59, 69, 70) (**Figure 2**). Furthermore, MA-specific T cells identified in the blood and lungs of tuberculosis (TB) patients displayed markers of effector and central memory cells, and persisted several months after successful treatment, indicating generation of a persistent memory T cell compartment (70). Lipoglycan-reactive T cells obtained from bronchoalveolar lavage of TB patients showed potent cytotoxic properties, and inhibited growth of intracellular mycobacteria (71). The recent detection of CD1b<sup>+</sup> macrophages within lung granulomas of TB patients further suggests the importance of CD1-mediated immunity in this infection (72). In mice transgenic for the full human CD1 locus, both infection with mycobacteria or immunization with mycobacterial lipids elicited a slow primary CD1-restricted T cell response and very rapid secondary responses (73), similar to what was observed for peptide-specific T cells. Finally, studies in guinea pigs indicated that immunization with mycobacterial lipids or purified Ac2SGL conferred protections when challenged with *M. tuberculosis* (74–77).

The frequency of group 1 CD1-restricted T cells remains a poorly investigated issue. Two independent studies revealed that, in the blood of healthy donors, a surprisingly high frequency of

T cells reacted to CD1-overexpressing targets in the absence of foreign antigens. A major fraction of these T cells were restricted to CD1a and CD1c molecules (78, 79). In one study, it was found that CD1-autoreactive T cells at birth were mainly contained in the naïve CD45RA<sup>+</sup> compartment, while in adult blood their frequency among CD45RO<sup>+</sup> effector/memory cells increased (78). These phenotypes are consistent with a progressive transition from naïve to effector/memory, typical of adaptive peptide-specific T cells. In addition, CD1a-restricted T cells expressed the skinhoming receptors CCR4 and CCR10 and could be isolated from skin biopsies (79). Their capacity to release IL-22 further suggested an immunological role in the skin (79, 80) (**Figure 2**). In addition, these autoreactive T cells could promote monocyte-derived dendritic maturation in a CD1c- or CD1d-dependent manner (81), thus attributing them a helper-like function (**Figure 2**). Selfreactive T cells might also act as sentinels for cell stress and inflammation. Indeed, APC may accumulate antigenic endogenous lipid antigens after microbial stimulation, and thus become very efficient in stimulating self-lipid-specific T cells (82) (**Figure 2**).

In individuals affected by Grave's disease or Hashimoto thyroiditis, two autoimmune diseases of the thyroid, CD1a and CD1c self-reactive T cells infiltrated thyroid glands and were capable of lysing thyroid cells *in vitro* (83), possibly contributing to gland destruction. CD1c-autoreactive T cell clones isolated from systemic lupus erythematous (SLE) patients were able to provide pathogenic CD1c-dependent help to B cells *in vitro* (84). This study also showed that clones from healthy donors promoted IgM response in B cells, whereas cells isolated from SLE patients also elicited IgG production by the same B cells (84). These data suggested a role of CD1c self-reactive T cells in the genesis of the detrimental autoantibody responses that characterize this autoimmune disease. High frequency of circulating CD1 restricted T cells recognizing diverse self-glycosphingolipids were detected in multiple sclerosis patients (60, 66). Such clones preferentially recognized sulfatides made of long acyl chains, which are highly enriched in brain plaque lesions, thus showing a correlation between antigen specificity and lipid accumulation at sites of disease. Together these findings suggested that CD1 autoreactive T cells might participate in the pathologic process of myelin disruption.

CD1a-self-reactive T cells, which preferentially home to skin in healthy donors (79, 80), have been indicated as being capable of promoting inflammatory and autoimmune reactions of the skin (**Figure 2**). These cells accumulated in individuals with psoriasis and atopic dermatitis (85, 86). In both cases, the reactivity of these T cells depended on PLA2 secreted by mast cells in psoriatic lesions (85) or released by house dust mites in atopic dermatitis (86). PLA2 activity probably participates in generating CD1apresented neoantigens, most likely lysophospholipids and free fatty acids, from the pool of skin phospholipids (85, 86). PLA2 is also a component of bee venom, and when injected sub-cutaneously induced local activation of CD1a-restricted polyclonal T cells (87, 88). In another study, urushiol, a molecule found in poison ivy and able to bind CD1a, induced CD1a-restricted T cells, which in a mouse model and in psoriasis patients amplified the local inflammation (17). Furthermore, contact sensitizers, including common cosmetic compounds, could unleash or potentiate the capacity of APCs to induce self-lipid-specific autoreactivity of skin-associated CD1a-restricted and CD1d-restricted T cells (89). All these data indicated CD1a, expressed at high levels on skin-resident Langerhans cells, as potential therapeutic target for skin inflammatory diseases.

# T CELLS RESTRICTED TO MR1

A second population of T cells, which recognize non-peptidic antigens, is constituted by MR1-restricted T cells. The MR1 antigen-presenting molecule is also non-polymorphic and has structural similarities to MHC class I molecules as it is displayed on the cell surface as an heterodimer composed of a heavy chain non-covalently associated with β2 microglobulin (90). MR1 tissue distribution also resembles that of MHC class I molecules, as is almost ubiquitous (91), thus indicating that specific T cells might be activated by many different cell types.

MR1, like CD1 genes, is non-polymorphic. In addition and in contrast to CD1, the MR1 protein sequence is very conserved among species (91), which is not the case for CD1 genes. Both these features raise the question, what are the selection mechanisms that keep the MR1 structure conserved? Among the various possibilities, requirements for binding unique categories of antigens and/or a mandatory interaction with conserved molecules different from TCR might occur.

An intriguing MR1 feature is its antigen-binding pocket. This is formed by two interconnected cavities decorated by hydrophilic and hydrophobic aminoacids (92). The cavities are large enough to allocate molecules larger than the ones identified so far. Variability in ligand sizes might suggest that MR1 evolved the capacity of presenting antigens of different origins and chemical structure. This latter possibility is supported by the nature of the MR1-binding molecules that have been identified, so far. Among them are those formed by non-enzymatic condensation of 5-ribityl amino uracil, a precursor of riboflavin which is a typical bacterial molecule, with methylglyoxal or glyoxal carbonyls (3). Importantly, the resulting molecules activate MAIT cells in a very efficient manner. While recognition of microbial antigens by MAIT cells was anticipated by previous studies (93, 94), it was a surprise that the stimulatory antigens were neither peptides nor lipids. Structural studies also showed that formation of a covalent bond with MR1 is mandatory for stable binding and MAIT cell stimulation (3). Whether this bond has additional implications, such as a prolonged persistence of the MR1–antigen complex within APC and a reduced possibility of antigen exchange during MR1 recycling need further studies.

In another series of experiments, it was found that the MR1 antigen-binding pocket can also allocate small molecules, such as drugs, including salicylates and diclofenac, drug metabolites, and drug-like molecules (95). These compounds bound in different orientations, outlining the adaptability of the MR1-binding pocket. Intriguingly, all these molecules appeared to occupy only the A′ pocket of MR1, thus suggesting that bigger molecules might also occupy the F′ pocket.

Recently, a second type of MR1-restricted T cells was identified. These cells, defined as MR1T cells, express polyclonal TCR chains, are CD8+ or CD4/CD8 double negative and can be isolated from circulating pool of healthy donors (96) (**Table 1**). MR1T cells showed preferential recognition of tumor cells and not of normal cells, even when the levels of expressed MR1 were physiologically very low. In addition, the presence of multiple antigens was suggested by differential recognition of tumor cells and by chromatographic separation of at least two antigens. Although the stimulatory molecules remain unknown, they are shared with mouse and hamster tumors. Upon antigen recognition, MR1T cells showed a variety of functions, including release of Th1 or Th2 cytokines, of vascular endothelial growth factor or platelet-derived growth factor AA, and expressed different transcriptional programs (96). These findings indicate that MR1T cells represent a novel population of functionally different T lymphocytes recognizing non-microbial antigens accumulating in tumor cells and showing characteristic of adaptive T cells (**Figure 3**).

As MR1T cells have been identified very recently, there is no literature concerning their role in diseases. The fact that these

Frontiers in Immunology | www.frontiersin.org June 2018 | Volume 9 | Article 1365

phenotype of the clones isolated so far.

cells preferentially recognize tumor cells will promote new studies in patients harboring tumors expressing MR1 molecules.

In contrast, MAIT cells have been recognized for a decade and several studies have addressed their potential role in diseases. We briefly describe these findings below.

The identification of the bacterial antigens stimulating MAIT cells complemented a series of studies indicating a role of MAIT cells during bacterial infections. Both human and MAIT cells showed anti-mycobacterial effects (97, 98), and released a variety of cytokines upon recognition of bacteriainfected APC (97). MAIT cells accumulated in the lungs of mice infected with *Salmonella typhimurium*, upon stimulation with antigen and a toll-like receptor agonist and participated in local inflammation by secreting IL-17 and IFNγ (99). Studies in patients with severe bacterial infections showed an early decrease in MAIT cell count (100) and a significant positive correlation among non-streptococcal bacterial infections and MAIT depletion. Patients with persistent decreased MAIT cell numbers showed increased susceptibility to intensive care unit-acquired infections. A reduced number of circulating MAIT cells and their parallel increase in the lung were reported in patients with TB (93, 94). Reduced numbers of circulating MAIT cells were described in cystic fibrosis patients with *Pseudomonas aeruginosa* infection (101). In donors infected with *Plasmodium falciparum* sporozoites MAIT cells increased up to 6 months after infection clearance (102). A role of MAIT cell in protection during bacterial infection was also indicated by impaired protection during infection with *Francisella tularensis* in MAIT-depleted mice (103) and a positive effect in mice transgenic for a TRAV1-TRAJ33 TCR chain (93). However, in chronic infections MAIT cells might also participate in tissue lesions, as found in a mouse model of gastric infection with *Helicobacter pylori* (104). Their secretion of IL-17 and cytotoxic activity were considered important in establishing tissue lesions in this model.

Mucosal-associated invariant T cells might also contribute to immune response during viral infections. In patients with HIV infection the number of circulating MAIT cells varies to different extents (105–107). These effects might be ascribed to concurrent bacterial infections and migration to peripheral tissues.

A recent study also outlined a role of MAIT cells in graft versus host disease (GVHD) (108). MAIT cells accumulated in different GVHD target organs, contributed to intestinal mucosal integrity, released large amounts of IL-17, and limited the expansion of alloreactive T cells in the graft. As MAIT cells are much more abundant in humans than in rodents, it will be important to perform detailed studies in patients receiving bone marrow transplantation.

Mucosal-associated invariant T cells have also been studied in several autoimmune diseases. MR1-deficient mice develop severe experimental autoimmune encephalomyelitis (109) and also show loss of gut integrity and worsened diabetes (110). Several studies showed alterations in the number of circulating MAIT cells in patients with SLE (111), and in obese and type II diabetes patients (112). MAIT cells detected in psoriatic plaques (113) secreted large quantities of IL-17, thus implicating a possible role in local inflammation. In inflammatory bowel disease, circulating MAIT cells were diminished, while their number was increased in the inflamed versus healthy mucosal tissue (114). MAIT cells from colon released large amounts of IL-17 and IL-22. Combined, these findings indicate that MAIT cells may contribute to local inflammation in various diseases. They might become active upon TCR triggering or in response to local high levels if IL-12 or IL-18 (115, 116), thus behaving as amplifiers of inflammation also in the absence of antigen recognition.

# EXPLOITATION OF NON-PEPTIDE-SPECIFIC T CELLS IN THERAPY

Non-peptidic specific T cells are being used in different clinical settings. The recent explosion of T cell immunotherapy in cancer prompted exploitation of CD1-restricted T cells in tumor patients.

Invariant natural killer T cells were the first non-conventional T cell population utilized. Several clinical trials are being performed in cancer using α-GalCer, a strong agonist of iNKT cells [recently reviewed in Ref. (41, 42, 117)]. Other studies took advantage of the adjuvant capacity of iNKT cells and utilized soluble CD1d-α-GalCer complexes conjugated with tumor-specific antibodies (118), or α-GalCer conjugated with tumor-associated peptides (119, 120) or administration of dendritic cells pulsed with both α-GalCer and long peptides from the melanomaassociated NY-ESO-1 protein (121). These treatments induced significant amelioration both in animal models and in clinical settings.

An important therapeutic strategy is based on tumor-specific recognition by group 1 CD1-restricted T cells. As lipids and in particular glycosphingolipids, are altered in tumor cells (122), tumor-associated lipids are of immunotherapeutic interest. mLPA is an example as it is preferentially expressed in leukemic cells and is presented by CD1c (67). mLPA-specific T cells recognized and killed CD1c<sup>+</sup> leukemia targets and limited leukemia cell spread in a mouse model (67). Immunotherapeutic strategies might be transfer of mLPA-specific TCR genes and vaccination with mLPA to expand specific T cells in leukemia patients. Important bonuses of these approaches, which are not applicable in case of peptide antigens, are (i) the non-polymorphic nature of CD1c, which allows application to the entire human population, and (ii) the lipid structure preservation under selective pressure. Whether additional tumor-associated lipids stimulate other CD1-restricted T cells remains a very important area of research.

MR1-restricted T cells might also be considered in tumor immunotherapy. The adjuvant properties of MAIT cells might be exploited similarly to iNKT cells, although no studies are available at present. MR1T cells might offer an additional possibility based on specific recognition of tumor-associated antigens (96). Transfer of MR1T TCR genes and vaccination with MR1T-stimulatory antigens might parallel the approaches using mLPA-specific T cells.

A large number of studies identified microbial lipids stimulating group 1 CD1-restricted T cells [reviewed in Ref. (4)] and suggested their use in novel vaccines. Two studies showed that mouse (123) or guinea pig (77) vaccination with group 1 CD1 binding lipids confers protection in *M. tuberculosis* infection models. These promising results, together with the finding that lipid-specific T cells expand in TB patients (52, 69) will prompt investigations to formulate novel lipid-based TB vaccines.

### CONCLUDING REMARKS

Non-peptide-specific T cells are abundant in blood, have a frequency similar to MHC-restricted T cells and can be considered important protagonists of adaptive immunity. Their functions, tissue distribution and capacity to generate memory populations are revealing unforeseen immunological roles. The identification of the antigenic repertoire stimulating both CD1- and MR1 restricted T cells will be instrumental to depict their participation in immune homeostasis and in diseases.

#### REFERENCES


# AUTHOR CONTRIBUTIONS

ML, LM and GDL wrote and revised the manuscript and approved its final version for publication.

# ACKNOWLEDGMENTS

The authors are grateful to present and past collaborators of the Experimental Immunology laboratory for their contribution to the topics of this review and to Paula Cullen for reading the manuscript. The authors thank the support of Swiss National Foundation (310030-173240), the Swiss Cancer League (KFS-3730-08-2015), and the European Union's Research and Innovation Program (TBVAC2020 643381) to GDL.


**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 Lepore, Mori and De Libero. 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.*

*Yue Ren1,2, Etsuko Sekine-Kondo1 , Midori Tateyama1,3, Thitinan Kasetthat1,4, Surasakadi Wongratanacheewin4 and Hiroshi Watarai1 \**

*1Division of Stem Cell Cellomics, Center for Stem Cell Biology and Regenerative Medicine, Institute of Medical Science, University of Tokyo, Tokyo, Japan, 2Department of Neurology, The Neurological Institute of Jiangxi Province, Jiangxi Provincial People's Hospital, Nanchang, China, 3Department of Immunology, Kitasato University School of Medicine, Sagamihara, Japan, 4Department of Microbiology, Khon Kaen University, Khon Kaen, Thailand*

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Nyambayar Dashtsoodol, RIKEN Center for Integrative Medical Sciences (IMS), Japan Weiming Yuan, University of Southern California, United States*

*\*Correspondence:*

*Hiroshi Watarai hwatarai@ims.u-tokyo.ac.jp*

#### *Specialty section:*

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

*Received: 28 March 2018 Accepted: 24 May 2018 Published: 14 June 2018*

#### *Citation:*

*Ren Y, Sekine-Kondo E, Tateyama M, Kasetthat T, Wongratanacheewin S and Watarai H (2018) New Genetically Manipulated Mice Provide Insights Into the Development and Physiological Functions of Invariant Natural Killer T Cells. Front. Immunol. 9:1294. doi: 10.3389/fimmu.2018.01294*

Invariant natural killer T (iNKT) cells are a unique T cell subset that exhibits characteristics of both innate immune cells and T cells. They express Vα14-Jα18 (*Trav11*-*Traj18*) as an invariant chain of the T cell receptor (TCR) and are restricted to the MHC class I-like monomorphic antigen presenting molecule CD1d. iNKT cells are known as immune regulators that bridge the innate and acquired immune systems by rapid and massive production of a wide range of cytokines, which could enable them to participate in immune responses during various disease states. Thus, *Traj18*-deficient mice, *Cd1d*deficient mice, or iNKT cell-overexpressing mice such as iNKT TCRα transgenic mice and iNKT cell cloned mice which contain a Vα14-Jα18 rearrangement in the TCRα locus are useful experimental models for the analysis of iNKT cells *in vivo* and *in vitro*. In this review, we describe the pros and cons of the various available genetically manipulated mice and summarize the insights gained from their study, including the possible roles of iNKT cells in obesity and diabetes.

Keywords: invariant natural killer T cells, CD1d, Traj18, iPSC, obesity, adipose tissue, cloned mice, thymic differentiation

# INTRODUCTION

Invariant natural killer T (iNKT) cells (also called type I NKT cells) are characterized by the expression of an invariant T cell receptor (TCR), Vα14-Jα18 (*Trav11*-*Traj18*) paired with Vβ8.2 (*Trbv13-2*), Vβ8.1 (*Trbv13-3*), Vβ7 (*Trbv29*), or Vβ2 (*Trbv1*) in mice and the Vα24-Jα18/Vβ11- Dβ2-Jβ2.7 (*TRAV10*-*TRAJ18*/*TRBV25-1*-*TRBD2*-*TRBJ2-7*) pair in humans. The iNKT cells differ from classical αβ T cells in recognizing (glyco)lipid antigens (Ags) in conjunction with the monomorphic MHC class I-like CD1d molecule (1, 2). The prototypic Ag recognized by iNKT cells is the glycosphingolipid α-galactosylceramide (α-GalCer), originally isolated from the marine sponge *Agelas mauritianus*. It was identified from structure–activity relationship studies around the glycosphingolipid Agelasphin 9b by the pharmaceutical division of Kirin Brewery Co. Ltd. in a screen for naturally occurring molecules that prevented tumor metastases in mice *in vivo* (3). The synthetic derivative compound, also known as KRN7000 (α-GalCer C26:0), retains the activity of Agelasphin 9b while being much easier to synthesize (4). α-GalCer and its derivatives have been used in many different studies and are highly potent iNKT cell modulators both in humans and in mice (5, 6). Recent studies have demonstrated that iNKT cells, even though they all express the same invariant Ag receptor, can be classified into different functional subtypes, interferon (IFN)-γ-producing iNKT1, interleukin (IL)-13/IL-4-producing iNKT2, and IL-17A-producing iNKT17 (7). When activated by α-GalCer, iNKT cells rapidly produce these various types of cytokines, resulting in bystander immune modulating functions leading to activation and inhibition of various immune effector cells, including NK cells, macrophages, granulocytes, dendritic cells (DCs), basophils, and eosinophils in the innate system as well as CD4<sup>+</sup> T and CD8<sup>+</sup> T cells in the acquired system. Therefore, iNKT cells participate in broad spectrum regulation of immune homeostasis and in various disease states including infection, autoimmunity, allergy, antitumor responses, metabolic disorders, allograft rejection, and graft-vs-host disease (8, 9).

Numerous studies investigating the role of iNKT cells have utilized mouse models of iNKT cell deficiency. One such model directly targets Jα18 (*Traj18*−/−) (10), which is required for iNKT-TCR formation. However, the overall TCR repertoire diversity is impaired in *Traj18*<sup>−</sup>/<sup>−</sup> mice, in which *Traj18* was replaced with a PGK-Neor cassette, which had inadvertent but substantial effects on transcription and TCRα gene rearrangements (11). Another model makes use of mice deficient in CD1d (*Cd1d1*<sup>−</sup>/<sup>−</sup>) (12), which prevents the development of any CD1d-restricted T cells including iNKT cells. However, in mice, although not in humans, there has been a gene duplication event and so two homologous genes, *Cd1d1* and *Cd1d2*, encode CD1d proteins. Even though *Cd1d1*<sup>−</sup>/<sup>−</sup>*Cd1d2*<sup>−</sup>/<sup>−</sup> mice have also been developed (13, 14), the role of CD1d2 in iNKT cell development and function is still unclear. Consequently, any changes in immunological activity attributed to iNKT cells that are based on studies of *Traj18*<sup>−</sup>/<sup>−</sup> or *Cd1d*<sup>−</sup>/<sup>−</sup> mice need to be reassessed.

On the other hand, mice that have been genetically manipulated to have elevated numbers of iNKT cells are also useful tools for iNKT cell study. Therefore, rearranged Vα14-Jα18 and Vβ8 genes were introduced into recombination-activating gene-deficient mice, and there was preferential generation of iNKT cells but no mature B and T lymphocytes (15). iNKT-TCRα transgenic mice that overexpressed iNKT-TCRα (mVα14- Jα18) were firstly generated by Bendelac et al. (16) resulting in preponderance of iNKT cells, while abnormal development of other immune cells were also observed. A human iNKT-TCRα (hVα24-Jα18) transgenic mouse has also been developed by similar approach (17). Moreover, iPS cell lines obtained by reprogramming of mature iNKT cells (iNKT-iPSC) from C57BL/6 (B6) mice preferentially generate iNKT cells but no conventional αβ T or γδ T cells, NK cells, DCs or B cells *in vitro* (18). Furthermore, mice generated from the iNKT-iPSC had a much larger number of iNKT-like cells (19) compared to mice with a rearranged Vα14-Jα18 transgene (16). It is therefore important to compare the development and function of iNKT cells and their subtypes that differentiate *in vivo* in these iNKT cell overexpressed mice.

# *Traj18*-DEFICIENT AND *Cd1d*-DEFICIENT MICE

Because iNKT cells are highly conserved among species including mice and humans, mouse models of iNKT cell deficiency represent useful tools for the analysis of iNKT cell biology. However, as described above, the originally generated *Traj18*<sup>−</sup>/<sup>−</sup> mice (10) lack transcripts not only of *Traj18* but also of genes encoding Jα regions upstream of *Traj18*, resulting in an almost 60% reduction in the diversity of the TCRα repertoire (11). It is possible that the lower overall αβ TCR diversity of the original *Traj18*<sup>−</sup>/<sup>−</sup> mice contributed to the divergent results that have been reported by some of the studies that used the mice. Recently, four new *Traj18*<sup>−</sup>/<sup>−</sup> mouse lines have been established by different research groups including ours. Two lines were generated by classical P1 bacteriophage cyclization recombination/locus of crossover in P1 (Cre/ loxP) technology (20, 21), a third was generated by transcription activator-like effector nuclease (TALEN) methodology (TALEN-*Traj18*<sup>−</sup>/<sup>−</sup>) (22), and a fourth by using the clustered regularly interspaced short palindromic repeat (CRISPR)/ Cas9 technology (CRISPR-*Traj18*<sup>−</sup>/<sup>−</sup>) (23). All four groups analyzed TCRα diversity in CD4<sup>+</sup>CD8<sup>+</sup> double-positive (DP) thymocytes by next-generation sequencing and found that the usage frequency of Jα gene segments in these new *Traj18*<sup>−</sup>/<sup>−</sup> mouse lines was comparable with WT B6 mice (20–23).

In addition to canonical Vα14-Jα18 iNKT cells, another minor α-GalCer/CD1d reactive subset of T cells harboring *Trav10*-*Traj50* was recently described as type Ib NKT cells (24). However, type Ib NKT cells were discovered in mice that lack expression of *Traj* gene segments upstream of *Traj18* (10). We (23) and Chandra et al. (20) could not detect any type Ib NKT cells in the new mouse strains lacking iNKT cells. By contrast, Zhang et al. (22) did find type Ib NKT cells in their TALEN-*Traj18*<sup>−</sup>/<sup>−</sup> mice. However, these mice express *Trav11*-*Traj18* mRNA with a partial deletion, indicating that a mutant iNKT-TCRα has the unexpected ability to recognize α-GalCer/CD1d. Based on these results, we should rethink the existence of type Ib NKT cells.

It is known that iNKT cells are restricted by CD1d molecules, but that two CD1d isoforms, CD1d1 and CD1d2, are present in mice. Two gene manipulated lines has been developed, *Cd1d1*<sup>−</sup>/<sup>−</sup> (12) and *Cd1d1*<sup>−</sup>/<sup>−</sup>*Cd1d2*<sup>−</sup>/<sup>−</sup> (13, 14). iNKT cells are severely impaired in both lines, indicating that CD1d2 cannot substitute for CD1d1 to support iNKT cell development. CD1d1 and CD1d2 in 129/Sv mice share 93% amino acid identity. Although CD1d2 on thymocytes cannot substitute for the development of iNKT cells (25), we cannot eliminate the potential role of CD1d2 in the development or function of iNKT cells. Sundararaj et al. (26) recently reported that the structure of the CD1d2 A′-pocket was restricted in size compared with CD1d1 in complex with endogenous lipids or a truncated acyl-chain analog of α-GalCer. They found that the majority of iNKT cells in the *Cd1d1*<sup>−</sup>/<sup>−</sup> mice showed an increase in the iNKT2 and iNKT17 populations and a concomitant decrease in iNKT1 compared with WT mice (26). A small but consistent increase in the proportion of cells using the Vβ8 gene segment, concomitant to a reduction in Vβ7 gene usage, was also observed for CD1d2-selected iNKT cells compared with CD1d1-selected iNKT cells (26). B6 mice, but not BALB/c or 129/Sv mice, harbor a two-nucleotide insertion in exon 4 of *Cd1d2*, which encodes the α3 domain (27). This frameshift mutation introduces a stop codon, abolishing surface expression but possibly still allowing expression of a soluble CD1d2 molecule (**Figure 1A**). These results indicate that the CD1d2 molecule can present different sets of self-antigen(s) in the thymus of different mouse strains, thereby potentially impacting the development of iNKT cells. Even though *Cd1d2*<sup>−</sup>/<sup>−</sup> mice have not yet been established, they should provide an answer to this controversy.

CD1d is also an Ag presenting molecule for cells other than iNKT cells. Another type of CD1d-restricted cell is the type II or variant NKT cell, which has a more diverse TCR repertoire and appears to recognize various lipid Ags including sulfatides, also known as 3-*O*-sulfogalactosylceramide, which are a class

presenting molecule.

antigen(s). The structure of CD1d2 is different among species in mice due to the frameshift mutation. The CD1d2 molecule in B6 mice has MHC class I-like domain required for the presentation of glycolipid ligand(s) but lacks immunoglobulin C1 domain. It is still unclear whether the soluble form of CD1d2 works as an antigen

Ren et al. Gene Manipulated Mice for iNKT

of glycolipids that contain a sulfate group (28). Unfortunately, no tools are yet available that can be used to analyze the entire population of type II NKT cells. Thus, when we discuss type II NKT cells, it is important to understand the advantages and limitations of each experimental tool, as well as the precise definition of type II NKT cells being analyzed (29). There is also a report that the homeostasis of liver-resident IL-17A-producing γδT (γδT-17) cell depends on hepatocyte-expressed CD1d that presents lipid Ag (30); however, it is unclear whether CD1d is required for the development of γδT-17 cells in the thymus (**Figure 1B**).

Despite a high degree of conservation, subtle but important differences exist between the CD1d Ag presentation pathways of humans and mice. Wen et al. (31) have generated a human CD1d knock-in mouse (hCD1d-KI) which substitute mouse *Cd1d1* to human *CD1D* locus. Reduced numbers of iNKT cells were observed, but at an abundance comparable to that in most humans. They further generated human iNKT-TCRα chain knock-into the hCD1d-KI (32). Similar to humans, the mice developed a subset of CD8αβ+ iNKT cells among other human-like iNKT subsets. The results support human *CD1D* is functionally and phenotypically ortholog of mouse *Cd1d1*. These hCD1d-KI mice will allow more accurate *in vivo* modeling of human iNKT cell responses as some human pathogens specifically target human CD1D for pathogenicity and will facilitate the preclinical optimization of iNKT cell-targeted immunotherapies.

#### iNKT CELLS AND OBESITY

Obesity research is an illustrative example of how the different genetically engineered animals have been employed to study the role of iNKT cells in a complex disease. Both the original *Traj18*<sup>−</sup>/<sup>−</sup> (10) and the *Cd1d*<sup>−</sup>/<sup>−</sup> (12–14) mice have been used by many different research groups to study the role of iNKT cells and/or type II NKT cells in obesity-related pathologies in high fat diet (HFD)-induced obesity models. However, these studies have reported very conflicting results, with some groups finding no effect (33, 34), some protection (35–37) and others finding promotion (38, 39) of obesity-associated disease.

Among these studies, only one paper, that published by Lynch et al. (33), has focused on the protective role of iNKT cells in obesity by studying HFD-induced obese *Traj18*<sup>−</sup>/<sup>−</sup> mice on a B6 background. In this study, they found that when mice lacking iNKT cells were placed on an HFD they showed enhanced weight gain, larger adipocytes, fatty livers, and insulin resistance. By contrast, many other research groups suggested a pathogenic role of iNKT cells in obesity by showing an ameliorated metabolic phenotype in HFD-induced obese *Traj18*<sup>−</sup>/<sup>−</sup> or *Cd1d*<sup>−</sup>/<sup>−</sup> mice. In the studies that used B6 background *Traj18*<sup>−</sup>/<sup>−</sup> mice, Wu et al. (39) reported ameliorated hepatic steatosis, glucose tolerance, and insulin sensitivity, as well as reduced tissue inflammation in *Traj18*<sup>−</sup>/<sup>−</sup> mice on an HFD. Pathological roles of NKT cells in obesity were also reported by Satoh et al. (38) in *Cd1d1*<sup>−</sup>/<sup>−</sup> mice; however, no difference in the metabolic parameters between *Traj18*<sup>−</sup>/<sup>−</sup> and WT B6 mice on an HFD was observed in their study, arguing for a pathogenic role of type II NKT rather than iNKT cells in these pathologies. Similarly, Kotas et al. (40) and Lee et al. (41) have also reported a minor role of iNKT cells in the development of obesity, by comparing *Traj18*<sup>−</sup>/<sup>−</sup> and *Cd1d1*<sup>−</sup>/<sup>−</sup>*C d1d2*<sup>−</sup>/<sup>−</sup> mice with WT B6 mice on an HFD.

Many reasons for these divergent results have been proposed and discussed, including the age, gender, and background of the mice, HFD type and duration, and the gut flora or environmental microbial distribution among the animals employed by the different research groups. Nevertheless, if we focus only on the results obtained from B6 background *Traj18*<sup>−</sup>/<sup>−</sup> mice, it is interesting to note that most of the studies have used an HFD of 60% fat calories, and none of them have reported a decreased level of weight gain in *Traj18*<sup>−</sup>/<sup>−</sup> mice as compared with WT B6 mice (36–40). It seems that a consensus has been reached that iNKT cells do not participate in promoting the development of obesity, at least as measured by gain in bodyweight.

To exclude the possible effect of impaired TCR repertoire diversity on diet-induced obesity observed in the original *Traj18*<sup>−</sup>/<sup>−</sup> mice (10, 11), we re-investigated the contribution of iNKT cells to the development of obesity using our CRISPR-*Traj18*<sup>−</sup>/<sup>−</sup> mice (23) with an unbiased TCR repertoire. The results were clear cut, obese CRISPR-*Traj18*<sup>−</sup>/<sup>−</sup> mice gained less body weight and had smaller visceral fat-pads and adipocytes, less fat deposits in the liver, and ameliorated glucose tolerance and insulin resistance (23). The ameliorated levels were almost equivalent to those seen in obese *Cd1d1*<sup>−</sup>/<sup>−</sup> mice, indicating that iNKT cells play a pathogenic role in diet-induced obesity and that the impact of CD1d deficiency on metabolism is iNKT cell dependent.

It is notable that one of the T cell populations affected by impaired TCR repertoire diversity in the original *Traj18*<sup>−</sup>/<sup>−</sup> mice, the mucosal-associated invariant T (MAIT) cells that use the invariant Vα19-Jα33 (*Trav1-Traj33*) chain in mice (42), were recently reported to have an altered distribution and cytokine productions in obese patients, and were found to be positively associated with insulin resistance (43, 44). Considering the potential role of MAIT cells and other T cell subsets in obesity, results obtained with the original *Traj18*<sup>−</sup>/<sup>−</sup> mouse model should be interpreted with caution.

## iNKT-TCR**α** TRANSGENIC AND iNKT CELL CLONED MICE

Mice that have been genetically manipulated to have elevated numbers of iNKT cells were first attempted to generated by overexpressed Vα14-Jα18 iNKT TCRα (mVα14-Jα18) as a transgene (16). Consistent with the results from *Cd1d*-deficient mice (12–14), the mVα14-Jα18 transgenic mice exhibited increased IL-4 and immunoglobulin (Ig) E in serum, indicating that mouse iNKT cells are one of the important sources of IL-4 and IgE. Because both human and mouse iNKT cells are restricted to α-GalCer/CD1d, a human iNKT TCRα (hVα24-Jα18) transgenic mouse has also been developed (17). Interestingly, analysis of the mice and derived hVα24-Jα18<sup>+</sup> T cells revealed that type 1 diabetes [insulin-dependent diabetes mellitus (IDDM)] is associated with an extreme T helper (Th) 1 phenotype of hVα24- Jα18<sup>+</sup> T cells, suggesting that there is a strong link between hVα24-Jα18<sup>+</sup> T cells and human type 1 diabetes. On the other hand, there is evidence that IL-4 exerts a dominant-negative effect on the progression to IDDM in non-obese diabetic (NOD) mice (45–47), and NOD mice with the mVα14-Jα18 transgene were protected from diabetes (48), indicating that not only the number but also the phenotype of iNKT cells influences the incidence of diabetes both in humans and mice. The fact that the gut microbiota can impact iNKT cell development and functions (49–51) and is associated with diabetes onset, regulatory imbalance, and IFN-γ levels in NOD mice should be also considered (52). Another iNKT cell-overexpressing mouse line was derived from iPSCs. iPSCs hold tremendous potential for applications not only in drug discovery, regenerative medicine, and cell replacement therapy (53–55), but also in basic biology. We have succeeded in reprogramming splenic iNKT cells from WT B6 mice (18). These iPSC-iNKT cells could be differentiated into iNKT cells *in vitro* and secreted large amounts of IFN-γ. Importantly, iPSC-iNKT cells recapitulated the known adjuvant effects of natural iNKT cells and suppressed tumor growth *in vivo*. These studies demonstrate the feasibility of expanding functionally competent iNKT cells *via* an iPSC phase, an approach that may be adapted for iNKT cell-targeted therapy in humans (56, 57). We further succeeded in generating iNKT cell cloned *Trav11*- *Traj18<sup>+</sup>/<sup>+</sup>* mice from iPSC-iNKT cells through germline transmission and breeding with WT B6 mice (19). The absolute numbers and percentages of α-GalCer/CD1d dimer<sup>+</sup> TCRβ+ cells in the thymus and periphery of *Trav11*-*Traj18<sup>+</sup>/<sup>+</sup>* mice were elevated by 10–20 fold compared to B6 mice and 10–20-fold compared to B6 mice and by 3–10-fold compared to iNKT-TCRα transgenic mice due to the bypass of TCRα rearrangement at the double-negative (DN) stage. They lacked γδ T cells due to the deletion of the δ locus and had reduced numbers of αβ T cells while NK, B, and DC numbers were normal. However, the surface phenotype of α-GalCer/CD1d

Figure 2 | CD1d restricted cells in iPSC-invariant natural killer T (iNKT)-derived cloned mice and iNKT cell subtypes in the thymus of B6 mice. (A) Percentage of CD1d-restricted α-GalCer/CD1d dimer+TCRβ+ cells positive for the indicated cell surface molecules in WT B6, *Trav11*-*Traj18+/+* and *Cd1d1*−/−*Trav11*-*Traj18+/<sup>+</sup>* mice. (B) The iNKT cell subtypes previously characterized in the thymus of B6 mice. Their phenotypes and developmental pathways in the thymus are also shown. Function of iNKT cells is acquired through the development in the thymus distinct from conventional αβ T cells. All of the iNKT subtypes may arise from the CD1d-restricted α-GalCer/CD1d dimer+TCRβ+ cells in *Cd1d1*−/−*Trav11*-*Traj18+/+* mice in panel (A), while it still remains to be elucidated which signals control the divergence of iNKT1, iNKT2, and iNKT17 subsets.

dimer<sup>+</sup> TCRβ+ cells in *Trav11*-*Traj18<sup>+</sup>/<sup>+</sup>* mice was different from that in WT B6 mice; there was a partial reduction of CD44<sup>+</sup> cells and changes in the CD4<sup>+</sup>:NK1.1<sup>+</sup> ratio (19). We think this is due to the shortage of CD1d molecules in the face of an excess number of α-GalCer/CD1d dimer+TCRβ+ cells because the surface phenotype of the iNKT cells changed into the mature phenotype as seen in WT B6 when these cells were sorted and transferred into *Traj18*<sup>−</sup>/<sup>−</sup> mice (58). Generation of *Trav11*-*Traj18<sup>+</sup>/<sup>+</sup>* mice carrying a *Cd1d1* transgene should clarify this point.

*Trav11*-*Traj18<sup>+</sup>/<sup>+</sup>* mice on a *Cd1d1*<sup>−</sup>/<sup>−</sup> background have also been generated (59). Interestingly, these mice have thymic CD1d-restricted α-GalCer/CD1d dimer<sup>+</sup>TCRβ+ cells, which are considered to be iNKT cells before CD1d selection. The frequency of positive cells of CD44, CD4, and NK1.1 by thymic α-GalCer/ CD1d dimer<sup>+</sup>TCRβ+ cells from *Cd1d1*<sup>−</sup>/<sup>−</sup>*Trav11*-*Traj18<sup>+</sup>/<sup>+</sup>* mice is further lower than those from *Trav11*-*Traj18<sup>+</sup>/<sup>+</sup>* mice (**Figure 2A**), suggesting that CD1d plays a role in the induction of these surface molecules on iNKT cells.

### iNKT CELL DEVELOPMENT IN THE THYMUS

Until recently, the iNKT cell field had embraced a sequential lineage developmental model (60) in which "developmental intermediates" produce Th2-type cytokines and "mature" NK1.1<sup>+</sup> iNKT cells produce Th1 cytokines. However, based on the finding of the expression of distinct transcription factors, T-bet (*Tbx21*), PLZF (*Zbtb16*), and RORγt (*Rorc*) (61) and surface markers, CD4, NK1.1, and IL-17RB (7) in iNKT cell subsets, we considered an alternative "lineage diversification" model for iNKT cells (62), analogous to the differentiation of effector Th cells (63) and innate lymphoid cells (64, 65). Three major subsets of iNKT cells (iNKT1, iNKT2, and iNKT17) that produce distinct cytokines have been defined (7, 61, 66, 67) (**Figure 2B**), and these represent diverse lineages and not developmental stages, as previously thought. In fact, it was recently reported that some iNKT1 cells developed through an alternative DN pathway that bypasses the DP pathway (68), supporting the above findings that iNKT subtypes possibly arise from different precursors in the thymus. Of note, thymic α-GalCer/CD1d dimer<sup>+</sup>TCRβ+ cells from *Cd1d1*<sup>−</sup>/<sup>−</sup>*Trav11*-*Traj18<sup>+</sup>/<sup>+</sup>* mice described above exhibit the precursors of all iNKT cell subtypes. The poised effector state is

#### REFERENCES


acquired during development in the thymus, where iNKT precursors differentiate into one of three distinct subsets, while it has still been unclear which signals control the divergence of iNKT1, iNKT2, and iNKT17. The precise molecular mechanisms should be clarified that are important for iNKT lineage diversification but are dispensable for conventional αβ T cell development. Taken collectively, it can be proposed that the acquisition of diverse functional characteristics by iNKT subtypes might be dependent on the environment providing an appropriate cytokine milieu, as well as on the cytokine receptor signaling in precursor cells undergoing CD1d selection.

# CONCLUSION

The gene manipulated mice described here will reveal more insights into mouse iNKT cell development and function, and these insights should also be applicable to human iNKT cell studies. Overall, some reported differences between *Cd1d*<sup>−</sup>/<sup>−</sup> and *Traj18*<sup>−</sup>/<sup>−</sup> mice are likely due to the loss of some T cell populations including MAIT cells in the original *Traj18*<sup>−</sup>/<sup>−</sup> mice. *Traj18*<sup>−</sup>/<sup>−</sup> mice with unbiased TCR diversity will inform us of the actual role of iNKT cells and type II NKT cells. *Cd1d1*<sup>−</sup>/<sup>−</sup>*Trav11*-*Traj18<sup>+</sup>/<sup>+</sup>* mice may further reveal differences in iNKT cell thymic development and may account for the observed mouse strain specific differences. iNKT cells hold great promise for treatment of a myriad of diseases, and these gene manipulated mice will be invaluable in deciphering the role of iNKT cells in health and disease.

# AUTHOR CONTRIBUTIONS

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

# ACKNOWLEDGMENTS

This work was supported mainly by a Grant-in-Aid for challenging Exploratory Research (16K14591) by the Japan Society for the Promotion of Science to HW and in part by the Foundation for Young Scholars of Jiangxi Province, China (20161BAB215259) and Key Research and Development Project of Jiangxi Province, China (20161ACG70018) to YR. We thank Ms. Kazuko Okunuki for secretarial assistance.


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

*Copyright © 2018 Ren, Sekine-Kondo, Tateyama, Kasetthat, Wongratanacheewin and Watarai. 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.*

# Regulatory Roles of invariant natural Killer T Cells in Adipose Tissue inflammation: Defenders Against Obesity-induced Metabolic Complications

*Yoon Jeong Park1,2, Jeu Park1 , Jin Young Huh1,3, Injae Hwang1 , Sung Sik Choe1 and Jae Bum Kim1,2\**

*1Department of Biological Sciences, Center for Adipose Tissue Remodeling, College of Natural Sciences, Institute of Molecular Biology and Genetics, Seoul National University, Seoul, South Korea, 2Department of Biophysics and Chemical Biology, Seoul National University, Seoul, South Korea, 3Department of Medicine, University of California San Diego, San Diego, CA, United States*

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Masashi Satoh, Kitasato University School of Medicine, Japan Hiroshi Watarai, University of Tokyo, Japan*

> *\*Correspondence: Jae Bum Kim jaebkim@snu.ac.kr*

#### *Specialty section:*

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

*Received: 20 April 2018 Accepted: 28 May 2018 Published: 11 June 2018*

#### *Citation:*

*Park YJ, Park J, Huh JY, Hwang I, Choe SS and Kim JB (2018) Regulatory Roles of Invariant Natural Killer T Cells in Adipose Tissue Inflammation: Defenders Against Obesity-Induced Metabolic Complications. Front. Immunol. 9:1311. doi: 10.3389/fimmu.2018.01311*

Adipose tissue is a metabolic organ that plays a central role in controlling systemic energy homeostasis. Compelling evidence indicates that immune system is closely linked to healthy physiologic functions and pathologic dysfunction of adipose tissue. In obesity, the accumulation of pro-inflammatory responses in adipose tissue subsequently leads to dysfunction of adipose tissue as well as whole body energy homeostasis. Simultaneously, adipose tissue also activates anti-inflammatory responses in an effort to reduce the unfavorable effects of pro-inflammation. Notably, the interplay between adipocytes and resident invariant natural killer T (iNKT) cells is a major component of defensive mechanisms of adipose tissue. iNKT cells are leukocytes that recognize lipids loaded on CD1d as antigens, whereas most other immune cells are activated by peptide antigens. In adipose tissue, adipocytes directly interact with iNKT cells by presenting lipid antigens and stimulate iNKT cell activation to alleviate pro-inflammation. In this review, we provide an overview of the molecular and cellular determinants of obesity-induced adipose tissue inflammation. Specifically, we focus on the roles of iNKT cell-adipocyte interaction in maintaining adipose tissue homeostasis as well as the consequent modulation in systemic energy metabolism. We also briefly discuss future research directions regarding the interplay between adipocytes and adipose iNKT cells in adipose tissue inflammation.

Keywords: adipocytes, invariant natural killer T cells, obesity, inflammation, CD1d

# INTRODUCTION

White adipose tissue (WAT) is a central controller of lipid and glucose homeostasis that communicates locally and with distant tissues. The mass of WAT expands or reduces dynamically in response to nutritional states. WAT actively senses nutritional changes and accordingly stores extra energy in the form of triglycerides or supplies nutrients to other organs (1, 2). Generally, WAT expands both by hyperplasia (an increase in mature adipocyte number) and hypertrophy (an increase in mature adipocyte size) (3, 4). White adipocytes are the major cell type of WAT and normally contain a single large lipid droplet. White adipocytes crosstalk with multiple cell types in both local and remote tissues *via* the secretion of a variety of signaling molecules (1, 2).

Traditionally, the immune system has been considered central to the elimination of pathogenic microbes and toxic or allergenic molecules that threaten the normal homeostasis of the host. A more recent addition to the broad discussion of immunity in health and diseases is the role of the interplay between immune response and metabolism (5). In particular, the roles of this interplay in obesity and metabolic diseases have been suggested by the findings that the immune program is intimately linked to physiological and pathological changes in WAT (6–9). One example is the inappropriately active and/or overactive immune responses in WAT in obesity and its related metabolic diseases. Along with enhanced WAT expansion, obesity induces both quantitative and qualitative changes in WAT immunity, which potentiates the dysfunction of adipose tissue as well as systemic energy homeostasis (10–12).

Among the resident immune cells in WAT, invariant natural killer T (iNKT) cells are regarded as one of the key players linking dynamic changes in adipocyte metabolisms to WAT homeostasis (13). In the following review, we briefly discuss different molecular and cellular factors involved in the control of WAT immunity in obesity. In particular, we emphasize the roles of the interaction between iNKT cells and adipocytes in maintaining WAT homeostasis as well as whole body energy metabolism.

#### WAT IMMUNITY IN OBESITY

Obesity is defined as the massive expansion of WAT due to the imbalance between caloric intake and energy expenditure. In adult obesity, WAT expansion features by dramatic increases in the number of hypertrophic adipocytes that are significantly related to detrimental changes in WAT, including hypoxia, oxidative stress, and insulin resistance (3, 4). Obesity is strongly associated with interrelated metabolic diseases, including insulin resistance, type 2 diabetes, and cardiovascular disease, which impose a high social burden in terms of quality of life (3, 4). Given that WAT is the major organ for energy storage and mobilization, most previous obesity-related studies focused on finding abnormalities in adipocyte physiology and metabolism in effort to understand the link between obesity and metabolic diseases (14). However, the recent discovery of adipokines, an array of mediators secreted by adipose tissue, has revised the concept of WAT being merely a fat storage depot (3, 4). Instead, it has become clear that WAT is a dynamic endocrine system that is crucial in the regulation of systemic energy homeostasis.

Adipokines include angiogenic proteins, metabolic regulators, and inflammatory mediators. Most adipokines including leptin and adiponectin act as the bridge between the functional status of WAT and other organs, modulating systemic energy metabolism (3, 4). Among various adipokines, the identification of inflammatory mediators has clarified the connection between immunity and obesity and its related metabolic diseases (15). The first study that established the reframing of obesity as an inflammatory condition demonstrated the detrimental effect of tumor necrosis factor alpha (TNF-α), an inflammatory mediator secreted by adipose tissues, on insulin resistance in many animal models of obesity (16). Subsequent studies enforced the idea that alterations in WAT immunity are closely associated with dynamic changes in energy homeostasis in obesity and metabolic diseases (8, 9).

One hallmark characteristic of WAT immunity in obesity is chronic low-grade inflammation, which leads to a modest increase in circulating pro-inflammatory factors (8, 9). In a lean state, WAT immunity is skewed toward the anti-inflammatory phenotype, which supports tissue expansion (3, 4). In obesity, nutritional stresses promote the secretion of inflammatory cytokines and acute-phase reactants including TNF-α, interleukin (IL)-6, and serum amyloid A in WAT. Although WAT simultaneously increases the release of anti-inflammatory cytokines, including IL-4, IL-10, and IL-2 to counteract the unfavorable effects of inflammation, WAT immunity eventually shifts toward an inflammatory state, leading to prolonged inflammation in obesity (8, 9, 13).

### KEY PLAYERS OF WAT INFLAMMATION IN OBESITY

White adipose tissue is a heterogeneous organ composed of white adipocytes, mural endothelial cells, fibroblasts, and various immune cells, including macrophages, T cells, B cells, and NKT cells. These cells are engaged in maintaining the well-being of adipocytes, clearance of apoptotic cells, and retaining healthy physiological functions of WAT. Particularly, obesity-induced multiple insults, including epigenetic malfunction, hypoxia, and oxidative stress have complex impacts on adipose tissue inflammation by altering inter-cellular interaction between adipocytes and immune cells. For instance, recent studies have underscored the important roles of epigenetic modulators in the progression of adipokine dysregulation and subsequent adipose tissue inflammation (17, 18). In obese WAT, hypoxia and oxidative stress work in concert to promote dysfunction of adipocytes and lead to the stimulation of inflammatory signaling pathways in neighboring immune cells (19–22).

Among various cells in WAT, adipocytes act as both sensors and messengers that form an early warning network of WAT immunity (**Figure 1**). In response to excessive nutritional overload, adipocytes undergo both metabolic and immunologic reprogramming, which includes dramatic changes in metabolite and lipid compositions (23, 24). Following reprogramming, adipocytes alert neighboring immune cells to eliminate such stresses through the secretion of an array of cytokines and presentation of certain types of antigens that reflect dynamic alterations in WAT under stressful conditions (25–28).

Macrophages are another key player in adipose tissue inflammation. These cells are the primary source of pro-inflammatory cytokines in obese WAT (8, 9). Monocytes differentiate into classically activated macrophages (M1) or alternatively activated macrophages (M2) according to the stimuli. M1 macrophages are pro-inflammatory, whereas M2 macrophages are anti-inflammatory (29). The balance between M1 and M2 macrophages is

crucial to maintain WAT homeostasis, and the disturbance in this balance triggers the pathologic dysfunction of WAT (**Figure 1**). In obesity, the proportion of M1 macrophages is significantly increased compared to M2 macrophages, which confers a vicious cycle of WAT inflammation through having multiple impacts on other cells (29–31).

Simultaneously, WAT also promotes anti-inflammatory responses in effort to alleviate pathologic dysfunction of adipose tissue (8, 9). In WAT, there are many types of cells involved in anti-inflammatory responses, including M2 macrophages, eosinophils, regulatory T (Treg) cells, and iNKT cells. Depletion of anti-inflammatory cells in animal models of obesity accelerates WAT inflammation and consequently aggravates metabolic disorders including insulin resistance (26, 32–36). These studies suggest that, despite of a dominant role of pro-inflammatory response, anti-inflammatory response is still required to dampen WAT inflammation.

# DISTINCT CHARACTERISTICS OF NKT CELLS

NKT cells are innate-like T lymphocytes that function similarly to innate cells, displaying less specificity and more rapid activation compared to adaptive immune cells (37, 38). NKT cells can be activated by exogenous or endogenous lipid antigens, and by cytokines produced by antigen-presenting cells (APCs). Upon activation, NKT cells rapidly secrete a variety of cytokines. Moreover, NKT cells express cytotoxic granules containing perforin and granzyme, and induce apoptosis of target cells (37). NKT cells are largely categorized into three types: iNKT cells (type I), diverse NKT cells (dNKT, type II), and NKT-like cells (37). Although both type I and type II NKT cells recognize lipid antigens loaded on the MHC class I-like family protein CD1d, they are activated by distinct types of lipid antigens *via* the expression of different repertoires of T-cell receptors (37, 38). For instance, iNKT cells express a conserved semi-invariant TCR and are potently stimulated by α-galactosylceramide (α-GC), a marine sponge-derived glycolipid (39). Type II NKT cells express a broader TCR repertoire and sulfatide is one of the major antigens for these cells (40, 41). Among subsets of the NKT cell population, iNKT cells have been suggested to modulate WAT immunity in both lean and obese individuals (25, 26, 34, 35, 42). One of the interesting features of iNKT cells is their remarkable functional plasticity with both pro- and anti-inflammatory characteristics. Upon activation signaling, iNKT cells secret either robust Th1-type or Th2-type cytokines according to the nature of activating stimuli and types of APCs and cytokines (43).

# ADIPOCYTES AS PROFOUND APCs FOR iNKT CELLS

While conventional naïve T cells are mostly localized to immune organs, iNKT cells reside in many tissues, with a relatively high abundance in the liver and WAT (25, 44). Along with iNKT cells, such tissues appear to have distinct pools of APCs that rapidly process and present lipid antigen to confer tissue-specific function to iNKT cells. Generally, dendritic cells, macrophages, and B cells are considered "professional" APCs as they express various components required for lipid antigen synthesis and presentation. They function as the major APCs that modulate the differentiation and activity of iNKT cells in lymphatic organs, including the thymus and spleen (38, 43). "Non-professional" APCs are not conventional APCs, but express CD1d. Interestingly, iNKT cells residing in metabolic tissues are promptly activated by non-professional APCs (25, 44). Among the APCs in WAT, adipocytes seem to be a efficient non-professional APCs with the highest levels of CD1d in parallel with expression of other factors required for lipid antigen presentation (13, 25). As obesity is closely associated with major changes in the lipid repertoire of adipocytes, considerable attention has been directed to the role of adipocyte-derived lipids in the control of WAT inflammation (45–47). Owing to the primary function of adipocytes being endocrine cells, most studies related to obesity have focused on the link between the immune system and secreted lipids (45, 46). However, given that adipocytes express the highest level of CD1d among the resident APCs in WAT and since lipid metabolites can act as "antigenic" lipids after being loaded on CD1d, it is very likely that endogenous lipid metabolites derived from adipocytes may act as antigenic lipids (13, 25). Indeed, adipocyte cell line and primary adipocytes isolated from both mouse and human activate iNKT cells and stimulate cytokine secretion (26, 35). Moreover, adipocytes are able to promote cytokine secretion of iNKT cells without exogenous lipid antigens, such as α-GC, implying the presence of adipocyte-derived lipid antigens (28, 35). Very recently, we and other groups have reported the *in vivo* role of adipocytes as APCs by the use of an adipocyte-specific CD1d knockout (CD1dADKO) mouse model (25, 28). In these studies, lean CD1dADKO mice exhibit reduced number of iNKT cells in WAT and have different cytokine profiles in adipose iNKT cells compared to control mice in obesity (25).

In nature, iNKT cells recognize a vast range of lipid antigens, which includes microbial lipids and self-lipid antigens. Generally, such lipid antigens are composed of sugar moieties linked to a lipid backbone that can either be based on a ceramide or a diacylglycerol (48–51). Recently, several reports demonstrated that obesity induces the activation of enzymes involved in ceramide synthesis in conjunction with the elevation of cellular ceramides in mouse and human WAT (52, 53). Ceramide is found in high concentrations within cell membranes and is used as a precursor molecule for the synthesis of glycolipids. Given that many self-lipid antigens contain ceramide backbones and that these antigens are more potent than antigens based on diacylglycerol, an increase in ceramide-mediated glycolipids might contribute to the enrichment of lipid antigen pools in adipocytes as well as subsequent activation of adipose iNKT cells (54). However, the structural basis of adipocyte-derived antigens including types of lipid backbone and sugar moieties has not been fully explored and is a promising avenue of investigation. Also, further analyses of the effects of nutrient stresses on characteristics of adipocytederived lipid antigens that modulate expansion of iNKT cell population and specific Th1 or Th2 cytokine profiles could have promising therapeutic potential concerning WAT inflammation.

# ADIPOSE iNKT CELLS AND THEIR ROLES IN WAT INFLAMMATION

White adipose tissue harbors a distinct pool of cells of the immune system (**Figure 1**). The characteristics of the cells are often governed by adipose tissue-specific cues including antigens. In lean WAT, adipose iNKT cells account for 1–20% of the resident T-cell pool (34). The majority of adipose iNKT cells are tissue-resident lymphocytes, whereas a small portion of the cells is infiltrated into WAT (42). Adipose iNKT cells produce anti-inflammatory cytokines, such as IL-4 and IL-10, and regulate the function of M2 macrophages and Treg cells, which contribute to the maintenance of WAT homeostasis (**Figure 1**). Recent reports described several characteristics unique to adipose iNKT cells. The transcriptome of adipose iNKT cells differs from the transcriptome of iNKT cells residing in other tissues (42). Among surface markers defining iNKT cells, the expression of CD4 and NK1.1 is relatively low in adipose iNKT cells (42). Also, adipose iNKT cells are less dependent on promyelocytic leukemia zinc finger (PLZF), a key transcription factor responsible for iNKT activation (42). Adipose iNKT cells express little PLZF compared to other iNKT cells and the quantity of adipose iNKT cells is not affected in *Plzf*<sup>+</sup>/<sup>−</sup> mice (42). Instead, the levels of T-bet, GATA-3, and E4BP4 are high in adipose iNKT cells (42). Finally, adipose iNKT cells seem to be chronically activated, while iNKT cells in the rest of body are in a poised state that requires an additional signal for rapid cytokine production (42). Such a distinct activation state of adipose iNKT cells results from special microenvironments, including lipid antigens, cytokines, and adipokines in WAT.

In obesity, WAT undergoes dramatic changes in the immune system favoring a pro-inflammatory environment. Notably, the number of adipose iNKT cells significantly declines in parallel with elevation of inflammation in WAT (26, 34, 55). Recent reports from several groups including ours have shown that iNKT-cell-deficient mouse models (*J*α*18<sup>−</sup>/<sup>−</sup>* and *CD1d<sup>−</sup>/<sup>−</sup>* mice, which are deficient in iNKT cells and both iNKT cells and type II NKT cell, respectively) are more susceptible to obesity, adipose tissue inflammation, as well as insulin resistance on a high-fat diet regimen (26, 34). These phenotypes are reversed by the adoptive transfer of iNKT cells or specific activation of iNKT cells with α-GC, supporting the protective role of iNKT cells in obesity (26, 34). Furthermore, the impaired induction of arginase-1, an M2 macrophage marker gene, was reported in *CD1d<sup>−</sup>/<sup>−</sup>* mice (33). Collectively, these studies reveal that adipose iNKT cells are crucial to maintain WAT homeostasis due to their ability to secrete anti-inflammatory cytokines. Thus, the dramatic decrease in adipose iNKT would constitute an important initiator of the microenvironment favorable for inflammation in WAT. However, whether adipose iNKT cells act only as anti-inflammatory cells in obesity is debatable. Other studies suggest that iNKT cell deficiency leads to a decrease in obesity-induced adipose tissue inflammation and insulin resistance (56, 57). For instance, Satoh et al. demonstrated that CD1dADKO mice exhibit improved insulin sensitivity and adipose tissue inflammation (28). There are also other reports suggesting that iNKT cells are dispensable for adipose tissue inflammation as well as systemic energy homeostasis (58, 59). Multiple factors could be attributable to such a difference in adipose iNKT polarization in relation to obesity and adipose tissue inflammation. These factors include the type of diet, duration of diet intervention, types of control mouse groups, and gut microbiota (25, 26, 28, 34, 56). Particularly, the potential effect of gut microbiota on adipose iNKT cells appears to be interesting. A very recent study demonstrated that glucagon-like peptide-1 (GLP-1), a gut hormone that is used to treat obesity and diabetes, activates adipose iNKT cells and enhances the secretion of antiinflammatory cytokines including IL-10 in WAT (60). Like other gut hormones, serum GLP-1 level is sensitively modulated by the composition of gut microbiota in response to changes in the nutritional status (61). Thus, it is probable that different composition of gut microbiota among different laboratories would impact on GLP-1-adipose iNKT axis, accounting for the differences in the function of adipose iNKT cells that have been reported in obesity.

## FUTURE RESEARCH DIRECTIONS IN ADIPOSE iNKT CELLS

White adipose tissue is characterized by a unique immune system that dynamically responds to nutritional stresses. With respect to iNKT cells, WAT provides a special microenvironment that is enriched in profound APCs and diverse activating stimuli (lipid antigens) and cytokines. Although recent reports have demonstrated the distinct characteristics and physiological roles of adipose iNKT cells, the interconnected mechanisms of interplay between adipose iNKT cells and other cells in WAT need to be elucidated (**Figure 2**). The kinds of endogenous lipid antigens presented by adipocytes and the underlying molecular mechanisms that mediate dynamics of lipid antigen presentation by adipocytes remain to be determined (**Figure 2A**). Although it has been reported that several regulators of adipocyte differentiation including peroxisome proliferator-activated receptor gamma and CCAAT/enhancer-binding protein (C/EBP)-β and -δ control CD1d expression, the regulatory mechanisms of other necessary machineries involved in lipid antigen synthesis and presentation

number of iNKT cells in WAT. However, the regulatory mechanisms that modulate interplay between adipocytes and adipose iNKT cells in obesity are still unclear. (A) Function of adipose iNKT cells is influenced by (1) "lipid antigens" loaded on CD1d and (2) "co-stimulatory molecules." In obesity, a variety of factors can lead to dynamic changes in both types of lipid antigens and combination of costimulatory molecules, resulting in alteration of the functionality of adipose iNKT cells. (B) In addition to cytokine production, iNKT cells have the ability to directly or indirectly induce apoptosis. iNKT cells can express fasL and activate NK cells to kill target cells. NK cells activated by IFN-γ secrete perforin/granzyme to promote cell death processes. One of the characteristics of obese WAT is an increase in the number of hypertrophic adipocytes that are susceptible to apoptosis. Therefore, it would be interesting to determine whether adipose iNKT cells can contribute to obesity-induced death of hypertrophic adipocytes in WAT.

in adipocytes are still unclear (26, 62). Moreover, it has been proposed that types of co-stimulation can affect iNKT cell functions. For instance, certain types of co-stimulatory molecules, such as CD40/CD40L and CD28/B7.2, are required for the Th1-skewed responses of iNKT cells (63, 64). Therefore, it is worth studying that the dynamics of a repertoire of co-stimulation in WAT in response to nutritional stress in obesity. Finally, even though the major functions of iNKT include cytokine production and cytolytic activity, most studies involving adipose iNKT cells have focused on the contribution of iNKT cell-derived cytokines to adipose tissue inflammation. Given that hypertrophic adipocytes are prone to apoptosis in obese WAT, it would be both interesting and important to investigate the involvement of adipose iNKT cells in the clearance of hypertrophic adipocytes in obesity (**Figure 2B**).

## CONCLUSION

The incidence of obesity and metabolic diseases has risen dramatically during the past few decades. Adipose tissue inflammation

## REFERENCES


is considered to be one of the key mechanisms linking obesity and metabolic diseases. A growing body of evidence indicates that the crosstalk between adipocytes and adipose iNKT cells is crucial to regulate WAT homeostasis as well as for adipose tissue inflammation. In this regard, understanding of regulatory mechanisms that modulate interplay between adipocytes and adipose iNKT cells will provide a new approach to control adipose tissue inflammation and metabolic diseases.

# AUTHOR CONTRIBUTIONS

All authors listed have made a substantial contribution to the work.

# FUNDING

This work was supported by the National Research Foundation (NRF) funded by the Korean government (the Ministry of Science, ICT & Future planning) [2011-0018312].


**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 Park, Park, Huh, Hwang, Choe and Kim. 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.*

# Role of natural Killer T Cells in the Development of Obesity and insulin Resistance: insights From Recent Progress

*Masashi Satoh and Kazuya Iwabuchi\**

*Department of Immunology, Kitasato University School of Medicine, Sagamihara, Japan*

Natural killer T (NKT) cells play important roles in adipose tissue inflammation, and thus influence the development of diet-induced obesity and insulin resistance. The interactions between cluster of differentiation (CD)1d and NKT T cell receptor are thought to be critical in this process, as demonstrated in two NKT cell-deficient mouse models systemic CD1d gene knockout (KO) and prototypic Jα18 KO mice. The latter lacks some repertoires besides invariant (i)NKT cells due to manipulation of the Jα18 gene segment; therefore, the role of iNKT vs. variant NKT cells must be reinterpreted considering the availability of new Jα18 KO mice. NKT cells have varied roles in the development of obesity; indeed, studies have reported contradictory results depending on the mouse model, diet, and rearing conditions, all of which could affect the microbiome. In this mini-review, we discuss these points considering recent findings from our laboratory and others as well as the role of NKT cells in the development of obesity and insulin resistance based on data obtained from studies on conditional CD1d1 KO and new Jα18 KO mice generated through gene editing.

Keywords: natural killer T cell, cluster of differentiation 1d, adipocyte, lipid, obesity, insulin resistance, adipose tissue inflammation

# INTRODUCTION

#### Obesity as a Chronic Inflammatory Disorder

Inflammation in adipose tissue (AT) is induced by hypertrophy of adipocytes that secrete inflammatory cytokines and chemokines (1) and thus recruit various immune cells such as macrophages, T cells [αβ, γδ, regulatory T cells (Tregs), and natural killer T (NKT) cells], B cells, NK cells, and leukocytes that exist in a steady state in immune organs (2, 3). Fat accumulation is a major factor contributing to meta-inflammation and metabolic dysfunction (1, 4). Obesity alters the microenvironment in AT from an anti-inflammatory to a pro-inflammatory state, leading to impaired immune balance (5, 6). Visceral (V)AT in the lean state predominantly contains M2 macrophages, eosinophils, and Tregs that suppress inflammation and maintain insulin sensitivity (7, 8). By contrast, VAT in obese individuals has more M1 macrophages, cluster of differentiation (CD)8<sup>+</sup> T cells, NK cells, B cells, and neutrophils that enhance inflammation and reduce insulin sensitivity (9–13). Notably, chronic low-grade inflammation accompanied by obesity is implicated in the etiology of lifestylerelated diseases such as atherosclerosis, type 2 diabetes, and various cancers (14).

#### *Edited by:*

*Yun-Cai Liu, Tsinghua University, China*

#### *Reviewed by:*

*Toshinori Nakayama, Chiba University, Japan Koji Yasutomo, Tokushima University, Japan*

#### *\*Correspondence:*

*Kazuya Iwabuchi akimari@kitasato-u.ac.jp*

#### *Specialty section:*

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

*Received: 03 April 2018 Accepted: 28 May 2018 Published: 11 June 2018*

#### *Citation:*

*Satoh M and Iwabuchi K (2018) Role of Natural Killer T Cells in the Development of Obesity and Insulin Resistance: Insights From Recent Progress. Front. Immunol. 9:1314. doi: 10.3389/fimmu.2018.01314*

# NKT Cells

Natural killer T cells are a unique T cell subset that recognize lipid antigen presented by CD1d (15, 16). α-Galactosylceramide (α-GalCer) is a prototypical ligand recognized by invariant (i)NKT cells that harbors an invariant T cell receptor (TCR) α-chain (Vα14-Jα18 in mouse and Vα24-Jα18 in human) (17). Another type of NKT cell known as variant (v)NKT cells express diverse TCRs that are presumed to recognize various lipid antigens including sulfatide (18). Activated NKT cells secrete large amounts of cytokine that modulate immune balance, implying that they can either enhance or suppress inflammatory and immune responses. NKT cells have been reported to exacerbate, protect against, or have no role in the development of obesity through modulation of AT inflammation (19).

Here, we summarize the correlation between the CD1d/ NKT cell axis and obesity with a focus on AT inflammation and discuss factors that may contribute to the discrepancies among reports considering recent progress.

# OPPOSING FUNCTIONS OF NKT CELLS IN THE DEVELOPMENT OF OBESITY

Many studies have examined whether NKT cells play a role in diet-induced obesity (DIO) and have reported variable results.

#### NKT Cells as an Aggravator of DIO

Ohmura *et al*. first demonstrated that iNKT cells induce AT inflammation and glucose intolerance in β2-microglobulin (β2m) knockout (KO) mice fed a high-fat diet (HFD) and treated with the NKT cell stimulator α-GalCer (20). Since β2m KO mice also lack CD8+ T cells, the role of NKT cells in obesity has been examined using CD1d KO mice fed an HFD. However, two subsequent studies showed that NKT cell deficiency is insufficient to protect against or aggravate DIO (21) and that CD1d is important for the modulation of metabolic functions *via* a non-NKT cell-mediated mechanism (22). By contrast, we showed that CD1d KO mice lacking both iNKT and vNKT cells showed a reduced body weight (BW) gain along with improved AT inflammation and insulin resistance (23). Meanwhile, Jα18 KO mice lacking only iNKT cells demonstrated similar pathology to wild-type (WT) mice, suggesting that vNKT cells may contribute to DIO in the absence of iNKT cells (23). Wu *et al*. reported that iNKT cells responded to lipid excess and produced pro-inflammatory cytokines that promoted AT inflammation and insulin resistance (24).

#### iNKT vs. vNKT Cells

We investigated whether iNKT cells (24) or vNKT cells (23) contribute to the exacerbation of DIO, since distinct measures must be taken to control either subset. We speculated that vNKT cells contribute to the development of DIO in the absence of iNKT cells based on the aforementioned results (i.e., no difference in BW between WT and Jα18 KO mice on an HFD) and some additional observations (23): (1) the NK1.1<sup>+</sup>TCRβ+ population in AT was activated upon consumption of an HFD and contained more CD8+ but fewer CD4−CD8− subsets in Jα18 KO (referred hereafter as Jα281 KO) (25) mice, which differed from observations in either WT or CD1d KO mice; (2) WT mice harbored more non-iNKT (=vNKT) cells in AT; and (3) hepatic mononuclear cells from Jα281 KO mice [which are enriched in vNKT cells including CD1d<sup>+</sup> antigen-presenting cells (APCs)] transferred insulin resistance to CD1d KO hosts.

However, the Jα281 KO strain was shown to exhibit a marked reduction in TCR diversity, which could affect immune responses (26). Four novel Jα18 KO mouse strains were independently generated after the report (26) by deleting only the *T-cell receptor alpha joining* (*Traj*)*18* locus and leaving the remaining Jα repertoire unperturbed using novel technologies [clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein-9 nuclease or transcription activator-like effector nuclease] (27–30). New Traj18 KO (referred hereafter as simply Traj18 KO) mice gained less weight and had heightened sensitivity to insulin compared with WT mice, suggesting that iNKT cells play a pathogenic role in DIO (30). In that study, the mice were fed the same HFD (HFD-32; CLEA Japan, Tokyo, Japan) as those in our experiments, and Jα281 KO mice fed this diet showed similar BW gain to WT mice. The interpretation of the results from Traj18 KO mice was that iNKT but not vNKT cells exacerbate the development of DIO. Experiments using Vα14-Jα18 transgenic mice lacking low-density lipoprotein receptor also demonstrated that the abundance of iNKT cells increased adiposity by inducing metabolic abnormalities and AT inflammation (31). The DIO results from Traj18 KO mice also imply that reduced TCR diversity or the lack of particular T cell subsets in Jα281 KO but not Traj18 KO mice account for the discrepancy among reports on the involvement of iNKT vs. vNKT cells. Mucosal-associated invariant T (MAIT) cells that utilize Jα33 may be lost in Jα281 KO mice and may thus affect the development of obesity, as was suggested in studies of human subjects (32, 33). However, the actual role of MAIT cells in obesity and their involvement (or that of other T cell subsets) in DIO in Jα281 KO mice require further investigation.

#### Protective Role of NKT Cells Against Obesity

Some studies have reported that NKT cells play a protective role against obesity. Regulatory cytokines such as interleukin (IL)-4 and -10 produced by AT iNKT cells prevented the development of DIO (34, 35) and insulin resistance even in mice fed a low-fat diet (36). IL-13-producing innate immune cells such as type 2 innate lymphoid cells (ILC2s), iNKT cells, and vNKT cells were shown to prevent DIO by suppressing inflammation in AT (37). AT-resident iNKT cells express transcriptional repressor E4-binding protein (E4BP) 4 (also known as nuclear factor, IL-3regulated) but not promyelocytic leukemia zinc finger protein (PLZF), unlike iNKT cells in other tissues, reflecting their antiinflammatory phenotype (38); moreover, IL-10-producing iNKT cells (NKT10) are enriched in subcutaneous white (W)AT (39). Interestingly, an F108Y substitution in TCRβ altered NKT cell development to an adipose-like phenotype (40) without affecting TCR activation nor its ability to bind CD1d–ligand complexes, suggesting that a hydrophobic patch created after TCRα–TCRβ pairing is essential for the development of a distinct NKT cell population (40). iNKT cells with TCRβ F108Y express E4BP4 but not PLZF, similar to AT-resident NKT cells (38). These results

Satoh and Iwabuchi NKT Cell and Obesity

suggest that NKT cells in AT constitute a specialized subset and are not regular iNKT cells that localize there as passers-by.

# Mechanism of Fat Reduction *via* Thermogenesis and Relationship With Protective NKT Cells

In the development of obesity, the inflammatory environment created by NKT cell activation leads to insulin resistance and impaired glucose tolerance, which further accelerates metabolic changes that promote weight gain through increased fat mass. Meanwhile, recent studies on the suppression of obesity have provided insight into how NKT cells prevent obesity other than by producing anti-inflammatory cytokines. Fat mass is actively reduced in brown (B)AT through thermogenensis (41). BAT contains thermogenic mitochondria that express uncoupling protein (UCP) 1 and contribute to energy expenditure, in contrast to WAT (42). UCP1-expressing adipocytes with thermogenic capacity known as beige or brite cells—also develop in WAT in response to various stimuli (43). The relationship between iNKT cells and thermogenesis was demonstrated by the finding that activated iNKT cells enhanced fibroblast growth factor 21 production and increased the number of beige cells in WAT, which in turn increased thermogenesis and weight loss (44). Several recent studies have demonstrated that innate immune cells play an important role in the induction of beige cells. vNKT cells and ILC2s induced by IL-25 produce IL-13 and regulate glucose homeostasis to protect against obesity (37). ILC2s also sustain eosinophils that produce IL-4, which stimulates M2 macrophages in VAT (45). IL-4 further stimulates M2 macrophages to secrete catecholamines for the induction of thermogenic gene expression in BAT and lipolysis in WAT (46). IL-33 is also critical for the maintenance of ILC2s in the induction of beige cells in WAT and regulation of energy expenditure. ILC2s produce methionine– enkephalin peptides that can act directly on adipocytes to upregulate UCP1 expression and promote beiging (47). These findings indicate that the innate immune system—including iNKT cells, macrophages, and ILCs—in AT controls thermogenesis by inducing beige cells, which is an important mechanism for the regulation of obesity and insulin resistance in addition to the control of AT inflammation *via* production of antiinflammatory cytokines.

# APC FOR NKT CELLS IN AT

Natural killer T cells in DIO act as NKT1 or NKT2 (or AT-resident NKT) cells through interactions with CD1d-expressing cells in AT. Many cell types in AT express CD1d including macrophages, dendritic cells, adipocytes, and possibly others. Recent studies have shown that adipocytes can activate T cells and NKT cells through antigen presentation (48, 49). CD1d expressed on the surface of adipocytes can induce helper T cell (Th)1 and Th2 cytokine release by iNKT cells depending on the co-expression of microsomal triglyceride transfer protein and CCAAT/enhancerbinding protein-β and -δ even in the absence of exogenous ligands (48), suggesting that adipocytes express ligands that are recognized by NKT cells. To determine whether interaction between NKT cells and adipocytes influence DIO, we analyzed mice with adipocyte-specific CD1d1 deletion (adipoqcre-CD1d1fl/fl) and found that they gained less weight than control mice fed an HFD (50), consistent with our findings from conventional CD1d KO mice (24). A decrease in IFN-γ and concomitant increase in adiponectin was observed following disruption of the NKT cell/ adipocyte interaction, which ameliorated AT inflammation and insulin resistance. On the contrary, another study showed that adipocyte-specific CD1d1 deletion reduced IL-4 expression in adipose iNKT cells and increased AT inflammation and insulin resistance (51), in accordance with an earlier report (49). The fact that these studies reported opposite results using the same conditional (c)KO mice provides strong evidence that adipocytes are the APCs for NKT cells in modulating AT inflammation, and that different HFDs can explain the discrepancy in the reported roles of NKT cells in the development of obesity (50, 51).

# CD1d2-Restricted NKT Cells

The fact that opposite results were obtained using the same cKO mice is critical, because it excludes the possibility that the results simply reflect the use of either pro-aggravating [CD1d1 KO; (52)] or pro-ameliorating [CD1d1/d2 KO; (53, 54)] mice (**Table 1**). Although it was reported that CD1d2 does not specify a specific NKT cell population (55), CD1d2 may affect the development of obesity in CD1d1 KO mice. Indeed, it was recently reported that CD1d2 can present distinct species of glycosylceramide (GlyCer) and affect the differentiation of NKT cells (56). Thus, the possible contribution of CD1d2-restricted NKT cells to the development of obesity remains to be determined, although contradictory results were obtained regarding DIO using the same cKO mice that express neither CD1d1 nor CD1d2 on adipocytes (50, 51).

In addition to studies of genetically engineered mice, other factors affecting the development of obesity have been investigated, including microbiota—especially those in the gut—and fat composition, both of which are related to diet and influence the presentation of ligands to NKT cells.

# OTHER FACTORS THAT AFFECT THE DEVELOPMENT OF DIO

#### Microbiota

The findings that gut microbiota composition is a critical factor in the development of obesity come from studies using germ-free (GF) animals. Conventionally raised mice have higher total body fat than GF mice, although the latter consume more food (57). When the two types of mice are fed a sugar-rich HFD, GF mice are protected from DIO owing to increased fatty acid (FA) oxidation and AMP-activated protein kinase activity (58). On the other hand, pathogenic alterations in gut microbiome profiles (i.e., dysbiosis) in obesity affect energy metabolism (59). In fact, the abundance of *Firmicutes* is increased whereas that of *Bacteroidetes* is decreased in *ob/ob* mice with a mutation in the gene encoding leptin; on the contrary, lean *ob/*+ mice fed a polysaccharide-rich diet predominantly harbor *Bacteroidetes* (60). Similar differences in gut microbiota composition are also observed between obese and lean human subjects (61). Furthermore, GF mice inoculated with microbiota from obese twin donors developed increased adiposity when compared with those receiving transplants from



*NKT cells regulate.*

*CDld but not NKT may regulate.*

*NKT cells promote.*

*NKT cells are neutral.*

lean twin donors and did not develop increased adiposity when they were cohoused with the latter mice (62). It is unclear whether microbiota or diet (calorie excess) is responsible for obesity.

Although gut microbiotas are transmissible and can be altered by diet, they may have the ability to directly alter systemic energy metabolism and thereby control weight gain. Several studies have demonstrated that NKT cells play a central role in maintaining homeostasis at mucosal surfaces (63, 64). CD1d KO mice exhibit altered gut microbiome profiles, which exacerbate intestinal inflammation induced by dextran sodium sulfate treatment and even in the steady state. Compared to non-littermate B6 mice, these mice have a higher abundance of segmented filamentous bacteria that can induce Th17 cells but reduced levels of *Akkermansia*, which may protect mice from developing colitis (65). A recent study using CD1dfl/flCD11cCre cKO mice also showed that CD1d expression on CD11c<sup>+</sup> cells contributes to the maintenance of intestinal homeostasis by regulation of the immunoglobulin A repertoire and induction of Tregs in the gut (66). *Bacteroides fragilis*, a prominent gut bacterial species, produces the glycosphingolipid α-GalCerBf, which is structurally related to the prototypic ligand α-GalCer or KRN7000 (67). α-GalCerBf stimulates iNKT cells in the context of CD1d, suggesting that alterations in the abundance of *B. fragilis* caused by obesity can affect NKT cell homeostasis. Disruption of the NKT cell/ CD1d interaction—which is required to maintain intestinal homeostasis—may affect energy consumption and fat storage by modulating gut microbiota composition.

#### Fat Composition

α-Galactosylceramide is a potent activator of NKT cells, and various analogs have been synthesized that elicit distinct cytokine responses (68). For instance, α-C-GalCer and RCAI-56 promote Th1-biased responses (69, 70), whereas OCH and 20:2 promote Th2-biased responses (71). Natural ligands for CD1d have also been identified. Several mammalian glycosphingolipids such as isoglobotrihexosylceramide and β-glucosylceramide (β-GlcCer) were shown to act as self-lipid antigens (72). However, a recent study showed that a small quantity of stimulatory α-GlyCer was present in β-GlcCer preparations (73). Accordingly, pure β-GlcCer may not activate iNKT cells, which can respond to a minor fraction of α-GlyCer. Phospholipids (PHLs) that are a major component of mammalian cell membranes including phosphatidylinositol (PI), phosphatidylethanolamine (PE), and phosphatidylglycerol (PG) are natural antigens recognized by NKT cells (74). Lysophosphatidylcholine is also a natural ligand that is recognized not only by human iNKT cell clones (75) but also by Vα24<sup>−</sup>/Vβ11<sup>−</sup> vNKT cells (76). Characteristic lipid abnormalities observed during the course of obesity include an increase in triacylglycerol and cholesterol levels in the low-density lipoprotein fraction, with a corresponding decrease in high-density lipoprotein cholesterol. In addition, obesity-related changes of serum lipids such as FAs, PHLs, and their oxidation products as well as oxylipins, sphingolipids, and their metabolites contribute to the health status and risk of comorbidities in obese patients (77). A lipidomics analysis demonstrated that changes in PHL concentrations may contribute to the development of insulin resistance and metabolic syndrome (78). Elevated circulating levels of phosphatidylcholine, PI, PE, and PG have been detected in subjects with when compared with those without non-alcoholic steatohepatitis (79). Although the molecular basis for the correlation between NKT cell activation and altered PHL levels in obese subjects remains unclear, some PHLs may affect NKT cell biology, based on the observations that the concentration of Cer species (C18:0, C20:0, and C24:1) and total Cer level was higher in type 2 diabetes; insulin sensitivity was inversely correlated with C18:0, C20:0, C24:1, C24:0, and total Cer levels; and increased tumor necrosis factor (TNF)-α concentration was correlated with the levels of C18:1 and C18:0 ceramide (80).

The mechanism of insulin resistance in obese patients with an elevated Cer concentration may involve inflammation induced

by NKT cell activation, since certain Cer species stimulate NKT cells. FAs are the major components of fat and mediate immune responses. In AT, free FAs secreted by adipocytes especially saturated FAs (SAFAs) such as palmitate and laurate—activate macrophages *via* toll-like receptor 4 to induce TNF-α expression, whereas polyunsaturated (PU)FAs such as linolenate and eicosapentaenoic acid do not have this effect (81). SAFAs, but not PUFAs, stimulate the expression of inflammatory cytokines such as IL-6 and TNF-α in adipocytes (82) that further promote metabolic syndrome. The composition and concentration of FA in sera that are altered and elevated in obese subjects are determined based on endogenous synthesis rates and dietary fat characteristics (83, 84). Thus, dietary fats may affect AT inflammation by modulating the functions of immune cells and adipocytes, suggesting that HFDs with different compositions of FA species presumably affect NKT cell response to either promote or suppress AT inflammation and obesity.

#### CONCLUSION

Obesity-associated inflammation in AT contributes to metabolic syndrome and is controlled by adipocytes and NKT cells with other immune cells, as discussed in this review. NKT cells appear to respond to lipid antigens on adipocytes and modulate inflammation (either by ameliorating or by aggravating this process) depending on the input—i.e., dietary lipids and ligands derived from the microbiome (**Figure 1**). Although the critical factors that give rise to the distinct outcomes of NKT cells remain elusive, future investigations should focus on two mutually interactive topics: (1) gut microorganisms that regulate energy consumption and modulation/maintenance by NKT cells and (2) diet/fat composition that can alter gut microbiota, the balance of lipid species, and the synthesis of endogenous lipid antigens that affect NKT cell activation. Furthermore, elucidating the mechanism of BAT maintenance and WAT beiging by NKT cells can provide a basis for the development of strategies to reverse metabolic dysregulation and reduce fat mass.

# AUTHOR CONTRIBUTIONS

MS wrote the first draft. MS and KI edited the manuscript.

#### FUNDING

This study was supported by the Japan Society for the Promotion of Science (Grant-in-Aid for Young Scientists [B] [no. 16K21347 to MS] and Grant-in-Aid for Scientific Research [C] [no. 17K08791 to KI and MS]); by an A-MED grant (no. 16ek0109076s0202 to KI); and by MEXT (Grant-in-Aid for Research Branding Project; Agromedicine at Kitasato University to KI and MS).

# REFERENCES


colonization in the intestine of mice. *J Clin Invest* (2009) 119:1241–50. doi:10.1172/JCI36509


macrophages. *Arterioscler Thromb Vasc Biol* (2007) 27:84–91. doi:10.1161/01. ATV.0000251608.09329.9a


**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 Satoh and Iwabuchi. 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.*

# Mucosal-Associated invariant T Cells in Autoimmune Diseases

*Asako Chiba\*, Goh Murayama and Sachiko Miyake\**

*Department of Immunology, Juntendo University School of Medicine, Tokyo, Japan*

Mucosal-associated invariant T (MAIT) cells are innate T cells restricted by MHCrelated molecule 1 (MR1). MAIT cells express semi-invariant T-cell receptors TRAV1- 2-TRAJ33/12/20 in humans and TRAV1-TRAJ33 in mice. MAIT cells recognize vitamin B2 biosynthesis derivatives presented by MR1. Similar to other innate lymphocytes, MAIT cells are also activated by cytokines in the absence of exogenous antigens. MAIT cells have the capacity to produce cytokines, such as IFNγ, TNFα, and IL-17, and cytotoxic proteins, including perforin and granzyme B. MAIT cells were originally named after their preferential location in the mucosal tissue of the gut, but they are also abundant in other peripheral organs, including the liver and lungs. In humans, the frequency of MAIT cells is high in peripheral blood, and these cells constitute approximately 5% of circulating CD3+ cells. Their abundance in tissues and rapid activation following stimulation have led to great interest in their function in various types of immune diseases. In this review, first, we will briefly introduce key information of MAIT cell biology required for better understating their roles in immune responses, and then describe how MAIT cells are associated with autoimmune and other immune diseases in humans. Moreover, we will discuss their functions based on information from animal models of autoimmune and immunological diseases.

Keywords: mucosal-associated invariant T cells, multiple sclerosis, systemic lupus erythematosus, inflammatory arthritis, inflammatory bowel diseases, diabetes, asthma

# INTRODUCTION

Two subsets of T cells express semi-invariant T-cell receptors (TCRs). The first subset includes the thoroughly studied invariant natural killer T (iNKT) cells that uniquely recognize lipid antigens presented by CD1d, a homolog of the MHC molecule. TCRα rearrangement in iNKT cells includes Vα24–Jα18 (TRAV10-TRAJ18) in humans and Vα14–Jα18 (TRAV 11–TRAJ 18) in mice. The second subset, mucosal-associated invariant T (MAIT) cells, are restricted by the MHC-related protein 1 (MR1) and express Vα7.2–Jα33 (TRAV1-2–TRAJ33) in humans and Vα19–Jα33 (TRAV1–TRAJ33) in mice (1). Vα7.2–Jα33 rearrangement was discovered by Porcelli et al. along with Vα24–Jα28 during analysis of the TCR repertoire of human CD4<sup>−</sup>CD8<sup>−</sup> (double-negative; DN) T cells (2). Later, Tilloy et al. discovered homologous Vα19–Jα33 in mice (1). MAIT cells were originally named after their preferential location in the gut lamina propria. Their absence in germ-free mice also indicated their association with mucosal immunity (3). In 2009, Martin et al. generated a monoclonal antibody against human Vα7.2 TCR and demonstrated that Vα7.2TCR<sup>+</sup> cells with high expression of CD161 were MAIT cells (4). Human MAIT cells are abundant in peripheral blood and constitute up to 10% of blood CD3<sup>+</sup> cells. Because the frequency of iNKT cells in human peripheral blood is 0.01–1%, MAIT cells are 10- to 1,000-fold more frequent than iNKT cells.

#### *Edited by:*

*Kazuya Iwabuchi, Kitasato University School of Medicine, Japan*

#### *Reviewed by:*

*Paula M. Oliver, University of Pennsylvania, United States Agnes Lehuen, Institut National de la Santé et de la Recherche Médicale (INSERM), France*

#### *\*Correspondence:*

*Asako Chiba a-chiba@juntendo.ac.jp; Sachiko Miyake s-miyake@juntendo.ac.jp*

#### *Specialty section:*

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

*Received: 30 March 2018 Accepted: 29 May 2018 Published: 11 June 2018*

#### *Citation:*

*Chiba A, Murayama G and Miyake S (2018) Mucosal-Associated Invariant T Cells in Autoimmune Diseases. Front. Immunol. 9:1333. doi: 10.3389/fimmu.2018.01333*

**340**

Recent studies using MR1 tetramers revealed that not all of the TCR usage of human MAIT cells is restricted to TRAV1-2– TRAJ33 (5). Approximately, 30% of MR1-restricted TRAV1-2<sup>+</sup> cells use TRAV1-2 joined with TRAJ20 or TRAJ12 gene segments (5). Moreover, subsets of MR1-resticted T cells do not express TRAV1-2, and their features are discussed elsewhere (6). TRAV1-2–TRAJ33 are mostly paired with TRBV6-6 and TRBV20 (1, 5, 7). In mice, only TRAV1–TRAJ33 (Vα19–Jα33) has been reported as a murine MAIT TCR paired with TRBV13-3 (Vβ8.1), TRBV 13-2 (Vβ8.2), and TRBV19 (Vβ6) (1, 5, 8). The usage of different MAIT TCRs might be related to the tissue distribution of MAIT cells. Vα7.2–Jα33 is the dominant MAIT TCR Vα transcript in human peripheral blood, but the percentages of Vα7.2–Jα12 transcripts are higher than those of Vα7.2–Jα33 transcripts in kidney and intestine biopsies from some individuals (7).

#### MAIT CELL PHENOTYPE

Mucosal-associated invariant T cells in adult blood exhibit the effector memory phenotype (CD95hiCD62LloCD45RO+CD45RAlo CD27<sup>+</sup> CD122<sup>+</sup>) (4, 9). In the thymus and cord blood, MAIT cells display a naïve phenotype and are present at very low numbers (4). However, these MAIT cells already express CD161 in the thymus and CD161 and IL-18Rα in cord blood (10) and produce TNFα in response to *Mycobacterium tuberculosis*-infected cells (11). MAIT cells also express PLZF, a master regulator of innatelike T cells (8, 12). PLZF expression appears important in the development of MAIT cells because PLZF deficient mice have a significantly lower frequency of MAIT cells. Approximately 95% of human MAIT cells are DN or CD8<sup>+</sup>. Most MHC-restricted CD8<sup>+</sup>T cells express the CD8αβ heterodimer, but CD8<sup>+</sup> MAIT cells express CD8αα homodimers, and some of them coexpress the CD8αβ heterodimer (4, 5, 13). Approximately 60% of MAIT cells were CD4<sup>−</sup>CD8<sup>−</sup> in most tissues of C57BL6/J mice except for lymph nodes where 40% of MAIT cells were CD4<sup>+</sup>. CD8<sup>+</sup> MAIT cells were more frequent in Balb/c mice than in C57Bl/6J mice, but CD4<sup>+</sup>MAIT cells were also enriched and constituted half of MAIT cells in lymph nodes (8).

Human peripheral blood MAIT cells are CCR5<sup>+</sup>CCR6<sup>+</sup>CC R7<sup>−</sup>CCR9<sup>+</sup>/<sup>−</sup> CXCR3<sup>−</sup>CXCR4<sup>+</sup>/<sup>−</sup> CXCR6<sup>+</sup> (10, 14, 15). Mouse MAIT cells are CCR6<sup>+</sup>/<sup>−</sup> CCR9<sup>+</sup>/− CXCR6hi but negative for CCR4, CCR7, CXCR1, CXCR3, and CXCR4 (8). Lack of CCR7 and CD62L expression indicates their poor ability to migrate into lymph nodes *via* high endothelial venules, and expression of CCR9 and CXCR6 suggests their ability to migrate into the intestine and the liver. In fact, human MAIT cells are abundant in peripheral blood and enriched in tissues such as the liver (20–50% of CD3<sup>+</sup> cells), intestine (1–10% of CD3<sup>+</sup> cells), and lung (2–4% of CD3<sup>+</sup> cells) (5, 10, 16–21). Human MAIT cells are also detected in other tissues, including female genital mucosa, kidney, prostate, and ovary (7, 22). FTY720, an agonist of sphingosine-1-phosphate receptors, inhibits the egress of naïve and central memory T and B cells from lymph nodes. FTY720 has been used for treatment of patients with multiple sclerosis (MS). FTY720 treatment decreased the total lymphocyte count but increased MAIT cell frequency; it also reduced DN cells and increased CD8hi and CD4+cells among MAIT cells (23). This finding indicates that MAIT cells are indeed rare in lymph nodes, and tissue distribution may differ among subsets of MAIT cells. Activated MAIT cells may obtain more migrating capacity because IL-18 stimulated MAIT cells express very late antigen-4 (VLA-4), an integrin important for migration into the site of inflammation (24). No antibody against murine Vα19TCR is available, and the frequency of MAIT cells in mice was unknown until the recent development of MR1 tetramers (8). Compared with iNKT cells, MAIT cells are relatively rare in laboratory strains of mice except for CAST/EiJ mice (1, 3, 25). The average frequency of MAIT cells among C57BL/6 mouse lymphocytes is 3.3, 0.7, 0.6, 0.2, 0.08, and 0.05% in the lung, lamina propria, liver, lymph nodes, spleen, and thymus, respectively (8).

#### MAIT CELL ACTIVATION MECHANISMS

Early studies demonstrated that MAIT cells are deficient in germ-free mice and activated by antigen-presenting cells in the presence of bacteria in an MR1-dependent manner (3, 26, 27). These findings suggested that MAIT cells might recognize microbial antigens presented by the MR1 molecule. Microbes that activated MAIT cells included various types of bacterial species and yeast. In 2012, Kjer-Nielsen et al. described several MR1-restricted antigens. They identified 6-formylpterin (6-FP), a photodegradation product of folic acid (vitamin B9), as an MR1 ligand. 6-FP upregulated surface expression of MR1 but failed to activate MAIT cells. The researchers found that reduced 6-hydroxymethyl-8-d-ribityllumazine (rRL-6-CH2OH) derived from the bacterial riboflavin (vitamin B2) biosynthetic pathway is a MAIT cell-activating MR1 ligand (28). Later, Corbett et al. revealed that some potent MR1 ligands, including 5- (2-oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU), are produced by an interaction between early intermediates in the bacterial riboflavin synthesis pathway and either glyoxal or methylglyoxal, and these antigens are unstable unless they are captured and stabilized by the MR1 molecule (29). More recently, several MR1 ligands have been reported among drugs and drug metabolites, such as diclofenac and methotrexate (30). A photodegraded product of aminopterin or methotrexate captured by the MR1 molecule inhibited MAIT cell activation by 5-OP-RU, whereas diclofenac and its metabolites stimulated MAIT cells.

Similar to iNKT cells, MAIT cells are activated by cytokines in an MR1-independent manner (**Figure 1**). MR1 expression is indispensable for the development of MAIT cells but not for the effector functions of these cells. Our group demonstrated that MAIT cells exacerbated joint inflammation in arthritis models, and MAIT cells exerted their effector function even when they were adoptively transferred into MR1-deficient mice (31). A MAIT cell-enriched population from V19iTCR transgenic (Vα19iTg) mice produced IL-17 after exposure to IL-23 and proliferated upon IL-1β stimulation (31). Inhibition of bacterial growth of *Mycobacterium* by MAIT cells was more dependent on IL-12-mediated activation of these cells rather than on MR1 antigen recognition by MAIT cells (32). Human MAIT cells express high levels of IL-18Rα and are activated to produce IFNγ

by IL-12 plus IL-18 (33–37). MAIT cells are also activated by type I IFN (33, 34). The kinetics of MAIT cell activation upon different types of stimuli might differ as activation of MAIT cells at early time points after incubation with *E. coli* was MR1-dependent, and IL-12 + IL-18-mediated activation took more time (35). MAIT cells are activated by TCR signals (anti-CD3/CD28) when they are stimulated in the presence of other peripheral blood mononuclear cells, but sorted MAIT cells (CD4<sup>−</sup>CD8<sup>+</sup>CD56<sup>+</sup> CD16<sup>−</sup>CD161hiVα7.2<sup>+</sup> cells) did not respond to TCR signals. However, sorted MAIT cells produced IFNγ and granzyme B when they were activated with TCR signals in the presence of pro-inflammatory signals provided by monocytes activated by TLR agonists. Sorted MAIT cells also produced cytokines, such as IFNγ and TNFα, in response to IL-12/15/18 stimulation (37). Because MAIT cells are enriched in mucosal tissue where MR1 antigens produced by commensal bacteria are present, they may be programed to respond only when they are exposed to such antigens together with inflammatory signals to avoid unwanted tissue inflammation.

# MAIT CELL FUNCTION

Upon stimulation with phorbol 12-myristate13-acetate (PMA) and ionomycin or anti-CD3 and anti-CD28, human MAIT cells produce IFNγ, TNFα, IL-17, IL-2, and granzyme B (10, 14, 26). Mouse spleen MAIT cells produce high levels of IL-17 and MIP-1α and low levels of IL-10, IFNγ, and TNFα following stimulation with PMA and ionomycin or anti-CD3 and anti-CD28 (8). This bias toward IL-17 production was also observed in MAIT cells in various types of organs, such as the thymus and lung (8). Unsurprisingly, mouse MAIT cells express high levels of retinoic acid-related orphan receptor (RORγτ) and low levels of T-bet (8). Human MAIT cells also express Rorc and T-bet; as expected, the expression of T-bet is higher than that of Rorc (10, 38). MAIT cells in different tissues may vary in their cytokine-producing capacity. Mouse thymus MAIT cells also produce other cytokines, such as GM-CSF, IL-4, and IL-13 (8). Mouse MAIT cells from iVα19– Vβ6 transgenic mouse spleen produced IL-2 after stimulation by *E. coli*-infected dendritic cells (26). In humans, adipose tissue MAIT cells but not peripheral blood MAIT cells produce more IL-10 than IL-17. Interestingly, MAIT cells in adipose tissue from obese individuals produced more IL-17 and less IL-10 (39).

# FACTORS AFFECTING MAIT CELL FREQUENCY AND FUNCTION

Mucosal-associated invariant T cell numbers are very low in peripheral blood at birth, and their frequency increases with age up to 40–50 years of age (9, 10, 40). Novak et al. studied MAIT cells from individuals at different ages and demonstrated that the frequency of MAIT cells is highest in women of fertile age and significantly declines in elderly individuals (40). MAIT cells are decreased in patients with type 2 diabetes (T2D) and/or obesity (39, 41, 42). Upon stimulation with PMA and ionomycin, MAIT cells from T2D patients produced higher levels of IL-17, and MAIT cells from obese T2D patients produced even higher levels of other cytokines, including IL-2, granzyme B, and IFNγ. In obese patients, MAIT cell frequency was higher in omental adipose tissue than in peripheral blood; moreover, MAIT cell frequency was increased, and cytokine-producing ability was decreased after bariatric surgery. Smoking appears to reduce the frequency of peripheral blood MAIT cells (43). Circulating MAIT cells are affected by not only systemic but also inhaled administration of corticosteroids (43). Some drugs and drug metabolites have been reported as MR1 ligands (30). MR1 ligands derived from microorganisms should be abundantly present in the gut. Although whether the frequency and function of MAIT cells are affected by drugs or indigenous microbes remains unknown, taking the possible influence of these factors into account when studying human MAIT cells might be important.

## MAIT CELLS IN AUTOIMMUNE AND IMMUNOLOGICAL DISEASES

The role of MAIT cells in immunological disorders was largely unknown until our group described their protective role against experimental autoimmune myelitis (EAE), an animal model of MS, and their pro-inflammatory roles in arthritis models. Since anti-Vα7.2 TCR monoclonal antibody has become available, many groups, including ours, have conducted studies on MAIT cells in autoimmune and immunological diseases. MAIT cells appear to be involved in various types of diseases (**Figure 2**), but their contribution to each pathology is currently unknown. This has been mostly due to the lack of good tools to study murine MAIT cells. MAIT cells are usually rare in mice and there is no monoclonal antibody against murine MAIT cell TCR. New technical approaches such as MR1 tetramers and CAST/EiJ mice may overcome these issues. Here, we review MAIT cells in human immunological diseases and the corresponding animal models. In human studies, MAIT cells are identified by the Vα7.2 TCR and high CD161 expression unless otherwise specified.

#### Multiple Sclerosis

Multiple sclerosis is an inflammatory demyelinating disease affecting the central nervous system (CNS). Although the etiology of MS is not fully understood, MS is considered an autoimmune disease against the myelin component of the CNS. Illes et al. investigated invariant TCR expression in the CNS lesions of patients with MS by using the single-strand

conformation polymorphism clonotype method (44). They found very low Vα24–Jα18 expression in MS CNS samples but observed Vα7.2–Jα33 expression in half of MS CNS samples and in most cerebral spinal fluid (CSF) samples. Later, the presence of Vα7.2TCR<sup>+</sup>CD161<sup>+</sup> in MS lesions was confirmed by other groups (24, 45, 46). CD8<sup>+</sup>MAIT cells are present in MS brain lesions, and approximately 5% of CD8<sup>+</sup>T cells were Vα7.2TCR<sup>+</sup>CD161<sup>+</sup> cells in acute and chronic active MS lesions, suggesting infiltration of MAIT cells into MS lesions (24). These studies were performed by using MS autopsies, and a more recent study demonstrated Vα7.2–Jα33 transcripts in the brain lesions of a MS patient with newly onset disease (47). There are several conflicting reports regarding the frequency of circulating MAIT cells in MS patients. Most reports demonstrate the reduction of all or subsets of MAIT cells in MS except for one report showing an increase in CD161hi CD8<sup>+</sup>T cells, and most of these cells are usually MAIT cells (48). Two reports demonstrated the reduction of V7.2TCR<sup>+</sup> CD161 high cells or CD161high memory CD8<sup>+</sup>T cells in MS (14, 24). Other groups showed that the frequency of MAIT cells was comparable between MS patients and healthy volunteers (23, 46), but MAIT cells were reduced in patients with progressive disease (46), and CD8hi cells among MAIT cells were decreased in MS (23). IFNβ treatment did not affect the frequency of MAIT cells in MS patients (23), but the frequency of MAIT cells in patients with relapse was increased along with clinical recovery after steroid treatment. Several findings suggest migration of MAIT cells into the CNS. *In vitro*, IL-18 increased surface expression of VLA-4 on CD8<sup>+</sup> MAIT cells, and the frequency of CD8<sup>+</sup>MAIT cells was inversely correlated with the serum level of IL-18 in MS patients but not in healthy individuals (24). MS MAIT cells overexpress P-selectin glycoprotein ligand-1 (PSGL-1) and CD11a (part of the lymphocyte function-associated antigen 1), which are important for cell rolling and homing though the blood–brain barrier (46).

Therefore, what role do MAIT cells play in the pathogenesis of MS? IFNγ and TNFα production by MAIT cells was decreased in untreated MS patients, but FTY720 treatment recovered the cytokine-producing capacity of MAIT cells (23). Successful treatment of MS with autologous hematopoietic stem cell transplantation was accompanied by depletion of CD8<sup>+</sup>MAIT cells, whereas regulatory T cells and CD56high natural killer cells were increased in the peripheral blood of MS patients (45). These findings indicated a pro-inflammatory role for MAIT cells in MS, whereas MAIT cells played a protective role in EAE, an animal model of MS (49). The disease development and progression were suppressed in Vα19iTg mice, and MR1-deficient mice developed more severe EAE than did control mice. Adoptive transfer of T cells enriched with a MAIT cell population protected wildtype mice from EAE. Inhibition of EAE in Vα19iTg mice was associated with decreased Th1 and Th17 responses against myelin oligodendrocyte glycoprotein and increased secretion of IL-10. Cytokines produced by Vα19iT cells are different from those produced by human MAIT cells. IL-10 is heavily produced by Vα19iT cells, but human MAIT cells produce very little IL-10. Whether MAIT cells play a protective role in the development of human MS is unknown. As depletion of MAIT cells (CD5<sup>+</sup>CD19<sup>−</sup> TCRγδ− CD161highVα7.2TCR<sup>+</sup> cells) increased IFNγ production by T cells *in vitro*, human MAIT cells may also have suppressive effect on other T cells (14).

#### Systemic Lupus Erythematosus (SLE)

Systemic lupus erythematosus is a systemic autoimmune disease that affects various types of organs, including the skin, kidneys, and CNS. The most characteristic features of SLE are the production of autoantibodies targeting nucleic acids and immune activation by the generation of nucleic acid-containing immune complexes. Thus, the impaired tolerance of T and B cells has been considered one of the major causes of the disease. However, abnormalities in function and number have been reported in innate lymphocytes including natural killer cells and iNKT cells in patients with SLE (50–56). The frequency of MAIT cells was also reduced in the peripheral blood of SLE patients, and the reduction of these cells was more profound than that of γδT cells and iNKT cells (33). We confirmed that the reduction of MAIT cells in SLE was not a result of downregulation of surface markers by single-cell PCR for the expression of Vα7.2–Jα33 TCR. Moreover, the reduction of MAIT cells was not due to the use of corticosteroids in SLE. Lupus MAIT cells were less responsive to stimuli and prone to death, and there were more apoptotic cells among circulating MAIT cells in SLE. Cho et al. also reported the decrease of circulating MAIT cell in SLE patients (57). They showed impaired IFNγ production by lupus MAIT cells that was accompanied by elevated PD-1 expression and an intrinsic defect in the Ca2<sup>+</sup>/ calcineurin/NFAT1 signaling pathway of these cells. In our study, MAIT cells from SLE patients with active disease expressed high levels of CD69, and the activated status of MAIT cells positively correlated with disease activity. Thus, MAIT cells in SLE patients appear to be activated and lost due to activation-induced cell death *in vivo*; moreover, the remaining MAIT cells are less responsive to stimuli. We elucidated two possible mechanisms of MAIT cell activation. First, monocytes from SLE patients exerted higher MR1 antigen-presenting capacity to MAIT cells. Second, elevated IFNα appeared to be associated with activation of MAIT cells in SLE. Overexpression of type I IFNs and IFNinducible genes has been reported in SLE patients, and type I IFN is thought to play a central role in the pathogenesis of lupus (58). CD69 expression on MAIT cells positively correlated with serum levels of IL-18 and IFNα in SLE, and exposure to IFNα-induced MAIT cell activation *in vitro*, suggesting that these cytokines may also contribute to the activation of MAIT cells in SLE. MAIT cells migrate into inflamed tissues including kidneys (59). MAIT cells constitutively express chemokine receptors, and exposure to IL-18 upregulates the surface expression of VLA-4 (24), which mediates T-cell migration through an interaction with vascular cell adhesion molecule (VCAM-1). Urinary levels of IL-18 and VCAM-1 were increased and associated with nephritis activity in SLE (60, 61). DN T cells infiltrated the kidneys in lupus, and the majority of these cells were neither γδT cells nor iNKT cells (62). Thus, MAIT cells may migrate into inflamed tissues in SLE.

#### Inflammatory Arthritis

Rheumatoid arthritis (RA) is the most common inflammatory arthritis that typically affects the small joints of the hands and feet, and the synovium is the primary site of inflammation. RA is characterized by production of rheumatoid factor (RF) and anti-citrullinated protein antibody. Spondyloarthritis (SpA) is a group of disorders comprising ankylosing spondylitis (AS), psoriatic arthritis, reactive arthritis, arthropathy of inflammatory bowel disease (IBD), and undifferentiated SpA. The features of SpA include the absence of RF and association with *HLA-B27*; the main targets are the enthesis and axial skeleton. Neutralizing antibodies against the TNFα and IL-6 signaling pathways are widely used in RA treatment; in addition to TNF inhibitors, blocking the IL-23/IL-17 axis is beneficial in AS. Circulating MAIT cells are reduced in patients with RA and SpA including AS (63–65). MAIT cells displayed enhanced IL-17-producing capacity and activated status of these cells correlated with disease activity in AS (63, 64). Cell death of circulating MAIT cells was not enhanced, but they were accumulated in the synovial fluid (SF) in AS. SF MAIT cells displayed high levels of CD69 and enhanced producing capacity of IL-17 and granzyme B in AS. Additionally, SF MAIT cells are enriched in RA (57). Interestingly, IL-17 production by SF MAIT cells was higher in AS than in RA, but TNFα- and IFNγ-producing SF MAIT cells in AS were comparable to those in RA (64). IL-7R polymorphisms are associated with AS, and IL-7 primes MAIT cells (20, 66). Gracey et al. demonstrated that IL-7R expression is increased on AS MAIT cells, and exposure to IL-7 exacerbated the IL-17-producing capacity of AS MAIT cells. Considering their capacity to produce inflammatory cytokines at the site of tissue inflammation, MAIT cells appear to contribute to tissue inflammation in arthritis. In animal models of inflammatory arthritis, MAIT cells enhanced arthritic inflammation (31). DBA1J mice immunized with type II collagen (CII) develop collagen-induced arthritis (CIA), and MR1 deficiency attenuated the disease severity of CIA. Because MR1 deficiency had little effect on T and B cell responses against CII, MAIT cells appeared to contribute to the effector phase of arthritis. In fact, MAIT cell deficiency reduced the disease severity of collagen antibodyinduced arthritis (CAIA), and adoptive transfer of a T-cell population enriched with iVα19 TCR<sup>+</sup> cells from iVα19 TCR T g mice enhanced CAIA in MR1-deficient mice. This iVα19 TCR<sup>+</sup> cell population was activated by IL-1β or IL-23 in the absence of exogenous antigens. Therefore, MAIT cell activation in the CAIA model may be mediated by cytokines and does not require TCR stimulation.

#### Inflammatory Bowel Diseases

Inflammatory bowel diseases are chronic relapsing disorders of the gastrointestinal tract, comprising Crohn's disease (CD) and ulcerative colitis (UC). CD affects the distal ileum and colon, and UC involves only the colon. Inflammation in UC is superficial and includes the mucosa and submucosa, whereas CD involves transmural inflammation. The etiology of IBD is not fully understood, but the clinical efficacy of neutralizing antibodies specific for TNFα indicates the role of cytokine-producing immune cells in IBD (67). Both innate and adaptive immune systems appear to contribute to the pathogenesis of IBD. CD is thought to be mediated by Th1 and Th17 cell responses against gut commensal microbiota. UC is believed to be mediated by Th2 responses; however, anti-IL-13 therapy was not beneficial, and cytokines involved in the pathogenesis of UC appear to be more complicated (68, 69). MAIT cells are reduced in the peripheral blood of patients with CD and UC (18, 70–72). However, IL-17 production by MAIT cells was increased in UC patients (71). CD69 expression on MAIT cells was associated with disease activity. Increased IL-17 production by MAIT cells and correlation of CD69 expression on these cells with disease activity indicated the association of MAIT cells with the pathogenesis of UC. T cells expressing CD161, IL-23R and RORγt are enriched in intestinal mucosa from patients with IBD; thus, MAIT cells may be associated with tissue inflammation in IBD (67, 73, 74). In fact, MAIT cells accumulated in the inflamed mucosa of patients with CD and UC (18, 71, 72). Plasma IL-18 levels were positively correlated with CD69 expression on MAIT cells in UC. Thus, the reduction of circulating MAIT cells may be a result of the recruitment of these cells to the inflamed tissue. A report by Hiejima et al. showed reduced MAIT cell frequency in intestinal mucosa from IBD patients (70). They demonstrated enhanced cell death of MAIT cells in peripheral blood in IBD patients and inflamed mucosa of those with CD. Further studies are required to understand the differences among reports and their function in IBD pathology. One report demonstrated that adoptive transfer of Jα33+ cells into mice reduced the severity of intestine inflammation in 2,4,6-trinitrobenzene sulfonic acid (TNBS) colitis, suggesting their protective role in an animal model of CD (75).

# Type 1 Diabetes (T1D)

Harms et al. investigated CD161bright CD8<sup>+</sup> T cells in patients with juvenile T1D whose clinical onset was within 12 months and demonstrated there was no reduction of these cells in the peripheral blood of patients (76). The frequency of CD27<sup>−</sup>MAIT cells was increased in patients, and these cells displayed more enhanced IL-17-producing capacity. CD161bright CD8<sup>+</sup>T cells increased with age in control individuals but not in juvenile T1D patients, suggesting that circulating MAIT cells may be decreased in patients with long-standing T1D. More recently, a lower frequency of MAIT cells was detected in children with recent onset T1D than in control children (77). MAIT cells from T1D patients expressed increased levels of CD25 and PD-1 and displayed enhanced cytokine production, including TNFα and granzyme B. Because human MAIT cells exerted cytotoxic activity against a pancreatic β-cell line, MAIT cells may contribute to β-cell destruction in T1D.

In non-obese diabetic (NOD) mice, the frequency and number of MAIT cells was lower in the spleen and pancreatic lymph nodes of NOD mice than in those of C57BL/6 mice. MAIT cells were present in pancreatic islets and enriched in the ileum of NOD mice, and these MAIT cells exhibited a more activated phenotype. The recruitment of MAIT cells into the ileum decreased and that into the pancreas increased with aging. IFNγ and granzyme B production by islet MAIT cells was already observed in prediabetic mice and further increased in diabetic mice. These findings indicate that MAIT cells are recruited to the pancreas and contribute to tissue inflammation. However, MR1 deficiency increased the rates of diabetes in NOD mice and the streptozotocin-induced T1D model (77). MR1 deficiency increased intestinal permeability, and this issue was associated with increased infiltration of lymphoid cells into the lamina propria and more bacterial translocation from the gut to pancreatic lymph nodes. Thus, MAIT cells appear to be important for the maintenance of tissue integrity, but they contribute to tissue damage once inflammation occurs.

#### Asthma

Asthma encompasses chronic airway inflammation characterized by increased airway hypersensitivity to various types of antigen-specific and non-specific stimuli. Th2 cytokines, such as IL-4, IL-5, and IL-13, play important roles in activating other cells including eosinophils. MAIT cell frequency was reduced in patients with asthma in peripheral blood, sputum, and endobronchial biopsy specimens (78, 79). The reduction in MAIT cell frequency was associated with disease severity, inhaled corticosteroid dose, respiratory function, and disease duration. Moreover, CD69<sup>+</sup>MAIT cells were associated with respiratory function (80). MAIT cells are present in the lung at similar or higher frequencies than those in peripheral blood and are enriched in the lung under inflammatory conditions, including infection (6, 81). Thus, MAIT cells may be involved in the pathogenesis of asthma, but it is difficult to determine the role MAIT cells play in asthma because human MAIT cells mostly produce Th1 and Th17 type cytokines. Therefore, studies of MAIT cells at the site of inflammation or using animal models are required to understand their role in asthma.

# CONCLUDING REMARKS

Mucosal-associated invariant T cells appear to be involved in various types of immune disorders, and circulating MAIT cells were reduced in most diseases. We speculate that these findings are due to their unique characteristics. MAIT cells are very sensitive to stimuli, can be activated by antigens and cytokines, and have the capacity to migrate to inflamed tissues. There are several conflicting findings regarding MAIT cell frequencies in some diseases among different research groups. This discrepancy may be due to the methodology used to identify MAIT cells. Other potential reasons include the influence of factors, such as age, gender, obesity, and smoking, on MAIT cells (43). MR1 ligands derived from commensal microbes or drugs could modify MAIT cell function or frequency. Thus, these factors might have influenced these different findings. Studies using new tools such as MR1 tetramers and MR1 ligands may answer these questions and determine the potential of MAIT cells as a therapeutic target in immune diseases.

# AUTHOR CONTRIBUTIONS

AC, GM, and SM wrote the manuscript. Both AC and SM contributed equally to this work. GM drew schematics.

# FUNDING

This study was supported by the Japan Society for the Promotion of Science [Grant-in-Aid for Scientific Research (C) 17K09983 (to AC) and Grant-in-Aid for Scientific Research (B) 17H04218 (to SM)].

# REFERENCES


bowel diseases. *Clin Exp Immunol* (2014) 176(2):266–74. doi:10.1111/cei. 12277


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

*Copyright © 2018 Chiba, Murayama and Miyake. 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.*

# Functions of CD1d-Restricted invariant Natural Killer T Cells in Antimicrobial immunity and Potential Applications for infection Control

*Yuki Kinjo\*, Shogo Takatsuka, Naoki Kitano, Shun Kawakubo, Masahiro Abe, Keigo Ueno and Yoshitsugu Miyazaki*

CD1d-restricted invariant natural killer T (*i*NKT) cells are innate-type lymphocytes that

*Department of Chemotherapy and Mycoses, National Institute of Infectious Diseases, Tokyo, Japan*

#### *Edited by:*

*Kazuya Iwabuchi, Kitasato University School of Medicine, Japan*

#### *Reviewed by:*

*Moriya Tsuji, Aaron Diamond AIDS Research Center, United States Laurent Brossay, Brown University, United States*

*\*Correspondence: Yuki Kinjo* 

*ykinjo@niid.go.jp*

#### *Specialty section:*

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

*Received: 02 April 2018 Accepted: 22 May 2018 Published: 06 June 2018*

#### *Citation:*

*Kinjo Y, Takatsuka S, Kitano N, Kawakubo S, Abe M, Ueno K and Miyazaki Y (2018) Functions of CD1d-Restricted Invariant Natural Killer T Cells in Antimicrobial Immunity and Potential Applications for Infection Control. Front. Immunol. 9:1266. doi: 10.3389/fimmu.2018.01266*

express a T-cell receptor (TCR) containing an invariant α chain encoded by the *Vα14* gene in mice and *Vα24* gene in humans. These *i*NKT cells recognize endogenous, microbial, and synthetic glycolipid antigens presented by the major histocompatibility complex (MHC) class I-like molecule CD1d. Upon TCR stimulation by glycolipid antigens, *i*NKT cells rapidly produce large amounts of cytokines, including interferon-γ (IFNγ) and interleukin-4 (IL-4). Activated *i*NKT cells contribute to host protection against a broad spectrum of microbial pathogens, and glycolipid-mediated stimulation of *i*NKT cells ameliorates many microbial infections by augmenting innate and acquired immunity. In some cases, however, antigen-activated *i*NKT cells exacerbate microbial infections by promoting pathogenic inflammation. Therefore, it is important to identify appropriate microbial targets for the application of *i*NKT cell activation as a treatment or vaccine adjuvant. Many studies have found that *i*NKT cell activation induces potent adjuvant activities promoting protective vaccine effects. In this review, we summarize the functions of CD1d-restricted *i*NKT cells in immune responses against microbial pathogens and describe the potential applications of glycolipid-mediated *i*NKT cell activation for preventing and controlling microbial infections.

Keywords: invariant natural killer T cell, CD1d, glycolipid, adjuvant activity, microbial infection

#### INTRODUCTION

Natural killer T (NKT) cells are innate-type lymphocytes that recognize glycolipid antigens presented by the MHC class I-like molecule CD1d (1–7). The CD1d molecule is critical for both thymic development and the effector functions of NKT cells (1–7). These CD1d-restricted NKT cells are classified into two subpopulations, invariant NKT (*i*NKT) cells or type I NKT cells and type II NKT cells. In this review, we focus on *i*NKT cell functions in immunity and potential therapeutic applications, while the features and functions of type II NKT cells are summarized in several recent reviews (8, 9). The *i*NKT cell subtype expresses an invariant T-cell receptor (TCR) α chain encoded by the *V*α*14-J*α*18* gene in mice and *V*α*24-J*α*18* gene in humans. This TCRα chain is paired with a restricted repertoire of TCRα chains, such as Vβ8, 7, and 2 in mice and Vβ11 in humans (1–7). Upon TCR stimulation by microbe-derived glycolipid antigens or the potent synthetic lipid antigen α-galactosylceramide (α-GalCer), *i*NKT cells rapidly activate and produce large amounts of cytokines, including interferon-γ (IFNγ), interleukin-2, IL-4, IL-13, and IL-17A, and stimulate innate immune responses by activating antigen-presenting cells (APCs) and NK cells (1–6). *i*NKT cells activated by glycolipid antigens not only produce cytokines but also express CD40 ligand (CD40L), which stimulates the maturation of APCs, such as dendritic cells (DCs), leading to the augmentation of acquired immune responses (1–6). Through these unique signaling functions, CD1d-restricted *i*NKT cells participate in both innate and acquired immune responses against a variety of microbial pathogens, including bacteria, fungi, viruses, and protozoan parasites (10–15). In this review, we summarize the contributions of CD1d-restricted *i*NKT cells to immune responses against microbial pathogens by focusing on selected microbial infections. We also describe the potential applications of glycolipid-mediated *i*NKT cell activation for the development of new therapies and vaccines against infectious diseases.

# *i*NKT CELLS CONTRIBUTE TO INNATE IMMUNE RESPONSES AGAINST MICROBIAL PATHOGENS

Invariant natural killer T cells participate in the early phase of the immune response against various microbes through recognition of microbial components and stimulation of innate immune cells (10–15). Following infection by *Aspergillus fumigatus*, a major cause of invasive fungal infection in immunocompromised patients, *i*NKT cells produce IFNγ in response to recognition of endogenous antigens presented by CD1d, while APCs such as DCs release IL-12 in response to stimulation by β-glucan, resulting in the promotion of fungal clearance (16). Conversely, CD1d-deficient mice that lack *i*NKT cells exhibit delayed fungal clearance following infection by *A. fumigatus* (16).

*Streptococcus pneumoniae* (Pneumococcus) is the major cause of community-acquired pneumonia and meningitis, and is responsible for more than one million deaths annually. In the early phase of pneumococcal infection, *i*NKT cells contribute to neutrophil recruitment and bacterial clearance in the lungs through the release of neutrophil-recruiting cytokines such as tumor necrosis factor (TNF) and macrophage inflammatory protein-2 (17). Cell transfer experiments suggest that IFNγ produced by *i*NKT cells is essential for neutrophil recruitment (18). Other studies have reported that the *i*NKT cell response to Pneumococcus involves recognition of pneumococcal glycolipids (19, 20) and production of cytokines including IFNγ. The production of these cytokines is enhanced by IL-12 released from APCs stimulated by toll-like receptor (TLR) ligands of Pneumococcus (21).

CXCR6<sup>+</sup> NKT cells, which consist mainly of *i*NKT cells, patrol liver sinusoids for signs of bacterial infection (22). When the TCR is stimulated by αGalCer, an anti-CD3 antibody, or bacterial glycosphingolipid, these CXCR6+ NKT cells stop crawling (22, 23). These observations suggest that *i*NKT cells play a major role in liver surveillance and arrest when stimulated. It has been demonstrated that Kupffer cells in liver sinusoids capture intravenously injected *Borrelia burgdorferi*, a causative bacterium of the inflammatory disorder Lyme disease (24, 25). *i*NKT cells then accumulate around *Borrelia*-ingested Kupffer cells and form clusters, a response dependent on the stimulation of the cytokine receptor CXCR3 (26). *Borrelia*-ingested Kupffer cells express CD1d and can, therefore, activate *i*NKT cells (26). *i*NKT cells have been shown to recognize a *B. burgdorferi* glycolipid presented by CD1d (27, 28), and the ensuing activation contributes to bacterial clearance and prevention of joint and heart inflammation (24, 25). Consistent with these observations, mice deficient in *i*NKT cells or depleted of Kupffer cells exhibited bacterial dissemination to bladder, joints, and heart (26). These results indicate that *i*NKT cells contribute to the immune response against *B. burgdorferi* during the early phase of infection by recognizing bacterial glycolipids presented by Kupffer cells in liver sinusoids, thereby preventing bacterial dissemination to other tissues.

Invariant natural killer T cells also participate in host protection against post-stroke bacterial infection, a major cause of stroke-related death. In mice, the number of crawling or CD69-expressing *i*NKT cells rapidly declined following transient middle cerebral artery occlusion, while the number of *i*NKT cells producing IL-10 (but not IFNγ or IL-4) increased (29). At 24 h after stroke, mice exhibited systemic infection as evidenced by the detection of endogenous bacteria in multiple organs and tissues (29). In contrast, the activation of *i*NKT cells by α-GalCer induced bacterial clearance from these organs and tissues. Furthermore, stroke-induced bacterial infection was prevented by the administration of propranolol, a nonspecific β-adrenergic receptor blocker, or by 6-hidroxydopamine, a neurotoxin that depletes peripheral neuronal terminals of noradrenaline, through recovery of *i*NKT cell functions such as crawling and IFNγ production and by shifting to a Th1-dominant response (29). Intriguingly, protection from bacterial infection by propranolol has not been observed in CD1d-deficient mice that lack *i*NKT cells. These results imply that stroke-associated infection is mediated by the suppression of *i*NKT cell function.

Collectively, these results indicate that *i*NKT cells play an important role in host defense against the early phase of microbial infection through the recognition of microbial glycolipids and stimulation of innate immune cells.

## MECHANISMS OF *i*NKT CELL RESPONSES AGAINST MICROBIAL PATHOGENS

Previous studies have identified at least three mechanisms that trigger *i*NKT cell response to microbial pathogens: microbial glycolipid-mediated TCR activation, endogenous antigenmediated weak TCR stimulation with concomitant inflammatory cytokine-mediated stimulation, and activation solely by inflammatory cytokines (2, 10, 12, 15).

Several microbial lipid antigens have been identified that activate *i*NKT cells through CD1d presentation to the TCR. For example, mouse and human *i*NKT cells recognize α-linked glycosphingolipids (GSLs) containing either a galacturonic acid or a glucuronic acid derived from commensal *Sphingomonas* species of the intestine (30–32). The structures of these glycolipids are very similar to α-GalCer, but with subtle differences such as the carbohydrate moiety and a shorter C14 acyl chain replacing the C26 acyl chain of α-GalCer (30, 31, 33). In addition to GSLs, *i*NKT cells also recognize glycerol-containing glycolipids. *B. burgdorferi* expresses a diacylglycerol containing α-linked galactose called *B. burgdorferi* glycolipid-II (BbGL-II). A BbGL-II isoform containing a palmitic acid (C16:0) and an oleic acid (C18:1) potently stimulated mouse *i*NKT cells (27, 28). Human *i*NKT cells respond more strongly to BbGL-II isoforms containing fatty acids with greater unsaturation, such as oleic acid (C18:1) and linoleic acid (C18:2) (27, 28). *Streptococcus pneumoniae* express an α-linked diacylglycerol containing a glucose (Glc-DAG). The Glc-DAG containing a palmitic acid (C16:0) and a vaccenic acid (C18:1) is recognized by mouse and human *i*NKT cells (19). These *Sphingomonas*, *B. burgdorferi*, and *S. pneumoniae* glycolipids act as antigens that stimulate mouse and human *i*NKT cell TCRs and induce cytokine release. One intriguing question is how the *i*NKT cell TCR with an invariant α chain recognize different antigens. Recent structural analyses of the *i*NKT cell TCR−glycolipid−CD1d ternary complex revealed that *i*NKT cell TCR induces conformational changes to both the bacterial glycolipid antigen and CD1d, thereby allowing the recognition of different glycolipid antigens by a conserved binding orientation (20, 33).

Activation of *i*NKT cells by combined endogenous antigenmediated weak TCR stimulation and inflammatory cytokinemediated stimulation (2, 10, 12) is exemplified by *Salmonella typhimurium*, a Gram-negative bacterium expressing lipopolysaccharide (LPS). *S. typhimurium* has been shown to stimulate IFNγ release from *i*NKT cells despite the absence of a recognized glycolipid antigen (34). The activation of *i*NKT cells by *S. typhimurium* is mediated by IL-12 released from APCs stimulated by LPS through TLR4 and myeloid differentiation primary response 88 signaling (34). In addition, *i*NKT cell activation is partially dependent on CD1d (34), suggesting that *i*NKT cell activation during *S. typhimurium* infection requires a combination of weak TCR stimulation by an endogenous antigen and stimulation by inflammatory cytokines released by APCs in response to *S. typhimurium*.

In other cases, *i*NKT cells are activated solely by inflammatory cytokines (10, 12, 15). In the early phase of murine cytomegalovirus (MCMV) infection, a substantial number of *i*NKT cells produce IFNγ (35). However, this MCMV-associated cytokine production is independent of CD1d, but highly dependent on IL-12 and partially dependent on type I IFN (35). *i*NKT cells also amplify IFNγ release from NK cells and contribute to host protection against MCMV infection (35). In Nur77gfp reporter mice harboring T cells that express green fluorescent protein (GFP) upon antigen-mediated TCR stimulation, but not inflammatory cytokines, MCMV infection induced IFNγ production by *i*NKT cells without GFP expression (36). Collectively, these results show that the *i*NKT cell response to MCMV is independent of TCR stimulation but dependent on inflammatory cytokines. In contrast to MCMV, *S. pneumoniae* and *Sphingomonas paucimobilis* induced the expression of GFP and IFNγ in *i*NKT cells, indicating that these species activate *i*NKT cells through TCR stimulation (36). Alternatively, *S. typhimurium* and LPS did not induce GFP expression by *i*NKT cells, although these cells did produce IFNγ (36). These results suggest that inflammatory signals play an important role in *i*NKT cell activation in response to microbes that do not possess glycolipid antigens, greatly expanding the spectrum of *i*NKT cell-activating pathogens.

*Cryptococcus neoformans* is a fungal pathogen that causes pulmonary infection and can also disseminate to the central nervous system and cause meningitis, especially in immunocompromised individuals such as those with acquired immune deficiency syndrome. Following pulmonary infection of mice with *C. neoformans*, *i*NKT cells accumulated in the lungs, a response dependent on monocyte chemoattractant protein-1 (37). Jα18-deficient mice lacking *i*NKT cells exhibited delayed fungal clearance due to a weak Th1 response, normally a key immune response against *C. neoformans* infection (37). These results suggest that *i*NKT cells contribute to protection against cryptococcal infection through the stimulation of Th1 response. It was subsequently reported, however, that Jα18-deficient mice also exhibit defects in the rearrangement of Jα segments upstream of Jα18 (38). Therefore, the results obtained in Jα18-deficient mice may not be solely due to *i*NKT cell deficiency, especially under conditions involving the adaptive immune response.

A recent study has revealed an important role of *i*NKT cells in the initial formation of germinal centers during influenza infection. *i*NKT cells are a major source of early IL-4 release during infection, which is essential for the induction of germinal center B cells and ensuing IgG1 production (39). This enhanced IL-4 release is triggered by CD1d and IL-18 stimulation from CD169<sup>+</sup> macrophages (39). Furthermore, the transcriptomic analysis of lymph nodes in Zika virus-infected macaque monkeys revealed that IL-4 and NKT cell signatures, but not the Tfh cell signature, was strongly correlated with neutralizing antibody titer in the early phase of infection (39). These results suggest that *i*NKT cells promote initial germinal center formation and IgG production during the early stage of viral infection through macrophageinduced IL-4 release.

Taken together, these results highlight the importance of *i*NKT cells in host defense against various microbial infections through the stimulation of both innate and acquired immunity.

## GLYCOLIPID-MEDIATED ACTIVATION OF *i*NKT CELLS ENHANCES ANTIMICROBIAL IMMUNITY

As discussed, *i*NKT cells contribute to neutrophil recruitment during pneumococcal infection (17, 18). For instance, the activation of *i*NKT cells by α-GalCer promoted bacterial clearance through the recruitment of neutrophils and protected mice from lethal pneumococcal infection (17). Respiratory DCs, especially CD103<sup>+</sup> DCs, promote *i*NKT cell activation through the release of IFNγ and IL-17A, key cytokines conferring protection against pneumococcal infection (40). Stimulation of *i*NKT cells by α-GalCer also induces macrophage activation. During lung infection by *Pseudomonas aeruginosa*, α-GalCer treatment increased IFNγ and TNF in bronchoalveolar lavage fluid and the phagocytosis of bacteria by alveolar macrophages, resulting in rapid recovery from pneumonia (41). It has also been shown that α-GalCer treatment significantly inhibits malaria infection at the liver stage, but not at the blood stage, in an IFNγ-dependent manner (42). Alpha-C-galactosylceramide (α-C-GalCer), a C-glycoside analog of α-GalCer, induces longer IFNγ production and lower IL-4 production than α-GalCer (43), and α-C-GalCer has been shown to exhibit superior antimicrobial efficacy during the liver stage of malaria infection compared with α-GalCer (43). This superior effect of α-C-GalCer is dependent on IL-12, which is necessary for IFNγ production by NK cells, a major source of IFNγ. These results indicate that glycolipid-mediated *i*NKT cell activation enhances innate immune responses, resulting in a greater control of microbial infections at the early phase.

Glycolipid-activated *i*NKT cells augment the induction of effector CD4T cells and CD8T cells through the activation of APCs such as DCs. During *Chlamydophila pneumoniae* infection, α-GalCer-activated *i*NKT cells upregulated CD40 expression and IL-12 production by DCs, leading to the expansion of IFNγ-producing CD4T cells and IFNγ-producing CD8T cells and ultimately decreasing the bacterial burden in lungs (44, 45). It has also been shown that α-GalCer treatment enhances the Th1 response and fungal clearance during *C. neoformans* infection in an IFNγ-dependent manner (46). In the absence of IL-18, the increased IFNγ production and inhibition of fungal growth induced by α-GalCer were further enhanced through a greater production of IL-12 and IL-4 (47). Alpha-GalCer treatment also increases the memory CD4T cell pool size and alters the function of memory Th2 cells for increased IFNγ production (48). Further, α-GalCer treatment promotes the differentiation of central memory CD8T cells. During MCMV infection, α-GalCer treatment rapidly induced IFNγ and IL-4 production and decreased viral titers in spleen and liver (49). These α-GalCer-treated mice also exhibited greater numbers of MCMV antigen-specific central memory CD8T cells (49). These results suggest that glycolipidmediated *i*NKT cell activation may be an effective strategy to augment the induction of effector and memory CD4T cells and CD8T cells that contribute to host protection against microbial infections.

## *i*NKT CELLS CONTRIBUTE TO THE PATHOGENESIS OF SOME MICROBIAL INFECTIONS

In contrast to these documented benefits, *i*NKT cells play a detrimental role against the host during certain microbial infections by the induction or augmentation of inflammation, which results in the exacerbation of infection or causes severe acute or chronic inflammatory diseases (12, 13). *Candida* species colonize the skin and gastrointestinal and genitourinary mucosal surfaces and are a major cause of bloodstream infections among inpatients, with mortality rates from candidemia and invasive candida infections as high as 30−40% (50, 51). *i*NKT cells contribute to the pathogenesis of *C. albicans* infection, the most frequent *Candida* species. Following systemic *C. albicans* infection, Jα18-deficient mice lacking *i*NKT cells exhibited a higher survival rate and a lower fungal burden in various organs than wild-type (WT) mice because of the increased accumulation of macrophages and neutrophils in the peritoneal cavity (52). Consistent with the amelioration of infection by *i*NKT cell depletion, IL-10 levels were lower and IL-12p40 levels were higher in the serum of *C. albicans*-infected Jα18-deficient mice than infected WT mice. Conversely, NKT cell transfer exacerbated *C. albicans* infection in Jα18-deficient mice concomitant with reduced accumulation of macrophages and neutrophils (52). Furthermore, IL-10 treatment exacerbated *C. albicans* infection in Jα18-deficient mice, and transfer of IL-10-deficient NKT cells into Jα18-deficient mice significantly increased survival following *C. albicans* infection compared to the transfer of WT NKT cells (52). However, another study found no difference in susceptibility to *C. albicans* infection between Jα18-deficient and WT mice (53). This discrepancy is probably because of the different *C. albicans* strains employed and distinct routes of infection. It should also be reiterated that the difference in infection response by Jα18-deficient mice may not be due to *i*NKT cell deficiency alone, as these mice also show deficits in the rearrangement of Jα segments upstream of Jα18 (38).

Alpha-GalCer-mediated *i*NKT cell activation also exacerbates *C. albicans* infection. Alpha-GalCer-treated mice exhibited higher fungal burden in kidneys, higher IL-6 levels in serum and kidneys, wider dissemination of fungi, and shorter survival than control-infected mice (54). The number of neutrophils, the main effector cells controlling *C. albicans* infection, was significantly decreased in *C. albicans* infected and α-GalCer-treated mice, and this difference was IFNγ-dependent (54). It is thought that some bacterial species can disseminate to blood from the intestine in immunocompromised patients and activate *i*NKT cells. Furthermore, this mode of *i*NKT cell activation may exacerbate certain infections. Mice pre-infected with *Sphingomonas* bacteria, which are commensal and possess glycolipid antigens for *i*NKT cells (30–32), prior to *C. albicans* exposure exhibited enhanced IFNγ-dependent *i*NKT cell activation, increased production of inflammatory cytokines, and greater fungal burden (54). Collectively, these results indicate that *i*NKT cells participate in the pathogenesis of *C. albicans* infection and that *i*NKT cell activation by glycolipid antigens or bacterial infection can exacerbate *C. albicans* infection.

## GLYCOLIPID-ACTIVATED *i*NKT CELLS EXHIBIT EFFECTIVE ADJUVANT ACTIVITIES TO PREVENT MICROBIAL INFECTIONS

Many studies have demonstrated the potential adjuvant activities of glycolipid-activated *i*NKT cells for protection against microbial infections (11, 12, 14). For instance, immunization with malarial antigens and α-GalCer inhibited the liver stage of malaria and prevented parasitemia more effectively than malarial antigen alone. Immunization with malarial antigens and α-GalCer also increased the number of IFNγ-producing antigen-specific CD8T cells, major effector cells controlling the liver stage of malaria infection (55). Mice sublingually immunized with the *Mycobacterium tuberculosis* antigens Ag85B and ESAT-6 together with α-GalCer exhibited stronger antigen-specific CD4T- and CD8T-cell responses than mice immunized with Ag85B and ESAT-6 alone, and resulted in a significantly lower organ bacterial burden (56). Immunization with bacillus Calmette–Guérin (BCG)-incorporated α-GalCer or α-C-GalCer, an analog with a C-glycoside, induced a greater number of antigen-specific IFNγproducing CD8T cells than unmodified BCG through increased maturation of DCs by *i*NKT cells (57). Mice immunized with glycolipid-incorporated BCG also exhibited reduced bacterial loads in lungs and spleen compared with mice receiving unmodified BCG immunization (57). These vaccine effects were more evident with α-C-GalCer than α-GalCer, probably due to the lower IL-4 production and prolonged IL-12 production induced by α-C-GalCer compared to α-GalCer (43).

Figure 1 | Activation of CD1d-restricted invariant natural killer T (*i*NKT) cells augments both innate and acquired immunity to control microbial infection. The T-cell receptor (TCR) of *i*NKT cells recognizes glycolipid antigens presented by CD1d on antigen-presenting cells (APCs). In response, activated *i*NKT cells produce cytokines, including interferon-γ (IFNγ), interleukin-4 (IL-4), and IL-17A, that stimulate innate immune responses such as neutrophil (Neu) recruitment. Glycolipid-activated *i*NKT cells also express CD40 ligand (CD40L), which promotes APC maturation. *i*NKT cells provide cognate help to B cells to promote antibody production when glycolipidconjugated antigens are presented by B cells. Through cytokine release and CD40L–CD40 interaction, *i*NKT cells stimulate dendritic cells (DCs), triggering DC production of cytokines such as IL-12. These DC-derived cytokines stimulate IFNγ production by *i*NKT cells, which in turn enhances microbial clearance by stimulating macrophages. Activated *i*NKT cells also induce maturation of DCs that prime IFNγ-producing effector CD4T and CD8T cells, resulting in the clearance of microbes. The mature DCs induced by activated *i*NKT cells enhance the differentiation not only of effector T cells but also of memory T cells, conferring long-term protection against microbial infection.

Glycolipid-mediated *i*NKT cell activation also augments antibody production by B cells (58, 59). Intranasal administration of α-GalCer and influenza hemagglutinin (HA) vaccine or formalin-inactivated whole-virion vaccine induced higher titers of mucosal IgA and systemic IgG compared to influenza vaccine alone (60–62). Co-administration of α-GalCer and influenza vaccine also protected mice from lethal influenza virus infection, including H5N1 influenza virus infection (62), through enhanced viral clearance (60–62). Glycolipid-mediated *i*NKT cell activation also has adjuvant activity in swine. Indeed, α-GalCer showed excellent adjuvant activity with UV-inactivated influenza virus for increasing virus-specific antibody titers and IFNγ-producing cells in swine, and this immunization strategy protected against pandemic H1N1 influenza infection (63).

Although it is well known that follicular helper T (TFH) cells play a critical role in the stimulation of germinal center B cells for high-affinity antibody production, as well as differentiation of memory B cells and long-lived plasma cells, recent studies have demonstrated that follicular helper NKT (NKTFH) cells contribute to augmented IgG antibody production by vaccines containing an *i*NKT cell glycolipid antigen (59, 64–67). Immunization of mice with liposomes containing pneumococcal capsular polysaccharide (CPS) and PBS57, an α-GalCer analog, or CPS-α-GalCer conjugate vaccine induced NKTFH cells expressing PD-1 and CXCR5 or PD-1 and ICOS (66, 67). Mice treated with these vaccines showed an enhanced IgG1 production, indicating that cognate B cells are activated by T cells. Intriguingly, IgG1 production was dependent on CD1d expression by B cells and DCs, indicating that IgG1 production is induced by the cognate interaction of *i*NKT cells and B cells (66). These vaccines containing pneumococcal CPS and a glycolipid induced germinal center formation and CPS-specific memory B cells and long-lived plasma cells, which provided long-term protection against pneumococcal infection (67).

Collectively, glycolipid-mediated *i*NKT cell activation provides excellent adjuvant activities for inducing effector T-cell responses, promoting high-affinity antibody production by B cells, and for augmenting memory T- and B-cell responses (**Figure 1**).

# CONCLUDING REMARKS

Due to space limitations, this review focused on only a few selected studies of *i*NKT cell responses to microbes. However, numerous studies have demonstrated that CD1d-resticted *i*NKT cells contribute to immune responses against a broad spectrum of pathogenic microbes despite constituting only a small fraction of the leukocyte population. Due to the capacity for rapidly producing large quantities of cytokines in response to TCR stimulation, glycolipid-activated *i*NKT cells can augment both innate and acquired immunity, thereby providing protection against disparate microbial pathogens. However, in some cases, glycolipid-mediated *i*NKT cell activation may contribute to the pathogenesis of infection by exacerbating inflammation. Therefore, it is critical to distinguish which microbial targets are suppressed by *i*NKT cell activation for treatment or vaccine development. Considering the accumulating evidence for excellent adjuvant activities of glycolipid-mediated *i*NKT cell activation and the strong evolutionary conservation of *i*NKT cell responses to glycolipid antigens among experimental animals (usually mice) and humans (10–12), the inclusion of glycolipids inducing *i*NKT cell activation in vaccination regimens may be an effective strategy to prevent and control microbial infections in humans. It should be noted, however, that human and mouse *i*NKT cell responses to some glycolipids may differ as previously demonstrated by two C-glycoside analogs of α-GalCer (68) and *Borrelia* glycolipids (27). Therefore, careful consideration is needed when choosing a glycolipid antigen for clinical application of glycolipid-mediated *i*NKT cell activation.

## AUTHOR CONTRIBUTIONS

All authors contributed to this work and approved submission for publication.

#### REFERENCES


#### ACKNOWLEDGMENTS

The authors thank Yasuko Takatsuka for the preparation of the figure. The authors thank the NIH tetramer core facility for providing the CD1d tetramer.

# FUNDING

This study was supported by grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (KAKENHI 16H05349); the Japan Agency for Medical Research and Development, AMED (18im0210107j0002); the Ministry of Health, Labour and Welfare of Japan (H28 Shinko-Gyosei-005); the Yakult Bio-Science Foundation; the Takeda Science Foundation; the Life Science Foundation of Japan; and the Astellas Foundation for Research on Metabolic Disorders.


C-glycoside analogues against human versus murine invariant NKT cells. *J Immunol* (2009) 183:4415–21. doi:10.4049/jimmunol.0901021

**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 Kinjo, Takatsuka, Kitano, Kawakubo, Abe, Ueno and Miyazaki. 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.*

#### *Shin-ichiro Fujii\*, Satoru Yamasaki, Yusuke Sato and Kanako Shimizu*

*Laboratory for Immunotherapy, RIKEN Center for Integrative Medical Sciences (IMS), Yokohama, Japan*

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Mark L. Lang, University of Oklahoma Health Sciences Center, United States Tonya J. Webb, University of Maryland, Baltimore, United States*

> *\*Correspondence: Shin-ichiro Fujii shin-ichiro.fujii@riken.jp*

#### *Specialty section:*

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

*Received: 31 March 2018 Accepted: 22 May 2018 Published: 04 June 2018*

#### *Citation:*

*Fujii S, Yamasaki S, Sato Y and Shimizu K (2018) Vaccine Designs Utilizing Invariant NKT-Licensed Antigen-Presenting Cells Provide NKT or T Cell Help for B Cell Responses. Front. Immunol. 9:1267. doi: 10.3389/fimmu.2018.01267*

Vaccines against a variety of infectious diseases have been developed and tested. Although there have been some notable successes, most are less than optimal or have failed outright. There has been discussion about whether either B cells or dendritic cells (DCs) could be useful for the development of antimicrobial vaccines with the production of high titers of antibodies. Invariant (i)NKT cells have direct antimicrobial effects as well as adjuvant activity, and iNKT-stimulated antigen-presenting cells (APCs) can determine the form of the ensuing humoral and cellular immune responses. In fact, upon activation by ligand, iNKT cells can stimulate both B cells and DCs as *via* either cognate or non-cognate help. iNKT-licensed DCs generate antigen-specific follicular helper CD4<sup>+</sup> T cells, which in turn stimulate B cells, thus leading to long-term antigen-specific antibody production. Follicular helper iNKT cell-licensed B cells generally produce rapid, but short-term antibody. However, under some conditions in the presence of Th cells, the antibody production can be prolonged. With regards to humoral immunity, the quality and quantity of Ab produced depends on the APC type and the form of the vaccine. In terms of cellular immunity and, in particular, the induction of cytotoxic CD8+ T cells, iNKT-licensed DCs show prominent activity. In this review, we discuss differences in iNKT-stimulated APC types and the quality of the ensuing immune response, and also discuss their application in vaccine models to develop successful preventive immunotherapy against infectious diseases.

#### Keywords: NKT, Tfh, NKTfh, dendritic cell, vaccines

# INTRODUCTION

The success of vaccination strategies against viral infection or cancer depends on the efficient generation of appropriate antigen-specific T and B cell responses. Adequate antibody (Ab) responses of appropriate specificity elicited by vaccination are required to control and protect against many viral pathogens, such as influenza, human immunodeficiency virus, and human papilloma virus (1). Not only the suitable form of antigen, e.g., the commonly used inactivated virus, live attenuated virus, and recombinant viral protein, but also the optimal adjuvant are required for a successful vaccine. The development of ideal vaccine systems has been intensively explored to enhance the efficacy of weak antigens and broaden the immune response profile, leading to generation of high titer broadly neutralizing anti-viral antibodies.

Vaccines targeting B cells are essentially, of two types, T-dependent and T-independent, based on the requirement for T-cell help for Ab production (2). T-independent B cell responses are usually elicited by non-protein antigens that are unable to stimulate Th cells. Multimeric haptens or polysaccharides are typical T-independent B cell antigens that are recognized *via* the B cell antigen receptor (BCR). T-independent antigens generally induce robust and rapid B cell antibody responses, but with a low level of somatic hypermutation and thus affinity maturation, and limited isotype switching. T-dependent responses are typically induced by protein antigens and, as the term implies, there is cognate T-cell help for the antigen-specific B cells (3), which is provided by a specialized subset of CD4<sup>+</sup> T cells called T follicular helper (Tfh) cells. When antigens contact B cells in the follicles of secondary lymphoid organs, the antigen is internalized by the B cells upon binding to antigen-specific BCRs. The antigen is then processed and antigen-derived peptides are presented in the context of MHC class II (MHC II) molecules. Subsequently, the activated B cells are recruited to the border of the T cell and B cell zones, in which Tfh cells are generated following interacting with dendritic cells (DCs) presenting the same antigen. For the generation of Tfh cells, upregulation of the transcriptional repressor Bcl-6, costimulation by CD28, and stimulation with IL-21 have been reported as important factors (3). Also, by upregulating CXCR5, Tfh cells in turn localize to the boundary of the T and B cell zone (3), which is critical location for B cells to encounter Tfh cells.

Besides these classical T-dependent and T-independent vaccines, NKT cell-mediated vaccines have also been tested as a third vaccine candidate. NKT cells constitute approximately 0.05–0.2% of lymphocytes among human peripheral blood mononuclear cells and are classified into two groups: type I NKT cells express the invariant Vα14-Jα18 TCRα chain paired with either Vβ2, Vβ7, or Vβ8 in mice and Vα24-Jα18/Vβ11 in humans (4). The type I, invariant NKT cells (hereafter iNKT) recognize glycolipids, such as α-GalCer. By contrast, type II NKT cells display more diverse αβ-TCR pairings and respond to sulfatide, but do not to α-GalCer (5). Several reports have shown that iNKT cells can deliver helper signals to B cells directly or indirectly. In infectious diseases, neutralizing Ab production induced by vaccines represents a major protection mechanism against pathogens. Here, we review the features of iNKT cell-mediated Ab production, particularly by interacting directly or indirectly with B cells. We also discuss how these two pathways, i.e., vaccines utilizing iNKT cell help for B cells or iNKT cell help for DCs, augment efficient antigen-specific Ab production in the development of vaccination strategies against infectious diseases.

## THE ROLE OF iNKT CELLS IN INFECTIOUS DISEASES

Realization of the importance of iNKT cells in protection from infectious diseases has largely been based studies of the responses of Jα18- or CD1d-deficient mice, both of which lack iNKT cells, to viruses, bacteria, and parasites (6, 7). The outcome of most of these infectious models is worse in the iNKT-deficient animals. In studies of viral infections, iNKT cells play a protective role against influenza virus and cytomegalovirus (8, 9), herpes simplex virus type 1, and hepatitis B virus (10). In bacterial infection models, iNKT cells have been shown to be important against *Pseudomonas aeruginosa*, *Streptococcus pneumoniae*, *Mycobacterium tuberculosis* (11), *Chlamydia pneumoniae*, *Sphingomonas paucimobilis*, and *Staphylococcus aureus* (12). The protective responses of iNKT cells during infections are mediated by two mechanisms. First is the direct activation by stimulation of the NKT TCR by iNKT cell ligands expressed on various pathogens. Second is indirect activation of iNKT cells is through other immune cells and is due to the cytokine milieu and toll-like receptors (TLRs) agonists. In the first type of response, iNKT cells directly recognize glycolipids and lipoproteins, highly abundant in cell walls of many pathogens. These include glycosphingolipid in Gram-negative bacteria *Sphingomonas*, diacylglycerol in *Borrelia burgdorferi,* phosphatidylinositol mannoside in *Mycobacterium tuberculosis*, and glycosphingophospholipid in the protozoa *Leishmania donovani* (13). In the second type of response, iNKT cells are activated through macrophages during microbial infections. When infected, antigen-presenting cells (APCs) recognize bacterial signals *via* innate receptors, such as TLRs of macrophages, e.g., TLR4, TLR7, and TLR9. These allow macrophages to produce inflammatory cytokines, e.g., IL-12, which activate iNKT cells (14).

### Two Types of iNKT Based Vaccines by Focusing on iNKT-Licensed B Cells and iNKT-Licensed DCs to Induce Humoral Immunity

B cell response is generally defined as being cognate helper T (Th), namely Tfh-dependent (TD) or Tfh-independent (TI) (3). In terms of iNKT-mediated vaccines directed toward B cell responses, there are two strategies using "iNKT mediated-DC therapy" or "iNKT mediated-B cell therapy" (**Figure 1**). For the induction of humoral responses, iNKT cell-licensed DCs (non-cognate help) can induce Tfh cells, resulting in long-lasting antibody responses. On the other hand, iNKT cell-licensed B cells (cognate help) generally result in rapid and robust, but short-term responses. However, under some conditions, iNKT cell and B cell interactions seem to be more flexible and may be dependent on the experimental models. These are further discussed below.

#### Vaccines Using iNKT-Licensed B Cell Responses to Induce iNKTfh-Mediated Humoral Immunity

A lipid antigen component, e.g., a hapten or a lipid antigen conjugated to CD1d-binding glycolipids, can induce T-independent Ab production (15, 16). In addition to lipid antigens, a coadministration of T-dependent protein antigen plus α-GalCer, certain protein antigen conjugates, including protein–α-GalCer conjugates or protein incorporated into α-GalCer-liposomes, have been introduced as vaccine formulations for cognate iNKT help (15).

When such immunogens are taken up *via* antigen-specific BCRs, they promote extensive BCR cross-linking and enhance BCR internalization. Simultaneously, the iNKT cell ligand from

phagocytosed aAVC-HA to CD4+ T cells. (v) On the other hand, B cells also capture HA antigen. B cells can then be stimulated by antigen-specific CD4+ T cells, resulting in PB expansion, GC formation, and long-term Ab production.

the immunogen is loaded on CD1d inside of B cells and then expressed on the cell surface, resulting in a strong cognate iNKT– B cell interaction (16–19) (**Figure 1A**). Without incorporating the iNKT cell ligand, a T-independent response usually elicits only short-lived IgM antibody without germinal center (GC) formation, affinity maturation, and class switch recombination (18). However, with an iNKT cell ligand, even as a T-independent response, iNKT cells are converted to iNKTfh cells in a BCL6dependent manner. iNKT cells help B cells to proliferate and differentiate into extrafollicular plasma cells. They also robustly and rapidly induce GC formation, low-affinity antibody maturation, secretion of high titers of specific IgM, and early class-switched antibodies. On the other hand, this humoral immunity is shortlived and does not form a memory response (20–22).

In studies of the mechanism of iNKT-licensed B cell-dependent Ab production by lipid- and/or protein-conjugated complexes, several factors have turned out to be essential. Since iNKT cells in steady state are usually located in the marginal zone in spleen or interfollicular or medulla areas in LNs (23, 24), iNKT cells can more easily be activated by contacting marginal zone B cells than by follicular B cells after administration of α-GalCer (16, 21). The cognate help of iNKT cells for B cells presenting CD1d-lipid is dependent on IFN-γ, IL-21, CD40/CD40L, and B7-1/2 (16, 20, 21). SAP expression in iNKT cells plays a key role in the cognate help for antigen-specific B cells, although this is not the case for noncognate helper functions (25).

In terms of vaccine strategy, several types of vaccines have been described including T-dependent protein antigens combined with α-GalCer and these in nanoparticle liposome formulations (20–22). These showed the direct interaction between B cells and iNKT cells in response to T-dependent antigen, however, these strategies did not bring about high or long duration Ab production (20–22). On the other hand, other investigators have used different conditions and could demonstrate long-term responses. Lang et al. showed that immunization with T-dependent protein antigen together with α-GalCer in the presence of Th cells resulted in long-lasting Ab titers (26–28). A model using coadministration of NP-KLH and α-GalCer elicits Ig class-switched Ab production and requires Th cells and CD1d on B cells, but not on DCs. Here, BAFF- and APRIL-secreting NKT cells play a role in plasma cell longevity (27, 28). Also, liposomal nanoparticles displaying both lipid and polysaccharide antigens induce interactions with DCs for iNKT cell activation and then can elicit long-term B cell memory (29). Thus, the successful induction of iNKT cell-licensed B cells *in situ* demands three signals (**Figure 1**), BCR cross-linking (signal 1), costimulatory molecules, such as CD28-CD80/CD86, SLAM-family receptor signaling, and CD40-CD40L (signal 2), and inflammatory cytokines (IFN-γ and IL-21) (signal 3).

#### Vaccines Using iNKT-Licensed DC Responses to Induce Tfh-Mediated Humoral Immunity

We previously showed that coadministration of OVA protein antigen plus α-GalCer elicited both CD4<sup>+</sup> and CD8<sup>+</sup> T cell responses *via* DC maturation (30). Similar responses have been demonstrated in some cancer and infectious disease models (31–33). We elucidated the mechanism of iNKT-licensed DCs for T cell response; activated iNKT cells promote DC maturation *via* CD40/40L signaling and cytokines (IFN-γ and TNF-α). DCs *in situ* capture the protein antigen and α-GalCer-simultaneously, then present α-GalCer on CD1d to iNKT cells and the peptide on MHC class II to CD4<sup>+</sup> T cells or class I to CD8<sup>+</sup> T cells (34, 35). The antigen-specific CD4<sup>+</sup> Tfh cells that are derived from the helper CD4<sup>+</sup> T cells stimulate antigen-specific B cells that had already taken up the protein *via* the BCR and presented peptide in MHC class II (18, 36) (**Figure 1B**). Therefore, expression of both CD1d and MHC class II on DCs and MHC, II but not CD1d on B cells is essential for Ab production (37). Such a vaccine elicits a strong Ab response, i.e., characterized by GC formation, high affinity maturation, primary class-switched Ab, plasma cells, and memory B cells (37, 38).

In the steady state, DCs express CD80/86 to some extent. After activation by iNKT cells, expression of costimulatory molecules, i.e., CD80/86 and CCR7, on DCs is promptly upregulated. The mature DCs are key players for priming CD4<sup>+</sup> Tfh cells (3). When focused on DC subsets, we and others showed that the XCR1<sup>−</sup> DC (CD8<sup>−</sup> DC) subset is superior to the CD8<sup>+</sup> DC subset for CD4<sup>+</sup> Tfh cell priming (39). Shin et al. demonstrated this by using mAb to identify specific DC subsets, i.e., anti-DEC205 for the CD8<sup>+</sup> DC subset or anti-DCIR2 antibodies for the CD8<sup>−</sup> DC subset, and also showed that the CD8<sup>−</sup> DCs subset is superior to the CD8<sup>+</sup> DC subset because of its dominant ICOSL and OX40L expression (40).

The location of DC and iNKT cells before and after iNKT cell activation has recently been clarified. As discussed previously, in the steady state, iNKT cells are localized in the marginal zone (MZ) of the spleen or medulla in LN (41), whereas XCR1<sup>−</sup> DCs (CD8<sup>−</sup> DC or 33D1<sup>+</sup> DCs) are localized in the bridging channel (BC), a unique region of the spleen that spans the interface between the red pulp (RP) and white pulp (WP) and XCR1<sup>+</sup> DCs reside in the MZ. After i.v. immunization with antigens and/or adjuvants, the majority of both DC subsets migrate into the WP, but XCR1<sup>+</sup> DCs preferentially go to the CD8<sup>+</sup> T cell area and XCR1<sup>−</sup> DCs prefer to go to the CD4<sup>+</sup> T cell area (42, 43). After activation by α-GalCer or ligand-containing cells, activated iNKT cells accumulate in the MZ or BC and can be in close contact with DCs that have already taken up antigen and α-GalCer (44, 45). iNKTlicensed DCs then relocate to each T cell area.

Germinal center formation in secondary lymphoid organs is considered key for inducing better Ig class switching, somatic mutation, affinity maturation, and long-lasting Ab responses (46). In addition to Bcl6fl/flCD4-cre mice, LTα−/<sup>−</sup> mice and Lyn<sup>−</sup>/<sup>−</sup> mice are deficient in GC formation. When LTα−/<sup>−</sup> mice and Lyn<sup>−</sup>/<sup>−</sup> mice are immunized with antigen plus adjuvant, they respond with long-lasting Ab production, probably due to the generation of long-lived plasmablasts (19, 47–51). However, judging from the data using Bcl6fl/flCD4-cre mice, the formation of GC is essential for iNKT cell-mediated Ab production. When NKTfh and Tfh cells are compared, iNKTfh cells induce earlier GC formation, but it is more short-lived compared to GCs induced by Tfh cells (4). On the other hand, Tfh cells help in the induction of mature GCs better than iNKTfh cells (4). The expression of Bcl6 by Th cells is apparently crucial for efficient Ab production. Thus, iNKT celllicensed DCs induce antigen-specific CD4<sup>+</sup> Tfh cells and drive them into the B cell zone (**Figure 1B**). Then, Tfh cells are engaged in cognate interactions with B cells, resulting in the formation of early GCs and also leading to the long-lasting production of antigen-specific Abs.

The development of effective vaccines is a critical need. As discussed above, immunization by coadministration of protein antigen together with an iNKT cell ligand clearly generates potent Ab production. Traditional immunization protocols usually require a high dose of protein, e.g., 100–500 µg OVA protein per mouse is typically injected (15) yet, even so, Ab production is not impressively high (52). We have reported the artificial adjuvant vector cell (aAVC) system as an efficient vaccine strategy that can potently induce innate and adaptive immunity and this will be discussed in detail in the next section. But, to summarize this section on the induction of antigen-specific Tfh cells and Ab, effective iNKT cell-licensed DCs *in situ* are required (30, 34, 35, 53). The DCs require three factors: (i) expression of the appropriate antigen peptide–MHC complex, (ii) upregulation of costimulatory and chemokine molecules, including CD80/CD86, ICOSL, and OX40L, and (iii) production of inflammatory cytokines and chemokines, such as IL-6, IL-12, and CCL17.

#### An Efficient Strategy Using iNKT-Licensed DCs to Induce Humoral Immunity

We have established the aAVC system, comprised of a CD1d<sup>+</sup> cellular vaccine incorporating foreign protein antigen plus an iNKT cell glycolipid antigen. We chose "adjuvant vector cell" as the name for this cellular vaccine to describe the fact that the "vector like cells" deliver the antigen as well as an iNKT cell adjuvant to host DCs. We used NIH3T3 cells for mouse and HEK293 cells for humans as vector cells (54, 55). These cells are co-transfected with CD1d and antigen mRNA, and then loaded with α-GalCer for use (54–56). The aAVC, therefore, express the α-GalCer–CD1d complex on their surface and antigen protein intracellularly. The aAVCs directly activate iNKT cells *via* the α-GalCer ligand, and iNKT cells producing IFN-γ can then simultaneously activate NK cells. The combination of innate killer iNKT/NK cells capable of producing IFN-γ then eliminates the adjuvant vector cells, which is not syngeneic with the recipient. Subsequently, the killed aAVC are taken up by DCs (CD8<sup>+</sup> or XCR1<sup>+</sup> DCs) *in situ*, thereby several immunogenic features of DCs are engaged. The aAVC captured by DCs in lung, liver, and spleen undergo maturation due to interaction with CD40L on iNKT cells and then produce inflammatory cytokines. The aAVC vaccine can efficiently generate antigen-specific CD8<sup>+</sup> T cells and memory T cells (36). Interestingly, in aAVC-vaccinated mice, antigen-specific CD4<sup>+</sup> Tfh primed by XCR1<sup>−</sup> DCs and GC were both generated, resulting in induction of long-term Ab production (39). Killed aAVCs are phagocytosed, mainly by XCR1<sup>+</sup> DC (CD8<sup>+</sup> DC) cells and, presumably, the separated protein antigen and α-GalCer are endocytosed simultaneously by the XCR1<sup>−</sup> DC subset. MHC II<sup>−</sup>/<sup>−</sup> mice do not have CD4<sup>+</sup> T cells but do have iNKT cells. Immunization with aAVC-OVA elicited no detectable specific Ab response in these mice (39), indicating that iNKT celllicensed DC strategies require MHC II for CD4<sup>+</sup> T cell-mediated humoral immunity. Although both Tfh and iNKTfh co-exist in the aAVC-mouse models, Tfh cells are superior to iNKTfh cells for inducing Ab production (39).

# APPLICATION OF AN iNKT CELL-TRIGGERED DC-DESIGNED VACCINE "ARTIFICIAL ADJUVANT VECTOR CELL" FOR INFLUENZA INFECTION

Influenza virus is a member of the orthomyxoviridae family that contains a segmented negative-sense single-stranded RNA genome (57). Influenza infection is a major global health problem,

which is initially caused by viral infection of the respiratory tract. Upon viral infection, iNKT cells in the follicular areas of LNs facilitate IL-4 signaling to B cells which triggers the seeding of GC cells and B cell immunity (23). However, to provide sufficient protection against influenza virus in humans, both adaptive T cell and Ab-producing B cells need to be established and maintained as immunological memory.

Immunity against influenza virus is largely mediated by neutralizing antibodies that target the major surface glycoprotein hemagglutinin (HA) (58), in particular, the immunodominant head region of HA, or the viral neuraminidase (NA) (59). Preexisting neutralizing Ab, rather than the recall of virus-specific CTL, is thought to account for memory anti-viral protection. Unfortunately, however, the antiviral antibodies generated by immunization or natural infection are only effective against a limited number of viral strains.

Several studies have addressed the possibility of using a combined vaccine approach, i.e., coadministration of inactivated influenza virus (IIV) together with α-GalCer, to enhance protective efficacy *via* subcutaneous or intranasal administration (60–62). We established CD1d<sup>+</sup> HA mRNA-transfected cells loaded with α-GalCer (aAVC-HA) and demonstrated that this is a more efficient strategy for generating antigen-specific Ab production than coadministration of antigen plus α-GalCer (**Figure 2**) (39). The efficacy of aAVC appears to depend on the GC and Tfh. We could easily modify the HA protein, depending on the circulating virus strain, and the manufacturing process for this vaccine has a much shorter timeframe than others, e.g., egg-derived vaccines, which would be especially valuable during a flu pandemic. The aAVC vaccine thus holds great promise as a potential broad spectrum prophylactic or therapeutic agent and for the development of a universal influenza or other viral vaccine

#### REFERENCES


(39). As a future study, if the optimal stem region antigen were incorporated into an iNKT cell-mediated vaccine, a universal HA stem antigen-expressing aAVC could be developed.

# CONCLUSION

iNKT cells play an immunomodulatory role during immune responses. To utilize this capacity as an adjuvant in terms of novel vaccine strategies, it is essential to understand the interaction of iNKT cells with APCs in the host. We here summarized details of the relationship of iNKT cell-triggered B cells and DCs with Ab production and compared them in terms of vaccine development. Now we need to consider the future directions and challenges in translating these findings from experimental data obtained from mice to use in the clinic.

## AUTHOR CONTRIBUTIONS

SF and KS conceptualized, wrote, and edited the manuscript. SY and YS wrote and edited the manuscript.

## ACKNOWLEDGMENTS

The authors are grateful to Dr. Peter Burrows for peer-reviewing and helpful comments in the preparation of the manuscript. This work is supported by grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan (to KS and SF, grant number 15K09590) and the Japan Agency for Medical Research and Development (translational research network program) (to SF, grant number 15lm0103002j0004).


19. Vomhof-DeKrey EE, Yates J, Leadbetter EA. Invariant NKT cells provide innate and adaptive help for B cells. *Curr Opin Immunol* (2014) 28:12–7. doi:10.1016/j.coi.2014.01.007

20. King IL, Fortier A, Tighe M, Dibble J, Watts GF, Veerapen N, et al. Invariant natural killer T cells direct B cell responses to cognate lipid antigen in an IL-21dependent manner. *Nat Immunol* (2012) 13(1):44–50. doi:10.1038/ni.2172


memory CTL generation and protective immunity. *Proc Natl Acad Sci U S A* (2009) 106(9):3330–5. doi:10.1073/pnas.0813309106

**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 Fujii, Yamasaki, Sato and Shimizu. 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.*

*Francesca A. Ververs1 , Eric Kalkhoven2 , Belinda van't Land3,4, Marianne Boes1,4 and Henk S. Schipper1,5\**

*<sup>1</sup> Laboratory for Translational Immunology, University Medical Center Utrecht, Utrecht, Netherlands, 2Department of Molecular Cancer Research, Center for Molecular Medicine, University Medical Center Utrecht, University Utrecht, Utrecht, Netherlands, 3Department of Immunology, Nutricia Research, Utrecht, Netherlands, 4Department of Pediatric Immunology, Wilhelmina Children's Hospital, University Medical Center Utrecht, Utrecht, Netherlands, 5Department of Pediatric Cardiology, Wilhelmina Children's Hospital, University Medical Center Utrecht, Utrecht, Netherlands*

Invariant natural killer T (iNKT) cells are lipid-reactive T cells with profound immunomodulatory potential. They are unique in their restriction to lipid antigens presented in CD1d molecules, which underlies their role in lipid-driven disorders such as obesity and atherosclerosis. In this review, we discuss the contribution of iNKT cell activation to immunometabolic disease, metabolic programming of lipid antigen presentation, and immunometabolic activation of iNKT cells. First, we outline the role of iNKT cells in immunometabolic disease. Second, we discuss the effects of cellular metabolism on lipid antigen processing and presentation to iNKT cells. The synthesis and processing of glycolipids and other potential endogenous lipid antigens depends on metabolic demand and may steer iNKT cells toward adopting a Th1 or Th2 signature. Third, external signals such as toll-like receptor ligands, adipokines, and cytokines modulate antigen presentation and subsequent iNKT cell responses. Finally, we will discuss the relevance of metabolic programming of iNKT cells in human disease, focusing on their role in disorders such as obesity and atherosclerosis. The critical response to metabolic changes places iNKT cells at the helm of immunometabolic disease.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Jae B. Kim, Seoul National University, South Korea Hiroshi Watarai, University of Tokyo, Japan*

#### *\*Correspondence:*

*Henk S. Schipper h.s.schipper-3@umcutrecht.nl*

#### *Specialty section:*

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

*Received: 31 March 2018 Accepted: 14 May 2018 Published: 28 May 2018*

#### *Citation:*

*Ververs FA, Kalkhoven E, van't Land B, Boes M and Schipper HS (2018) Immunometabolic Activation of Invariant Natural Killer T Cells. Front. Immunol. 9:1192. doi: 10.3389/fimmu.2018.01192*

Keywords: immunometabolism, NKT, obesity, atherosclerosis, sphingolipid, AMPK, mTOR

#### INVARIANT NATURAL KILLER T (iNKT) CELLS CENTER STAGE IN IMMUNOMETABOLIC DISEASE

Immunometabolic diseases such as obesity, type 2 diabetes, and cardiovascular disease (CVD) are the major health burdens of our time and illustrate the intricate web between metabolic dysregulation and inflammation (1). The links between metabolism and inflammation may be explained from an evolutionary perspective. An effective immune defense critically depends on efficient energy storage and release, as reflected by the co-evolution of the immune system and metabolism in *Drosophila* fat bodies, and the reminiscent immune cell functions of adipocytes in humans and other higher organisms (2). Unfortunately, evolution could not foresee the endemic nutritional overload in 21st century Western societies, causing glucotoxicity and lipotoxicity, and propagating local and systemic inflammation (3).

NKT cells were identified as important players in immunometabolism due to their unique response to lipid antigens and hybrid qualities of both the innate and adaptive immune system (4). NKT cells readily produce copious amounts of Th1, Th2, and/or Th17 cytokines upon activation, which resembles an innate activation scheme (5). Similar to T cells, NKT cells develop in the thymus and undergo positive and negative thymic selection. However, instead of interacting with MHC class

**365**

2 molecules, iNKT cells are selected by CD1d-expressing thymocytes. Two NKT cell subtypes have been defined: type 1 signifies CD1d-restricted iNKT cells carrying an invariant T cell receptor that recognizes the prototypic ligand alpha-galactosylceramide, while type 2 signifies CD1d-restricted iNKT cells carrying different T cell receptors not recognizing alpha-galactosylceramide (6). This review focuses on type 1 NKT cells, also known as iNKT cells, which represent the most studied NKT cell subset.

Invariant natural killer T cell frequency in peripheral blood is low, but they are highly enriched in adipose tissue (AT) in mice and humans (7, 8). Functionally, AT-resident iNKT cells have an anti-inflammatory phenotype by secreting IL-4, which contributes to prevention of insulin resistance and AT inflammation (7, 9). In obesity, the protective IL-4 production by iNKT cells is lost, and total iNKT cell numbers in AT and peripheral blood decrease, making leeway for adipose tissue inflammation, insulin resistance, and type 2 diabetes to develop (7–10). The same phenomenon is observed in other metabolic disorders. When comparing human identical twins, of which only one sibling developed type 1 diabetes, diabetic siblings show lower frequencies of iNKT cells. When multiple iNKT clones were compared from the twins, all clones isolated from diabetic siblings produced only IFN-γ upon stimulation, while all clones isolated from the healthy twin produced both IL-4 and IFN-γ (11). In atherosclerosis, a similar decrease in iNKT cell numbers and production of IL-4 is observed in established CVD (12). Notably, iNKT cell numbers in peripheral blood seem to increase in the earliest phase of atherosclerosis, accompanied by an increase in IL-4 production, GATA3- and CD69 expression, and increased proliferative capacity (13). This model, in which iNKT cells play an anti-inflammatory or pro-homeostatic role early in disease development, seems widely applicable for human disease (14), and begs the question: what do iNKT cells see when trouble starts stirring?

#### iNKT CELL ACTIVATION BY SPHINGOLIPID LIGANDS

In the early 1990s, it was discovered that iNKT cells can be activated by glycosphingolipids (GSL) following identification of alpha-galactosylceramide, a potent marine sponge sphingolipid antigen identified in a cancer antigen screen (15). Since then, endogenous sphingolipids have been scrutinized as potential lipid antigens for iNKT cells.

Sphingolipids are synthesized either *via* the *salvage pathway,* by degradation and re-usage of existing sphingolipids, or *via de novo* synthesis in the endoplasmic reticulum (ER), by attachment of a fatty acid to a sphingosine base (16). Spingomyelinases and glucosidases are important enzymes in the *salvage pathway*, converting membrane sphingomyelin and glucosylceramides back to ceramide within the lysosome (17). Serine palmitoyl transferase (SPT) and ceramide synthases are important for *de novo* synthesis. *De novo* synthesis is orchestrated by six different ceramide synthases (CerS), which determine the length of the fatty acid chain attached to the sphingosine base. Sphingosine with one fatty acid attached is called ceramide, which is the central metabolite in sphingolipid metabolism. More complex sphingolipids such as GSL are generated in the Golgi by addition of different headgroups by UDP-glucose ceramide glucosyltransferase (UGCG) and other glycosyltransferases (18). Translocation to the Golgi is facilitated by ceramide transfer proteins (CERT) (17). The simplest glycosphingolipid has only one sugar residue attached, either glucose or galactose. The sugar headgroup can be attached to ceramide in a beta- or alpha-anomeric fashion. To date, only beta-anomeric GSL have been identified in humans. Some studies reported iNKT cell reactivity to beta-linked GSL, but this was disputed later as contamination of alpha-linkages was found in the preparations (19–22). The alpha-anomeric linkage remains one of the key determinants for antigenicity (20, 23, 24). Enter the search for endogenous lipid ligands continues as, unfortunately, developing a robust method for isolation of these ligands is technically challenging. In the meantime, extensive studies on the effect of various synthetic alpha-galcer analogs on iNKT function were performed, including analogs with truncated alkyl chains, varying saturation status, or the presence of aromatic structures (24–26). These efforts revealed that analogs with a shorter alkyl chain can elicit an IL-4 response without prior IFN-γ induction in mice *in vivo* (alpha-GalCer C10:0, alpha-GalCer C20:2, alpha-GalCer C20:4, OCH, alpha-GalCer-PGB1) (26). However, in human iNKT cells, even though almost all glycolipid analogs elicit a potent cytokine response, there is hardly any Th2-polarization (24, 26). Enter differences between mouse and human ligand-mediated activation are abound: there are differences in potential endogenous ligands and where the ligands derive from, considering that human CD1d and mouse CD1d1 travel to different subcellular compartments for endogenous ligand extraction (27–32). The secretory route from the ER, *via* the Golgi, to the plasma membrane is similar for human CD1d and mouse CD1d1. Upon folding in the ER and association with beta-2-microglobulin, lipid transfer proteins such as microsomal transfer protein mediate loading of chaperone lipids in the ER and/or endogenous lipid antigens in the Golgi (33–35). The endolysosomal recycling route, however, is different for human CD1d and mouse CD1d1. On the basal side of the membrane, CD1d has a short cytoplasmic tail carrying a sorting motif. The sorting motif binds to the adaptor protein complex 2 upon which membrane internalization is mediated to enter the early endosome (36). Only mouse CD1d1, but not human CD1d, can also bind adaptor protein 3, which then targets late endosomes and lysosomes (31). Considering the observed co-localization of human CD1d with the lysosomal membrane protein LAMP1, the lack of a lysosomal sorting motif does not preclude lysosomal transportation of human CD1d (31, 32). Nevertheless, differences in endolysosomal trafficking may result in loading of different lipid antigens. LDL receptor (LDLR)-mediated uptake of GSL for example, is processed in the endosomal compartment (37), while the *salvage pathway* of plasma membrane GSL starts in the lysosomal compartment (38) (**Figure 1**). These differences are important to keep in mind when studying iNKT cells and Th1/ Th2 skewing in mouse models.

#### SPHINGOLIPIDS IN IMMUNOMETABOLIC DISEASE

Sphingolipids play a key role in immunometabolic disease, which supports their potential relevance as iNKT cell antigens (17, 18, 39).

Figure 1 | CD1d lipid loading at the crossroads of glycosphingolipid metabolism. In the ER–Golgi pathway that is similar for mouse and human, CD1d heavy chains assemble with β2M and chaperone lipids in the ER before transit to the cell surface. Alternatively, the Golgi complex produces GSL that are loaded onto CD1d by microsomal transfer protein, and may serve as endogenous lipid antigens. Ceramide precursors are transported to the Golgi by ceramide transfer protein (CERT). In the Golgi, UGCG and other glycosyltransferases convert ceramide into GSL. These GSL endproducts can be loaded onto CD1d, or transported to the plasma membrane in membrane-bound transport carriers. In the endolysosomal pathway, mouse CD1d1 may target the endosome and lysosome directly *via* interaction with sorting motifs adaptor protein 2 (AP2), which targets the endosomal compartment, and adaptor protein 3 (AP3), which targets the lysosomal compartment. Human CD1d is internalized *via* AP2 but cannot bind AP3. However, human CD1d can still be found in the lysosomal compartment. Possibly, ER-resident CD1d proteins gain access to the endolysosomal compartment *via* an auxiliary pathway, in conjunction with MHC class II-associated invariant chain (li). In the endolysosomal compartments, CD1d proteins are loaded with exogenous or endogenous lipid antigens, orchestrated by a variety of lipid transfer proteins including GM2 activator (GM2a), saposins A-D (Sap), and Niemann-Pick type C2 protein. Exogenous lipid antigens are delivered to the endosomal compartment *via* endocytosis of LDLR-associated glycolipids, MR-associated microbial lipids, and other scavenger receptors. Some exogenous lipids require processing into antigenic lipids before CD1d-loading, for example, through lipid hydrolases (Hy). Endogenous lipid antigens are delivered to the endolysosomal compartment *via* endocytosis of membrane-associated GSL, which can be loaded onto CD1d or degraded in lysosomes by glycohydrolases (Gly-Hy) and accessory proteins, before recycling to the ER (salvage pathway). Upon lipid antigen loading in the endolysosomal compartments, CD1d–lipid complexes recycle back to the cell surface for interaction with the invariant TCR on invariant natural killer T cells. Abbreviations: AP, adaptor protein; β2M, β2-microglobulin; ER, endoplasmic reticulum; Gly-Hy, glycohydrolase; GM2a, GM2 activator; GSL, glycosphingolipids; Hy, hydrolase; LDLR, LDL-receptor; Li, MHC class II-associated invariant chain; MR, mannose receptor; Sap, saposins; TCR, T-cell receptor; UGCG, UDP-glycose ceramide glucosyltransferase; VLDL, very-low-density lipoprotein.

The six CerS involved in *de novo* sphingolipid synthesis are differentially expressed, allowing for tissue- and cell type-dependent variation in ceramide acyl chain length profiles (16). Importantly, differences in sphingolipid chain length may orchestrate glucose metabolism and mitochondrial homeostasis, and play a key role in obesity and type 2 diabetes. For example, reduction in C16 sphingolipid levels increases beta-oxidation and improves glucose metabolism, with a 30–50% reduction in C16 levels being sufficient to prevent diet-induced obesity and insulin resistance (40). Furthermore, acyl chain length may determine cell fate. While C22-24 ceramides prevent apoptosis, C16 ceramides can induce apoptosis *via* activation of the intrinsic mitochondrial apoptotic pathway (16, 41–43). In CVD, C16 ceramides are considered harmful, as Cer(d18:1/16:0)/Cer(d18:1/24:0) ratios predict cardiovascular death (44). Intriguingly, sphingolipids such as glucosylceramide, lactosylceramide, ceramide, dihydroceramide, sphingomyelin, and sphingosine-1-phosphate (S1P) amass in human atherosclerotic plaques. All except S1P induce apoptosis *in vitro*, and are associated with plaque instability (39, 45, 46). Consequently, D-PDMP, an inhibitor of glucosylceramide synthase and lactosylceramide synthase, has an astounding protective effect on atherosclerosis development in ApoE−/− mice. Treatment led to complete prevention of intima media thickening and arterial stiffening measured as aortic pulse-wave velocity (47). Likewise, treatment with the SPT inhibitor myriocin was shown to ameliorate insulin resistance and atherosclerosis in mouse and rat models (18).

It is tempting to speculate that the pathophysiological role of sphingolipids in immunometabolic disease is partly explained by their role as iNKT cell ligands. In order to identify sphingolipid antigens potentially involved in immunometabolic disease, several approaches may be explored. First, the intracellular crossing between the sphingolipid metabolism and the iNKT cell lipid loading pathways can be scrutinized (48, 49). The Golgi and lysosomal compartment facilitate exchange of chaperone lipids bound to CD1d for antigenic lipids and are, therefore, important crossroads in iNKT cell lipid antigen loading and sphingolipid metabolism (**Figure 1**). Alternatively, animal or cellular models of naturally occurring disorders in sphingolipid metabolism may be exploited to identify metabolic intermediates or end products in lipid antigen presentation. For example, the mouse model for Fabry disease, alpha-galactosidase A knock out, combined with globoside 3 synthase- or isogloboside 3 synthase double knock out, revealed that globosides, but not isoglobosides, are responsible for iNKT cell deficiency in Fabry disease (50). Hexb knock out mice, a model for Tay Sach and Sandoff disease, also show severe iNKT cell deficiency (51). The iNKT cell deficiency in these lysosomal storage disease mouse models suggests that the glycosphingolipid synthetic pathways involved may contain endogenous lipid antigens for iNKT cells. Alternatively, glycosphingolipid accumulation may hinder antigen presentation similarly to acLDL accumulation or cholesterol accumulation following NPC1 deficiency (50), and possibly NPC2 deficiency (52), regardless of the glycosphingolipid involved (50, 53). The latter model aligns with the lipid raft hypothesis, which proposes that iNKT cell activating lipids may either function as *bona fide* lipid antigens, *or* may impact CD1d loading, stabilization or clustering on the cell membrane, and in that way enforce iNKT cell activation (50, 54–56). Finally, immunometabolic diseases may serve as a starting point to identify sphingolipid antigens (18). In CVD, for example, lipoprotein particles that enter the cell *via* LDL receptor- and scavenger receptor-mediated uptake are important carriers of glycosphingolipid species (39). The increased uptake of oxidized lipoproteins *via* class A scavenger receptors in atherosclerosis may potentially induce a different iNKT cell effector response due to co-transported glycosphingolipid species (37). In conclusion, sphingolipids are promising candidate antigens from an immunometabolic perspective. However, translation of the changes in sphingolipid metabolism to iNKT cell activation remains technically challenging.

#### INDIRECT ACTIVATION OF iNKT CELLS

At present, two principal ways of iNKT cell activation have been described. As discussed before, high affinity lipid antigens may induce a strong T-cell receptor (TCR) signal and activate iNKT cells directly. Alternatively, innate activation of an antigenpresenting cell (APC) leads to presentation of an endogenous lipid ligand with low affinity, followed by a weak TCR signal that can fully activate iNKT cells in combination with cytokine co-stimulation secreted by the activated APC (5, 57–60). Innate activation of the APC can either be due to inflammatory or metabolic cues. For example, LPS can trigger iNKT-cell activation. This activation is CD1d- and APC dependent. However, this activation is also IL-12 dependent, in both mice and in human *in vitro* models (57, 58). The current model is that iNKT cells are first triggered *via* their TCR to upregulate CD40L, to enhance APC–iNKT cell interaction, and maintain proximity for paracrine IL-12 co-stimulation, which is induced by CD40:CD40L interaction (23, 61, 62). It was postulated that the duration of TCR triggering determines CD40L upregulation. Duration of TCR triggering, in turn, depends on the alkyl chain length and stabilization of the CD1d–glycosphingolipid-complex (24, 25). Furthermore, IL-12 ultimately drives a Th1-biased iNKT cell response (62). *In vivo*, IL-4 can be detected 2 h after intraperitoneal lipid agonist injection, while IFN-γ is measured after 6 h, as is IL-12 (61). The relatively slow IFN-γ response, which also requires prolonged and enhanced APC–iNKT cell interaction, suggests that the IFN-γ response requires *de novo* IFN-γ protein synthesis, while IL-4 is pre-synthesized and can, therefore, be released instantly even upon weak or short TCR stimulation. In fact, binding affinity of glycolipids to CD1d correlates very well with IFN-γ production but not at all with IL-4 production by human iNKT cells (63). Transgenic mouse studies revealed that the Notch and RBP-J pathway might be responsible for the IL-4 response by iNKT cells, mainly regulated by the conserved noncoding sequence-2 enhancer (CNS-2). As Notch- and TCR signaling synergistically contribute to T cell activation, this could explain why a weak TCR signal still allows for a relatively high IL-4 production by iNKT cells (64). This leaves us with a model in which iNKT cells are potent effector memory IL-4 producers upon homeostatic, weak antigenic stimulation, which can become highly inflammatory IFN-γ producing cells in an environment in which either high affinity ligands are available, or where IL-12 or CD40:CD40L co-stimulation are more easily established, either directly or due to activation of APCs.

## IMMUNOMETABOLIC iNKT CELL ACTIVATION

The intracellular sphingolipid pool and subsequent CD1d ligand loading may be affected by TLR-activation or altered metabolism in the APC (65, 66). Several mechanisms were recently reported. For example, blocking glycolysis and increasing fatty acid oxidation (FAO) *via* AMPK provokes a CD1d-mediated iNKT cell cytokine response (65). AMPK is a nutrient-sensing kinase that is activated under low glucose conditions and blocks cellular glycolysis while promoting cell-sparing oxidative phosphorylation (67). Adiponectin, an adipokine produced by lean adipocytes that promotes insulin sensitivity, can directly activate AMPK (65, 68). Adiponectin overexpression in *ob/ob* obese mice protects against insulin resistance and AT inflammation (69), perhaps activating iNKT cells in a Th2-skewed manner through direct or indirect iNKT cell modulation. Conversely, TLR signaling leads to increased glycolysis, reduced FAO, and AMPK inhibition in the APC (70), but again leads to iNKT cell activation (58, 71). TLR-induced glycolysis is established *via* HIF-1α upregulation despite normoxic conditions, analogous to the Warburg effect (70). This pathway may be potentiated by mild hypoxia (72). In early obesity, relative hypoxia arises following adipocyte hypertrophy and hyperplasia and has been dubbed one of the initiating events in AT inflammation (1, 73). Indeed, hypoxia is also an important factor in cancer and in atherosclerosis (72, 74). iNKT cells are sensitive to HIF-1α activation and respond with a CD1d-mediated cytokine response (65). If and how the iNKT cell response is skewed toward an anti- or pro-inflammatory phenotype in these experiments, and whether different ligands are presented, remains to be determined. Notably, TLR4 signaling enhances atherosclerosis development in ApoE<sup>−</sup>/<sup>−</sup> mice in an iNKT cell-dependent manner (75). TLR4 signaling can be activated by LPS but also by excess free fatty acids, suggesting that nutrient overload mimics infection with regard to its downstream effects. Furthermore, during obesity, adipocytes produce the adipokine leptin to flag nutrient excess and diminish food intake. Leptin contributes to an iNKT cell response that results in anergy and PD-1 upregulation by directly triggering the leptin receptor expressed by iNKT cells (76, 77). Importantly, leptin-mediated iNKT cell activation still requires TCR triggering (77). These findings support the view that changing metabolic conditions determine the ligand pool and steer the iNKT cell response (**Figure 2**).

Besides the sphingolipid ligand pool, metabolic changes may also affect co-stimulatory molecules involved in iNKT activation, including CD40 and CD40L. For example, the amount of surfaceexpressed CD40 on macrophages and smooth muscle cells in human plaques correlates with the stage of atherosclerosis development (78). Possibly, ox-LDL signaling *via* LOX-1, a receptor for ox-LDL, is responsible for the CD40/CD40L upregulation (79). In addition to establishing a firm APC–iNKT cell interaction for Th1 skewing, CD40:CD40L signaling induces LDLR upregulation in human B-cells, enhancing iNKT cell activation (80). Low glucose conditions and AMPK activation causes lower baseline expression of CD40 by dendritic cells, decreased CD40 upregulation, and decreased IL-12 production upon LPS challenge (81).

Figure 2 | Immunometabolic activation of invariant natural killer T (iNKT) cells. Schematic representation of metabolic re-programming of iNKT cell function in immunometabolic disease. Metabolic changes in the antigen-presenting cell may affect the GSL ligand pool or alter the availability of co-stimulatory molecules. AMPK versus mTOR are depicted as the main metabolic regulators of the cellular metabolic program, they can each inhibit the other. AMPK is activated in low glucose conditions and can be activated by adiponectin, the adipokine secreted by lean adipocytes. AMPK activation drives fatty acid oxidation and is associated with cellular longevity. In lean fat- or homeostatic conditions, the iNKT cell response is mainly Th2 skewed (IL-4), but changes to Th1 in metabolic disease (IFN-γ). Whether AMPK or mTOR activation can be linked directly to Th1 or Th2 skewing of the iNKT cell response has not yet been studied to our knowledge. In dyslipidemic conditions, the hallmark of metabolic disease, FFA may activate mTOR *via* TLR4 signaling. The TLR4-driven glycolysis observed in metabolic disease is reminiscent of glycolysis observed under normoxic conditions in cancer (the Warburg effect). mTOR blocks AMPK and in that manner removes at least one of the brakes on CD40 upregulation and IL-12 production, the co-stimulatory requirements for a Th1 iNKT cell response. In late stage metabolic disease, iNKT cells have been described as anergic, while simultaneously upregulating PD-1. Blocking the leptin receptor on iNKT cells may reverse this anergic state. Abbreviations: GSL, glycosphingolipid; FFA, free fatty acids.

Moreover, AMPK inhibition leads to enhanced IL-12 production after LPS stimulation (81). IL-12 contributes to the formation of early atherosclerotic lesions in ApoE−/− mice and correlates positively with pulse wave velocity in healthy individuals, supporting a role for IL-12 in early atherogenesis in humans (82, 83).

#### IMPLICATIONS FOR IMMUNOMETABOLIC DISEASE

Immunometabolic diseases such as obesity, type 2 diabetes, and CVD are increasingly considered to be the downside of co-evolution of the immune system and metabolism (2). The growing body of immunometabolic diseases and the intricate web between metabolic dysregulation and inflammation emphasizes the need to understand metabolic programming of immune cells.

Current treatment for immunometabolic disease often targets dyslipidemia, with statins as mainstream therapy. However, we are learning now that the definition of dyslipidemia should extend far beyond cholesterol and triglyceride ratios, since ceramide and sphingolipid metabolism are closely involved in dyslipidemia and its consequences (17, 18, 39, 46). We are just starting to adequately link sphingolipid metabolism directly to metabolic disease and iNKT cell function, but the findings highlighted in this review indicate that the sphingolipid–iNKT cell axis holds promise for new treatment strategies. iNKT cells are unique in their restriction to lipid antigens and seem to possess all qualities required for immune-modulation. However, it is essential to unravel

#### REFERENCES


the underlying mechanism(s) directing the iNKT cell cytokine response, and to finally identify the endogenous lipid ligands involved. To this end, we may further explore the microbiome for extraction of potential lipid antigens, as iNKT cell–microbiome interaction has been firmly established (84–87). Additionally, we may further investigate naturally occurring iNKT cell subsets that skew toward Th1 or Th2 cytokine production, Th1/Th2 tissue distribution, plasticity, and the role of epigenetic memory (88, 89). Finally, co-stimulatory molecules, cytokines and adipokines involved in iNKT cell activation may be equally important in modulating iNKT cell function in immunometabolic disease.

#### AUTHOR CONTRIBUTIONS

FV, MB, and HS wrote the manuscript. FV and HS created the figures. BL and EK provided critical evaluation and offered insightful suggestions to improve the content. All edited the manuscript and approved the final version for submission.

#### FUNDING

This work was supported by the Wilhelmina Children's Hospital Research Fund, Dutch Topsector Life Sciences and Health TKI fund, and the Nutricia Research Fund. HS was supported by a Fellowship Clinical Research Talent of the University Medical Center Utrecht, and a VENI-NWO Innovational Research Incentive.


glucosylceramide fraction. *Proc Natl Acad Sci U S A* (2014) 111:13433–8. doi:10.1073/pnas.1415357111


production of inflammatory cytokines by bystander cells in vivo. *Int Immunol* (2005) 17:1619–29. doi:10.1093/intimm/dxh342


leptin modulates susceptibility for T cell mediated hepatitis. *J Hepatol* (2014) 60:175–82. doi:10.1016/j.jhep.2013.08.008


**Conflict of Interest Statement:** None of the authors have a competing financial interest in relation to the presented work. BL is employed by Nutricia Research and is leading a strategic alliance between University Medical Centre Utrecht/ Wilhelmina Children's Hospital and Nutricia Research, as indicated by the affiliations.

*Copyright © 2018 Ververs, Kalkhoven, van't Land, Boes and Schipper. 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.*

# Functional Invariant Natural Killer T Cells Secreting Cytokines Are Associated With Non-Progressive Human Immunodeficiency Virus-1 Infection but Not With Suppressive Anti-Retroviral Treatment

*Dharmendra Singh1 , Manisha Ghate2 , Sheela Godbole3 , Smita Kulkarni <sup>4</sup> and Madhuri Thakar <sup>1</sup> \**

*1Department of Immunology, National AIDS Research Institute, Pune, India, 2Department of Clinical Sciences, National AIDS Research Institute, Pune, India, 3Department of Epidemiology and Biostatistics, National AIDS Research Institute, Pune, India, 4Department of Virology, National AIDS Research Institute, Pune, India*

### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Dominic Paquin Proulx, United States Military HIV Research Program, United States Jagannadha K. Sastry, University of Texas MD Anderson Cancer Center, United States*

> *\*Correspondence: Madhuri Thakar mthakar.nari@gov.in, mthakar@nariindia.org*

#### *Specialty section:*

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

*Received: 10 March 2018 Accepted: 08 May 2018 Published: 24 May 2018*

#### *Citation:*

*Singh D, Ghate M, Godbole S, Kulkarni S and Thakar M (2018) Functional Invariant Natural Killer T Cells Secreting Cytokines Are Associated With Non-Progressive Human Immunodeficiency Virus-1 Infection but Not With Suppressive Anti-Retroviral Treatment. Front. Immunol. 9:1152. doi: 10.3389/fimmu.2018.01152*

Background: CD1d restricted invariant natural killer T (iNKT) cells are important in the activation and regulation of immune responses. Limited information is available regarding the functional role of iNKT cells in the human immunodeficiency virus (HIV) disease progression.

Methodology: α-GalCer stimulated iNKT cells were characterized for their functionality in terms of cytokine production (IFN-γ, TNF-α, IL-2, IL-4, and IL-21) and CD107a expression in HIV-1 infected [23 long-term non progressors (LTNPs), 28 progressors, 18 patients before and after suppressive anti-retroviral treatment (ART)] along with 25 HIV-1 negative subjects using multicolor flow cytometry.

Results: The functional profile of α-GalCer stimulated iNKT cells was similar in LTNPs and healthy controls. The number of LTNPs showing functional response in terms of secretion of cytokines (IFN-γ/IL2/TNF-α) and CD107a expression was significantly higher than seen in the progressors. The cytokine secretion by the stimulated iNKT cells was predominantly Th1 type. The frequencies of iNKT cells showing secretion of IFN-γ or IL2 or TNF-α or expression of CD107a were higher in LTNPs (*p* < 0.05 for all) and also significantly associated with lower plasma viral load (*p* value ranged from 0.04 to 0.003) and higher CD4 count (*p* value ranged from 0.02 to <0.0001). The functional profile of the iNKT cells before and after ART did not differ significantly indicating absence of restoration of iNKT cells functionality after suppressive ART. The IL-4 and IL-21 secreting iNKT cells were rare in all study populations.

Conclusion: The presence of functional iNKT cells secreting number of cytokines in nonprogressive HIV infection could be one of the multiple factors required to achieve HIV control and hence have relevance in understanding the immunity in HIV infection. The failure of restoration of the iNKT functionality after ART should be potential area of future research.

Keywords: invariant natural killer T cells, cytokines, CD1d, long-term non progressors, human immunodeficiency virus

#### INTRODUCTION

Understanding the innate and acquired immune mechanism in human immunodeficiency virus (HIV) infection is important in designing strategies for prevention of the HIV infection and also for immune therapies in infected individuals. The invariant natural killer T (iNKT) cells are one of the important innate effector cells which get readily activated either upon stimulation of their TCR by CD1d presented glycolipid antigen, or by cytokines in a TCR-independent manner (1, 2) and rapidly produce an array of regulatory and pro-inflammatory cytokines (3, 4), that can subsequently activate and regulate a variety of innate and adaptive immune cells, such as dendritic cells, natural killer cells, and CD4+ and CD8+ T cells (5, 6). These cells have shown to play a role in cancer (7–9), autoimmune diseases (10, 11), and various infectious condition (12–14).

The role of iNKT cells in viral infection has been emphasized by number of studies in mice and humans. The mice deficient in iNKT cells show increased susceptibility or have impaired immune response to several viruses (15). The herpes viruses shown to manipulate CD1d expression to escape iNKT cell surveillance to establish lifelong latency in humans (16). The chronic HIV infection has shown selective depletion of iNKT cells (17–20). The depletion of iNKT cells is reported to be because of either direct infection by HIV as they expressed both CD4 and CCR5 or due to Fas-mediated activation induced cell death (13).

The chronic immune activation is a hallmark of HIV infection. The iNKT cells are known to influence immune activation. Ibarrondo et al. showed that the loss of CD4+ iNKT cells in gut mucosa of HIV-infected individuals was associated with systemic immune activation (21). The iNKT cells known to work early in the disease course as a bridge between innate and acquired immune responses. In HIV infection these cells showed to recognize the HIV-infected DCs early in HIV infection which are then actively targeted by Nef- and Vpu-dependent viral immune evasion mechanism (22). Various studies have reported variable degree of restoration of functionality of these cells after successful anti-retroviral treatment (ART) (23–28). The level of iNKT cell activation in HIV-infected individuals is associated with disease progression and the frequencies and functionality of iNKT cells are preserved in non-progressive HIV infection, such as HIV-1 infected elite controllers and long-term non progressors (LTNPs) (26, 29, 30). After the exposure to SIV, the AIDS resistant mangabeys also showed higher frequency of iNKT cells as compared to the SIV susceptible macaques (31). Our previous study has shown preserved frequencies of iNKT cells with proliferating capacity and lower expression of exhaustive markers in HIV-1-infected LTNPs (26). Since the iNKT cells carry out multiple functions through secretion of number of cytokines, it is important to understand whether this preserved iNKT cell population is also functionally sound. In this study, we assessed the functional characteristics of the iNKT cells in terms of multiple cytokine secretion after stimulation with α-GalCer in individuals with non-progressive HIV infection (LTNPs) and compared with the iNKT cell functionality in progressive HIV infection and also in the individuals with successful ART.

#### MATERIALS AND METHODS

#### Study Subjects

23 (9M/14F) LTNPs and 28 (12M/16F) progressors were enrolled from the out-patient clinics of the National AIDS Research Institute, Pune. 18 (10M/8F) of the 28 progressors were initiated on ART. These 18 patients were followed up for 12 months post-ART. The ART was initiated when CD4 count was dropped below 350 cells/mm3 ; as per the National criteria of the ART initiation at the time of enrollment. Additionally 25 (13M/12F) HIV-1 seronegative healthy individuals (HCs) were enrolled in the study. The definitions and clinical details of the study population have been reported earlier (26) and the demographic details of the study participants are given in **Table 1**.

20 ml whole blood sample was collected from each study participants. The plasma and peripheral blood mononuclear cells (PBMCs) were separated as previously described (32), and stored at −80°C and −196°C, respectively until tested. The study was carried out in accordance with the institutional ethics committee (NARI Ethics Committee: Registration No: ECR/23/INST/MH/2013/ RR-16). The study (Protocol Number: NARI/EC Protocol No.: 2013-07) was approved by the NARI Ethics committee. All subjects gave written informed consent with the Declaration of Helsinki.

#### CD4 Count and Viral Load Estimation

CD4 T cell counts were quantified by Flow cytometry (FACS Calibur, Becton-Dickinson, CA, USA) using TruCount kit (Becton-Dickinson, CA, USA) as described previously (26), and plasma viral load (pVL) was measured as RNA copies/ml by


*HIV, human immunodeficiency virus; iNKT, invariant natural killer T; ND, not done; IQR, inter quartiles range; HCs, healthy controls; LTNP, long-term non progressors.*

Abbott m2000rt HIV-1 Real-time PCR according to the manufacturer's instructions.

#### Functionality of iNKT Cells

The functional capacity of iNKT cells was assessed for intracellular secretion of multiple cytokines, such as IFN-γ, TNF-α, IL-2, and/or IL-4, and CD107a expression after stimulation with α-GalCer using multicolor flow cytometry. Briefly, the cryopreserved PBMCs were revived and rested overnight in RPMI with 10% fetal bovine serum (FBS), at 37°C with 5% CO2. The next day, the cells were stimulated with 100 ng/ml α-GalCer (Funakoshi Tokyo, Japan) in the presence of anti-CD107a PerCpCy5.5 (BD Biosciences) for 1 h at 37°C in 5% CO2, followed by incubation with Brefeldin A (10 µg/ml; Sigma Aldrich, USA) and GolgiStop (monensin) (1.5 µg/ml; BD Biosciences) for 5 h. Unstimulated cells were also included to assess the background response. The cells were resuspended in 1 ml PBS, washed, and stained with violet amine reactive dye (Invitrogen, Carlsbad, CA, USA) for 30 min at room temperature for differentiation of dead and live cells. The cells were washed and incubated with a cocktail of antibodies; anti-Vα24 PE and anti-Vβ11 FITC (Beckman Coulter, Marseilles, France), anti CD3 PETR (Invitrogen, Carlsbad, CA, USA) for 30 min at room temperature in dark. After washing, the cells were fixed and permeabilized with Perm2 (BD Biosciences, San Jose, CA, USA) according to the manufacturer's instruction and incubated with a mixture of antibodies; anti-IFN-γ PECy7, anti-TNF-α PECy7, anti-IL-2 APC, anti-IL-4 PerCpCy5.5, and anti-IL-21 APC (all from BD Biosciences, San Jose, CA, USA) for 30 min in the dark at room temperature. The cells were washed and stored at 4°C in the dark until acquisition on FACSAria-I (BD Biosciences, USA) and analyzed using FACSDiva software version 6.1.3 (BD Biosciences, USA). The gating strategy is depicted in **Figure 1**.

Since the frequency of iNKT cell was very low in PBMCs, particularly in HIV-1-infected individuals, the study participants having more than 0.04% of iNKT cells were included in the

functional analysis to ensure an accurate assessment of iNKT cell cytokine production and also for measurement of proliferating ability, expression of exhaustion and senescence markers as previously described (26).

To determine cytokine secreting iNKT cells, lymphocytes were first gated on the basis of forward and side-scatter and second gate was set on live lymphocytes using side scatter and violet amine reactive viability dye. Minimum 100,000 events of live CD3+ T lymphocytes were analyzed. A third gate was set on Vα24+Vβ11+ live lymphocyte and percentage of iNKT cells was calculated from CD3+ live lymphocytes. These iNKT cells were further drilled down to gate on cytokines secreting iNKT cells. Single stained controls were used to set compensation parameters and the unstimulated cells were used to set the gate for cytokine secreting cells.

The background response in unstimulated iNKT cells from the study participants was subtracted from the response shown by stimulated cells and the response above the background level was considered as a response to α-GalCer stimulation.

# Expression of CD57 or PD1**+**ve iNKT Cells

The expression of immune exhaustion (PD-1) and senescence (CD57) markers was assessed as described previously (26). Briefly, after revival and resting overnight, PBMCs were resuspended in 1 ml PBS and then stained with violet amine reactive dye (Invitrogen, Carlsbad, CA, USA) for 30 min at room temperature for differentiation of dead and live cells. The samples having more than 90% viability were considered for further analysis. The cells were washed again and incubated with a cocktail of antibodies [anti-Vα24 PE, anti-Vβ11 FITC (Beckman Coulter, Marseilles, France), anti-CD57 APC (Biolegend, USA), anti-CD3 PETR (Invitrogen, Carlsbad, CA, USA), anti-CD3 APC, and anti-PD1 PerCpCy5.5 (BD Biosciences)] for 30 min at room temperature. After washing, the cells were fixed in 3% paraformaldehyde, acquired on FACSAria-I (BD Biosciences, USA) and analyzed using FACSAria-I (BD Biosciences, USA) and analyzed using FACSDiva software version 6.1.3 (BD Biosciences, USA).

#### Assessment of Proliferation Ability of iNKT Cells

The proliferation ability of α-GalCer stimulated iNKT cells was assessed as described previously (26, 28). Briefly, after revival and resting overnight, 1 × 106 PBMCs were incubated in triplicate in RPMI 1640 with 10% FBS, 100 ng/ml α-GalCer (Funakoshi Tokyo, Japan), and 50 IU/ml recombinant human IL-2 (Roche Diagnostics, USA). The medium was replenished at day 3 and 7, and the culture was analyzed for iNKT cells frequency at day 0 and 13 on flow cytometry as described previously (26, 28).

#### Statistical Analysis

GraphPad Prism version 5.01 software was used for statistical analyses. Differences in variables between the study groups were analyzed with Mann–Whitney *U* test and spearman test was used for the correlation analysis. The mean of triplicate experiments for proliferation assessment was considered for the analysis. Changes in the parameter over the time (before and after ART) were analyzed with paired *t*-test. *p* Value of <0.05 was considered as significant.

#### RESULTS

## Cytokine Secretion Profile of iNKT Cells in LTNPs Was Similar to That Seen in Healthy Controls

The ability of α-GalCer stimulated iNKT cells from HIV uninfected and infected individuals to secrete IFN-γ, IL-2, TNF-α, IL-4, or IL-21 was assessed along with their ability to express CD107a as a marker of cytotoxicity using flow cytometry. We observed that the iNKT cells from 24 out of 25 (96%) HCs, 22 of 23 LTNPs (95.65%) and 20 of 28 progressors (71.46%), and 11 of 18 ART-treated (61.11%) individuals responded to α-GalCer stimulation and secreted one or more cytokines.

Heatmaps of iNKT cells secreting cytokines (IFN-γ, IL-2, TNF-α, IL-4, or IL-21) or expressing CD107a among the responders from all study groups (**Figure 2A**: each raw is a single participant) demonstrated that LTNPs and HCs showed similar pattern of α-GalCer stimulated cytokine secretion and CD107a expression. The functional profile in HCs (39/144 observations) and LTNPs (24/132 observations) showed higher magnitude of 3+ and 4+ grade (corresponding to 10–20% and 20–30% of iNKT cells secreting particular cytokine or expressing CD107a, respectively) whereas such a high magnitude was rarely observed in the progressor group (5/120 observations). The IL-4 and IL-21 secreting iNKT cells were rare in all study populations (**Figure 2A**).

The CD107a expression and IFN-γ secretion was found to the most frequent and strong function of iNKT cells from LTNPs and HCs as compared to the progressors. Among the positive responses, the IFN-γ is found to be secreted by the iNKT cells of all HCs (100%), 18/22 LTNPs (81.8%), and 11/20 (55%) progressors. Similarly, among the positive responses the CD107a expressing iNKT cells were observed in 23/24 (95%) HCs, 20/22 LTNPs (90.9%), and 18/20 (90%) progressors. When the magnitude of the functionality (% of stimulated iNKT cells secreting cytokine/s or expressing CD107a) was assessed, the frequencies of IFN-γ secreting or CD107a-expressing iNKT cells were significantly higher in LTNPs as compared to the progressors (*p* < 0.05) but lower than seen in HCs (*p* > 0.05) (**Figure 2B**).

The frequencies of both IFN-γ secretion and CD107a expression were associated with lower pVL and higher CD4 count in HIV-infected individuals (**Figures 2C,F**). Overall, the ability to secrete the cytokines or express CD107a was associated with higher CD4 count [IFN-γ+ iNKT cells (*r* = 0.41; *p* = 0.0016), CD107a+ iNKT cells (*r* = 0.55; *p* < 0.0001), TNF-α+ iNKT cells (*r* = 0.37; *p* = 0.008), IL-4+ iNKT cells (*r* = 0.30; *p* = 0.04), IL-2+ iNKT cells (*r* = 0.40; *p* = 0.007), and IL-21+ iNKT cells (*r* = 0.38; *p* = 0.017)] (**Figures 2C,D**), and lower pVL [IFN-γ+ iNKT cells (*r* = −0.37; *p* = 0.03), for CD107a+ iNKT cells (*r* = −0.37; *p* = 0.01), IL2+ iNKT cells (*r* = −0.39; *p* = 0.04), and for TNFα+ iNKT cells (*r* = −0.38; *p* = 0.03)] (**Figures 2E,F**), whereas no significant association was observed with IL-4+ iNKT cells (*r* = −0.19; *p* = 0.38) or IL-21+ iNKT cells (*r* = −0.24; *p* = 0.1).

#### The Functionality of iNKT Cells Was Not Restored After Suppressive ART

The pattern of α-GalCer stimulated cytokine secretion and CD107a expression was similar in the 18 study participants before and at

FIGURE 2 | Assessment of the cytokines secretion profile of invariant natural killer T (iNKT) cells. (A) Heatmaps of iNKT cells secreting cytokines (IFN-γ, IL-2, TNF-α, IL-4, or IL-21) or expressing CD107a from all study groups [HCs, long-term non progressors (LTNPs), progressors, before, and post anti-retroviral treatment at 12 months]. Only responders are represented in the heatmaps (each row is single participants). For each particular cytokine, the functional response has been graded from 1+ to 4+ according to the magnitude (% of iNKT cells secreting cytokines or expressing CD107a). (B) The bar diagram shows the response (% of iNKT cells showing intracellular cytokine secretion/CD107a expression after stimulation with α-GalCer) on the *Y*-axis in different groups of study participants (HCs, LTNPs, and progressors). (C) The scatter plot shows the correlation analysis for percent of IFN-γ+, IL-2+, and CD107a+ iNKT cells (*Y*-axis) with corresponding CD4 T cell counts on the *X*-axis and (D) plasma viral load (pVL) (RNA copies/ml) on the *X*-axis. (E) The scatter plot shows the correlation analysis for percent of TNF-α+, IL-4+, and IL-21+ iNKT cells (*Y*-axis) with corresponding CD4 T cell counts on the *X*-axis and (F) pVL (RNA copies/ ml) on the *X*-axis.

12 months after ART (**Figure 2A**). The iNKT cells secreting IFN-γ/ IL-2/TNF-α and/or expressing CD107a were rarely observed (only in 5/18 participants) and the magnitude was always of 1+ or 2+ grade which was not changed significantly after suppressive ART. The grade 3+ and 4+ responses were observed only in case of CD107a expression in three participants and the magnitude of response was changed in case of IFN-γ from grade 2+ to 3+ in two participants (**Figure 2A**).

There was no significant increase in cytokine secreting and/ or CD107a expressing iNKT cells at 12 months post ART as compared to the percentages before initiation (by Wilcoxon's signed-rank test) (*p* > 0.05 for all) (**Figure 3**). The mean frequency of cytokine secreting or CD107a expressing iNKT cells at 12 months post ART was still significantly lower than those observed in LTNPs (*p* > 0.05 to all) (**Figure 3**). Only IFN-γ+ iNKT cells were found to be significantly associated with higher

term non progressors and at 12 months post ART.

CD4 counts (*r* = 0.71; *p* = 0.02) at 12 months post ART (data not shown).

#### The Cytokine Secretion by iNKT Cells Was Associated With Higher Proliferating Ability of the iNKT Cells

Earlier we have shown that the iNKT cells from LTNPs have higher proliferating capacity as compared to the iNKT cells from progressors (26). Further, we wanted to assess whether this proliferating ability is associated with the functionality of iNKT cells. As expected, the proliferating ability was significantly associated with the higher frequencies of cytokine secreting and degranulating iNKT cells [IFN-γ+ iNKT cells (*r* = 0.37; *p* < 0.0001), CD107a+ iNKT cells (*r* = 0.58; *p* < 0.0001), TNF-α+ iNKT cells (*r* = 0.37; *p* = 0.028), IL-2+ iNKT cells (*r* = 0.47; *p* = 0.01), and IL-21 + iNKT cells (*r* = 0.55; *p* = 0.003)] (**Figures 4A,B**).

## CD57**+** and PD1 Expression Was Associated With Poor Functionality

In our previous study in the same study population we observed that the progressive HIV infection is associated with higher expression of CD57 and PD1 expressing iNKT cells (26). We observed that the frequencies of both CD57 or PD1 expressing iNKT cells were associated with lower frequencies of cytokine secreting and degranulating iNKT cells [in case of CD57 expressing iNKT cells: for IFN-γ+ iNKT (*r* = −0.47; *p* = 0.001), for CD107a+ iNKT cells (*r*=−0.55; *p*< 0.0001), for IL-2+ iNKT cells (*r*=−0.38; *p*= 0.01), for TNF-α+ iNKT cells (*r* = −0.41; *p* = 0.002) (**Figures 4C,D**),

and in case of PD-1-expressing iNKT cells for IFN-γ+ iNKT (*r* = −0.49; *p* = 0.004), for CD107a+ iNKT cells (*r* = −0.50; *p* < 0.0001), for IL-2+ iNKT cells: *r* = −0.33; *p* = 0.02, for TNF-α + iNKT cells (*r* = −0.44; *p* = 0.001), and for IL-21 + iNKT cells (*r* = −0.35; *p* = 0.03)] (**Figures 4E,F**).

#### DISCUSSION

The present study indicate that the multiple cytokine secreting functional iNKT cells are associated with non-progressive HIV infection and the pattern of cytokine secretion is similar with that seen in the HIV negative healthy controls. We used LTNPs as a model of non-progressive HIV infection to assess the functionality of α-GalCer stimulated iNKT cells in HIV infection. Since iNKT cells can regulate both adaptive and innate immune responses through the rapid production of a vast array of cytokines, we determined the functionality of iNKT cells in terms of secretion of multiple cytokines and expression of CD107a as a marker of cytotoxicity. The heatmap analysis showed similar pattern of cytokine secretion by the iNKT cells from LTNPs and healthy controls after α-GalCer stimulation. This finding supports our previous observation of preserved frequencies of iNKT cells in non-progressive HIV infection (26) and further confirms that the iNKT cells retain the functionality in terms of the multiple cytokine secretion and cytotoxicity in LTNPs. Our observation of higher number of functional iNKT cells secreting multiple cytokines in non-progressive HIV infection contributes to the available information on HIV pathogenesis stating that the functional innate effector response could also be important in virus control in HIV infection. Further, the cytokine response was predominantly Th1 type as the IFN-γ and IL-2 secretion were the most frequently observed functions of α-GalCer stimulated iNKT cells with a weak IL-4 secretion. The mucosal iNKT cells secreting IL-4 (Th2 type) were shown be associated with lower immune activation (29). The differences in the observation might be due to the analysis of iNKT cells from different sites. The Th1 iNKT cells have been shown to have better prognosis in chronic lymphocytic leukemia (33). The observation in this study highlighted the importance of Th1 profile of iNKT cells in chronic infection-like HIV. The predominant secretion of Th1 cytokine could induce potent CD8 and NK cell response resulting in killing of HIV-infected cells. We observed proliferating ability of iNKT cells was associated with higher cytokine secretion whereas the CD57 and PD-1 expressing iNKT cell frequencies were associated with reduced ability to secrete cytokines. Hence although it was an expected phenomenon for iNKT cells, the presence of lower expression of CD57 and PD-1 on iNKT cells with more proliferating and cytokine inducing ability indicates the sound and competent immune response in LTNPs which might be responsible for efficient virus control and halting the disease progression. It might also be possible that the functional iNKT cells help in facilitating the antitumor activity in HIV infection. The PD-1 blockade using anti PD-1 antibodies along with α-GalCer has shown improvement in iNKT cell functions leading to persistent anti-metastasis response in mouse model (34). Similar strategy can be worth exploring in case of HIV infection.

The ART has shown to have a profound benefit in improving the HIV control, quality of life and life expectancy. The early ART initiation has shown better restoration of iNKT cells frequencies (23, 27) and functionality (24). However, in our study the ART has been initiated late in the course of infection (CD4 < 350), it could be the reason for no improvement in the functionality of the iNKT cells even after suppressive ART for 12 months (**Figures 2A** and **3**). Previously we have shown that the ART in the same study population partially restored the quantity of iNKT cells (26). It would be interesting to explore the functionality of iNKT cells in the LTNPs who are initiated on ART in the test and treat era. It is possible that the failure of restoration of iNKT functionality after ART might lead to susceptibility of HIV-associated cancers and other infections. Although the benefits of ART could not undermine the strategies to improve the iNKT functionality with ART would be worth exploring.

#### REFERENCES


In conclusion, our observation of presence of functional iNKT cells secreting IFN-γ, IL-2, and having degranulating ability in non-progressive HIV infection in indicator of importance of sound immune system in achieving HIV control. The restoration of the iNKT functionality after ART could be potential area of future research especially considering the long-term benefits of ART.

#### ETHICS STATEMENT

The study was carried out in accordance with the institutional ethics committee (NARI Ethics Committee: Registration No: ECR/23/INST/MH/2013/RR-16). The study (Protocol Number: NARI/EC Protocol No.: 2013-07) was approved by the NARI Ethics committee. All subjects gave written informed consent with the Declaration of Helsinki.

#### AUTHOR CONTRIBUTIONS

DS designed the protocol and performed all the lab experiments. MG and SG developed the enrollment criteria, sample collection, and follow up the samples. SK has been done viral load testing and data analysis. MT and DS designed the hypothesis and manuscript writing and data analysis. MT has been planned and monitored of all experiments of the study. All authors read and approved the manuscript.

#### ACKNOWLEDGMENTS

The authors would like to thank the study participants and the staff of the Clinic, Immunology (Mr. Amol Kokare and Mrs. Shubhangi Bichare) and Virology laboratory (Mr. Rajkumar londhe and Mrs. Vaishali Chimanpure) for their help. The investigators wish to thank the Indian Council of Medical Research (ICMR), Government of India for supporting DS with Senior Research Fellowship.

# FUNDING

Institutional funding by National AIDS Research Institute, Pune.


**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 Singh, Ghate, Godbole, Kulkarni and Thakar. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner 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.*

# CD1-Restricted T Cells During Persistent virus infections: "Sympathy for the Devil"

*Günther Schönrich\* and Martin J. Raftery*

*Berlin Institute of Health, Institute of Virology, Charité—Universitätsmedizin Berlin, Humboldt-Universität zu Berlin, Berlin, Germany*

Some of the clinically most important viruses persist in the human host after acute infection. In this situation, the host immune system and the viral pathogen attempt to establish an equilibrium. At best, overt disease is avoided. This attempt may fail, however, resulting in eventual loss of viral control or inadequate immune regulation. Consequently, direct virus-induced tissue damage or immunopathology may occur. The cluster of differentiation 1 (CD1) family of non-classical major histocompatibility complex class I molecules are known to present hydrophobic, primarily lipid antigens. There is ample evidence that both CD1-dependent and CD1-independent mechanisms activate CD1-restricted T cells during persistent virus infections. Sophisticated viral mechanisms subvert these immune responses and help the pathogens to avoid clearance from the host organism. CD1-restricted T cells are not only crucial for the antiviral host defense but may also contribute to tissue damage. This review highlights the two edged role of CD1-restricted T cells in persistent virus infections and summarizes the viral immune evasion mechanisms that target these fascinating immune cells.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Johan K. Sandberg, Karolinska Institute (KI), Sweden Randy Brutkiewicz, Indiana University Bloomington, United States*

> *\*Correspondence: Günther Schönrich*

*guenther.schoenrich@charite.de*

#### *Specialty section:*

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

*Received: 18 January 2018 Accepted: 02 March 2018 Published: 19 March 2018*

#### *Citation:*

*Schönrich G and Raftery MJ (2018) CD1-Restricted T Cells During Persistent Virus Infections: "Sympathy for the Devil". Front. Immunol. 9:545. doi: 10.3389/fimmu.2018.00545*

Keywords: human CD1 molecules, antigen presentation, persisting viruses, viral immune evasion, NKT cells

# INTRODUCTION

The majority of virus infections are self-limiting. Immunocompetent hosts often eliminate the invading pathogen without causing permanent tissue damage. Some viruses, however, can resist clearance, and persist in the host organism in the face of intact antiviral immune responses. Persisting viruses belong to different RNA and DNA virus families and represent a global threat to human health (1). As a strategy for persistence, they utilize chronic infection, latency, or both. Chronic infection is characterized by the continuous generation of infectious virus particles as observed during infection with hepatitis B virus (HBV) (2) or hepatitis C virus (HCV) (3). During latent infection, production of viral progeny is put on ice while the viral genome replicates with the host DNA. In the latent stage, viruses are nearly invisible for the host defense but can reinitiate production of infectious viral particles. Primarily human papillomavirus (HPV) (4) and the human herpesviruses (HHVs) (5) establish latency. Human immunodeficiency virus (HIV) establishes both chronic and latent infection (6). The immunological challenges associated with persisting virus infections are distinct from acute self-limiting infections (7).

Virus persistence is largely the result of an evolutionary arms race between the host immune system and viral immune evasion mechanisms (8). For example, persisting viruses have learned Schönrich and Raftery CD1-Restricted T Cells and Persisting Viruses

to subvert the attack of cytotoxic T cells, which recognize viral peptides presented by major histocompatibility complex (MHC) class I molecules (9). Cluster of differentiation 1 (CD1) molecules are structurally related to MHC class I heavy chain molecules and also associate with β2-microglobulin (10–12). They are expressed in mammals, birds and reptiles, although isoforms, antigen-binding sites, recycling motifs and genomic locations are not well conserved (13, 14). In striking contrast to MHC class I molecules, CD1 molecules are non-polymorphic and present hydrophobic, primarily lipid antigens to specialized T cells (15, 16). These CD1-restricted T cells are at the frontline of the human immune response against pathogenic microbes including viruses (17). In this review, we discuss lipid-driven T cell responses during persistent virus infections and the corresponding viral counter measures.

#### CD1-RESTRICTED T CELLS

In humans CD1 molecules are divided into group 1 (CD1a, CD1b, CD1c) and group 2 (CD1d) by sequence homology (18). Group 1 genes can be induced in a coordinated fashion primarily in professional antigen-presenting cells such as dendritic cells (DCs). In contrast, CD1d is constitutively expressed in a wider range of hematopoietic and non- hematopoietic cells (19). CD1 restricted T cells belong to the "unconventional" T cells that do not recognize MHC-bound peptides, and which often show rapid effector functions and orchestrate other immune cells (20–23).

Group 1 CD1-restricted T cells express diverse αβ T-cell receptors (TCRs). They can undergo clonal expansion in the periphery after recognition of stimulatory self-lipids or exogenous lipid antigens derived from bacteria such as *Mycobacterium tuberculosis* (24–26). CD1d-restricted T cells are referred to as natural killer T (NKT) cells because they usually—but not always—express NK1.1 (CD161), a NK cell activating C-type lectin. These cells are further divided into type 1 and type 2 NKT cells.

Type 1 NKT or invariant NKT (iNKT) cells express a semiinvariant αβ TCR, defined by expression of the Vα14-Jα18 TCR in mice and Vα24-Jα18 TCR in humans. They have been analyzed in great detail [recently reviewed in Ref. (17, 27–30)]. The iNKT cells are optimally stimulated by α-galactosylceramide (αGalCer), a glycolipid antigen derived from marine sponge (31). Physiological ligands include cellular and microbial lipids (32). It has been shown that iNKT cells contribute to antiviral responses although the relevant lipid antigens have not yet been defined [recently reviewed in Ref. (33–37)].

Type 2 NKT cells are not stimulated by αGalCer and show much more TCR sequence variability than iNKT cells (38–40). They are more prevalent in humans than iNKT cells, show features of both conventional T cells and iNKT cells and also influence the outcome of infections with persisting viruses (38–41).

#### Group 1 CD1-Restricted T Cells

Group 1 CD1-restricted T cells have been analyzed much less intensively than NKT cells. They can be CD4+, CD8+, or double negative (DN) (42, 43). *In vitro*-studies revealed that after recognition of self or foreign lipid antigens group 1 CD1-restricted T cells become cytoloytic and secrete large amounts of cytokines such as IFN-γ and TNF-α (44–46). These cytokines have a profound antiviral effect and play a pivotal role in controlling persistent virus infections (47). Self-reactive group 1 CD1-restricted T cells are common and the frequency is comparable to alloreactive T cells (estimated to be 1/10 to 1/300 of circulating T cells) (42, 48). They can induce TNF-α dependent DC maturation (49). Intriguingly, self-reactive CD1b-restricted T cells can acquire the phenotype of T helper 17 (Th17) cells and recruit neutrophils (50). The autoreactivity of group 1 CD1-resticted T cells is enhanced by stimulation through pattern recognition receptors (PRRs) (51). It is probable that viruses also trigger activation of self-reactive group 1 CD1-restricted T cells through virus-sensing PRRs and that this contributes to antiviral immunity or virus-induced immunopathogenesis. This notion is supported by the finding that HHV-5, also called human cytomegalovirus (HCMV) can interfere with localization of group 1 CD1 molecules to the cell surface especially CD1b (52).

#### CD1d-Restricted T Cells iNKT Cells

As part of the innate immune system iNKT cells perform multiple effector functions rapidly after being activated. Human iNKT cells comprise phenotypically and functionally distinct subsets: CD4<sup>+</sup>, a small fraction of CD4<sup>−</sup>CD8αα+ and DN cells (53). The CD4<sup>+</sup> subset can produce Th2-type cytokines whereas both the CD4<sup>+</sup> and CD4<sup>−</sup> subsets can secrete Th1-type cytokines and cytotoxic molecules such as perforin or granzyme B (54–56). The iNKT cells can contribute to both protection from pathogenesis and enhancement of disease (57). In mice, functionally polarized subsets of iNKT cells are generated in the thymus. They secrete Th1-, Th2-, or Th17-like cytokines (NKT1, NKT2, and NKT17 subsets) similar to MHC class II-restricted CD4<sup>+</sup> T cells and innate lymphoid cells (58–60). In fact, these subsets and the corresponding MHC-restricted T cell subsets share the same transcription factors that regulate their function (61). Further subsets of iNKT cells that are potentially relevant during viral infections have been defined: follicular helper-like iNKT cells (62), IL-9-producing iNKT cells (63), regulatory (Foxp3<sup>+</sup>) iNKT cells (64, 65), and IL-10-producing iNKT cells called NKT10 cells (66).

The iNKT cells can lyse CD1d<sup>+</sup> target cells, mostly through the interaction between CD95 (Fas) and CD178 (FasL) (67, 68). Virus-infected cells are eliminated by iNKT cells in a CD1ddependent manner (69). In line with this notion, iNKT cells limit the number of Epstein–Barr virus (EBV)-transformed human B cells *in vitro* by a mechanism requiring direct contact with EBVinfected cells (70). EBV, a HHV that infects more than 90% of the human population worldwide, is associated with tumors, such as Burkitt lymphoma, Hodgkin disease, and lymphomas, in immunosuppressed patients (71). The iNKT cell frequency in human tissue is very low (approximately 0.1% in peripheral blood and spleen) (72). Additionally, in contrast to mice iNKT cells are not enriched in the human liver (73, 74). This suggests that human iNKT cells may be more important in helping orchestrate the antiviral immune response rather than in killing virus-infected cells.

There is ample evidence that iNKT cells attract, stimulate, and regulate other innate cells with antiviral effector functions such as NK cells and neutrophils (21, 22). NK cells are found in most compartments of the human organism at a higher frequency than iNKT cells and are critically involved in protection from persisting viruses (75, 76). The iNKT cells transactivate NK cells through the release of IFN-γ in mice (77–80) or IL-2 in humans (81). Moreover, iNKT cells can directly or indirectly recruit and activate neutrophils. These innate cells contribute to both antiviral defense and virus-induced immunopathogenesis (82–85). NK1.1-negative iNKT lymphocytes can directly recruit neutrophils through preferential IL-17 secretion (86, 87). The iNKT cells can also indirectly promote neutrophil responses by interacting with monocyte-derived DCs resulting in prolonged Ca<sup>+</sup> influx and release of inflammatory mediators such as PGE2 (88). In mice infected with murine cytomegalovirus (MCMV), a well-established model of persistent herpesvirus infection, IL-22 secreted by iNKT cells contributes to recruitment of antiviral neutrophils expressing TNF-related apoptosis-inducing ligand (TRAIL) (85). Moreover, iNKT cells can reverse the suppressive phenotype of neutrophils that is induced by acute-phase reactant serum amyloid A-1 (89). *Vice versa*, neutrophils may reduce inflammatory iNKT cell responses by a contact-dependent mechanism thereby contributing to reduction of inflammation (90). These examples illustrate how intensely iNKT cells and other innate immune cells cross-regulate each other.

Another important antiviral function of iNKT cells is the regulation of adaptive immunity (91). For example, iNKT cells help B cells to proliferate and produce high titers of antibodies (92, 93). B cells present bacterial glycolipids through CD1d to iNKT cells which in turn differentiate into follicular helper-like iNKT cells that provide cognate help to B cells (62, 94–96). Recent evidence suggests that iNKT cells are also critical for production of specific antibodies during viral infections despite the absence of exogenous lipid antigens (97, 98). The underlying mechanism, however, is neither based on follicular helper-like iNKT cell differentiation nor CD1d-mediated cognate B cell help by iNKT cells but on tightly regulated early cytokine secretion by iNKT cells at the interfollicular area (97). On the other hand, iNKT cells licensed by a specialized neutrophil subset restrict activation of autoreactive B cells during inflammasome-driven inflammation through FasL (67).

In addition to helping B cells, iNKT cells promote T cell responses directed against intracellular pathogens such as viruses (79, 99–103). This is accomplished by iNKT-mediated DC maturation, which is of central importance for an efficient antiviral host defense (99, 104–106). It has also been reported that activated CD4<sup>−</sup>CD8αα+ iNKT cells curb expansion of antigen-stimulated conventional T cells by CD1d-dependent killing of DCs (107). Thus, iNKT cell subsets regulate adaptive immune responses at different stages.

The importance of crosstalk between iNKT cells and monocytederived DCs is underlined by the observation that both cell types are programmed to migrate to inflamed peripheral tissue (55, 108–110). Many persistent viruses such as herpes simplex virus type 1 (HSV-1), HSV-2 and HCMV target monocytederived DCs to induce apoptosis or impair antigen presentation through MHC class I molecules (111–113). As a consequence antiviral CD8<sup>+</sup> T cells have to be activated by uninfected bystander DCs that take up apoptotic debris containing viral antigens and cross-present it through MHC class I molecules (114, 115). This cross-presentation, also called cross-priming, requires pathogenassociated molecular patterns (PAMPs) and/or specialized subsets of Th cells that mature or "license" the cross-presenting DCs (116). It has been shown that uninfected DCs can be licensed for cross-priming by iNKT cells (117). On the other hand, iNKT cells suppress CD8<sup>+</sup> T cell responses to viral proteins expressed in skin epithelial cells by blocking cross-priming in the draining lymph nodes (118). This suppressive iNKT cell effect may increase the persistence of viruses such as HPV and HSV-1 in the skin. The effect of iNKT cells on CD8<sup>+</sup> T cell priming (activating versus inhibitory) may be dependent on the type of antigen and the requirement for CD4<sup>+</sup> T cell helper epitopes (118).

A recent study has shown that iNKT cells release the brake on the adaptive immune response against influenza virus by interfering with myeloid-derived suppressor cells (MDSCs) (119). MDSCs are in fact a heterogeneous population of myeloid progenitor cells including immature DCs, immature macrophages, and granulocytes (120). The iNKT cells abolish the suppressive activity of MDSCs in a crosstalk that requires CD1d and CD40. The finding by De Santo et al. is important as MDSCs facilitate the development of persistent virus infections by impairing antiviral functions of T cells, NK cells, and APCs (121).

#### Type 2 NKT Cells

Evidence that type 2 NKT cells play a protective role in virus infections came from a study that compared virus-induced disease severity between wild-type mice, Jα18 KO mice (lacking only type 1 NKT cells) and CD1d KO mice (lacking both type 1 and type 2 NKT cells) (122). In comparison to iNKT cells little is known about the function of type 2 NKT cells due to the lack of suitable animal models and the difficulty in tracking this poorly defined cell type *in vivo* (41). However, there is evidence that type 2 NKT cell subsets with distinct functional profiles exist (41, 123).

#### CLINICAL OBSERVATIONS

A number of clinical observations suggest involvement of CD1 restricted T cells in either the control of persisting viruses or virus-induced immunopathogenesis.

# Group 1 CD1-Restricted T Cells

Group 1 CD1-restricted T cells from patients with active tuberculosis expand after reexposure to cognate antigen similar to adaptive MHC-restricted T cells (124). In HIV-infected patients, CD1b-restricted T cells recognizing mycobacterial glycolipids are strongly reduced (125). The reduced frequency of mycobacteriaspecific CD1b-restricted T cells may contribute to the increased incidence of tuberculosis in this group. The kinetics of group 1 CD1-restricted T cells stimulated by virus-induced self-lipids is not understood. Self-reactive group 1 CD1-restricted T cells show an adaptive-like population dynamics (42). It is possible but has not yet been proven that group 1 CD1-restricted T cell reacting to virus-induced stimulatory self-lipids expand in a similar way (**Figure 1A**). On the other hand, self-reactive CD1b-restricted T cells have been described that are more like innate T cells (126). These cells may contribute to early antiviral host defense similar to iNKT cells.

## CD1d-Restricted T Cells

Patients with defective iNKT cell responses suffer from severe infection with HHVs (127). For example, boys with X-linked lymphoproliferative disorder have severely impaired iNKT cell development and T cell function that results in life-threatening lymphoproliferation after primary infection with EBV or Kaposi's sarcoma-associated herpesvirus (KSHV) (128–131). A similar disease was observed in girls with a homozygous mutation in the IL-2-inducible T cell kinase (*ITK*) gene on chromosome 5q31–5q32 (132). Moreover, patients with profound iNKT cell deficiency develop severe VZV-associated disease after vaccination with the live attenuated varicella vaccine (133, 134). Hermansky–Pudlak syndrome type 2, an autosomal recessive disease caused by mutations of the AP3B1 gene, encoding for the beta3A subunit of AP-3, also show reduced numbers of iNKT cells (135). Intriguingly, iNKT cell numbers decrease regardless of the HIV-status in patients with multicentric Castleman disease (MCD), a rare polyclonal lymphoproliferative disorder that is associated with KSHV (136). This observation suggests that a deficiency in iNKT cells contributes to the pathogenesis of KHSV-associated MCD. Although deficient iNKT cells can be

Figure 1 | Dynamics of CD1-restricted T cells after persistent virus infections. (A) Group 1 CD1-restricted T cells are thought to undergo clonal expansion in response to an acute infection followed by a contraction phase leaving an increased population of memory cells. (B) The frequency of invariant NKT (iNKT) cells in the blood is maintained at a stable level over time in individuals infected with human herpesviruses (HHVs) or human papillomavirus (HPV). Human immunodeficiency virus (HIV) infection results in a loss of iNKT cells and a decrease in functionality. Hepatitis B virus (HBV) and hepatitis C virus (HCV) may result in a decrease in iNKT cell frequency due to redistribution into the liver.

regarded as the common denominator of these primary immundeficiencies it has to be kept in mind that other immune cells are affected as well.

In humans with an intact immune system, distinct dynamics of circulating iNKT cells are observed after infection with different persisting viruses. The frequency of peripheral blood iNKT cells after infection with persisting viruses that establish latent infection such as HHVs and HPV remain unaltered (**Figure 1B**) (98). In contrast, it has been reported that the frequencies of circulating iNKT cells are significantly decreased during chronic HBV and HCV infection and recover during antiviral therapy (**Figure 1B**) (137, 138). This decrease may be due to trafficking of iNKT cells to the liver as intrahepatic enrichment of iNKT cells is observed in patients with chronic viral hepatitis (137). In another study, however, iNKT cell frequencies between patients with chronic viral hepatitis and healthy individuals were not different (139). CD1d is strongly upregulated on hepatocytes and other liver cells during chronic viral hepatitis (137, 140). In fact, CD1d-restricted T cells contribute to virus-induced liver injury by killing hepatocytes and production of pro-inflammatory cytokines, which promote liver fibrosis and cirrhosis (74, 137, 141–144).

Several studies have shown that iNKT cells are depleted during HIV infection (145–152) (**Figure 1B**). The remaining iNKT cells display an exhausted phenotype with reduced functionality that may improve with early antiretroviral therapy (152–157). In fact, the level of iNKT cell activation in HIV-infected patients is associated with disease progression markers and the number and functionality of iNKT cells are preserved in non-progressive HIV infection (155, 158). Intriguingly, both CD4<sup>+</sup> and CD4<sup>−</sup> iNKT cells are depleted in HIV-infected individuals (149) although CD4<sup>−</sup> iNKT cells are resistant to HIV infection (148, 150). This could be due to apoptosis of uninfected bystander cells. Triggering of both extrinsic and intrinsic death pathways in T cells due to systemic immune activation has been described in many studies (159).

## CD1-RESTRICTED T CELLS IN ANIMAL MODELS OF PERSISTENT VIRUS INFECTIONS

#### Group 1 CD1-Restricted T Cells

The function of group 1 CD1-restricted T cells has proven difficult to address *in vivo* as laboratory mice do not express group 1 CD1 molecules (160–162). In contrast, mice with a humanized immune system express human group 1 CD1 molecules (163) and open up a new avenue for studying human immune responses to viral pathogens (164). Similarly, human group 1 CD1 transgenic mice have recently been developed (126, 165, 166). Future studies have to investigate a possible role of group 1 CD1-restricted T cells in persisting virus infections by using suitable animal models and tetramers.

#### CD1d-Restricted T Cells

The protective and pathogenic role of CD1d-restricted T cells has been investigated in several animal models of persistent virus infections (see **Table 1**).



*Ad-HBV, HBV-expressing adenoviral particles; HBV, hepatitis B virus; HSV-1, herpes simplex virus type 1; HSV-2, herpes simplex virus type 2; MCMV, mouse cytomegalovirus; SIV, simian immunodeficiency viruses; NKT, natural killer T; iNKT, invariant NKT.*

#### Herpesviral Infection

Studies in CD1d KO and iNKT cell-deficient (Jα18 KO) mice indicate that CD1d-restricted T cells have a protective role after infection with HSV-1 (98, 167, 168) or HSV-2 (169). In addition, intravaginal pretreatment of C57BL/6 mice with αGalCer lowers HSV-2 disease and inhibits viral replication (170). One report did not confirm an antiviral role for NKT cells, most likely because mice were infected with a less virulent HSV-1 strain (171).

In mice infected with MCMV, activation of iNKT cells by exogenous αGalCer reduces viral replication (172). Furthermore, it has been shown in the MCMV model that iNKT cells can protect from MCMV-induced myelosuppression (173). In C57BL/6 mice NK cells can compensate for iNKT deficiency (172, 174, 175) whereas in BALB/c mice NK cells are less well activated by MCMV and thus iNKT deficiency results in higher viral loads (176).

#### Chronic Viral Hepatitis

In contrast to HHVs, HBV, and HCV infect target cells without causing cytopathic effects. Injection of αGalCer into HBV transgenic mice activates intrahepatic resident iNKT cells and NK cells resulting in noncytopathic control of HBV replication through secretion of type I and II IFN (80, 177). Hepatocytes directly control iNKT cell homeostasis through modulating the balance between activating and non-activating lipids presented by CD1d molecules (178). Accordingly, iNKT cells act as an early warning system for HBV infection, which profoundly alters lipid metabolism (179). For example, type 2 NKT cells recognize CD1d-bound lysophospholipids, antigenic self-lipids that are generated in hepatocytes by a HBV-induced secretory phospholipase (180). The resulting activation of type 2 NKT cells also leads to IL-12 mediated transactivation of iNKT cells and plays a pivotal role in virus control (180). However, type 2 NKT cells can also cause liver injury in HBV transgenic mice (181, 182). It has to be kept in mind that there are major differences between human and murine liver NKT cells when extrapolating these results to humans (73, 74). Nevertheless, human NKT cell lines are stimulated in a CD1d-dependent manner by human hepatocytes (180). Taken together, these findings support clinical observations suggesting a role for CD1d-restricted T cells in antiviral immunity and virus-induced immunopathogenesis in human liver.

#### HIV Infection

In a humanized mouse model of HIV-1 infection activation of type 2 NKT cells inhibits viral replication and prevents virus-induced pancytopenia (183). In Asian macaques, which develop acquired immunodeficiency syndrome (AIDS) during persistent simian immunodeficiency virus (SIV) infection, NKT cell depletion inversely correlates with viral load (184). Comparative studies of Asian macaques and sooty mangabeys, which serve as a natural host and do not develop AIDS during persistent SIV infection, suggest that NKT cells protect from SIV-induced immune activation and immunodeficiency (185).

#### ACTIVATION OF CD1-RESTRICTED T CELLS

In contrast to other microbes such as bacteria, viruses have not been investigated with regard to CD1 ligands, although viral infection can stimulate CD1-restricted antiviral T cells. Activation of CD1-restricted T cells during viral infection could be triggered by CD1 molecules presenting antigenic self-lipids. In addition, CD1-independent mechanisms that activate CD1-restricted T cells such as cytokine release have been described.

#### Group 1 CD1-Restricted T Cells

Autoreactive group 1 CD1-restricted T cells are stimulated by DCs that express group 1 CD1 molecules and present increased amounts of antigenic self-lipids upon activation by bacterial PAMPs through TLRs (186, 187). It is very likely but has not yet been shown that group 1 CD1-restricted autoreactive T cells are activated in a similar fashion during viral infections upon activation of DCs through virus-sensing PRRs.

## CD1d-Restricted T Cells CD1d-Dependent Activation

The limited repertoire of microbial and endogenous ligands presented by CD1d is expanding (188–190). However, the identity of self-lipids presented during viral infection is still enigmatic. Stimulatory self-lipids could be channeled into the CD1d-restricted antigen presentation pathway by several distinct mechanisms. First, there is ample evidence that TLR stimulation increases CD1d presentation of antigenic self-lipids resulting in iNKT cell activation (191–197). Besides modulating the repertoire of self-lipids, viruses also increase CD1 surface expression through triggering PRRs (198). Second, viral invasion and inflammation is associated with pathological hypoxia that activates hypoxia-inducible factor (HIF), a "master regulator" of adaptive immune responses (199, 200). HIF alters lipid metabolism in such a way that self-reactive NKT cells are activated (201, 202). Third, viruses profoundly rewire host lipid metabolism and remodel lipid distribution to boost in a coordinated fashion viral entry, replication, assembly, and egress (203–208). Disturbance of the normal lipid trafficking patterns within the cell due to enveloped virus production (HHVs, HBV, HCV) or even naked virus (HPV) will allow access of otherwise sequestered lipids or altered lipids to the relevant CD1 presenting compartment. Furthermore, after assembly and release from the infected host cell the envelope of the newly built viral particle may carry the ligand with it and deliver it during infection to another host cell. Taken together, several distinct mechanisms facilitate CD1ddependent iNKT cell activation during different stages of the viral life cycle.

Of note, viral immune evasion mechanisms targeting other immune components may alert iNKT cells. For example, MHC class I downregulation by virus-encoded immunevasins induces CD1d upregulation and enhances NKT cell activation (209). This finding is supported by other observations. First, low pH stripping of MHC class I molecules augments CD1d surface expression and activates iNKT cells (210). Second, APCs from transporter associated with antigen presentation 1 (TAP1) deficient mice are defective in MHC class I antigen presentation but show an increased capacity to stimulate iNKT cells (210, 211). This reverse correlation between CD1d and MHC class I surface expression may enable iNKT cells to detect the loss of MHC class I molecules, a situation called "missing self " (212). The underlying mechanism is not yet fully understood but may involve masking of CD1d by MHC class I molecules on the cell surface (210). Increased iNKT cell activation during persistent virus infections may also result from viral blockers inhibiting the autophagic machinery, which downregulates iNKT cell responses through CD1d internalization (213, 214).

#### CD1d-Independent Activation

CD1d-restricted T cells can also be stimulated independently of the TCR. First, there is ample evidence that pro-inflammatory cytokines (IL-12, IL-18, or type I IFN) released from APCs after stimulation through PRRs activate iNKT cells predominantly in a CD1d-independent manner (215, 216). It may be that an initial CD1d-mediated TCR signal is still required (217). CD1d-independent iNKT cell activation was observed in mice infected with MCMV, and was driven by type I IFN and IL-12 after DC stimulation (176, 218–220). The factors that determine whether pro-inflammatory cytokines are sufficient for activation of antiviral iNKT cells are unclear.

Second, engagement of NKG2D on NKT cells results in CD1d-independent activation and subsequent killing of ligandexpressing target cells (182, 221). NKG2D is an activating receptor expressed on CD4<sup>−</sup> iNKT cells and is triggered by stress ligands. The latter are upregulated during viral infections after stimulation of virus-sensing PRRs such as RIG-I and MDA-5 (222, 223). In addition, NKG2D can also co-stimulate TCR-mediated activation of CD4<sup>−</sup> iNKT cells (221).

Third, iNKT cells can be activated by apoptotic cells through TIM-1, a member of the T-cell immunoglobulin mucin (TIM) family of cell surface proteins (224). TIM-1 is constitutively expressed by NKT cells and serves as a receptor for phosphatidylserine, an important marker of cells undergoing apoptosis (225). This mode of iNKT cell activation may be relevant during persistent virus infections because viruses frequently drive their host cells into apoptosis (226).

# VIRAL EVASION OF CD1-RESTRICTED T CELLS

During the evolutionary arms race with the host immune system persisting viruses have developed multilayered defense strategies to interfere with antiviral CD1-restricted T cells (17, 33, 35, 36, 227). They target the CD1 antigen presentation machinery either on the transcriptional, posttranscriptional, or posttranslational level.

# Group 1 CD1-Restricted T Cells

The group 1 CD1 molecules are expressed on professional APCs in response to certain cytokines such as granulocyte-macrophage colony-stimulating factor (228). The anti-inflammatory cytokine IL-10 prevents upregulation of group 1 CD1 molecules on professional APCs such as monocyte-derived DCs (229–231). In fact, the activity of IL-10 is crucial for establishing viral persistence (232). Large DNA viruses such as herpesviruses have acquired numerous genes from their host including viral homologs of IL-10 (vIL-10) to subvert the host immune response (233, 234). For example, vIL-10 from HCMV and EBV not only prevent upregulation of MHC class I/II molecules but also group 1 CD1 molecules (52, 235). Recent reports suggest that vIL-10 strongly induces the expression of its cellular counterpart in monocytederived cells thereby potentiating its immunomodulatory effect (236, 237). In addition, viral proteins with no homology to cellular proteins such as HCMV-encoded pUL11 or EBV-encoded latent membrane protein 1 highjack cellular signaling pathways to induce expression of cellular IL-10 (238, 239).

Group 1 CD1 molecules are downregulated from the cell surface during the early phase of HCMV infection with CD1b being especially sensitive to this effect (52). This block is not performed by known HCMV-encoded MHC class I-blocking molecules and results in the intracellular accumulation of group 1 CD1 molecules (52). The underlying mechanism has not yet been described. In addition, HIV-1 negative factor (Nef) downregulates CD1a after infection of immature DCs with HIV-1 and impairs stimulation of CD1a-restricted T cells (240, 241). In order to modulate host membrane trafficking pathways HIV-1 Nef has to interact with various host proteins through distinct motifs (242). Nef-mediated CD1a downregulation might require the interaction of Nef with hematopoietic cell kinase and p21-activated kinase 2, which are both expressed in immature DCs (241). Collectively, these findings support the idea that group 1 CD1 molecules contribute to lipid-driven antiviral immune responses.

# CD1d-Restricted T Cells

#### Viral Downregulation of CD1d Gene Transcription

In contrast to group 1 CD1 molecules, CD1d is constitutively expressed not only in myeloid cells such as B cells, macrophages, and DCs and but also in epithelial tissue. Latent EBV infection that is associated with transformation of B cells results in complete shutdown of CD1d mRNA expression and the lack of iNKT cell activation (70). This is due to increased binding of lymphoid enhancer-binding factor 1 to the distal region of the CD1d promotor (**Figure 2**). Moreover, CD1d transcription is downregulated during severe primary HCMV infection (243).

#### Virus-Induced Degradation of CD1d mRNA

Replicating α- and γ-herpesviruses induce global mRNA degradation to shut off host protein synthesis and reallocate cellular resources to their own need by different mechanisms (244). It has been shown that EBV-encoded shutoff protein BGLF5, a lytic phase protein, downregulates multiple immune components including CD1d (**Figure 2**) (245, 246).

#### Viral Interference With Trafficking of CD1d Molecules

After translation, CD1 heavy chains bind to β2-microglobulin and are loaded with self-lipids in the endoplasmic reticulum (ER). This trimeric complex travels then to the cell surface through the secretory pathway. Subsequently, a tyrosine motif within their cytoplasmic tail enables human CD1 molecules to bind the μ-subunit of adaptor protein complex (AP)-2 (CD1b-d) and AP-3 (CD1b and murine CD1d). In contrast, CD1a lacks AP-2 and AP-3 sorting motifs. Because of these differences CD1 molecules traffic through distinct intracellular compartments for acquiring lipids and present these lipids to CD1-restricted T cells (247–249). Persistent viruses utilize various posttranslational mechanisms to interfere with antigen presentation through CD1d molecules in different cellular compartments.

Figure 2 | Evasion of CD1d antigen presentation. Persisting viruses evade invariant NKT (iNKT) cell activation by interfering with CD1d biosynthesis and CD1d trafficking. (1) CD1d gene transcription is downregulated by increased binding of Epstein–Barr virus (EBV)-encoded lymphoid enhancer-binding factor 1 (LEF-1) to the distal region of the CD1d promotor. (2) Viral host shutoff factors such as EBV BGLF5, an early lytic phase protein, inhibit protein synthesis by degrading mRNA. (3) Human papillomavirus (HPV) E5 translocates CD1d into the cytosol for proteasomal degradation. (4) EBV gp150 provides an abundantly sialylated glycan shield for CD1d on the cell surface preventing recognition by iNKT cells. (5) Herpes simplex virus type 1 (HSV-1) glycoprotein B/US3 and human immunodeficiency virus (HIV) negative factor (Nef) redirect CD1d to the trans-Golgi network (TGN). (6) In contrast, HIV Nef retains CD1d in the early endosome (EE). (7) Kaposi's sarcomaassociated herpesvirus (KSHV) K3 redirects CD1d from the late endosome (LE) to the lysosome.

#### Virus-Induced Proteasomal Degradation of CD1d in the Cytosol

HPV subverts immune responses through expression of E5 (250). Clinical samples of HPV-infected cervical epithelium show decreased CD1d expression (251). Further analysis of transfected cell lines revealed that HPV E5 interacts with calnexin, a chaperone involved in CD1d folding in the ER, and finally targets CD1d to the cytosolic proteolytic pathway (**Figure 2**) (251).

#### Viral Shielding of Surface CD1d Molecules

A novel broadly acting mechanism of herpesviral subversion of antigen presentation has been discovered recently (252). EBVencoded gp150, which is heavily glycosylated and expressed in the late phase of EBV replication, inhibits antigen presentation by MHC class I and II as well as CD1d molecules. The EBV-encoded gp150 does not interfere with recycling of CD1d molecules but prevents their detection by iNKT cells on the cell surface through an abundantly sialylated glycan shield (**Figure 2**) (252). Of interest, the Ebola virus glycoprotein masks MHC class I molecules by a similar mechanism but an effect on CD1d has not yet been tested (253). Further studies have to elucidate whether other herpesviruses have developed are similar strategy to mask antigen-presenting molecules on the cell surface.

#### Viral Modulation of CD1d Recycling

Persisting viruses can also subvert CD1 antigen presentation by interfering with CD1 recycling. For example, CD1d surface expression is downregulated during productive infection with HHVs such as HSV-1 (209, 254) and KSHV (255). This effect was dependent on the viral titer used for infection of CD1dexpressing cells (209).

Further attempts identified several viral proteins interfering with CD1d recycling. For example, KSHV-encoded modulator of immune recognition proteins 1 and 2 (also known as K3 and K5) function as membrane-bound E3-ubiquitin ligases and reduce expression of a number of immunologically important surface molecules including MHC class I and CD1d (256). In fact, K3 and K5 have been pilfered from the host genome and belong to the *M*embrane *A*ssociated *R*ING-*CH* family of E3 ligase (MARCH) proteins. They reroute both MHC class I and CD1d to the lysosomal compartment by ubiquitinylating a unique lysine residue on the cytoplasmic tail (**Figure 2**) (255). In contrast to MHC class I, however, the CD1d complexes are resistant to lysosomal degradation resulting in reduced CD1d levels on KSHV-infected B cells although the total cellular amount of CD1d is not altered (255).

A recent study shows that CD1d molecules on the APC surface build nanoclusters and that these structures are important for iNKT cell activation (257). Intriguingly, the actin cytoskeleton prevents CD1d nanoclustering (257). This finding is in accordance with a previous study showing that disruption of actin filaments increases CD1d antigen presentation (258). It may also explain why HSV-1 lacking VP22 (UL49), a viral tegument protein that stabilizes microtubules, lacks the ability to inhibit CD1d recycling (259). Although VP22 is necessary for HSV-1 induced CD1d downregulation, other viral proteins are additionally required. The type II kinesin motor protein KIF3A, which transports proteins along the microtubule network, is necessary for CD1d surface expression (260). Of relevance, KIF3A is phosphorylated by US3, a HSV-1 encoded protein that induces downregulation of CD1d by suppressing its recycling (261). Another HSV-1 protein, glycoprotein B (gB), binds to CD1d within the ER and remains stably associated throughout CD1d trafficking (261). Both US3 and gB seem to be required for relocalization of CD1d to the trans-Golgi network (TGN) thereby reducing CD1d surface expression and iNKT cell activation (**Figure 2**) (261). In contrast, a previous study has reported that CD1d is trapped in lysome-like structures during HSV-1 infection (254). Similarly, another study reported that CD1d molecules are degraded in the lysosomes after HSV-1 induced phosphorylation of a dual residue motif (T322/S323) in the cytoplasmic tail of CD1d (262, 263). However, under normal conditions CD1d complexes are resistant to lysosomal degradation (255, 264).

HIV-1 encodes two proteins, Nef and viral protein U (Vpu) which target numerous immunologically important surface molecules including CD1d (265). In HIV-1 infected DCs, Vpu neither induces endocytosis nor rapid degradation but suppresses CD1d recycling by retaining CD1d in the early endosome (196, 266, 267). In contrast, Nef increases CD1d internalization and re-localizes CD1d molecules to the TGN (196, 268, 269). Thus, Vpu and Nef interfere with iNKT cell activation by complementary mechanisms (**Figure 2**). As a result, the capacity of HIV-1 infected DCs to stimulate iNKT cells is impaired (196).

The unique short (US) region of HCMV encodes several proteins including US2 and US11 that interfere with antigen presentation through MHC class I molecules (270). US2 and US11 induce rapid translocation of MHC class I heavy chains from the ER into the cytosol and subsequent degradation by the proteasome (271, 272). US2 and US11 also physically interact with CD1d but do not downregulate CD1d surface expression (273, 274). However, reduced activation of iNKT cells after stimulation with αGalCer-pulsed US2-expressing APCs has been reported but the underlying mechanism is unclear (274).

#### Viral Disruption of the iNKT Cell–DC Axis

The bidirectional interaction between DCs and iNKT cells is crucial for an efficient antiviral immune response (28, 275). During infection triggering of a combination of PRRs stimulates IL-12 release from DCs and increases the presentation of antigenic self-lipids by CD1d (191, 195, 197). In fact, signaling through PRRs inhibits degradation of glyosylceramides in the lysosome thereby increasing the amount of lipids that stimulate iNKT cells (193, 276). Once activated iNKT cells further enhance IL-12 production by DCs through CD40–CD40 ligand stimulation. This amplification loop then results in NK cell transactivation and increased responses of MHC-restricted CD4<sup>+</sup> as well as CD8<sup>+</sup> T cell responses. Persisting viruses such as HHVs use several different strategies to blunt DC function (277). For example, VZV interferes with TLR signaling in DCs thereby decreasing IL-12 production (278). Moreover, herpesviral IL-10 homologs can decrease the ability of uninfected DCs to secrete IL-12 (234). Virus-infected DCs that do not produce biologically active IL-12 fail to activate the antiviral functions of iNKT cells (279). Moreover, HCMV-infected DCs upregulate on the surface death ligands such as TRAIL that kill T cells (113). Taken together, it is likely that multilayered viral countermeasures severely impair activation and function of iNKT cells thereby disrupting the important iNKT cell–DC axis.

#### Viral Interference With CD1d-Independent Activation of CD1d-Restricted T Cells

Persistent viruses such as HHVs also interfere with NKG2D ligand upregulation on virus-infected cells thereby reducing the likelihood of NKG2D-mediated activation of antiviral iNKT and type 2 NKT cells (280, 281). Moreover, HSV-1-infected keratinocytes block cytokine-dependent activation of iNKT cells (282).

#### Virus-Induced Functional Impairment of CD1d-Restricted T Cells

Persisting viruses not only target CD1 antigen presentation and interfere with activation of CD1-restricted T cells but can also induce a state of unresponsiveness in CD1-restricted T cells. It has been demonstrated *in vitro* that HSV-1-infected keratinocytes impair TCR signaling in iNKT cells in a contact-dependent manner (282). In contrast to DCs, CD1d is not downregulated on keratinocytes after infection with HSV-1. This finding nicely illustrates that depending on the infected cell type one and the same virus use different strategies to evade CD1d-restricted T cell responses. Moreover, iNKT cells from patients with KSHVassociated MCD, a severe B-cell lymphoproliferative disorder, show impaired proliferation when stimulated with αGalCer (136). This hyporeactive state is indicative of virus-induced iNKT cell anergy or exhaustion. In accordance, induction of anergy in iNKT cells has been described in many reports and is required for prevention of uncontrolled inflammation and tissue destruction (283). On the other hand, it also allows viruses to replicate and spread more efficiently. The underlying mechanisms of virus-induced iNKT cell anergy are unknown but may involve viral modulation of costimulatory molecules and other signaling molecules on APCs (284).

#### REFERENCES


Recently, a regulatory iNKT cell subset called NKT10 has been described that bear similarities to Tregs (66). Tregs suppress protective immune responses thereby supporting virus persistence and at the same time also reducing the inflammation-mediated tissue damage (285). The NKT10 cells occur naturally in mice and humans and expand after stimulation (66). It would be interesting to investigate whether persisting viruses increase the frequency and activity of NKT10 cells to curb antiviral immune responses.

# CONCLUSION AND FUTURE DIRECTIONS

The risk for the host in terms of severe disease from a persistent virus infection ranges from low in case of well-adapted viruses such as the HHVs to ineluctable in case of a recent émigré such as HIV-1. CD1-reactive cells play a significant role in this process. Although the role of self-reactive NKT cells in antiviral immunity and immunopathogenesis has been analyzed in the clinical setting and in experimental models, it is unclear how group 1 CD1 restricted T cells contribute to antiviral immunity. Future efforts should define in detail the antigenic self-lipids that are induced and loaded on CD1 molecules during viral infections and how they program antiviral immunity. This knowledge can then be harnessed to develop novel vaccines and adjuvants for protection from persistent virus infection.

#### AUTHOR CONTRIBUTIONS

Both authors contributed to the conception, writing, and critical revising of this review.

#### FUNDING

The authors acknowledge support from the German Research Foundation (DFG) and the Open Access Publication Fund of Charité—Universitätsmedizin Berlin.


activation of Valpha14+ NKT cells promotes the breakage of CTL tolerance. *Int Immunol* (2008) 20(7):869–79. doi:10.1093/intimm/dxn046


chronic hepatitis B. *PLoS One* (2011) 6(12):e28871. doi:10.1371/journal. pone.0028871


molecules, not only class I MHC but also CD1a, in immature dendritic cells. *Virology* (2004) 326(1):79–89. doi:10.1016/j.virol.2004.06.004


**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 Schönrich and Raftery. 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.*

# CD1d-Restricted Type ii nKT Cells Reactive with endogenous Hydrophobic Peptides

#### *Yusuke Nishioka1 , Sakiko Masuda2 , Utano Tomaru3 and Akihiro Ishizu2 \**

*1Graduate School of Health Sciences, Hokkaido University, Sapporo, Japan, 2 Faculty of Health Sciences, Hokkaido University, Sapporo, Japan, 3Department of Pathology, Faculty of Medicine, Graduate School of Medicine, Hokkaido University, Sapporo, Japan*

NKT cells belong to a distinct subset of T cells that recognize hydrophobic antigens presented by major histocompatibility complex class I-like molecules, such as CD1d. Because NKT cells stimulated by antigens can activate or suppress other immunocompetent cells through an immediate production of a large amount of cytokines, they are regarded as immunological modulators. CD1d-restricted NKT cells are classified into two subsets, namely, type I and type II. CD1d-restricted type I NKT cells express invariant T cell receptors (TCRs) and react with lipid antigens, including the marine sponge-derived glycolipid α-galactosylceramide. On the contrary, CD1d-restricted type II NKT cells recognize a wide variety of antigens, including glycolipids, phospholipids, and hydrophobic peptides, by their diverse TCRs. In this review, we focus particularly on CD1d-restricted type II NKT cells that recognize endogenous hydrophobic peptides presented by CD1d. Previous studies have demonstrated that CD1d-restricted type I NKT cells usually act as pro-inflammatory cells but sometimes behave as anti-inflammatory cells. It has been also demonstrated that CD1d-restricted type II NKT cells play opposite roles to CD1d-restricted type I NKT cells; thus, they function as anti-inflammatory or pro-inflammatory cells depending on the situation. In line with this, CD1d-restricted type II NKT cells that recognize type II collagen peptide have been demonstrated to act as anti-inflammatory cells in diverse inflammation-induction models in mice, whereas pro-inflammatory CD1d-restricted type II NKT cells reactive with sterol carrier protein 2 peptide have been demonstrated to be involved in the development of small vessel vasculitis in rats.

Keywords: CD1d, NKT cell, hydrophobic peptide, sulfatide, glycolipid

#### INTRODUCTION

NKT cells are first reported as T cells that share surface markers of NK cells and recognize antigens presented by the major histocompatibility complex (MHC) class I-like molecule CD1d (1). Because subsequent studies have demonstrated that there are CD1d-restricted T cells that do not express NK markers and that conventional T cells express NK markers when activated, CD1d restriction rather than the expression of NK markers is considered to be a critical feature of NKT cells (2, 3). Currently, NKT cells are defined as T cells that recognize antigens presented by MHC class I-like molecules, including CD1d.

#### *Edited by:*

*Kazuya Iwabuchi, Kitasato University School of Medicine, Japan*

#### *Reviewed by:*

*Vipin Kumar, University of California, San Diego, United States Masaki Terabe, National Cancer Institute (NIH), United States*

> *\*Correspondence: Akihiro Ishizu aishizu@med.hokudai.ac.jp*

#### *Specialty section:*

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

*Received: 09 December 2017 Accepted: 05 March 2018 Published: 15 March 2018*

#### *Citation:*

*Nishioka Y, Masuda S, Tomaru U and Ishizu A (2018) CD1d-Restricted Type II NKT Cells Reactive With Endogenous Hydrophobic Peptides. Front. Immunol. 9:548. doi: 10.3389/fimmu.2018.00548*

CD1d is a CD1 family member that can present hydrophobic antigens (e.g., glycolipids) (4). Whereas the CD1 family includes CD1a, CD1b, CD1c, CD1d, and CD1e in humans, rodents express only CD1d but not CD1a, CD1b, CD1c, and CD1e (5, 6).

CD1d-restricted NKT cells stimulated by antigens can activate or suppress other immunocompetent cells through an immediate production of a large amount of cytokines. Thus, these cells are regarded as immunological modulators. They are classified into two subsets, namely, type I and type II, according to the different usage of T-cell receptors (TCRs) (7, 8).

CD1d-restricted type I NKT cells are also called invariant NKT cells because of their usage of the limited α-chain (Vα24-Jα18 in humans and Vα14-Jα18 in rodents) with no or very few nucleotide insertions in the complementarity-determining region 3 that constitutes TCR. These cells can recognize a glycolipid, α-galactosylceramide (α-GalCer) (9–11). Based on this evidence, α-GalCer-loaded CD1d tetramers are employed to detect CD1drestricted type I NKT cells. Up to now, several microbial and endogenous lipids have been identified as antigens recognized by CD1d-restricted type I NKT cells (12).

On the contrary, CD1d-restricted type II NKT cells use diverse TCRs (13) and react with a wide variety of antigens, including microbial and endogenous glycolipids and phospholipids and endogenous hydrophobic peptides, but do not recognize α-GalCer. Due to the difficulty in preparation of ligand-loaded CD1d tetramers to detect CD1d-restricted type II NKT cells specifically, studies on these cells fall behind those on CD1drestricted type I NKT cells. However, in terms of the abundance of CD1d-restricted type II NKT cells compared to CD1d-restricted type I NKT cells in humans, it is worthy of understanding their physiological and pathological roles (14, 15). In this review, we focus particularly on CD1d-restricted type II NKT cells that recognize endogenous hydrophobic peptides presented by CD1d.

#### ANTIGENS RECOGNIZED BY CD1d-RESTRICTED TYPE II NKT CELLS

Previous studies have demonstrated that CD1d-restricted type II NKT cells recognize microbial and endogenous glycolipids and phospholipids. Tatituri et al. identified *Mycobacterium tuberculosis*-derived phosphatidylglycerol (PG), diphosphatidylglycerol (cardiolipin), and phosphatidylinositol as exogenous antigens recognized by CD1d-restricted type II NKT cells (16). More recently, Wolf et al. have demonstrated that Gram-positive *Listeria monocytogenes*-derived PG is also recognized by CD1drestricted type II NKT cells (17). It has been demonstrated that CD1d-restricted type II NKT cells can recognize lysophosphatidylcholine species (18) and sulfatides, which are endogenous glycolipids abundant in the central nervous system (19). Nair et al. identified two types of CD1d-restricted type II NKT cells in patients with Gaucher's disease (20). These cells recognize β-glucosylceramide and glucosylsphingosine. Rhost et al. have demonstrated that β-GalCer is also recognized by CD1drestricted type II NKT cells (21).

Although CD1d mainly presents lipid antigens, such as glycolipids, earlier studies depicted that it can present hydrophobic peptides (22, 23). In 2011, Liu et al. reported that type II collagen peptide could be an autoantigen recognized by CD1d-restricted NKT cells (24). Although they did not declare the subtype of type II collagen-reactive CD1d-restricted NKT cells in the paper, it appears to be a type II phenotype in terms of the antigen property. More recently, our group demonstrated that the hydrophobic peptide derived from sterol carrier protein 2 (SCP2), an intracellular lipid transporter, could be presented by CD1d and then recognized by CD1d-restricted type II NKT cells (25). Girardi et al. conducted X-ray analyzes of binding formation of CD1d and peptide antigens (26). The results demonstrated that the binding mode of peptides and CD1d is obviously different from that of glycolipids and CD1d but rather resemble that of peptides and MHC.

Castano et al. reported that peptide antigens that contain a motif of [F/W]XX[I/L/M]XXW can bind to CD1d (22). However, neither type II collagen peptide PPGANGNPGPAGPPG (24) nor SCP2 peptide FFQGPLKITGNMGLA (25) contains this motif. Furthermore, the membrane proximal external region of HIV-1 glycoprotein 41 peptide that contains this motif cannot bind to CD1d (26). The collective findings suggest that it is not easy to generalize the motif that represents the binding potential to CD1d. Not only amino acid sequences but also electric charge and peptide length could influence the binding to CD1d.

### FUNCTION OF CD1d-RESTRICTED TYPE I NKT CELLS

It has been demonstrated that the development of experimental autoimmune encephalomyelitis (EAE) and uveitis is enhanced in CD1d-restricted type I NKT cell-deficient Jα18 knockout mice (27, 28). In addition, the activation of CD1d-restricted type I NKT cells protects non-obese diabetes (NOD) mice from developing insulitis (29). These findings suggest an anti-inflammatory aspect of CD1d-restricted type I NKT cells in these murine autoimmune diseases. On the contrary, Jahng et al. reported that the simultaneous activation of CD1d-restricted type I NKT cells and myelin-reactive T cells exacerbated the progression of EAE (30), and Griseri et al. reported that CD1d-restricted type I NKT cells accelerated insulitis in the CD8 T cell-mediated diabetes model (31). These controversial results might be attributed to the experimental conditions. Interestingly, it has been demonstrated that CD1d-restricted type I NKT cells protect from Th1-mediated inflammation (32–34) but exacerbate Th2-mediated inflammation (35–37).

The following studies suggest a rather pro-inflammatory than anti-inflammatory phenotype of CD1d-restricted type I NKT cells. Chiba et al. reported that CD1d-restricted type I NKT cell number was increased in arthritic joints in the collageninduced arthritis model (33). In another antibody-induced arthritis model, CD1d-restricted type I NKT cells augmented inflammation in the joints by suppressing the production of the anti-inflammatory transforming growth factor β (TGF-β) (34). CD1d-restricted type I NKT cells accelerated inflammation also in cholangitis model (38). Furthermore, Kumar et al. reported that CD1d-restricted type I NKT cells acted as an inflammation promoter in liver injury by activating NK cells that kill hepatocytes through Fas/Fas ligand (FasL) interaction as well as by producing the pro-inflammatory cytokines (39–41). The more recent study demonstrated that the activation of CD1d-restricted type I NKT cells in the lungs by *Francisella tularensis* induced tularemia-like disease in mice (42).

In tumor immunity, CD1d-restricted type I NKT cells are also associated with the promotion of immune response against tumors (43). For instance, it has been demonstrated that the activation of CD1d-restricted type I NKT cells increased survival in mice bearing B16 melanoma (44, 45). Subsequent studies have revealed that a large amount of interferon-γ released from

sometimes results in the injury of own tissues. Under such situation, injured cells then present hydrophobic autoantigens, probably peptides, on their CD1d to activate CD1d-restricted type II NKT cells. Thereafter, activated CD1d-restricted type II NKT cells function to diminish inflammation by producing anti-inflammatory cytokines and by inducing apoptosis of effector cells *via* Fas/FasL interaction. (B) When CD1d-restricted type II NKT cells activated by endogenous hydrophobic peptides produce pro-inflammatory cytokines, the inflammation is exacerbated.

activated CD1d-restricted type I NKT cells is pivotal for tumor protection (46, 47).

# FUNCTION OF CD1d-RESTRICTED TYPE II NKT CELLS

The function of CD1d-restricted type II NKT cells has been investigated mainly by the following methods: (1) *in vivo* and/or *in vitro* stimulation by sulfatides; (2) observation of the difference in phenotype between CD1d knockout mice, which lack whole CD1d-restricted NKT cells, and Jα18 knockout mice, which solely lack CD1d-restricted type I NKT cells; and (3) use of 24αβ transgenic mice that carry the CD1d-restricted type II NKT cellderived TCR gene. The stimulation of CD1d-restricted type II NKT cells by sulfatides resulted in anti-inflammatory effects on liver injury (39, 40). Kwiecinski et al. demonstrated that sulfatidestimulated CD1d-restricted type II NKT cells attenuated sepsis induced by *Staphylococcus aureus* (48). Concerning these mechanisms, some studies have demonstrated that sulfatide-stimulated CD1d-restricted type II NKT cells suppressed the activation of pro-inflammatory type I NKT cells (39, 49, 50).

Terabe et al. and Renukaradhya et al. independently conducted experiments employing CD1d knockout and Jα18 knockout mice, and they both demonstrated that CD1d-restricted type II NKT cells downregulated cancer immunosurveillance (51, 52). Furthermore, other experiments that employed CD1d knockout and Jα18 knockout mice revealed that CD1d-restricted type II NKT cells attenuated the development of graft-versus-host disease after bone marrow transplantation (53).

Cardell et al. generated the CD1d-restricted type II NKT cell hybridoma VIII24 from MHC class II knockout mice (54). Skold et al. developed 24αβ mice that carried the Vα3.2-Vβ9 gene derived from the TCR of VIII24 hybridoma (55). Duarte et al. transduced the Vα3.2-Vβ9 gene into NOD mice and established 24αβ/NOD mice (56). These mice exhibited a decrease in the incidence of diabetes compared to the parent NOD mice. Furthermore, Liao et al. generated 24αβ/CD1dTg mice that overexpressed CD1d, and demonstrated that these mice spontaneously developed colitis underlying dysregulated differentiation of CD1d-restricted Vα3.2-Vβ9<sup>+</sup> type II NKT cells in the thymus (57).

The study published by Liu et al. (24) is noteworthy. They reported that type II collagen peptide-reactive CD1d-restricted NKT cells suppressed autoimmune arthritis by producing TGF-β, an anti-inflammatory cytokine, and by inducing apoptosis of effector cells through Fas/FasL interaction. This report encouraged us to make the following hypothesis: preceding inflammation sometimes results in the injury of own tissues. Under such situation, injured cells then present hydrophobic autoantigens, probably peptides, on their CD1d to activate CD1d-restricted type II NKT cells. Thereafter, activated CD1d-restricted type II NKT cells function to diminish inflammation by producing antiinflammatory cytokines and by inducing apoptosis of effector cells *via* Fas/FasL interaction (**Figure 1A**).

## INVOLVEMENT OF CD1d-RESTRICTED TYPE II NKT CELLS IN IMMUNE-RELATED INFLAMMATORY DISEASES

It has been demonstrated that CD1d-restricted type II NKT cells play critical roles in the development of liver injury in an acute hepatitis B transgenic murine model (58). In addition, the activation of CD1d-restricted type II NKT cells is accompanied by conventional T cell activation and proinflammatory cytokine production, leading to an enhancement of hepatic injury in murine autoimmune hepatitis models (59).

In patients with ulcerative colitis (UC), sulfatide-reactive CD1d-restricted type II NKT cells in lamina propria mononuclear cells are increased compared to healthy controls and patients with Crohn's disease (60). Moreover, sulfatide stimulation induced pathogenic interleukin (IL)-13 production and IL-13Rα2 expression on CD1d-restricted type II NKT cells from UC patients but not from healthy controls or patients with Crohn's disease. These findings are consistent with the previous report indicating that CD1d-restricted type II NKT cells with dysregulated differentiation is pathogenic in the murine colitis model (57). A recent study using CD1d-deficient (CD1d knockout) and CD1d-restricted type I NKT cell-deficient (Jα18 knockout) mice has also demonstrated that pro-inflammatory type II NKT cells are involved in dextran sulfate sodium-induced colitis in mice (61).

More recently, our group has revealed the involvement of proinflammatory type II NKT cells that are reactive with the endogenous SCP2 peptide in the pathogenesis of small vessel vasculitis in rats (25, 62) (**Figure 1B**). CD1d-restricted type II NKT cells activated by the SCP2 peptide function to enhance inflammation by producing pro-inflammatory cytokines. Although further studies are needed to clarify the precise mechanism, the involvement of pro-inflammatory CD1d-restricted type II NKT cells that recognize endogenous hydrophobic peptide is worthy of attention in the pathogenesis of immune-related inflammatory diseases.

# AUTHOR CONTRIBUTIONS

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

# FUNDING

This study was supported by a Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan with grant number 26293082 (AI), a grant from the Japan Research Committee of the Ministry of Health, Labor, and Welfare for Intractable Vasculitis (AI), and a grant from the Japan Agency for Medical Research and Development with grant number 15ek0109121 (AI).

# REFERENCES


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

*Copyright © 2018 Nishioka, Masuda, Tomaru and Ishizu. 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.*

# Therapeutic Potential of invariant natural Killer T Cells in Autoimmunity

*Luc Van Kaer\* and Lan Wu*

*Department of Pathology, Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN, United States*

Tolerance against self-antigens is regulated by a variety of cell types with immunoregulatory properties, such as CD1d-restricted invariant natural killer T (iNKT) cells. In many experimental models of autoimmunity, iNKT cells promote self-tolerance and protect against autoimmunity. These findings are supported by studies with patients suffering from autoimmune diseases. Based on these studies, the therapeutic potential of iNKT cells in autoimmunity has been explored. Many of these studies have been performed with the potent iNKT cell agonist KRN7000 or its structural variants. These findings have generated promising results in several autoimmune diseases, although mechanisms by which iNKT cells modulate autoimmunity remain incompletely understood. Here, we will review these preclinical studies and discuss the prospects for translating their findings to patients suffering from autoimmune diseases.

#### *Edited by:*

*Hongbo Chi, St. Jude Children's Research Hospital, United States*

#### *Reviewed by:*

*Yuan Zhuang, Duke University, United States John R. Lukens, University of Virginia, United States*

*\*Correspondence:*

*Luc Van Kaer luc.van.kaer@vanderbilt.edu*

#### *Specialty section:*

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

*Received: 09 February 2018 Accepted: 28 February 2018 Published: 13 March 2018*

#### *Citation:*

*Van Kaer L and Wu L (2018) Therapeutic Potential of Invariant Natural Killer T Cells in Autoimmunity. Front. Immunol. 9:519. doi: 10.3389/fimmu.2018.00519*

Keywords: invariant natural killer T cells, CD1d, immunotherapy, autoimmunity, type 1 diabetes, multiple sclerosis, rheumatoid arthritis, systemic lupus erythematosus

# INTRODUCTION

Autoimmunity develops when tolerance against self-tissues becomes compromised. Such a breakdown of the normal mechanisms that promote self-tolerance may be triggered in a genetically susceptible host by environmental factors. Traditional treatments for autoimmune diseases have predominantly relied on immunosuppressive and anti-inflammatory agents that often only provide short-term relief. A more enticing outcome of immunotherapy would be to prevent clinical manifestations or to stop disease progression after its initiation. One potential way to accomplish this would be to re-establish tolerance by targeting immunoregulatory cell networks. A lot of the work in this field has centered on CD4<sup>+</sup>Foxp3<sup>+</sup> regulatory T cells (Tregs), whereas other studies have concentrated on invariant natural killer T (iNKT) cells, a subset of glycolipid-reactive T cells. Here, we review the preclinical studies with iNKT cell antigens in mouse models of autoimmune disease, discuss proposed mechanisms for their therapeutic efficacy, and consider the hurdles faced in translating these findings to patients with autoimmune diseases.

# A BRIEF PRIMER ON iNKT CELLS AND THEIR FUNCTIONS

Invariant natural killer T cells express a semi-invariant T cell receptor (TCR), Vα14-Jα18/Vβ8.2, -7, or -2 in mice or Vα24-Jα18/Vβ11 in humans, and multiple surface markers associated with activated/memory T cells or natural killer (NK) cells [reviewed in Ref. (1–5)]. The semi-invariant TCR of iNKT cells recognizes glycolipid antigens presented in the context of the MHC class I-related protein CD1d. Relevant antigens for iNKT cells include both exogenous and endogenous glycolipids, many of which are glycosphingolipids. A common antigen employed in the iNKT cell field is the α-galactosylceramide (α-GalCer) KRN7000, which is a synthetically optimized version of a glycolipid originally isolated from a marine sponge (6).

In mice, functional subsets of iNKT cells, called iNKT1, iNKT2, iNKT10, and iNKT17 cells, have been identified that are characterized by production of the signature cytokines IFN-γ, IL-4, IL-10, and IL-17, respectively (7–11). These subsets are generated in the thymus, are characterized by expression of signature transcription factors, and are enriched in specific tissues, i.e., liver and spleen for iNKT1 cells; lungs and intestine for iNKT2 cells; adipose tissue and spleen for iNKT10 cells; and lungs, skin, and lymph nodes for iNKT17 cells. Additionally, a subset of iNKT cells that produces IL-21 and is specialized to interact with B cells to regulate humoral immunity (called iNKTFH cells) has also been identified (12). In humans, similar subsets of iNKT cells expressing select transcription factors and cytokines have not yet been fully characterized.

Within hours of lipid antigen activation, iNKT cells can mount an effector response characterized by cytokine production and cytotoxicity (1–5). These cells therefore represent a critical component of the innate arm of the immune response. Activation of iNKT cells in this way also leads to the transactivation a variety of other immune cell types (13, 14). Consequently, iNKT cells can either promote or suppress immune responses in different diseases (15). They promote natural immunity to cancer, protect the host against some infections, typically suppress autoimmunity, and contribute to the development of a variety of inflammatory diseases (1–5, 15, 16). In both mice and humans that are predisposed to the development of autoimmunity, iNKT cells often are reduced in number and exhibit an IFN-γ-biased cytokine production profile (16–18), providing indirect evidence for a role of these cells in curbing autoimmunity.

Considering their immunoregulatory functions, the therapeutic activities of iNKT cells in disease have been examined (16, 17). These studies have provided evidence for therapeutic efficacy against tumors, infectious agents, and autoimmune and inflammatory diseases.

## *IN VIVO* RESPONSE OF iNKT CELLS TO GLYCOLIPID ANTIGENS

Most studies investigating the *in vivo* response of iNKT cells to glycolipid antigen activation have employed KRN7000, which when injected by the intraperitoneal route in mice, results in systemic iNKT cell activation (19, 20). Activation of iNKT cells in this way results in the following series of events: (a) KRN7000 is presented to iNKT cells by CD1d-expressing antigen-presenting cells, predominantly CD8α+ dendritic cells (DCs) (21). (b) iNKT cells become activated within hours, resulting in the induction of activation markers such as CD25, CD69, and ICOS. (c) iNKT cells rapidly but transiently produce cytokines, with an initial burst of IL-4 (1–8 h), followed by IFN-γ (12–36 h activation) (16). (d) These cells transiently (between 8 and 30 h after treatment) downregulate their TCRs (22). (e) They also downregulate surface expression of the NK cell marker NK1.1, which occurs as early as 24 h after treatment and can last for an extended time period (over 1 month) (22). (f) iNKT cells upregulate expression of the programmed death-1 (PD-1) inhibitory receptor, which is evident as early as 2–3 days after KRN7000 treatment and may last for an extended time period (up to 2 months) (23–25). (g) iNKT cells rapidly expand in multiple tissues (spleen, peripheral blood, bone marrow, and liver), which peaks around 3 days after treatment (22, 26). (h) The iNKT cell population returns to pre-treatment levels within 2–3 weeks, which is mediated by activation-induced cell death (22, 26, 27). (i) While iNKT cells lack classical immunological memory, these cells exhibit long-term alterations in immune responsiveness following lipid antigen stimulation. Specifically, *in vivo*-activated iNKT cells acquire a hyporesponsive or anergic phenotype, resulting in reduced proliferation and IFN-γ production in response to glycolipid antigen restimulation (27, 28). Such hyporesponsiveness was noted as early as 3 days until up to 2 months after KRN7000 treatment. Repeated intraperitoneal injection of KRN7000 is particularly powerful in inducing longterm iNKT cell anergy. While the physiological significance of this response remains uncertain, it may prevent persistent cytokine production in order to avoid chronic inflammation during situations where glycolipids are present for an extended time period (28). Mechanistic studies revealed a role for the PD-1/PD-L pathway in this process (23–25, 29, 30). It was also noted that these hyporesponsive iNKT cells exhibit regulatory properties due to their capacity to produce residual amounts of IL-4 (28) and increased levels of IL-10 (10), thereby suggesting that they might have adopted a phenotype characteristic of iNKT10 cells.

Administration of KRN7000 *via* the intraperitoneal or intravenous routes predominately activates iNKT1 and to a lesser extent iNKT2 cells in spleen and liver, but does not activate iNKT2 cells in lymph nodes (9). However, oral administration of KRN7000 stimulates iNKT2 cells in mesenteric lymph nodes (9). The latter manner of administration also avoids induction of iNKT cell anergy (31), as does administration *via* the intradermal (32) and intranasal (31) routes, in the context of strong co-stimulation (28, 33), blockade of the PD-1/PD-L pathway (23, 24, 34), nanoparticles (35), or recombinant CD1d molecules (36). Due to differences in the distribution of tissue-specific iNKT cell subsets, different mouse strains induce divergent responses to KRN7000, with BALB/c mice generating IL-4-biased iNKT cell responses and SJL/J mice generating IFN-γ-biased responses as compared with C57BL/6 mice (9, 37).

Although information is limited, studies with human subjects have shown that KRN7000 and related glycolipids can promote iNKT cell cytokine production and expansion (38). Additionally, repeated KRN7000 treatment caused progressively lower iNKT cell responses in these patients (39), thereby suggesting anergy induction. When KRN7000 was delivered to patients pre-loaded on DCs, such iNKT cell dysfunction was avoided (40).

The cytokine production profile of iNKT cells can be modulated by a variety of means, such as the strength and quality of co-stimulation, the presence of cytokines, as well as the nature of the glycolipid antigen employed (16, 41, 42). Structural variants of KRN7000 have been identified that deviate iNKT cell responses toward T helper (Th)1 or Th2 cytokine production (16, 41, 42), or that fail to induce iNKT cell anergy (43). These methods to modulate iNKT cell cytokine responses have been exploited for the development of improved iNKT cell-based therapeutics.

# IMPACT OF iNKT CELL ANTIGENS ON INNATE AND ADAPTIVE IMMUNE RESPONSES

Invariant natural killer T cells are engaged in extensive crosstalk with other immune cell types, which greatly impacts their therapeutic activities (16). Glycolipid-activated iNKT cells activate and enhance cytokine production by DCs and macrophages, modulate the functions of neutrophils, and influence the generation, recruitment, and functions of myeloid-derived suppressor cells (MDSCs). Glycolipid-activated iNKT cells also induce IFN-γ production and cytotoxicity in NK cells (44). iNKT cell antigens also influence adaptive immune responses, including CD8 and CD4 T cell responses, as well as B cell and antibody responses. Most evidence suggests that KRN7000 enhances Th2 immunity, especially when administered repeatedly (16, 45, 46). Structural variants of KRN7000 that further bias adaptive immune responses toward Th2 cytokine production (e.g., OCH and C20:2) or that instead promote Th1 immunity (e.g., α-C-GalCer) have been identified (16). Additionally, iNKT cell antigens can enhance the generation and suppressive properties of CD4<sup>+</sup>Foxp3<sup>+</sup> Tregs (16, 47). These effects of glycolipid-activated iNKT cells on immune responses formed the scientific premise for investigating the therapeutic activities of KRN7000 and related glycolipids in a variety of diseases, including autoimmune diseases.

# PRECLINICAL STUDIES OF iNKT CELL ANTIGENS IN AUTOIMMUNITY

The immunomodulatory activities of KRN7000 and related iNKT cell antigens have been investigated in mouse models of autoimmunity (16, 18, 48). Key studies in select autoimmune diseases are reviewed here and potential mechanisms will be discussed in the next section.

# Autoimmune Diabetes

Several research groups have evaluated the effects of KRN7000 in non-obese diabetic (NOD) mice (16), a tractable model to study type 1 diabetes. Repeated injection of KRN7000 partially prevented insulitis and protected against diabetes (46, 49–51). This treatment was most effective when started early, during the initial stages of insulitis. KRN7000 was also protective when diabetes development in NOD mice was accelerated by treatment with cyclophosphamide (49), and following transplantation of freshly diabetic NOD mice with pancreatic islets (49). Repeated injection of the Th2-biasing α-GalCers OCH or C20:2 exhibited improved outcomes in the prevention of diabetes in NOD mice as compared with KRN7000 (52, 53).

# Multiple Sclerosis-Like Disease

The effects of KRN7000 and related glycolipids in the experimental autoimmune encephalomyelitis (EAE) model of multiple sclerosis-like disease in mice have been investigated using multiple myelin-derived autoantigens, mouse strains, and treatment protocols (54). Repeated subcutaneous or intraperitoneal injection of KRN7000 at the time of immunization of C57BL/6 mice with a myelin oligodendrocyte glycoprotein (MOG) peptide protected the animals against EAE (55–57). The Th2-biasing α-GalCer variant OCH was more effective than KRN7000 in protecting C57BL/6 mice against MOG-induced EAE, and was also effective when administered to mice *via* the oral route (58). In PL/J mice, treatment with KRN7000 at the time of EAE induction with myelin basic protein (MBP) also ameliorated disease (56). However, in MBP peptide-induced disease in B10.PL mice, a similar co-treatment protocol exacerbated disease, whereas KRN7000 treatment prior to disease induction was protective (56). Finally, in SJL/J mice, which contain low numbers and Th1-biased iNKT cells, KRN7000 exacerbated MBP-induced EAE (55).

# Autoimmune Arthritis

OCH but not KRN7000 was shown to protect C57BL/6 mice against collagen-induced arthritis (59). KRN7000 similarly protected against collagen-induced arthritis in DBA/1 mice when the animals were treated early but not late during the disease process (60, 61). Surprisingly, a single injection of the Th1-biasing analog α-C-GalCer was also effective in protecting DBA/1 mice against arthritis (61). KRN7000 also protected against arthritis in a model induced by immunization of DBA/1 mice with a glucose-6-phosphate isomerase peptide (62).

# Lupus-Like Autoimmunity

The impact of iNKT cell activation on both spontaneous and induced models of lupus-like autoimmunity has been explored (63). Repeated injection of KRN7000 to lupus-prone MRL/lpr mice ameliorated skin inflammation but did not affect kidney inflammation (64). In the NZB/W model of spontaneous lupus development, treatment at a young age resulted in disease amelioration (65), whereas treatment at an older age resulted in disease exacerbation (66), thereby suggesting that iNKT cell activation might have different effects when administered during different stages of the disease. Repeated KRN7000 treatment also ameliorated lupus-like disease induced by the natural hydrocarbon oil pristane in BALB/c mice, but exacerbated disease in SJL/J mice (37).

## MECHANISMS OF AUTOIMMUNE MODULATION BY iNKT CELL ANTIGENS

Because the pathogenesis of autoimmune diseases is diverse, mechanisms responsible for the immunomodulatory effects of iNKT cell antigens in autoimmunity are likely diverse as well. Nevertheless, general themes by which iNKT cells exert their therapeutic activities have emerged (**Figure 1**).

First, disease protection afforded by KRN7000 often correlates with increased Th2 and/or reduced Th1/Th17 responses against the targeted autoantigens (37, 46, 49–51, 55–57, 60, 61, 64). Conversely, in cases where disease was exacerbated, the opposite profile was usually seen (37, 55, 56, 66). These findings suggest that a KRN7000-induced shift in the autoantigen-specific Th cell profile contributes to disease protection. This possibility is further supported by studies showing that Th2-biasing KRN7000

analogs often have superior therapeutic efficacy than the original compound (52, 53, 58, 59). Likewise, co-administration of KRN7000 with Th2-biasing anti-CD86 antibodies (67) or with the Th2-biasing cytokine IL-7 (49) also enhanced therapeutic efficacy. Such immune deviation might be imparted directly by Th2-biasing cytokines (e.g., IL-4, IL-10, and IL-13) produced by iNKT cells, or indirectly, by inducing immunoregulatory cells that subsequently promote tolerance (see below). At least in some studies, these effects of glycolipid-activated iNKT cells on autoimmune responses correlated with the capacity of the antigen to induce iNKT cell anergy (28, 29, 68, 69).

Second, KRN7000 treatment studies performed with an experimental model of autoimmune myasthenia gravis (70) and with the NOD model of diabetes (71) have shown a critical role of Foxp3<sup>+</sup> Tregs in disease protection. Cytokines such as IL-2, IL-10, and TGF-β, produced by KRN7000-stimulated iNKT cells may all contribute to the generation of Foxp3<sup>+</sup> Tregs in this system (47). Additionally, it is possible that induction of Foxp3<sup>+</sup> Tregs involves immune suppressive myeloid lineage cells induced following iNKT cell activation (see below).

Third, there is ample evidence that KRN7000 can induce tolerogenic properties in myeloid lineage cells, including MDSCs (29, 72), DCs (50, 68, 73), macrophages (74), and neutrophils (75). MDSCs are a heterogeneous population of myeloid progenitor cells induced during inflammation by cytokines such as IL-4, IL-10, IL-13, IFN-γ, and GM-CSF (76) that can all be produced by iNKT cells. In steady-state conditions, these cells quickly differentiate into various mature myeloid cells such as macrophages, DCs and granulocytes. However, during inflammatory conditions, they rapidly expand, retain their immature phenotype, and acquire immunosuppressive properties (77). These cells employ a variety of mechanisms to suppress T cell function, including arginase-1 and inducible nitric oxide synthase activity, reactive oxygen species, immunosuppressive cytokines such as IL-10 and TGF-β, and induction of Foxp3<sup>+</sup> Tregs (76, 77). The latter property of MDSCs provides a potential link between the capacity of KRN7000 to induce both MDSCs and Foxp3<sup>+</sup> Tregs. Studies with the EAE model have shown a critical role of immunosuppressive MDSCs in disease protection mediated by KRN7000 (72). Similarly, a role for tolerogenic DCs was shown in EAE (68) and the NOD model of type 1 diabetes (50), and for tolerogenic M2-phenotype macrophages in EAE (74). Because MDSCs can differentiate into mature myeloid cells (76), tolerogenic DCs, and macrophages induced by KRN7000 might be derived from MDSCs (**Figure 1**).

Finally, in the studies with Th1-biasing KRN7000 analogs, IFN-γ was key in suppressing Th1 and Th17 responses, without promoting Th2 responses (61, 78). Although the relevant source of IFN-γ was not explored, both antigen-activated iNKT cells and transactivated NK cells produce IFN-γ under such conditions. While it remains unclear how IFN-γ might lead to suppression of pathogenic T cells, it can inhibit Th17 responses that play key roles in many autoimmune diseases (79), a possibility that is consistent with the suppressive role of iNKT cells toward Th17 responses (80). Additionally, IFN-γ might suppress pathogenic T cells by inducing anergy in them, a possibility that is supported by studies on the suppressive activities of iNKT cells against diabetogenic T cells in NOD mice (81, 82).

While each of the listed mechanisms likely contributes to the protective effects of KRN7000 on autoimmunity, immune deviation, and induction of immunosuppressive myeloid cells are likely to be most critical for the Th2-biasing KRN7000 analogs, whereas anergy induction in pathogenic T cells is likely the dominant mechanism for the Th1-biasing analogs (**Figure 1**).

Because these proposed mechanisms of protection involve both Th1 and Th2 cytokines, it is perhaps not surprising that IL-4, IL-10, and IFN-γ (37, 46, 49–53, 55–57, 59–61, 64, 65, 80, 83, 84), all have been shown to play a role in the tolerogenic properties of KRN7000. Even more striking, in some experimental models KRN7000 can protect against autoimmunity in an IL-4- and/or IL-10-independent manner (84, 85). Thus, multiple mechanisms might be at play that are influenced by the particular animal model of autoimmunity employed and by a variety of other factors such as the amount, timing, frequency, and route of glycolipid treatment. Another variable that likely contributes to some of the divergent findings in this field is the profound effect of the natural microbiota on the functions and therapeutic properties of iNKT cells (86).

#### FUTURE OUTLOOK

Although KRN7000 and related iNKT cell antigens can protect against autoimmunity in many experimental models, there are also several examples of disease exacerbation. Considering the large number of variables that affect their therapeutic efficacy, a better understanding of the immunomodulatory properties of iNKT cell antigens is required, not only in mice but especially so in humans. While treatment with iNKT cell antigens in human subjects has favorable safety profiles (38), it remains uncertain

#### REFERENCES


if this will also be the case in patients at risk for or with autoimmunity. In particular, iNKT cells from human patients with autoimmunity are commonly reduced in numbers and often exhibit a Th1-biased cytokine production profile as compared with healthy subjects (15, 16, 18), which might be challenging to overcome during therapy. Translating the animal studies to humans is further complicated by differences in the prevalence, tissue distribution, and functions of iNKT cells between mice and humans (1–5), which likely all impact therapeutic potential. Nevertheless, the studies reviewed here have revealed cooperative interactions between immune suppressive iNKT cells, Tregs, and myeloid lineage cells, which could be exploited in combination therapies.

#### AUTHOR CONTRIBUTIONS

LK wrote the first draft, and LK and LW edited the manuscript.

### FUNDING

Work in the authors' lab was supported by grants from the National Institutes of Health (DK104817), the Department of Defense (W81XWH-15-1-0543), the National Multiple Sclerosis Society (60006625), and the Crohn's and Colitis Foundation of America (326979).


of glycolipid-activated invariant NKT cells. *J Immunol* (2009) 182:2816–26. doi:10.4049/jimmunol.0803648


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

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

# The influence of invariant natural Killer T Cells on Humoral immunity to T-Dependent and -independent Antigens

*Mark L. Lang\**

*Department of Microbiology and Immunology, University of Oklahoma Health Sciences Center, Oklahoma City, OK, United States*

Vaccination with CD1d-binding glycolipid adjuvants and co-administered protein, lipid, and carbohydrate antigens leads to invariant natural killer T (NKT) cell-dependent enhancement of protective B cell responses. NKT cell activation boosts the establishment of protein antigen-specific B cell memory and long-lived plasma cell (LLPC) compartments. NKT cells may exert a similar effect on some carbohydrate-specific B cells, but not lipid-specific B cells. The mechanisms of action of NKT cells on B cell responsiveness and subsequent differentiation into memory B cells and LLPC is dependent on CD1d expression by dendritic cells and B cells that can co-present glycolipids on CD1d and antigen-derived peptide on MHCII. CD1d/glycolipid-activated NKT cells are able to provide help to B cells in a manner dependent on cognate and non-cognate interactions. More recently, a glycolipid-expanded subset of IL-21 secreting NKT cells known as NKT follicular helper cells has been suggested to be a driver of NKT-enhanced humoral immunity. This review summarizes established and recent findings on how NKT cells impact humoral immunity and suggests possible areas of investigation that may allow the incorporation of NKT-activating agents into vaccine adjuvant platforms.

Keywords: CD1d, Natural Killer T, vaccine, humoral immunity, pathogen

#### INTRODUCTION

Several research groups have demonstrated that CD1d-restricted natural killer T (NKT) cells influence the humoral immune response to viruses, bacteria, their toxins, parasites, and fungi (**Table 1**). Typically prophylactic immunization of a mammal with a vaccine antigen or other pathogen product in combination with a CD1d-binding, NKT-activating adjuvant such as the α-galactosylceramide (α-GC) glycolipid has resulted in the enhancement of pathogen-specific Ab responses. These NKT-enhanced Ab responses are associated with, or contributory to enhanced protection against lethal challenges with pathogens or their toxins. The NKT-enhanced Ab responses are also typified by Ig class switch (1–4), establishment of B cell memory (Bmem) (2, 5), and long-lived plasma cells (LLPC) (6, 7), all hallmarks of a desirable vaccine response.

These findings support the notion that NKT cells could be harnessed following prophylactic vaccination to improve existing vaccines or contribute to the development of new vaccines.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*Laurent Brossay, Brown University, United States Paolo Dellabona, Scientific Institute San Raffaele (IRCCS), Italy*

> *\*Correspondence: Mark L. Lang mark-lang@ouhsc.edu*

#### *Specialty section:*

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

*Received: 11 January 2018 Accepted: 05 February 2018 Published: 22 February 2018*

#### *Citation:*

*Lang ML (2018) The Influence of Invariant Natural Killer T Cells on Humoral Immunity to T-Dependent and -Independent Antigens. Front. Immunol. 9:305. doi: 10.3389/fimmu.2018.00305*

**Abbreviations:** Bmem, memory B cells; LLPC, long-lived plasma cells; NKTfh, NKT follicular helper cells; α-GC, alpha-galactosylceramide.



*In response to the pathogens indicated by superscript "a," some groups observed that NKT activation enhanced humoral immunity, while others reported that NKT cells were dispensable for the response.*

Arguably, to understand how best to harness NKT cells during vaccination, and/or how to appropriately direct a humoral immune response, the intersection of NKT cell and B cell biology needs to be understood. In this article, we discuss what is known about the mechanisms by which invariant NKT cells influence humoral immunity. We also discuss whether NKT-activating adjuvants can or should be incorporated into vaccines. Type II NKT cells expressing diverse TCRs (dNKT) are fully discussed elsewhere (37, 38), but briefly described herein in the context of vaccination.

## MECHANISMS REGULATING NKT CELL INFLUENCE ON T-DEPENDENT HUMORAL IMMUNITY

As mentioned, co-administration of a protein Ag and α-GC leads to enhanced humoral immunity against the protein Ag in a manner that is CD1d-dependent, and NKT cell-dependent (39). A model for how the humoral response is initiated is shown in **Figure 1**. In this model, professional APCs including classical CD11c<sup>+</sup> dendritic cells (DCs) capture both the Ag and α-GC by

FIGURE 1 | Model for natural killer T (NKT) cell influence on humoral immunity (A) CD1d+/+ dendritic cells (DCs) are able to capture, internalize, process, and present peptide Ag on MHCII and glycolipid Ag on CD1d and do so in a coordinated fashion. As a result, Th cell priming occurs, as does NKT activation and/or NKT follicular helper cell (NKTfh) differentiation. (B) B cells capture Ag *via* the BCR, but also capture complexed CD1d-binding glycolipid, or internalize it by endocytosis. B cells are, thus, able to coordinately present peptide on MHCII and glycolipid on CD1d. Consequently, B cells are able to receive help from DC primed or activated classical Th/Tfh cells as well as NKT/NKTfh cells. The additional help from NKT/NKTfh cells enhances the establishment of a Bmem compartment and the generation of long-lived plasma cells.

endocytic mechanisms. This allows the internalization and trafficking of Ag and adjuvant (α-GC) into late endosomal processing compartments known as MIIC (MHC Class II compartments). It is in these compartments that protein-derived peptides and α-GC intersect with MHC II and CD1d, respectively (40, 41). Using well-defined mechanisms, peptide is loaded on MHCII and α-GC on CD1d [reviewed in Ref. (42, 43)]. The MHCII/ peptide and CD1d/α-GC complexes are then transported to the cell surface for presentation to classical CD4<sup>+</sup> T cells and NKT cells, respectively. Evidence also suggests that presentation of MHCII/peptide and CD1d/α-GC is facilitated by plasma membrane micro-domains or "rafts" (44, 45).

In the model (**Figure 1A**), Th priming by DCs is concordant with initial activation of NKT cells. In previous studies, our laboratory generated mixed bone marrow chimeric mice in which 50% of DCs expressed the diphtheria toxin receptor (DTR) under control of the CD11c promoter and the other 50% of cells were non-transgenic and CD1d<sup>+</sup>/+ or CD1d-/- (46). Administration of DT temporarily ablated DTR transgenic CD1d<sup>+</sup>/<sup>+</sup> DCs, leaving non-transgenic CD1d<sup>+</sup>/+ or CD1d-/- DCs intact. In those experiments, Ab titers were similar between the groups. However, complete ablation of DTR<sup>+</sup>; CD1d<sup>+</sup>/<sup>+</sup> DCs delayed the α-GC-enhanced Ab response, suggesting a contribution by CD1d<sup>+</sup>/<sup>+</sup> DCs (46). Since that experiment, a Cre-Lox system has been employed by the Bendelac group to permanently ablate only CD1d<sup>+</sup>/<sup>+</sup> DCs, showing a definitive contribution of these DCs to the humoral response to pneumococcal capsular polysaccharides (29). Although, a direct contribution of CD1d<sup>+</sup>/<sup>+</sup> DCs to T-dependent humoral responses has not been formally demonstrated, it appears likely that they are required for NKTenhanced responses.

In the model (**Figure 1B**), B cells specific for the immunizing Ag capture native Ag *via* the BCR and internalize α-GC by endocytosis, leading to MHCII and CD1d co-presentation by B cells. This will allow B cells to receive classical T cell help from Th cells and additional help from NKT cells. As a result of coordinated Th- and NKT-mediated B cell help, germinal center entry, Ig class switch, Bmem differentiation, and establishment of LLPC compartments are enhanced. Our laboratory performed adoptive transfers of CD1d<sup>+</sup>/<sup>+</sup> and CD1d<sup>−</sup>/<sup>−</sup> B cells into recipient μMT mice and demonstrated that B cell CD1d expression was essential for NKT-enhanced responses to the co-administered protein Ag (47). Co-presentation on MHCII and CD1d was further supported by Barral and colleagues who used liposomes containing Ag and α-GC for immunization (48).

These results raised the question of whether cognate interactions between B cells and NKT cells were occurring and dependent on CD1d and Vα14 TCR expression, respectively. In support of a direct B: NKT interaction and possible cognate interaction is our previous study adoptively transferring CD1d<sup>+</sup>/<sup>+</sup> and CD1d<sup>−</sup>/<sup>−</sup> B cells (47). Chang and colleagues used intra-vital microscopy to demonstrate direct interaction between HEL-specific MD4 B cells and NKT cells *in vivo* (49). The interactions lasted for 4–50 min suggesting a direct but time-limited interaction. The van den Elzen group showed that a combination of retinoic acid and α-GC led to reduced expression of CD1d by B cells, arguing for a constrained time window for B:NKT interaction (50). The Terhorst laboratory have also reported that signaling lymphocyte activation molecule associated protein (SAP) is expressed by NKT cells, but seems to be dispensable for initial B cells responses such as IgM production, but contributes to germinal center responses and, thus, class switch and somatic hyper-mutation (51). It should also be noted that Tonti and colleagues have observed cognate and non-cognate interactions between CD1d<sup>+</sup>/<sup>+</sup> B cells and NKT cells (52). This suggests that the particular Ag, the dose and formulation (particulate versus soluble or linked versus separate Ag and adjuvant), and perhaps the route of immunization could influence the degree to which enhanced Ab responses rely on B cell CD1d expression. However, on balance, the evidence that CD1d<sup>+</sup>/<sup>+</sup> B cells directly interact with NKT cells, and that this is required for NKT-enhanced humoral immunity is quite compelling.

Fewer studies have addressed whether there is direct communication between Th/Tfh and NKT/NKT follicular helper cells (NKTfh) cells during a humoral response. Our studies showed a temporal relationship between Th/Tfh and NKT/ NKTfh production of IL-4 and IL-21, with the NKT/NKTfh compartments providing an early source of IL-21 (27). However, we did not detect any direct dependence of one cell type upon the other with regard to cytokine secretion.

While there is good evidence in support of a CD1d-dependent mechanism for B cell stimulation of NKT cells, it is somewhat less clear how the NKT provides help to the B cell. For example, using mixed bone marrow chimeras in which NKT cells were either CD40L<sup>+</sup>/<sup>+</sup> or CD40L<sup>−</sup>/<sup>−</sup>, equal Ab responses to Ag and α-GC were observed (4). ICOS could not be studied in a similar manner because it is required for peripheral NKT survival (53), but *in vitro* assays suggested its requirement for NKT activation of marginal zone B cells (54). Given the propensity of marginal zone B cells to respond to T-independent Ags, its role in NKT-enhanced T-dependent responses remains unclear. It is difficult to envision CD40L and ICOS having no role to play in NKT-enhanced humoral responses, but experimental systems whereby these ligands are missing from the cell surface may be compensated by the same signals derived from Th cells. Alternatively, these co-receptor signals may be genuinely dispensable for NKT-mediated B cell help. If so, then the mechanisms of NKT- and Th-mediated B cell help are distinct.

Some evidence supports a role for NKT-derived soluble factors in B cell responses. The NKT cellular compartment is prolific in its rapid IFNγ and IL-4 section following α-GC activation, yet in the context of additional Th-mediated cytokine responses, NKT-derived cytokines may play a fairly limited role in influencing isotype switch. In bone marrow chimeras whereby NKT cells lacked IFNγ or IL-4, there were only modest effects on Ig class switch (3). A new study, however, reported that IL-4 secreting NKT cells positioned at the edge of the B cell follicle can promote germinal center entry, perhaps providing a mechanism of NKT-enhanced B cell memory (55). However, different laboratories have reported that α-GC leads to differentiation and expansion of a subset of NKT cells that display the hallmarks of T follicular helper cells (Tfh) and are, therefore, referred to as NKTfh cells (49, 56–58). This phenomenon explains the previous identification of an IL-21-secreting NKT subset (59), which is now known to express high levels of the master transcriptional regulator Bcl6, and upregulate the chemokine receptor CXCR5, and the PD1 molecule. The NKT subset may provide an early source of IL-21 (27) and perhaps accelerate Ig class switch, an effect that may have been missed in earlier studies examining cytokine contributions (3). The NKT-enhanced IgG response to T-dependent Ag is typically IgG1-dominated and this makes sense given the pivotal role of Tfh-derived IL-21 in IgG1 class switch (60, 61).

As mentioned, NKT activation is associated with increased numbers of LLPC (6, 7). Some mechanistic insights have been gained through bone marrow chimera experiments in which NKT cells lacked expression of either B cell activating factor (BAFF), a proliferation-inducing ligand (APRIL), or both BAFF and APRIL. While NKT-derived BAFF was dispensable for LLPC responses, APRIL made a modest contribution to longevity. However, the combination of BAFF and APRIL were critical for LLPC survival. In controls, bone marrow plasma cell numbers were maintained over around 90 days after immunization with minimal attrition. In the absence of NKT-derived BAFF and APRIL, there was a ~90% loss with 26 days (7). These data suggest a direct effect of NKT-derived plasma cell survival factors on the endurance of a humoral immune response.

### MECHANISMS REGULATING NKT CELL INFLUENCE ON T-INDEPENDENT HUMORAL IMMUNITY

Studies by our group demonstrated that Abs complexed to a biotinylated α-GC could be used to stimulate BCR-dependent uptake, trafficking, loading, and presentation by CD1d (41). This Ag presentation pathway resulted in 100- to 1,000-fold more efficient activation of NKT hybridoma cells and suggested a hypothesis that such pathways could stimulate NKT-driven production of glycolipid-specific Abs. Indeed, the Brenner group demonstrated that anti-nitrophenol (NP) hapten Abs could be produced in a CD1d-/NKT-dependent manner following immunization with an NP-modified α-GC (62). The humoral response to NP-α-GC was examined and found to stimulate short-lived IgM responses without the establishment of Ab recall responses and B cell memory (62). In a further study, the B cell response to glycolipids was attributed to NKTfh cells (58). Therefore NKT (and NKTfh cells) cells may be able to boost Bmem responses to T-dependent Ags but not T-independent lipid Ags.

The Bendelac group, however, demonstrated a role of NKT/ NKTfh cell-driven anti-polysaccharide responses (29). In a study involving immunization with capsular pneumococcal polysaccharides and α-GC, class-switch recombination, affinity maturation, and B cell memory were observed and there was a limited induction of NKTfh cell responses (29). In some unpublished studies from our laboratory, we have been unable to observe convincing Ab recall responses to T-independent carbohydrate Ags co-administered with α-GC, although there is a good adjuvant effect on primary responses (*Lang, unpublished observation*).

Clearly, information on the influence of NKT and NKTfh cells on humoral immunity to T-independent Ags is limited. More study is warranted in this area, particularly with regard to Ags associated with pathogenic bacteria.

#### CONSIDERATIONS FOR USING NKT CELL-ACTIVATING VACCINES

The α-GC adjuvant has been valuable in helping delineate mechanisms of action by which NKT cells impact humoral immunity. However, several questions remain as to how best to move forward to incorporating NKT activation strategies into vaccines. The α-GC adjuvant is particularly potent *in vivo* and has the potential to initially activate all Type I NKT cells expressing the Vα14 TCR. There have been numerous reports detailing NKT cell anergy whereby a single treatment with α-GC can induce long-term NKT hypo-responsiveness to further stimulation (63–65). However, route of immunization may be contributory to this effect. Intradermal, subcutaneous, and mucosal vaccination routes allow repeat immunization and NKT responsiveness whereas intravenous and intraperitoneal delivery tends to result in anergy (6, 66–68). Some of the mechanisms underlying NKT anergy have been delineated and there are signaling pathways, such as CARMA1 and PD-1 that can be targeted to minimize anergy in mouse models (69, 70). While PD-1 blockade might be of practical value in cancer immunization, it is likely impractical for routine prophylactic vaccination in the field. A study in mice whereby α-GC was administered by the intra-tracheal route led to airway NKT cell activation and exacerbated airway hyper-reactivity and inflammation which is worth considering as a potential caveat to intranasal administration (71). These studies demonstrate that a combination of adjuvant selection, formulation, route of delivery, and perhaps mitigation of anergy-driving mechanisms may have to be considered when incorporating CD1d ligands into vaccines.

There are now several variants based on the α-GC molecule that can attenuate or enhance NKT activation [reviewed in Ref. (72)]. The α-GC molecule can be modified in its acyl chain, sphingosine chain, or sugar head-group and there is, therefore, considerable room for manipulating its effects on NKT cells. Furthermore, the Th1/Th17 to Th2 balance can be modulated by altering the α-GC molecule. Depending on the type of immune response that is desired, a different α-GC-derived adjuvant could be used for vaccination, perhaps with weaker anergy-inducing effects.

The vaccine formulation itself should be considered. Physically linking or associating vaccine antigens with α-GC (or a derivative thereof) is more likely to ensure that the same DCs and B cells that capture the vaccine, and coordinately present peptide on MHCII/HLA-2 and α-GC on CD1d. Small soluble complexes may result in different outcomes from larger (~100 nm) particles where extra-follicular B cell responses were observed in mice. Selection of the best particle size for ensuring that follicular and perhaps germinal center responses are worth considering.

Several studies have shown that α-GC is safe and welltolerated when administered intravenously to cancer patients either in free form, or as part of a DC vaccine (73–77). It is, therefore, likely to be safe for inclusion in vaccines, but a few studies in mice implicated administration of α-GC during the third trimester in pregnancy loss, late preterm birth, and neonatal mortality (78–80). This issue, therefore, warrants additional attention to determine if α-GC adjuvants should be avoided in pregnancy.

This article has focused heavily on Type I invariant NKT cells. Type II NKT cells exhibit diverse TCR usage and respond to a growing list of CD1d-binding molecules. Available information on Type II NKT cells is reviewed elsewhere (37). We observed that the Th2 cytokine response to Alum adjuvant was attenuated by around 65% in mice lacking Type II NKT cells but not Type I NKT cells (81). This observation warrants further investigation and in the context of protection against pathogenic challenges. However, it should perhaps also be considered that Alum is safe, and stimulates excellent Th2 responses and poor Th1 responses. The inclusion of Type I NKT cell-activating adjuvants into vaccines containing Alum could potentially result in coordinated Type I and Type II NKT responses and give a broader response to existing vaccines.

#### REFERENCES


#### CONCLUDING REMARKS

The α-GC adjuvant has made possible valuable insights into how NKT cell and B cell biology intersect and provides a good jumping off point for the inclusion of similar adjuvants in vaccines. Potentially, derivatives of α-GC could be used to enhance, broaden, and extend the protective humoral response to a variety of protein and non-protein antigens.

#### AUTHOR CONTRIBUTIONS

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

#### ACKNOWLEDGMENTS

Research discussed herein is supported by NIH award AI125708 and AI134719 to ML. I thank the various members of my laboratory whose research has helped shape several of the ideas discussed herein.

protective immunity against influenza A virus. *Arch Virol* (2017) 162(5):1251– 60. doi:10.1007/s00705-017-3230-7


IL-21-dependent manner. *Nat Immunol* (2012) 13(1):44–50. doi:10.1038/ ni.2172


**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 a potential conflict of interest.

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

# Mucosal-Associated invariant T Cells: New insights into Antigen Recognition and Activation

*Xingxing Xiao1,2 and Jianping Cai1,2\**

*1State Key Laboratory of Veterinary Etiological Biology, Key Laboratory of Veterinary Parasitology of Gansu Province, Lanzhou Veterinary Research Institute, Chinese Academy of Agricultural Sciences, Lanzhou, China, 2 Jiangsu Co-Innovation Center for Prevention and Control of Animal Infectious Diseases and Zoonoses, Yangzhou, China*

Mucosal-associated invariant T (MAIT) cells, a novel subpopulation of innate-like T cells that express an invariant T cell receptor (TCR)α chain and a diverse TCRβ chain, can recognize a distinct set of small molecules, vitamin B metabolites, derived from some bacteria, fungi but not viruses, in the context of an evolutionarily conserved major histocompatibility complex-related molecule 1 (MR1). This implies that MAIT cells may play unique and important roles in host immunity. Although viral antigens are not recognized by this limited TCR repertoire, MAIT cells are known to be activated in a TCR-independent mechanism during some viral infections, such as hepatitis C virus and influenza virus. In this article, we will review recent works in MAIT cell antigen recognition, activation and the role MAIT cells may play in the process of bacterial and viral infections and pathogenesis of non-infectious diseases.

#### *Edited by:*

*Luc Van Kaer, Vanderbilt University, United States*

#### *Reviewed by:*

*S. M. Mansour Haeryfar, University of Western Ontario, Canada Derek G. Doherty, Trinity College, Dublin, Ireland*

> *\*Correspondence: Jianping Cai caijianping@caas.cn*

#### *Specialty section:*

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

*Received: 16 September 2017 Accepted: 30 October 2017 Published: 10 November 2017*

#### *Citation:*

*Xiao X and Cai J (2017) Mucosal-Associated Invariant T Cells: New Insights into Antigen Recognition and Activation. Front. Immunol. 8:1540. doi: 10.3389/fimmu.2017.01540*

Keywords: mucosal-associated invariant T cells, antigens, recognition, activation, diseases

# INTRODUCTION

T lymphocyte activation is best explained by conventional T cells using a very diverse repertoire of T cell receptors (TCRs), which recognize various antigenic peptides presented in the context of classical major histocompatibility complex (MHC) class I or class II molecules (1, 2). This response is adaptive or acquired and is different from the innate immune response, which is MHC and TCR independent. In humans and mice, however, there are also other important groups of T cells, known as non-conventional T cells, which can be activated in a mechanism distinct from conventional T cells. Unlike conventional T cells, which recognize peptide antigens, invariant natural killer T (iNKT) cells and mucosal-associated invariant T (MAIT) cells recognize lipids and vitamin B metabolites, respectively (3, 4). Once activated in a TCR-dependent and/or -independent manner, these unconventional T cells produce cytokines and some cytotoxic effector molecules, such as tumor necrosis factors (TNFs), perforin, and granzymes (Gzms) (5–8). Because these cells share properties of both the adaptive and innate responses, they are termed innate-like T cells, which express TCR with limited diversity and respond rapidly to relatively conserved antigen challenge, and can be said to represent a bridge between the adaptive and innate immune systems (8–11).

Like the intensively studied innate-like T cells, iNKT cells are restricted by CD1d and express a semi-invariant TCR Vα24-Jα18 in humans and Vα14-Jα18 in mice with a biased repertoire of Vβ chains (6, 12–15). In mice, iNKT cells are abundant in the liver, bone marrow and adipose tissue, but less in other tissues, such as spleen, blood and lung (16–18); however, in humans, they are rare in blood and liver compared to mice (19). Nevertheless, a recent study shows that iNKT cells are abundant in human omentum (20). iNKT cells recognize synthetic, self, and microbial lipid-based antigens presented by CD1d (6, 21, 22). Upon TCR or cytokinemediated activation, iNKT cells rapidly produce cytokines and chemokines, exert cytotoxic activity by high expression of Gzms, perforin, and FasL, and regulate other immune cells (6, 23). As such, iNKT cells play important roles in combating infectious diseases and are also implicated in host responses to autoimmune diseases, chronic inflammatory diseases, and cancers (17, 24).

Mucosal-associated invariant T cells, a new subpopulation of innate-like T cells, have many similarities to iNKT cells, both in immunological properties and functions. Both cell types express semi-invariant TCRs, are restricted by monomorphic molecules, and can rapidly produce proinflammatory cytokines. With an increasing understanding of the antigens recognized by MAIT cells and their activation, the roles MAIT cells play in the immune response have been gradually revealed. Here, this review will summarize the discovery and phenotypic characterization of MAIT cells, then focus on MAIT cell antigen recognition and activation, and finally, discuss the roles of MAIT cells in infectious and non-infectious diseases.

#### MAIT CELL DISCOVERY AND PHENOTYPE

In 1993, Porcelli et al. found that peripheral blood αβ CD4<sup>−</sup>CD8<sup>−</sup> [double negative (DN)] T cells from healthy people preferentially expressed two invariant TCRα chains (15). One invariant TCRα chain, comprised of the Vα24 and Jα18 gene segments (Vα14 and Jα18 in mice), was expressed by what we now know as iNKT cells, while the other invariant TCRα chain used the Vα7.2 and Jα33 gene segments (Vα19 and Jα33 in mice) (12, 14, 15). Then, in 1999, a seminal study by Tilloy et al. showed that this invariant TCRα Vα7.2/Vα19-Jα33 chain was expressed by a new type of T cell subset present in humans, mice, and cattle, indicating conservation between mammalian species (25). In 2003, Treiner et al. observed that this novel subset of T cells was preferentially located in the intestinal lamina propria (LP) of humans and mice, and therefore, named these cells "mucosalassociated invariant T (MAIT) cells" (26). They also showed that MAIT cells were restricted by MR1, a monomorphic class I-related MHC molecule that is highly conserved in mammalian species (26, 27).

Although originally described as αβ DN T cells, MAIT cells, which are CD3<sup>+</sup>, may also be CD4<sup>+</sup> and/or CD8<sup>+</sup> (25, 28, 29). In humans, only 2–11% of MAIT cells in blood express CD4, while in Vα19iTg-Cα−/<sup>−</sup> C57BL/6 (B6) mice, up to 50% of splenic MAIT cells are CD4<sup>+</sup>, implying different coreceptor requirements between humans and transgenic (Tg) mice (29). In wild-type (WT) mice (B6 or BALB/c), MAIT cells also consist of three subsets (DN, CD4<sup>+</sup>, and CD8<sup>+</sup>), and their frequencies vary in different tissues and strains (28). CD8<sup>+</sup>MAIT cells are present in different frequencies in humans and mice and are a mixture of CD8αα and CD8αβ cells (28–31). Contrary to iNKT cells, MAIT cells are abundant in humans, but rare in mice (17, 32). Moreover, as with iNKT cells, MAIT cells also display a restricted TCR repertoire. In humans, the majority of MAIT cells express a TCR comprised of a canonical Vα7.2-Jα33 TCRα rearrangement paired with limited TCRβ chains (predominantly Vβ2 or Vβ13), and less frequent usage of the non-canonical Vα7.2-Jα12/Jα20 TCRα rearrangement (15, 25, 29, 33). Mouse MAIT cell TCRs utilize an invariant Vα19-Jα33 TCRα chain predominantly paired with a Vβ6 or Vβ8 chain (25, 28). The variability of MAIT TCRβ chains may contribute to the functional heterogeneity of MAIT cells as immune effectors against various microbes (34).

Mucosal-associated invariant T cells express several cell surface proteins that can be used for identification and characterization (**Table 1**). Both human and mouse MAIT cells have a memory phenotype and may express transcription factors promyelocytic leukemia zinc finger (PLZF), retinoic acid-related orphan receptor γt (RORγt), and T-bet (28, 35–37). Detection of other cell surface proteins in various combinations has been used to detect MAIT cells. Both human and mouse MAIT cells express homing receptors, such as CXCR6, CCR9, α4β7, and/or CD103, which are consistent with their ability to migrate to the skin, liver, lung, and gut LP (28, 30). MAIT cells also express cytokine receptors, such as IL-7Rα, IL-12R, and IL-18Rα (28, 38–40).

Initially, the reagent used to specifically identify and study human MAIT cells (CD3<sup>+</sup>CD161hiVα7.2<sup>+</sup>T cells) was an antibody mix including anti-Vα7.2 and anti-CD161 mAb (35); however, because Vα7.2 TCR is also expressed by conventional T cells and CD1b-restricted germline-encoded, mycolyl lipid-reactive (GEM) T cells (Vα7.2-Jα9) (46, 47), an anti-Vα7.2 mAb alone is no longer considered an appropriate tool to study MAIT cells in humans. MAIT cells are phenotypically CD161hi but CD161 is downregulated on MAIT cells in HIV patients (48), so CD161 is not a useful marker when examining specimens from HIV positive patients and potentially other cases. Because MAIT cells are rare in mice and there is no anti-Vα19 specific mAb available, researchers generated B6 Tg mice expressing only the TCR Vα19-Jα33 (Vα19i) chain to study MAIT cell development, phenotype, and antigen specificity (35, 41, 49). Unfortunately, maybe because of the developmental differences in MAIT cells between species, MAIT cells in Tg mice do not reflect the properties of MAIT cells in WT mice or humans (28, 41, 43). Recently, researchers have developed two new tools to study human and mouse MAIT cells: MR1-antigen tetramers (29, 50) and the B6-MAITCAST mouse, whose augmented MAIT cells are phenotypically very similar to human and WT mouse MAIT cells (**Table 1**) (38). These two new tools have facilitated the characterization of "natural" mouse MAIT cells, and the use of MR1-antigen tetramers circumvents the overlap/limitations of surrogate phenotypes to define MAIT cells in humans.

#### MAIT CELL ANTIGENS

Mucosal-associated invariant T cells are known as nonconventional T cells in part because they recognize non-peptide antigens presented by the non-polymorphic MR1 molecule (51). MR1 has a standard MHC-I fold, and the antigen-presenting process of MR1 is similar to MHC-I and CD1d molecules, leading Huang et al. to speculate that MR1 could present peptide or lipid antigens to MAIT cells (52). Kjer-Nielsen et al. later showed


#### Table 1 | Surface phenotype of MAIT cells in mouse and human.

*n.d., not described.*

that MAIT cells could recognize small molecule metabolites, including reduced 6-hydroxymethyl-8-d-ribityllumazine (rRL-6-CH2OH), 7-hydroxy-6-methyl-8-d-ribityllumazine (RL-6-Me-7-OH), and 6,7-dimethyl-8-d-ribityllumazine (RL-6,7-diMe), derived from the riboflavin [vitamin B2 (VB2)] biosynthetic pathway of bacteria (4). These compounds all have a bicyclic structure and a ribityl tail, but differ in their potency for activating MAIT cells (**Table 2**). Further research found that the most potent agonists were unstable pyrimidine intermediates, including 5-(2-oxoethylideneamino)-6-d-ribitylaminouracil (5-OE-RU) and 5-(2-oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU), which are also derived from the VB2 biosynthetic pathway and are formed by the spontaneous non-enzymatic reactions between 5-amino-6-d-ribitylaminouracil (5-A-RU) and glyoxal or methylglyoxal, respectively (50). Although these three pyrimidines have the same single-ring structure and ribityl tail, 5-A-RU does not stimulate MAIT cells because MR1 is not refolded efficiently with 5-A-RU alone (50, 53). Because 5-OE-RU and 5-OP-RU are very unstable in water, Mak et al. preformed 5-OP-RU in DMSO-*d*6 and then diluted it in aqueous buffer, and as a result, improved its purity, quantity, and stability (54). They also designed and synthesized a new reagent **11**, a very potent analogue of 5-OP-RU, which was completely stable in water and showed a similar functional profile as 5-OP-RU, making it a valuable antigen analogue that can be used for MAIT cell research. Presently, six distinct antigens have been identified that are capable of activating MAIT cells, five VB2 metabolites and 5-OP-RU analogue. Although similar in structure, these antigens differ in their abilities to activate MAIT cells (**Table 2**) (4, 50, 54). In addition to these six agonists, MAIT cell inhibitors have also been identified, such as 6-formylpterin (6-FP) and acetyl-6-formylpterin (Ac-6-FP), which are folic acid [vitamin B9 (VB9)] derivatives with a bicyclic structure and a formyl group (4, 53, 55, 56); Ac-6-FP is a stronger inhibitor than 6-FP, which is the first MR1 ligand described (4, 53).

Another reason MAIT cells are termed non-conventional is that they use a semi-invariant TCR and are restricted by a highly conserved MR1, implying a MAIT cell antigen repertoire with limited diversity (56). Known MAIT cell antigens only fill the A′ pocket of MR1 yet the heterogeneity of MAIT TCR chain Jα and Vβ usage suggested that MAIT cells may recognize additional antigens (17, 29, 30, 56). Therefore, finding more antigens will be helpful to understand MAIT cell recognition and activation. For example, the five known VB2 metabolite antigens, which are only synthesized by plants as well as many bacteria and yeast (56, 58), may reflect a primitive mechanism of self/non-self discrimination (59, 60). Recently, Keller et al. found that aside from VB9 derivatives and VB2 precursors, MR1 also presents drugs and drug-related molecules with diverse chemical structures (57), indicating that the pocket of MR1 has sufficient plasticity to accommodate a variety of molecules, including agonists or inhibitors of MAIT cell activation, such as diclofenac (DCF), 5-hydroxy DCF (5-OH-DCF) and 4′-hydroxy DCF (4′-OH-DCF), and 3-formylsalicylic acid (3-F-SA), 5-formylsalicylic acid (5-F-SA), 2-hydroxy-1-naphthaldehyde (2-OH-1-NA) and 2,4-diamino-6-formylpteridine (2,4-DA-6-FP), respectively. Therefore, we can modulate MAIT cell function by some of these drugs. Taken together, researchers have reported several agonists and inhibitors derived from VB metabolites and drugs or drug-related molecules. Of them, some molecules, such as DCF, DCF metabolites, and 5-F-SA, only activate specific subsets of MAIT cells with specific TCRβ chain usage (57), suggesting that the use of MAIT TCRβ chains "fine tunes" the responsiveness to certain antigens. Indeed, the effector responses of MAIT cells against *Escherichia coli* (*E. coli*) and *Candida albicans* (*C. albicans*) exhibit TCRβ



*#: a potent analogue of 5-OP-RU.*

*\*: also as an agonist of activation of MAIT cells expressing a TRBV6-4.*

+*: the potency for activating.*

−*: the potency for non-activating.*

chain bias. MAIT cells that are hyporesponsive to *E. coli* infection express Vβ8, Vβ13.1, and Vβ13.6, while MAIT cells that are responsive to *C. albicans* infection have a different TCRβ chain bias (34). To date, there are no reports about the nature of endogenous MAIT cell antigens; the antigens involved in MR1-mediated MAIT cell activation are all exogenous, which is inconsistent with iNKT cells (6, 24).

The identification of specific MAIT cell antigens has resulted in the generation of MR1-antigen tetramers. The first generation of tetramers was generated by loading of MR1 with rRL-6-CH2OH (29), which had lower affinity for staining MAIT cells. Now, however, the second generation of MR1 tetramers is prepared with 5-OP-RU, which is the most potent MAIT cell activator so far (50). Although MR1-antigen tetramers have facilitated studying and understanding mouse and human MAIT cell research, the use of MR1 tetramer staining has some limitations. Several authors have shown that MR1-antigen tetramer<sup>+</sup> T cells are not all MAIT cells (10, 61) and contain 1–4% Vα7.2<sup>−</sup> T cells, one subset of which detects infection with the riboflavin auxotroph *Streptococcus pyogenes* (*S. pyogenes*) in a TCR-dependent manner (61), implying that MR1 can present different ligands to MR1-antigen tetramer<sup>+</sup> T cells depending on the TCRα chain usage. Indeed, Lepore et al. discovered a novel population of MR1-restricted T cells (MR1T), which are MR1 antigen tetramer<sup>+</sup> T cells but are not MAIT cells, have variable TCRα chains, and can be activated by self-antigens (62). These antigens are purified from two tumor cell lines, and can activate MR1T cells without forming a Schiff base with MR1. However, the nature of self-antigens still need to be further studied.

#### MAIT TCR-MR1-ANTIGEN INTERACTIONS

Several reviews are available that discuss MAIT TCR-MR1 antigen interactions (30, 51, 56, 59); therefore, we will focus on the difference between recently discovered antigens, drugs and drug-related metabolites, and riboflavin metabolites in interaction with TCR-MR1. Comparisons of the ternary structures of MAIT TCR-MR1-6-FP and MAIT TCR-MR1-RL-6-Me-7-OH suggested that MR1 has plasticity in its ability to accommodate ligands (59, 60). Consistent with this, there are more MR1 ligands discovered recently (50, 57). Whether it be VB2 metabolites or drugs and drug-related metabolites, the MAIT TCR docks with MR1-antigen complexes in an approximately orthogonal and central manner (50, 57, 63). However, based on crystallography studies, DCF and 5-OH-DCF presentation is dramatically different from that of VB2 antigens, which are presented in the MR1-binding cleft with the plane of the central phenylacetic acid ring basically perpendicular to that of the phenylacetic acid ring of DCF when it interacts with MR1-TCR (57).

Crystal structures of the MR1-antigen-MAIT TCR complex revealed that Tyr95α (Y95α) on MAIT TCRα chain interacts with ribityl tail of VB2 derived antigens, which plays a vital role in activating MAIT cells (29, 53, 63, 64); however, Y95α does not interact with DCF and 5-OH-DCF, which lack a ribityl group (57). DCF and 5-OH-DCF activate MAIT cells when the Y95α aromatic ring piles against the phenylacetic acid ring forming van der Waals forces with E99β from TCRβ chain (57). In contrast, inhibitors, such as 3-F-SA, 2-OH-NA, and 2,4-DA-6-FP, do not directly interact with the MAIT TCR (57), which is also the case with VB9-derived inhibitors (53). Therefore, small organic molecules acting as MAIT cell antigens directly contact the MAIT TCR, and vice versa.

#### MAIT CELL ACTIVATION

Mucosal-associated invariant T cells are known to be activated by some bacteria and yeasts, but not by viruses, in an MR1 dependent manner (11, 65). However, recent studies have shown that not only in viral diseases (7, 66), but also in some non-infectious diseases (67–69), MAIT cells can be activated in an MR1-independent manner. In either the MR1-dependent or -independent manner, upon activation, MAIT cells rapidly proliferate, secrete proinflammatory cytokines and other factors resulting in lysis of the infected cells, and have the capacity for B cell help (5, 70, 71).

Microbes which utilize the riboflavin biosynthetic pathway activate MAIT cells in an MR1-dependent manner (**Figure 1A**) (11, 37, 65, 72, 73). Human MAIT TCRs recognize VB2-derived antigens presented by MR1, and MAIT cells then upregulate the expression of CD25, CD69, and CD161, secrete Th1-type cytokines (IFN-γ and TNF-α) and Th17-type cytokines (IL-17 and IL-22), but do not secrete Th2-type cytokines (30, 40, 74), which is consistent with their expression of transcription factors like T-bet and RORγt (36, 44, 75). Mouse MAIT cells secrete high levels of IL-17, but lower levels of IL-4, IL-10, IL-13, IFN-γ, TNF-α, and GM-CSF upon activation with anti-CD3/ CD28-coated beads (28, 30, 38). In addition to the secretion of proinflammatory cytokines, MAIT cells produce Gzms (GzmA, GzmK, and GzmB) and perforin, which function to effectively lyse infected cells (5, 76). Primary human MAIT cell activation is inefficient after *in vitro* stimulation with soluble ligands in an MR1-dependent manner, but also requires toll-like receptor (TLR) signaling and antigen-presenting cell (APC) activation (77). Consistent with this, accumulation and enrichment of MAIT cells *in vivo* not only requires VB2-derived antigens but also costimulatory signals, such as TLR agonists (37). Therefore, to establish a murine model of bacterial infection for MAIT cell studies, mice can be inoculated with synthetic antigens and TLR agonists such as CpG and poly I:C first, to promote MAIT cell accumulation and proliferation.

Like iNKT cells, MAIT cells can also be activated in an MR1 independent manner (**Figure 1B**). Several studies have shown that MR1-independent MAIT cell activation is dependent on IL-18 in synergy with other inflammatory mediators (7, 45, 66), which is consistent with the high expression of IL-18Rα on MAIT cells (30, 60). Upon cytokine-mediated activation in viral infections, MAIT cells can produce IFN-γ and GzmB, which

can be suppressed or even abrogated by anti-IL-18 but not by MR1-blocking (7, 66, 78). In contrast, MAIT cells are activated via different mechanisms in response to different viral infections. In dengue virus (DENV) infection, MAIT cells respond to both IL-12 and IL-18, while in hepatitis C virus (HCV) and influenza virus (IAV) infection, MAIT cell activation is predominantly dependent on IL-18, which is produced by CD14<sup>+</sup> monocytes (66, 75, 78). Moreover, IFN-α/β may also play a role in MAIT cell activation during viral infection (7). Corresponding to activation of MAIT cells in viral infection, previous studies showed that TLR8 and TLR3 are potent activators of MAIT cells, promoting the secretion of IL-12 and IL-18 by APCs (7, 45, 79). In addition to viral infections, MAIT cells also respond to infections with some bacterial species, such as *Mycobacterium tuberculosis*, *Mycobacterium bovis* bacillus Calmette–Guérin, and *Enterococcus faecalis*, in an MR1-independent manner (45, 80, 81). So, although there are two pathways of MAIT cell activation in response to bacterial infections, MR1-dependent and MR1 independent, the relative contribution of the two pathways is still unclear. Interestingly, MAIT cells are also activated in some non-infectious diseases. In patients with systemic lupus erythematosus (SLE), MAIT cells can be activated by IFN-α, IL-15, and IL-12 plus IL-18 in the absence of exogenous antigens (69). The positive correlation between the plasma concentration of these cytokines and the expression levels of CD69 on MAIT cells suggests that proinflammatory cytokines may activate MAIT cells and may play a role in the pathogenesis of SLE and possibly other inflammatory processes.

Interestingly, recently a new study showed that, apart from MR1-dependent and cytokine-mediated activation, MAIT cells can also be activated by superantigens (SAgs) produced by *Staphylococcus aureus* and *S. pyogenes* in a TCR Vβ-dependent manner (**Figure 1C**) (82), following which Sandberg et al. wrote a commentary to highlight this new discovery (83). Moreover, SAgs also activate MAIT cells through IL-18/IL-12 signaling, which is dominant over the TCR Vβ-dependent pathway of MAIT cell activation (**Figure 1C**). MAIT cell activation also requires MHC-II interaction with SAgs, which can activate conventional T cells through binding to TCR Vβ chains, and conventional T cells then promote the production of IL-18 and IL-12 through release of inflammatory mediators (82, 84). Upon activation by SAgs, MAIT cells make significant contributions to the cytokine storm via rapid production of proinflammatory cytokines but then are anergized to subsequent bacterial challenge through upregulation of inhibitory receptors such as lymphocyte-activation gene 3, demonstrating that MAIT cells also play a role in pathogenesis in some bacterial infection (82, 83).

#### MAIT CELLS AND DISEASES

In 2010, two studies reported that MAIT cells reacted to infected cells (11, 65). Since that time, there has been a growing body of research describing the role of MAIT cells in disease. Many have suggested that MAIT cells play important roles in infectious diseases, including bacterial and viral diseases, and non-infectious diseases, including autoimmune diseases and cancer; this topic has been reviewed recently (42, 74, 85–88), so here, we will focus on more recently published articles.

Many studies have described a role of MAIT cells in bacterial infections (38, 89). For example, human CD8<sup>+</sup>MAIT cells are important in combating *Salmonella enterica serovar Typhi* (*S. typhi*) infection (73) and recently using a mouse model of *Salmonella* Typhimurium infection, MAIT cells have been shown to accumulate in the lungs of infected mice (37). Similarly, in response to *S. typhi*, MAIT cells are activated and home to affected tissues, such as the gut (73). MAIT cells can also be activated and functionally impaired in patients with scrub typhus, caused by *Orientia tsutsugamushi* (90).

As mentioned above, MAIT cells may also be involved in the clearance of some viral infections (45, 80). In patients with HCV and DENV infections, MAIT cells are present at a lower frequency in blood than in healthy controls, and can be activated in a cytokine-mediated manner to upregulate the expression of IFN-γ and GzmB (7, 78, 91). Furthermore, there is a negative correlation between the degree of depletion of the intrahepatic MAIT cells and the severity of liver inflammation and fibrosis in HCV-infected patients (78). In IAV infections, human MAIT cells can resist infection by IL-18-dependent activation (7, 66). Moreover, the function and frequency of MAIT cells can be impaired and reduced, respectively, in human T lymphotropic virus type 1 infection (92).

Apart from infectious diseases, MAIT cells also play potential roles in inflammatory diseases and cancers. Patients with ankylosing spondylitis have a lower frequency of MAIT cells in peripheral blood but a higher frequency in synovial fluid (SF) compared with the healthy controls, and MAIT cells in SF of patients with an exaggerated IL-17 phenotype are primed by IL-7 (93). Moreover, human MAIT cells are reduced and functionally altered in the peripheral blood of patients with common variable immunodeficiency (CVID) (94) and primary Sjogren's syndrome (pSS) (68); however, the alteration of MAIT cell function in CVID and pSS patients is different. In CVID patients, MAIT cells are more activated and produce cytokines at higher frequencies than MAIT cells in healthy controls, while in pSS patients, MAIT cells are activated at a lower level and produce less cytokines than MAIT cells in healthy controls. In some cancers, such as gastric, colon, and lung cancers, MAIT cells can migrate from the peripheral blood to affected tissues, express GzmB and perforin, and thus have the potential to kill cancer cells in patients with mucosal-associated cancers (67). Moreover, MAIT cells heavily infiltrate the hepatic metastases in patients with colorectal carcinoma (CRC), and the function of tumor-infiltrating and tumor-margin MAIT cells are impaired irrespective of preoperative chemotherapy, implying the attractiveness of therapeutic targeting of MAIT cells in CRC (95).

However, MAIT cells also play a pathogenic, not protective, role in SAg-mediated illnesses (82, 83). MAIT cells activated by SAgs not only contribute to cytokine storm but also acquire an anergic phenotype, indicating a role for them in immunopathology and immunosuppression and also implying that MAIT cells can be used as an efficacious therapeutic target of SAg-mediated illnesses. Collectively, based on an ever increasing understanding of the mechanisms involved in MAIT cell activation, MAIT cells continue to be implicated in more and more diseases, often with differing mechanisms of activation.

Given their ability to rapidly produce cytokines after stimulation, direct lysing of microbe-infected cells, and their potential to regulate immune responses, MAIT cells similar to iNKT cells and Vγ9Vδ2 T cells, which have been used as targets in immunotherapy research (96–101), may also be an effective target for disease treatment. For example, due to their potential role in cancers and infectious diseases, adoptive MAIT cell transfer therapy may be a promising approach to treat diseases (102). Moreover, with a growing number of MAIT cell antigens to be found, the potent antigens can be administered alone or pulsed on APCs to enhance immune responses, or can be used as an effective adjuvant to boost vaccine efficacy. Furthermore, as their pathogenic role in SAg-mediated illnesses, blocking the function of MAIT cells may alleviate the pathological damage caused by SAgs (82). However, there is still no one immunotherapy based on MAIT cells to be used in clinical trials. Therefore, our better understanding of the biology of MAIT cells will accelerate the use of MAIT cell immunotherapy in diseases.

## CONCLUSION

Since the original description in 1993 of a population of unconventional DN T cells with a limited repertoire, much has been learned about the role of MAIT cells in immune response, including a better understanding of MAIT cell phenotypes, restricting molecules, MAIT cell development, antigen recognition, molecular interactions of the TCR-MR1-antigen complex, and their roles in diseases. It was previously believed that in human T cells expressing Vα7.2-Jα33<sup>+</sup> TCRα chain were exclusively MAIT cells, but later it was found that, in addition to canonical Vα7.2-Jα33 TCRα, MAIT cells also express non-canonical Vα7.2-Jα12/20 combinations (29). Similarly, it was originally believed that MR1-restricted T cells were MAIT cells, but later it was confirmed that MR1-restricted T cells are not all MAIT cells, which also contain Vα7.2<sup>−</sup> T cells (10, 61). Presently MAIT cells are defined as those T cells which express a Vα7.2 TCRα chain and are restricted by MR1. It is now well known that MAIT cell antigens are not only confined to VB2 metabolites. Drugs and drug-related molecules can also serve as MAIT cell agonists (57). MAIT cell activation can also be triggered by SAgs in an MR1 independent manner, but dependent on the certain TCR Vβ and/ or cytokine-mediated pathway (82, 83).

With more and more antigens being described that are recognized by MAIT cells, along with new and different mechanisms

#### REFERENCES


of activation, our understanding of MAIT cells continues to expand. However, some questions still remain unanswered. Firstly, although the number of exogenous antigens continues to grow, are there endogenous antigens that activate MAIT cells? And what is their nature? Secondly, even though the intracellular trafficking pathway of the MR1-antigen complexes within the cell has been shown (103, 104), what intracellular machinery is responsible for retaining empty MR1 molecule in the endoplasmic reticulum to load the extracellular antigens? Thirdly, as MAIT cells are recognized as a potential subset of tissue-resident T cells, what is their contribution to the pathogenesis of tissue or organ (such as liver and lung) diseases? Although some work has described MAIT cells in animals other than human and mouse, what are the roles of MAIT cells play in other animal species? Fourthly, how can MAIT cell responses be modulated, either through small molecules or immunotherapy, to improve human health? Finally, almost all published human data are generated by *in vitro* assays with a short period of experimental time. Given all the caveats of any *in vitro* experiments, the findings/observations remain to be verified by *in vivo* studies in mouse work, albeit the species difference.

Mucosal-associated invariant T cells, the once obscure cell population, represent up to 10% of T cells in the human peripheral blood, therefore, it is imperative that we continue to identify and characterize antigens and mechanisms that activate MAIT cells to better understand the whole scope of the immune response against microorganisms and disease pathogenesis. New insights will lead to improved therapeutics for infectious and non-infectious diseases controlled in part by MAIT cells.

#### AUTHOR CONTRIBUTIONS

JC designed the outline, organized the text, and critically revised the manuscript. XX drafted the manuscript. All authors reviewed and approved the final version of the manuscript.

#### ACKNOWLEDGMENTS

We thank Dr. Patricia Wilkins at Parasitology Services, USA, for editorial assistance. This work was supported by the Innovative Special Project of Agricultural Sci-Tech (grant no. CAASASTIP-2014-LVRI-09) and Fundamental Research Program of Chinese Academy of Agricultural Sciences (grant no. 0032160017).


in mucosal-associated invariant T cells. *J Exp Med* (2013) 210:2305–20. doi:10.1084/jem.20130958


antimicrobial MR1-restricted MAIT-cell population in chronic HIV-1 infection. *Blood* (2013) 121:1124–35. doi:10.1182/blood-2012-07-445429


cells administration to the nasal submucosa in unresectable or recurrent head and neck cancer. *Cancer Immunol Immunother* (2008) 57:337–45. doi:10.1007/s00262-007-0373-5


**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 Xiao and Cai. 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.*